Cyclin C regulates adipogenesis by stimulating transcriptional activity of CCAAT/enhancer-binding protein α

Brown adipose tissue is important for maintaining energy homeostasis and adaptive thermogenesis in rodents and humans. As disorders arising from dysregulated energy metabolism, such as obesity and metabolic diseases, have increased, so has interest in the molecular mechanisms of adipocyte biology. Using a functional screen, we identified cyclin C (CycC), a conserved subunit of the Mediator complex, as a novel regulator for brown adipocyte formation. siRNA-mediated CycC knockdown (KD) in brown preadipocytes impaired the early transcriptional program of differentiation, and genetic KO of CycC completely blocked the differentiation process. RNA sequencing analyses of CycC-KD revealed a critical role of CycC in activating genes co-regulated by peroxisome proliferator activated receptor γ (PPARγ) and CCAAT/enhancer-binding protein α (C/EBPα). Overexpression of PPARγ2 or addition of the PPARγ ligand rosiglitazone rescued the defects in CycC-KO brown preadipocytes and efficiently activated the PPARγ-responsive promoters in both WT and CycC-KO cells, suggesting that CycC is not essential for PPARγ transcriptional activity. In contrast, CycC-KO significantly reduced C/EBPα-dependent gene expression. Unlike for PPARγ, overexpression of C/EBPα could not induce C/EBPα target gene expression in CycC-KO cells or rescue the CycC-KO defects in brown adipogenesis, suggesting that CycC is essential for C/EBPα-mediated gene activation. CycC physically interacted with C/EBPα, and this interaction was required for C/EBPα transactivation domain activity. Consistent with the role of C/EBPα in white adipogenesis, CycC-KD also inhibited differentiation of 3T3-L1 cells into white adipocytes. Together, these data indicate that CycC activates adipogenesis in part by stimulating the transcriptional activity of C/EBPα.

Brown adipose tissue is important for maintaining energy homeostasis and adaptive thermogenesis in rodents and humans. As disorders arising from dysregulated energy metabolism, such as obesity and metabolic diseases, have increased, so has interest in the molecular mechanisms of adipocyte biology. Using a functional screen, we identified cyclin C (CycC), a conserved subunit of the Mediator complex, as a novel regulator for brown adipocyte formation. siRNA-mediated CycC knockdown (KD) in brown preadipocytes impaired the early transcriptional program of differentiation, and genetic KO of CycC completely blocked the differentiation process. RNA sequencing analyses of CycC-KD revealed a critical role of CycC in activating genes co-regulated by peroxisome proliferator activated receptor ␥ (PPAR␥) and CCAAT/enhancer-binding protein ␣ (C/EBP␣).

