SOX6 Attenuates Glucose-stimulated Insulin Secretion by Repressing PDX1 Transcriptional Actvity and Is Down-regulated in Hyperinsulinemic Obese Mice*

In obesity-related insulin resistance, pancreatic islets compensate for insulin resistance by increasing secretory capacity. Here, we report the identification of sex-determining region Y-box 6 (SOX6), a member of the high mobility group box superfamily of transcription factors, as a co-repressor for pancreatic-duodenal homeobox factor-1 (PDX1). SOX6 mRNA levels were profoundly reduced by both a long term high fat feeding protocol in normal mice and in genetically obese ob/ob mice on a normal chow diet. Interestingly, we show that SOX6 is expressed in adult pancreatic insulin-producing β-cells and that overexpression of SOX6 decreased glucose-stimulated insulin secretion, which was accompanied by decreased ATP/ADP ratio, Ca2+ mobilization, proinsulin content, and insulin gene expression. In a complementary fashion, depletion of SOX6 by small interfering RNAs augmented glucose-stimulated insulin secretion in insulinoma mouse MIN6 and rat INS-1E cells. These effects can be explained by our mechanistic studies that show SOX6 acts to suppress PDX1 stimulation of the insulin II promoter through a direct protein/protein interaction. Furthermore, SOX6 retroviral expression decreased acetylation of histones H3 and H4 in chromatin from the promoter for the insulin II gene, suggesting that SOX6 may decrease PDX1 stimulation through changes in chromatin structure at specific promoters. These results suggest that perturbations in transcriptional regulation that are coordinated through SOX6 and PDX1 in β-cells may contribute to the β-cell adaptation in obesity-related insulin resistance.

In obesity-related insulin resistance, pancreatic islets compensate for insulin resistance by increasing secretory capacity. Here, we report the identification of sex-determining region Y-box 6 (SOX6), a member of the high mobility group box superfamily of transcription factors, as a co-repressor for pancreatic-duodenal homeobox factor-1 (PDX1). SOX6 mRNA levels were profoundly reduced by both a long term high fat feeding protocol in normal mice and in genetically obese ob/ob mice on a normal chow diet. Interestingly, we show that SOX6 is expressed in adult pancreatic insulin-producing ␤-cells and that overexpression of SOX6 decreased glucosestimulated insulin secretion, which was accompanied by decreased ATP/ADP ratio, Ca 2؉ mobilization, proinsulin content, and insulin gene expression. In a complementary fashion, depletion of SOX6 by small interfering RNAs augmented glucose-stimulated insulin secretion in insulinoma mouse MIN6 and rat INS-1E cells. These effects can be explained by our mechanistic studies that show SOX6 acts to suppress PDX1 stimulation of the insulin II promoter through a direct protein/protein interaction. Furthermore, SOX6 retroviral expression decreased acetylation of histones H3 and H4 in chromatin from the promoter for the insulin II gene, suggesting that SOX6 may decrease PDX1 stimulation through changes in chromatin structure at specific promoters. These results suggest that perturbations in transcriptional regulation that are coordinated through SOX6 and PDX1 in ␤-cells may contribute to the ␤-cell adaptation in obesity-related insulin resistance.
Insulin resistance is tissue insensitivity to the regulatory effects of insulin and is the leading cause of type 2 diabetes (1,2). Most affected individuals with insulin resistance do not directly develop diabetes but rather adapt to chronic insulin resistance by expanding pancreatic ␤-cell mass and/or insulin secretory capacity. To provide the required amount of insulin to maintain normal glucose levels, ␤-cell mass increases by islet neogenesis, ␤-cell replication, and ␤-cell hypertrophy. Pancreatic ␤-cells eventually fail to compensate for the increased insulin demand created by insulin resistance, leading to type 2 diabetes (1-6).
Pancreatic-duodenal homeobox factor-1 (PDX1), 4 a homeodomain transcription factor, and the insulin/insulin-like growth factor signaling pathway are critical for ␤-cell replication and the compensatory response to insulin resistance (7). PDX1 is expressed in ␤-cells of the islets of Langerhans and is involved in regulating the expression of a number of key ␤-cell genes. It plays a pivotal role in the development of the pancreas and islet cell ontogeny (8). In a mouse model, inactivation of both pdx1 alleles results in pancreas agenesis, whereas heterozygous pdx1 ϩ/Ϫ mice or animals carrying a ␤-cell-specific mutation of the gene exhibit glucose intolerance (9 -11). Mutations in the human PDX1 gene are associated with maturity onset diabetes of the young (MODY4) and predispose to late onset type II diabetes (12)(13)(14). Although these results show that PDX1 plays a key role in the development of the pancreas and glucose-stimulated insulin secretion (GSIS) from ␤-cells, its functional role in the ␤-cell adaptation seen in chronic insulin resistance is poorly understood.
PDX1 is a 284-amino acid protein consisting of 1) an NH 2 -terminal transactivation domain of 144 amino acids, 2) a homeodomain of 60 amino acids, and 3) a COOH-terminal domain of 80 amino acids. PDX1 binds through its homeodomain to target sequences called A-boxes (A/T-rich elements) of the insulin gene promoter (15). The NH 2 -terminal activation domain of PDX1 recruits the coactivator p300 and stimulates insulin gene expression synergistically with E12 and E47, which bind to E-boxes that are also located in the insulin gene promoter (16 -19). Interestingly, p300 is recruited to the insulin gene promoter only when cells are cultured in high glucose media (20).
To identify additional factors that may contribute to the ␤-cell adaptation in insulin resistance, we have been characterizing genes that are selectively regulated in the islets of mice fed a high fat diet (HFD) using microarray analysis. Through the evaluation of transcriptional changes by microarray and quantitative real time PCR analyses, we found that one of the sex-determining region Y-box (SOX) transcription factors, SOX6, is markedly down-regulated in the islets of HFD-fed mice and normal chow fed ob/ob mice. Functional analyses with pancreatic ␤-cell line MIN6 cells revealed that SOX6 reduces GSIS by inhibiting PDX1 transcriptional activity, and our evidence indicates this occurs through a direct interaction between SOX6 and PDX1 proteins. We further show that overexpression of SOX6 results in decreased expression of genes involved in mitochondrial metabolism, including the NADH dehydrogenase complex of the mitochondrial respiratory chain, ATP synthase, and a subunit of cytochrome c oxidase. Taken together, the current data suggest that SOX6 is a key protein in the regulation of GSIS and that, together with PDX1, it contributes to the adaptive compensation of ␤-cells during the progression of obesity-related insulin resistance.
