Gene-specific Transcriptional Activity of the Insulin cAMP-responsive Element Is Conferred by NF-Y in Combination with cAMP Response Element-binding Protein*

Cyclic AMP stimulates insulin gene transcription through a cAMP response element (CRE). In the present study the insulin CRE-binding proteins and their functions were investigated. A mutational analysis of nuclear protein binding in electrophoretic mobility shift assays in combination with specific antisera showed that in the CRE of the rat insulin I gene the imperfect CRE octamer-like sequence TGACGTCC interacts weakly with CREB and overlaps with two sequence motifs (TTGTTGAC and CCAAT) that bind winged helix-like proteins and the transcription factor NF-Y, respectively. Transient transfection of wild-type and mutant insulin CRE-reporter fusion genes and the inactivation of cellular CREB or NF-Y by overexpression of the dominant negative mutants KCREB or NF-YA29, respectively, indicate that cAMP inducibility of the insulin CRE is mediated by CREB or closely related proteins; however, NF-Y binding to the insulin CRE confers constitutive, basal activity and decreases the ability of CREB to mediate cAMP-stimulated transcription and calcium responsiveness. Results from these studies demonstrate that NF-Y binds to the insulin CRE and modulates the function of CREB. Together with the nonpalindromic sequence of the CRE octamer motif, the interaction of NF-Y with CREB may be responsible for the gene-specific transcriptional activity of the insulin CRE and explain why it has considerable basal activity but is less responsive to cAMP stimulation than others.

Although CREB is a ubiquitous protein, gene-specific differences exist in the responsiveness of different CREs to transcriptional activation (1)(2)(3)(12)(13)(14)(15). In addition to CREB and the closely related proteins ATF-1 and CREM, the CRE motif is recognized by a number of proteins of the CREB/ATF family of transcription factors that are not activated by cAMP but may compete with CREB for binding to the CRE (1)(2)(3). CREs in which the DNA binding or function of CREB is modulated by the presence of another protein at adjacent or overlapping sites have been described (1)(2)(3)13) and in some cases the accessory proteins have been identified including the glucocorticoid receptor (14,16), YY1 (17), and C/EBP proteins (15, 18 -21). Therefore, a given CRE may have unique functional properties depending on its specific DNA sequence both within the core octamer and nucleotides flanking this motif.
Gene-specific transcriptional activities have been reported also for the CRE of the rat insulin I gene (22), although the molecular basis for its unique functional properties has remained unclear. The peptide hormone insulin is synthesized in ␤ cells of the islets of Langerhans. It is a key regulator of blood glucose concentration, and inappropriate regulation of insulin production and secretion causes diabetes mellitus. Cyclic AMP stimulates insulin gene transcription (22,23). In the absence of other signals, cAMP inducibility of the insulin gene is modest (22)(23)(24). However, cAMP stimulates insulin gene transcription synergistically with glucose (25)(26)(27)(28). Potent stimulators of ␤ cell cAMP levels are hormones including glucagon-like peptide-1 (29), which appears to be a physiologically important hormonal mediator of the "incretin effect" on insulin secretion and has been proposed as a new therapeutic agent for the treatment of non-insulin-dependent diabetes mellitus (29). Therefore, similar to the glucose competence concept of insulin secretion (29), a synergistic interaction between the hormonally regulated cAMP-dependent signaling system and the glucose-regulated signaling system may give ␤ cells the ability to match the ambient concentration of glucose to an appropriate transcriptional response of the insulin gene (24). Despite its suggested physiological significance, the molecular mechanism of cAMP-induced insulin gene transcription is poorly understood.
Studies of the transcriptional activity of reporter fusion genes in insulinoma cells demonstrated that a CRE in the 5Ј-flanking region of the rat insulin I gene is required and sufficient for cAMP induction (22,23). However, the insulin CRE shows unique functional properties; it (i) has considerable basal transcriptional activity, (ii) gives a weak cAMP response, and (iii) is not responsive to membrane depolarization and calcium influx (22). These activities distinguish the insulin CRE from typical CREs that share a perfect 5Ј-TGACGTCA-3Ј octamer, bind CREB, and fully respond to cAMP and depolarization-induced calcium influx in insulinoma cells (22, 30 -32). The special properties of the insulin CRE may be explained by its particular DNA sequence and distinct pattern of nuclear protein binding (22). The insulin CRE octamer, 5Ј-TGACGTCC-3Ј (between Ϫ178 and Ϫ185 relative to the transcriptional start site), differs in one position from the CRE octamer consensus motif. Cellular CREB binds weakly to the insulin CRE (22), such that CREB cannot be detected among the nuclear proteins binding to the insulin CRE in an electrophoretic mobility shift assay with labeled insulin CRE used as probe (22). Furthermore, the insulin CRE contains a CCAAT box motif that in some other genes has been shown to confer cAMP inducibility (33)(34)(35)(36). Therefore, the role of CREB in the cAMP responsiveness of the insulin CRE was unclear, and the identities of the other insulin CRE-binding proteins as well as their functional significance were unknown. The present study addressed these questions.