Overexpression of PPAR␥2 or addition of the PPAR␥ ligand rosiglitazone rescued the defects in CycC-KO brown preadipocytes and efficiently activated the PPAR␥-responsive promoters in both WT and CycC-KO cells, suggesting that CycC is not essential for PPAR␥ transcriptional activity. In contrast, CycC-KO significantly reduced C/EBP␣-dependent gene expression. Unlike for PPAR␥, overexpression of C/EBP␣ could not induce C/EBP␣ target gene expression in CycC-KO cells or rescue the CycC-KO defects in brown adipogenesis, suggesting that CycC is essential for C/EBP␣-mediated gene activation. CycC physically interacted with C/EBP␣, and this interaction was required for C/EBP␣ transactivation domain activity. Consistent with the role of C/EBP␣ in white adipogenesis, CycC-KD also inhibited differentiation of 3T3-L1 cells into white adipocytes. Together, these data indicate that CycC activates adipogenesis in part by stimulating the transcriptional activity of C/EBP␣.
Adipose tissues play a pivotal role in energy homeostasis (1,2). Because of the increased epidemic of obesity and metabolic diseases, a better understanding of the molecular events in adipocyte biology is important for treating this global health problem. As the major cell types in adipose tissues, white adipocytes are best known for energy storage, whereas uncoupling protein 1 (UCP1)-expressing 4 brown/beige adipocytes dissipate energy in the form of heat via dissociation of cellular respiration from ATP synthesis (3,4). It has been shown that increased amounts and/or activity of brown/beige adipose tissues reduce metabolic defects in mice (5)(6)(7)(8) and correlate with leanness in humans (9 -11). Despite extensive studies of adipogenesis in the past, our understanding of the development of brown/beige adipocytes remains incomplete.
Transcriptional regulation is critical in adipogenesis; extracellular adipogenic stimuli activate a cascade of core transcription factors that control the differentiation of all types of adipocytes (2,12,13). Within this core transcriptional hierarchy, PPAR␥ is the master regulator, as it is both necessary and sufficient for adipogenesis (14 -17). The induced early transcription factors include C/EBP␤, C/EBP␦, the glucocorticoid receptor (GR), STAT5A, and cAMP response element-binding protein (CREB) (2,12,13). These transcription factors then activate downstream transcription factors, including PPAR␥ and C/EBP␣, which positively activate each other and coordinately regulate the expression of a large group of genes that define adipocyte phenotypes (2,12,13). Other transcription factors that modulate the expression or functions of the core adipogenic factors often affect adipogenesis. For example, EBF1 and EBF2 promote adipogenesis by activating the transcription of Pparg and Cebpa (18). ZFP423 determines white adipocyte cell fate by activating Pparg gene expression and suppressing EBF2 and PRDM16 functions (19,20). EBF2 determines and maintains the identity of brown adipocytes by facilitating PPAR␥ binding to brown fat-specific binding sites during differentiation (21). The activities of these adipogenic transcription factors can be further regulated by various transcription cofactors or non-coding RNAs. For example, the PPAR␥ cofactor PGC1␣ controls mitochondrial biogenesis and respiration through induction of Ucp1 and nuclear respiratory factors (22), and another PPAR␥ cofactor, PRDM16, has emerged as a brown and beige adipocyte determination factor (23,24). Moreover, the microRNAs miR-133 and miR-155 regulate brown and beige adipocyte differentiation by targeting PRDM16 and C/EBP␤, respectively (25,26). The long noncoding RNA Blnc1 in complex with EBF2 activates the thermogenic genes (27). Thus, overlapping and distinct transcriptional networks are involved in controlling white, brown, and beige adipogenesis.
The multisubunit protein complex Mediator is a transcriptional cofactor that was originally discovered in yeast (28 -30) and is conserved in eukaryotes (31)(32)(33). The Mediator complex can be recruited by numerous transcription factors, including PPAR␥ (34 -37) and C/EBP␤ (38 -40), and integrates cellular signals with the basal transcription machinery to activate or repress gene transcription (41)(42)(43)(44). The Mediator complex as a whole may regulate a large number of genes, but each subunit displays remarkable context-dependent or pathway-specific functions (41)(42)(43)(44). It has been shown that the Mediator subunits MED1 and MED14 directly bind to PPAR␥ and activate PPAR␥-dependent transcription and adipogenesis (36,37,45). More recent studies show that MED1 activates the Ucp1 gene in brown adipocytes by interacting with PRDM16 (46,47). The Mediator subunit MED23 promotes white adipogenesis through ELK1 (40,48). These studies suggest the involvement of the Mediator complex in adipogenesis, but the underlying mechanisms are distinct for different subunits. It is currently unclear whether other Mediator subunits also regulate adipogenesis.
In this study, we identify cyclin C (CycC) as a novel regulator of adipogenesis. Unlike classical cyclins, CycC does not directly regulate the cell cycle but acts as a conserved subunit of the Mediator complex. Here we show that CycC is essential for adipogenesis. In contrast to MED1, MED14, and MED23, CycC functions primarily to support C/EBP␣-mediated gene expression in adipogenesis.

Identification of novel regulators of brown adipogenesis
To study brown adipocyte biology, we established mouse brown preadipocyte cell lines according to protocols published previously (49,50). Selection of single clones was based on the efficiency of lipid accumulation as assayed by oil red O staining (Fig. 1A) as well as the induction of marker genes (i.e. the common adipocyte marker Fabp4, Fig. 1B, and the brown adipocyte-specific marker Ucp1, Fig. 1C) as detected by qRT-PCR in the presence of differentiation mixture (49,50). To identify novel regulators that control the formation of brown adipocytes, we transduced immortalized brown preadipocytes with a lentiviral library of shRNA, which included a total of 4551 protein-coding genes (Fig. 1D), and six or more independent sequences of shRNA were designed to target each gene. Before induction of differentiation, a small portion of those transduced cells was collected to determine the initial abundance of each shRNA in the library (i.e. control), and the remaining cells were cultured to confluence, followed by induction for 8 days (Fig. 1D). Taking advantage of the floating nature of lipid-filled adipocytes, we were able to isolate undifferentiated cells, which had little lipid droplets and were adherent, by two rounds of trypsin digestion and replating (Fig. 1D). We then extracted the genomic DNA from undifferentiated cells that were collected before and after induction. The gene identities were determined according to shRNA barcode sequences that were amplified by PCR and analyzed by Illumina next-generation sequencing (Fig. 1D).
Because shRNA-mediated knockdown (KD) may have off targets, we used the following cutoff to define a positive hit from this screen. Compared with the control, at least two different shRNA clones were enriched more than 1.5-fold, but none of the shRNA clones for that gene had less than a 0.3-fold reduction. Using these criteria, we identified a total of 543 genes (or 11.9%) as positive hits (data not shown). Functionally, those genes are involved in several biological processes, including cellular metabolism, macromolecule biosynthesis, regulation of gene expression, and protein modifications (Fig. 1E). Several genes that have been reported previously as regulators of adipogenesis, including Cebpa (17), Cebpb (51,52), Ebf1 (18), Rbl1 (53), and Stat5a (54,55), were among the positive hits (data not shown), indicating the validity of this screen.

