Wnt3a disrupts GR-TEAD4-PPARγ2 positive circuits and cytoskeletal rearrangement in a β-catenin-dependent manner during early adipogenesis

Adipogenesis is a process which induces or represses many genes in a way to drive irreversible changes of cell phenotypes; lipid accumulation, round cell-shape, secreting many adipokines. As a master transcription factor (TF), PPARγ2 induces several target genes to orchestrate these adipogenic changes. Thus induction of Pparg2 gene is tightly regulated by many adipogenic and also anti-adipogenic factors. Four hours after the treatment of adipogenic hormones, more than fifteen TFs including glucocorticoid receptor (GR), C/EBPβ and AP-1 cooperatively bind the promoter of Pparg2 gene covering 400 bps, termed “hotspot”. In this study, we show that TEA domain family transcription factor (TEAD)4 reinforces occupancy of both GR and C/EBPβ on the hotspot of Pparg2 during early adipogenesis. Our findings that TEAD4 requires GR for its expression and for the ability to bind its own promoter and the hotspot region of Pparg2 gene indicate that GR is a common component of two positive circuits, which regulates the expression of both Tead4 and Pparg2. Wnt3a disrupts these mutually related positive circuits by limiting the nuclear location of GR in a β-catenin dependent manner. The antagonistic effects of β-catenin extend to cytoskeletal remodeling during the early phase of adipogenesis. GR is necessary for the rearrangements of both cytoskeleton and chromatin of Pparg2, whereas Wnt3a inhibits both processes in a β-catenin-dependent manner. Our results suggest that hotspot formation during early adipogenesis is related to cytoskeletal remodeling, which is regulated by the antagonistic action of GR and β-catenin, and that Wnt3a reinforces β-catenin function.


Introduction
The mouse 3T3-L1 cells have been widely used as an in vitro model system to investigate molecular mechanisms of adipogenesis 1 . The 3T3-L1 cells can be differentiated into mature adipocytes by two-day exposure to 3-isobutyl-1-methylxanthine (IBMX/M), dexamethasone (Dex/D), and insulin (I) [2][3][4] . Recently, genome wide analyses of DNase I hypersensitive regions revealed that after 4 h MDI treatment, approximately three times more DNA regions became accessible 5 . ChIP-seq analyses revealed that during such dynamic chromatin remodeling (within 4 h after MDI treatment), hundreds of DNA sequences covering 400 bps, termed "hotspots", including the promoter of Pparg2, are cooperatively occupied with at least five TFs including C/EBPβ/δ, GR (also known as NR3C1), STAT5A, and other TFs. These early TFs recruit coregulators to induce chromatin remodeling and form early enhanceosomes at the hotspots 1,5,6 . Hotspots are enriched with enhancer histone marks, namely, H3K4me1, H3K4me2, and H3K27ac, suggesting that hotspots are key enhancers 1,[7][8][9] . Similarly, intracellular and extracellular structures were also remodeled during adipogenesis when preadipocytes changed to round and lipid-laden adipocytes 4,10-12 . Within 24 h, MDI treatment rearranged the actin cytoskeleton from stress fibers to cortical structures in preadipocytes 13 . Thus, MDI dramatically rearranged both chromatin and cytoskeleton in this short period.
Previous studies on the anti-adipogenic effects of Wnt/β-catenin signaling focused on the inhibition of PPARγ activity by β-catenin [29][30][31][32][33][34] . However, how Wnt signaling prevents adipogenic hormones from derepressing and activating the transcription of Pparg2 during early adipogenesis (within 48 h after MDI treatment) remains unclear. Here, we showed for the first time that canonical Wnt signaling inhibits not only hotspot formation of Pparg2, but also cytoskeletal rearrangement in a β-catenin-dependent manner. These two events are regulated by the antagonistic actions of GR and β-catenin, and Wnt3a reinforces β-catenin function.
Chromatin immunoprecipitation (ChIP) and formaldehydeassisted isolation of regulatory elements (FAIRE) ChIP analyses were performed as described previously 38 . FAIRE analyses were performed using ChIP lysates (30 μg chromatin) as described previously 39 . Briefly, sonicated chromatin lysates were phase-separated by two rounds of phenol/chloroform extraction. Nucleosome-free DNA in the upper aqueous phase was obtained using ethanol precipitation. DNA was further treated with 10 μg RNase A and 20 μg proteinase K, and extracted using the QIAquick PCR purification kit (QIAGEN, Chatsworth, CA, USA). The isolated genomic DNAs were used for FAIRE-qPCR. The Ct value of a target gene in the isolated DNA sample of a ChIP or FAIRE experiment was normalized to the Ct value of the target gene in the input DNA (ΔCt = Ct sample −Ct input ). The percentage of input indicates the value of 100 × 2 ΔCt .