Animals, Diets, and Pancreatic Islet Preparation-10-Week male C57BL/6J mice and ob/ob mice were purchased from Charles River and housed in a temperature-and humidity-controlled (26.5°C and 35%) facility with a 12-h light/dark cycle (09:00 to 21:00 h). Mice were fed with a normal chow diet (NCD) (CE-2; CLEA, Osaka, Japan) or a high fat diet (HFD) (24) ad libitum for 9 weeks and then sacrificed for islet preparations at 11:00. Mouse pancreatic islets were isolated by a standard collagenase digestion method as described previously (22,25,26).
Quantitative Real Time PCR (QRT-PCR) and Affymetrix Oligonucleotide Microarray-The methods for microarray and QRT-PCR have been described (23,24). We used Affymetrix Genechip MOE430-A and -B arrays that contain probe sets for Ͼ30,000 mouse genes. All primer sequences used in this article are available by request.
Immunohistochemistry-For light microscopy of paraffin-embedded sections, mouse pancreatic tissues were fixed with 10% (w/v) formalin at room temperature for 20 h. The samples were dehydrated with an alcohol series and embedded in paraffin. Antigen retrieval was performed by heating the sections in an autoclave at 121°C for 15 min. The sections were incubated with anti-SOX6 antibody (ab-12054) (1:2000 dilution) for 16 h at 4°C. Bound antibody was detected with the Simple Stain MAX-PO (Multi) reagent (Nichirei), an amino acid polymer coated with goat anti-rabbit IgG (FabЈ) and peroxidase using 3,3Ј-diaminobenzidine (Dojindo, Kumamoto, Japan) as a substrate, and a hematoxylin counterstain was applied. For double immunofluorescence of adult mouse pancreas, fixed frozen tissues were permeabilized with 0.2% Triton X-100 for 20 min at 4°C and stained with an anti-SOX6 antibody (ab-12054) (1:2000 dilution) labeled with Zenon Alexa Fluor 594 labeling kit and an anti-insulin or an anti-PDX1 antibody (KR059) (1:2000 dilution). For the detection of PDX1 and insulin, Alexa Fluor 488 antirabbit IgG (1:2000 dilution) and Alexa Fluor 488 anti-guinea pig IgG were used as a secondary antibody, respectively. Control experiments were carried out by omitting the primary antibody. Immunofluorescence was captured with a confocal laser scanning microscope (Fluoview FV500, Olympus, Japan).
Expression Plasmids-Retroviral expression vectors encoding mouse SOX6 and other SOX genes were generated by PCR and insertion of the cDNAs into the pMX, a cytomegalovirus (CMV) promoter-driven retroviral expression vector (provided by Dr. Toshio Kitamura at University of Tokyo) (27). pCMV-PDX1, a pcDNA3-based plasmid encoding mouse PDX1, was obtained from Dr. Kazuya Yamagata at Osaka University (28), and to create pcDNA3-based plasmids encoding mutant PDX1, the deletion sequences of PDX1 (amino acids 1-205, 1-144, and 145-284) were amplified by PCR and ligated into pcDNA3 (Invitrogen). To create pcDNA3-based plasmids encoding full-length and mutant SOX6, the full-length and deletion sequences of SOX6 (amino acids 181-827, 263-827, 617-827, 697-827, and 617-696) were generated by PCR amplification and ligated into pcDNA3. pCMV-⌬HMG-SOX6 encodes an internal deletion mutant form of SOX6 in which a 265amino acid region containing the high mobility group (HMG) domain (amino acids 563-827) was deleted. This was constructed by digestion of pCMV-SOX6 with ApaI to remove the ApaI-ApaI 0.8-kbp fragment containing sequences for the HMG domain, and the plasmid was subsequently religated. A GAL4-PDX1 fusion construct, pBIND-PDX1, and a GAL4-E47 fusion construct, pBIND-E47, were constructed by inserting each cDNA fragment into a polylinker site of pBIND plasmid (Promega), which contains the DNA binding domain of the yeast GAL4 protein.
Reporter Plasmids-pINS(Ϫ872)-luc is the rat insulin II gene promoter-luciferase reporter construct that spans Ϫ872 to Ϫ176 relative to the translation initiation site. pINS(Ϫ552)-luc and pINS(Ϫ413)-luc are 5Ј-deletion mutants of pINS(Ϫ872)-luc, and each contains a deletion with the 5Ј-end denoted in parentheses and the same 3Ј-end point at Ϫ176. pINS(Ϫ413mut)-luc is identical to pINS(Ϫ413)-luc except that the potential SOX binding site (nucleotides Ϫ248 to Ϫ242) is deleted. pINS(Ϫ370mut)-luc was constructed in an identical manner to pINS(Ϫ413mut)-luc, but starting from position Ϫ370. The 5Ј-flanking region of the rat insulin II gene (Ϫ872 to Ϫ176) (29) was amplified by PCR using a forward primer starting from Ϫ872 (5Ј-TATAGGTAC-CCCCAACCACTCCAA-3Ј) and a reverse primer OLi(Ϫ176R) (5Ј-TATACCCGGGGGTTACTGAATCC-3Ј) and cloned into pGL3-basic. pINS(Ϫ552)-luc and pINS(Ϫ413)-luc were constructed in a similar manner to pINS(Ϫ872)-luc using the respective forward primers starting from the positions Ϫ552 (5Ј-TATAGGTACCTGTGAAACAA-CAGTTCAAGGG-3Ј) and Ϫ413 (5Ј-TATAGGTACCTTCATCAG-GCCACCCAGGAG-3Ј) and coupled with a common reverse primer OLi(Ϫ176R). p(E5 ϩ E2 ϩ E3) 4 -luc is a luciferase reporter plasmid driven by a promoter consisting of four tandem copies of the E47-responsive element (5Ј-ACACCTGCAGCAGCTGGCAGGAAGCAGGTCATGT-GGCA-3Ј) from the mouse IgH promoter (30). It was constructed by annealing the oligonucleotides for the top and bottom strands and subsequent ligation into the MluI and BglII sites of pGL3-basic. All plasmid constructs were verified by restriction endonuclease mapping and DNA sequencing. pG5luc is a luciferase reporter construct driven by a promoter consisting of five copies of GAL4 binding sites plus the adenovirus E1B TATA box (Promega).