CycC is down-regulated during brown adipogenesis and aging
Consistent with the critical role of gene transcription in adipogenesis, DNA-binding proteins or chromatin modifiers were particularly enriched among positive hits in this screen (a total of 156 genes or 28.7%, data not shown). Among them, 14 genes are transcription cofactors (Fig. 1F), including Ccnc, which encodes CycC. Although previous studies have reported that the Mediator subunits MED1 (36,45), MED14 (37), and MED23 (40, 48) regulate either white adipogenesis or brown adipocyte-specific genes, the role of CycC in adipogenesis has not been investigated.
First, we examined the protein levels of CycC and its known partners CDK8 and CDK19, which are mutually exclusive subunits of the Mediator complex (56), during the differentiation of brown preadipocytes. As shown in Fig. 2A, during differentiation, CycC and CDK19 proteins were down-regulated, but CDK8 proteins had a trend of increase. Because aging affects brown fat, we compared the protein abundance in BAT between young (3 months old) and aged (24 months old) C57Bl/6J mice. As shown in Fig. 2, B and C, the protein levels of CycC and CDK19, but not CDK8, were significantly lower in BAT from aged mice. As expected, aging also reduced the UCP1 proteins (Fig. 2, B and C). Cold exposure is known to enhance BAT activity. Thus, we also examined CycC levels upon cold exposure. As shown in Fig. 2D, cold exposure at 4°C for 6 h induced the expression of Ucp1 and Pgc1a in BAT, as expected, but did not affect Ccnc, Cdk8, and Cdk19 compared with room temperature (22°C). At the protein level, cold expo-sure slightly down-regulated CycC and CDK19 but increased CDK8 protein levels (Fig. 2E). Thus, different from what we observed in the liver (57), in BAT, CycC and CDK19, but not CDK8, appear to be coordinately regulated.
Although the best-known function of CycC is to be a subunit of the Mediator complex, recent studies have shown that CycC can also function in the cytoplasm to regulate mitochondrial fission (58,59). Therefore, we examined the subcellular location of CycC in brown adipocytes by immunostaining. As shown in Fig. 2F, CycC proteins were exclusively located in the nucleus in brown preadipocytes. Although a small portion of CycC proteins became diffused in the cytoplasm after differentiation, the majority of CycC proteins were still in the nucleus (Fig. 2F). These data suggest that, in brown adiopocytes, CycC acts predominantly in the nucleus.

