Thyroid hormone receptor alpha sumoylation modulates white adipose tissue stores

Thyroid hormone (TH) and thyroid hormone receptor (THR) regulate stem cell proliferation and differentiation during development, as well as during tissue renewal and repair in the adult. THR undergoes posttranslational modification by small ubiquitin-like modifier (SUMO). We generated the THRA (K283Q/K288R)−/− mouse model for in vivo studies and used human primary preadipocytes expressing the THRA sumoylation mutant (K283R/K288R) and isolated preadipocytes from mutant mice for in vitro studies. THRA mutant mice had reduced white adipose stores and reduced adipocyte cell diameter on a chow diet, compared to wild-type, and these differences were further enhanced after a high fat diet. Reduced preadipocyte proliferation in mutant mice, compared to wt, was shown after in vivo labeling of preadipocytes with EdU and in preadipocytes isolated from mice fat stores and studied in vitro. Mice with the desumoylated THRA had disruptions in cell cycle G1/S transition and this was associated with a reduction in the availability of cyclin D2 and cyclin-dependent kinase 2. The genes coding for cyclin D1, cyclin D2, cyclin-dependent kinase 2 and Culin3 are stimulated by cAMP Response Element Binding Protein (CREB) and contain CREB Response Elements (CREs) in their regulatory regions. We demonstrate, by Chromatin Immunoprecipitation (ChIP) assay, that in mice with the THRA K283Q/K288R mutant there was reduced CREB binding to the CRE. Mice with a THRA sumoylation mutant had reduced fat stores on chow and high fat diets and reduced adipocyte diameter.

www.nature.com/scientificreports/ Preadipocytes proliferation in mutant mice. We investigated whether the reduced fat mass in THRA K283Q/K288R mutant mice was associated with reduced preadipocyte proliferation. To determine the preadipocyte proliferation status in live mice, we performed EdU labeling of 3 week old mice. The EdU labeled-nuclei represent cells in S phase, a phase of active DNA synthesis. In adipose tissue, dividing preadipocytes are concentrated in the vascular stroma. The EdU labeled-nuclei were reduced 27% ± SD 5.2% (p < 0.02) in the vascular stroma of inguinal adipose tissue in THRA mutant mice, compared to Wt mice (Fig. 4A,B). We also isolated preadipocytes from the stromal vascular fraction of iWAT from 6 week old Wt and THRA mutant mice. Cells were synchronized to G 0 /G 1 cell cycle by serum withdrawal for 36 h and then stimulated to enter the cell cycle by addition of a 10% serum supplement. The proliferation of preadipocytes isolated from THRA K283Q/K288R mutant mice, as determined by EdU-DNA incorporation, was reduced 24.3% ± SD 3.7% (p < 0.011) in mutant mice compared to Wt mice (Fig. 4C,D). The number of preadipocytes isolated from SVF of the iWAT of the mutant mice was reduced 23.3% ± SD 3.3% (p < 0.0035) from 3 individual experiments (n = 6 mice/experiment) compared to that of Wt mice (Fig. 4E). The mRNA levels of Cyclin D1 and D2 and cyclin-dependent kinase 2 (Cdk2) were reduced in the mutant preadipocytes, compared to Wt mice (Fig. 4F).
