Adipocyte-specific Expression of Murine Resistin Is Mediated by Synergism between Peroxisome Proliferator-activated Receptor γ and CCAAT/Enhancer-binding Proteins*

Resistin antagonizes insulin action in mouse, making it a potential therapeutic target for treating metabolic diseases such as diabetes. To better understand how mouse resistin gene (Retn) expression is restricted to fat tissue, we identified an adipocyte-specific enhancer located ∼8.8-kb upstream of the transcription start site. This region contains a binding site for the master adipogenic regulator peroxisome proliferator-activated receptor γ (PPARγ), and binds endogenous PPARγ together with its partner retinoid-X receptor α (RXRα). It also contains three binding sites for CCAAT/enhancer-binding protein (C/EBP), and is bound by endogenous C/EBPα and C/EBPβ in adipocytes. Exogenous expression of PPARγ/RXRα and C/EBPα in non-adipocyte cells synergistically drives robust expression from the enhancer. Although PPARγ ligands repress Retn transcription in adipocytes, rosiglitazone paradoxically stimulates the enhancer activity, suggesting that the enhancer is not directly involved in negative regulation. Unlike expression of Retn in mouse, human resistin (RETN) is expressed primarily in macrophages. Interestingly, the region homologous to the mouse Retn enhancer in the human gene contains all three C/EBP elements, but is not conserved for the sequence bound by PPARγ. Furthermore, it displays little or no binding by PPARγ in vitro. Taken together, the data suggest that a composite enhancer binding both PPARγ and C/EBP factors confers adipocyte-specific expression to Retn in mouse, and its absence from the human gene may explain the lack of adipocyte expression in humans.

We have previously reported a functional C/EBP␣-binding site at the proximal promoter of Retn, and that a proximal 264-bp fragment of Retn was sufficient for expression in adipocytes (42). However, the expression level of the mouse Retn promoter was quite modest in comparison with the robust expression of endogenous Retn in adipocytes. Moreover, unlike Retn, expression of C/EBP␣ is not restricted to adipose tissue (30,43), and thus C/EBP␣ alone is not sufficient to explain the adipocyte-specific expression of Retn. A functional Sp1 binding site has also been reported in the proximal promoter of the Retn gene, but Sp1 is a ubiquitously expressed transcription factor (44). We have also reported that Retn is dramatically induced by retroviral infection of constitutively active PPAR␥ (fusion of viral transcriptional activator VP16 and PPAR␥) in 3T3-L1 preadipocytes (45), though PPAR␥ failed to activate a reporter gene containing Ϫ6.2 kb of the Retn 5Ј-flanking sequence in our previous study (42). Based on these observations, we hypothesized that additional cis-regulatory elements, presumably PPAR␥-binding sites, might exist outside the region we have previously analyzed to dictate and enhance adipocyte-specific expression of Retn.
Here, we have combined analysis of epigenetic marks and transcription factor binding by chromatin immunoprecipitation to identify a novel adipocyte-specific enhancer element at ϳ8.8-kb upstream to the transcription start site (TSS) of Retn. It contains a binding site for PPAR␥ and three C/EBP-binding sites, and endogenous PPAR␥, C/EBP␣, and C/EBP␤ bound to this region in adipocytes. Simultaneous expression of C/EBP␣ and PPAR␥ synergistically and robustly activated the Retn enhancer in non-adipocytic cells. Interestingly, the human resistin gene (RETN) contains a sequence homologous to the mouse enhancer, but lacks the functional PPAR␥-binding site. Thus, murine-specific cooperation between C/EBP␣ and PPAR␥ dictates adipocyte specific expression of mouse Retn.

MATERIALS AND METHODS
Chromatin Immunoprecipitation Followed by Genomic Chip (ChIP-chip) Analysis-ChIP-chip was performed as described previously (46). In short, input and chromatin IP DNA were amplified linearly using in vitro transcription, and hybridized to a custom, densely tiled DNA microarray. ChIP signals (IP/input) were normalized to total histone H3 occupancy, and plotted as a moving average of 19 contiguous probes spanning 510 bp.
