Role of Tissue-Specific Transcription Factor LFB3 in a Cyclic AMP-Responsive Enhancer of the Urokinase-Type Plasminogen Activator Gene in LLC-PK₁ Cells

Abstract

A cyclic AMP (cAMP)-inducible enhancer in the pig urokinase-type plasminogen activator gene located 3.4 kb upstream of the transcription initiation site is composed of three protein-binding domains, A, B, and C. Domains A and B each contain a CRE (cAMP response element)-like sequence but require the adjoining C domain for full cAMP responsiveness. A tissue-specific transcription factor, LFB3/HNF1β/vHNF1, binds to the C domain. Mutation analyses suggest that the imperfect CRE and LFB3-binding sequences are required for tight coupling of hormonal and tissue-specific regulation. CREB and ATF1 bind to domains A and B, and this binding is enhanced upon phosphorylation by cAMP-dependent protein kinase (protein kinase A [PKA]). Analysis in a mammalian two-hybrid system revealed that CREB/ATF1 and LFB3 interact and that transactivation potential is enhanced by PKA activation. Interestingly, however, phosphorylation of CREB at Ser-133 does not contribute to its interaction with LFB3. The region of LFB3 involved in its interaction with CREB/ATF1 lies, at least partly, between amino acids 400 and 450. Deletion of this region removed the ability of LFB3 to mediate cAMP induction of the ABC enhancer but did not impair its basal transactivation activity on the albumin promoter. Thus, the two activities are distinct functions of LFB3.

Many proteins are phosphorylated by cyclic AMP (cAMP)-dependent protein kinase (protein kinase A [PKA]), leading to the initiation of various cellular events (reviewed in reference 48). Several transcription factors, including multiple variants of CREB (44, 49), CREM (11), and ATF1 (27, 41), have been implicated as nuclear targets for PKA. CREM and ATF1 are highly related to CREB and able to dimerize with it (11, 17). All mediate transcriptional modulation by binding to cAMP response element (CRE) sites (TGACGTCA). CREB is the best-studied CRE-binding protein and is the basis of current concepts of CRE regulation (reviewed in reference 33). Two signals are essential for the activation of a target gene by CREB: a phospho-Ser-133-dependent interaction of CREB with RNA polymerase II via the coactivator CBP and a glutamine-rich-domain interaction with TFIID via TAF_II130 (7, 12, 36). In most cases, the CRE site is present in the promoter not as an isolated, single copy but usually as a multimer or in close proximity to other cis-acting elements (10, 38, 43). The additional cis elements may recruit other transcription factors, which are often not ubiquitously expressed and, therefore, provide the basis for tissue-specific gene expression (reviewed in reference 26).

We have studied urokinase-type plasminogen activator (uPA) gene regulation in LLC-PK₁ cells, a cell line derived from pig kidney epithelia (28, 31, 47). In these cells, uPA gene expression is regulated by many independent signaling pathways induced, e.g., by cAMP (47), cytoskeletal reorganization (5, 18, 25), tumor promoter phorbol esters (25), or okadaic acid (24, 35). The cAMP-dependent induction of the uPA gene is cell specific, as cAMP is unable to induce the uPA gene in U937 (6), HeLa (31) or F9 (28) cells. The pig uPA gene has a cAMP-responsive enhancer located 3.4 kb upstream of the transcription initiation site and composed of three domains, A, B, and C. Both the A domain and the B domain contain a CRE-like sequence (TGACG), which is essential for cAMP induction, but require the adjoining C domain to confer full cAMP inducibility on a heterologous promoter (31, 47). The C domain is distinct from the A and B domains in that it contains no CRE. We have purified the protein binding to the C domain and found it to be the pig equivalent of LFB3 (31), also known as vHNF1 (3) or HNF1β (30), which is highly expressed in kidney. LFB3 is related to the liver-specific transcription factor HNF1α (also termed LFB1); both factors recognize the same consensus HNF1 sequence in vivo and in vitro (9). The C domain sequence contains two imperfect HNF1 recognition sequences (47).

In the present study, we characterized the ABC enhancer and the role of LFB3 in the coupling of tissue-specific and hormonal gene regulation.

MATERIALS AND METHODS

Reagents.