THRA sumoylation-defective mutant K283Q/K288R reduces nuclear localization of cyclin D1, cyclin D2, Cdk4 and Cdk2 in human primary preadipocytes. The influence of the sumoylation defective THRA on proliferation was studied in human primary preadipocytes cells, transfected with plasmids expressing THRA, THRA K283Q/K288R, or an empty vector (control), synchronized to G 0 by serum withdrawal for 48 h. Serum withdrawal resulted in more than 75% of the cells synchronized to G 0 phase, as assessed by flow cytometry. Serum withdrawal prolonged the G 1 and S 1 phases, compared to growth in serum conditions, due to recovery from the lower cellular metabolism associated with serum starvation 16 . Cells were, therefore, supplemented with serum for 9 h and analyzed for the expression of G 1 phase regulators. In the G 1 phase, cyclin D1 forms a complex with cyclin-dependent kinase 4 (Cdk4) The cyclin D1/Cdk4 complex is then transported from the cytoplasm to the nucleus by p27kip Cdk inhibitor 17 . In cells transfected with THRA K283Q/K288R, compared to those transfected with empty vector (control) and THRA, the immunofluorescent (IF) staining demonstrated that Cdk4 protein was predominantly in the cytoplasm but not the nucleus (Fig. 5A). The cell lysate was fractionated into cytoplasmic and nuclear fractions and immunoprecipitated (IP) with antibodies against  www.nature.com/scientificreports/ www.nature.com/scientificreports/ Cdk4 in order to quantitate the subcellular distribution of Cdk4. The quality of fractionation was confirmed by alpha tubulin staining for the cytoplasmic fraction and lamininB1 for the nuclear fraction (see Supplementary  Fig. S5 online). In the cytoplasmic fraction, Cdk4 was similar among different samples. In the nuclear fractions, Cdk4 was reduced in cells expressing THRA K283Q/K288R, compared to control and THRA-transfected cells (Fig. 5B). Cdk4 enters the nucleus in the form of a cyclin D1/Cdk4 complex, carried by p27 kip , so the reduced nuclear Cdk4 could be the result of reduced nuclear cyclin D1. To investigate this possibility, we performed a co-IP with anti-p27 kip and WB with anti-cyclin D1. We found that in the cells transfected with THRA K283Q/ K288R, p27 kip -bound nuclear cyclin D1 was only 45% of the control levels ( Fig. 5C). Cyclin D1 mRNA was reduced 28% (p < 0.033) but CDk4 mRNA was not changed (Fig. 5B,C far right panels), indicating that reduction of the cyclin D1/Cdk4 complex is associated with reduced cyclin D1 gene expression. Cyclin D2 is a G 1 /S phase-specific cyclin and forms a complex with Cdk4/6 that mediates the G 1 /S transition. We detected reduced nuclear cyclin D2 protein by immunofluorescence and IP of cytoplasmic and nuclear fractions in THRA K283Q/K288R-transfected cells, compared to THRA and control ( Fig. 6A-C). Cyclin D2 mRNA was decreased 36% in mutant cells compared to control (p < 0.034) (Fig. 6D). Cdk2 is another important factor regulating the G 1 /S phase transition. Cdk2 moves into the nucleus and, together with Cdk4, sequentially phosphorylates retinoblastoma protein (Rb). Cdk2, as shown by immunofluorescence staining and IP, localized to the nucleus in control and THRA-transfected cells. In THRA K283Q/K288R-tranfected cells, however, there was significantly less Cdk2 protein in the nucleus (Fig. 6E-G). Cdk2 mRNA levels were reduced 20% in THRA K283Q/K288R-transfected cells (Fig. 6H), indicating that reduced nuclear Cdk2 protein was associated with impaired Cdk2 gene expression and reduced nuclear translocation. These data show that expression of THRA K283Q/K288R is associated with slowing of the G 1 /S transition. We observed reduced nuclear localization of the cyclinD-Cdk complex and reduced mRNA levels of the cell cycle regulating genes, Cyclin D1, Cyclin D2 and Cdk2 in the G1 phase.
