Fad24, a mammalian homolog of Noc3p, is a positive regulator in adipocyte differentiation

Obesity is a major risk factor for diseases such as cardiovascular disease, type-2 diabetes mellitus, hypertension, stroke, dyslipidemia and some cancers (Must et al., 1999; Visscher and Seidell, 2001) and the prevalence of obesity is increasing globally (Must, 1999; Visscher and Seidell, 2001). It is now clear that the adipose tissue secretes proteins that have various physiological roles (Fruhbeck et al., 2001; Rosen, 2002): for example, leptin and adiponectin regulate energy homeostasis through food intake (Zhang et al., 1994) and insulin sensitivity (Hu et al., 1996b; Maeda et al., 2002). Plasminogen activator inhibitor-1 has an influence on blood coagulation by inhibiting the function of plasminogen activator (Samad et al., 1996). Angiotensinogen affects blood pressure via the renin-angiotensin system (Saye et al., 1989). Tumor necrosis factor-α, (TNF-α), interleukin-6 and adipsin, all of which affect immune function, are also secreted (Cook et al., 1987; Hotamisligil et al., 1993). When an individual becomes obese, the normal balance of proteins secreted from adipose tissue is upset. This disruption can lead to the diseases indicated above.

Obesity is the result of an over-accumulation of adipose tissue resulting from both the swelling of individual adipocytes and an increase in their number. Consequently, to elucidate the molecular mechanisms of obesity, analyses of both lipid accumulation in adipocytes and the development of adipocytes are important. Several key transcription factors have already been identified and three classes of transcription factor are involved in adipocyte differentiation (Rosen, 2002; Rosen et al., 2000). These include peroxisome proliferator-activated receptor γ (PPARγ), the CCAAT/enhancer-binding protein (C/EBP) family and sterol regulatory element-binding protein-1 (SREBP-1), also called adipocyte determination and differentiation factor 1 (ADD1). The molecular mechanisms of expression and function of these genes during adipogenesis have been well characterized by studies with mouse 3T3-L1 cells, a model for the adipogenesis leading to mature adipocytes caused by stimulation with inducers. C/EBPβ and C/EBPδ are expressed comparatively early in adipogenesis. These expression levels reach a maximum within two days of induction (Yeh et al., 1995; Cao et al., 1991). C/EBPα and PPARγ are induced in turn by C/EBPβ and C/EBPδ. The expression levels of C/EBPα and PPARγ peak several days after induction. Moreover, these transcription factors are also known to influence adipocyte differentiation in vivo. Thus, the adipocyte differentiation from metaphase to anaphase is comparatively well understood. However, the early phase expected to be a trigger of adipogenesis remains unknown. Recently, some interesting studies have been done on this period of adipogenesis. Wnt-10b signaling activity was found to inhibit adipocyte differentiation at the very beginning of the differentiation (Ross et al., 2000). TNF-α (Hu et al., 1996a), preadipocyte factor-1 (Smas et al., 1997), and transforming growth factor-β1 (Choy et al., 2000) are also known to regulate adipocyte differentiation by inhibiting the differentiation at the earliest stage.

In our laboratory, we have been focusing on the earliest stage of adipogenesis. Using mouse 3T3-L1 fibroblastic cells, we have already isolated 102 genes whose expression increases in the early stage of the differentiation process using the polymerase chain reaction (PCR) subtraction-cloning method (Imagawa et al., 1999; Kitamura et al., 2001; Nishizuka et al., 2002). Of these 102 genes, we analyzed two genes, regulator of G protein signaling 2 (RGS2) and TC10-like/TC10βLong (TCL/TC10βL). We found that the expression of RGS2 (Nishizuka et al., 2001) and TCL/TC10βL (Nishizuka et al., 2003) peaks at the very beginning of the differentiation and that RGS2 and TCL/TC10βL have the ability to independently regulate adipogenesis via the PPARγ pathway.