CycC is essential for brown preadipocyte differentiation
To verify the shRNA screen results, we transfected brown preadipocytes with a pool of siRNAs that were designed to target four different sequences of Ccnc mRNA for reducing off targets. Nonspecific (NS) siRNA was used as a control. The CycC-KD efficiency was determined by qRT-PCR ( Fig. 3A) and immunoblotting (Fig. 3B). The siRNA-treated cells were cultured until confluent and induced with differentiation mixture. After 2 days of induction, CycC-KD significantly inhibited the expression of adipocyte-enriched aP2, FAS, and ATGL proteins, whereas other markers, such as PPAR␥1, C/EBP␣, and Caveolin, were less affected (Fig. 3C). As shown in Fig. 3D, CycC-KD significantly inhibited cellular lipid accumulation, as examined by oil red O staining. Interestingly, CycC-KD resulted in a mixed pattern of gene expression. For common adipocyte markers, initially CycC-KD cells expressed significantly lower levels of Pparg, Fabp4 (encoding aP2), Cebpa, Adipoq (encoding Adiponectin), and Cd36 during 2 days of induction (Fig. 3, E and F), but after 5 days of induction, these differences vanished (Fig. 3, E and F). These data suggest that CycC may act primarily at the early stage(s) of adipogenesis. For lipogenic genes, such as Srebf1c and Fasn, and other genes, including Acc1, Atgl, Hsl, and Slc2a4 (encoding GLUT4), CycC-KD reduced their expression throughout the differentiation process (Fig. 3F). Moreover, CycC-KD also significantly reduced the expression of most brown adipocyte-specific marker genes, including Ucp1, Cidea, Otop1, and Fgf21 after 5 days of differentiation (Fig. 3G), suggesting that CycC is required to activate the brown adipocyte transcription pro-gram. Interestingly, Prdm16, the key activator of the brown adipocyte transcription program, was not affected by CycC-KD (Fig. 3G).
Because of the transient nature of siRNA-mediated KD and partial reduction of CycC proteins, we generated CycC-KO brown preadipocytes by treating brown preadipocytes isolated from Ccnc flox/flox mice with lentiviruses expressing Cre proteins. CycC-KO was verified by qRT-PCR using a pair of exonspecific primers (Fig. 4A). As shown in Fig. 4B, CycC-KO cells completely lost their adipogenic capability, as examined by oil red O staining for lipid accumulation. Consistently, gene expression analyses by qRT-PCR revealed that, in CycC-KO cells, the differentiation mixture was unable to induce general adipocyte makers, including Pparg, Cebpa, Fabp4, and Adipoq, brown adipocyte-specific markers, and mitochondrial genes, including Ucp1, Cidea, Otop1, Fgf21, Cox5b, and Cox8b, and lipogenic genes, including Srebf1c and Fasn (Fig. 4, C and D). At the protein level, CycC-KO almost completely abolished the induction of key brown and general adipocyte markers, including PPAR␥, C/EBP␣, UCP1, aP2, FAS, ATGL, GLUT4, and Caveolin (Fig. 4E). These results indicate that CycC is essential for brown preadipocyte differentiation in vitro.
To determine whether CycC overexpression in CycC-KO cells could rescue the adipogenic defects, we treated cells with retroviruses expressing CycC. The levels of CycC were exam- To determine whether CycC overexpression in WT brown preadipocytes could conversely enhance adipogenesis, we treated cells with CycC-expressing retroviruses. Surprisingly, although the mRNA levels of Ccnc were increased by 2.5-fold, adipogenesis was not significantly affected (Fig. 4, K-M). Immunoblotting revealed that CycC overexpression reduced endogenous CycC and, therefore, restricted the increase in the total amount of CycC proteins (Fig. 4N).

CycC activates the core adipogenic program
To understand the molecular mechanism(s) underlying CycC regulation of brown adipogenesis, we performed RNAseq analyses to determine the effects of CycC-KD on genomewide gene expression at day 2 of differentiation. As shown in Fig. 5A, of a total of 10,852 transcripts that were detected, CycC-KD resulted in up-regulation (Ͼ1.5-fold as cutoff) of 789 genes (or 7.3%) and down-regulation (Ͻ0.5-fold as cutoff) of 633 genes (or 5.8%). These data indicate that, although the Mediator complex as a whole may control the expression of many genes, the CycC subunit only regulates a small subset of genes in brown adipocytes. Biological pathway analyses revealed that the most significantly up-regulated pathways  included regulation of cell migration, cell surface receptor signaling, cell motility, and cell development (data not shown), and the most significantly down-regulated pathways were involved in lipid metabolism and fat cell differentiation (Fig.  5B), indicating that gene down-regulation is the primary cause of the phenotype observed after CycC-KD. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses revealed that the most significantly down-regulated pathway was the PPAR signaling pathway (Fig. 5C). In fact, CycC-KD reduced the expression of many known PPAR␥-target genes, including Pparg, Plin1, Cd36, Adipoq, and Lpl (Fig. 5D). Down-regulation of PPAR␥ target genes was also verified by qRT-PCR analyses (Fig. 5E).