Plasmids-pGL3-basic and pGL4.27 were purchased from Promega (Madison, WI). Various lengths of the 5Ј-flanking region of mouse Retn were inserted into MluI and XhoI sites of pGL3-basic using bacterial artificial chromosome (BAC) recombination methods as described previously (47,48). Sequence information was obtained from the UCSC genome browser. The mouse Retn enhancer element (Ϫ8872 to Ϫ8676 and Ϫ8843 to Ϫ8695 bp) and human genomic DNA fragment corresponding to the mouse Retn enhancer (Ϫ13262 to Ϫ13110 bp) were amplified by PCR and subcloned into BglII and XhoI sites of the pGL4.27 vector.
Transfection and Luciferase Assay-3T3-L1 adipocytes and preadipocytes were transfected by electroporation (Nucleofector II, AMAXA). Briefly, either preadipocytes or mature adipocytes (days 8 -10 after differentiation) were detached from culture dishes with 0.25% trypsin, washed twice with phosphate-buffered saline, and resuspended in electroporation buffer (solution V, AMAXA). Approximately 1 ϫ 10 6 cells were electroporated (electroporation program T-020) with equimolar amount of reporter vector (0.25ϳ1.0 g) and ␤-galactosidase expression vector (0.3 g), and seeded into triplicate wells of a 24-well plate. HEK 293T and HepG2 cells were transfected using Lipofectamine 2000 (Invitrogen). Medium was replaced with fresh culture medium after 16 h of transfection. The cells were harvested, and luciferase activities were measured by the luciferase assay system (Promega) after 48 h. ␤-Galactosidase activity was measured as described elsewhere (50). Light units were normalized to ␤-galactosidase activity. Fold activations relative to control were calculated, and means and standard errors from triplicate samples were plotted.
Genome-wide ChIP-chip analysis for PPAR␥and C/EBP␣binding sites in adipocytes was described elsewhere (51).
Preparation of Nuclear Extract-Nuclear extract from 3T3-L1 preadipocytes and adipocytes were prepared as described previously (52) with slight modifications. In brief, ϳ2 ϫ 10 7 cells were washed twice with ice-cold phosphatebuffered saline, scraped in 3 ml of phosphate-buffered saline, and spun briefly at 3,000 ϫ g. They were resuspended in 10 -20packed cell volumes of cold Buffer A (10 mM HEPES-KOH pH 7.9, 1.5 mM MgCl 2 , 10 mM KCl, 0.5 mM dithiothreitol (DTT), 0.2 mM phenylmethanesulfonyl fluoride (PMSF), and Complete protease inhibitor mixture (Roche Applied Science)), and allowed to swell on ice for 10 min. Then, the cells were subjected to a tissue homogenizer with a B pestle, monitoring lysis of the cytosol fraction with a microscope. The nuclei were pelleted down at 3,000 ϫ g at 4°C for 5 min. The nuclei from adipocytes were further washed twice with the same volume of Buffer A to get rid of residual lipid. The pellet was resuspended with 3-4 packed nuclear volumes of cold Buffer C (20 mM HEPES-KOH pH 7.9, 25% (vol/vol) glycerol, 1.5 mM MgCl 2 , 420 mM KCl, 0.5 mM DTT, 0.2 mM PMSF, and 1ϫ Complete), rocked at 4°C for 30 min, and spun at 15,000 ϫ g for 2 min at 4°C. The supernatant fraction containing nuclear protein was collected and subjected to dialysis in dialysis buffer containing 20 mM HEPES-KOH pH 7.9, 25% (vol/vol) glycerol, 1.5 mM MgCl 2 , 50 mM KCl, 0.5 mM DTT, 0.2 mM PMSF, and 1ϫ Complete to reduce the concentration of KCl.
2 g of nuclear extract or 2 l of in vitro translated mouse PPAR␥ and RXR␣, synthesized using the TNT T7-coupled transcription/translation system (Promega), were used for the binding reaction. The supershift assay was performed by adding 0.1 g of rabbit polyclonal IgG specific for C/EBP␣, C/EBP␤, C/EBP␦, PPAR␥, or control IgG (sc-61, 746, 636, 7196, and 2027, Santa Cruz Biotechnology) to each reaction. All the reactions were carried out at room temperature, and each reaction was preincubated 10 min before the addition of probe (10,000 cpm). To resolve the complexes, the reactions were applied to 4 or 5% native acrylamide gels (acrylamide-bisacrylamide, 29:1) in 0.5ϫ TBE buffer (44.5 mM Tris, 44.5 mM boric acid, 1 mM EDTA) and 5% (vol/vol) glycerol.