Delayed G 1 /S transition. We performed cell cycle analysis by Flow Cytometry to evaluate S-phase entry ( Fig. 7A). Cells were synchronized by serum withdrawal and then cells were released to the G 1 phase by serum supplementation. At baseline, 0 h (before adding serum), the number of cells in G 1 phase was not significantly different among the different plasmid-transfected cells (Fig. 7B). The fraction of cells in S phase, 16 h after serum supplementation, was 44.7% in control and 47.5% in THRA-expressing cells, but reduced to 37.5% (p < 0.01) in cells expressing THRA K283Q/K288R (Fig. 7C). The S phase lag persisted at 24 h in THRA K283Q/K288Rtransfected cells, compared to control and THRA-transfected cells. The late S-phase entry delayed DNA synthesis and the expansion of cellular content, resulting in a reduced proliferation rate. Figure 5. Mutation of the THRA sumoylation site, K283Q/K288R, associated with reduced nuclear localization of cyclin D1-Cdk4 complex, cyclin D2 and Cdk2 in G 1 phase. (A) Human preadipocytes were transfected with empty vector (control) or plasmids expressing THRA or THRA K283Q/K288R. Cells were synchronized to G 0 / G 1 phase by serum starvation for 48 h. Cells were then allowed to re-enter the cell cycle by supplementing the medium with 10% calf serum. After 9 h of serum supplementation, the cells were analyzed for nuclear content of cyclin D1-Cdk4 complex, cyclin D2 and Cdk2. Immunofluorescent (IF) staining of cyclin D1 and Cdk4 nuclear localization in cells. Scale bar, 50 μm. Circled inset shows higher magnification. (B) The nuclear and cytoplasmic fractions were analyzed for Cdk4 protein expression by IP with anti-Cdk4 Ab, followed by immunoblot (IB) with anti-Cdk4 and quantified Cdk4 IB signals using Li-Cor Image Studio Lite. (C) Co-IP detection of cyclin D1 in the cytoplasm and nuclear fractions. Anti-p27kip antibody was used in IP and anti-cyclin D1 antibody in IB. Quantification of WB was done using Image Studio Lite. q-PCR quantification of Cdk4 and cyclin D1 (CCND1) mRNA expression, presented as the mean value and SD, fold above control. *p < 0.05 compare to control by One-way Anova. IP-immunoprecipitation, IB-immunoblot. The blots (B,C) were cut prior to hybridization and the origin blots are provided (see Supplementary Fig. S7  www.nature.com/scientificreports/ Influence of the THRA K283Q/K288R on CREB-and AP1-mediated gene expression. To determine in an unbiased fashion the signaling pathways disrupted by the desumoylated THRA, we screened the transcription factors important in the G 1 phase regulation, using a functional reporter assay. CREB, P53, C-myc, SMAD, FoxO1, SP1 and AP1 are critical transcription factors for expression of genes important for cell cycle regulation [18][19][20][21][22][23][24][25][26][27][28][29][30] . CREB, C-myc, SP1 and AP1 are positive regulators of cell cycle progression [30][31][32][33] and P53, SMAD and FoxO1 are negative regulators, promoting cell cycle exit [34][35][36] . We used luciferase reporters carrying multiple (6x) consensus response elements for each transcription factor and determined the functional activity of these transcription factors in cells transfected with empty vector (control), and vectors expressing THRA and THRA K283Q/K288R. The luciferase activity was not different between control and THRA-transfected cells for all transcription factor studied. However, in THRA K283Q/K288R-transfected cells, luciferase transcription-mediated by CREB and AP1 were reduced 46% (p < 0.05) and 53% (p < 0.05), respectively and P53 and SP1 were increased 31% (p < 0.05) and 53% (p < 0.05), respectively, compared to control cells (Fig. 8A). The transcription activity mediated by c-Myc, SMAD and FoxO1 was barely detectable. CREB, AP1 and SP1 are known to regulate the expression of CCND1 (cyclin D1), CCND2 (cyclin D2) and Cdk2 and the response elements for these factors have been identified 19,22,27 . THRs has been previously shown to antagonize CREB-mediated transcription 37-39 . Based www.nature.com/scientificreports/ on these findings, we focused on the effects of the desumoylated THRA on CREB-mediated transcription for these key cell cycle genes.
Altered THRA-CREB interaction in preadipocytes isolated from THRA K283Q/K288R mutant mice. THR does not bind directly to the CREB response element (CRE) but has been shown, in previous studies, to interact with CREB, as demonstrated by pull-down and co-IP assays 37,40,41 . We postulated that the sumoylation-defective THRA K283Q/K288R influenced protein-protein interactions between THRA and CREB, resulting in reduced expression of cell cycle genes. There are four potential CREs in the mouse CCND1 gene, numbered CCND1-site1 to-site 4, equivalent to the human orthologue CCND1 gene 32,42 . To detect the influence of THRA K283Q/K288R on protein-DNA and protein-protein interaction, we isolated preadipocytes from THRA K283Q/K288R mutant and wild type mice (abbreviated as mutant-preadipocytes and wt-preadipocytes). The mouse preadipocytes were synchronized by serum withdrawal for 36 h and then cultured with 10% serum for 8 h prior to harvesting nuclear protein for the ChIP assay. The initial ChIP assay, using anti-CREB antibody, showed that CREB bound to two of the four putative CREs (CCND1-site1 and CCND1-site2) in the G 1 phase (Table1). In the remainder of the studies, only these two active sites were evaluated. In addition, the effects of THRA on CREB binding to CREs in the CCND2, Cdk2 and Cullin3 (Cul3) gene promoters were investigated by ChIP assay (Table 1). CREB enrichment to the CCND1-site 2 was 7-times greater than that to CCND1-site 1 in wt-preadipocytes, indicating that the CCND1-site 2 is the primary CREB binding site (Fig. 8B,C). THRA from Wt-preadipocytes did not interact with CREB in either CCND1-site1 or site2 (Fig. 8B,C, orange bars). In mutant-preadipocytes, a very small amount of CREB was enriched to the CCND1-site1 and none to the CCND1-site 2. The THRA K283Q/K288R from mutant preadipocytes had significant enrichment (7.8%) to the CCND1-site 2 (Fig. 8C), indicating THRAK283Q/K288R is associated with reduced CREB DNA-binding.
