ErbB3-binding protein 1 (EBP1) represses HNF4α-mediated transcription and insulin secretion in pancreatic β-cells

HNF4α (hepatocyte nuclear factor 4α) is one of the master regulators of pancreatic β-cell development and function, and mutations in the HNF4α gene are well-known monogenic causes of diabetes. As a member of the nuclear receptor family, HNF4α exerts its gene regulatory function through various molecular interactions; however, there is a paucity of knowledge of the different functional complexes in which HNF4α participates. Here, to find HNF4α-binding proteins in pancreatic β-cells, we used yeast two-hybrid screening, a mammalian two-hybrid assay, and glutathione S-transferase pulldown approaches, which identified EBP1 (ErbB3-binding protein 1) as a factor that binds HNF4α in a LXXLL motif–mediated manner. In the β-cells, EBP1 suppressed the expression of HNF4α target genes that are implicated in insulin secretion, which is impaired in HNF4α mutation-driven diabetes. The crystal structure of the HNF4α ligand-binding domain in complex with a peptide harboring the EBP1 LXXLL motif at 3.15Å resolution hinted at the molecular basis of the repression. The details of the structure suggested that EBP1's LXXLL motif competes with HNF4α coactivators for the same binding pocket and thereby prevents recruitment of additional transcriptional coactivators. These findings provide further evidence that EBP1 plays multiple cellular roles and is involved in nuclear receptor–mediated gene regulation. Selective disruption of the HNF4α–EBP1 interaction or tissue-specific EBP1 inactivation can enhance HNF4α activities and thereby improve insulin secretion in β-cells, potentially representing a new strategy for managing diabetes and related metabolic disorders.

tion or tissue-specific EBP1 inactivation can enhance HNF4␣ activities and thereby improve insulin secretion in ␤-cells, potentially representing a new strategy for managing diabetes and related metabolic disorders.
HNF4␣ (hepatocyte nuclear factor 4␣) is a unique member of the nuclear receptor (NR) 3 superfamily and plays a critical role in early vertebrate development and metabolic regulation (1). It is highly expressed in the liver, kidney, intestine, and pancreas, and its crucial role in these vital organs has been proven by a genome-wide expression profiling study (2) and conditional inactivation of its gene in mice (3)(4)(5). HNF4␣ regulates expression of a wide variety of essential genes, including those involved in the liver and pancreatic cell differentiation, glucose metabolism, lipid homeostasis, and amino acid metabolism (6 -8). In pancreatic ␤-cells, it regulates the expression of numerous genes involved in the insulin secretion signaling pathway (3,4,9,10). As such, mutations in HNF4␣ cause a dominantly inherited form of diabetes referred to as MODY1 (maturity onset diabetes of the young 1) (11), further underscoring its pivotal role in human pancreatic ␤-cell function and metabolic regulation (4,9).
As a member of the NR superfamily, HNF4␣ is comprised of distinctive modular domains and exerts its function through various molecular interactions via combinatorial recruitment of multiprotein complexes, including transcriptional coregulators and mediators that in turn facilitate remodeling of the chromatin structure. Well-known transcriptional coregulators of HNF4␣ include p160/SRC coactivators such as SRC1 and GRIP1/NCoA2 (13), CBP (14), and PGC-1␣ (15) and NR corepressors such as NCoR, SMRT, and SMILE (16,17). In addition, components of the mediator complex such as MED1 and MED25 have been hitherto identified as transcriptional binding partners of HNF4␣ (18 -20). However, its entire protein recruiting network and their physiological roles have not been well-characterized, especially in ␤-cells, and many more key regulators still remain to be uncovered. Thus, to identify additional functional binding partners of HNF4␣ in ␤-cells, we conducted yeast two-hybrid experiments using various constructs of HNF4␣ as bait and a ␤-cell library as prey and identified, among others, EBP1 (ErbB3 binding protein 1) as a novel functional binding partner of HNF4␣.
EBP1 is a member of the PA2G4 family of proliferation regulated proteins, and its multifunctional roles have been reported for regulation of gene transcription (21)(22)(23), regulation of ribosome assembly and rRNA processing through RNA interactions (24,25), transduction of growth regulatory signals through interaction with the ErbB3 receptor (26), and growth inhibition and the induction of differentiation of human cancer cells (26,27). However, its major physiological role appears to be transcriptional regulators through interaction with transcription factors and histone modifiers such as Six1 and Sin3A (23,28). In particular, EBP1 is a well-known transcriptional corepressor for another NR, androgen receptor (AR)-regulated genes, through its interaction with histone deacetylases (28 -30) and serves as a coregulator of cancer-related gene expression (31)(32)(33). In this work, we present the evidence that EBP1 also acts as a repressor of HNF4␣-mediated transcription through interacting with the LXXLL motif-binding pocket of HNF4␣, which has a direct link to glucose-stimulated insulin secretion in ␤-cells that is impaired in the HNF4␣ mutationdriven diabetes.