12-O-Tetradecanoylphorbol-13-acetate (TPA) was obtained from Sigma, 8-bromo-cAMP (Br-cAMP) was obtained from Boehringer Mannheim, and [α-³²P]dATP (3,000 Ci/mmol) was obtained from Amersham. The oligonucleotides used for electromobility shift assays (EMSA) were as follows (only upper strands are given): domain A, 5′-AATTCTGTGCCTGACGCACAG-3′; domain B, 5′-AATTCGCCCATGACGAACACTGGG-3′; domain C, 5′-GTGAATGAATAAAGGAATAAATGAATGATTTCACA-3′; somatostatin CRE, 5′-AATTCGCCTCCTTGGCTGACGTCAGAGAGAGAG-3′.

Cell culture.

LLC-PK₁ (16), 293, and COS-1 cells were cultured in Dulbecco’s modified Eagle’s medium (Gibco-BRL) supplemented with 10% (vol/vol) fetal calf serum (AMIMED), 0.2 mg of streptomycin per ml, and 50 U of penicillin per ml at 37°C in a humidified CO₂ (5%) incubator. NIH 3T3, F9, and HeLa cells were cultured in Dulbecco’s medium supplemented with 5% calf serum (NIH 3T3), 5% fetal calf serum (F9), or 3% fetal calf serum and 3% newborn calf serum (HeLa). All cell lines were plated directly on plastic dishes except F9 cells, which were plated on gelatin-coated plastic dishes.

Plasmids.

In pTATA (provided by A. E. Sippel), the firefly luciferase gene is linked to a minimal promoter of the thymidine kinase gene (−46 to +52) containing only the TATA box and the transcription initiation site. Derivatives of pTATA containing the entire sequence or parts of the ABC region at the AccI-BamHI sites immediately 5′ of the TATA box have been described previously (28). The sequences used to construct these derivatives are shown in Fig. 1A. In p3AP1-TATA, three tandem repeats of the collagenase AP1 element were inserted at the same AccI-BamHI sites of pTATA (5′-cgacCGGCTGACTCATCACGGCTGACTCATCACGGCTGACTCATCAa-3′). In pCRE-TATA, the rat somatostatin CRE sequence was used (5′-cgacCGCCTCCTTGGCTGACGTCAGAGAGAGAGTTTa-3′). Similar constructs with nonmutated sequences, but with the simian virus 40 early gene promoter, and the pig LFB3 expression vector have been described previously (31). pALB-luc was derived by subcloning the 190-bp HindIII-BglII fragment of the rat albumin proximal promoter from pALB-cat (15), which was provided by M. Yaniv, into HindIII-BglII sites of pGL2-basic (Promega). pBGO-ATG-CREB and pBGO-ATG-ATF1 were constructed by cloning the respective cDNAs amplified with PCR into pBGO-ATG (46), provided by P. Matthias as SmaI/XbaI fragments. The following primers were used for PCR: ATF13′, 5′-CGTCTAGACTTTCTTAGGAATCAAACAC-3′; ATF15′, 5′-ATGGAAGATTCCCACAAGAGTACCAC-3′; CREB3′, 5′-CGTCTAGATAATCTGATTTGTGGCAGTAAAGG-3′; and CREB5′, 5′-ATGACCATGGAATCTGGAGC-3′.

FIG. 1.

Significance of the deviation of protein-binding sequences in the ABC enhancer from respective consensus sequences. (A) Wild-type and mutated sequences derived from the ABC enhancer located at −3.4 kb of the pig uPA promoter were cloned in front of a reporter gene. The mutated sequences are shown in lower case (in reverse type), the half-palindromic CREs are boxed, and domains A, B, and C are separated by vertical lines. (B) Different constructs were transiently transfected into LLC-PK₁ cells; 16 h later, cells were induced by 1 mM Br-cAMP for 6 h and luciferase activity was measured. (C) pABC-TATA and pABHNF-TATA were transiently transfected into F9 cells with or without expression vectors for the catalytic subunit of PKA (CEV) and LFB3. After 16-h transfection, cells were collected and luciferase activity was measured.

To prepare an expression vector for His-tagged LFB3, the EcoRI-BamHI PCR fragment of the LFB3 coding region was subcloned into XhoI/BamHI-digested pMT-PKA (51), where the EcoRI end of the fragment and the XhoI end of the vector had been blunt ended by using the DNA polymerase I Klenow fragment. pMT-PKA was provided by T. Wirth.