However, a complex series of transcription, signal transduction, and morphological events underlies adipocyte differentiation and it is not possible to explain the entire phenomenon with the factors known at present. Of 102 genes isolated, 46 had no significant similarity with genes of known function listed in the GenBank/EMBL databases. The first step in this pursuit was to isolate the full-length open reading frames (ORFs) of the unknown genes. Therefore, in this study, we focused on one of these genes, fad24 (factor for adipocyte differentiation, clone number 24), and found that FAD24 contains a basic leucine zipper (bZIP)-like domain and a NOC domain at its C-terminal end. Recently, yeast protein Noc3p (nucleolar complex-associated protein), with a sequence sharing about 30% similarity with fad24 was reported (Milkereit et al., 2001). At present, a mammalian homolog of Noc3p is not listed in the database and we concluded that fad24 is a mammalian homolog of yeast Noc3p. The expression level of fad24 was elevated during the adipogenesis of mouse 3T3-L1 cells. Moreover, the ectopic expression of the fad24 gene changes with adipocyte differentiation in the presence of BRL49653, which is a ligand for PPARγ. The expression of PPARγ and its targets, adipocyte fatty acid-binding protein 2 (aP2) and lipoprotein lipase (LPL) were also elevated. The results strongly indicate that fad24 is closely involved in adipocyte differentiation.

Total RNA was extracted with TRIzol (Gibco BRL Life Technologies, Grand Island, NY) according to the manufacturer's instructions. For northern blot analyses, 15-25 μg total RNA was separated on a 1% agarose gel containing 2% formaldehyde and transferred to a Hybond-N⁺ nylon membrane (Amersham Biosciences). For northern blot analysis of human tissues, a Multiple Tissue Northern (MTN™) blot containing ∼2 μg of poly(A)⁺ RNA per lane was purchased (BD Biosciences Clontech, Palo Alto, CA). Hybridizations were performed under stringent conditions with Express Hyb (BD Biosciences Clontech), according to the manufacturer's directions. Each probe was labeled with [α-³²P]dCTP using a BcaBEST labeling kit (Takara Biomedicals, Kusatsu, Japan). The subtracted fragment (SU; 2141-2501 bp) of mouse cDNA was used as a probe for detection of mouse fad24. For northern blot analysis of human fad24, a cDNA fragment corresponding to a part of ORF (1124-1603 bp; GenBank accession number AB077992) was used as a probe.

As mouse fad24 cDNA was isolated as a small 361 bp fragment, 5′-rapid amplification cDNA ends (5′-RACE) and reverse transcriptase-polymerase chain reaction (RT-PCR) were used for cloning the full-length cDNA. 5′-RACE was performed using a Marathon cDNA Amplification Kit (BD Biosciences Clontech) following the instructions of the manufacturer. Total RNA was prepared from 3T3-L1 cells 3 hours after induction. mRNA was isolated from total RNA using oligotexdT30 (Daiichi Pure Chemicals, Tokyo, Japan) according to the manufacturer's directions. Single strand cDNA was amplified with oligo-(dT) primer and AMV reverse transcriptase. The second strand cDNA was synthesized using a second-strand enzyme cocktail containing RNase H, E. coli DNA polymerase I and E. coli DNA ligase. The resultant double stranded cDNA was ligated to a Marathon cDNA adapter using T4 DNA ligase. PCR was performed using the forward primer AP-1: 5′-CCATCCTAATACGACTCACTATAGGGC-3′ and a fad24-specific reverse primer: 5′-TCGGGCTTGAGGGCCTCAGAGCC-3′. RT-PCR was carried out with ReverTra Dash (Toyobo, Osaka, Japan) according to the manufacturer's directions. Total RNA was isolated from 3T3-L1 cells 3 hours after induction as described above. The single strand cDNA was synthesized using a random primer and ReverTra Ace (Toyobo). PCR was performed with KOD plus and a fad24-specific forward primer: 5′-CGAGTCTCAGGGCAGCGGGGTGTT-3′ and reverse primer: 5′-GGACTGCCTCCACAGACTAG-3′. The fragments obtained from 5′-RACE and RT-PCR were subcloned into a T-added EcoRV site of pBluescript KS⁺.

The sequence was determined with an automated sequencer DSQ-1000 (Shimadzu Corp., Kyoto, Japan) and an ABI PRISM 310 (Applied Biosystems, Foster City, CA). The human ortholog of fad24 was predicted using the human genome database. Searches for the human ortholog in the human genome databases were performed using BLAST programs accessed via the NCBI homepage (http://www.ncbi.nlm.nih.gov).