Initial identification of HNF4␣-EBP1 interactions
To search out additional key binding partners of HNF4␣ in pancreatic ␤-cells, we performed yeast two-hybrid screens (utilizing the service provided by the Yeast Model System Genomics facility at Duke University, which is now closed) and identified EBP1 as one of its binding partners. The bait vectors containing various constructs of HNF4␣ were made and screened against the pretransformed mouse pancreatic library as prey. When full-length HNF4␣ or the ligand-binding domain (LBD) only was used as bait, EBP1 was identified as one of the putative binding partners, together with the well-known HNF4␣ transcriptional coregulators such as PGC-1␣, GRIP1/ NCoA2, and the mediator component MED25 of RNA polymerase II (the original figures of the plates with proper controls were not provided to us except the table of potential binding partners and their sequence information). DNA sequence analysis of the positive clones revealed four cDNAs derived from the same gene, encompassing the 197-373 or 286 -367 region of EBP1 (NP_035249) containing the LXXLL motif (positions 354 -358). This mouse sequence shares 99% sequence identity with its human counterpart (only 4 amino acid differences of 394).

Further confirmation of the interaction between HNF4␣ and EBP1 and interaction domain mapping
To confirm the initial yeast two-hybrid findings, mammalian two-hybrid assay was carried out with HeLa cells. As shown in Fig. 1A, HNF4␣-EBP1 mutual interaction was evident even in the absence of additional external ligand, and this interaction was greatly reduced when the EBP1 mutant (EBP1-NR) in which the sole LXXLL sequence was mutated to LXXAA was used, suggesting its involvement in this interaction. A weak residual interaction by EBP1-NR suggests a possible involvement of other regions of EBP1 for HNF4␣ interactions, although the LXXLL motif plays a major role.
To map the critical domains of HNF4␣ and EBP1 for their mutual interactions, mammalian two-yeast assay and GST pulldown experiments were performed with a series of HNF4␣ truncation mutants and EBP1 full-length protein. The crystal structures of EBP1 and HNF4␣ are available, and they indicate that the entire structure of EBP1 is made of a single well-folded domain (34,35), whereas HNF4␣ is made of several functional modular domains (36 -38). Therefore, several different constructs of HNF4␣ encompassing various regions of the protein Figure 1. EBP1 binds directly to HNF4␣ in a LXXLL motif-mediated manner. A, initial yeast two-hybrid findings were confirmed by mammalian two-hybrid assay. This interaction was greatly reduced when the EBP1-NR mutant (a mutation of the LXXLL motif of EBP1 to LXXAA) was used. Interaction between the two proteins, as GAL4 and VP16 fusion constructs, results in an increase in firefly luciferase expression over the negative controls. B, crude minimal interaction domain mapping by the same experiment with a series of HNF4␣ truncation mutants when EBP1 full-length was used.

EBP1 acts as a novel transcriptional repressor for HNF4␣
were made and used for examination, whereas a single construct of EBP1 full-length was used. As shown in Fig. 1B and Fig.  S1, all the truncated mutants containing the HNF4␣-LBD (positions 151-377) made interactions, and the LBD alone proved to be sufficient for interaction with EBP1. These findings are in line with the initial yeast two-hybrid screening outcomes in which the protein products corresponding to the amino acid 197-373 or 286 -367 region of EBP1 containing the LXXLL motif (positions 354 -358) showed interactions with both HNF4␣-LBD alone and the full-length HNF4␣. Taken together, these binding analysis results prove that EBP1 physically interacts with HNF4␣ in living cells and present EBP1 as a potential transcriptional regulator.