Different Gal4-LFB3 mutants were constructed by cloning the respective PCR fragments, digested by EcoRI/XbaI, into the vector pM (Clontech) in frame at the C terminus of the Gal4 DNA-binding domain (amino acids [aa] 1 to 147). The following primers were used for PCR: for LBF3(1-314) (Gal4-A), M5 (5′-CGGAATTCATGGTGTCCAAGCTCACG-3′) and M6 (5′-CGTCTAGATCACAGCTTCTGCCGGAACGC-3′); for LFB3(315-559) (Gal4-B, M7 (5′-CGGAATTCGCCATGGACGCCTACAGC-3′) and M8 (5′-CGTCTAGATCACCAGGCTTGGAGAGG-3′); for LFB3(394-559) (Gal4-C), M8 and M9 (5′-CGGAATTCCACAATCTCCTCTCACCTG-3′); for LFB3(477-559) (Gal4-D), M8 and M10 (5′-CGGAATTCTCCCAGCAGCTGCACAGC-3′); for LFB3(315-393) (Gal4-E), M7 and M11 (5′-GCTCTAGATCAGCCCGGGTCCAGGCTGGC-3′); for LFB3(315-476) (Gal4-F), M7 and M12 (5′-GCTCTAGATCAGAACTGCACGGGCTGCAG-3′). The full-length Gal4-LFB3 was constructed by subcloning the EcoRI/XbaI fragment prepared from pcDNA3-LFB3 into the pM vector. pVP16 (Clontech) and PCR fragments containing the coding sequences of CREB and ATF1 were used to generate plasmids expressing VP16-CREB and VP16-ATF1 fusion proteins. PCR products were subcloned into pVP16 as EcoRI/XbaI fragments. Primers used for PCR were CrebEcoRI5′ (5′-CGGAATTCATGACCATGGAATCTGGAGC-3′), CREB3′ (5′-CGTCTAGATAATCTGATTTGTGGCAGTAAAGG-3′), ATF1EcoRI5′ (5′-CGGAATTCATGGAAGATTCCCACAAGAGTACCAC-3′), and ATF13′ (5′-CGTCTAGACTTTCTTAGGAATCAAACAC-3′).

To change the codon for Ser-133 (TCC) in CREB into Ala (GCC), we applied the QuickChange site-directed mutagenesis kit (Stratagene) using two complementary oligonucleotides bearing the following mutation (underlined) in the forward oligonucleotide: 5′-CCTTTCAAGGAGGCCTGCCTACAGGAAAATTTTG-3′. The resulting plasmid was called pVP16-CREB (S133A). pVP16-CREB and pVP16-CREB (S133A) were further used to generate plasmids expressing a CREB mutant lacking 71 C-terminal amino acids corresponding to the basic leucine zipper region. For this purpose, pVP16-CREB and pVP16-CREB(S133A) were first digested with HinfI, followed by blunt-end formation using the DNA polymerase I Klenow fragment, and then digested with EcoRI to recover deleted CREB cDNA. The fragments were then cloned into pVP16 digested with EcoRI/SmaI, resulting in pVP16-CREBΔ and pVP16-CREBΔ(S133A), respectively.

To generate a vector expressing LFB3 tagged with the hemagglutinin (HA) epitope at its N terminus, we first ligated a KpnI/EcoRI double-stranded oligonucleotide containing the sequence encoding the HA epitope (sense strand oligonucleotide, 5′-CCCACCATGGCTTACCCATACGATGTTCCAGATTACGCTG-3′) into KpnI/EcoRI-digested pcDNA3 (Invitrogen), resulting in pcDNA3-HA. Subsequently, we cloned LFB3 cDNA into pcDNA3-HA using EcoRI/XbaI restriction sites. pcDNA3-HALFB3Δ(400-450) was constructed by ligating two LFB3 PCR fragments comprising the sequences coding for aa 1 to 400 and 450 to 559 into pcDNA3-HA digested with EcoRI/XbaI. The following primers were used for PCR amplification: for LFB3(1-400), M5 and 3del2a (5′-AGGTGAGAGGAGATTGTG-3′); for LFB3(450-559), M8 and 5del2a (5′-AACTCCTCCCAAGCTCAG-3′). An expression vector for His-tagged LFB3Δ(400-450) was constructed by linking a PCR fragment generated from pcDNA3-HALFB3Δ(400-450) with M5 and M8 as primers to pQE-30 (Qiagen). All constructs were verified by sequencing.