EXPERIMENTAL PROCEDURES
Plasmid Construction-The plasmid Ϫ85InsLuc was prepared by cloning a fragment of the rat insulin I gene promoter (from Ϫ85 to ϩ49) with 5Ј-XhoI and 3Ј-BglII ends into the XhoI-BglII sites of pXP2 (37). For 4xInsCRE(Ϫ85Ins)Luc, 4xm1-InsCRE(Ϫ85Ins)Luc, 4xm2-InsCRE (Ϫ85Ins)Luc, 4xm3-InsCRE(Ϫ85Ins)Luc, 4xm4-InsCRE(Ϫ85Ins)Luc, 4xE␣-CAAT(Ϫ85Ins)Luc, and 4xG3B(Ϫ85Ins)Luc, four copies of the synthetic oligonucleotides with 5Ј-GATC overhangs (for sequences see Fig. 1 or further below) were cloned in the forward orientation into the BamHI site of Ϫ85InsLuc. The plasmid Ϫ410InsLuc was prepared by cloning a fragment of the rat insulin I gene promoter (from Ϫ410 to ϩ49) with 5Ј-XhoI and 3Ј-BglII ends into the XhoI-BglII sites of pXP2 (37). Four bases in the CRE octamer-like sequence (from Ϫ183 to Ϫ180) were selectively deleted inside the insulin promoter with the restriction enzyme AatII and T4 DNA polymerase, yielding the construct Ϫ410(⌬Ϫ183/Ϫ180)InsLuc. Subcloning and plasmid isolation were performed by standard procedures. All constructs were sequenced by the enzymatic method to confirm the identity and the orientation of the inserts.
Cell Culture and Transfection of DNA-The pancreatic islet cell line HIT-T15 (38) was grown in RPMI 1640 medium supplemented with 10% fetal calf serum, 5% horse serum, 100 units/ml penicillin, and 100 g/ml streptomycin. Cells were trypsinized and transfected in suspension by the DEAE-dextran method as described (30) with 2 g of indicator plasmid per 6-cm dish. Rous sarcoma virus-chloramphenicol acetyltransferase plasmid (0.4 g/6-cm dish) was added as a second reporter to check for transfection efficiency. When indicated, 2 g of RSV-KCREB (39) or, unless indicated otherwise, 3 g of an expression vector encoding the dominant negative mutant of NF-Y, NF-YA29 (40), were co-transfected per 6-cm dish. These co-transfections were done with a constant DNA concentration, which was maintained by adding Bluescript (Stratagene, La Jolla, CA). Cells were stimulated with forskolin (10 M) or high KCl (45 mM final concentration) for 6 h before harvest. Cell extracts (30) were prepared 48 h after transfection. A chromatographic chloramphenicol acetyltransferase assay (41) and the luciferase assay (30) were performed as described previously. Thin layer chromatography plates were analyzed with a Fuji PhosphorImager.
Nuclear Extracts-Nuclear extracts were prepared from HIT cells by the method of Dignam et al. (42) with the modification described (22).
Electrophoretic Mobility Shift Assay-Using 15 g of protein from nuclear extracts, the electrophoretic mobility shift assay was performed as described (41). In some binding reactions 2.5 l of an antiserum directed against the B subunit of NF-Y, ␣-NF-YB (43), 2 l of a specific anti-CREB antiserum (R1090) (44), or an equal volume of preimmune serum was added to the binding reaction, and the assay was then performed as described (22).
Oligonucleotides-The sequences of the CRE oligonucleotides of the rat insulin I gene (wild type and mutants 1 to 4) and the rat somatostatin gene are shown in Fig. 1. The sequences of other oligonucleotides were as described previously or read as follows (only one strand with the 5Ј-GATC overhang is shown): TTR-HNF-3, containing a fragment of the transthyretin promoter from Ϫ111 to Ϫ85 that includes an HNF- Materials-A stock solution of forskolin (100 mM) was prepared in dimethyl sulfoxide and further diluted in cell culture medium. Controls received the solvent only.