Western blot analyses, nuclear extraction, and reporter assay Western blot analyses were performed as described previously 35 . To obtain nuclear extracts, the cells were washed twice with ice-cold phosphate buffered saline (PBS), harvested, and then lysed with hypotonic buffer (20 mM Tris-HCl (pH 8.0), 10 mM NaCl, 0.2% NP-40, 10 mM β-glycerophosphate, 10 mM NaF, 1 mM Na 3 VO 4, and protease inhibitors) and incubated for 10 min on ice. The supernatant (cytosol extracts) was removed and the nuclear pellet was washed with hypotonic buffer and lysed with NETN buffer (20 mM Tris-HCl (pH 8.0), 140 mM NaCl, 0.5% NP-40, 10 mM β-glycerophosphate, 10 mM NaF, 1 mM Na 3 VO 4, and protease inhibitors), followed by 10 cycles of sonication (1 cycle; 30 s on, 30 s off). Nuclear extracts (supernatant) were obtained by centrifugation at 13,000 × g for 10 min at 4°C. The reporter plasmids, C/ EBP-Luc and GRE-Luc, contained the luciferase gene under the regulation of three copies of C/EBP binding sequences (GTTGCGCAAG) and one copy of glucocorticoid responsive element (GRE) (AGAA-CACTGTGTTCT), respectively. Reporter assays were performed using Lipofectamine reagent as described previously 40 . The pRL-TK plasmid encoding Renilla luciferase was cotransfected for normalizing transfection efficiency.

Immunofluorescence of F-actin
The cellular F-actin was stained with fluorescent phalloidin conjugates (25 μΜ) for 40 min at room temperature prior to Hoechst staining. The stained cells were observed under a Zeiss LSM510 inverted confocal microscope according to the manufacturer's instructions. F-actin structures in individual cells were categorized into three groups; S (stress) fiber, where F-actin stress fibers were observed both in the nucleus and cytoplasm; T (transition state) fiber, where F-actin stress fibers were observed in the cytoplasm but not in the nucleus; C (cortical structure), where F-actin stress fibers were observed neither in the nucleus nor in the cytoplasm, but were observed as cortical structures near the cellular membrane. Cells  in each treatment were observed and categorized into three groups.

Statistical analysis
All quantitative measurements were performed in at least three independent experiments. Two-tailed unpaired Student's t-tests were used to compare the data between controls and indicated experimental groups. *p-values < 0.05; **p-values < 0.01; ***p-values < 0.001 were considered statistically significant.

Wnt3a inhibits early induction of Pparg2
We treated 3T3-L1 preadipocytes with adipogenic hormones (MDI) in the presence of recombinant Wnt3a (Fig. 1a). MDI induced the protein levels of both PPARγ and C/EBPα, but reduced β-catenin proteins. Two ng/ml of Wnt3a was sufficient to block the induction of both PPARγ and C/EBPα proteins and lipid accumulation but increased β-catenin proteins (Fig. 1b, c, and S1A). Although Wnt3a repressed both PPARγ1 and PPARγ2 proteins, PPARγ2 is major target for C/EBPβ and GR in response to MDI (Fig. S1C). Early temporal treatment of Wnt3a (5 ng/ml, for 0-2 days) was sufficient to block the early and later processes of adipogenesis, whereas more Wnt3a (>25 ng/ml) was required to block adipogenesis during the later processes (4-6 days time points) ( Fig. 1d-g and S1B). Although late temporal treatment (for 4-6 days) of Wnt3a (25 ng/ml) increased the mRNA levels of Axin2, a Wnt target gene, it did not effectively reduce the mRNA and protein levels of Pparg2 and Cebpa (Fig. 1e, f). These results confirmed the previous findings that the early period (within 0-2 days) is more sensitive to Wnt3a inhibition than the late period 17 .