The Cdk2 gene contains two potential CREs, but only one site had detectable CREB binding in our initial ChIP assays, which is referred as Ckd2 CRE (see Table 1 for location and sequence). In Wt-predipocytes, both THRA and CREB, were enriched at the Cdk2 CRE with similar enrichment, 1.3% and 1.2%, respectively (Fig. 8D). In mutant-preadipocytes, overall enrichment of CREB and THRA K283R/K288Q to this site was significantly reduced to 0.093% and 0.26% respectively. We speculate that fully sumoylated THRA facilitates CREB DNA binding on CDK2 CRE and THRA K283Q/K288R reduced CREB binding to the site. In the CCDN2 gene promoter region, low CREB enrichment was detected and THRA enrichment was negligible (Fig. 8E). Although the CCND2 mRNA levels was reduced in mutant preadipocytes, it was not due to reduced activation by CREB but may be the result of reduced AP1. There is an AP1 binding site located in the CCND2 promoter 23 . THRB is Normal rabbit IgG was used as the negative control in ChIP assays. All ChIP data shown are after deducting the negative control and expressing as % input enrichment. Input was 10% of total pre-cleared lysate. Statistical analysis was performed using Student t-test. CREB, cAMP response element binding protein; P53 tumor suppressor protein, c-Myc cellular homolog of myc oncogene, FoxO1 forkhead box protein 1, Sp1 specificity protein 1, AP1 activator protein 1. www.nature.com/scientificreports/ known to directly interact with AP1 protein 43,44 . The reporter assays demonstrated that AP1-mediated luciferase activity was reduced in mutant preadipocytes, consistent with reduced CCND2 mRNA due to the influence of THRA K283Q/K288R on AP1-mediated transcription. Cullin 3 (Cul3) is the core component of cullin-RING-based BCR (BTB-Cul3-RBX1) E3 ubiquitin-protein ligase complex, which targets cyclin E for ubiquitin-directed degradation and controls S-entry 45 . The S entry delay and reduced Cul3 mRNA seen in human primary preadipocytes transfected with sumoylation-defective THRA K283Q/K288/R may be associated with altered CREB-mediated Cul3 gene transcription. We examined the CREB binding to CRE of the Cul3 gene. In wt-preadipocytes, the enrichment of CREB and THRA to the CRE was 4% and 12%, respectively (Fig. 8F). THRA enrichment was greater than CREB, suggesting that THRA may be able to bind in the vicinity of the CRE and facilitate Cul3 mRNA expression. In mutant-preadipocytes, the enrichment of THRA K283Q/K288R and CREB to the Cul3 gene was eliminated, resulting in a reduction in Cul3 mRNA by 36% (p < 0.045) (Fig. 8G). The ChIP results indicate that in the presence of THRA K283Q/ K288R CREB-mediated gene expression of CCND1, Cdk2 and Cul3, was altered, consistent with the finding of delayed G 1 /S transition in THRA K283Q/K288R-transfected human preadipocytes. THRA K283Q/K288R is associated with reduced CREB binding to DNA and recruitment of NCoR. NCoR is a key regulator in THR-mediated transcription repression. NCoR protein contains two regions that specifically interact with unliganded THR 46 . Although THRA K283Q/K288R does not reduce T3 binding, as shown in a reporter assay, reduced sumoylation may alter the interface of THRA with other regulatory proteins. Using ChIP assays, we investigated whether an abnormal interaction between THRA K283Q/ K288R and NCoR alters CREB-mediated transcriptions in Wt-preadipocytes. At the CCND1-site1, neither THRA nor NCoR were enriched (Fig. 8H). At the CCND1-site2 (Fig. 8I), the enrichment of NCoR and THRA was detectable in very low levels in wt-preadipocytes, consistent with the lack of interaction between THRA and CREB on the CCND1 gene (Fig. 8C). In contrast, in mutant-preadipocytes, both THRA K283Q/K288R and NCoR were highly enriched and CREB reduced on the CCND1-site2.