Mouse 3T3-L1 (ATCC CL173) preadipocyte cells were maintained in Dulbecco's modified Eagle's medium (DMEM) containing 10% calf serum. For the differentiation experiment, medium was replaced with DMEM containing 10% fetal bovine serum (FBS), 10 μg/ml insulin, 0.5 mM 3-isobutyl-1-methylxantine (IBMX) and 1 μM dexamethasone 2 days post-confluence. After another 2 days, the medium was changed to DMEM containing 5 μg/ml insulin and 10% FBS and was replaced every 2 days. Mouse NIH-3T3 (clone 5611, JCRB 0615) fibroblast cells were maintained in DMEM containing 10% calf serum. NIH-3T3 cells overexpressing fad24 were induced in the same medium as 3T3-L1 cells 2 days post-confluence, with or without 0.5 μM BRL49653. HeLa cells were maintained in minimal essential medium (MEM) containing 10% FBS. C2C12 mouse myoblasts were maintained in DMEM containing 10% FBS. For the differentiation experiment, the medium was replaced with DMEM containing 2% horse serum.

Fat cells were prepared as described previously (Shimba et al., 2003). In brief, epidermal fat pads were isolated from 6-week-old male C57Bl/6J mice, washed with sterile PBS, minced and washed with Krebs-Ringer bicarbonate (KRB) buffer (pH 7.4). Then, the minced tissue was digested with 1.5 mg/ml of collagenase type II (Sigma Aldrich, Saint Louis, MO) in KRB buffer, containing 4% bovine serum albumin at 37°C for 1 hour on a shaking platform. The undigested tissue was removed with a 250 μm nylon mesh and the digested fraction was centrifuged at 500 g for 5 minutes. The adipocytes were obtained from upper layer, washed with buffer and centrifuged to remove other cells. The stromal vascular fraction was resuspended in erythrocyte lysis buffer (150 mM NH₄Cl, 25 mM NH₄HCO₃ and 1 mM EDTA, pH 7.7), filtered through 28 μm nylon mesh and then precipitated at 500 g for 5 minutes.

The five target regions: (1) 207-227 bp; (2) 230-250 bp; (3) 538-558 bp; (4) 565-585 bp; (5) 1069-1089 bp were selected according to Qiagen siRNA online design tool (http://sirna.qiagen.com/) for the RNAi of fad24. A 19 nucleotide short-hairpin RNA (shRNA) coding fragment with a 5′-TTCAAGAGA-3′ loop was subcloned into the ApaI/EcoRI site of pSilencer 1.0-U6 (Ambion, Inc. Austin, TX). As a negative control, the scrambled fragment of region 2 was generated. Transfection of shRNA expression vectors into 3T3-L1 cells were performed by Nucleofector (Amaxa, Cologne, Germany) using Cell Line Nucleofector Kit V (Amaxa). 3T3-L1 cells were harvested and resuspended in Nucleofector solution at 1.0×10⁶ cells/100 μl. After addition of 9 μg shRNA expression vectors, the cells were transfected by program `T-20' of Nucleofector. Then, the cells were plated on sixor 24-well plates. The transfected 3T3-L1 cells were applied to differentiation experiments 3 days after the transfection. Differentiation experiments were carried out using the same medium as described above.

Isolation and reverse transcription of total RNA were done as described above. ABI PRISM 7000 sequence detection system (Applied Biosystems) was used to perform Q-PCR. The pre-designed primers and probe sets of PPARγ, C/EBPα, C/EBPβ, C/EBPδ, SREBP-1, LPL and 18S rRNA were obtained from Applied Biosystems. The primers and probe set for fad24 were designed with forward primer: 5′-TGCACAGAACACCGCACTGT-3′ and reverse primer: 5′-GCAAACCTTCGCACAATGG-3′; the probe was 5′-(FAM)CACGCTTCGGAGACAT(MGB)-3′. For β-actin detection, we used forward primer: 5′-ACGGCCAGGTCATCACTATTG-3′ and reverse primer: 5′-ATGGATGCCACAGGATTCCA-3′; with probe: 5′-(FAM)ACGAGCGGTTCCGAT(MGB)-3′. The reaction mixture was prepared using a TaqMan Universal PCR Master Mix (Applied Biosystems) according to the manufacturer's instructions. The mixture was incubated at 50°C for 2 minutes and at 95°C for 10 minutes, and then the PCR reaction was performed at 95°C for 15 seconds and at 60°C for 1 minute for 40 cycles. The relative standard curves were generated in each experiment to calculate the input amounts of the unknown samples.