Mutational Analysis of Nuclear Protein
Binding to the Insulin CRE-Using an electrophoretic mobility shift assay, it has been shown previously that three unidentified nuclear proteins bind to the rat insulin I gene CRE (complexes 1, 2, and 3) (Ref. 22; see also Fig. 2A, lane 3). As a first approach to characterize these insulin CRE-binding proteins a mutational analysis was performed. The sequences of the oligonucleotides used are shown in Fig. 1. Oligonucleotides containing the insulin CRE wild-type sequence or mutants 1 to 4 were labeled and incubated with nuclear extracts from the insulin-producing ␤-cell line HIT. The labeled CRE of the rat somatostatin gene was used for comparison. As shown in Fig. 2A, complex 1 binding to the insulin CRE was also detected using mutant 2 as probe, whereas complexes 2 and 3 were no longer formed. In contrast, when mutants 3 or 4 were used as probes, complex 1 was not detectable, whereas complexes 2 and 3 persisted ( Fig. 2A). It has been shown previously (22) and was confirmed in the present study (not shown) that complexes 1, 2, and 3 are not formed with labeled mutant 1. Note that a new protein complex was detected with labeled mutant 3 that comigrated with complex A on the labeled somatostatin CRE (Fig. 2A, compare lanes  2 and 1). A band of similar mobility but lower intensity was detected with labeled mutant 4 ( Fig. 2A, lane 5). Labeled mutant 4 showed also an additional band migrating more slowly than complex 1 ( Fig. 2A, lane 5); this band was not further investigated. The results of this mutational analysis are summarized in Fig. 2B, which indicates the mutations that abolish the binding of complex 1 or complexes 2 and 3, respectively. The bases that are required for the binding of bands 2 and 3 fall within a sequence that shows high similarity with a consensus motif for binding sites of HNF-3 proteins and related members FIG. 1. Sequences of the CRE oligonucleotides used. The oligonucleotides contain the CREs of the rat insulin I (InsCRE, from Ϫ193 to Ϫ168) or rat somatostatin genes (SomCRE, from Ϫ58 to Ϫ31) with 5Ј-GATC overhang. m1 to m4, mutant 1 to 4. The CRE octamer motif or related sequence is boxed. The bases marked by a dot indicate the 5Јand 3Ј-ends of the gene sequence. The mutated bases are underlined. The one-base deletion in mutant 1 of the insulin CRE is indicated by a dash.
of the winged helix family of transcription factors (match of 10 bases out of 12) (52) (Fig. 2B). The bases that are critical for the binding of complex 1 are very similar to a consensus NF-Ybinding site (match of 8 bases out of 10) (53) (Fig. 2B).
Evidence That Complexes 2 and 3 Represent Winged Helixlike Proteins-The significance of these sequence similarities was further investigated. As shown in Fig. 3A, nuclear protein binding to labeled insulin CRE forming complexes 2 and 3 was selectively competed away by an oligonucleotide (TTR-HNF-3) that contained a well characterized binding site for HNF-3 proteins in the transthyretin promoter (46,52). It was, however, not competed away by an oligonucleotide (TTR-HNF-4) that contained another fragment from the transthyretin promoter with an HNF-4 binding site (Fig. 3A). TTR-HNF-3 did not compete for protein binding to labeled m2-InsCRE (which lacks complex 2 and 3 binding) but did selectively compete for protein binding to labeled m4-InsCRE that comigrated with complexes 2 and 3 of labeled InsCRE (Fig. 3A). TTR-HNF-3 also competed for protein binding to labeled m3-InsCRE which comigrated with complex 3 of labeled InsCRE (Fig. 3A); at the same time the intensity of the band comigrating with complex 2 was only somewhat reduced by TTR-HNF-3 ( Fig. 3A), revealing that the mutation in mutant 3 allowed the binding of two new proteins, one comigrating with complex A of labeled Som-CRE as mentioned earlier ( Fig. 2A) and one that, although it does comigrate with, is distinct from complex 2; noteworthy, this complex does comigrate with complex B of labeled Som-CRE (see Fig. 2A). The sequence similarity of their binding sites within the insulin CRE to an HNF-3 consensus site and the competition by TTR-HNF-3 suggest that the binding specificity of the proteins forming complex 2 and 3 is related to that of winged helix proteins. Winged helix proteins share a conserved DNA-binding domain and thus exhibit related but distinct DNA-binding specificities (52,54,55). The winged helix protein HNF-3␤ is expressed in islets and also in HIT cells (56 -58). However, the addition of antisera directed against HNF-3␤, -3␣, or -3␥ to the binding reaction did not affect nuclear protein binding to the insulin CRE (data not shown). Evidence has been presented recently that another, not yet identified member of the winged helix family of transcription factors binds to the glucagon and somatostatin genes (45); it interacts with domain B of the glucagon G3 element and the FIG. 2. Mutational analysis of sequences within the insulin CRE required for nuclear protein binding as revealed by the electrophoretic mobility shift assay. A, nuclear extracts from the insulin-producing islet cell line HIT were incubated with the labeled oligonucleotides indicated. Specific protein complexes formed with In-sCRE (lane 3) are indicated as complexes 1-3. Specific protein complexes formed with SomCRE (lane 1) are indicated as A and B. F, free probe. B, the binding sites of complex 1 and complexes 2 and 3 within the insulin CRE show high similarity with consensus sites for NF-Y and HNF-3/winged helix proteins, respectively. The position of the mutations in mutant 1 (q), mutant 2 (ϩ), mutant 3 (f), and mutant 4(*) that abolished the binding of complex 1 or complexes 2 and 3 are indicated above or below the insulin CRE wild-type sequence, respectively. The CRE octamer-like sequence is underlined. Match to consensus binding sites for NF-Y (53) and HNF-3/winged helix proteins (52) are indicated by marks and given in parentheses.

