Pygo2 Regulates Adiposity and Glucose Homeostasis via β-Catenin–Axin2–GSK3β Signaling Pathway

Adipose tissue plays crucial roles in the regulation of energy metabolism through storing and releasing fuel as an energy depot and through secreting hormones and cytokines as an endocrine organ (1,2). Obesity can increase the risk of diabetes, hypertension, cardiovascular diseases, and several types of cancers and has become a major public health problem in industrialized countries (1,3). It is accepted that obesity results from a positive imbalance between energy intake and output. Two potential mechanisms explain the development of metabolic complications resulting from obesity, including 1) lipotoxicity, where excess lipids overflow from an incompetent and dysfunctional adipose tissue (4), and 2) specific patterns of adipokines and proinflammatory cytokines released from dysfunctional adipose tissue (2). These raise a new question: why do adipocytes change from a healthy storage tissue to one that seemingly promotes the metabolic syndrome? Indeed, the “adipose expandability hypothesis” (5,6) suggests that the total adipose tissue in an individual may have a maximal fixed capacity for safely storing fat. This capacity may be determined by multiple factors, including the number of pre-existing preadipocytes, genetic programs of adipogenesis, programs of vasculogenesis, or functionality of other cellular components within the adipose tissue. Therefore, understanding the molecular mechanisms that regulate the above factors might provide novel strategies to prevent obesity and related metabolic complications. The study of adipogenesis has been greatly facilitated by the development of immortalized preadipocyte lines, such as 3T3-L1 and 3T3-F442A cells (7–10), that differentiate into adipocytes and recapitulate many metabolic and endocrine functions of adipocytes in vivo. In response to stimulators of adipogenesis, two transcription factors, C/EBPβ and C/EBPδ, are transiently and rapidly induced. These proteins stimulate expression of key adipogenic transcription factors, namely, C/EBPα and peroxisome proliferator–activated receptor γ (PPARγ), which then work together to induce the expression of the adipocyte-specific genes necessary to produce the adipocyte phenotype (11,12). Identification of new factors that regulate progression through this adipogenic program is crucial for us to gain new insight into this biological process.

Wnt/β-catenin signaling is crucial in cell proliferation and differentiation throughout vertebrate embryonic development and tumorigenesis (13–15). β-Catenin plays a central role as a transcriptional coactivator in this process. In the absence of upstream signaling stimulation, cytoplasmic β-catenin is phosphorylated by glycogen synthase kinase 3β (GSK3β) in a complex containing Axin and the tumor suppressor adenomatous polyposis coli and is targeted for ubiquitin-mediated proteasomal degradation. Stimulation by upstream signaling leads to the inhibition of phosphorylation and degradation of β-catenin, which subsequently translocates into the nucleus and binds to the lymphoid-enhancing factor 1/T-cell factor (Lef1/Tcf) family of transcription factors. In turn, the β-catenin–Lef1/Tcf complex regulates the expression of downstream target genes involved in diverse cellular processes (16,17). Previous studies have shown that Wnt/β-catenin signaling could suppress adipocyte differentiation through inhibition of C/EBPα and PPARγ (18,19). Although a subset of Wnt/β-catenin signaling components have shown to be regulated during induction of adipogenesis (19–22), additional Wnt/β-catenin components that govern Wnt/β-catenin activity in this process have yet to be defined. Moreover, as β-catenin usually acts as a transcriptional activator, the mechanism by which β-catenin signaling blocks C/EBPα and PPARγ gene expression is still not clear.

Drosophila genetic studies have identified Pygopus (Pygo) as a nuclear component of Wnt/β-catenin signaling (23,24). Pygo proteins (Pygo1 and Pygo2 in mammals) are thought to promote β-catenin-Lef1/Tcf transcriptional activation through the regulation of β-catenin nuclear retention and/or by facilitating β-catenin to recruit coactivators (25–31). Recent studies have revealed that Pygo2 mediates β-catenin activity in a gene- and tissue-dependent manner in mammalian cells (31,32). We have previously reported the function of Pygo2 in mammary gland development (31), breast cancer stem cells (30), and breast cancer chemoresistance (33). In the current study, we showed that Pygo2 is a newly identified activator of Wnt/β-catenin signaling in the regulation of adipocyte differentiation. Furthermore, we identified a mechanism that mediates the inhibition of C/EBPα and PPARγ expression by Pygo2 and β-catenin during adipogenesis. Our study reveals an association between C/EBPα/PPARγ expression and Wnt/β-catenin activation, which highlights the role of Wnt/β-catenin in the regulation of adiposity and glucose homeostasis.