CycC regulation of brown adipogenesis is not dependent on PPAR␥ transcriptional activity
Previous studies have demonstrated that the Mediator subunits MED1 (36,45) and MED14 (37) are required to activate PPAR␥-mediated gene transcription. To determine whether CycC is also required for PPAR␥-dependent gene expression in brown adipocytes, we overexpressed PPAR␥2 in WT and CycC-KO brown preadipocytes by retroviral transduction (Fig.  6A). As shown in Fig. 6B, overexpression of PPAR␥2 was able to efficiently rescue CycC-KO defects of adipogenesis, as examined by oil red O staining. Consistent with the rescue of lipid accumulation, overexpression of PPAR␥2 also rescued key marker gene expression (Fig. 6C). Similarly, addition of the PPAR␥ ligand rosiglitazone to the culture medium also rescued the CycC-KO-caused defects in adipogenesis (Fig. 6D) and key marker gene expression (Fig. 6E). These results suggest that CycC is not required to maintain PPAR␥ transcriptional activity. Indeed, PPAR␥2 overexpression could similarly induce endogenous PPAR␥ target genes, such as Cd36 and Adipoq, in the presence or absence of CycC (Fig. 6F). Furthermore, PPAR␥2 overexpression by transient transfection could activate a synthetic PPAR-responsive (PPRE) promoter in both WT and CycC-KO brown preadipocytes (Fig. 6G). Interestingly, overexpressed PPAR␥2 was even more active in KO cells (Fig.  6G). Together, these data indicate that CycC is not an activator of PPAR␥-dependent transcription.

CycC is required for C/EBP␣-dependent transcription in brown preadipocytes
A previous study has shown that PPAR␥ and C/EBP␣ co-regulate a series of genes, including adipocyte markers, in adipogenesis (60). We compared CycC-regulated genes with those defined PPAR␥ and/or C/EBP␣ target genes (60). As shown in Fig. 7A, nearly half of PPAR␥-C/EBP␣ co-regulated genes were down-regulated upon CycC-KD. Thus, during brown adipogenesis, CycC is critically required to activate a subset of genes that are co-regulated by PPAR␥ and C/EBP␣. Because CycC-KO did not attenuate PPAR␥ transcriptional activity, we asked whether CycC regulates the transcriptional activity of C/EBP␣. To this end, we examined the ability of C/EBP␣ to activate its target genes by overexpression (Fig. 7B). As shown in Fig. 7C, C/EBP␣ overexpression by retroviral transduction robustly induced endogenous C/EBP␣ target genes such as Pparg2 in WT, but not CycC-KO, brown preadipocytes, although C/EBP␣ was overexpressed similarly in terms of -fold increase. These data suggest that CycC may be important for the transcriptional activity of C/EBP␣. Consistent with gene expression, C/EBP␣ overexpression could not rescue the differentiation defects in CycC-KO cells (Fig. 7, D and E). Thus, our results indicate that CycC stimulates brown adipogenesis at least in part through activating C/EBP␣-mediated gene expression.
Similar to CycC regulation of endogenous genes, overexpression of C/EBP␣ in CycC-KO cells was significantly less efficient to activate a synthetic C/EBP-responsive promoter (3ϫCBE-Luc) in luciferase reporter assays (Fig. 7F). By co-immunoprecipitation (co-IP) analyses, we found that CycC physically interacted with C/EBP␣ (Fig. 7G). Consistent with the transcriptional cofactor role of CycC, the transactivation domain (TAD) of C/EBP␣ is about 50% as active in CycC-KO cells as in WT cells in luciferase reporter assays (Fig. 7H). CycC-KO in brown preadipocytes did not affect the TAD activities of C/EBP␦ and SREBP1c but slightly reduced C/EBP␤-TAD activity and increased VP16-TAD activity (Fig. 7H). Together, our data suggest that CycC is required for C/EBP␣-TAD activity.
C/EBP␤ is an important transcription factor upstream of C/EBP␣ in adipogenesis. We wondered whether overexpression of C/EBP␤ could rescue the defects in CycC-KO cells. To this end, we transduced brown preadipocytes with retroviruses expressing C/EBP␤ (LAP2) (Fig. 8A). As shown in Fig. 8, B and C, C/EBP␤ overexpression was unable to rescue the CycC-KOcaused defects in brown adipogenesis.

CycC regulates white adipogenesis
Because C/EBP␣ is an important activator for white adipocyte formation, we speculate that CycC may be also required for white adipogenesis. To test this hypothesis, we transfected 3T3-L1 cells with the Ccnc or NS siRNA pool, and CycC-KD efficiency in 3T3-L1 cells was analyzed by qRT-PCR (Fig. 8D) and immunoblotting (Fig. 8E). The siRNA-treated cells were  cultured until confluent and then treated with differentiation mixture. As shown in Fig. 8F, CycC-KD significantly inhibited the formation of adipocytes, as examined by oil red O staining. Quantitative measurements further confirmed that CycC-KD reduced lipid accumulation during 3T3-L1 cell differentiation (Fig. 8G). Similar to what we observed in brown adipocytes, CycC-KD in 3T3-L1 cells also reduced the expression of key adipocyte markers, such as Pparg, Fabp4, Cebpa, and Adipoq, at the early stages (day 2) of adipogenesis (Fig. 8H). Thus, our data indicate that CycC is also required for white adipogenesis.