FIG. 3. Competition experiments show a binding specificity of complexes 2 and 3 that is related to winged helix proteins.
Nuclear extracts were incubated with the labeled oligonucleotides indicated. Specific protein complexes formed with InsCRE are indicated as complexes 1-3. F, free probe. A, competitors containing binding sites for HNF-3 or HNF-4 from the transthyretin promoter (TTR) were added at a 500-fold molar excess, except for lane 2 (1500-fold molar excess). B, cross-competition of the insulin CRE and domain B of the glucagon G3 element (G3B) for binding of complexes 2 and 3. Competitors were added at a 5-, 50-, and 500-fold molar excess (from left to right).
somatostatin upstream element (45). When an oligonucleotide containing domain B of the glucagon G3 element (G3B) was used as probe, protein complexes were formed that comigrated with complexes 2 and 3 of the insulin CRE (Fig. 3B). Crosscompetition of the insulin CRE and G3B for binding to these complexes was observed using a similar molar excess of the competitors (Fig. 3B). Thus, based on their binding specificity, the proteins forming complexes 2 and 3 on the insulin CRE could be members of the winged helix family of transcription factors distinct from HNF-3␤ but the same as those binding to the glucagon G3 element.
Complex 1 Represents Binding of NF-Y-The binding site of complex 1 within the insulin CRE contains a CAAT motif with strong similarity to an NF-Y consensus site (Fig. 2B). Se-quences related to "CAAT boxes" can, however, be recognized by diverse transcription factors with distinct DNA-binding domains including C/EBP (59) and CTF/NF-I (48). As shown in Fig. 4A, complex 1 with the insulin CRE is not competed away by oligonucleotides containing a C/EBP-binding site (Glu-C/ EBP) (47) or an CTF/NF-I consensus site (49). Complex 1 binding is, however, competed away by CAAT boxes from the ␣ subunit of the choriogonadotropin gene (CG␣-CAAT) (60), the long terminal repeat of the Moloney murine sarcoma virus (MSV-CAAT) (50,51), and the E␣ gene from the murine class II genes of the major histocompatibility complex (E␣-CAAT) (51) (Fig. 4A), all of which are recognized by NF-Y (50, 51, 60). CG␣-CAAT competed also for complex 2 and 3 binding (Fig.  4A), consistent with the fact that CG␣-CAAT is recognized by the winged helix-like proteins binding to G3B (45). Cross-competition was observed between InsCRE and E␣-CAAT as increasing amounts of InsCRE and E␣-CAAT competed for complex 1 with labeled InsCRE (Fig. 4B) and protein binding to labeled E␣-CAAT (Fig. 4C). Irrespective of whether InsCRE or E␣-CAAT were used as probe, E␣-CAAT was a stronger competitor than InsCRE (Fig. 4, B and C), suggesting that E␣-CAAT binds with somewhat higher affinity. Competition for protein binding to E␣-CAAT was also seen with m2-InsCRE, whereas m4-InsCRE (NF-Y site mutated) did not compete (Fig.  4C), indicating that the competition by InsCRE was specific. The addition of an antiserum directed against NF-Y to the binding reaction abolished complex 1 of labeled InsCRE, whereas complexes 2 and 3 remained unaffected (Fig. 4D). Protein binding to labeled E␣-CAAT comigrated with complex 1 binding to the insulin CRE and was markedly reduced by the anti-NF-Y antiserum (Fig. 4D). These results, when taken together, strongly suggest that complex 1 represents the binding of NF-Y or a closely related protein to the insulin CRE.