The generation and genotyping of null and floxed Pygo2 alleles were performed as previously described (34). PDGFRα-Cre mice were purchased from The Jackson Laboratory (stock no. 013148) and then were back-crossed into the FVB background for more than six generations. For adipocyte precursor–specific deletion of Pygo2, PDGFRα-Cre mice were crossed with Pygo2^flox/flox mice to generate PDGFRα-Cre;Pygo2^flox/+ mice, which were then intercrossed to generate PDGFRα-Cre;Pygo2^flox/flox (Pygo2^pre−/−) mice. Genetically obese (ob/ob) mice were purchased from Shanghai SLAC Laboratory Animal Co., Ltd., and the origin is 000632-B6.Cg-Lep^ob of The Jackson Laboratory. Animals were maintained on a 12-h light/12-h dark cycle in a temperature-controlled barrier facility with free access to water and food. Age-matched male mice were used for all experiments for uniformity and to avoid the hormone impact of females. Animal studies were approved by the Institutional Animal Care and Use Committee of Xiamen University.

For glucose tolerance test (GTT), we deprived the mice of food for 14 h and then injected them intraperitoneally with 2 g glucose per kilogram body weight. For insulin tolerance test (ITT), mice were deprived of food for 4 h with ad libitum access to water and then they were injected intraperitoneally with 0.75 units human insulin per kilogram body weight. We collected blood samples and measured the glucose concentration with a glucometer (One Touch Ultra; LifeScan Inc., Milpitas, CA). The level of serum insulin was determined using a mouse ELISA kit (ALPCO, Salem, NH).

The body compositon of mice was performed on a small animal composition analyzer (MesoMR21-60H-I; Shanghai Niumag Corporation, Shanghai, China) with the following parameters: SFO1 (MHz) = 21.77, P1 (us) = 11, P2 (us) = 22, and SW (KHz) = 20. MR images were acquired on a 0.55 T MRI Instrument (MesoMR23-060H-I; Shanghai Niumag Corporation). The T1 weighted imaging parameters were customized as follows: SFO1 (MHz) = 23.31, FOV Read (mm) = 100, FOV Phase (mm) = 100, Matrix = 256 × 256, TR (ms) = 400, TE (ms) = 18, eight slices, and a slice thickness of 2 mm with in-plane resolution of 0.4 × 0.4 mm (voxel size). Mice were anesthetized with isoflurane to ensure stability throughout the experiment.

The white adipose tissue (WAT) was from inguinal adipose depots of 12-week-old mice and human subjects that underwent elective thigh surgery. We washed and minced WAT with PBS, digested for 1 h at 37°C with Type II collagenase (C6685-1G; Sigma-Aldrich) solution, passed the mixture through a nylon filter (70 μm) to remove undigested materials, and centrifuged at 1,200 rpm for 10 min at 4°C. We recovered the floating mature adipocytes and the pelleted stromal vascular fraction (SVF) respectively. Human adipose tissue was collected after informed consent and in agreement with the Institutional Review Board of The First Affiliated Hospital of Xiamen University.

The lentiviral pBobi vector was used to express hemagglutinin (HA)-Pygo2, β-catenin, dnTCF4 (dominant-negative TCF4, a β-catenin binding domain deletion mutant), and Axin1. Axin2 and Axin2-ΔGSK3β (Axin2 mutant without GSK3β binding domain 353–410) were cloned into the lentiviral vector pLV-CS2.0. The lentiviral pLL3.7 vector was used to express short hairpin RNA (shRNA) directed against mouse Pygo2, Axin2, or a nonsilencing scrambled control sequence. The target sequences were provided in Supplementary Table 1. β-Catenin knockout or Pygo2 knockout 3T3-L1 cells were constructed with lentiCRISPRv2 vector by CRISPR/Cas9 technique. The target site was designed as follows: β-catenin 5′-CAGCGTCAAACTGCGTGGAT-3′ and Pygo2 5′-CCTGCGCCCCCCACTTTAG-3′. The generation of the lentivirus was performed by cotransfecting pBobi, pLV-CS2.0, pLL3.7, or lentiCRISPRv2 carrying the expression cassette with helper plasmids pVSV-G and pHR into HEK 293T cells using Turbofect reagent (Thermo Fisher Scientific, Waltham, MA). The viral supernatant was collected 48 h after transfection. 3T3-L1 cells were infected with viral supernatants containing 10 mg/mL Polybrene at 40–50% confluency and were given fresh medium 24 h later.

Western blotting was performed according to standard procedures. The following antibodies were used: Pygo2, a rabbit polyclonal antibody used as previously described (30); p-Ser96-Snail, a rabbit polyclonal antibody raised against a synthetic peptide (amino acids 89–105 with p-Ser96 modified); and p-Ser184-C/EBPβ antibody (gift from Xi Li, Biology Science Institutes, Chongqing Medical University, Chongqing, China) (35). The other commercially available antibodies are listed in Supplementary Table 2.