The ChIP data suggests that reduced THRA sumoylation may play a role in THRA interaction with other transcription factors, such as CREB and could impact the expression of CCND1, Cdk2 and Cul3 genes. Additionally, THRA sumoylation may be important for interaction with NCoR.

Discussion
TH is essential for cell proliferation during development and in regeneration after injury, as shown in a range of cellular and animal models 2 . TH regulates stem cell proliferation in the intestine, heart, brain, bone and skeletal muscle. In most tissues the stem cell proliferation is mediated by THRA, however, In the liver, regeneration of hepatic cells is regulated by THRB. THRA modulates the proliferation of the intestinal epithelial progenitors at the bottom of the crypts 3,47,48 . THRA/TH regulates cell growth and differentiation in the myocardium by www.nature.com/scientificreports/ stimulating specific genes and the AKT/mTOR pathway [49][50][51] . In the adult mouse brain, disruption of thyroid hormone signaling reduces neuroprogenitor survival, proliferation and differentiation [52][53][54] . In skeletal muscle THRA mediates myogenesis and tissue repair in response to injury 55 . THRA regulates bone growth and turnover via regulating differentiation of chondrocytes and osteoblasts and ossification 56,57 . We used the THRA K283Q/K288R mouse model to investigate the impact of THRA sumoylation on adipocytes and fat deposition. In human primary preadipocytes, exogenous expression of mutant THRA (K283Q/ K283R), was associated with reduced nuclear protein content of cyclin D1, cyclin D2 and Cdk2 and downregulation of gene expression, leading to the delay of G1/S transition. In our mouse model, mutation at the THRA-SUMO conjugation site, K283Q/K288R, was associated with reduced WAT stores and reduced proliferation of preadipocytes as assessed by in vivo and in vitro labeling studies.
In general, an increased fraction of adipose stem cells, preadipocytes, is characteristic of metabolically healthy fat with the ability to expand fat storage by hyperplasia and reduce fat deposition in organs 58 . Adipocyte diameter varies by fat depot as well as differences in fat depot characteristics between genders. Expansion of fat stores by adipocyte hypertrophy and larger adipocytes diameter has been associated with insulin resistance and metabolic diseases. A number of studies, however, have associated smaller adipocyte diameter with insulin resistance and Type 2 diabetes 59 . A clinical study with 40% overfeeding for 8 weeks showed that those with smaller adipocyte diameter at baseline had a greater increase in insulin resistance that those with larger adipocytes 60 . Single cell expression profiling of adipose progenitors has shown at least three distinct populations of pre-adipocytes, further showing the complexity of assessing the role of preadipocytes in expansion of adipocyte stores 61 . Our mice with a THRA sumoylation site mutation showed a reduction in fat stores and serum leptin, although did not show significant adverse metabolic consequences of reduced preadipocytes. Addition of a high fat diet resulted in a greater difference between mutant and wt mice in the fraction of body fat and fat stores. Reduced fat stores may also be the result of enhanced energy expenditure. The impact of the desumoylated THRA on thermogenesis and metabolic rate will need to be directly evaluated.
