Suppression of the Pancreatic Duodenal Homeodomain Transcription Factor-1 (Pdx-1) Promoter by Sterol Regulatory Element-binding Protein-1c (SREBP-1c)*

Overexpression of sterol regulatory element-binding protein-1c (SREBP-1c) in β cells causes impaired insulin secretion and β cell dysfunction associated with diminished pancreatic duodenal homeodomain transcription factor-1 (PDX-1) expression in vitro and in vivo. To identify the molecular mechanism responsible for this effect, the mouse Pdx-1 gene promoter (2.7 kb) was analyzed in β cell and non-β cell lines. Despite no apparent sterol regulatory element-binding protein-binding sites, the Pdx-1 promoter was suppressed by SREBP-1c in β cells in a dose-dependent manner. PDX-1 activated its own promoter. The E-box (−104/−99 bp) in the proximal region, occupied by ubiquitously expressed upstream stimulatory factors (USFs), was crucial for the PDX-1-positive autoregulatory loop through direct PDX-1·USF binding. This positive feedback activation was a prerequisite for SREBP-1c suppression of the promoter in non-β cells. SREBP-1c and PDX-1 directly interact through basic helix-loop-helix and homeobox domains, respectively. This robust SREBP-1c·PDX-1 complex interferes with PDX-1·USF formation and inhibits the recruitment of PDX-1 coactivators. SREBP-1c also inhibits PDX-1 binding to the previously described PDX-1-binding site (−2721/−2646 bp) in the distal enhancer region of the Pdx-1 promoter. Endogenous up-regulation of SREBP-1c in INS-1 cells through the activation of liver X receptor and retinoid X receptor by 9-cis-retinoic acid and 22-hydroxycholesterol inhibited PDX-1 mRNA and protein expression. Conversely, SREBP-1c RNAi restored Pdx-1 mRNA and protein levels. Through these multiple mechanisms, SREBP-1c, when induced in a lipotoxic state, repressed PDX-1 expression contributing to the inhibition of insulin expression and β cell dysfunction.

The homeobox protein PDX-1 serves as the master control switch for expression of exocrine and endocrine pancreatic developmental programs. In adults, PDX-1 expression is restricted to the ␤ and ␦ cells of the pancreas and is required for regulation of insulin (1,2), glucose transporter 2 (3), islet amyloid polypeptide (4 -6), and glucokinase (6) expression. In mice, ␤ cell-selective disruption of Pdx-1 leads to the development of diabetes with increasing age (7). Mice heterozygous for Pdx-1 were found to be glucose-intolerant (7,8) with increased islet apoptosis, decreased number of islets, and abnormal islet architecture (9). In humans, the mutation of Pdx-1 has been associated with the MODY4 locus (10,11), linking PDX-1 to a form of diabetes known as maturity-onset diabetes of the young. Thus, impaired PDX-1 expression appears to be associated with diabetes (12)(13)(14).
Transfection and Luciferase Assays-HIT (3.5 ϫ 10 5 cell/ well) and HepG2 cells (4.2 ϫ 10 4 cell/well) were plated on 12-well plates. The expression plasmids, luciferase reporter plasmid (0.5 g), and internal control pSV-␤-gal (0.5 g) were cotransfected using SuperFect transfection reagent (Qiagen) for HepG2 cells and JetPEI-Man (Polyplus transfection) for HIT cells. The total amount of DNA in each transfection was adjusted to 1.5 g/well. The luciferase activity in transfectants was measured by MicroLumat Plus (EG&G Berthold) and normalized to the ␤-Gal activity measured using a standard kit (Promega).
Adenovirus Infection Studies-INS-1 cells were plated at a density of 4 ϫ 10 6 cells/75-cm 2 dish for 48 h before viral treatment. Cells were transduced with adenoviral vectors encoding the green fluorescent protein (Ad-GFP) as the control or human nuclear SREBP-1c(1-436) (Ad-SREBP-1c) at the indicated multiplicities of infection by incubation for 48 h at 37°C (21). The virus-containing medium was then aspirated, and cells were cultured in 25 mM glucose medium for 6 h. Total RNA was isolated, and 10 g of each sample was subjected to electrophoresis.