Cell Culture and Retroviral Infection-MIN6 cells (a line of mouse pancreatic ␤-cells) (31) and INS-1E cells (a clone of parental rat ␤-cell line INS-1E cells (32) selected for insulin content and adequate proliferation (33)) were kind gifts from Dr. Jun-Ichi Miyazaki (Osaka University) and Dr. Pierre Maechler (University Medical Center at Switzerland), respectively. MIN6 cells were grown in Dulbecco's modified Eagle's medium containing 25 mM glucose, 5.5 M ␤-mercaptoethanol, 100 units/ml penicillin, and 100 g/ml of streptomycin sulfate, supplemented with 15% fetal bovine serum at 37°C in 5% CO 2 . INS-1E cells were cultured in RPMI1640 containing 11.6 mM glucose, 10 mM HEPES, pH 7.4, 1 mM sodium pyruvate, 50 M ␤-mercaptoethanol, 2 mM glutamine, 100 units/ml penicillin, and 100 g/ml streptomycin sulfate, supplemented with 5% fetal bovine serum at 37°C in 5% CO 2 . Retroviral infection to MIN6 cells was performed as previously described (23) using pMX plasmids (27,34). Human embryonic kidney 293 cells and BHK21 cells (a line of hamster kidney cells) were obtained from the Cell Resource Center for Biomedical Research at Tohoku University (Sendai, Japan) and maintained in Dulbecco's modified Eagle's medium containing 100 units/ml penicillin and 100 g/ml of streptomycin sulfate, supplemented with 10% fetal bovine serum at 37°C in 5% CO 2 .
Transient Transfection Assays-MIN6, HEK293, or BHK21 cells were plated on day 0 at a density of 5 ϫ 10 4 cells/24-well plates. On day 1, cells were transfected with luciferase reporter plasmid, expression plasmids, and pCMV␤ (Stratagene), a ␤-galactosidase reference gene, using Lipofectamine PLUS reagent (Invitrogen) as previously described (23,(35)(36)(37). The total amount of DNA in each transfection was adjusted to 0.2-0.7 g/well. On day 2, the cells were harvested and assayed for firefly luciferase activity and normalized to ␤-galactosidase activity using kits from Promega and BD Biosciences, respectively.
Insulin Secretion, Content, and Adenine Nucleotide Determinations and an Intracellular Ca 2ϩ Assay-The secretory responses to glucose and other secretagogues were tested in MIN6 cells and INS-1E cells between passages 16 -35 and 54 -95, respectively (31,39). Before the experiments, MIN6 or INS-1E cells were washed twice with phosphatebuffered saline and preincubated for 30 min at 37°C in glucose-free Krebs-Ringer bicarbonate HEPES buffer (KRBH) of the following composition: 129 mM NaCl, 4.7 mM KCl, 5.0 mM NaHCO 3 , 1.2 mM KH 2 PO 4, 1.2 mM MgSO 4 , 2.0 mM CaCl 2 , and 10 mM HEPES, pH 7.4. Bovine serum albumin (0.1%) was added as an insulin carrier. Next, cells were washed once with glucose-free KRBH and then incubated for 1 h in KRBH and stimuli as indicated. Incubation was stopped by putting the plates on ice, and the supernatants were collected for insulin secretion. Cellular insulin was extracted with acid-ethanol (0.4 M HCl in 74% ethanol) overnight at 4°C as described previously (22). Insulin secretion and content were determined by a rat insulin enzyme-linked immunosorbent assay kit (Shibayagi Co., Shibukawa, Japan). ATP and ADP content in MIN6 cells were determined using the ATP assay system (Toyobo-Net, Tokyo, Japan) as previously described (22,40). To determine the intracellular Ca 2ϩ levels, MIN6 cells were loaded with 2 M fura-2/AM (Dojindo) in KRBH containing 10 mM glucose at room temperature for 1 h as described (26). The loading solution was removed and then applied to a Functional Drug Screening System 6000 (Hamamatsu Photonics, Shizuoka, Japan). Intracellular Ca 2ϩ concentration was measured by the ratio of emission fluorescence of 510 nm by excitation at 340 and 380 nm.
Chromatin Immunoprecipitation (ChIP) Assay-A commercially available assay kit (Upstate Biotechnologies, Charlottesville, VA) was used for ChIP studies according to the manufacturer's protocol. Approximately 2 ϫ 10 6 MIN6 cells were cross-linked for 15 min at 37°C with formaldehyde (1% final concentration) in Dulbecco's modified Eagle's medium, subsequently washed twice with phosphate-buffered saline containing proteinase inhibitors. Cells were scraped, centrifuged, and resuspended in 0.5 ml of lysis buffer (50 mM Tris-HCl at pH 8.1, 10 mM EDTA, 1% SDS). The cells were then sonicated on ice 10 times using 30-s pulses using a Sonifier cell disrupter model Micro-150 (GENEQ Inc., Montreal, Canada), and then the debris was removed by centrifugation. Supernatants were collected and used for immunoprecipitation. To reduce the nonspecific binding, samples were preincubated with protein A-Sepharose and sonicated salmon sperm DNA for 1 h at 4°C. After centrifugation, the supernatant was incubated with specific antibodies (anti-PDX1 (sc-14664), anti-SOX6 (sc-17332), anti-acetyl-histone H3 and H4), or control rabbit IgG overnight at 4°C followed by incubation with Protein A-Sepharose for 1 h. After centrifugation and washing, the immunocomplexes were eluted twice with 250 l of elution buffer (1% SDS and 0.1 M NaHCO 3 ) for 15 min at room temperature. The cross-linking was then reversed by adding 20 l of 5 M NaCl and 1 l of 10 mg/ml RNase A and by incubating at 65°C for 6 h. After treating with 1.5 l of proteinase K (10 g/l), the DNAs were extracted with phenol/chloroform, subsequently ethanol-precipitated using 20 g of glycogen carrier, and dissolved in 50 l of distilled water. The amount of DNA recovered from immunoprecipitates with specific antibodies or control IgG was quantified by QRT-PCR, which was performed in triplicate. Data were represented as -fold change over DNA input. The primers used to amplify the insulin II promoter sequences were 5Ј-GGAACTGTGAAA-CAGTCCAAGG-3Ј and 5Ј-CCCCTGGACTTTGCTGTTTG-3Ј.