EBP1 represses HNF4␣-mediated transactivation in the LXXLL motif-dependent manner
Because EBP1 is a well-known transcriptional corepressor for another NR, AR-regulated genes (28,29), we first tested its involvement in HNF4␣-mediated transcription by overexpressing both proteins and measuring the changes in the reporter gene expression level by HNF4␣ luciferase assays with HeLa cells. The reporter vector pGL3-HNF1␣ contained one copy of the HNF4␣ response element (Ϫ64 to Ϫ52) within the promoter of human HNF1␣ (Ϫ298 to the first AGT) (20). As shown in Fig. 2A, EBP1 substantially reduced the expression of the reporter gene (lanes 3 and 7), whereas this effect is completely masked by the EBP1 LXXAA mutant (EBP1-NR) or shEBP1 (lanes 4 and 8 or lanes 5 and 9, respectively), suggesting a direct involvement of EBP1 in HNF4␣-mediated transactivation (see Fig. S2 for validation of EBP1 knockdown by shRNA). To assure a direct role by EBP1, we also measured the HNF4␣ expression levels when the cells were singly transfected or cotransfected with EBP1, in which the HNF4␣ expression levels remained the same (Fig. S3A), indicating that this is not due to reduced HNF4␣ protein expression but is mainly attributed to the direct repressive activity of EBP1.
To investigate whether EBP1 also serves as a repressor for other NRs, in addition to AR and HNF4␣, a selective set of representative NRs have been tested by similar overexpression and transcription reporter assays. We chose PPAR␥, ER␣, PR, RAR␣, and RXR for the studies, among which all but ER␣ showed repressive outcomes upon the overexpression of EBP1 Figure 2. Effects of EBP1 on HNF4␣ and other NR-mediated transcription. A, overall transcriptional activity measured by standard luciferase-based transcriptional reporter assays on HNF4␣-responsive elements. The LXXLL motif mutant (EBP1-NR) and gene knockdown experiment with shRNA were also performed to test the mutational and specific protein effects, respectively (lanes 4 and 5 and lanes 8 and 9). The left five lanes are without HNF4␣ transfection (thus empty vector, single transfection of EBP1-expressing vectors, or with shEBP1), and the right four lanes correspond to the ones with HNF4␣ transfection (thus single or double transfections of HNF4␣-and EBP1-expressing vectors or with shEBP1). The scatter plots with individual data points (n ϭ 6) are shown. The midlines indicate the average (or mean) values, and the vertical lines indicate the S.E. * indicates a p value Ͻ0.001 with respect to each other. ns, nonsignificant. All data have been normalized against firefly Renilla luciferase activity. B-F, overall transcriptional activity measured by standard luciferase-based transcriptional reporter assays on other representative NR-responsive elements (PPAR␥, ER␣, PR, RAR␣, and RXR␣). For ER␣ and PPAR␥, 10 nM estradiol, a potent ligand for ER␣, and 1 M troglitazone, an agonist for PPAR␥, were added, respectively. EBP1 showed repressive effects on these NR-mediated transactivations except ER␣ (C).

EBP1 acts as a novel transcriptional repressor for HNF4␣
( Fig. 2, B-F). This is quite contrast to our previous findings in which HNF4␣ and ER␣ displayed the same positive response to the overexpression of MED25 (mediator complex subunit 25), whereas other NRs did not (20). Thus, it appears that coregulatory recruitments by NRs are member-specific, and each coregulator or mediator component works for its own cognate members of the NR superfamily (39,40).

EBP1 reduces HNF4␣ target gene expression involved in glucose-stimulated insulin secretion in ␤-cells
MODY patients are mainly characterized by a severe impairment of insulin secretion, and the mutations in the MODY gene products, including HNF4␣, are monogenic causes of an insulin secretion defect resulting in diabetes onset (41,42). Thus, to probe the involvement of EBP1 in HNF4␣-specific target gene expression and insulin secretion in ␤-cells, we next tested whether EBP1 is required for HNF4␣ transcriptional activation of previously known HNF4␣ target genes that are directly involved in ␤-cell insulin secretion such as PPAR␣ (3), L-pyruvate kinase (L-PK) (9, 10), GLUT2 (9, 10), and Kir6.2 (3,4).
These proteins are involved in the insulin secretion signaling pathway at a respective critical point such as glucose sensing and transport (GLUT2), TCA, or Krebs cycle, i.e. ATP production by mitochondrial enzymes (L-PK), ATP-dependent potassium channel (Kir6.2), or transcriptional regulation of additional gene products along the insulin secretion pathway (PPAR␣).
Target gene expression levels were measured by means of transient transfection in MIN6 cells followed by quantification of reverse-transcribed and amplified DNA products by real time PCR and Q-PCR. Our results showed that EBP1 represses the activation of the majority of HNF4␣-specific target genes involved in insulin secretion. As shown in Fig. 3 (A and B), expression of aforementioned HNF4␣ target genes were mostly reduced upon transfections of HNF4␣ and EBP1 (nearly 50%), whereas Kir6.2 showed no response. Although the ATP-dependent potassium channel is one of the central players in glucosestimulated insulin secretion, and altered potassium channel activity is related to the impaired insulin secretion in MODY patients (43,44), there are contradicting data on whether or not * indicates a p value Ͻ0.001 with respect to each other. They all equally went down, and these results are in good agreement with the Q-PCR data shown in A. C, ChIP assay using the anti-HNF4␣ antibody was performed on chromatin extracted from HNF4␣ and/or EBP1 overexpression vector-treated MIN6 cells, and the specific GLUT2 promoter region containing the HNF4␣ recognition site was amplified using PCR. Lanes 2 and 5 represent endogenous HNF4␣ binding to GLUT2 promoter. EBP1 partially inhibited HNF4␣ binding to the GLUT2 promoter region. D, effects of EBP1 WT, EBP1-NR mutant, and shEBP1 on insulin secretion shown by color change (right panel) and its quantification by absorption at 450 nm (left panel). * indicates a p value Ͻ0.001 with respect to each other. ns, nonsignificant. Darker colors (right panel) indicate more insulin secretion as shown by higher data points in their actual numerical value plots. These data show a direct correlation between HNF4␣ target gene expression and insulin secretion.