The THRA sumoylation mutant was associated with reduced preadipocyte proliferation in the in vitro and in vivo model, so we investigated the pathways that may be involved. In the cell cycle G0/G1 transition, D-type cyclins D1 and D2 are rapidly synthesized to meet the cellular proliferation demand. cAMP response element binding protein (CREB), is an important regulator of the cell cycle through direct regulation of genes, including cyclin D1 (CCND1), cyclin D2 (CCND2) and cyclin-dependent kinase 2 (Cdk2) 19,22,27 . THRs has been previously shown to antagonize CREB-mediated transcription by direct protein-protein interaction, independent of T3, as has been demonstrated previously in pituitary specific-Pit1 and interleukin (IL)-6 promoters [37][38][39] . A recent characterization of a desumoylated THRB showed a direct disruption of CREB-mediated gene transcription in the thyroid on genes important for thyroid hormone synthesis 15 . A functional screen of transcription response element promoters in our in vitro model, indicated reduced CREB signaling in the presence of desumoylated THRA.
SUMO conjugation alters receptor DNA-binding as well as interaction with binding partner and other transcription factors, altering gene expression and influencing cellular growth, proliferation and cellular function 62 . We utilized a ChIP assay to determine how the presence of the THRA sumoylation mutant influenced CREB signaling. The association of desumoylated THR with CREB inhibits CREB phosphorylation and ligand-dependent recruitment of coactivators, which impairs the CREB-mediated transcription and reduces cell proliferation 37,38 .
The role of THR/T3 in development has spatial and temporal specificity. THR may interact with various signals to facilitate the cellular demand in stem cells proliferation, although such activity does not occur in mature lineages. CCND1, CCND2 and Cdk2 mRNA level were reduced in G1 phase of mutant-preadipocytes. However, there was no change in mature adipocytes (WAT). The gene expression of D-type cyclins and Cdks are regulated by multiple transcription factors. For example, in addition to the two CREB binding sites, the CCND1 promoter region, within 200 bp of transcription starting site (TSS) also contains binding sites for AP1, Ets, SP1, Egr1, TCF, NF k B and SATA5 63 . These transcription factors may act in spatial and temporal manner. CREB is involved in G1 phase regulation.
The mutation in THRA sumoylation site was associated with alterations in CREB binding, with an observed reduction in gene expression. The THRA mutant mouse model permitted us to study the physiologic function of THRA-SUMO conjugation, without disrupting the DNA binding and T3-mediated signaling properties of THRA. In a previous report, a mutation in the second zinc finger of THRB disrupted THRB augmentation of PI3K signaling, independent of T3, but did not impair direct TH action or THR binding to DNA 64 . THR residues that do not mediate direct nuclear signaling are important for modulating cell signaling pathways. A number of factors, independent of T3, have been shown to regulate cytoplasmic and nuclear localization of THR, including posttranslational modification by acetylation 65 . This nuclear-cytoplasmic shuttling may be importat in cell cylce regulation. THRA sumoylation may play a role in preadipocyte proliferation though CREB-mediated gene expression of G1 regulators.
The mouse with a THRA gene mutation and reduced sumoylation had smaller fat pads and adipocyte diameter, as well as a marked reduction in serum leptin. These differences in fat deposition between mutant and wt mice was increased by placing mice on a high fat diet. Alterations in metabolic rate may contribute to the reduced fat stores. We measured a reduction of preadipocyte proliferation in vitro and in an in vivo model. We hypothesize that reduced THRA sumoylation disrupted CREB signaling and interfered with cell cycle regulation. Although the role of preadipocytes likely varies by fat depot, a reduced fraction of preadipocytes are characteristic of obese individuals and those with insulin resistance. THR sumoylation may play a role in the regulation of preadipocyte proliferation and maintainance of cardiometabolic health. www.nature.com/scientificreports/

Materials and methods
Animal use approval. Animal care and research protocol were approved by the institutional IACUC of VA Greater Los Angeles Healthcare System and in accordance with the Guide for the Care and Use of Laboratory Animals. All methods were carried out in accordance with ARRIVE Guidelines.