Mutational Analysis of Sequences within the Insulin CRE That Are Required for Function-As a first approach to study the functional significance of the proteins binding to the insulin CRE, the transcriptional activity of the insulin CRE and the mutants 1 to 4 was investigated in transient transfection experiments. Four copies of the insulin CRE wild-type sequence or mutants 1 to 4 were cloned in front of the truncated insulin promoter (from Ϫ85 to ϩ49) fused to the luciferase reporter gene. These fusion genes were transiently transfected into HIT cells. As has been shown before (22), the insulin CRE conferred basal activity (12.6-fold increase) as well as modest cAMP responsiveness to the minimal promoter (2-3-fold stimulation) ( Fig. 5; see also Fig. 6). Mutant 1 did not show any transcriptional activity (Fig. 5). Mutant 2 conferred basal activity as did the wild-type sequence but did not respond to cAMP stimulation (Fig. 5). The mutations in mutants 4 and 3 reduced and abolished, respectively, basal activity, whereas they enhanced cAMP responsiveness (Fig. 5). Forskolin stimulated transcrip-tion through m4-InsCRE 12-fold and through m3-InsCRE 148fold (Fig. 5).
Transcriptional Activity of Binding Sites for NF-Y and Winged Helix-like Proteins in Islet Cells-The above results obtained with mutant 2, which lacks complex 2 and 3 binding, and with mutants 3 and 4, which lack NF-Y binding, suggest that basal activity may be conferred to the insulin CRE by NF-Y; NF-Y may not, however, mediate cAMP responsiveness. Consistent with this assumption the NF-Y-binding site E␣-CAAT (binding complex 1) conferred basal activity to the minimal promoter but not cAMP responsiveness (Fig. 6). Basal activity of E␣-CAAT was less than that of InsCRE (Fig. 6), which may be explained by a different spacing of the NF-Ybinding sites in the oligomerized E␣-CAAT construct which allows functional synergism in a somewhat less efficient way. G3B, which binds the winged helix-like proteins (complexes 2 and 3), did not show any transcriptional activity (Fig. 6). These data suggest that proteins others than NF-Y and the winged helix-like proteins of complexes 2 and 3 confer cAMP responsiveness to the insulin CRE.
The Mutations in Mutants 1 to 4 Alter CREB Binding-It has been shown before that the transcription factor CREB is not detected among the nuclear proteins that bind to the labeled insulin CRE in the electrophoretic mobility shift assay (22). This is confirmed in Fig. 7A showing that the addition of an antiserum directed against CREB to the binding reaction had no effect on nuclear protein binding to the insulin CRE (compare lanes 5 and 6). The anti-CREB antiserum did, however, abolish the binding of complexes A and B to the labeled somatostatin CRE used as a positive control (Fig. 7A, compare lanes  1 and 2). The somatostatin CRE is a well characterized high affinity CREB-binding site (1-3). The base substitution in mutant 3 converts the CRE octamer-like sequence of the insulin CRE (TGACGTCC) into a perfect CRE octamer consensus sequence (TGACGTCA) (Fig. 1). It was noted above that two new protein complexes were detected that bind to labeled m3-In-sCRE but not to labeled InsCRE and that comigrate with complexes A and B of the labeled somatostatin CRE (see Figs. 2A, 3A, and also 7A). As shown in Fig. 7A, binding of these complexes to m3-InsCRE was abolished by the addition of anti-CREB antiserum (compare lanes 3 and 4). This shows that the mutation in m3-InsCRE increases the affinity of CREB such that the binding of proteins with CREB-like immunoreactivity becomes detectable in the electrophoretic mobility shift assay with labeled m3-InsCRE. This offers an explanation for the marked cAMP responsiveness of m3-InsCRE (Fig. 5).