The expression vector for the catalytic subunit of PKA (pCEV) was provided by S. McKnight, the pJ6-ATF1 expression vector was provided by P. Verde, and the pmcCREB vector was provided by G. Schütz.

Transient transfection assays.

For transient transfection experiments, cells were seeded and transfected as described elsewhere (28). In brief, cells were transfected by calcium phosphate-mediated precipitation and treated 20 h later with 1 mM Br-cAMP for 6 h. Where cells were induced by coexpression of the catalytic subunit of PKA, they were collected after 20 h of transfection. Cell extracts were assayed for luciferase activity as described elsewhere (31) by using a luminometer (Autolumat LB 953; Berthold) or for chloramphenicol acetyltransferase (CAT) activity by using a CAT-ELISA assay kit (Boehringer Mannheim).

Nuclear extracts and EMSA.

Nuclear extracts were prepared from LLC-PK₁ cells, and EMSA were performed as previously described (25). Oligonucleotide probes were radiolabeled by using the DNA polymerase I Klenow fragment and [α-³²P]dATP. Supershift experiments using anti-ATF1 (FI-1; Santa Cruz), anti-CREB (X-12; Santa Cruz), or anti-glutathione S-transferase (Z.5; Santa Cruz) antibodies were performed as described elsewhere (8). Briefly, nuclear extracts were incubated for 4 h at 4°C with 2 μl of the appropriate antiserum prior to the addition of the radioactive probe. Thereafter, samples were run on a 0.25× Tris-borate-EDTA gel at 200 V for 2 h and exposed to Kodak X-Omat AR film with an intensifying screen at −70°C.

EMSA-Western analysis.

Proteins binding to specific DNA sequences were detected by a method which is a composite of EMSA and Western blot (39). Nuclear extracts (25 μg of protein) from LLC-PK₁ cells were incubated with 10 pmol of nonradioactive double-stranded oligonucleotides corresponding to domain A or B or somatostatin CRE for 30 min at room temperature in binding buffer as above. Samples were run on 0.25× Tris-borate-EDTA at 200 V for 2 h. Proteins were blotted and analyzed for CREB/ATF1 with anti-human CREB rabbit antibody or anti-human ATF1 sheep antibody (Upstate Biotechnology).

Coprecipitation assays.

Recombinant LFB3 and LFB3Δ(400-450) proteins with six N-terminal histidine residues were prepared by transformation of BL21(DE3)pLysS and M15 bacteria (Qiagen), respectively, with the corresponding expression vectors and subsequently purified as described elsewhere (50). In vitro-translated ATF1 and CREB were obtained by use of pBGO-ATG-ATF1 or pBGO-ATG-CREB, respectively, as template in the TNT reticulocyte lysate system (Promega). Ten microliters of in vitro-translated [³⁵S]methionine-labeled ATF1 or CREB was incubated with His-tagged LFB3 or LFB3Δ(400-450) protein coupled to Ni-chelating Sepharose (Pharmacia) in Dignam buffer D (20 mM HEPES [pH 7.9], 20% glycerol, 100 mM KCl) for 2 h at 4°C. Samples were then washed three times with 20 mM HEPES (pH 7.9)–20% glycerol–500 mM KCl–0.5% Triton X-100–10 mM imidazole and once with TTBS (10 mM Tris HCl [pH 7.6], 50 mM NaCl, 0.2% Tween 20), eluted in sodium dodecyl sulfate (SDS) sample buffer, and subjected to SDS-polyacrylamide gene electrophoresis. Radioactive bands were visualized by fluorography.

RESULTS

Coupling of hormonal and tissue-specific gene regulation.

To characterize the ABC enhancer, we introduced a series of mutations (Fig. 1A) and cloned these mutated enhancers in front of a luciferase reporter gene with a minimal thymidine kinase promoter. The sequences in the ABC domains responsible for cAMP induction are imperfect CRE and HNF1-binding sequences. First, we examined the significance of their deviation from the consensus CRE and HNF1-binding sequences. When the imperfect CRE sequence was converted to the consensus CRE sequence, mutated AB domains (termed A*B*) mediated cAMP inducibility even in the absence of the adjoining C domain. The conversion of the domain C sequence to the consensus HNF1-recognition sequence (termed ABHNF) led to a strong increase in basal promoter activity (Fig. 1B).