Histone H3 Lysine Acetylation and Methylation Profiles Identify a Peak of Intergenic Enrichment 9-kb Upstream of the Retn Transcription Start Site in Mouse
Adipocytes-To identify potential cis-acting regulatory sequences that may reside far from the Retn proximal promoter, we examined histone modifications along the region of mouse chromosome 8 containing Retn and several of its neighboring genes. This strategy was based on recent observations from genome-wide mapping studies indicating that the chromatin of transcriptionally active genes is enriched for specific post-translational histone modifications. Although the combinatorial patterns of histone modifications at active genes can be quite complex (54), it is often the case that regions with high histone H3 lysine 4 trimethylation (H3K4me3) tend to be TSSs, whereas DNA elements containing high levels of H3 acetylation and H3K4me2, but low H3K4me3, tend to be enhancers (55)(56)(57). Thus, active enhancers can be enriched for certain chromatin marks, raising the possibility that they can be identified from histone modification profiles.
We used ChIP-chip to map H3K4me3, H3K4me2, and H3K9 acetylation (H3K9ac) at Retn and its 5Ј-and 3Ј-flanking regions extending 50 kb from the TSS. ChIP was performed in an established tissue culture model of adipogenesis, whereby 3T3-L1 preadipocyte cells can be induced to differentiate into adipocytes (57) to generate chromatin profiles of Retn in its transcriptionally inactive (preadipocyte) and active (adipocyte) states. Fig. 1 shows that H3K4me3 levels at Retn are similar to background in preadipocytes, while a peak of enrichment is present in adipocytes. Retn profiles for H3K4me2 and H3K9ac are similar such that both display background modification levels in preadipocytes and two distinct peaks of enrichment in adipocytes. As expected, one of the peaks is at the Retn TSS. The other is located 9-kb upstream of the TSS, and it has the characteristics of a tissue-specific enhancer given its low level of H3K4me3 and high levels of H3K4me2 and H3K9ac in adipocytes but not preadipocytes, and its location far from the proximal promoter. The evidence for cell type specificity of the enriched regions at or near Retn is strong considering the histone profiles at the neighboring gene Xab2, a pre-mRNA splicing factor. Expression of Xab2 is not adipocyte-specific, and consistent with this each histone mark is enriched similarly at the 5Ј-end of Xab2 in both cell types. Taken together, the data suggest that cis-acting regulatory sequences enhancing Retn transcription may reside ϳ9-kb upstream of the TSS.
Adipocyte-specific Enhancer Element Is Located between Ϫ8872 and Ϫ8676 of the Mouse Retn Gene-To determine whether the region at Ϫ9 kb regulates Retn transcription and to potentially identify additional cis-acting sequences, we fused 5Ј-fragments of the Retn promoter upstream of a luciferase reporter and transfected them into 3T3-L1 adipocytes. Truncations between Ϫ13,598 and Ϫ9004 bp led to minor incremental decreases in reporter activity, while truncation from Ϫ9004 to Ϫ6698 bp decreased activity 8-fold ( Fig. 2A). Promoter studies with a finer series of truncations further defined the region between Ϫ8872 and Ϫ8694 bp to be important for enhanced reporter activity ( Fig. 2A). This small region, ϳ200-bp long, was subsequently fused to a luciferase reporter containing a minimal promoter and transfected into several different cell lines. Compared with the control vector, it increased reporter activity 19-fold in 3T3-L1 adipocytes, but not in HEK 293T and HepG2 cells (Fig. 2B). These results demonstrate that the region of Retn between Ϫ8872 and Ϫ8676 bp is sufficient to activate transcription from a heterologous promoter in a cell type-specific manner, and together with the ChIP-chip analysis, strongly suggest that it functions as an adipocyte-specific enhancer for Retn expression.