Weak binding of cellular CREB or closely related proteins to the insulin CRE can be detected by the electrophoretic mobility shift assay in competition experiments (22). This is shown in Fig. 7B. The binding of nuclear proteins with CREB-like immunoreactivity to the labeled somatostatin CRE (complexes A and B) is competed away by unlabeled somatostatin CRE and at a higher molar excess also by the insulin CRE (Fig. 7B). Mutant 1 had lost CREB binding (Fig. 7B, compare lanes 6 and  4), and mutant 2 competed somewhat less efficiently for complexes A and B than InsCRE wild-type sequence (Fig. 7B,  compare lanes 9 and 4). Mutant 4 was a stronger competitor for CREB binding than InsCRE wild-type sequence (Fig. 7B, compare lanes 10 and 3). These results show that the mutations in mutants 1 to 4 of the insulin CRE change the affinity of the binding to CREB or closely related proteins; mutant 2 and, more so, mutant 1 show a decrease in binding whereas mutants 4 and 3 show an increase in binding.
Functional Significance of CREB and NF-Y Binding to the Insulin CRE as Assessed by Using Dominant Negative Mutants-The binding of CREB or closely related proteins to the insulin CRE, although weak, could be functionally significant. To further explore this possibility co-transfection experiments were performed using an expression vector encoding a dominant negative mutant of CREB, KCREB (39). Whereas forskolin stimulated insulin CRE-mediated transcription in controls, forskolin had no stimulatory effect when in addition KCREB was overexpressed (Fig. 8A), indicating that CREB or closely related proteins are required for cAMP responsiveness of the insulin CRE. Forskolin slightly decreased activity in the presence of KCREB (Fig. 8A), which is similar to what was found using m1-and m2-InsCRE (Fig. 5) and remains unexplained. When taken together with the results from the mutational analysis of the insulin CRE showing that mutations to the insulin CRE that increase CREB binding enhance cAMP responsiveness, whereas mutations to the insulin CRE that decrease CREB binding prevent cAMP responsiveness (see above), the present data strongly suggest that cAMP responsiveness is conferred to the insulin CRE by CREB or closely related proteins.
To further explore the role of NF-Y, a dominant negative mutant of NF-Y, NF-YA29 (40), was used. When increasing amounts of an expression vector encoding NF-YA29 were cotransfected with the insulin CRE reporter fusion gene, basal activity conferred by the insulin CRE (11.7 Ϯ 1.2 relative to the minimal promoter) was decreased to 8.9 Ϯ 0.6 (co-transfection of 0.025 g/dish of the NF-YA29 plasmid), 8.0 Ϯ 0.8 (0.150 g), 7.0 Ϯ 1.0 (0.5 g), and 3.9 Ϯ 0.1 (3 g) (n ϭ 6 each). These results independently confirm the conclusion drawn from the results of the mutational analysis of the insulin CRE and from the transcriptional activity of E␣-CAAT (see above) that NF-Y binding confers basal activity to the insulin CRE. Co-transfection of increasing amounts of the NF-YA29 expression vector enhanced cAMP responsiveness of the insulin CRE (Fig. 8B). While forskolin stimulated insulin CRE-mediated transcription 3.5-fold in controls, forskolin stimulated transcription up to 11-fold in the presence of the dominant negative NF-Y mutant (Fig. 8B). This effect was specific, because co-transfection of increasing amounts of the NF-YA29 expression vector had no effect on the stimulation by forskolin of transcription mediated by m4-InsCRE which lacks NF-Y binding (Fig. 8B). These data suggest that NF-Y binding to the insulin CRE not only confers basal activity but also decreases cAMP responsiveness.
It has been shown before that the insulin CRE does not respond to membrane depolarization and calcium influx, in contrast to the CREs of the rat glucagon and rat somatostatin genes that bind CREB with higher affinity (22). This is confirmed in Fig. 8C which shows that high potassium-induced membrane depolarization did not stimulate insulin CRE-mediated transcription. When the binding of NF-Y to the insulin CRE was inhibited by either mutation of the insulin CRE (m4-InsCRE reporter) or by overexpression of the dominant negative NF-Y mutant, membrane depolarization stimulated transcription (Fig. 8C). The stronger depolarization responsiveness of m4-InsCRE as compared with InsCRE in the presence of NF-YA29 (Fig. 8C) could be explained by a better binding of CREB to m4-InsCRE than to InsCRE or, alternatively, by the assumption that overexpression of NF-YA29 inhibits NF-Y binding to the insulin CRE less efficiently than the mutations in mutant 4. These data suggest that NF-Y binding to the insulin CRE not only confers basal activity and decreases cAMP responsiveness but also inhibits depolarization responsiveness.