Wnt3a significantly reduced the nuclear protein level of GR, but not C/EBPβ during early adipogenesis, although Wnt3a did not reduce the total protein levels of either GR or C/EBPβ (Fig. 1h, i and S2A). Interestingly, Wnt3a did not reduce GR protein level in the nuclei of the 3T3-L1 cells treated with only Dex, suggesting that Wnt3a specifically reduced nuclear GR only during early adipogenesis (Fig. 1j). Consistently, Wnt3a reduced GR binding at two major hotspots (−0.3 or +2.6 kb) located near the Pparg2 transcription start site (TSS) (Fig. 1k). Although Wnt3a did not reduce C/EBPβ protein level, it inhibited the binding of C/EBPβ on the hotspots in Pparg2 (Fig. 1l). We selected seven genes (Acsl1, Hp, Tpcn2, Slc10a6, Krt13, Tsc22d3 and Megf9), the promoters (−5 to +1 kb from TSS) of which contain GR binding peaks, as identified in the ChIP-seq analyses 5 . We found that Wnt3a reduced GR occupancy in the promoter of all seven genes, which are also occupied by C/EBPβ; interestingly, Wnt3a also reduced C/EBPβ binding to these promoters

Wnt3a disrupts cooperation between C/EBPβ and GR on the Pparg2 promoter
To investigate whether the reduction in nuclear GR destabilized C/EBPβ binding to the Pparg2 promoter, we generated C/β-NIH cells that ectopically express FLAGtagged C/EBPβ (Fig. 2a). Dex alone (without IBMX) can induce Pparg2 expression in C/β-NIH cells, as IBMX is required for the induction of C/EBPβ (Fig. 2b). Furthermore, the constitutively expressed C/EBPβ can bind the Pparg2 promoter only when GR binds it in response to Dex treatment (Fig. 2c), and vice versa, GR binds to the Pparg2 promoter only in C/β-NIH cells but not in EV-NIH cells. These results confirmed that C/EBPβ and GR interdependently bind the Pparg2 promoter. Similar to Wnt3a, a GR antagonist, RU486, prevented not only GR but also C/EBPβ from binding to the Pparg2 promoter ( Fig. 2c).
In addition to the Pparg2 promoter, C/EBPβ and GR can bind to different sets of their own target sequences. Analyses of the luciferase reporter gene show that Wnt3a does not inhibit C/EBPβ to induce the luciferase gene regulated by C/EBP binding sites (Fig. 2d). Similarly, RU486, but not Wnt3a, prevented Dex from inducing the luciferase gene, which was under the regulation of the glucocorticoid response element (GRE) (Fig. 2e). These results suggest that Wnt3a blocks neither C/EBPβ nor GR binding to their consensus motifs but disrupts the cooperativity between C/EBPβ and GR on the promoter of Pparg2 by reducing GR level in the nuclei during early adipogenesis. We investigated whether Wnt3a also disrupted the cooperative binding of other hotspot TFs. We found that Wnt3a also reduced binding of STAT5, KLF4, and c-Jun to Pparg2 (Fig. 2f). The cooperative bindings of several TFs significantly increased the recruitment of enhancer-associated coregulators, p300 and CBP, to form enhanceosomes 8,41 and increased the accessibility of the chromatin structure of Pparg2 1 . As expected, Wnt3a blocked the recruitment of p300 and CBP on Pparg2 in the MDI-treated cells (Fig. 2g). FAIRE (that detects nucleosome-depleted regions in the genome) and H3 ChIP analyses showed that MDI treatment reduced histone occupancy at the Pparg2 promoter, but not in the presence of Wnt3a, suggesting that Wnt3a inhibits MDIinduced chromatin opening of Pparg2 (Fig. 2h, i).

Wnt3a prevents GR from inducing Tead4, a novel hotspot TF of Pparg2
Starick et al. showed that TEA domain transcription factors (TEADs) reinforced GR binding to a subset of GR target genes as a heterodimer binding partner of GR 42 . For the first time, we found that Dex is responsible for inducing the mRNA and protein levels of Tead4 in a GRdependent manner (Fig. 3a-e and S3A to S3C). Interestingly, Wnt3a completely blocked MDI from inducing the mRNA and protein levels of Tead4, but did not block Dex's ability to induce Tead4 (Fig. 3f, g). These results are consistent with the findings that Wnt3a inhibits GR function in MDI-treated cells with higher sensitivity than in Dex-treated 3T3-L1 cells (Fig. 1i, j). ChIP revealed that both GR and TEAD4 occupied their putative binding sites in Tead4 after MDI treatment (Fig. 3h, i). Interestingly, we found that GR bound to not only GRE (−0.9 kb) but also to TEAD binding elements (TBE) (+0.3 kb) in Tead4 (Fig. 3h, i), and that knockdown of Tead4 reduced GR binding on the TBE of Tead4 (Fig. 3j). These results suggest that GR and TEAD4 can cooperatively bind to the TBE of Tead4, and that Tead4 is the target of TEAD4 itself as well as GR. Wnt3a prevented both GR and TEAD4 from binding to the GRE and the TBE of Tead4 during early adipogenesis (Fig. 3i).