C/EBP␣, C/EBP␤, and PPAR␥ Bind to the Retn Adipocyte Enhancer-We have recently reported genome-wide binding locations for PPAR␥ and C/EBP␣ based on ChIP-chip in 3T3-L1 adipocytes (51). For Retn, PPAR␥ and C/EBP␣ were found to occupy a region near the TSS and another overlapping with the region identified as the adipocyte enhancer at Ϫ8.8 kb (Fig. 3A). To confirm binding to these sites, we used quantitative PCR to analyze ChIP DNA isolated from PPAR␥, C/EBP␣, or C/EBP␤ immunoprecipitations. Compared with insulin, which does not contain C/EBP-and PPAR␥-binding sites, the enhancer and promoter showed substantial enrichment in C/EBP␣, C/EBP␤, and PPAR␥ immunoprecipitations from adipocytes, but not preadipocytes, indicating that all three factors bind to these regions in adipocyte cells (Fig. 3B).
Sequence analysis of the Retn enhancer identified three putative binding sites for C/EBP factors, and a region containing moderate homology to the direct repeat-1 (DR-1) PPAR␥binding site (Fig. 4A). To determine if C/EBPs and PPAR␥ can bind these sites and if other trans-acting factors target the Retn enhancer, we performed EMSA. The enhancer was divided into  five fragments of ϳ55 bp, and these were incubated with extract prepared from 3T3-L1 preadipocytes and adipocytes. As shown in Fig. 4C, binding activities abundant in adipocytes, but not preadipocytes, were able to shift probes B, C, and E, but not A and D.
As probes B, C, and E contain the putative binding sites for C/EBP and PPAR␥, we examined whether their gel-shift complexes contain C/EBP factors and/or PPAR␥ by performing EMSA in the presence of antibodies recognizing these proteins. It is often the case that antibody binding to a sequence-specific transcription factor will reduce the abundance of its gel-shift complex either by blocking DNA binding or by supershifting the complex to a higher position in the gel. Indeed, addition of C/EBP␣ antibody reduced a part of gel-shift complexes with probe B and caused a new, slowly migrating complex to appear that likely consists of C/EBP␣, probe B, and C/EBP␣ antibody (Fig. 4D, compare lanes 2 and 3). Addition of C/EBP␤ antibody reduced a part of the gel-shift complexes for probe B, and produced a supershift distinct from that generated by the C/EBP␣ antibody (Fig. 4D, lane 4). These data suggest that protein-DNA complexes for probe B contain C/EBP␣ and C/EBP␤. In contrast, addition of C/EBP␦ antibody had little effect on the abundance or mobility of the gel-shift complexes (Fig. 4D, lane 5). When all three C/EBP antibodies were included in the binding reaction, all of the original gel-shift complexes disappeared, with some of the signal appearing as supershifted complexes (Fig. 4D, lane 6). Similar antibody effects were observed for probe C (Fig. 4D, lanes 7-12), indicating that most, if not all, of the protein-DNA complexes observed with probes B and C are associated with C/EBP␣ and C/EBP␤. Furthermore, the presence of C/EBP␣ and C/EBP␤ in multiple complexes suggests that they may interact with these sites as both homo-and heterodimers. These data agree with earlier in vitro studies demonstrating that C/EBP isoforms are capable of binding to the same site as homo-and heterodimers in all intrafamilial combinations and that the majority of the binding activity in 3T3-L1 adipocytes is associated with C/EBP␣ and C/EBP␤ (59,60).
Consistent with the conclusion that C/EBP␣ and C/EBP␤ are the main adipocyte nuclear proteins bound to probes B and C, addition of PPAR␥ antibody did not alter the gel-shift complexes (data not shown). However, the gel-shift complex generated with probe E was blocked by antibody against PPAR␥, but not C/EBP␣ (Fig. 4E). Probe E contains a putative PPAR␥-binding site, and Fig. 4F reveals that in vitro translated PPAR␥ can bind to it as a heterodimer with RXR␣. As before, antibody specific to PPAR␥ blocked binding. As a whole, the EMSA and ChIP studies make a compelling case that C/EBP␣, C/EBP␤, and PPAR␥ bind to the Retn enhancer in 3T3-L1 adipocytes.

C/EBP␣ and PPAR␥/RXR␣ Function Synergistically to Transactivate Transcription Driven by the Retn Enhancer-To
dissect roles for C/EBP␣ and PPAR␥ in the Retn enhancer function, we transfected 293T cells with the reporter construct from Fig. 2B, whereby the Retn enhancer drives reporter transcription from a heterologous promoter, and with expression plasmids for C/EBP␣ and PPAR␥. In the absence of exogenously expressed protein, the enhancer construct displayed similar transcriptional activity to the control carrying only the heterologous promoter and reporter gene, indicating that 293T cells are devoid of activities that function at the enhancer (Fig. 5).