Generation of sumoylation defective THRA K283Q/K288R mutant mice. The mutant mice were generated using FLP/FRT recombinase-mediated cassette exchange techniques (see Supplementary Fig. S1 online). The mutant mice were generated in C57BL/6NTac background. ES cells was derived from C57BL/ NTac mice and manipulated to create the desired targeted mutation in the mouse Thra gene. The ES cells with the Thra gene mutation were used as donors and injected to the acceptor embryos of the C57BL/6NTac mice. The embryos were implanted back to C57BL/NTac female mice. The pups, then, have the exact C57BL/6NTac genome except the specific Thra mutation. The mutation is located at exon 8 of THRA gene. The mutations are located at 1367nt (A to C) and 1381 nt (A to G) based on NCBI accession # NM_178060.4. The homozygous mutation was confirmed by direct DNA sequencing and High Resolution Melting curve. The homozygous mice were then bred to the same strain for at least three generations prior to using them in experiments. We used inhouse bred C57BL/6NTac mice with identical genetic background and living environment as the mutant mice for controls. Isolation and culture of mouse primary preadipocytes. Mouse primary preadipocytes were isolated from subcutaneous adipose tissue of wild type and THRA K283Q/K288R mutant mice using the method described previously 66 . In brief, 6 week-old male mice (n = 6) were euthanized. Subcutaneous fat pads were dissected, washed with PBS, and minced in 1 mL of advance DMEM/-F12 medium containing dipeptidase II (2.4 U/mL/10 mM CaCl 2 ) and collaginase D (1.5 U/mL). After 1 h digestion at 37 °C with agitation, the digestion was stopped by adding complete DMEM medium containing 10% FBS. After centrifuge for 10 min at 700 × g, the pellet (stromal vascular fraction) was collected, resuspended in 1 mL PBS, and filtered through a 40 μM cell strainer, and then plated to 1% collagen coated plates. After 1 h of plating, cells were carefully washed with warm PBS to remove unattached debris and fresh medium was added. The isolated cells were used immediately in subsequent experiments.
Chromatin immunoprecipitation (ChIP) assay. Preadipocytes isolated from inguinal fat were synchronized to G 1 phase as described the above. After the cells were re-fed with 10% calf serum for 8 h, cells were fixed with formaldehyde (final concentration 1%) for 10 min. Fixation was then stopped with the addition of glycine. The remainder of the procedure followed the manufacturer's instructions for ChIP Kit (Qiagen Inc). For each ChIP reaction, 2 million cells were used. The antibodies used were anti-THRA (ab2743 ChIP grade, rabbit, Abcam Inc.), anti-NCoR (17-10260, ChIP grade, rabbit, Millipore Inc.), and anti-CREB (ab31387, ChIP grade, rabbit, Abcam Inc.). Normal rabbit IgG was used as negative control. The Input was 10% pre-cleared total lysate. The purified DNA was PCR amplified for 35 cycles with triplicates. The data shown was deducted from negative control and expressed as % Input enrichment. ChIP q-PCR primers were Qiagen pre-designed and are shown ( Table 1).
In vivo EdU (5-ethynyl-2'-deoxyuridine) labeling of dividing preadipocytes in mice. EdU (Invitrogen) was prepared 50 μg/mL in saline and was administered to 3 week old mice (n = 2) at 50 μg/g BW (i.p.). Twelve hours after injection, mice were euthanized and inguinal fat was fixed in 4% paraformaldehyde in 4 °C overnight. The frozen tissue blocks were sectioned (12 μM thickness). The EdU was detected by Click-IT chemistry following the manufacturer instructions (Thermofisher Scientific Inc. www.nature.com/scientificreports/ magnification for vascular stromal areas and then these areas were examined in a × 20 magnification field. EdUincoporated dividing cells were quantified by counting the green fluorescent labeled nucleus in three consecutive fields with 20X magnification in each mouse.

EdU labeling of proliferating preadipocytes cells isolated from mutant mice. Cells were isolated
from the subcutaneous WAT of 6 week old wild type and mutant mice, grown to contact inhibition (48 h post 100% confluence), and then re-plated to 4-chamber slides at 2 × 10 4 cells per chamber. EDU (Clik-it EDU kit) was added to culture medium at a final concentration of 10 μM and incubated for 4 h in the culture incubator. Following fixation, cells were stained with Alexa Flour (594 or 488) for imaging. The number of EdU incorporated cells were quantified by counting total visible cells in a slide for 6 randomly selected frames and divided by the total visible DAPI labeled cells in these frames and presented as % DAPI stained cells.