GST Pull-down Assay-Glutathione S-transferase (GST) fusion constructs containing full-length PDX1 (amino acids 1-284) and fulllength SOX6 (amino acids 1-827) were created in the bacterial expression vector pGEX-4T-2 (Amersham Biosciences), expressed in BL21 bacteria, and purified as previously described (21,23). pcDNA3 constructs, a T N T quick coupled transcription/translation system (Promega), and L-[ 35 S]methionine (1000 Ci/mmol; Amersham Biosciences) were used for synthesizing 35 S-labeled in vitro proteins. Purified GST or GST fusion proteins were incubated with 35 S-labeled proteins for 2 h at 4°C and washed five times before SDS-PAGE was carried out.
Statistical Analysis-For unpaired comparisons of two groups, Student's t test was used. For multiple comparisons, one-way analysis of variance, followed by Tukey's honestly significant difference test was used.

Identification of SOX6 as a Down-regulated Transcription Factor in
Hyperinsulinemic Animals-In an attempt to uncover the mechanism underlying hyperinsulinemia and the ␤-cell adaptation phenotype in obesity-induced insulin resistance, we examined the gene expression profile in pancreatic islets of hyperinsulinemic obese mice that were fed an HFD. Using the Affymetrix mouse gene chip MOE430 and QRT-PCR of 59 transcription factors from the data base of the Beta Cell Biology Consortium (on the World Wide Web at www.cbil.upenn.edu/ EPConDB/index.shtml), we identified 17 transcription factors whose expression levels were altered either more than 2 or less than 0.5 in response to HFD (supplemental Table 1) and subsequently examined their properties in insulinoma ␤-cell line MIN6 cells (supplemental Fig.  1). Through these processes, we identified the unique SOX transcription factor, SOX6, as a specifically down-regulated gene in these HFD-fed mice, which also has the ability to modulate GSIS (supplemental Fig. 1). As shown in Fig. 1A, the islet expression of SOX6 mRNA was severalfold lower in both the HFD-induced obese mice and genetically obese ob/ob mice than in normal mice. This result suggested that the levels of SOX6 mRNA might be negatively regulated by insulin or glucose. To test this hypothesis, we carried out fasting/refeeding experiments where insulin levels change to regulate blood glucose. As shown in Fig. 1B, the pancreatic levels of SOX6 RNA were essentially unchanged either by fasting or subsequent refeeding after 12 h of fasting, whereas expression of stearoyl-CoA desaturase 2, a well characterized insulin-regulated gene, was down-and up-regulated by fasting and refeeding, respectively. These results indicate that the expression of SOX6 may not be simply regulated by acute changes in insulin or blood glucose levels but instead by the prolonged hyperglycemia, hyperinsulinemia, and/or insulin resistance generated in the course of developing obesity.
Next, we examined whether SOX6 protein was expressed in the pancreatic islet. Nuclear extracts were prepared from mouse pancreatic islets and insulinoma MIN6 cells and immunoblotted with an antibody raised against SOX6. A strong signal was detected at the predicted molecular weight in the nuclear extract from mouse islets and MIN6 cells (Fig. 1C). This signal was diminished by the application of siRNA for SOX6 (see Fig. 2D), indicating that this antibody specifically recog-FIGURE 1. SOX6 gene expression in pancreatic islets from diet-induced obese and genetically obese ob/ob mice. A, relative amounts of SOX6 mRNA in pancreatic islets from C57BL/6J male mice on either an HFD or NCD or those from male ob/ob mice on NCD. Male C57BL/6J mice (8 weeks of age) and ob/ob mice were fed with either HFD or NCD for 9 weeks as described under "Experimental Procedures." B, relative amounts of mRNAs for SOX6 and stearoyl-CoA desaturase 2 in islets from mice subjected to fasting and refeeding. C57BL/6J male mice, 8 weeks of age, were divided into three groups: fed, fasted, and refed. The fed group was fed ad libitum, the fasted group was fasted for 24 h, and the refed group was fasted for 24 h and then refed a normal chow diet for 1 h prior to study. A and B, total RNA from pancreatic islets were prepared and subjected to QRT-PCR as described. Mouse 36B4 (NM_007475) mRNA was used for the invariant control for QRT-PCR. Values represent the amount of mRNA relative to those in NCD mice (A) or in the fed condition experiment (B), which is arbitrarily defined as 1. Each bar and symbol represent mean Ϯ S.E. of three independent experiments performed in triplicate. *, p Ͻ 0.01 compared with control. SCD2, stearoyl-CoA desaturase 2. C, immunoblot analysis of nuclear extracts prepared from mouse pancreatic islets and MIN6 cells. Nuclear extracts were prepared as previously described (36,41), and aliquots (20 g) were subjected to SDS-PAGE and immunoblot analysis with anti-SOX6 antibody (top). The filter was exposed for 30 s. The position of SOX6 is indicated by an arrow. D, immunohistochemistry of paraffin-embedded section from control C57BL/6J (left) and ob/ob (right) mice using anti-SOX6 antibody. E, double immunofluorescence of fixed frozen sections of adult mouse pancreas for SOX6 and insulin. F, double immunofluorescence of fixed frozen sections of adult mouse pancreas for SOX6 and PDX1.