Cotransfection of either C/EBP␣ or PPAR␥/RXR␣ stimulated the Retn enhancer activity 7-and 2-fold, respectively, indicating that the exogenously expressed factors can function at the Retn enhancer in the context of the reporter construct. Of note, simultaneous expression of both C/EBP␣ and PPAR␥/RXR␣ stimulated the Retn enhancer activity by 22-fold. These results demonstrate that ectopic expression of C/EBP␣ and PPAR␥/ RXR␣ is sufficient for the Retn enhancer activity in non-adipocyte cells, and that both factors working together produce more activity than the sum of activities of either on its own.
The Retn Enhancer Is Not Directly Involved in the Suppression of Retn mRNA by PPAR␥ Ligand-PPAR␥ ligands inhibit Retn expression (4,42), but the molecular mechanism remains to be elucidated. We examined the effect of rosiglitazone, an antidiabetic thiazolidinedione (TZD) drug that functions as a PPAR␥ agonist on the adipocyte-specific Retn enhancer. Rosiglitazone actually stimulated enhancer activity by ϳ3-fold in adipocytes as well as in 293T cells cotransfected with PPAR␥/ RXR␣ and C/EBP␣ (Fig. 6, A and B). The Retn-13-kb luciferase reporter was also stimulated by rosiglitazone (data not shown).
Because the suppression of Retn mRNA could be mediated both transcriptionally and post-transcriptionally, we assessed whether rosiglitazone affected the stability of Retn mRNA. Adi-pocytes were pretreated with DMSO or rosiglitazone (1 M) 48 h, then exposed to 5 g/ml actinomycin D to block transcription. Endogenous Retn mRNA was suppressed by rosiglitazone, as previously reported. Importantly, the treatment with rosiglitazone did not alter Retn mRNA stability (Fig. 6C). Similar effects were seen after simultaneous treatment with rosiglitazone and actinomycin D (data not shown). This suggests that the reduction of Retn mRNA by TZD occurs at the level of transcription. Indeed, Retn precursor mRNA (pre-mRNA) was also suppressed by rosiglitazone (Fig. 6D). These data strongly suggest that Retn mRNA is transcriptionally suppressed by TZDs. However, the mechanism does not directly involve the adipocyte-specific enhancer.
Human RETN Lacks a Functional PPAR␥-binding Site-The molecular explanation for why resistin expression is restricted to adipocytes in rodents and macrophages in humans is unknown. One possibility is that the cis-acting regulatory elements differ for the various resistin orthologs. To explore this idea, we investigated whether the mouse Retn enhancer is conserved in humans. Comparative genomic analysis from the UCSC genome website shows that ϳ150 bp of the mouse Retn enhancer are conserved in the rat at approximately Ϫ7.1 kb, and in the human at approximately Ϫ13.3 kb, relative to the TSS (Fig.  7A). Overall similarity was roughly 66% between human and mouse and 88% between rat and mouse. The three C/EBP-binding sites from the mouse gene were very well conserved for human and rat. Consistent with this, C/EBP factors from 3T3-L1 extract bound to the human and mouse sequences with comparable affinities (data not shown). On the other hand, the PPAR␥-binding site is well conserved in rat, but not human. Indeed, the human sequence displayed little or no binding of in vitro translated PPAR␥ and RXR␣ by EMSA (Fig.  7B). As expected from the binding assays, exogenous expression of C/EBP␣ significantly activated the human sequence in 293T cells, but PPAR␥/RXR␣ did not, either alone or in com- A, sequences with similarity to C/EBP-and PPAR␥-binding motifs are underlined in the mouse Retn enhancer region. Comparison of the putative binding sites to the consensus C/EBP and PPAR␥ motifs is shown. B, schematic representation of the oligonucleotide probes used for this study, with putative C/EBP-and PPAR␥binding sites indicated. C-E, EMSA of activities within 3T3-L1 preadipocyte (Pread) or adipocyte (Ad) nuclear extract (NE) and the indicated probes. F, EMSA with in vitro translated PPAR␥/RXR␣. For the indicated reactions, antibody (Ab) was added to the nuclear extract or in vitro translated protein ϳ10 min before the addition of probe DNA. bination with C/EBP␣ and PPAR␥/RXR␣ (Fig. 7C). Furthermore, the human sequence conferred minimal transcriptional activity in 3T3-L1 adipocytes (Fig. 7D). Thus, human RETN lacks the PPAR␥-binding site that is critical for enhancer function in mouse, which likely accounts for its low expression in adipocytes.