Discussion
In this study, we performed a loss-of-function screen to search for novel regulators of brown adipogenesis. Because this screen was based on the floating nature of lipid-rich adipocytes (Fig. 1D), the resulting positive hits may represent various mechanisms of action that could disturb the formation of mature brown adipocytes, including regulation of cell differentiation, fatty acid biosynthesis, and lipid droplet formation. The validity of this screen is evidenced by the presence of numerous reported regulators of adipogenesis among positive hits (data not shown) as well as the relatively low enrichment rate of positive hits (ϳ12%). From the screen, we identified a number of genes that are potential regulators for BAT development. Particularly enriched are transcription factors, cofactors, and regulators of protein modifications, such as ubiquitination (data not shown). Thus, future studies are necessary to determine whether those genes play a role in the development of BAT.
One positive hit of the screen was the CycC subunit of the Mediator complex. Independent experiments confirmed that CycC-KD in brown preadipocytes impairs brown adipogenesis (Fig. 3). More strikingly, CycC-KO in brown preadipocytes completely abolished the ability to differentiate into mature brown adipocytes (Fig. 4), demonstrating the essential role of CycC for brown adipocyte formation in vitro.
The Mediator complex functions primarily as a transcriptional cofactor in eukaryotic cells (41,44). Recent studies show that CycC also has functions outside of the Mediator complex (58,59). However, our data suggest that CycC in brown preadipocytes functions most likely in the context of the Mediator complex. First, CycC proteins are predominantly located in the nuclei in brown fat cells (Fig. 2F). Second, it is difficult to increase CycC protein levels in brown preadiocytes (Fig. 4N), as CycC overexpression results in a reduction of endogenous CycC, likely because of the degradation of both endogenous and overexpressed CycC proteins. These data suggest that only CycC proteins in the Mediator complex are relatively stable, like in cancer cells (61).
The best known binding partners of CycC are CDK8 and CDK19 (56). Interestingly, our data suggest that CycC regulation of brown adipogenesis may be through the CDK19-Mediator complex, as CycC and CDK19, but not CDK8, in brown adipocytes are changed similarly at the protein level during in vitro differentiation and aging or cold exposure in mice. However, whether CDK19 and/or CDK8 are also involved in adipogenesis needs to be investigated in the future.
To elucidate the molecular mechanism(s) underlying CycC regulation of brown adipogenesis, we analyzed gene expression by RNA-seq. Our RNA-seq data show that CycC-KD affects ϳ13% of the expressed transcripts in brown adipocytes during differentiation (Fig. 5A). These data indicate that CycC regulation of gene transcription is relatively specific. Further supporting this conclusion, CycC depletion did not significantly affect brown preadipocyte growth (data not shown). Although previous studies have implicated the Mediator subunits MED1 (36,45), MED14 (37), and MED23 (40,48) in the adipogenic transcriptional program, the regulatory mechanisms are clearly distinct for different subunits. Therefore, the Mediator complex likely plays roles at multiple stages of adipogenesis through different transcription factors.
Among CycC-regulated transcripts, the PPAR pathway was the top down-regulated pathway. Nearly half of the known PPAR␥ target genes were significantly down-regulated upon CycC depletion. One assumption is that, like MED1 and MED14, CycC is required for the transcriptional activity of PPAR␥. However, overexpressed PPAR␥2 was able to efficiently activate both endogenous and synthetic PPAR␥-responsive promoters in CycC-KO cells, indicating a dispensable nature of CycC for PPAR␥-dependent gene expression. In fact, CycC-KO resulted in an increase of overall PPAR␥ transcriptional output (Fig. 6G). Although the mechanism(s) underlying such an increase are currently unclear and need to be further investigated, CycC regulation of PPAR␥ is not likely to be the cause of defects in adipogenesis that results from CycC-KO. Supporting this conclusion, overexpression of PPAR␥2 or addition of the PPAR␥ ligand rosiglitazone could efficiently rescue the differentiation defects in CycC-KO cells. Interestingly, CycC-KD did not affect the expression of those same genes at a later stage of differentiation (Fig. 3E). One possibility is that self-activation of PPAR␥ and/or increased overall PPAR␥ transcriptional activity in CycC-KO cells resulted in this "catch-up expression" of PPAR␥ target genes. Alternatively, it is also possible that, in mature brown adipocytes, CycC may regulate both positive and negative events of gene transcription, neutralizing the overall outcome. Considering the fact that CycC proteins are down-regulated during brown preadipocyte differentiation ( Fig. 2A), CycC may be important primarily at the early stages of differentiation. Nevertheless, different from MED1 or MED14 regulation of adipogenesis, multiple lines of evidence argue against a primary role of PPAR␥ in CycC regulation of brown adipogenesis.
Because CycC regulation of PPAR␥ could not explain CycC-KO phenotypes in brown adipogenesis, an alternative mechanism is that CycC regulates PPAR␥ target genes through other transcription factors, as any given promoter is often controlled by multiple transcription factors. It has been shown that most PPAR␥ target genes in adipogenesis are also co-regulated by C/EBP␣ (60). We examined whether CycC regulates C/EBP␣. Indeed, CycC-KO significantly reduced C/EBP␣ transcriptional activity (Fig. 7, C and F). Mechanistically, CycC (likely in the context of the Mediator complex) physically binds to C/EBP␣ (Fig. 7G) and supports its TAD activity, although CycC may also regulate other transcription factors, such as C/EBP␤ (Fig. 7H). Because C/EBP␣ can activate itself, its mRNA levels were also decreased in CycC-KO cells. Moreover, overexpression of C/EBP␣ could induce known PPAR␥-C/ EBP␣ target gene expression only in WT cells but not in CycC-KO cells. Different from the results of increasing PPAR␥ level/activity, overexpression of C/EBP␣ could not rescue CycC-KO-caused differentiation defects in brown preadipocytes, suggesting that CycC regulation of C/EBP␣-mediated gene expression is required for brown adipocyte formation. In agreement with the key role of C/EBP␣ in activating the adipogenic gene program, overexpression of an upstream transcription factor, C/EBP␤, could not rescue CycC-KO-caused defects either. Furthermore, consistent with the critical role of C/EBP␣ in adipogenesis and the role of CycC in regulating C/EBP␣, CycC-KD also impairs white adipogenesis, as demonstrated using 3T3-L1 cells (Fig. 8F).
In summary, this study demonstrates that CycC is required for adipogenesis. Mechanistically distinct from MED1 and MED14, CycC regulation of adipogenesis is not through PPAR␥. Although additional mechanisms cannot be excluded, our data indicate that CycC promotes adipogenesis by activating C/EBP␣-dependent transcription.