EBP1 acts as a novel transcriptional repressor for HNF4␣
one of its subunits, Kir6.2, is a direct target of HNF4␣. Although one group reported that Kir6.2 is a target gene of HNF4␣ in ␤-cells through conditional HNF4␣ knockout experiments (4), another group reported that the expression level of Kir6.2 in HNF4␣ knockout mice was unchanged as compared with control mice (3). Our findings partially support the latter observation, although we cannot rule out the possibility that EBP1 involvement is for only a subset of endogenous HNF4␣ target genes and that other repressors might be involved in HNF4␣mediated Kir6.2 transactivation. We subsequently performed ChIP assays using antibodies against HNF4␣ and the PCRbased amplification product of the target gene promoter, which further confirmed that HNF4␣ binds to the GLUT2 promoter region, and EBP1 partially inhibited HNF4␣ binding to the promoter region, resulting in reduced target gene expression (Fig. 3C).
The involvement of EBP1 in HNF4␣-mediated transcription and insulin secretion was further tested using the EBP1 LXXAA mutant (EBP1-NR) and the knockdown of EBP1 using shRNA. In keeping with the results from shRNA treatments for the luciferase assays ( Fig. 2A), the overall insulin secretion levels in MIN6 cells were recovered to the normal level by both the EBP1-NR mutant and EBP1 shRNA treatment (Fig. 3D), indicating that this repression is EBP1-specific, and once again the C-terminal LXXLL motif in EBP1 is important for this interaction. The EBP1-NR expression level was very comparable with that of EBP1 WT (Fig. S3, A and B), indicating that this reduced effect is not due to an altered protein expression level of EBP1-NR. Taken together, these data indicate that EBP1 is a critical transcriptional repressor involved in HNF4␣-mediated transcription through a direct protein-protein interaction that promotes glucose-stimulated insulin secretion in pancreatic ␤-cells.

Structure determination of the HNF4␣-EBP1 complex
We further pursued delineation of the molecular basis for EBP1 repression against HNF4␣ by determining the crystal structure of the complex. Although crystallization attempts with the full-length EBP1 were unsuccessful, its fragment containing the key interaction LXXLL motif in complex with HNF4␣-LBD produced the crystals for X-ray diffraction data analysis. Details of the crystallization and structure determination are provided under "Experimental procedures," and the typical crystals/diffraction patterns and final refinement statistics are shown in Fig. S4 and Table 1, respectively.
Although this particular crystal form contains an unusually high number of molecules in the asymmetric unit (8 dimers or 16 monomers in the asymmetric unit), this space group assignment was correctly confirmed by the program Zanuda (45) provided by the CCP4 program suits (46). Multiple independent molecules in the asymmetric unit is not uncommon, and 8 -16 monomers in the asymmetric unit have been observed in several previous structures (47). In our crystal, each dimer contains both open and closed conformations of HNF4␣ with a bound ligand (long-chain fatty acid, modeled as lauric acid) in each monomer and the EBP1 peptide only for the closed (active) conformation (Fig. 4A). Overall, the entire 16 monomers are packed in such a way that there are two sets of eight monomers (a set of two tetramers of functional dimer assembly following a local D2 symmetry) in different orientations. However, there are few structural variations among the eight dimers in the asymmetric unit (Table S1). One set of eight monomers are well-packed in the crystals, whereas the other set is loosely packed, resulting in higher B-factors and relatively weak electron density for the proteins as well as bound ligands and peptides. As a result, the ligands and peptides were not modeled for the second set of eight protein molecules with weak electron densities.

Crystal structure of the complex and the competitive nature of EBP1 repression against HNF4␣
The overall structure of HNF4␣ and the binding mode of the EBP1 LXXLL motif are nearly identical to those observed in the HNF4␣-LBD-SRC1 peptide and HNF4␣-LBD-PGC-1␣ peptide complexes that we previously reported (38,48) (Fig. S5A). SRC1 and PGC-1␣ are well-known coactivators for HNF4␣ (13, 15), and the same binding mode used by EBP1, thus potentially blocking coactivator interactions, hinted at the competitive nature of repression. When the bound peptides are superimposed, the HNF4␣ proteins superimpose extremely well with root-mean-square deviations of C␣-atom positions 0.855 and 0.858 Å between the EBP1-bound protein and each of the coactivator-bound proteins (SRC1-and PGC-1␣-bound proteins, respectively) (Fig. S5B). Their binding modes define the canonical NR/LXXLL motif interactions that are comprised of leucine-mediated hydrophobic interactions within the hydrophobic groove and a "charge clamp" at both ends created by the hydrogen bonds between the backbone atoms of the peptides and the side chains of HNF4␣ (38, 48).