SOX6 Acts as a Co-repressor for PDX1
nized SOX6 protein. Immunostaining of the pancreas with the SOX6 antibody showed that SOX6 is localized in both the nucleus and cytosol of islets of Langerhans in a similar pattern as reported for SOX13 (42). Consistent with the decrease in SOX6 mRNA level shown in ob/ob mice in Fig. 1A, the number of cells expressing SOX6 protein was also reduced in ob/ob mice (Fig. 1D). To examine whether SOX6 is expressed in ␤-cells, double immunofluorescence for SOX6 and insulin was carried out, and this analysis showed that SOX6 was expressed in the majority of ␤-cells (Fig. 1E). Double staining with glucagon showed that SOX6 was also expressed in ␣-cells (data not shown). Since PDX1 is also expressed in the majority of ␤ cells, we examined whether SOX6 co-localized with PDX1 in ␤-cells by double immunofluorescence staining. Endogenous SOX6 and PDX1 were diffusely distributed in the nuclei of pancreatic ␤-cells and co-localized (Fig. 1F).
SOX6 Negatively Regulates Glucose-stimulated Insulin Secretion-We next examined the effects of SOX6 expression on GSIS. MIN6 cells were transduced with a recombinant retrovirus encoding SOX6, and insulin secretion in response to 5.6 or 16.7 mM glucose was examined as described under "Experimental Procedures." As a control experiment, a recombinant retrovirus encoding alkaline phosphatase (ALP) was transduced into companion dishes of MIN6 cells. As shown in Fig. 2A, retrovirus-mediated induction of SOX6 resulted in ϳ60% inhibition of GSIS evoked by 16.7 mM glucose as compared with the control ALPtransduced cells. Control experiments for these studies showed that both SOX6 and ALP RNAs were expressed in the transduced cells (data not shown).
The SOX gene family encodes a group of transcription factors defined by the conserved HMG DNA-binding domain and consists of more than 20 individual members (43). SOX proteins are classified into eight subgroups, A-H. SOX6 and its highly related proteins, SOX5 and SOX13, belong to group D, which is unique, because all three contain an additional specifically positioned leucine zipper motif. To evaluate the effects of these and other SOX proteins on GSIS, we selected SOX5, -9, -13, and -15 plus several other randomly selected SOX proteins and expressed each through retrovirus transduction in MIN6 cells. The mRNA expression of each SOX gene was confirmed by either QRT-PCR or immunoblotting or both. As shown in Fig. 2B, similar to SOX6, expression of the highly related SOX5 and SOX13 also attenuated GSIS, whereas expression of the other more distally related SOX proteins (SOX9 and SOX15) had minimal effects. The expression of SOX5, SOX9, and SOX15 was confirmed by immunoblotting (Fig. 2B, inset). In experiments not shown, we also determined that similar expression of six other more distantly related SOX proteins had minimal effects on GSIS. These studies demonstrate that SOX6 and its closest relatives all can potentially inhibit GSIS. However, unlike SOX6, the levels of the SOX5 and SOX13 transcripts are not down-regulated by the HFD or in ob/ob obese mice (data not shown), suggesting that SOX6 is the major SOX protein involved in modulating GSIS in obesity-induced pancreatic islets.
To complement the results obtained by retroviral expression, we next employed an RNA interference approach to knock down the expression of the endogenous SOX6 in mouse insulinoma MIN6 cells and rat insu-  NOVEMBER 11, 2005 • VOLUME 280 • NUMBER 45

JOURNAL OF BIOLOGICAL CHEMISTRY 37673
linoma INS-1E cells. Transfection of an siRNA duplex, corresponding to nucleotides 2249 -2267 of the mouse and rat SOX6 mRNAs, resulted in a reduction of SOX6 mRNA levels by ϳ50% in both MIN6 and INS-1E cells (Fig. 2C), whereas it had almost no effect on the levels of SOX5 or SOX13 (data not shown). Consistent with the decrease in mRNA level, immunodetectable SOX6 was reduced in both MIN6 and INS-1E cells treated with SOX6-specific siRNA (Fig. 2D). Under these conditions, insulin secretion in response to 16.7 mM glucose was significantly enhanced by the SOX6 siRNA (1.4-fold, p Ͻ 0.05) in both cell lines as compared with companion dishes where a control siRNA was transfected (Fig. 2E). In contrast, the increase of insulin secretion by SOX6 siRNA was not observed in 5.6 mM glucose. Together with the retroviral expression findings, these data indicate that SOX6 negatively regulates GSIS in insulin-secreting cells.
SOX6 Expression Reduces ATP Production and Insulin Gene Transcripts-To further evaluate the mechanism for the reduced GSIS mediated by SOX6, we analyzed the insulin secretory responses to ␣-ketoisocaproate (␣-KIC), tolbutamide, and KCl in MIN6 cells transduced with the SOX6-expressing retrovirus. These secretagogues were selected for their actions at different and specific stages in the insulin secretion pathway; after glucose, which is the primary event, ␣-KIC acts at mitochondrial metabolism, and KCl and tolbutamide act at membrane depolarization (44). Whereas SOX6 expression had almost no effect on the depolarization-induced insulin secretion by 0.3 mM tolbutamide or 20 mM KCl, it profoundly reduced insulin secretion by 10 mM ␣-KIC at glucose concentrations of 5.6 and 16.7 mM (Fig. 3A). This result suggests that mitochondrial ATP production from ␣-KIC is inhibited by the expression of SOX6. Consistent with the reduced insulin secretion from ␣-KIC by SOX6, the ATP content and ATP/ADP ratio in the presence of 16.7 mM glucose were significantly decreased (by 25-30%) in SOX6-expressing cells as compared with control ALPtransduced cells (Fig. 3B). Furthermore, glucose-induced intracellular Ca 2ϩ ([Ca 2ϩ ] i ) transients (as determined by changes in the 340/380-nm fluorescence ratio), the eventual trigger for the exocytosis of insulincontaining vesicle (Fig. 3C), and total cellular insulin content (80% of ALP control cells) were significantly decreased in SOX6-expressing cells (Fig. 3D).