Total RNA was isolated from cells or adipose tissues using Tripure reagent (Roche, Basel, Switzerland). First-strand cDNA was synthesized using a HiFi-MMLV cDNA Kit from Beijing CoWin Biotech (CWBIO, Beijing, China). UltraSYBR Mixture (CWBIO) was used for the real-time quantitative PCR (qPCR) reaction, and the expression levels were quantified using the ΔΔCT (where CT is threshold cycle) method. Primer sequences are provided in Supplementary Table 3.

3T3-L1 cells in six-well plates were transfected with the indicated plasmids using Lipofectamine 3000 (Invitrogen, Life Technology, Waltham, MA). The total amount of plasmid DNA transfected was made equivalent by adding empty vectors. Cells were harvested after 48 h and processed for luciferase and β-galactosidase assays, and data were normalized to β-galactosidase levels.

Chromatin immunoprecipitation (ChIP) was performed using a ChIP Assay Kit according to its biotechnology protocol (Upstate). In brief, chromatin samples were prepared and immunoprecipitated with IgG, C/EBPβ antibody (sc-150; Santa Cruz Biotechnology), and protein A/G beads. Purified DNA was subjected to real-time qPCR with primers flanking the C/EBPα and PPARγ promoter sequences of the C/EBPβ binding sites. Supplementary Table 4 lists the primer sequences.

Pieces of epididymal WAT from the mice were fixed in 10% neutral buffered formalin, dehydrated in ethanol, embedded in paraffin, and cut at a thickness of 4 μm. Sections were deparaffinized, rehydrated, and stained with hematoxylin-eosin. The adipocyte size was quantified using Image-Pro Plus software, and at least 600 cells per sample were measured.

Each experiment was repeated at least three times. Representative experiments are shown unless stated otherwise. Data were analyzed using the GraphPad software package. Data are presented as the means ± SD (in vitro experiments) or SEM (in vivo experiments) and were analyzed by unpaired two-tailed Student t test. P < 0.05 was considered statistically significant. All data points were included in the final analysis, without exclusions. No randomization was used and the experimenter was blinded to genotypes.

To investigate a potential role of Pygo proteins in adipogenesis, we first analyzed the Pygo1 and Pygo2 protein expression levels during differentiation of 3T3-L1 preadipocytes into mature adipocytes. Interestingly, Western blotting demonstrated that the Pygo2 protein levels progressively declined in contrast to the levels of the adipocyte differentiation marker adipocyte fatty acid binding protein 4 (FABP4) after hormonal stimulation, whereas Pygo1 expression did not change significantly (Fig. 1A). Consistently, semi-qPCR analysis revealed a similar expression pattern for Pygo1 and Pygo2 during differentiation (Fig. 1B). We next examined the distribution of Pygo2 mRNA in mouse adipose tissue. As shown in Fig. 1C, Pygo2 mRNA was expressed in the SVF of the tissues containing committed preadipocytes and was essentially absent in mature adipocytes in subcutaneous adipose depots. A parallel experiment was also conducted on a human sample, and a result similar to that in mouse adipose tissue was achieved (Fig. 1D). We also examined the expression levels of Pygo1 and Pygo2 in genetically obese ob/ob mice. The expression level of Pygo2, but not Pygo1, was decreased in the WAT of ob/ob mice compared with that in wild-type (WT) mice (Fig. 1E). These data suggest that Pygo2 may play a role during adipocyte differentiation.

The expression pattern of Pygo2 in preadipocytes and adipose tissue suggested that overexpression of Pygo2 might impair the differentiation process. To test this, 3T3-L1 cells were infected with a lentivirus carrying mouse Pygo2 or a control lentivirus. These cells were then induced to differentiate into mature adipocytes in the presence of the full differentiation cocktail (MDI) for 7 days. As shown in Fig. 2A, the efficient expression of Pygo2 during cell differentiation was confirmed by immunoblotting using anti-HA and Pygo2 antibodies. As expected, 3T3-L1 preadipocytes infected with a control lentivirus underwent efficient differentiation, and fat accumulation was visualized by staining lipids with Oil Red O. In contrast, Pygo2 overexpression effectively inhibited the accumulation of lipid droplets (Fig. 2B), concomitant with reduced expression levels of C/EBPα, PPARγ, and FABP4 (Fig. 2C). These data indicate that Pygo2 may negatively regulate adipocyte differentiation. Therefore, knockdown of Pygo2 in preadipocytes may enhance the differentiation process. To test this hypothesis, we infected 3T3-L1 preadipocytes with lentivirus expressing Pygo2 shRNAs or a control virus, and the infected cells were induced by MDI for 4 days. As shown in Fig. 2D and E, knockdown of Pygo2 significantly accelerated the differentiation process when examined by both Oil Red O staining and the expression of differentiation markers. When the infected cells were allowed to proliferate to confluence and maintained without MDI treatment for 2 weeks, knockdown of Pygo2 was found to undergo spontaneous adipogenesis, whereas the control cells failed to differentiate (Fig. 2F). These observations further support our hypothesis that Pygo2 negatively regulates adipocyte differentiation.