Furthermore, TEAD4 also bound to the Pparg2 promoter during early adipogenesis. We found that Tead4 knockdown reduced Pparg2 induction and adipogenesis ( Fig. 4a-d). Interestingly, Tead4 knockdown did not reduce GR and C/EBPβ protein levels but reduced their (see figure on previous page) Fig. 1 Effects of Wnt3a on early induction of Pparg2. a Post-confluent 3T3-L1 preadipocytes were induced to undergo adipogenesis by treatment with adipogenic hormones, IBMX (M), dexamethasone (D), and insulin (I) in the presence or absence of recombinant mouse Wnt3a (W3a). The treated cells were harvested at the indicated time points after induction. b Western blot analyses of 3T3-L1 cells using the indicated antibodies. 14-3-3γ was used as the loading control. Arrows indicate two PPARγ proteins (γ1, 54 kDa; γ2, 57 kDa). c, d Optical densities (510 nm) of Oil Red-O stained lipid in the 3T3-L1 cells at 6 days after the hormone treatment. The images of the Oil Red-O stained 3T3-L1 cells are shown in the Supplementary Fig. S1. e qRT-PCR analyses of Pparg2, Cebpa, and Axin2 mRNA levels, which were normalized to 18S rRNA levels as described previously 35 . f Western blot analyses of 3T3-L1 cells using the indicated antibodies. g Relative mRNA levels of Pparg2 to 18S rRNA levels. h Western blot analyses of 3T3-L1 cells using the indicated antibodies. Arrows indicate two active forms of C/EBPβ (36 and 38 kDa respectively). i, j Western analyses of nuclear extracts (NE) of 3T3-L1 cells treated with either MDI or Dex (D; 2 μM) in the presence or absence of W3a. Lamin C was used as the loading control for the nuclear proteins. The relative band intensities of GR, C/EBPβ (36 kDa, the lower band with a black arrowhead), and lamin C were determined using the ImageJ software from four independent western analyses (details in Supplementary Fig. S2A). k, l ChIP-qPCR analyses of GR or C/EBPβ occupancy on Pparg2 (-0.3 kb or +2.6 kb from TSS) in 3T3-L1 cells. qPCR data show mean ± S.E. All data were repeated at least three independent same or similar experiments. *p < 0.05, **p < 0.01, and ***p < 0.001 by Students' t-test; ns, not significant binding to the Pparg2 promoter (Fig. 4e, f), suggesting that TEAD4 reinforced GR binding not only to the Tead4 promoter, but also to the Pparg2 promoter, for strong induction of both Pparg2 and Tead4. Although TEADs are major transcription factors that convey the Hippo signal by recruiting their coactivators TAZ/YAP, TEAD4 recruited neither YAP nor TAZ on Pparg2 (Fig. 4g-j). We also found that TEAD4 binds on the promoter of other  i ChIP-qPCR analyses of histone H3 on the -0.3 kb region from TSS of Pparg2 in 3T3-L1 cells. qPCR data show mean ± S.E. All data were repeated at least three independent same or similar experiments. *p < 0.05, **p < 0.01, and ***p < 0.001 by Students' t-test MDI-induced genes, but do not recruit TAZ or YAP (Fig. S4A). Furthermore, TAZ knockdown did not block Wnt3a inhibitory effects suggesting that TAZ is not essential for the anti-adipogenic function of Wnt3a ( Fig. 4k and Fig. S4B to S4E). We showed for the first time that TEAD4 and TAZ/YAP oppositely regulate the expression of Pparg2. These findings indicated that TEAD4 and GR form a positive circuit for induction of both Tead4 and Pparg2. Thus, Wnt3a disrupted two mutually related positive circuits by blocking GR binding to the promoters of Tead4 and Pparg2 (Fig. 4l).