SOX6 Negatively Regulates Transcripts of Genes for Insulin as Well as ATP Production in Mitochondria-To further examine the molecular mechanism for the reduction in ATP generation and cellular insulin content, we examined gene expression in MIN6 cells transduced with SOX6. Consistent with reduced levels of ATP production, genes for the NADH dehydrogenase complex subunit (complex I), the cytochrome bc1 complex subunit (complex III), cytochrome c oxidase complex subunit (complex IV), and the ATP synthase subunit (complex V) were reduced by 40 -60% compared with control ALP-expressing MIN6 cells (TABLE ONE). Importantly, consistent with the reduction of total cellular insulin content, the insulin I and II gene transcript levels in SOX6 cells were markedly decreased, as quantified by QRT-PCR (to 60% of ALP control cells) (Fig. 3E).
SOX6 Acts as a Co-repressor for PDX1 on the Insulin Promoter-Based on the reduced insulin I and II gene mRNA in SOX6-overexpressing cells, we examined the effects of SOX6 expression on insulin gene promoter activity using the promoter region of the insulin II gene. A DNA fragment extending from nucleotide Ϫ872 to Ϫ176 of the rat insulin II gene was subcloned into the promoterless luciferase reporter gene, pGL3-basic, to create the promoter reporter construct pINS(Ϫ872)-luc. This promoter reporter construct was transiently transfected into MIN6 cells along with increasing amounts of pCMV-SOX6. As shown in Fig. 4A, the normalized luciferase activities in MIN6 cells were decreased in proportion to the amounts of co-transfected pCMV-SOX6. There are three potential SOX elements in the rat insulin II gene (Fig.  4B). To analyze whether SOX6-mediated suppression of the insulin promoter is mediated by the direct binding of these SOX-like elements by SOX6, we constructed a series of deletion mutants, each lacking one of the three SOX-like elements (Fig. 4B). Co-transfection of these mutants with SOX6 revealed that deletion of any of these SOX elements did not affect the suppressive effects of SOX6. These results suggest that the suppression of the insulin promoter by SOX6 is either dependent on SOX6 binding to nonconsensus sites or independent of SOX6 binding to DNA. Consistent with previously reported observations (45,46), deletion of nucleotides Ϫ413 to Ϫ370 of the insulin II promoter abolished all of the promoter activity (Fig. 4B).
We next examined the effects of SOX6 on the insulin II promoter activation by either PDX1, E47 (an E2A gene product), or both. As shown in Fig. 4C and consistent with a previous report (17), PDX1 together with E47 synergistically activated the insulin II promoter, and increasing amounts of the SOX6 expression plasmid resulted in a dosedependent inhibition. Fig. 4D shows that SOX6 had almost no effect on the E47-mediated transactivation of a highly E47-responsive IgH enhancer reporter, p(E5 ϩ E2 ϩ E3) 4 -luc, suggesting that SOX6 does not specifically repress E47 function.
To evaluate whether SOX6 might inhibit activation mediated by PDX1, we fused the coding sequence of PDX1 and E47 to the GAL4 DNA-binding domain, and the transactivation potential of the resulting GAL4 fusions was examined in the absence and presence of SOX6. As shown in Fig. 4E, SOX6 strongly suppressed GAL4-PDX1 transactivation, whereas it had almost no effects on GAL4-E47.
These results suggest that SOX6 and PDX1 proteins may directly interact together to inhibit the insulin II promoter. To evaluate this possibility, we carried out GST pull-down assays using a GST-SOX6 fusion protein and in vitro-translated full-length PDX1, and to map the interaction domain we also evaluated a series of deletion mutants. As shown in Fig. 5A, i and ii, the full-length protein and the mutant lacking the COOH terminus (residues 206 -284) were coprecipitated with GST-SOX6, whereas the mutant lacking the homeodomain and the COOH terminus had weak binding to GST-SOX6 (Fig. 5A, iii), and deletion of the NH 2 -terminal 144 amino acids abolished the binding to GST-SOX6 (Fig. 5A, iv). These data indicate  NOVEMBER 11, 2005 • VOLUME 280 • NUMBER 45 that the NH 2 -terminal 144 amino acids of PDX1 were critical for the interaction with SOX6. As a complement to the GST-SOX6 pull-down experiment, we constructed a GST-PDX1 fusion protein and evaluated it for interaction with full-length SOX6 or a series of deletion mutants of SOX6. As shown in Fig. 5B, it is clear that the minimum PDX1 binding site of SOX6 is a region containing the functionally important HMG box of SOX6. Together with the GST-SOX6 pull-down assay findings, these data indicate that the HMG box of SOX6 interacts physically with the NH 2 -terminal 144 amino acids of PDX1, as schematically depicted in Fig. 5C.

SOX6 Acts as a Co-repressor for PDX1
In Fig. 5D, top, we show that elimination of the HMG domain from SOX6 resulted in a protein that was unable to attenuate GSIS. The protein expression of mutant SOX6 lacking the HMG domain (⌬HMG-SOX6) was confirmed by immunoblotting with an anti-SOX6 antibody (Fig. 5D, bottom). Furthermore, the deletion of the HMG domain of SOX6 abolished the SOX6 suppressive effect on the insulin II gene promoter (Fig. 5E), whereas expression of the NH 2 terminus (residues 1-144) of PDX1 reversed the SOX6-mediated inhibition (Fig. 5F). Taken together with the GST interaction studies, these data demonstrate that the SOX6 HMG domain suppresses the insulin II gene promoter by interacting physically with the NH 2 terminus of PDX1, the key transactivator protein for positive regulation of insulin gene transcription.
siRNA-mediated Knockdown of PDX1 Decreased SOX6 Occupancy of the Insulin II Promoter-To further determine the mechanism underlying insulin II promoter inhibition by SOX6, we examined the association of PDX1 and SOX6 with the insulin II promoter in SOX6 or control ALP retrovirus and PDX1-specific siRNA-treated MIN6 cells by ChIP assays. We used QRT-PCR to evaluate the results of the ChIP, and the specific primers were designed as schematically depicted in Fig. 6.