EBP1 acts as a novel transcriptional repressor for HNF4␣
Another prominent feature of the structure as previously mentioned is the mixed monomer conformations within the functional dimer of HNF4␣-LBD wherein one monomer adopts a closed (active) conformation, thus occupied by the EBP1 peptide, and the other monomer displays an open (inactive) conformation with the final helix H12 extended despite the presence of a bound fatty acid (Fig. 4, A and B). The LXXLL motif binding is known to fully induce closed (active) conformations for both monomers (38,48); however, despite the same binding mode in the crystal structures, EBP1 (or its LXXLL motif) was not able to lock both monomers into an active conformation. One possibility for this observation is that EBP1 peptide displays a relatively weak binding compared with that of well-known coactivators such as SRC1 and PGC-1␣ (Fig. 4C). Thus, we tested their relative binding capabilities by measuring the degrees of protein stability elevation when peptides are bound. As shown in Fig. 4D, EBP1 peptide binding caused a smaller degree of shift in protein melting temperature compared with those caused by coactivator peptide bindings.
Although not very quantitative, these data suggest that EBP1 binds HNF4␣ rather weakly yet effectively serves as a competitive repressor against the coactivators by preventing them from binding to the same pocket and recruiting the main transcriptional machinery. A similar means of repression utilizing the same LXXLL motif has been reported to other NR-repressors such as NCoR and RIP140 (49, 50).

Discussion
The current model of eukaryotic gene regulation is best described by the combinatorial recruitment involving multiple transcriptional regulators (51,52); however, the full extent of tissue-specific and protein-specific recruitment has not been well-characterized. Transcription factors initiate transcription by recognizing their target genes and mediating additional interactions with various proteins as part of their combinatorial recruitment involving multiple transcriptional regulators and mediators to recruit the remainder of the main transcriptional machinery (51,52). To gain additional molecular insights into HNF4␣ function, we performed yeast two-hybrid screens and identified EBP1 as a functional binding partner of HNF4␣ in pancreatic ␤-cells.
Our study was focused on ␤-cells because HNF4␣ is one of the culprit gene products for a dominantly inherited form of diabetes, MODY1, which is mainly characterized by the defect in insulin secretion from ␤-cells (41,42). The interaction between HNF4␣ and EBP1 was initially identified by yeast twohybrid analysis and further confirmed by mammalian two-hybrid assay and GST pulldown. Unlike other novel NR repressors that typically use the repressor-specific motifs such as LXX(I/ H)IXXX(I/L) (53), EBP1-HNF4␣ interaction utilizes the same LXXLL motif used by the activators, thus representing an atypical corepressor with binding properties of a coactivator and

. EBP1 LXXLL motif interaction with HNF4␣ and its effect on HNF4␣-LBD overall conformation.
A, overall structure of the complex and the close-up view of the LXXLL motif-mediated EBP1 binding to HNF4␣ represented by the final 2F o Ϫ F c map contoured at 1 (inset). The backbone of the bound peptide is shown as a purple wire in the inset, whereas the overall structure is shown as a ribbon diagram. The bound fatty acid (structural ligand) is also shown as sticks in the overall structure. B, HNF4␣-apo protein conformation (PDB access code 1M7W), coactivator-bound activated conformation (PDB access code 1PZL or 3FS1), and competitive-repressor-bound conformation (PDB access code 6CHT), which display two different states of dimeric assembly. Only the coactivators are capable of fully inducing active conformation in which the helix 12 (H12) of both monomers fold in. C, sequence alignment of the LXXLL motifs that can bind to HNF4␣ and used in the structure determinations (B). Among three LXXLL motifs of PGC-1␣, only the middle one with the highest binding affinity is included in the figure. D, normalized melting curves depicting shifted thermal stability of HNF4␣-LBD when forming complexes with the LXXLL motif-containing peptide of each protein (colored lines) from that of the HNF4␣-LBD apo protein (black line). A melting temperature (T m ) for apo protein and ⌬T m values for peptide-bound proteins are indicated in parentheses. PGC-1␣ peptide shows the best binding followed by SRC-1 and EBP1 peptides.