Previous studies have indicated that Pygo2 promotes β-catenin activity in a tissue- and gene-dependent manner in mammals. To directly assess whether Pygo2 facilitates β-catenin signaling in 3T3-L1 preadipocytes, we at first performed Lef1/Tcf luciferase reporter assays. Sole expression of either β-catenin or Pygo2 in 3T3-L1 cells stimulated the activity of SuperTOPFlash containing seven optimal Lef1/Tcf binding sites. Coexpression of β-catenin and Pygo2 resulted in a further increase in SuperTOPFlash activity (Fig. 3A). In contrast, knockdown of Pygo2 in 3T3-L1 cells resulted in significantly lower SuperTOPFlash activity (Fig. 3B). Consistently, Axin2, a constitutive downstream target of Wnt/β-catenin signaling that steadily expresses in almost all Wnt/β-catenin–activated cells, showed increased or reduced expression in both mRNA and protein levels upon Pygo2 overexpression or knockdown (Fig. 3C and D). Wnt/β-catenin signaling is rapidly suppressed upon induction of 3T3-L1 cell differentiation in response to 1-methyl-3-isobutylxanthine (IBMX) (19,36). We investigated whether knockdown of Pygo2 substitutes for the IBMX function. As shown in Fig. 3E and F, when cells could not be effectively induced to differentiate by omitting IBMX from the differentiation cocktail (DI), knockdown of Pygo2 led to higher lipid accumulation and expression of the adipogenic markers than in control. Moreover, we found that overexpression of dnTCF4 could not further enhance adipogenesis in the presence of Pygo2 shRNA under the DI condition (Fig. 3G and H). Our results demonstrate that Pygo2 can inhibit differentiation of 3T3-L1 cells through mediation of β-catenin signaling.

GSK3β is a regulator that phosphorylates nuclear transcription factors such as C/EBPβ and Snail to promote adipogenesis (37–39). Similar to Axin1, Axin2 is a cytoplasmic protein and contains a homologous GSK3β binding domain (40). Therefore, we reasoned that Axin2 might physically interact with GSK3β. To examine this possibility, we cotransfected Myc-tagged Axin2 or Axin2-ΔGSK3β and HA-tagged GSK3β to perform coimmunoprecipitation experiments. As shown in Supplementary Fig. 1A, Axin2, but not Axin2-ΔGSK3β, interacted with GSK3β. Next, we examined whether Axin2 directly associates with GSK3β. An in vitro pulldown assay was performed, and it demonstrated direct binding between Axin2 and GSK3β (Supplementary Fig. 1B). Finally, we investigated whether the endogenous proteins interact by performing immunoprecipitation experiments in 3T3-L1 cells. As shown in Supplementary Fig. 1C, immunoprecipitation of GSK3β indeed pulled down Axin2. We conclude that there is a direct interaction between Axin2 and GSK3β in preadipocytes. This result implies that Axin2 may also keep GSK3β in the cytoplasmic compartment by a retention mechanism. To evaluate this model directly, we transfected Myc-tagged Axin2 or Axin2-ΔGSK3β in 3T3-L1 cells induced by MDI for 36 h, and GSK3β nuclear localization was assessed. As shown in Fig. 4A, GSK3β localized to both the cytosolic and nuclear compartments in control cells, and the introduction of Axin2 dramatically reduced the nuclear distribution of GSK3β, whereas Axin2-ΔGSK3β failed to exert this function. Next, we investigated whether Pygo2 regulates the cytoplasmic-nuclear distribution of GSK3β via the β-catenin–Axin2 cascade during adipogenesis. To this end, we overexpressed β-catenin or dnTCF4 in 3T3-L1 cells induced by MDI for 36 h. As shown in Fig. 4B and C, overexpression of β-catenin resulted in increased cytoplasmic Axin2 and GSK3β but decreased nuclear GSK3β. In contrast, overexpression of dnTCF4 resulted in decreased cytoplasmic Axin2 and GSK3β but increased nuclear GSK3β. When we knocked down Pygo2 expression in 3T3-L1 cells and performed the same experiment, results similar to those achieved with the overexpression of dnTCF4 were observed (Fig. 4D). Because GSK3β phosphorylates C/EBPβ and Snail in nucleus, we investigated the impact of Wnt/β-catenin signaling and Pygo2 on C/EBPβ and Snail. Our results revealed that overexpression of β-catenin in 3T3-L1 cells resulted in decreased phosphorylation on Ser184 of C/EBPβ and Ser96 of Snail (Fig. 4E), whereas overexpression of dnTCF4 or knockdown of Pygo2 resulted in increased phosphorylation on the same sites (Fig. 4F and G). Consequently, activation of Wnt/β-catenin signaling by β-catenin overexpression resulted in decreased occupancy, whereas suppression of Wnt/β-catenin signaling by dnTCF4 overexpression or Pygo2 knockdown resulted in increased occupancy of the C/EBPβ protein on both the C/EBPα and PPARγ promoters (Fig. 5A–C). In parallel, expression of the Snail protein, but not mRNA, was either upregulated or downregulated by activation or inhibition of Wnt/β-catenin signaling in the above cells, respectively (Fig. 5D–F).