When SOX6 was expressed by retroviral transduction, the insulin promoter sequence was present at a higher level in the SOX6 immunoprecipitate (Fig. 6A, top, compare lanes 2 and 4), suggesting an increased association of SOX6 with the promoter under these conditions. In contrast, SOX6 expression did not alter the DNA binding of PDX1 to the insulin II promoter (Fig. 6A, top, compare lanes 6 and 8), indicating that SOX6 does not simply interfere with the DNA binding of PDX1. The immunoblot in the bottom panel shows that SOX6 protein levels were elevated (6.5-fold, as evaluated by LumiVisionPRO) in the SOX6 retroviral transduced cells (Pre) and that SOX6 was efficiently collected by the immunoprecipitation procedure with anti-SOX6 antibody (Post) in proportion to the overall levels (Fig. 6A, bottom; compare the gels labeled Pre for direct immunoblot with those labeled Post).
It has been previously demonstrated that PDX1 specifically binds the proximal insulin promoter of MIN6 cells by a ChIP assay (47). 48 h following PDX1-siRNA treatment, we observed an 80% reduction in PDX1 occupancy at the insulin II promoter in MIN6 cells (Fig. 6B, top,  compare lanes 2 and 4). Interestingly, a decrease in PDX1 at the insulin II promoter in MIN6 cells was also accompanied by a 60% reduction in the binding of SOX6 to the proximal insulin promoter (Fig. 6B, top,  compare lanes 6 and 8). Taken together with the GST pull-down assays,  2 and 4 -6). E, MIN6 cells were transfected with 0.1 g of pG5luc, the multimerized GAL4-responsive promoter luciferase reporter construct, and 0.02 g of pCMV␤ together with 0.02 g of GAL4 fusion constructs: pBIND-PDX1 (lanes 3-5) and pBIND-E47 (lanes 6 and 7) together with the indicated amounts of pCMV-SOX6 (lanes 2 and 4 -7). A-E, luciferase activity was measured and normalized to ␤-galactosidase activity as described. Each bar and symbol represent mean Ϯ S.E. of triplicate experiments. *, p Ͻ 0.01 compared with control. RLU, relative luciferase units.

SOX6 Acts as a Co-repressor for PDX1
these results indicate that endogenous SOX6 protein in eukaryotic cells interacts with DNA-bound PDX1 at the insulin promoter in vivo.
SOX6 Decreased Acetylation of Chromatin-associated Histones H3 and H4-PDX1 transcriptional activity is reportedly regulated by the recruitment of p300, a coactivator protein that possesses intrinsic histone acetylase activity (20,48), and similar to SOX6, p300 interacts directly with the NH 2 -terminal region of PDX1. We therefore evaluated the effects of SOX6 on the acetylation levels of histones H3 and H4 in chromatin at the insulin II promoter, using antibodies directed at the acetylated forms of either histone H3 or H4. Consistent with previous papers (20,48), cultivation of MIN6 cells with a high glucose concen-tration (30 mM) leads to a significant increase in acetylation of histone H4 at the insulin II promoter (Fig. 6D, compare lanes 2 and 4), whereas there was no effect of glucose on the degree of H3 acetylation (Fig. 6C,  compare lanes 2 and 4). In cells where SOX6 was expressed by retroviral transduction, the levels of both acetylated histones H3 and H4 at the insulin II promoter were profoundly reduced (Fig. 6, C, compare lanes 2  and 6, lanes 4 and 8, and D, compare lanes 4 and 8), indicating that SOX6 leads to a reduction in acetylation of both H3 and H4 in the chromatin on the insulin II promoter. Because p300 has histone acetylase activity, this suggests that SOX6 may interfere with the recruitment of p300 to PDX1 bound at the insulin II promoter. , respectively, at room temperature for 1 h. Purified GST was used as a negative control for nonspecific binding. After washing extensively, the proteins bound to beads and 5% of the input proteins were resolved on SDS-PAGE and visualized by a FUJIX BAS2000 imaging system (Fuji Film, Tokyo, Japan). C, schematic diagram showing HMG domain within SOX6 interacts with NH 2 -terminal transactivation domain of PDX1. D, the effect of insulin secretion by retrovirally transduced mutant SOX6 lacking the HMG domain. Three days after infection with retroviruses for ALP, SOX6, or mutant SOX6 lacking HMG domain (⌬HMG) to MIN6 cells, insulin secretion for 1 h in response to 5.6 and 16.7 mM glucose was measured (top). At the end of the GSIS experiment, medium was aspirated, these same cells used for 16.7 mM glucose concentration were lysed in SDS lysis buffer, and these cells were subjected to SDS-PAGE followed by immunoblotting with anti-SOX6 antibody (ab-12054), which also reacts with ⌬HMG-SOX6 (bottom). The migration of wild type and ⌬HMG-SOX6 are noted. E and F, luciferase reporter gene assay. MIN6 cells were transfected with 0.1 g of pINS(Ϫ872)-luc and 0.01 g of pCMV␤ together with 0.1 g of either pCMV-SOX6 or pCMV-⌬HMG-SOX6 (E). MIN6 cells were transfected with 0.1 g of pINS(Ϫ872)-luc and 0.01 g of pCMV␤ together with the indicated amounts of pCMV-SOX6 (lanes 3-6) and pCMV-PDX1 (1-144), a construct encoding an NH 2 -terminal domain (amino acids 1-144) of PDX1 (lanes 2 and 4 -6). F, firefly luciferase activity was measured and normalized to ␤-galactosidase activity. Each bar and symbol represents mean Ϯ S.E. of triplicate experiments. *, p Ͻ 0.01 compared with control. n.s., not significant.

DISCUSSION
In the current study, we demonstrated that SOX6 protein is expressed in adult pancreatic islets, where its expression is dramatically down-regulated both by HFD feeding and in genetically predisposed obese model (ob/ob) mice. In both cases, the mice are hyperinsulinemic with insulin resistance, and they have an increase in pancreatic islet mass. In contrast, SOX6 mRNA levels were not altered by fasting and refeeding experiments, suggesting that pancreatic expression of SOX6 is not regulated by acute changes of blood glucose or insulin levels. Based on the attenuation of GSIS by SOX6 and its down-regulation in hyperinsulinemic mice, we propose that changes in the ratio of SOX6 to PDX1 may be responsible for the compensatory islet hyperplasia that occurs in response to insulin resistance.