EBP1 acts as a novel transcriptional repressor for HNF4␣
functions as a repressor by direct competition. The minimal interaction domain of HNF4␣ has been identified, and the physiological implications of EBP1-mediated transactivation of HNF4␣ were investigated by reporter gene assays and insulin secretion assays in combination with gene silencing studies. Subsequent crystal structure determination of the HNF4␣-LBD in complex with a peptide harboring the EBP1 LXXLL motif proves the competitive nature of suppression against the coactivators harboring the same LXXLL motif.
Apparently, there are two different alternatively spliced EBP1 isoforms (p48 made of 394 amino acids and p42 made of 340 amino acids). p48 is 54 amino acids longer than p42 at the N terminus. Because of this difference, p42 is missing a few posttranslational modification sites that p48 has and interacts with a slightly different set of proteins from what p48 does (27). However, they both contain the LXXLL motif toward the C-terminal end and are expected to interact with HNF4␣. Thus, we only used the p48 isoform throughout the study because it has the complete structure, and p48 is the predominant form in mammalian cells and is found both in cytoplasm and nucleus, whereas p42 is mostly found in cytoplasm (27,54).
EBP1 has already been reported to be a non-sDNA-binding corepressor that can interact with and repress the transcriptional activity of various transcription factors such as Six1 and E2F1 (23,55). In particular, EBP1 interacts with AR, another nuclear receptor family member, through the LXXLL motif and represses AR-mediated transcription (29,30). In this report, we present further supporting evidence that EBP1 is functionally recruited by a nuclear receptor HNF4␣, indicating that EBP1 is involved in gene selective activation of target genes by diverse yet evolutionally related transcription factors. Although the LXXLL motif of transcription regulators are involved in interactions with many NR family members, this transcriptional regulatory effect by EBP1 appears to be NR-specific, and overexpression of EBP1 does not alter the level of ER␣-mediated transcription. These findings establish EBP1 as a transcriptional repressor of HNF4␣ in ␤-cells leading to insulin secretion and shed light on the molecular mechanisms for nuclear receptor-mediated transcriptional repression and various regulators involved in the process.
Recently, tumor suppressor or oncogenic activities of EBP1 have also been observed in various cancer types, and EBP1 can serve as a prognostic indicator for certain cancer types (56 -63). EBP1 is either down-regulated or up-regulated in these cells, and the transcriptional regulatory function of EBP1 is believed to contribute these cancer developments. EBP1 mutations have also been reported to be associated with certain cancer types such as colorectal cancers (64,65). Our data presented here could be applicable to these situations in various tissues where both HNF4␣ and EBP1 are expressed. In all cases, we believe our newly obtained results have advanced the current understanding of the molecular mechanisms underlying HNF4␣ function and suggested a novel strategy for targeting this protein for the discovery of small compounds that can antagonize the repressive complex formation, thus improving insulin secretion and overall ␤-cell function for diabetes treatment and potentially being used for other EBP1-associated cancer therapies.

Yeast two-hybrid
The initial screening using the lexA system was carried out through the Yeast Model System Genomics facility at Duke University. HNF4␣-LBD (positions 142-368) was cloned into the bait vector pGBKT7short (a modified version of pGBKT7 (Clontech) with tags between Gal4BD and bait removed), and the mouse pancreatic library (Clontech) was used as prey. Selfactivation was tested with an empty Gal4BD bait vector before the screening against the library. Two-hybrid screens were performed with a standard method using the yeast strain S. cerevisiae AH109. Primary isolates were restreaked on TrpϪ/LeuϪ/ HisϪ/3 mM 3-amino-1,2-4-triazol plates and grown several days for LacZ assays. Positive colonies that showed a color change in LacZ assays were picked for colony PCR or for isolation of DNA. Approximately 6 ϫ 10 6 independent transformations were screened, of which 27 clones were positive for LacZ assays. Each positive hit was retested for one to one interaction by examining the growth of the transformant and performing LacZ assays. Subsequently, the positive clones were sequenced using automated DNA sequence analysis (ABI), and homologies were identified using BLASTN/BLASTX (National Center for Biotechnology Information).

EBP1-NR mutant generation
The QuikChange multisite-directed mutagenesis kit (Stratagene) was used to generate the EBP1-NR (LXXAA) mutant construct. The plasmid templates used in the mutagenesis protocol were pCMV Sport6 EBP1-FL. All of the generated constructs with the mutated sequences were verified with DNA sequencing.

Mammalian two-hybrid assays
The DNAs, pBIND-HNF4␣ FL, AF1, DNA-binding domain, LBD, FD, pACT-EBP1 WT, NR, and pG5-luciferease, were combined as indicated in each lane in the total amount of DNA not exceeding more than 100 ng/well. The transfection was performed with TurboFect (Fermentas, Glen Burnie, MD) and following the manufacturer's recommended protocols (Promega E2440). The cells were disrupted by addition of 150 l of cell lysis buffer directly into each well of the 48-well plate, and then aliquots of 70 l were added to each well of a 96-well luminescence plate. The luminescence activity was measured automatically by microplate reader (SpectraMax M5; Molecular Devices). The relative luciferase activity was calculated and normalized based on the pG5-luciferase basal control. For assessment of transfection efficiency, the Renilla luciferase activity assay was used.

GST pulldown assays
MIN6 cells were transiently transfected with expression plasmids as indicated in the figures by using Metafectene Pro transfection reagent (Biontex). After 24 h, the cells were harvested and sonicated in 1 ml of the lysis buffer containing 50 mM Tris, pH 7.5, 5 mM EDTA, 150 mM NaCl, 1 mM DTT, 0.01% Nonidet P-40, 2 g/ml BSA, 0.2 mM phenylmethylsulfonyl fluoride plus protease inhibitors (50 g of aprotinin and 5 g of leupeptin/ ml). After centrifugation, 500 g of a freshly solubilized postnuclear supernatant was mixed with 25 g of GST-fused HNF4␣ (full-length or LBD-only) immobilized onto GSH-Sepharose in buffer (100 mM Tris, pH 7.5, 10 mM EDTA, 300 mM NaCl, 2 mM DTT, 0.02% Nonidet P-40, 4 g/ml BSA, 0.4 mM phenylmethylsulfonyl fluoride plus protease inhibitors). After incubation at 4°C for 2 h, the beads were extensively washed, and bound proteins were eluted by adding 20 mM GSH. The elution samples and 10% of the homogenate (labeled input in Fig. S1) were applied to SDS-PAGE prior Western blotting detection with antibodies against EBP1 (Santa Cruz Biotechnology).