Involvement of LFB3 in the activity of the ABC enhancer was confirmed by cotransfection in F9 cells lacking endogenous LFB3. We used the catalytic subunit of PKA to induce these cells, since undifferentiated F9 cells are not responsive to cAMP (17). As shown in Fig. 1C, cotransfection of an LFB3 expression vector had no effect on the ABC enhancer unless there was a cAMP signal but strongly enhanced luciferase expression from the ABHNF-driven promoter without a cAMP signal. These results suggest that the imperfect CRE sequences in domains A and B and the imperfect HNF1-binding sequences in domain C are required for tight coupling of hormonal and tissue-specific regulation; the ABC enhancer is active only in the presence of both a cAMP signal and LFB3.

The ABC enhancer does not mediate TPA-dependent gene induction.

The CRE-like elements in the A and the B domains differ in only 1 and 2 nucleotides, respectively, from the consensus AP1 site TGA(C/G)TCA, also called the TPA response element (TRE). This sequence similarity prompted us to ask if the wild-type ABC enhancer or the mutated ABC enhancer AP1C (Fig. 1A), where CRE-like sequences in the AB domains are converted to the consensus AP1 sequence, could mediate TPA induction. As shown in Fig. 2, TPA treatment led to only modest induction (1.5-fold) from the ABC or AP1C construct. This induction is comparable to that for the empty vector pTATA and, therefore, not significant. In contrast, a construct with three AP1-binding sites was approximately threefold inducible. Thus, the ABC enhancer cannot mediate TPA induction. Conversion of CRE-like elements into perfect AP1 sites did not make the cAMP-inducible enhancer TPA inducible. This suggests that LFB3 cooperates specifically with the AB-binding proteins but not with AP1 factors.

FIG. 2.

Specificity of LFB3 cooperation. LLC-PK₁ cells were transfected with different reporter constructs and 16 h later induced either with 1 mM Br-cAMP or with 100 ng of TPA per ml for 6 h. Cells were then collected, and luciferase activity was measured. pAP1-TATA contains three copies of the AP1 sequence from the collagen gene. In the pAP1C-TATA construct, CRE-like sequences were converted to AP1. For details of these constructs, see Materials and Methods. Luciferase activity was measured as for Fig. 1C, and values are plotted as fold induction relative to noninduced cells.

ATF-1 and CREB interact with the AB domains.

Several transcription factors have been reported to mediate cAMP-dependent gene activation, of which the best studied so far are CREB and ATF1. Using recombinant CREB protein, Nichols et al. (37) showed that CREB can bind both A and B domains and that the binding is enhanced by treatment with PKA. We examined the possible involvement of these proteins by testing the effects of antibodies against these molecules in EMSA using nuclear extracts from LLC-PK₁ cells. As shown in Fig. 3A, oligonucleotides of domain A and B sequences, like that of the somatostatin CRE sequence, gave rise to single bands which could be competed by excess cold oligonucleotide of the same sequence. Inclusion of anti-CREB-specific antiserum in the binding reaction with domain A and B oligonucleotides produced a band slower than the main band. Inclusion of anti-ATF1-specific antiserum produced two slow-migrating bands. In contrast, a nonspecific antiserum did not change the migration pattern. This suggests that the retarded bands produced by the anti-CREB- and anti-ATF1 antisera are specific and that CREB and ATF1 can bind both A and B domains. The two retarded bands produced by the anti-ATF1-antiserum may indicate different ATF1-containing complexes: an ATF1 homodimer and a heterodimer with another protein. To further confirm that CREB and ATF1 bind to both A and B domains, we performed EMSA-Western blot analysis (39). As shown in Fig. 3B, both CREB and ATF1 were detected in fractions that bound to domains A and B as well as somatostatin CRE.

FIG. 3.