NIH-3T3 cells stably expressing fad24 were established using a Retro-X-system (BD Biosciences Clontech) according to the manufacturer's instructions, except that pDON-AI (TaKaRa) was used as a retroviral vector. The full-length cDNA of fad24 was subcloned into the vector pDON-AI. pDON-AI fad24 or pDON-AI empty vector was transfected into PT67 packaging cells by calcium phosphate coprecipitation. The virus transiently expressed in medium was collected and introduced into the target NIH-3T3 cells in the presence of polybrene. Then, the infected cells were cultured in medium containing 0.5 mg/ml neomycin (G418). After 14 days of culture, cells that formed a single colony were isolated, stored individually and used for the adipocyte differentiation analyses.

The pEGFP-fad24 chimeric plasmid (EGFP-FAD24) was constructed by subcloning the coding region in-frame into the 3′-end of pEGFPC1 (BD Biosciences Clontech). EGFP-truncated FAD24-expressing plasmid, pEGFP-Δfad24 chimeric plasmid (FAD24 ΔZIP) was generated by removing amino acids 520-682 with appropriate restriction enzymes. NIH-3T3 cells or HeLa cells were plated onto cell disk (Sumitomo Bakelite, Tokyo, Japan) 1 day before transfection. The cells were transfected using Fugene6™ transfection reagent (Roche, Mannheim, Germany) and after a 24-hour incubation, the cells were fixed with 3% paraformaldehyde, 0.1 mM CaCl₂ and 0.1 mM MgCl₂ for 20 minutes at room temperature. α-amanitin was added at a final concentration of 50 μg/ml to the medium as necessary, and the cells were further incubated for 5 hours and fixed with 3% paraformaldehyde containing 0.1 mM CaCl₂ and 0.1 mM MgCl₂. After blocking of the aldehyde group with 50 mM NH₄Cl in PBS for 5 minutes at room temperature, the cell disk was incubated with primary antibody, mouse monoclonal SC35 antibody (S4045) (Sigma) at 1:50 dilution in PBS for 1 hour at room temperature. After washes with fresh PBS five times, the secondary antibody labeled with TRITC (Sigma) was added for 30 minutes at room temperature. After washes with fresh PBS five times, EGFP and TRITC signals were detected by fluorescence microscopy.

We previously isolated 102 clones that are expressed 3 hours after induction by the PCR-subtraction method. Of these, 46 clones contain currently unknown genes, as BLAST searches of databases identified no significant matches against proteins of known function (Imagawa et al., 1999; Kitamura et al., 2001; Nishizuka et al., 2002). Fad24 was one of these 46 unknown genes. In this study, we first determined the expression level of fad24 and various adipogenic markers at the early and late stages of adipocyte differentiation of mouse 3T3-L1 cells by northern blot analysis (Fig. 1). Expression of fad24 was undetectable in the cells before induction. However, its expression level was quickly elevated after induction reaching a peak between 3 and 6 hours after which it declined until 24 hours after induction (Fig. 1A). For exact quantification of the signal for fad24 mRNA, we next performed Q-PCR for fad24 during the early and late stages of adipogenesis of 3T3-L1 cells. Fad24 is expressed specifically and transiently at the early stage of adipocyte differentiation (Fig. 1B).

We also determined the expression of various adipogenic markers at the same time points (Fig. 1C). Expression of PPARγ, C/EBPα, SREBP-1 and LPL was detected from 2 to 4 days after stimulation with adipogenic inducers. C/EBPβ and C/EBPδ expression was quickly elevated, reached a peak 1-6 hours after induction, and declined slowly at the later stages. Thus, fad24 is expressed just after the expression of C/EBPβ and C/EBPδ, indicating that fad24 may function as a novel factor for the earliest stage of adipogenesis.

The cDNA fragments isolated by PCR-subtraction were only 300-500 bp long as the amplified fragments were digested with RsaI to prevent bias (Imagawa et al., 1999). The length of the cDNA fragment of fad24 obtained was 361 bp. Therefore, isolation of a full-length cDNA of fad24 was necessary and was done using 5′-RACE and RT-PCR as shown in Fig. 2. The cDNA prepared from 3T3-L1 3 hours after induction was used as a template for 5′-RACE and a 2.3 kb cDNA fragment was isolated. A BLAST search detected a 424 bp mouse expressed-sequence tag (EST) fragment in the 3′-flanking region of fad24. Primers for RT-PCR were therefore designed from the 5′-RACE fragment and mouse EST fragment to amplify the 3′-flanking region of the subtracted fad24 fragment and we identified a 693 bp fragment. The combined sequences of the subtracted fragment and the fragments obtained by 5′-RACE and RT-PCR finally resulted in a 2782 bp full-length cDNA fragment with an ORF of 807 amino acids (GenBank accession number AB077991).