DISCUSSION
In this work, we report a novel adipocyte-specific enhancer at ϳ8.8-kb upstream of the mouse Retn gene, where PPAR␥ and C/EBP factors synergistically enhance adipocyte-specific expression of mouse Retn. We discovered the enhancer by ChIP-chip mapping of chromatin marks, and it was independently confirmed by functional dissection of the 5Ј-flanking sequence by reporter assay, and genome-wide ChIP-chip analysis for PPAR␥and C/EBP␣-binding sites. PPAR␥ and C/EBPs are key transcription factors for adipogenesis, and have been shown to regulate a number of adipocyte genes including adipocyte P2 (aP2) (37)(38)(39), phosphoenolpyruvate carboxykinase (PEPCK) (61,62), and leptin (40,41). In mouse Retn, PPAR␥and C/EBP-binding sites reside very closely within the 200-bp enhancer sequence. This enhancer region alone revealed strong transcriptional activity in adipocytes, but not in other cell types. In a heterologous reporter assay system, PPAR␥/RXR␣ or C/EBP␣ alone only partially stimulated the transcriptional activity of the Retn enhancer in non-adipocytic cells. In contrast, simultaneous expression of PPAR␥/RXR␣ and C/EBP␣ synergistically augmented the robust transcriptional activity of the enhancer. Recent genome-wide ChIP-chip analysis determined that C/EBP␣ binds nearby at ϳ60% of PPAR␥-binding sites in 3T3-L1 adipocytes (51), and a simultaneous report of ChIP with deep genomic sequencing also noted C/EBP-binding sites at a high percentage of PPAR␥ binding locations (63). These data strongly suggest that cooperation between PPAR␥ and C/EBP factors might be important for other adipocyte genes beyond Retn. At this point, the specific mechanism of the synergism between PPAR␥ and C/EBP␣ is not clear, and further study is required.
Unlike rodent Retn, human RETN is only minimally expressed in fat, but predominantly expressed in mononuclear cells including macrophages (7,12,13). The mechanism defining the different expression profile was unclear. Although the proximal promoter sequence of human RETN is quite divergent from that of mouse, it seems to conserve binding sites for C/EBP factors, Sp1, and adipocyte determination-and differentiation-dependent factor (ADD1)/sterol regulatory element-binding protein 1c (SREBP1c) as in the mouse promoter (44,64). In this study, we found that human RETN possesses a sequence that has high homology to the fat-specific enhancer of mouse Retn at Ϫ13.3-kb upstream of the TSS. It conserves three C/EBP-binding sites, but lacks the functional PPAR␥-binding site, and revealed little transcriptional activity in 3T3-L1 adipocytes. This low level expression of the human homologous sequence in mouse adipocytes is presumably due to the lack of the PPAR␥-binding site, though it is also possible that interspecies differences in epigenetic machinery, cellular environment, or transcription factors affect the different expression pattern. A recent report formally demonstrated that the genetic sequence is largely responsible for directing transcriptional programs in homologous tissues (65). This supports our conclusion that the lack of the functional PPAR␥-binding site in human RETN could explain the minimal expression of human RETN in fat.
Retn was originally identified as a gene that is suppressed by anti-diabetic drugs, TZDs (4). Rosiglitazone reduced Retn expression with an ED 50 similar to its K d for binding to PPAR␥, suggesting that this effect is mediated through PPAR␥ (42). Here we demonstrated that rosiglitazone suppresses Retn mRNA transcriptionally by demonstrating that it does not destabilize Retn mRNA and suppresses Retn mRNA at pre-mRNA level. However, rosiglitazone stimulated the mouse Retn enhancer activity in 3T3-L1 adipocytes as well as in 293T cells cotransfected with C/EBP␣ and PPAR␥/RXR␣. Retn-13kb luciferase reporter activity was also stimulated by rosiglitazone. These observations suggest that the PPAR␥-binding site of mouse Retn functions as a positive response element, and the negative regulation of Retn transcription is not directly mediated by cis-regulatory elements present in the ϳ13-kb upstream region of Retn that is sufficient for robust adipocytespecific expression. Human RETN, in which the PPAR␥-binding site is not conserved, has also been reported to be repressed by TZDs (7, 66 -68), further suggesting that the PPAR␥-binding site in the mouse enhancer is not necessary for the down-regulation of Retn mRNA. Because, so far, no additional PPAR␥-binding sites have been found around the Retn locus by our genome-wide ChIP-chip analysis, the repression effect might be mediated by indirect binding of PPAR␥, or inducing transcriptional repressors, or modifying transcriptional activity of other transcription factors.