Transient transfection and transcription assays (luciferase reporter assays)
The full-length cDNA of human HNF4␣ WT or the mutant were subcloned into the pcDNA3(ϩ)/Neo vector (Invitrogen), and the reporter vector pGL3-HNF1␣ containing one copy of the HNF4␣ response element (-64 to Ϫ52) within the promoter of human Hnf1␣ (Ϫ298 to the first AGT) was constructed and used for luciferase assays in the absence or presence of EBP1. HeLa cells were transfected using Opti-MEM and Lipofectamine 2000 reagent (Thermo Fisher Scientific) according to the manufacturer's recommendations. Briefly, a total of 30 ng of pcDNA3 HNF4␣ and 150 ng of pCMV Sport6 PGC-1␣, 50 ng of pCMV Sport6 EBP1, 50 ng of pGL3-HNF1␣, and 10 ng of pRL-TK (control Renilla luciferase vector) were used for transfection of 1 ϫ 10 5 cells seeded on 24-well plate 1 day before transfection. For gene silencing experiments, 20 pmol of shRNA of EBP1 and HNF4␣ were used for inhibition. For ER␣ and PPAR␥ luciferase assays, 6 h after transfection, the cells were treated with 10 nM of estradiol or 1 M of troglitazone for 48 h. These additional luciferase vectors were kind gifts from Dr. Dan Noonan at the University of Kentucky. After transfection and incubation, the cells were washed with 1ϫ PBS and lysed with luciferase lysis buffer supplied with the Luciferase assay kit (Promega). Luciferase activity was measured using the Dual Luciferase assay system (Promega) and Luminoskan (Thermo Fisher Scientific). All values were normalized by the relative ratio of firefly luciferase activity and Renilla luciferase activity. At least four independent transfections were performed in duplicate.

Gene knockdown (shRNA) for cellular studies
We used shRNA for EBP1 interference in the real time PCR/ Q-PCR and insulin secretion assays. EBP1 shRNA was purchased from Origene with the following sequences: 5Ј-ATGT-GGATGGCTTCATCGCTAATGTAGCT-3Ј. Cells grown to 50% confluence were transfected using Metafectene pro transfection reagent (Biontex) with EBP1 or scrambled shRNA according to the manufacturer's instructions.

Real-time PCR
MIN6 cells were transiently transfected with HNF4␣ and EBP1 using Metafectene protransfection reagent (Biontex) for 24 h. Total RNA from the treated cells was prepared with the Tri reagent (Sigma-Aldrich), and 0.5 g of RNA was reverse transcribed using Onestep RT-PCR kit (Qiagen) and amplified by PCR whose product formation was monitored continuously during PCR using sequence detection system software (version 1.7; Applied Biosystems). Accumulated PCR products were detected directly by monitoring the increase of the reporter dye (SYBR). The expression levels of PPAR␣, L-PK, GLUT2, and Kir6.2 in the exposed cells were compared with those in control cells at each time point using the comparative cycle threshold (Ct) method. The quantity of each transcript was calculated as described in the instrument manual and normalized to the amount of actin, a housekeeping gene.

Q-PCR
MIN6 cells were transfected with HNF4␣ and EBP1 plasmids. 24 h after transfection, total RNA was extracted using Tri reagent (Sigma-Aldrich) according to the manufacturer's protocol. cDNA synthesis and semiquantitative PCR for PPAR␣, L-PK, Kir6.2, and GLUT2 mRNA were performed, and the PCRs were electrophoresed using a 2% agarose gel. The band

EBP1 acts as a novel transcriptional repressor for HNF4␣
intensities of the amplified DNA products were visualized using the SYBR Green I DNA gel stain kit (Invitrogen).

ChIP assay
EBP1 and/or HNF4␣ plasmids-transfected MIN6 cells were cross-linked with formaldehyde in PBS at room temperature for 10 min then sonicated to shear DNA strands into the size of 200 -500 bp. The sonicated chromatin-DNA complexes were precipitated with antibodies against HNF4␣ (Santa Cruz Biotechnology catalog no. sc-6556) or nonspecific mouse IgG. PCR analysis was performed using 2 l of purified DNA, oligonucleotide primers, and Platinum Taq DNA polymerase. After 40 cycles, the products were analyzed using a SYBR green I DNA gel stain kit (Invitrogen).