To predict the full-length human ortholog cDNA of fad24, the human genome database was used (International Human Genome Sequencing Consortium, 2001). BLAST searches with the mouse fad24 cDNA sequence against the human genome database revealed a genomic fragment at locus 10q24 of human chromosome 10. By splicing out the introns and combining the exons, the full-length human ortholog of fad24 was predicted. Then, in order to isolate the human ortholog of fad24 cDNA, primers for RT-PCR were designed with reference to the predicted human fad24 sequence and the template cDNA was produced from RNA extracted from HeLa cells. By RT-PCR we obtained a 2450 bp cDNA containing a full-length ORF of the fad24 human ortholog (GenBank accession number AB077992). Interestingly, the number of exons and introns in fad24 in mouse and human are the same, and the length of each exon is preserved in almost all of the exons in mouse and human fad24.

The cDNA clones containing the full-length ORF for mouse and human fad24 were isolated as described above. In the mouse fad24, an ATG at position 100 bp, at the 5′-end within the coding region of the cDNA, is likely to serve as the start codon for translation initiation, although the sequence flanking this ATG does not conform very well to Kozak's rule (Kozak, 1997). This codon initiates an open reading frame extending to position 2520 (Fig. 2), and we predicted that the protein sequence is 807 amino acids long (Fig. 3A). Likewise, in the human ortholog of fad24, ATG at position 26 bp seems to be the first methionine, and this codon initiates an open reading frame extending to position 2400, which will be translated into a sequence 800 amino acids long. Thus, it is likely that the human FAD24 is seven amino acids shorter than the mouse ortholog (Fig. 3A).

The deduced amino acid sequences of mouse and human FAD24 revealed the presence of a bZIP-like structure. Although these bZIP sequences are well conserved in mouse and human FAD24, they are slightly different from the typical bZIP sequence found in c-jun, c-fos, the C/EBP family and the ATF family including cAMP response element binding protein (CREB) (Fig. 3B).

The NOC domain is also conserved at the C-terminal end of both mouse and human FAD24 (Fig. 3C). Recently, the yeast protein Noc3p was reported as a factor required for replication initiation and pre-rRNA processing (Milkereit et al., 2001; Zhang et al., 2002). In these reports, a human EST fragment (GenBank accession number AK022882), which shows 100% similarity to position 977-2450 of human fad24, was reported as a human ortholog of Noc3p. Accordingly, there is a possibility that fad24 is a mammalian ortholog of Noc3p.

To determine the tissue expression patterns of human fad24 mRNA, we next performed a northern blot analysis. As shown in Fig. 4A, a single transcript of 4.4 kb was obtained in most adult tissues. It is of interest that human fad24 was highly expressed in skeletal muscle. A moderate level of expression was observed in heart, kidney, liver and placenta whereas the expression in brain, colon, thymus, small intestine, lung and peripheral blood leukocytes was weak.

Expression of fad24 in the mouse adipocyte was also determined. The adipose tissue isolated from mouse was fractionated into a stromal vascular fraction and adipocytes and the expression of fad24 in both fractions was analyzed by northern blotting. Expression of fad24 was undetectable in the stromal vascular cells and in the adipocytes (Fig. 4B).

As high expression was observed in the skeletal muscle, we next analyzed the expression level of fad24 during myogenesis of mouse C2C12 cells. Expression of fad24 was weak, but was detected in the late stage of the differentiating C2C12 cells (Fig. 4C). Interestingly, it was also transiently induced in the initiation step, although the expression level was very low. Indeed, 2 days after differentiation, fad24 expression in C2C12 cells was only about one-sixth of that observed by Q-PCR 6 hours after induction in 3T3-L1 cells (data not shown). Thus, although fad24 was highly expressed in skeletal muscle, the expression level during myogenesis of C2C12 cells seems to be low.

As described above, the expression of mouse fad24 is rapidly induced at an early stage in the differentiation of 3T3-L1 cells into adipocytes. In addition, yeast protein Noc3p is reported to participate in replication initiation and pre-rRNA processing (Milkereit et al., 2001; Zhang et al., 2002). However, the function of fad24, the mammalian ortholog of Noc3p, remains unknown. To characterize the functional roles of this gene during adipogenesis, we performed RNAi to knock down the expression level of fad24 during adipogenesis. The five shRNA expression vectors with different target regions were transfected into 3T3-L1 cells. Six hours after induction, total RNAs were isolated and the expression level of fad24 was determined by RT-PCR. As a result, one of the five target regions, region 2 had the highest reduction in activity (data not shown) and for further analysis we therefore used a shRNA expression vector for this region.