ChIP-A ChIP assay was performed as described previously (37). In the experiment shown in Fig. 5C, immunoprecipitated genomic DNA fragments were quantified by a StepOne TM real time PCR system (Applied Biosystems) using Fast SYBR Green Master Mix (Applied Biosystems). Primer sets for the Pdx-1 proximal E-box region were 5Ј-GCCTCCTTCTTA-AGGCA-3Ј (sense) and 5Ј-ATCGCTTTGACAGTTCTCC-3Ј (antisense). The PCR conditions were 20 s at 95°C followed by 42 cycles of 3 s at 95°C and 30 s at 60°C. In the experiment shown in Fig. 9D, primer sets for mouse Pdx-1 promoter area I regions were 5Ј-CCACTAAGAAGGAAGGCCAG-3Ј (sense) and 5Ј-CTGAGGTTCTTTCTCTGCCTCTCTG-3Ј (antisense). The PCR conditions were 5 min at 94°C followed by 31 cycles of 30 s at 94°C, 30 s at 50°C, and 1 min at 72°C. The PCR products were resolved on a 3% agarose gel.
Real Time Quantitative Reverse Transcription (RT)-PCR-Real time PCR was performed using StepOne TM real time PCR (Applied Biosystems) with a TaqMan Gene Expression Cells-to-CT TM kit (Ambion). Primers were purchased from Applied Biosystems. The primers used were rat SREBP-1c (Rn01495769-m1), rat PDX-1 (Rn00755591-m1), rat insulin 1 (Rn0212433-g1), and rat ␤-actin (4352340E). Real time PCR was duplicated for each cDNA sample. Each gene mRNA level was acquired from the value of the threshold cycle (Ct) of real time PCR as related to ␤-actin.
Statistics-Results are expressed as means Ϯ S.E., and the statistical significance was assessed using F tests and Student's t tests for unpaired data.

Overexpression of Adenoviral SREBP-1c in INS-1 Cells Inhibits Expression of Insulin and ␤ Cell-specific Transcription
Factors-To investigate the inhibitory effect of SREBP-1c on ␤ cell function, rat insulinoma INS-1 cells were infected with Ad-SREBP-1c or Ad-GFP as the negative control (Fig. 1A). Ad-SREBP-1c overexpression increased expression of an SREBP-1c target gene, fatty-acid synthase, in a dose-dependent manner.
In contrast, Pdx-1, Beta2, and insulin mRNAs were markedly decreased with the concomitant suppression of insulin gene expression. Moreover, the PDX-1 protein was notably suppressed by SREBP-1c (Fig. 1B). In ␤ cell-specific SREBP-1c transgenic mice, pdx-1 expression decreased, but it increased in SREBP-1c knock-out mice (supplemental Fig. 1) (22). These data indicate that SREBP-1c suppresses PDX-1 at the mRNA level and could contribute, at least partially, to the inhibitory effect of SREBP-1c on ␤ cell function.
SREBP-1c Suppresses the Endogenous Activity of PDX-1 in ␤ Cells but Not in Non-␤ Cells-SREBP-1c overexpression in ␤ cells caused impaired insulin secretion, suppression of insulin and pdx-1 expression, and reduction of ␤ cell mass (22,27). Next, we explored the effect of SREBP-1c on the Pdx-1 promoter. The mouse Pdx-1 extending from Ϫ2721 to ϩ50 bp ( Fig.  2A) was linked to the luciferase reporter plasmid (PDX-1 (Ϫ2.7 k)-Luc) and subjected to a reporter gene assay by transient transfection into hamster insulinoma HIT and non-␤ HepG2 cells (Fig. 2B). PDX-1 (Ϫ2.7 k)-Luc activity was robust in HIT and HepG2 cells (Fig. 2B). SREBP-1c overexpression dose-dependently suppressed the Pdx-1 promoter activity of the 2.7-kb region in HIT cells but not in HepG2 cells (Fig. 2B). This suggested that SREBP-1c probably interacts with ␤ cell-specific endogenous factors and inhibits the Pdx-1 promoter.