Overexpression of GR is sufficient for blocking the inhibitory effects of Wnt3a
We investigated whether overexpression of ectopic GR is sufficient to resume the MDI-mediated induction of . qPCR data show mean ± SE. All data were repeated at least three independent same or similar experiments. *p < 0.05, ** p < 0.01, and *** p < 0.001 by Students' t-test; ns, not significant Tead4 in Wnt3a-treated 3T3-L1 cells. We found that GR overexpression prevented Wnt3a from reducing the mRNA and protein levels of Tead4. These results confirmed that Wnt3a reduced the expression of Tead4 by limiting the GR nuclear protein level during early adipogenesis (Figs. 1i, 5a, b). Overexpression of ectopic TEAD4 further increased the mRNA and protein levels of Pparg2 in MDI-treated as well as untreated 3T3-L1 cells, confirming that TEAD4 is a positive regulator of Pparg2 (Fig. 5c, d). However, in TEAD4 overexpressing cells, Wnt3a still limited GR protein level in the nuclei and reduced MDI-mediated induction of Pparg2, suggesting that TEAD4 cannot substitute GR for Pparg2 induction (Fig. 5c-e). Interestingly, overexpression of GR reduced the amount of C/EBPβ protein, which limited the MDImediated induction of PPARγ (Fig. 5f, g). This finding is consistent with previous study, which showed that liganded GR inhibits cAMP-activated CREB, an important transcription factor required for the induction of Cebpb 43 . We overexpressed GR together with C/EBPβ in 3T3-L1  (Fig. 5h).
Although Wnt3a still reduced early induction of PPARγ, GR/Cβ-L1 cells maintained sufficiently high PPARγ protein level to enable completion of adipogenesis even in the presence of Wnt3a (Fig. 5h, i). ChIP analyses also showed with MDI in the presence or absence of W3a. H3 was used as the loading control for nuclear proteins. The relative band intensities of GR and H3 were determined using the ImageJ software from two independent western blot analyses. qPCR data show mean ± SE. All data were repeated at least three independent same or similar experiments. *p < 0.05, **p < 0.01, and ***p < 0.001 by Students' t-test; ns, not significant that in GR/Cβ-L1 cells, Wnt3a did not reduce GR bindings to the Pparg2 promoter (Fig. 5j). These results corroborated the observation that GR protein level remained high in the nuclei of GR/Cβ-L1 cells even in the presence of Wnt3a (Fig. 5k) and demonstrated that Wnt3a blocked MDI-mediated induction of both Tead4 and Pparg2 by limiting GR in nuclei during early adipogenesis.
β-Catenin is necessary for conveying the Wnt3a signal to the Pparg2 promoter Knockdown of β-catenin reduced the mRNA levels of its target genes such as Axin2 and Ccnd1 in 3T3-L1 cells (Fig. 6a, b) and completely blocked the inhibitory effects of Wnt3a on adipogenesis and induction of both Pparg2 and Tead4 (Fig. 6c-g). We found that Wnt3a did not reduce the nuclear amounts of GR proteins in shβ-cat-L1 cells (Fig. 6h). Consequently, in 3T3-L1 cells with βcatenin knockdown, Wnt3a did not block the occupancies of C/EBPβ and GR on the Pparg2 promoter (Fig. 6i) and did not inhibit the opening of chromatin structure (Fig. 6j, k). These results indicate that β-catenin is necessary for conveying the Wnt3a signal to the chromatin of Tead4 and Pparg2 by limiting the nuclear localization of GR.
Wnt3a blocks remodeling of actin cytoskeleton in a βcatenin-dependent manner β-Catenin is connected to the actin cytoskeleton via interactions with α-catenin and cadherin proteins in cytoplasm 26 . Upon exposure to adipogenic hormones, 3T3-L1 cells differentiate into round and lipid-laden adipocytes, accompanied by changes in the actin cytoskeleton from stress fibers to cortical structures 4 . Time course staining with fluorescent phalloidin, a probe specific for filamentous actin (F-actin), revealed that MDI reduced F-actin stress fibers in 50% cells from 4 h, whereas they disappeared in 86% cells within 48 h; instead, F-actin reorganized only at the cortical region. The amount of β-catenin protein decreased during the rearrangement of F-actin from stress fibers to cortical structures. Wnt3a prevented MDI from reducing βcatenin protein and disrupting F-actin stress fibers (Fig. 7a, b). However, this was not observed in shβ-cat-L1 cells, suggesting that β-catenin is required for Wnt3a to prevent F-actin rearrangement (Fig. 7c). However, we observed that active S37A-β-catenin which is constitutively present in the nuclei, is not sufficient to elicit the inhibitory effects of Wnt3a on remodeling of chromatin and cytoskeleton during early adipogenesis although at late phase of adipogenesis, ectopic expression of S37A-β-catenin can reduce the mRNA and protein levels of Pparg2 even in the absence of Wnt3a (Fig. 7d-j and S5A to S5C).