Another interesting feature of mouse Retn is that it is selectively expressed in white adipose tissue (WAT), but not in brown adipose tissue (BAT) (4,10,58). BAT is characterized by densely packed mitochondria, and serves to dissipate energy to produce heat in response to cold in contrast to WAT. PPAR␥ and C/EBPs cannot explain the WAT-specific expression of Retn, since these transcription factors appear to play roles in the differentiation of brown adipocytes as well (69). Recently, Spiegelman and co-workers have discovered PRDM16 (PRD1-BF1-RIZ1 homologous domain containing 16) as a master regulator of brown adipogenesis, which can potentially explain the low expression of Retn in brown fat. PRDM16 is enriched in BAT compared with WAT. Ectopic expression of PRDM16 in white fat cell progenitor could induce brown specific genes and suppress white specific genes including Retn. Conversely, depletion of PRDM16 in brown fat cells causes loss of brown fat characteristics (58). PRDM16 suppresses WAT-specific genes, including Retn, through binding to C-terminal-binding protein-1 (CtBP-1) and CtBP-2 (48). CtBPs are corepressors, and associate with multiple histone modifying enzymes including histone deacetylaces (HDAC1 and HDAC2), histone methyltransferases (G9a and GLP) and a histone methyl transferase (LSD1) (70). Indeed, transient expression of PRDM16 suppressed the Retn Ϫ13kb luciferase reporter gene in 3T3-L1 adipocytes (48). However, the molecular mechanism by which PRDM16 represses the expression of Retn in BAT is still unclear, especially which transcription factor is targeted by PRDM16. PRDM16 has two distinct DNA binding domains that recognize distinct binding motifs (71). Although PRDM16 R998Q mutant, which completely lacks DNA binding ability, could repress the resistin expression (58), FIGURE 7. Human RETN lacks a functional PPAR␥-binding site. A, sequence alignment of the mouse, rat, and human resistin genes based on comparative genomic analysis. Sequence coordinates are relative to the resistin TSS for each species. C/EBP and PPAR␥ motifs from the mouse adipocyte enhancer are underlined. B, EMSA examining binding of in vitro translated PPAR␥/RXR␣ to the mouse PPAR␥-binding site (mouse) or the corresponding region of human RETN (human). C, luciferase activity of transfected plasmids in 293T cells. The human sequence corresponding to the mouse enhancer (Ϫ13262 to Ϫ13110 bp, hRETN-Enh) is fused to a luciferase reporter vector and cotransfected with or without C/EBP␣ and PPAR␥/RXR␣ expression vectors. D, luciferase activity of transfected plasmids in 3T3-L1 adipocytes. Reporter gene expression is driven by the mouse enhancer (Ϫ8843 to Ϫ8695 bp, mRetn-EnhC) or the corresponding sequence of human RETN (hRETN-Enh).
it is still possible that PRDM16 directly binds to cis-elements through another DNA binding domain. More recently, Spiegelman and co-workers demonstrated that brown, but not white, fat cells arise from common progenitor cells to muscle by lineage-tracing experiment, and PRDM16 stimulates brown adipogenesis by directly binding to PPAR␥ and enhancing its transcriptional activity (72). Further studies are required to elucidate the mechanism of white adipocyte specific expression of Retn.
In summary, this study reports a novel adipocyte-specific enhancer at ϳ8.8-kb upstream of mouse Retn. PPAR␥ and C/EBP factors synergistically activate fat-specific expression of mouse Retn from the enhancer. The lack of the functional PPAR␥ site in human RETN may explain its low expression level in fat. Further studies are necessary to elucidate the mechanism of WAT-specific expression and down-regulation by TZDs.