Finally, we assessed whether Wnt/β-catenin signaling and Pygo2 regulates adipogenesis via Axin2. This time, we knocked out β-catenin or Pygo2 by CRISPR/Cas9 technique with or without overexpression of Axin2 or Axin2-ΔGSK3β to detect whether the above functions driven by β-catenin or Pygo2 could be rescued by Axin2. The results showed that simultaneous overexpression of Axin2 significantly compromised the β-catenin or Pygo2 knockout–induced increase in nuclear translocation of GSK3β, phosphorylated C/EBPβ (Fig. 6A and B), increase in C/EBPβ occupancy on the PPARγ promoter (Fig. 6C and D), increase in adipogenesis (Fig. 6E), and decrease in Snail protein expression (Fig. 6A and B), whereas Axin2-ΔGSK3β failed to exert this function. Similar to Axin2, Axin1 is a cytoplasmic protein and contains a homologous GSK3β binding domain. We assessed whether Axin1 exerts a similar function. Indeed, simultaneous overexpression of Axin1 to the level that could compensate for the reduction of Axin2 also significantly compromised the Pygo2 knockout–induced increase in phosphorylated C/EBPβ, increase in C/EBPβ occupancy on the PPARγ promoter, and decrease in Snail protein expression (Supplementary Fig. 2A and B). To further confirm that Pygo2 inhibits adipogenesis through Axin2 in the opposite direction, Pygo2 was overexpressed and Axin2 was knocked down to the control level. As shown in Fig. 6F and G, Axin2 knockdown could significantly restore the impaired adipogenesis by Pygo2 overexpression. Together, these results functionally establish a relationship between Pygo2 and the Wnt/β-catenin signaling–Axin2 cascade during adipogenesis. From these data, we propose the model for this mechanism as exhibited in Fig. 6H.

To further investigate the role of Pygo2 in adipogenesis in vivo, primary mouse embryonic fibroblasts (MEFs) were prepared from WT and Pygo2^−/− mouse embryos at embryonic day 13.5, cultured to confluence, and then maintained for 3 weeks. The results showed that whereas WT MEFs failed to differentiate into adipocytes, Pygo2^−/− MEFs exhibited spontaneous adipocyte differentiation, as verified by strong Oil Red O staining (Fig. 7A) and an increase in C/EBPα and PPARγ levels (Fig. 7B). We next constructed a floxed allele of mouse Pygo2 and introduced it into the PDGFRα-Cre transgene to generate the Pygo2 conditional knockout mice in adipocyte precursors (PDGFRα-Cre;Pygo2^flox/flox) (41–43). The resulting PDGFRα-Cre;Pygo2^flox/flox mice are henceforth referred to as Pygo2^pre−/− mice. Littermates lacking the Cre gene (Pygo2^flox/flox) were used as WT controls. It was reported that in addition to the Lin⁻CD29⁺CD34⁺Sca-1⁺CD24⁺ adipocyte precursors, PDGFRα also labels small populations of CD45⁺CD31⁺ cells in WAT (43,44) and oligodendrocyte precursor cells (OPCs) in the brain (45). To exclude the possibility that Pygo2 knockout may have effects on immune cells in WAT or on the CNS, we assessed Pygo2 expression in these cell types in the WT and Pygo2^pre−/− mice. Real-time qPCR results indicated that Pygo2 was expressed mainly in Lin⁻CD29⁺CD34⁺Sca-1⁺CD24⁺ adipocyte precursors, very little in CD45⁺ and CD31⁺ cells, and was effectively deleted in Pygo2^pre−/− mice (Supplementary Fig. 7C). Moreover, when we isolated neurons and OPCs from mouse brain and checked the Pygo2 expression, real-time qPCR results showed that Pygo2 was highly expressed in neurons but absent in OPCs (Supplementary Fig. 3A). Consistently, confocal immunofluorescence analysis of brain tissue revealed that Pygo2 was highly expressed in NeuN⁺ neuron cells but absent in oligo2⁺ oligodendrocyte lineage cells (Supplementary Fig. 3B). From these results, we conclude that Pygo2 was knocked out only in adipocyte precursors in Pygo2^pre−/− mice.