Role of CREB and NF-Y in the Regulation of the Intact Insulin Promoter-Within the insulin promoter (from Ϫ410 to ϩ49) 4 bases of the insulin CRE octamer-like sequence were selectively deleted (construct Ϫ410(⌬Ϫ183/Ϫ180)InsLuc, Fig.  9A) which abolishes CREB and NF-Y binding (see above). This internal promoter deletion decreased basal activity by 40% and abolished transcriptional activation by cAMP (Fig. 9A) indicating that insulin CRE-binding proteins contribute to basal activity and are required for the stimulation by cAMP of insulin promoter activity. These results extend previous mutational analyses (23,26). Co-transfection of expression vectors encoding the dominant negative mutants KCREB or NF-YA29 together with the wild-type insulin promoter construct (Ϫ410In-sLuc) inhibited or enhanced, respectively, cAMP-induced insulin gene transcription (Fig. 9B), suggesting a role of CREB and NF-Y in the regulation also of the intact insulin promoter. DISCUSSION In the present study the insulin CRE-binding proteins and their function were characterized. It is shown that in addition to CREB, which interacts with low affinity with the CRE octamer-like sequence, the transcription factor NF-Y and winged helix-like proteins bind to overlapping sites within the insulin CRE. This combination of CRE-binding proteins forms a composite CRE with gene-specific properties. Cyclic AMP responsiveness is mediated by CREB; however, NF-Y binding to the insulin CRE confers basal activity and modulates the function of CREB. The data offer an explanation why the insulin CRE has considerable basal activity but is less responsive to cAMP stimulation than others.
The winged helix protein HNF-3␤ is expressed in insulin-(56) and glucagon-producing islet cell lines (57,58,61) and regulates glucagon gene transcription through a binding site within the G2 element of the glucagon gene (57,61). The winged helix-like proteins binding to the insulin CRE are distinct from HNF-3␤. Although it is not excluded that they enhance responsiveness of the insulin CRE, the present study does not establish a role for these proteins and, thus, their identity and functional significance remains to be shown. Islet   (45) and is confirmed by cDNA cloning. 2 The cross-competition between the insulin CRE and G3B indicates that the winged helix-like proteins binding to the insulin CRE are the same as those binding to domain B of the G3 enhancer element of the glucagon gene. On this element the winged helix-like proteins synergize with the paired domain transcription factor Pax6 binding to an adjacent site and confer in this combination cell-specific transcriptional activity (41,45,(62)(63)(64).
Several lines of evidence strongly suggest that NF-Y binds to the insulin CRE. First, the insulin CRE contains a CCAAT box motif with high similarity to an NF-Y consensus site. Second, this CCAAT box motif is recognized by a nuclear protein (complex 1). Third, complex 1 binding is not competed for by binding sites of the CAAT box-binding proteins CTF/NF-I or C/EBP, but is competed for by several NF-Y binding sites. Fourth, there is cross-competition for comigrating complexes between the insulin CRE and a well characterized NF-Y-binding site, E␣-CAAT. Finally, complex 1 is recognized by an antiserum directed against NF-YB. NF-Y (also called CBF and CP1) is a ubiquitous CCAAT box-binding protein and extremely conserved in evolution (40,53,65,66). Like the TATA box, CCAAT boxes are widespread promoter elements typically located between Ϫ60 and Ϫ100 base pairs 5Ј to the start of transcription, suggesting a role in basal transcription (67). Most CCAAT boxes located at this position are NF-Y-binding sites (67). NF-Y has been shown to interact with the TATA box-binding protein (68), to be important for transcription re-initiation (43), and to facilitate in vivo recruitment of upstream DNA binding transcription factors (69). Thus, NF-Y may have a general role in the assembly of proximal promoter complexes. Although, in general, NF-Y is constitutively active, it has been implicated directly or indirectly in the regulation of some promoters by, for example, heme (70), sterol (71,72), calcium depletion (73), and the viral Tax protein (74). In the insulin promoter the NF-Ybinding CCAAT box is located further upstream at Ϫ179 and falls within an element that confers cAMP responsiveness. The CCAAT box motif has been implicated in the cAMP inducibility of several genes (33)(34)(35)(36). Some of these CCAAT boxes have been shown to bind NF-Y. Boularand et al. (36) reported that an inverted CCAAT box motif at Ϫ67 of the human tryptophan hydroxlase gene promoter was required and sufficient for cAMP induciblity in pinealocytes. This inverted CCAAT box in the mouse gene has been shown to bind NF-Y (75). In addition, it has been demonstrated that a CCAAT box that binds NF-Y can mediate induction by cAMP of the rat hexokinase II gene in L6 myotubes (35). However, the present study clearly demonstrates that NF-Y does not confer cAMP responsiveness to the insulin CRE in islet cells.