SREBP-1c Suppresses PDX-1-Positive Autoregulatory Loop Activity in HIT and HepG2
Cells-Previous studies have identified several transcriptional factors for the Pdx-1 promoter that consists of two important regulatory regions as follows: the proximal and distal enhancer regions ( Fig. 2A). In the proximal region, USFs bind to the E-box (32,33). In the distal enhancer region, located at approximately Ϫ1.9 to Ϫ2.7 kb, several ␤ cell-specific transcription factor-binding sites have been identified. To delineate the sites responsible for SREBP-1c suppression of Pdx-1, we generated a series of 5Ј-deletion constructs and analyzed these reporters by transfection into HIT cells (Fig.  2C). In the absence of SREBP-1c, the deletion of sequences from Ϫ2721 to Ϫ753 bp in PDX-1 (Ϫ2.7 k)-Luc reduced the promoter activity by ϳ50%, and further truncation of the Pdx-1 promoter from Ϫ163 to Ϫ93 bp severely attenuated the promoter activity, indicating that both the distal enhancer and proximal regions are required for complete Pdx-1 promoter activity as previously reported (32). SREBP-1c caused 70, 45, and 39% reduction in the activity of the Pdx-1 promoter containing 2721, 753, and 163 fragments, respectively (Fig. 2C). SREBP-1c-mediated inhibition was completely abolished in PDX-1 (Ϫ93)-Luc. These data suggested that SREBP-1c inhibition could occur at the distal enhancer and proximal regions (between Ϫ163 and Ϫ93 bp).
Several ␤ cell-specific transcription factors such as HNF-3␤/ Foxa2, HNF-1a, Pax6, RIPE3b/Maf, and PDX-1 have been identified in the distal enhancer region ( Fig. 2A). In our experimental setting, the 2.7-kb Pdx-1 promoter was activated by HNF-3␤/Foxa2 and PDX-1 but not by HNF-1a and RIPE3b/Maf in HIT and HepG2 cells. PDX-1 binds to the enhancer element in area I, suggesting a possible autoregulatory loop (hereinafter referred as auto-loop) as a mechanism for its ␤ cell-specific expression (28,40). As in vivo evidence for the requirement for a PDX-1 auto-loop mechanism, the mice homozygous for a targeted deletion of the area I-III enhancer region of Pdx-1 (Pdx-1 ⌬I-III/⌬I-III ) had severely impaired pancreas development (41). Heterozygous mice (Pdx-1 ϩ/⌬I-III ) showed hyperglycemia and reduced insulin secretion (41). Thus, in this study, we focused on the effect of the SREBP-1c on the PDX-1 auto-loop activity. The effects of SREBP-1c on the PDX-1 (Ϫ2.7 k)-Luc activities were examined in HIT and HepG2 cells (Fig. 3). In HIT cells, exogenously transfected PDX-1 slightly, but dose-dependently, up-regulated promoter activity, and SREBP-1c inhibited the activity (Fig. 3A). In HepG2 cells lacking endogenous PDX-1, exogenously transfected PDX-1 robustly induced Pdx-1 promoter activity, representing the PDX-1 auto-loop mechanism (Fig. 3B). Intriguingly, in the presence of PDX-1, SREBP-1c suppression that was not observed in HepG2 cells without PDX-1 emerged in a markedly competitive manner. These results demonstrate that SREBP-1c suppresses the PDX-1 auto-loop activity.
Proximal E-box Is Essential for PDX-1-positive Auto-loop Activity and SREBP-1c Suppression-To determine the essential functioning site for SREBP-1c inhibition of PDX-1 auto- loop activity, a sequential deletion study was performed (Fig.  4A). In HIT cells, PDX-1-Luc-containing fragments from Ϫ2721 to Ϫ163 bp were activated by exogenous PDX-1 (Fig.  4A) and suppressed to basal levels by SREBP-1c (data not shown). In HepG2 cells, promoters longer than PDX-1 (Ϫ93)-Luc were activated by PDX-1 to a similar extent, and SREBP-1c coexpression completely suppressed the activities down to basal levels. PDX-1 (Ϫ93)-Luc, which does not contain the E-box at Ϫ104 bp, completely lost the PDX-1 auto-loop as well as basal activities in both cell lines (Fig. 4A). These data demonstrate that the PDX-1 auto-loop and SREBP-1c inhibition were mediated primarily through the proximal site between Ϫ163 and Ϫ91 bp.