First, we checked the β-catenin signaling in adipocyte progenitors by sorting PDGFRα⁺ cells, and the real-time qPCR results indicated that Pygo2^pre−/− animals exhibited greatly reduced expression of Wnt target genes Axin2 and Lef1 than the WT littermates in these cells (Fig. 7C). It is noted that Pygo2 expression was dramatically reduced in PDGFRα⁺ adipocyte progenitors of Pygo2^pre−/− animals, which suggests that Pygo2 was effectively knocked out in these cells (Fig. 7C). We also assessed food intake, movement, and energy expenditure in control and knockout mice. On a normal chow diet, the Pygo2^pre−/− mutation did not affect food intake and movement (Supplementary Fig. 4A and B). However, energy expenditure, including O₂ consumption, CO₂ production, and heat production, was lower in Pygo2^pre−/− mice (Supplementary Fig. 4C–H). Then, we monitored the weight of WT and Pygo2^pre−/− mice. No difference was found between WT and Pygo2^pre−/− mice at birth. On a normal chow diet, 14-week-old Pygo2^pre−/− animals weighed more than their WT littermates (Fig. 7D). MRI examination showed increased accumulation of fat volume (Fig. 7E) and fat percentage (Fig. 7F) in Pygo2^pre−/− mice. These differences were also mirrored by the increased weight of the epididymal WAT depot in Pygo2^pre−/− mice (Fig. 7G). The increase in adiposity was associated with a significant enlargement in adipocyte volume observed in Pygo2^pre−/− mice compared with that in WT animals, as assessed by classic hematoxylin-eosin staining (Fig. 7H and I). Moreover, this effect was accompanied by a significant increase in the total content of genomic DNA in epididymal WAT, which suggests an increased number of adipocytes in Pygo2^pre−/− mice (Fig. 7J). To further examine whether the increased adiposity in Pygo2^pre−/− mice was associated with changes in adipocyte differentiation, we analyzed mature adipocyte–specific gene expression. The expression levels of C/EBPα, PPARγ1, and PPARγ2 were significantly increased in the WAT of Pygo2^pre−/− mice compared with those in WT mice. Moreover, the expression levels of PPARγ target genes, including those coding for lipogenic proteins, such as FABP4, fatty acid synthase (FASN), and fatty acid transport protein 1 (FATP1), as well as those coding for adipokines, such as adiponectin, resistin, and leptin, were significantly increased in the WAT of Pygo2^pre−/− mice compared with those in WT mice (Fig. 7K). Interestingly, the genes involved in fatty acid oxidation in WAT, such as acyl-CoA dehydrogenase medium chain (MCAD) and carnitine palmitoyltransferase 1b (Cpt1b), were significantly decreased in the WAT of Pygo2^pre−/− mice compared with those in WT mice (Fig. 7K), which may represent the cause of decreased energy expenditure in these mice. We also examined the adipose tissue inflammation status and brown adipose tissue (BAT) phenotype in Pygo2 knockout mice by detecting the expression of TNFα and IL-6, as well as the histology of interscapular BAT. The results showed significantly increased TNFα and IL-6 expression in Pygo2^pre−/− mice (Fig. 7K), but no significant difference in the BAT phenotype between Pygo2^pre−/− mice and WT controls (Supplementary Fig. 5). Finally, we checked the above adipose tissue phenotypes in Pygo2^+/− mice. When these adipose tissue phenotypes were not changed in Pygo2^+/− mice compared with those in WT mice under normal chow diet conditions, results similar to those observed in Pygo2^pre−/− mice were achieved when the Pygo2^+/− mice were treated under high-fat diet (HFD) feeding for 6 weeks (Supplementary Fig. 6). It is possible that Pygo2 deletion prevents preadipocyte proliferation and/or commitment, thereby limiting the number of committed preadipocytes. To exclude this possibility, we sorted Lin⁻CD29⁺CD34⁺Sca-1⁺CD24⁺ adipocyte precursors, and a similar population percentage was observed in both Pygo2^pre−/− and control mice (Supplementary Fig. 7A and B). Real-time qPCR detection of proliferation markers, including Ki67 and Pcna, exhibited similar expression levels (Supplementary Fig. 7D). These results indicate that Pygo2 deletion does not affect preadipocyte commitment and/or proliferation. Collectively, the above data suggest that increased adipogenesis and decreased energy expenditure may contribute to increased adiposity in Pygo2^pre−/− mice.