As Wnt3a reduced nuclear GR protein level in a βcatenin-dependent manner (Fig. 6h), we investigated whether GR is necessary for MDI to rearrange F-actin. In shGR-L1 cells with GR knockdown, MDI could neither rearrange F-actin nor induce the mRNA level of Pparg2 (Fig. 8a, b). GR knockdown reduced MDI-induced binding of C/EBPβ to the Pparg2 promoter (Fig. 8c). In contrast, in GR/Cβ-L1 cells without MDI treatment, F-actin stress fibers were observed only around the peripheral region but not in the nuclei of 3T3-L1 cells (Fig. 8d). Wnt3a did not recover F-actin stress fibers in GR/Cβ-L1 cells. Like 3T3-L1 cells, in C3H10T1/2, mouse mesenchymal stem cells, both Dex and MDI increased the mRNA and protein levels of TEAD4 and PPARγ which were also inhibited by Wnt3a (Fig. 8e, f). Furthermore, Wnt3a prevented MDI from rearranging F-actin stress fibers in C3H10T1/2 cells (Fig. 8g). These findings suggest that during early adipogenesis, GR is necessary both for the rearrangement of F-actin and hotspot formation on the Pparg2 promoter, and that Wnt3a blocks these two events by limiting the nuclear level of GR in a β-catenindependent manner (Fig. 8h).

Discussion
The findings that treatment of Dex followed by IBMX was sufficient for adipogenesis, but that IBMX treatment followed by Dex treatment did not recapitulate this effect suggested that Dex primed preadipocytes to a novel commitment state for adipogenesis 44 . Adipogenesis of mouse embryonic fibroblasts (MEFs) isolated from GR (Nr3c1) knockout and GR dimerization-defective mutant mice was impaired 45 . Knockdown of GR or omission of Dex from MDI did not induce Pparg2 in 3T3-L1 cells 8,46 . These previous studies indicated that GR is essential for Pparg2 induction.
Using extensive genome-wide profiling of 15 TFs at 4 h after MDI treatment, Mandrup and colleagues demonstrated that during early adipogenesis, hotspots cooccupied by more TFs recruited more coactivators such as p300/CBP to constitute super-enhancers 1 . Furthermore, enhancers with large numbers of TFs are more sensitive to small changes in TF concentration compared to those with smaller numbers of TFs [47][48][49] . This is consistent with our findings that reduction in nuclear GR level by β-catenin sensitively changes super-enhancer formation on Pparg2 and other genes (Fig. S2B). Our findings that GR amplifies MDI signals to Pparg2 by inducing TEAD4, a novel TF for Pparg2, highlighted that changes in GR activity affect Pparg2 expression (Fig. 3).
ChIP-exo sequencing using GR antibody in IMR90 cells revealed that TEAD4 and GR co-occupied GR target genes as a heterodimer and that TEAD knockdown decreased the expression of several GR target genes 42 .
Biochemical CAP-SELEX analyses which identify cooperative interactions between TF pairs and heterodimeric DNA motifs, revealed that TEAD4 is the most common partner that cooperatively binds diverse DNA sequences with 32 TFs among 100 tested TFs. TEAD4 and its partner TFs recognize composite sequences that were considerably different from the individual TF motifs, suggesting that in vivo functions of TEAD4 may  Fig. 6 Effects of β-catenin knockdown. a-k 3T3-L1 preadipocytes were infected with a lentivirus encoding shRNA against mouse β-catenin (shβcat-L1 cells) or control shRNA (shCtrl-L1 cells). a Western blot analyses showing β-catenin level. b qRT-PCR analyses showing relative mRNA levels of β-catenin, Ccnd1, and Axin2 to 18S rRNA levels. c-k The shCtrl-L1 or shβ-cat-L1 cells were induced to undergo adipogenesis by treating with adipogenic hormones for the indicated time points in the presence or absence of W3a (5 ng/ml). c Images and optical densities (510 nm) of Oil Red-O stained lipid. Scale bars, 200 μm. d, f Relative mRNA levels of Pparg2 and Tead4 to 18S rRNA levels. e, g Western blot analyses using the indicated antibodies. h Western blot analyses of nuclear extracts (NE) of the shCtrl-L1 or shβ-cat-L1 cells using the indicated antibodies. Lamin C was used as the loading control for nuclear proteins. i, k ChIP-qPCR analyses of C/EBPβ, GR, or H3 occupancy on the -0.3 kb region from TSS of Pparg2. j FAIRE-qPCR analyses on the -0.3 kb region from TSS of Pparg2. qPCR data show mean ± SE. All data were repeated at least three independent same or similar experiments. *p < 0.05, **p < 0.01, and ***p < 0.001, † p = 0.065 by Students' t-test; ns, not significant encompass diverse biological functions depending on its partner TFs 50 . This is the first study to show that TEAD4 cooperatively binds to Pparg2 together with other hotspot TFs, and that Tead4 knockdown reduced Pparg2 expression. Since GR and TEAD4 cooperatively bind and induce Tead4, GR is a key TF that persistently drives expression of both Tead4 and Pparg2. Wnt3a disrupted two mutually related positive circuits by limiting GR binding to the promoters of these two genes.