As RT-PCR is semi-quantitative, we confirmed the expression level of fad24 at each time point after induction using Q-PCR. The expression of fad24 in cells transfected with shRNA expression vector was reduced at each time point when compared with that of the control cells (Fig. 5A). Using these transfected cells, we carried out a differentiation experiment. Three days after transfection, the cells were treated with inducers. After 8 days of induction, the cells were fixed and stained with Oil red O and the amounts of triacylglycerol were determined (Fig. 5B,C). The number of Oil red O-stained cells was lower in the cells transfected with shRNA expression vector, and the accumulation of triacylglycerol was also reduced. Next, we determined the expression levels of adipogenic marker genes by Q-PCR (Fig. 5D). The expression levels of PPARγ, C/EBPα and LPL decreased in the sifad24 cells, indicating that the reduction of fad24 expression inhibits the adipogenesis of 3T3-L1 cells. Interestingly, the expression levels of C/EBPβ and C/EBPδ were unchanged in the sifad24 cells. Rather, C/EBPδ expression seemed to be upregulated. As C/EBPβ and C/EBPδ were expressed in the early stages of differentiation (Fig. 1), we also determined the effect of expression of C/EBPβ and C/EBPδ at 6-24 hours by Q-PCR, and found that the expression was unchanged (Fig. 5E). These results implied that the reduction of fad24 expression during adipogenesis impaired the differentiation ability of 3T3-L1 cells.

Fad24 knock down 3T3-L1 cells generated by RNAi failed to differentiate into adipocytes as indicated above. To characterize the role of fad24 during adipocyte differentiation, we established a stable NIH-3T3 non-differentiating cell line that overexpresses fad24, using a retroviral system. As shown in Fig. 6A, the 4.0 kb endogenous transcripts of fad24 were rarely detected in either cells overexpressing fad24 or control cells harboring empty vector. On the other hand, the 6.4 kb exogenous transcript of fad24 derived from retrovirus vector was detected in fad24-overexpressing cells.

Using these established stable transformants, we next tested whether these cells could differentiate into adipocytes. Stimulation of fad24-overexpressing cells or control cells at 2 days post-confluence with inducers that convert preadipocytes to adipocytes in 3T3-L1 cells, did not result in any morphological change (data not shown). However, when BRL49653, a ligand for PPARγ, was added together with insulin, IBMX, dexamethasone and FBS to fad24-overexpressing NIH-3T3 cells, the cells differentiated into mature adipocytes, and the oil droplets were stained with Oil Red O. In contrast, the control cells retained their fibroblastic morphology even after one week of treatment with inducers containing PPARγ ligand, and the oil droplets were not detected (Fig. 6B).

Next, we determined the expression profiles of adipocyte marker genes. The expression levels of PPARγ and SREBP-1, which are known to be master regulators of adipogenesis, gradually increased until day 8 in cells overexpressing fad24. LPL and aP2, which are targeted by PPARγ, were also expressed about one week after induction. On the other hand, the expression level of the C/EBP family did not change during adipogenesis of fad24-overexpressing cells (Fig. 6C). As the expression of C/EBPβ and C/EBPδ are elevated in the very beginning of adipogenesis as shown in Fig. 1, we determined the expression of these genes in the NIH-3T3 cells overexpressing fad24 and control cells 24 hours after induction. Interestingly, the expression level of C/EBPδ was greatly reduced in the NIH-3T3 cells (Fig. 6D). Taken together, these findings strongly suggest that the fad24 gene has some functional role in adipogenesis and that these effects seem to be linked to the PPARγ pathway rather than the C/EBP pathway.

As fad24 has a bZIP-like motif and Noc3p was reported to be expressed in the nucleus, we next determined the subcellular localization of FAD24 by transfecting EGFP-tagged fad24 into NIH-3T3 cells. The signals were detected by fluorescence microscopy. EGFP-FAD24 localized to the nucleus, whereas EGFP alone as a control was detected throughout the cell (Fig. 7A). Although FAD24 was found in the nucleus upon staining with 4′,6-diamidine-2′-phenylindole dihydrochloride (DAPI), interestingly the EGFP-FAD24 fusion protein localized in compartments within the nucleus.