The SOX family of transcription factors contains a DNA binding HMG box, which is highly conserved across species. SOX proteins are involved in a number of developmental processes, including tissue spec-  Ab (ab-12054). B, recovery of the insulin promoter fragment following ChIP using SOX6 and PDX1 Ab and extracts from siRNA-transfected cells. The bottom panel shows the presence of PDX1 in cell lysate before immunoprecipitation (Pre) and the immunoprecipitated samples with PDX1-Ab (Post) detected by immunoblot analysis with PDX1-Ab. C and D, recovery of the insulin promoter fragment following ChIP using anti-acetyl histone H3 and H4 antibodies and extracts from retrovirus-infected MIN6 cells. Retrovirus-infected MIN6 cells were grown on media containing either 3 or 30 mM glucose for 16 h. ChIP assays were performed using anti-acetyl histones H3 and H4 antibodies for immunoprecipitation. Rabbit IgG was used as a negative control for immunoprecipitation. All data represent recovery, in percent, of each DNA fragment relative to total input DNA. Each bar and symbol represents mean Ϯ S.E. of triplicate experiments. *, p Ͻ 0.01 compared with control. n.s., not significant. The data represent the average of at least three independent experiments. ification and differentiation, embryonic patterning, and maturation. Numerous members of the SOX gene family are also expressed during pancreas development, and they may have important and possibly redundant functions in pancreas development (49). Although SOX6 expression was not noticed in this previous study (49), our studies provide strong evidence that it is indeed expressed in islets and regulated by HFD. SOX6 and highly related SOX5 are co-expressed during mouse chondrogenesis and cooperatively activate the expression of the Col2a gene, a chondrocyte differentiation marker (50). Since SOX factors play important roles in key developmental processes, knockout mice tend to die in utero. Mice lacking both SOX5 and SOX6 have a severe generalized chondrodysplasia and die in utero (51). Therefore, the roles for the SOX proteins in adult tissues including pancreatic islets are poorly understood.
Our data suggest that SOX6 modulates GSIS by regulating PDX1 transcriptional activities in pancreatic islets. Our studies also indicate that a physical interaction between the HMG domain of SOX6 and the NH 2 -terminal activation domain of PDX1 is important for this repression. This same NH 2 -terminal activation domain of PDX1 is known to recruit the histone acetylase coactivator p300, and the HMG domain has been shown to be a common protein-protein interaction motif in other SOX proteins as well (52). Consistent with a model where SOX6 interacts with DNA bound PDX1, our studies show that SOX6 expression does not affect PDX1 DNA binding, and reduced PDX1 binding to the insulin promoter by PDX1-specific siRNA results in a decrease in the co-association of SOX6. Furthermore, SOX6 recruitment results in reduced histone acetylation at the insulin II promoter. Based on these observations, we hypothesize that the suppressive effect of SOX6 on PDX1 is mediated, in part, by decreasing the localized histone hyperacetylation that normally would occur upon recruitment of p300 to DNA-bound PDX1. It is also possible that SOX6 interacts with PDX1 to recruit a repressor complex to the promoter of PDX1 target genes.
Recently, Iype et al. (53) showed that decreased insulin transcription was associated with decreased occupancy of the insulin promoter by PDX1 and p300. These investigators also showed that, whereas recruitment of RNA polymerase II to the insulin coding region was significantly reduced, there was no corresponding change in the recruitment of RNA polymerase II to the proximal promoter. They suggested that PDX1 directly regulates insulin transcription through formation of a complex with transcriptional factors (E47, BETA2/NeuroD) and cofactors (p300) to lead to a physical and functional interaction with RNA polymerase II (53). Since SOX6 inhibits synergistic activation by PDX1 and E47, our data suggest that SOX may disrupt this PDX1-mediated increase in recruitment of the polymerase II transcriptional machinery to the insulin gene.
The decrease in insulin secretion by elevated SOX6 is likely to be mediated in part by the decreased insulin gene expression. We also showed that SOX6 overexpression resulted in a decrease in the ratio of ATP/ADP that was accompanied by a significant decrease in expression for several genes of oxidative metabolism (TABLE ONE). Because a decreased ATP/ADP ratio is also associated with decreased GSIS, our results are additionally consistent with a model where SOX6 represses GSIS through decreasing expression of genes involved in ATP production. Whether these genes are direct targets of PDX1 or whether SOX6 may decrease gene expression through other mechanisms requires further study. Interestingly, however, expression of a dominant negative form of PDX1 through adenovirus transduction in vitro reduced the expression of mitochondrial genes and also resulted in severe consequences in ␤-cell mitochondrial function (54).
Although SOX genes of the same group tend to be co-expressed at the same developmental stage and exhibit functional redundancy in development, it is intuitive that redundancy may not extend to key transcriptional regulatory decisions in adult tissues. Group D SOX proteins (SOX5, -6, and -13) are all detected in adult islets (49) (present work). SOX13 has been reported to be expressed in ␤-cells and functions as autoantigen in type 1 diabetes (42). Our data show that they are all capable of suppressing GSIS when overexpressed; however, only SOX6 gene expression was reduced by the HFD or in ob/ob mice. Since all three are expressed in islets, it is not immediately clear why downregulation of SOX6 alone by the HFD could result in an increase in GSIS. It is possible that the levels of endogenous SOX5 and SOX13 proteins are relatively low, so a decrease in SOX6 alone would be sufficient to initiate a response, or additional proteins that are also aberrantly expressed by the HFD augment the SOX6 effect specifically. Experiments are in progress to address these and related issues. It is conceivable that SOX genes in adult tissues are regulated in response to the circulating nutrients or hormones, such as glucose, insulin, or other nutrients. Thus, SOX6 may contribute to the insulin gene regulation in pathophysiological states where PDX1 function is compromised as observed in insulin resistance and diabetes. Therefore, the modulation of PDX1 function by SOX6 may provide a promising new therapeutic target for treatment of type II diabetes.