Goentoro and Kirschner have showed that fold-change, but not the absolute level of β-catenin, regulates Wnt signaling in an experimental model of Xenopus dorsalanterior development 51 . Our findings that the nuclear form of S37A-β-catenin failed to deliver Wnt3a signal imply that Wnt3a disrupts hotspot formation and cytoskeletal rearrangement but not by increasing the coactivator activity of β-catenin in the nuclei (Fig. 7d-j). Similar to β-catenin, the TAZ coactivator activated by the canonical Wnt pathway, inhibits adipogenesis by inhibiting PPARγ protein 52,53 . We found that the protein level of TAZ decreased during early adipogenesis, but not in the presence of Wnt3a, suggesting that TAZ can be a mediator that delivers the Wnt signal to Pparg2. However, our finding that TAZ knockdown did not block Wnt3a inhibitory effects indicates that TAZ is not essential for the anti-adipogenic function of Wnt3a (Fig. 4k).
Unlike lipogenesis in other cells such as hepatocytes and myocytes, adipogenesis requires dramatic cytoskeletal remodeling of F-actin stress fibers to cortical actin structures, which is required to hold a large lipid vacuole in the center, relocate the nucleus and other organelles, and attain a round shape. Cytoskeletal remodeling starts within 24 h after adipogenic hormone treatments 13 , suggesting that cytoskeletal remodeling is followed by complete induction of lipogenic genes as a feed-forward mechanism. These nuclear and cytoplasmic events should interdependently regulate each other to prevent metabolic and structural catastrophes during adipogenesis 54,55 . In addition to adipogenic hormones, the stiffness of extracellular matrices (ECM) is involved in commitment for adipogenesis. Primary preadipocytes embedded in stiffer matrices showed reduced rates of adipogenesis 10 . ECM stiffness increases tissue tension, which leads to increased actin and myosin fiber and cell stretching 56,57 . Interestingly, mesenchymal stem cells exposed to mechanical strain show increase in β-catenin level and could not differentiate into adipocytes, suggesting that Wnt/βcatenin conveys signals from intracellular tension to the nuclei 58 . In agreement with this result, we found that Wnt signal could not prevent MDI from changing F-actin stress fibers to cortical F-actin structures during adipogenesis in the absence of β-catenin. Interestingly, Wnt3a reduced GR level in the nuclei during the early phase of adipogenesis, but not in Dextreated preadipocytes, suggesting that the antagonistic effect of Wnt/β-catenin on GR is specific for adipogenesis (Figs. 1i, j, 6h). Extensive studies have revealed that the chaperone complex and intact cytoskeleton are required for the nuclear transport of GR 59 . Furthermore, GR is connected to actin filaments through HSP90, a main component of GR-chaperone complex suggesting that the nuclear transport of GR and cytoskeletal rearrangement are closely related 60 . Upon ligand binding, the GRchaperone complex recruits a motor protein dynein to form the liganded GR-HSP90-FKBP52-dynein complex, which is able to move along the microtubules through the nuclear pore complex (NPC) to the nucleus 61,62 . However, when the cytoskeleton is disrupted like adipogenesis, liganded GR simply diffuses in and out of the nucleus mainly via importins, exportins, and the RanGTPase (see figure on previous page) Fig. 7 Cytoskeletal rearrangement during early adipogenesis. 3T3-L1 cells were treated with MDI for the indicated time points in the presence or absence of W3a (5 ng/ml) as described in Fig. 1a. a Confocal microscopic images of cellular filamentous actin (F-actin) and β-catenin (upper panel). Factin structures in individual cells were categorized into three groups. S (stress fiber), where F-actin stress fibers were observed in both nuclei and cytoplasm; T (transition status), where F-actin stress fibers were observed in the cytoplasm but not in the nucleus; C (cortical structure), where F-actin stress fibers were observed neither in the nucleus nor in the cytoplasm, but F-actin was observed near the cellular membrane. Cells  in each treatment were observed and categorized into three groups. The graph indicates the percentage of cells in each category (lower panel). b Western blot analyses of 3T3-L1 cells using the indicated antibodies. c Confocal microscopic images of cellular F-actin and β-catenin (upper panel). The graph indicates the percentage of cells in each category (S, T, and C described in Fig. 7a) (lower panel). d-j 3T3-L1 preadipocytes were infected with retrovirus encoding HA-tagged S37A-β-catenin (S37A-β-L1 cells) or empty vector (EV-L1 cells) as a control. d Western blot analyses of the EV-L1 cells or the S37A-β-L1 cells using anti-HA and 14-3-3γ antibodies. e Confocal microscopic images of the EV-L1 cells or the S37A-β-L1 cells immunostained with either anti-β-catenin antibody or anti-HA antibody. The cells were treated with W3a (5 ng/ml) for 24 h. The nuclei were stained with Hoechst 33258 (blue). f qRT-PCR analyses showing relative mRNA levels of Axin2 and Ccnd1 to 18S rRNA. g-j The EV-L1 cells or the S37A-β-L1 cells were treated with MDI for the indicated time points in the presence or absence of W3a (5 ng/ml). g Relative mRNA levels of Pparg2 and Axin2 to 18S rRNA. h Western blot analyses using the indicated antibodies. i ChIP-qPCR analyses of C/EBPβ or GR occupancy on the -0.3 kb region from TSS of Pparg2. j Confocal microscopic images of the cells immunostained with fluorescent phalloidin conjugates (green) and with Hoechst 33258 (blue) (left panel). Graph indicating the percentage of cells in each category (S, T, and C described in Fig. 7a) (right panel). qPCR data show mean ± SE. All data were repeated at least three independent same or similar experiments  Fig. 8 Effects of GR on cytoskeletal rearrangement. The shCtrl-L1 or shGR-L1 cells were induced to undergo adipogenesis by treating with adipogenic hormones for the indicated time points in the presence or absence of W3a (5 ng/ml). a Confocal microscopic images of F-actin stress fibers (left panel). The graph indicates the percentage of cells in each category (S, T, and C described in Fig. 7a) (right panel). b qRT-PCR analyses of Pparg2 mRNA levels to 18S rRNA levels. c ChIP-qPCR analyses of C/EBPβ or GR occupancy on the -0.3 kb region from TSS of Pparg2. d Confocal microscopic images of cellular F-actin stress fibers (left panel) in the EV/EV-L1 or GR/Cβ-L1 cells. The graph indicates the percentage of cells in each category (S, T, and C described in Fig. 7a) (right panel). e-g C3H10T1/2 cells were treated with Dex (2 μM) or MDI for the indicated time points in the presence or absence of W3a (5 ng/ml). e qRT-PCR analyses of Tead4 and Pparg2 mRNA levels to 18S rRNA levels. f Western blot analyses showing TEAD4 and PPARγ protein levels. g Confocal microscopic images of cellular F-actin stress fibers (left panel) in C3H10T1/2 cells. The graph indicates the percentage of cells in each category (S, T, and C described in Fig. 7a) (right panel). h Schematic diagram showing the inhibitory effects of Wnt3a/βcatenin on positive circuits of GR-TEAD4-PPARγ2 and cytoskeletal remodeling during early adipogenesis (details in Results). qPCR data show mean ± SE. All data were repeated at least three independent same or similar experiments. ***p < 0.001 by Students' t-tests system 59,63 . The mechanism via which the liganded-GR translocates into the nucleus during adipogenic cytoskeletal rearrangement remains unknown. In contrast, βcatenin harbors armadillo repeats, which are similar to the importin-β HEAT repeats [64][65][66] . Therefore, β-catenin can pass through NPC like importin and equilibrates between the nucleus and the cytoplasm by passive diffusion. Therefore subcellular distribution of β-catenin in the nuclei, cytoplasm, and the membrane can be determined from its location and the levels of diverse interacting complexes, namely, the TCF/LEF transcription factors in the nucleus, APC and AXIN in the cytoplasm, and cadherin complex in the membrane 67 . It remains to be investigated whether GR and β-catenin compete with each other for nuclear translocation during adipogenic cytoskeletal rearrangement.

H G
In conclusion, this study provides insights into an intriguing question regarding chromatin remodeling and actin rearrangement, which are interdependently regulated during adipogenesis. Our findings that GR is necessary for the rearrangement of both cytoskeleton and chromatin, and that canonical Wnt3a inhibited both processes in a βcatenin-dependent manner, suggest that rearrangements of both chromatin and cytoskeleton are related in ways that involve the antagonistic activities of GR and β-catenin, and that Wnt3a reinforced β-catenin function.