Insulin secretion assays
This functional study was performed with the MIN6 cell line, which exhibits the characteristics of glucose metabolism and glucose-stimulated insulin secretion similar to those of normal islets (68). To quantify the amount of insulin secreted, MIN6 cells were grown on a 6-well dish (about 1 ϫ 10 6 cells) and transfected with various vectors harboring HNF4␣, EBP1 (WT or the NR mutant), or EBP1 shRNA and incubated for 24 h. After the incubation period, the cells were washed three times with KRB buffer (119 mM NaCl, 4.7 mM KCl, 2.5 mM CaCl 2 , 1.2 mM MgSO 4 , 1.2 mM KH 2 PO 4 , 25 mM NaHCO 3 , 10 mM Hepes, pH 7.4, and 0.1 g of BSA) and further incubated in 1 ml of prewarmed KRB buffer containing 1 mM glucose for 1 h at 37°C. The cell culture medium (total of 1 ml) was collected and used to measure the level of insulin release with the insulin ELISA kit (Mercodia) by means of an enzyme immunoassay followed by an optical density reading at 450 nm. The amount of stimulation (fold increase) refers to insulin secretion after various treatments relative to insulin secretion in 1 mM glucosetreated cells, which was set as 1-fold. The values are expressed as means Ϯ S.D. of data obtained from three independent experiments, each performed in duplicate (n ϭ 6).

Crystal data collection and structure determination
The crystals belong to the space group P21 with the unit cell dimensions a ϭ 139.727 Å, b ϭ 104.947 Å, c ϭ 139.573 Å, and ␤ ϭ 90.613°. The best crystals after optimizing freezing conditions diffracted to 3.15 Å resolution with the synchrotron radiation (collected at the beamline SER-CAT 22ID, Advanced Photon Source, Argonne, IL). There are 16 HNF4␣-LBD-ligand-EBP-1-NR-box-fragment complexes (eight functional dimeric complexes) in the crystal asymmetric unit resulting in 54.5% solvent content. Oscillation images (every 1.0°) were collected at 100 K, and the data were processed using the HKL2000 software package (70) ( Table 1).
Primary phasing was accomplished by molecular replacement using PHASER (McCoy, 2007). Our previous structure of human HNF4␣-LBD (PDB accession code 1PZL) (38), after deleting a bound ligand and the SRC-1 peptide, was used as a search model. The initial R value was 0.48 with a correlation coefficient of 0.57. The subsequent -weighted 2F o Ϫ F c map after rigid body refinement clearly revealed density corresponding to the bound EBP1 peptide that was not present in the search model. Further refinement was carried out with Refmacs as run by PHENIX (71) alternating with manual fitting in COOT (72) until convergence. A twin inspection utilizing the Xtriage data analysis tool within PHENIX revealed that there is an apparent twin law (l, Ϫk, h operator) given the crystal symmetry that really stands out over the remainder of the possible twin laws (R obs 0.115 versus Ͼ0.450 for the rest of possible operators). Thus, this particular twin target function was used throughout the refinement. Individual atomic coordinates, group B-factors, and NCS (non-crystallographic symmetry) constraints (eight open and eight closed conformations of HNF4␣ and eight bound EBP1 peptides) were utilized for initial rounds of refinement while enforcing a twin law of the l, Ϫk, h operator. Toward the end of refinement, individual B-factors, TLS refinement parameters, and NCS restraints were employed, with geometry and B-factor restraint weightings being optimized for each cycle. Solvent molecules were added manually in COOT. The final model was validated with AutoDepInputTool (74) and MolProbity (73) prior to deposition in the PDB. Data and refinement statistics are provided in Table 1. The figures were prepared with PyMOL (Schrödinger, LLC).

Fluorescence-based thermal shift assay
Purified recombinant HNF4␣-LBD protein was used to measure peptide interactions in a fluorescence-based thermal shift assay as described previously (12). HNF4␣-LBD protein was aliquoted into PCR tubes in a buffer containing 20 mM Tris, pH 8.0, and 200 mM NaCl. The final protein concentration in a 20-l reaction volume was 20 M. Peptides to be tested were added at either 5ϫ or 10ϫ concentration such that the DMSO concentration never exceeded 2%. SYPRO Orange dye (Invitrogen) was added last at a 5ϫ concentration. The PCR tubes were then sealed, centrifuged, and heated from 25 to 95°C degrees at a rate of 1°C/min on 7500 real-time PCR machine (Applied Biosystems). Raw data analysis and curve fitting to calculate T m values were performed as described by Niesen et al. (12).

Statistical analysis
Presented data are expressed as means Ϯ S.E. of at least three independent groups. Statistical significance was determined by one-way analysis of variance followed by Student-Newman-Keuls method using Sigma Stat 3.1 software (Systat Software,

EBP1 acts as a novel transcriptional repressor for HNF4␣
San Jose, CA). A probability value p Ͻ 0.05 was considered statistically significant.