Mutations in the E-box of PDX-1 (Ϫ2.7 k) E-box (m)-Luc and (Ϫ163) E-box (m)-Luc (data not shown) completely abolished the endogenous activities, and they exhibited no response to exogenous PDX-1 in both cell lines (Fig. 4B), confirming that the E-box in the proximal region is crucial for PDX-1-positive feedback regulation.
PDX-1 and USF1 Directly Interact in the Proximal Region-In the electrophoretic mobility shift assay (EMSA) using INS-1 and HepG2 nuclear extracts, only USF1 (or USF2) produced a DNA⅐protein complex in the range of Ϫ120 to Ϫ90 bp (Fig. 5A).
USFs are E-box-binding proteins and important components of  the regulatory apparatus (32). Because consensus cis elements for PDX-1 binding were not found in the proximal region from Ϫ163 to Ϫ93 bp, it is likely that PDX-1 could activate the promoter without DNA binding. Based on data from PDX-1 (Ϫ2.7 k) E-box (m)-Luc experiments (Fig. 4B), we examined the protein-protein interaction between USFs and PDX-1 using GST pulldown assays (Fig. 5B). In vitro translated [ 35 S]methioninelabeled USF1 was incubated with GST-PDX-1 fusion proteins or GST alone. 35 S-USF1 was pulled down with GST-PDX-1 (FL) and GST-PDX-1 (HD), but not GST-PDX-1 (⌬HD) and GST-PDX-1 (CT) with a deleted HD, indicating that USFs directly interact with the homeodomain of PDX-1.
To estimate the direct association between USFs and PDX-1 on the E-box of the endogenous Pdx-1 promoter of INS-1 cells, we performed ChIP and Re-ChIP assays using E-box primers (Fig. 5C, panels a and b). INS-1 cells were cultured in the absence or presence of 9-cis-RA and 22OH-Cho in 10 mM glucose. 9-cis-RA is a ligand for RXRs, and 22OH-Cho is a ligand for LXR. LXR⅐RXR heterodimers activate the SREBP-1c promoter, and these agonists enhance the SREBP-1c expression (42). Immunoprecipitated DNA fragments using IgG, anti-PDX-1, anti-USF1, and anti-USF2 antibodies were quantified by real time PCR. The ChIP assay confirmed USFs binding to the E-box (Fig. 5C, panel c). The results were not affected after SREBP-1c activation by 9-cis-RA and 22OH-Cho. In the Re-ChIP assay, DNA⅐protein complexes immunoprecipitated with anti-USF1 and anti-USF2 antibodies were removed from the beads and were re-immunoprecipitated with the anti-PDX-1 antibody or IgG. ChIP and Re-ChIP assays involving sequential immunoprecipitation confirmed the association of USFs with PDX-1 on the E-box. The signal decreased by 9-cis-RA and 22OH-Cho (Fig. 5C, panel d). These data suggest that PDX-1⅐USF complex formation on the E-box was involved in the PDX-1 auto-loop activity and that SREBP-1c expression interferes with the association between USFs and PDX-1 on the E-box.
In the distal enhancer region, EMSA analysis showed that potent binding of PDX-1 to the probe containing a PDX-1-binding site in area I (Ϫ2517 to Ϫ2492 bp) was dose-dependently inhibited by SREBP-1c (Fig. 6C). SREBP-1c did not bind to these probes. The SREBP-1c inhibition was potent as the dissociation of PDX-1 from the probe was still observed by addition of SREBP-1c after the PDX-1 binding to area I (data not shown).