Instead, cAMP responsiveness seems to be conferred by CREB. In the absence of other proteins, CREB binds with low affinity to the insulin CRE (Ref. 22 and this study). This may largely be due to the deviation in position 8 of the core octamerlike motif of the insulin CRE (TGACGTCc) from the consensus sequence (TGACGTCA) rather than to the bases flanking the octamer in the insulin CRE. This is suggested by mutant 3 which bound CREB much more strongly than the wild-type or mutant 4. The x-ray crystal structure of the AP-1⅐DNA complex indicates that the corresponding position in the AP-1 site is recognized on the opposite strand by alanine and cysteine residues of Fos and Jun that are conserved in CREB (76). Mutations of the insulin CRE that altered CREB binding changed cAMP responsiveness correspondingly. In addition a dominant negative mutant of CREB abolished cAMP inducibility of the insulin CRE. Thus, although CREB binds weakly to the insulin CRE, and although the insulin CRE contains a CCAAT box motif that has been shown to mediate cAMP inducibility in other genes (see above), the results of the present study, when taken together, strongly suggest that the cAMP responsiveness of the insulin CRE is conferred by CREB or a closely related protein.
Whereas NF-Y does not confer cAMP responsiveness, the 2 S. Herzig and W. Knepel, unpublished observation. present results define a role for NF-Y in insulin CRE-mediated transcription. Insulin CRE mutants that lack NF-Y binding still responded to cAMP but had lost basal activity. An NF-Ybinding site alone was not cAMP responsive but did confer basal activity in islet cells. NF-Y is a heterotrimeric protein and the assembly of the NF-Y subunits follows a specific pathway. NF-YB and NF-YC associate with each other to form a binary complex which then interacts with NF-YA. Formation of this ternary complex is required for binding to DNA (40,53,65,66,77). Mutants of NF-Y subunits, such as NF-YA29 (40), that still interact with the other subunits but do not bind DNA act as dominant negative mutants (40,66). Using this technique, the present study shows that inactivation of NF-Y in islet cells decreases basal activity of the insulin CRE while it enhances cAMP responsiveness. Thus, these data indicate that NF-Y modulates the function of CREB. NF-Y decreases the ability of CREB to mediate cAMP-induced gene transcription. It also inhibits depolarization responsiveness, and through NF-Y the insulin CRE gains constitutive basal activity that is conferred to the insulin promoter (22,23,26). The interaction of CREB with NF-Y at the insulin CRE could include direct proteinprotein interaction. The transcription factor YY1, that binds immediately downstream of the CRE of the mouse c-fos promoter, interacts directly with CREB, and this interaction has been suggested to lead to repression of CRE-dependent transcription (17). A direct interaction between CREB and NF-Y remains to be shown. However, as the basis of functional synergism, evidence has already been presented for complex formation between NF-Y and other transcription factors such as C/EBP proteins on the serum albumin promoter (78) and ATF-2 on the fibronectin promoter (79). Alternatively, CREB and NF-Y could compete for binding to the insulin CRE as their binding sites overlap. This is reminiscent of CREB-binding sites which overlap with binding sites for the glucocorticoid receptor in the human glycoprotein hormone ␣-subunit gene (14) or for C/EBP␤ and related proteins in the phosphoenolpyruvate carboxykinase gene (15, 18 -21). In these cases a specific negative regulation by glucocorticoids (14) and a tissuedependent activation by cAMP (15, 18 -21), respectively, is assumed to be achieved through these overlapping sites.
To maintain glucose homeostasis, insulin gene transcription appears to be regulated synergistically by glucose metabolism and the hormonally regulated cAMP pathway (24,25). Studies on many genes have taught that promoter activity depends on a synergistic interaction between multiple promoter-binding proteins. Transcriptional activation in response to extracellular signals thus involves the assembly or multiprotein complexes on enhancers and promoters induced by regulated transcription factors through interaction with other proteins. Consistent with this view, several specific cooperative effects of CREB have been shown including synergistic interactions with HNF-4 in the tyrosine aminotransferase gene (80), liver-enriched transcription factors in the phosphoenolpyruvate carboxykinase gene (15), and steroidogenic factor-1 in the aromatase CYP19 gene (81). Some of the insulin CRE-binding proteins identified in the present study are therefore likely to interact with constitutive or glucose-regulated transcription factors of the insulin promoter. In the absence of glucose, stimulation of insulin gene transcription by cAMP is modest. Together with the nonpalindromic sequence of the CRE octamer motif, the binding of NF-Y to the insulin CRE that inhibits the ability of CREB to mediate cAMP-induced transcription reported in our studies may be wholly or partially responsible for this modest response.