Increased WAT mass is usually associated with impairments in glucose metabolism and insulin sensitivity, leading to obesity (46–49). Therefore we examined the effect of Pygo2 ablation on glucose homeostasis. The levels of blood glucose and the concentration of fasting serum insulin under normal chow diet conditions were similar in Pygo2^pre−/− and WT mice (Fig. 8A and B). However, during a GTT, Pygo2^pre−/− mice showed significantly higher blood glucose concentrations after glucose challenge (Fig. 8C). Moreover, insulin secretion in response to glucose load also showed a delayed peak in Pygo2^pre−/− mice (Fig. 8D). The ITT was also conducted, which revealed that the glucose-lowering effect of insulin was more hampered in the mutant mice than it was in WT mice (Fig. 8E). We also conducted the above experiments in Pygo2^+/− mice. As described in Supplementary Fig. 8, the metabolic parameters were not changed in Pygo2^+/− mice compared with those in WT mice under normal chow diet conditions. However, results similar to those in Pygo2^pre−/− mice were achieved when the Pygo2^+/− mice were treated under HFD feeding for 6 weeks (Supplementary Fig. 9). Collectively, our in vivo observation that reveals an impaired glucose tolerance and decreased systemic insulin sensitivity in Pygo2-deficient mice highlights the important role played by Pygo2 in WAT and glucose homeostasis.

Previous studies have established that Wnt/β-catenin signaling inhibits adipogenesis, and some Wnt/β-catenin components are regulated in this process (19–22). Since there are so many components within the Wnt/β-catenin signaling cascade, there must be other factors that can govern the Wnt/β-catenin output level during adipogenesis. Indeed, our present study has identified Pygo2 as a newly identified activator of Wnt/β-catenin signaling in the regulation of adipocyte differentiation. We showed that Pygo2 exhibited a declined expression pattern during adipocyte differentiation, resulting in an attenuated Wnt/β-catenin signaling output level. Ectopic expression of Pygo2 attenuates adipocyte differentiation, whereas knockdown of Pygo2 enhances the differentiation process and induces spontaneous differentiation. Furthermore, MEFs lacking Pygo2 exhibited spontaneous adipocyte differentiation, and adipocyte precursor–specific Pygo2 deletion showed a significant increase in adipocyte differentiation. Thus, our gain- and loss-of-function studies both in vitro and in vivo have clearly indicated that Pygo2 is a novel factor required for the regulation of Wnt/β-catenin signaling during adipocyte differentiation. It is noted in this study that Pygo2 was also examined to express at a low level in the cytoplasm of 3T3-L1 cells (Fig. 4D), although it is originally identified as a nuclear component of Wnt/β-catenin signaling. Whether this protein is regulated to translocate between nucleus and cytoplasm during adipogenesis is worthy of further analysis.

Wnt/β-catenin signaling has been demonstrated to block adipogenesis by suppressing the expression of C/EBPα and PPARγ, whereas the expression of C/EBPβ and C/EBPδ was not affected (18,19). However, the underlying molecular mechanism by which Wnt/β-catenin signaling drives this regulation is still incompletely defined. The main challenge to connect this gap is that we must identify known or unknown Wnt/β-catenin downstream effectors that also directly control the expression of C/EBPα and PPARγ. Previous studies identified COUP-TFII as a negative regulator of adipogenesis (50) and a direct Wnt/β-catenin signaling downstream target that can directly inhibit PPARγ1 and PPARγ2 mRNA expression (51). However, the function of COUP-TFII in adipogenesis is controversial, since another report of an in vivo study showed that COUP-TFII inhibits the expression of Wnt10b and acts as a positive regulator of adipogenesis (52). In our system, we did not find the activation of COUP-TFII by either β-catenin or Pygo2 (Supplementary Fig. 10). Instead, we uncovered a new mechanism acting upstream of C/EBPβ to mediate this process (Fig. 6H). We found that inhibition of Wnt/β-catenin signaling by overexpressing dnTCF4 or knocking down β-catenin or Pygo2 resulted in the downregulation of Axin2, a constitutive Wnt target, in the cytoplasm. Consequently, the Axin2-bound GSK3β was released and translocated into the nucleus to phosphorylate C/EBPβ and Snail. C/EBPβ is the main upstream activator of C/EBPα and PPARγ (18), and Snail has been demonstrated to be a negative regulator of adipogenesis by inhibiting PPARγ expression (39). Moreover, phosphorylation of C/EBPβ by GSK3β results in an increase in DNA binding activity (37), whereas phosphorylation of Snail by GSK3β decreases its protein stability (38). Therefore, by this complex signal cascade, Pygo2 or β-catenin controls adipocyte differentiation. The original article on Wnt inhibition of adipogenesis showed that overexpression of Axin promotes adipogenesis (18), which seems to be inconsistent with our data. We reasoned that the dosage and the context are all different. First, in an overexpression study, people usually overexpress the objective protein to hundreds of times of the endogenous level. In our study, Axin2 is only overexpressed or knocked down within physiological dosage. Second, because Axin2 is a downstream target of Wnt/β-catenin signaling, we overexpressed or knocked down Axin2 upon Wnt/β-catenin signaling inhibition or activation to the original level to rescue the related phenotypes. Nevertheless, we have identified a novel mechanism for Axin2 in regulating adipogenesis.