Generation of CycC-KO brown preadipocytes
Ccnc-flox mice were generated and backcrossed with the C57Bl/6J strain for 10 generations. All mice were fed ad libitum with a standard chow diet (5053, LabDiet). The Albert Einstein College of Medicine Institutional Animal Care and Use Committee approved all animal experiments. Mouse brown preadipocyte cell lines were generated according to published protocols (49,50). For generation of CycC-KO brown preadipocytes, immortalized Ccnc flox/flox cells were transduced with the CMV-Cre-GFP lentivirus and selected with puromycin (2.0 g/ml). Single clones were selected.

Screen with the shRNA library
The Mouse Module I lentiviral shRNA library (Cellecta, Inc.) was packaged into lentiviruses in HEK293T cells using the calcium phosphate precipitation method. The viral supernatant was collected 48 h and 72 h after transfection, filtered, pelleted by ultracentrifugation, resuspended in 1ϫ PBS, and then frozen at Ϫ80°C. The final titer was 8 ϫ 10 6 transduction units per ml. In total, 2.2 ϫ 10 7 brown preadipocytes were plated at 50% confluence on 44 10-cm dishes and transduced with lentivirus stocks containing 4 g/ml Polybrene. The medium was replaced 1 day after transduction. Upon confluence, two dishes of cells were collected as controls, and other cells were induced for differentiation for 7 days. The cells were collected after trypsin digestion, dispersed by pipetting several times, and then replated in 10-cm dishes. In ϳ15 min, the undifferentiated preadipocytes and poorly differentiated cells attached to the dish bottom, whereas well differentiated adipocytes remained floating in the medium. The adherent cells were washed gently with 1ϫ PBS, followed by trypsin digestion and replating, and then collected. Genomic DNA was isolated using the Wizard genomic DNA purification kit and amplified by 10 rounds of PCR. The high-throughput sequencing was performed using a HiSeq 2500 instrument (Illumina) at the Einstein Genomics Core Facility. The enriched pathways and biological process Gene Ontology (GO) terms were obtained using the DAVID database (version 6.8).