PDX-1 Inhibits SREBP Target Genes-We tested whether the formation of the SREBP-1c⅐PDX-1 complex inhibits the activation of SREBP target genes by SREBPs in HepG2 cells (supplemental Fig. 3, A and B) using LDL receptor (SRE)-Luc or S14 (E-box)-Luc (43). Neither construct has any PDX-1-binding sites. PDX-1, ⌬ABC-PDX-1, and ⌬HD-PDX-1 had no effects on the basal activity of LDL receptor (SRE)-Luc or S14 (E-box)-Luc. LDL receptor (SRE)-Luc activation by SREBP-1a was inhibited with the coexpression of PDX-1 and ⌬ABC-PDX-1 but not ⌬HD-PDX-1. Similar results were obtained in the activation of S14 (E-box)-Luc by SREBP-1aM. Taken together with the data from Fig. 6C, these results suggest that the SREBP-1c⅐PDX-1 complex has negative effects on both targets. The mutant version ⌬HD-PDX-1 is located in the cytosol because of lack of the nuclear localization signal (RRMKWKK) (supplemental Fig. 3, C and D). Thus, the effect on the interaction between ⌬HD-PDX-1 and SREBP-1c in the nucleus could not be examined by this approach.
In addition, we investigated whether SREBP-1c knockdown could be reversed by increasing the level of the PDX-1 protein (supplemental Fig. 4). Thus, INS-1 cells were infected with Adsi-SREBP-1 or Ad-si-LacZ as a control in the absence or presence of 9-cis-RA and 22OH-Cho and analyzed using real time PCR (supplemental Fig. 4). In the control Ad-si-LacZ-treated cells, 9-cis-RA and 22OH-Cho induced endogenous SREBP-1c. In the Ad-si-SREBP-1-treated cells, this induction was suppressed by 20% (p ϭ 0.023). SREBP-1c induction significantly reduced Pdx-1 mRNA expression (p ϭ 0.032), but in the Ad-si-SREBP-1-treated cells, Pdx-1 mRNA expression was slightly restored (supplemental Fig. 4A). Western blotting analysis demonstrated that 9-cis-RA and 22OH-Cho increased the protein levels of membrane precursor and nuclear SREBP-1c in the control Ad-si-LacZ-treated cells, but they were decreased in the Ad-si-SREBP-1c-treated cells. Reduction of PDX-1 expression in the presence of 9-cis-RA and 22OH-Cho was slightly restored in the Ad-si-SREBP-1c-treated cells (supplemental Fig. 4B).

DISCUSSION
This study clearly demonstrated that SREBP-1c inhibits Pdx-1 promoter activity and represses its gene expression. The activation of SREBP-1c causes ␤ cell dysfunction, including impaired insulin expression and secretion in vitro (INS-1 cells) and in vivo (transgenic mice) (22,27). In both cases, reduction of PDX-1 was observed and was thought to contribute to ␤ cell dysfunction. To evaluate PDX-1 suppression by SREBP-1c in a more physiologically relevant setting, INS-1 cells were cultured with 9-cis-RA and 22OH-Cho in 10 and 25 mM glucose media. 9-cis-RA and 22OH-Cho increased SREBP-1c expression and maturation, leading to inhibition of PDX-1 expression at the mRNA and protein levels (Fig. 9, A-C). The physiological role of SREBP-1c in PDX-1 expression was confirmed by islets from SREBP-1-null mice and by RIP-nSREBP-1c transgenic mice (supplemental Fig. 1) (22). This molecular mechanism operates at the transcription level. Moreover, our data showed that PDX-1 expression is required for SREBP-1c suppression of the Pdx-1 promoter indicating that the target of SREBP-1c is PDX-1 auto-loop activation. To date, this mechanism was reported to operate at the TAAT motif at Ϫ2651 to Ϫ2648 bp in area I (28). The deletion of area I-III from the endogenous Pdx-1 locus results in severely reduced PDX-1 expression in the pancreas, indicating that this enhancer is crucial for PDX-1 expression in ␤ cells (41). In addition to this distal site confirmed by EMSA and ChIP assays, sequential deletion studies revealed another crucial site for the PDX-1 auto-loop, the proximal E-box. USFs play a key role in the transcription of pdx-1 through this E-box (Fig. 10A) (32,33). Considering the critical importance of the E-box, but modest stimulation of the Pdx-1 promoter by USF1 cotransfection (data not shown), it has been speculated that the action of USFs is mediated in synergy with another factor (33). Our data indicate that PDX-1 directly binds to USF1 and contributes to the transactivation of the Pdx-1 promoter. In chromatin, PDX-1 may facilitate or stabilize the formation of a large tran- scription complex by bending DNA by binding to both area I and the USFs/E-box as schematized in Fig. 10A. Other recently identified factors such as HNF3b, MafA, and Foxo1 could also be involved in this complex or mechanism. Using a GST-pulldown assay, we confirmed the direct interaction of HNF-3b⅐SREBP-1c and HNF-3b⅐PDX-1 (data not shown).