Cell diameter quantification. Mice (n = 3/group) were fed with regular chow diet. The tissues were fixed by trans-cardial perfusion with 4% paraformaldehyde. Subcutaneous WAT was collected and paraffin embedded to prepare sections. The histology sections were imaged using Apical Imager. The cell diameter was quantified using ImageScope at similar location of inguinal fat pads of Wt and mutant mice. For each genotype, three sections were quantified at maximum enlargement for three consecutive frames. A total of 360 cells were measured for each genotype.
Quantification of total number of preadipocyte isolated from subcutaneous WAT of each mice. Preadipocytes were isolated from stromal vascular fraction (SVF) of subcutaneous fat of each mice (n = 3/genotype) as described above. Cells were collected and diluted in DMEM at low density. The number of preadipocytes were counted, using a hemocytometer, for 6 aliquots from each mouse.
Cell culture and transfections. Human primary preadipocytes from non-diabetic donors (Lonza Inc) were maintained in PBM-2 medium with 10% FBS supplement (Lonza Inc). Cells were transfected with control plasmid (empty vector) and plasmids expressing hTHRA or hTHRA K283Q/K288R by electroporation, using a 4D Nucleofector device and primary cell-specific reagent (Lonza Inc). G 1 phase synchronization. Primary human preadipocytes (Lonza Inc) were grown to complete confluence and maintained for 48 h post-confluence. Cells were then trypsinized, washed twice with warm PBS and cultured in PBM-2 medium without serum for 48 h. Primary preadiocytes isolated from the inguinal fat of Wt and THRA K283Q/K288R mutant mice were grown in DMEM with 10% calf serum to 80% confluence, then cultured for 3 days in DMEM with 0.5% calf serum before serum withdrawn. We observed that human primary preadipocytes tolerated 72 h of serum withdrawal without signs of apoptosis, but primary preadipocytes isolated from mice tolerated only 36 h of serum withdrawal. Preadipocytes from mice were, therefore, serum starved for 36 h to achieve G 1 synchronization. The number of cells synchronized to G 0 /G 1 was measured by flow cytometry.
To release cells to the cell cycle, 10% serum was supplied for the period (0, 8, 16 and 24 h). Cells that underwent prolonged serum withdrawal required longer than the usual period in G 1 phase to recover the cellular content and cellular activity before entering S phase. For this reason, G 1 /S transition was analyzed 9 h after the cells were refed with serum.

Immunofluorescent (IF) staining and confocal imaging.
Human primary preadipocytes were plated in chamber slides. Cell culture conditions, serum withdrawal and refeeding conditions, were the same as described for cell synchronization studies. Cells, 9 h after serum refeeding, were fixed with 4% paraformaldehyde, permeabilized in 0.2% TritonX-100/PBS, and then blocked with 5% goat serum in PBS. Cells were incubated with primary antibodies at 1:100 concentration overnight and secondary antibodies at 1: 250 for 1 h. Cells were then washed, mounted with Prolong Gold/DAPI (Life Science Inc.) and imaged using Zeiss Laser Scanning Microscope (LSM 710). The details on the antibodies details, see the section describing Western blot and IP/Co-IP procedures.
Cell cycle analysis by flow cytometry. Human primary preadipocytes were synchronized by serum withdrawal for 48 h. Cells were collected at various time points (0, 8, 16 and 24 h), after supplementation with 10% serum, and fixed for flow cytometry analysis with propidium iodine.
Quantitative RT-PCR analysis. Total RNA was isolated from cells (Qiagen kit). Primers used in q-PCR were pre-evaluated by Qiagen. The data were normalized with four housekeeping genes (GAPDH, beta-Actin, beta2-microglobulin and HPRT) and shown as the average of triplicate determinations.
RNA-seq analysis of gene expression. Total RNA was isolated from Inguinal fat of 3 month old male mice (n = 3/genotype) and used in RNA-seq analysis. The library construction, RNA-seq and data analysis were performed by the Genomic Core Facility at UCLA.

Fractionation of nuclear and cytoplasmic proteins, immunoprecipitation (IP/co-IP) and
Western Blot. Nuclear and cytoplasmic proteins were isolated using a Nuclear and Cytoplasmic Extraction Reagent (NE-PER) Kit (ThermoFisher Inc.) following the manufacturer's instructions. The fractioned components were used for IP/co-IP experiments. After IP/Co-IP, protein was separated by 10% SDS-PAGE, blotted to a PVC