The most unexpected result of this study was that upon PPARγ upregulation, deletion of the Pygo2 gene in adipose tissue resulted in larger adipocytes as well as enhanced insulin resistance and impaired glucose tolerance, which seems to be opposite to the fundamental function of PPARγ. Indeed, it has been generally accepted that activation of PPARγ leads to amelioration of insulin resistance, since administration of thiazolidinedione, the PPARγ agonist, has been shown to increase insulin sensitivity in obese insulin-resistant animals and humans (53,54). However, another study in heterozygous PPARγ-deficient mice showed that these mice were protected from the development of insulin resistance due to adipocyte hypertrophy under an HFD and they suggested that PPARγ plays dual roles in the regulation of insulin sensitivity (55). Therefore, in some cases, activation of PPARγ ameliorates insulin resistance, whereas in other cases, the increase in the amount of the PPARγ gene product leads to insulin resistance. In our case, Pygo2 deletion results in increased adipogenesis and decreased energy expenditure. Because the adipose expandability for an individual is limited, the adipose tissue in Pygo2-deletion mice tends to differentiate to mature adipocyte but finally will reach the maximal fixed capacity for safely storing fat. During the adolescent period of mice, the increase in differentiated adipocytes still allows the storage of an increasing amount of fat. In adult mice, once the accumulation of fat is beyond the speed of adipocyte production, PPARγ activation results in adipocyte hypertrophy to generate large adipocytes. Once the adipocyte size is enlarged to the limit of its expandability, the excess lipids will overflow from the adipose tissue. Consequently, TNFα and IL-6 release from large adipocytes and should be the cause of glucose intolerance and insulin resistance. Although we have identified the decreased expression of fatty acid oxidation genes in Pygo2-deletion mice as the main cause of decreased energy expenditure, the underlying detailed mechanism needs to be further investigated in the future.

As discussed above, obesity has been implicated in glucose intolerance and insulin resistance (56,57). Indeed, we found a decreased expression level of Pygo2 in ob/ob mice in comparison with WT mice. When we deleted Pygo2 expression in mouse adipocyte precursors, an enlarged fat mass was observed even under normal chow diet conditions, which suggests an obesity tendency. Both GTT and ITT assays also showed abnormal blood glucose and insulin concentrations in Pygo2-deletion mice. It is noted that the glucose and insulin levels were synchronously regulated in WT mice, which exhibited the same peak value occurrence time at 15 min after glucose challenge. However, the insulin peak value occurrence time of Pygo2-deletion mice was delayed to 30 min after glucose challenge, which suggests impaired glucose-stimulated insulin secretion, and therefore perhaps a defect in β-cell function. Interestingly, the delayed insulin peak value in Pygo2-deletion mice was increased. We deduced that this might be the compensation for the functional abnormality of insulin in the Pygo2-deletion obese mice. Collectively, these data have linked the two important roles played by Pygo2 in adipose metabolism and glucose metabolism.

In summary, we have defined a new effector that determines the Wnt/β-catenin output level during adipogenesis. Moreover, we identified a novel molecular signal cascade that mediates Wnt/β-catenin signaling activity in adipose tissues, which established an association between C/EBPα/PPARγ expression and Wnt/β-catenin activation. Our study highlighted the role of Wnt/β-catenin in the regulation of adipogenesis and glucose homeostasis. This study also revealed an association between Pygo2 function and obesity or diabetes.

Funding. This work was supported by grants from the National Natural Science Foundation of China (grant numbers U1705284, 81472457, and 81772958 to B.-A.L. and 81472725 to W.M.) and “Project 111” sponsored by the State Bureau of Foreign Experts and Ministry of Education (grant number B06016).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. Y.-Y.X. conceived the study, designed and performed experiments, analyzed data, and wrote the manuscript. C.-L.M., Y.-H.C., W.-J.W., K.-K.Z., and W.M. performed experiments and analyzed data. X.-X.H., Q.-F.L., Y.-J.L., J.-J.H., T.H., and Z.-Z.Z. constructed various plasmids for the study. B.-A.L. conceived the study, designed experiments, and wrote the manuscript. B.-A.L. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.