Based upon the above notion regarding the PDX-1 auto-loop system, we propose molecular mechanisms by which SREBP-1c inhibits the Pdx-1 promoter in multiple ways. First, SREBP-1c and PDX-1 physically interact through their bHLH and HD, respectively (Fig. 6, A and B). This direct physical interaction could disrupt the PDX-1⅐USF complex on the E-box that we proposed. SREBP-1c⅐PDX-1 inhibits PDX-1 binding to area I for the distal enhancer region (Fig. 10B). EMSA and ChIP analysis supports this (Figs. 6C and 9D). The SREBP-1c⅐PDX-1 complex could also exhibit inhibitory effects on SREBP target genes (supplemental Fig. 3). Second, the Gal4 fusion protein reporter system demonstrated that the AD domain of PDX-1 was crucial for PDX-1 transactivation, and the HD portion with which SREBP interacts was regulatory (Fig. 7B). SREBP-1c strongly and dose-dependently inhibited the transactivation of PDX-1 in this system even in case of the Gal4-PDX-1 fusion protein that does not bind to SREBP-1c, leading us to another important mechanism, i.e. squelching of recruitment of PDX-1 coactivators (Fig. 7D). Consistently, DN-SREBP-1 lacking AD, but retaining the ability to bind to PDX-1, completely lost its inhibitory action on the Gal4-PDX-1 transactivation system (Fig. 7E); DN-SREBP-1 partially suppressed PDX-1's own activation in the PDX-1-luc assay (Fig. 8B). In the Gal4 system, Gal4-PDX-1(1-149) robustly activated the (Gal4) 8 -Luc, but Gal4-PDX-1(1-263) as well as Gal4-PDX-1(1-209) showed low activity (Fig. 7B). Thus, HD inhibits transactivation. DN-SREBP-1 could not bind to HD because of conformational hindrance or nuclear proteins binding to HD. Therefore, it is concluded that the mechanisms for SREBP-1 suppression of the PDX-1 autoregulatory loop could involve both direct physical interaction and squelching of the recruitment of the cofactors.
Although PDX-1 has been shown to interact with the histone acetyltransferase p300 and cAMP-response element-binding protein-binding protein (CBP) in insulin gene expression (45,46), SREBP isoforms SREBP-1a and SREBP-2 strongly interact with CBP and p300, but SREBP-1c shows only weak interaction (47,48). p300 or CBP overexpression did not rescue the inhibition of PDX-1 transactivation by SREBP-1c (data not shown). Frances et al. (49) have reported that PDX-1 physically associates with and recruits the H3-K4 methyltransferase SET9 to the insulin gene. Interaction of PDX-1 with SET9 may be required for the transactivation of the Pdx-1 promoter, and SREBP-1c may squelch the recruitment of SET9 to PDX-1. 35 S-SET9 binds not only to GST-SREBP-1c (nuclear form) and GST-SREBP-1c (bHLH) but also to GST-PDX-1 (FL) and GST-PDX-1 (HD) (supplemental Fig. 5). Cofactors like SET9 may be required for the Pdx-1 promoter activity. PDX-1 and SREBP-1c do not function as direct transcription factors but rather as modifiers of other factors. Furthermore, these data suggest that mutual interaction of the two transcription factors could mediate diverse effects on the transactivation of the functional gene.