Transfection of siRNA
ON-TARGETplus siRNA pools for Ccnc (047925) were purchased from GE Dharmacon. Non-targeting siRNA (001810) was used as a control. Preadipocytes were transfected with siRNA (25 pmol/2 ϫ 10 5 cells) using Dharma-FECT reagent by following the protocol of the manufacturer. The next day, fresh medium was added to allow cells to grow until confluent in 35-mm dishes or 12-well plates in DMEM with 20% FBS.

RNA-seq analysis
Total RNA was extracted from brown adipocytes transfected with NS-siRNA or Ccnc-siRNA and induced for 2 days (two biological replicates for each condition) using TRIzol reagent according to the instructions of the manufacturer. After removal of contaminating genomic DNA with RNase-free DNase I and quality analyses by Bioanalyzer (Agilent Technologies), RNA samples were processed at the Einstein Genomics Core Facility for library construction and sequencing using a HiSeq 2500 instrument (Illumina). Raw reads for each library were mapped using TopHat version 2.0.8 against the indexed mouse (mm9) genome, and transcripts were assembled using Cufflinks. Genes not expressed under both conditions (fragments per kilobase million (FPKM) Ͻ 1) were filtered out. Differentially expressed genes were analyzed using an RNA-seq processing tool, DESeq2, and selected by adjusted p Ͻ 0.05 and -fold change Ͼ1.5. The enriched pathways and biological process GO terms were obtained using the DAVID database (version 6.8), and the expressed genes were used as background. The heat map (18 down-regulated genes in the PPAR pathway) was generated based on the read count after variance-stabilizing transformation by DESeq2. The genes were ranked according to their p values.

Real-time PCR analysis
One g of total RNA was used for cDNA synthesis using the iScript cDNA synthesis kit (Bio-Rad). Real-time PCR was performed using FastStart Universal SYBR Green Master (Roche). Sequence information for specific primers used in this study is available upon request. All reactions were normalized to the Tbp level and performed in triplicate.

Co-IP
FLAG-C/EBP␣ (〈ddgene, 66978) was co-transfected with HA-GFP or HA-CycC. One day after transfection, cell lysates were prepared in the IP buffer as described previously (62). FLAG-tagged proteins were immunoprecipitated from the lysates using anti-FLAG antibody. After five washes with IP buffer, FLAG-tagged proteins were eluted with 3ϫ FLAG peptides. Proteins in elutes were detected by immunoblotting using anti-FLAG or anti-HA antibody.

Immunostaining
Immortalized brown preadipocytes or differentiated brown adipocytes were fixed with 4% formaldehyde (Fisher Scientific) in 1ϫ PBS for 10 min at room temperature, permeabilized with 0.1% Triton X-100 (Sigma-Aldrich) for 15 min, and blocked with 10% normal blocking serum for 30 min. The samples were then incubated with polyclonal anti-CycC antibody (AHF0072, Invitrogen, 1:200) in 1.5% normal blocking serum (ab166643, Abcam) for 2 h at room temperature, washed with 1ϫ PBS three times, incubated with Alexa Flour 488-labeled secondary antibody (A-21206, Invitrogen, 1:500) in 1.5% normal blocking serum for 45 min at room temperature in the dark, and washed with 1ϫ PBS three times. Nuclei were stained with Prolong Antifade reagent with DAPI (Invitrogen). Images were captured using a fluorescence microscope.

Oil red O staining
After various days of differentiation, cells were washed once with 1ϫ PBS and fixed with formaldehyde (Fisher Scientific) for 1 h. The staining stock was prepared by dissolving 0.35 g oil red O (Sigma) in 100 ml of isopropanol, stirred overnight, filtered through a 0.22-m filter, and stored at room temperature. For preparing the oil red O working solution, 6 ml of the stock was mixed with 4 ml of distilled water. After sitting at room temperature for 20 min, the staining solution was filtered again and then added to cells for 30 min. The staining solution was then removed, and cells were washed four times with distilled water. Images were captured using a digital camera and an inverted microscope.

Statistical analyses
The results are expressed as mean Ϯ S.E. Significance was determined by one-way analysis of variance. p Ͻ 0.05 was considered to be statistically significant. The statistical analyses were performed using SPSS15.0.  Note added in proof-In the version of this article that was published as a Paper in Press on March 28, 2017, the construct used for the C/EBP␤-TAD sequence in Fig. 7H was missing a few amino acids because of a PCR cloning error. The C/EBP␤-TAD clone has been corrected and the experiments were repeated. This error does not affect the results or conclusions of this work.