To pinpoint this localization, we investigated whether EGFP-FAD24 fusion protein colocalizes with PML nuclear bodies, which are the major spotted components in the nucleus. The preliminary result in transfected HeLa cells showed the EGFP-FAD24 fusion protein did not colocalize with PML bodies, although high quality data were not obtained (data not shown). Next, we investigated possible colocalization with nuclear speckles, which are other major spotted components in the nucleus. HeLa cells were transfected with EGFP-tagged fad24 or an EGFP-tagged truncated fad24, which lacks a bZIP-like domain. Then, immunofluorescent staining with anti-SC35 monoclonal antibody, which identifies nuclear speckles, was performed using these transfected cells. The foci produced by EGFP-FAD24 fusion protein colocalized with SC35 in HeLa cells. However, the EGFP-tagged truncated FAD24 protein did not colocalize with SC35 (Fig. 7B). Yeast NOC3P proteins were reported to concentrate in the nucleolus and nucleoplasm (Milkereit et al., 2001). When merged with the phase-contrast image, EGFP signals were also localized to the nucleolus as well as the nuclear speckles (Fig. 7C) as was the case with yeast.

Nuclear speckles are known to be sensitive to RNA polymerase II transcriptional activity (Misteli, 2000; Lamond and Spector, 2003). Therefore, we were interested in whether fad24 would still colocalize with nuclear speckles when RNA polymerase II activity was blocked by inhibitors. As observed before (Boronenkov et al., 1998), treatment with the RNA polymerase II inhibitor α-amanitin caused a morphological change in nuclear speckles where sizes increased and numbers decreased. When EGFP-tagged fad24-transfected HeLa cells were treated with α-amanitin, FAD24 still colocalized with SC35, but in larger speckles present in smaller numbers (Fig. 8). These results strongly indicate a physical association between fad24 and some components of nuclear speckles.

It is well documented that the transcription factors PPARγ, SREBP-1 and the C/EBP family function as master regulators of adipocyte differentiation (Rosen, 2002; Rosen et al., 2000). However, little is known about the early phase in which the adipogenesis is triggered. To provide new insights into this stage of the differentiation of adipocytes, we isolated the genes expressed at the beginning of the differentiation process (Imagawa et al., 1999; Kitamura et al., 2001; Nishizuka et al., 2002). Almost half of the genes isolated had no significant similarity with genes of known function listed in the database (Imagawa et al., 1999; Nishizuka et al., 2002).

In this study, we cloned cDNAs of mouse and human fad24. The mouse and human FAD24 proteins comprised 807 and 800 amino acids, respectively, and the amino acid sequences are highly conserved with 89.6% similarity between these two proteins. With the aim of identifying the genomic distribution of mouse fad24, a BLAST search of the mouse genome database (Mouse Genome Sequencing Consortium, 2002) was done with the mouse fad24 full-length cDNA sequence. The result indicated that mouse fad24 was located at 19c2 of mouse chromosome 19 and constituted 21 exons and 20 introns. In the sequences of the exon/intron junctions, the GT/AG rule was conserved in all cases.

To isolate the human ortholog of fad24, we predicted the human sequence using the Human Genome Database (International Human Genome Sequencing Consortium, 2001). Human fad24 was located at locus 10q24 of human chromosome 10 and was composed of 21 exons and 20 introns. In all the exon/intron junctions, the GT/AG rule was again conserved. By splicing out the introns, the human fad24 sequence was predicted. The isolated human fad24 cDNA was 100% similar to the sequence predicted using the Human Genome Database. Analyses of the deduced amino acid sequences revealed the existence of a bZIP-like domain at the C-terminal end. The bZIP domain was first reported in C/EBP (Landschulz et al., 1988). ATF family members such as CREB, and the nuclear oncogenes c-fos and c-jun also possess this structure (Gentz et al., 1989; Gonzalez et al., 1989). The bZIP domain consisted of a leucine-zipper region, which mediates protein dimerization, and a basic amino acid cluster, which functions as a DNA binding region.

We produced mouse FAD24 protein fused to glutathione S-transferase and examined the ability to bind single and double stranded DNAs. The preliminary results did not show any DNA binding activity. As the second leucine in bZIP was replaced with cysteine in FAD24, FAD24 may not form a homodimer or heterodimer. However, more experiments are needed to determine whether FAD24 acts as a transcription factor. Two different bands were observed in the northern blot analysis of mouse fad24. We could not isolate any splicing isoform in the series of mouse or human fad24 cloning. Moreover, northern blot analysis of human fad24 resulted in a single transcript. Therefore, it seems that these different sizes are derived from the use of different polyA additional signals, although we cannot exclude the possible use of a different promoter.

Recently, Noc3p a yeast protein sharing 30% sequence similarity with FAD24, was reported (Milkereit et al., 2001). In that report, FLJ12820 was isolated as a partial cDNA of the human homolog of Noc3p. By comparing the sequences of FLJ12820 and human fad24, FLJ12820 was found to utilize methionine at position 370 of fad24 as a first methionine for translation. Hence, there is a good possibility that fad24 is a mammalian homolog of Noc3p. Noc3p was first reported as one of the components of the machinery participating in the maturation of pre-ribosomes and transport from the nucleolus to cytoplasm. Recently, a unique feature of yeast NOC3P was reported (Zhang et al., 2002). NOC3P plays a direct role in the initiation of DNA replication by interacting with MCM proteins and ORC, as a cofactor for the establishment and maintenance of pre-RCs in the chromatin association of Cdc6p and MCM proteins, indicating multiple functions.

In our study, fad24 was highly expressed after 3 hours of adipocyte differentiation in 3T3-L1 cells. To investigate the function of fad24, its expression during adipocyte differentiation was blocked by RNAi. The results demonstrated the possibility of fad24 as an accelerating factor for adipocyte differentiation. Moreover, we established fad24-overexpressing NIH-3T3 cells. The adipocyte differentiation test revealed two unique features about this cell line. One is that NIH-3T3 cells overexpressing fad24undergo adipocyte differentiation only in the presence of the ligand PPARγ. The other is that PPARγ expression increased during differentiation of the established cell line, whereas C/EBPα expression did not change. As the PPARγ^-/- phenotype was not rescued by the ectopic expression of C/EBPα whereas the ectopic expression of PPARγ rescued the C/EBPα^-/- phenotype, it is now thought that the PPARγ pathway mainly regulates adipocyte differentiation (Rosen et al., 2002). Our data also support this because of the two reasons indicated above.

Noc3p is reported to localize to the nucleolus and nucleoplasm in yeast (Milkereit et al., 2001). Our data support this finding as fad24 was expressed in the nucleolus. However, it was found that fad24 also localizes to nuclear speckles. Furthermore, its localization was dramatically affected by the reorganization of nuclear speckles with the inhibition of RNA polymerase II. The morphology of the nuclear speckles is strongly linked with transcription activity. When transcription is active, the speckles become smaller and increase in number, because the nuclear speckles, consisting of proteins, small ribonucleoproteins and mRNA-splicing factors, diffuse to the active transcription sites. On the other hand, when the transcription activity is inhibited, the speckles become fewer and enlarged, because these factors accumulate (Misteli, 2000; Lamond and Spector, 2003). When HeLa cells transiently expressing EGFP-FAD24 fusion protein were treated with α-amanitin, a strong inhibitor of RNA polymerase II, FAD24 reorganized the nuclear speckles. These results suggest that fad24 may also function as a component of the splicing machinery.

In this paper, we have demonstrated the isolation and characterization of fad24. The data indicate an important role for FAD24 in adipogenesis, as FAD24 functions as an adipocyte differentiation accelerating factor in vitro, localizes to nuclear speckles and may participate in RNA processing. As fad24 has a NOC domain and localizes in the nucleolus and nuclear speckles, it is very likely that fad24 functions as a regulator for the mitotic clonal expansion, which occurs in the early stage of adipogenesis. To elucidate the biological and biochemical functions of FAD24 in adipocyte differentiation, more studies are required and we are now following this up by producing knockout mice.

We thank Drs B. M. Spiegelman, (Dana-Farber Cancer Institute, Harvard Medical School), S. L. McKnight (University of Texas Southwestern Medical Center) and R. Sato (University of Tokyo) for generously providing the plasmids containing cDNAs of PPARγ, C/EBPs, and SREBP-1, respectively. We also thank GlaxoSmithKline, for the gift of BRL49653. This study was supported in part by grants from the Ministry of Education, Culture, Sports, Science and Technology (MEXT), Japan, Japan Society for the Promotion of Science (JSPS) and ONO Medical Research Foundation, Japan.