Involvement of ATF1 and CREB in ABC enhancer-mediated cAMP induction. (A) EMSA. Nuclear extracts (NE) prepared from LLC-PK₁ cells were tested for binding to oligonucleotides containing sequences from the somatostatin promoter CRE region or from domain A or B. Radioactive oligonucleotides indicated at the bottom of each panel were incubated without (−) or with nuclear extracts plus anti-ATF1 (α-ATF1) antibody, anti-CREB antibody, nonspecific antiserum (immunoglobulin G) (ns. IgG), or a 100-fold excess of cold oligonucleotide of the same sequence. Reaction products were fractionated in nondenaturing acrylamide (5%) gel electrophoresis and exposed in a PhosphorImager or to X-Omat X-ray film with an intensifying screen at −70°C for 4 days for the A and B probes and 3 days for the somatostatin CRE probe. (B) Detection of CREB and ATF1 by blotting EMSA gels. Nuclear extracts (25 μg of protein) from LLC-PK₁ cells were incubated in the absence of oligonucleotide (control) or with oligonucleotides corresponding to domain A or B or somatostatin CRE or a nonspecific oligonucleotide (ns) and separated by nondenaturing gel electrophoresis. Proteins were analyzed by immunoblotting with polyclonal antisera to CREB and ATF1. Only regions corresponding to the main complex are shown. (C) Comparisons of different cell lines for ABC enhancer-mediated cAMP induction. Cell lines were transfected with 1 μg of pABC-TATA together with 0.5 μg (or 0.1 μg for COS-1 and NIH 3T3 cells) of pRSV-LFB3 and/or 0.5 μg of pCEV. After 16-h transfection, cells were collected and luciferase activity was measured. Assays were done in duplicate, and mean values are shown with error bars.

To date, ATF1 has been found in all cell lines examined (40, 41). We examined ABC enhancer-mediated cAMP induction in NIH 3T3, COS-1, and 293 cell lines; 293 has been shown to express ATF1 (41). In all cell lines, the ABC enhancer was inducible when the catalytic subunits of PKA and LFB3 were cotransfected, suggesting that the proteins binding to domains A and B are ubiquitous (Fig. 3C).

Binding of ATF1 and CREB to domains A and B is regulated by PKA-mediated phosphorylation.

We addressed the role of the phosphorylation of ATF1 and CREB by PKA with respect to their affinity for CRE sites. Nuclear extracts from LLC-PK₁ cells were treated with PKA, and DNA-binding activity was measured by EMSA in the presence of specific antisera against CREB and ATF1 (Fig. 4). Treatment with PKA resulted in increased DNA-binding activities of CREB and ATF1, suggesting that the binding of each protein is increased upon phosphorylation. In contrast, binding of ATF1 or CREB to consensus CRE was not affected by PKA treatment. This result is in agreement with previous reports that asymmetric CREs are only weakly bound by CREB unless they are phosphorylated, whereas symmetrical consensus CREs also have a high affinity for nonphosphorylated forms of CREB (37).

FIG. 4.

Increased DNA-binding activity of nuclear extracts (NE) from LLC-PK₁ cells by PKA treatment. Nuclear extracts were treated where indicated with 100 U of the catalytic subunit of PKA for 90 min at 37°C. Afterwards, samples were incubated with or without 2 μg of the indicated antibodies (Ab) for 4 h at 4°C and assayed for DNA-binding activity as described in the legend to Fig. 3. The gels were dried and exposed to X-Omat AR film for 1 day for the somatostatin CRE probe and 4 days for the domain B probe.

Interaction between ATF1 or CREB and LFB3.

Next, we examined the nature of the cooperation between the AB and C domains by inserting between them 5 or 10 nucleotides, corresponding to a half or a full turn, respectively, of the DNA helix (Fig. 1A). The insertion of 5 nucleotides (AB5C) reduced the inducibility to about 30% of the wild type (Fig. 5). Insertion of 10 nucleotides (AB10C) reduced the inducibility even further. These results suggest that the proximity and angular orientation of these proteins, LFB3 and the proteins binding to the AB domains, are important for their functional cooperation in mediating cAMP induction and indicate some form of physical interaction.

FIG. 5.

Interaction of LFB3 with CREB and ATF1. Effects of the insertion of 5 or 10 nucleotides between domains A, B, and C on ABC enhancer-mediated cAMP inducibility. Luciferase constructs (1 μg) were induced either with 1 mM Br-cAMP or by cotransfecting 0.5 μg of pCEV, a vector expressing a catalytic subunit of the PKA. Assays were done in duplicate, and mean values are shown with error bars. For details of the different constructs, see Materials and Methods and Fig. 1A.

We examined this interaction by the mammalian two-hybrid system (Clontech) in which the interaction was measured as CAT reporter gene activity in LLC-PK₁ cells. For this, we generated a panel of LFB3 deletion mutants fused to the DNA-binding domain of the yeast transcription factor GAL4 (Fig. 6A). As interaction partners, CREB and ATF1 were fused to the viral VP16 transactivation domain. As shown in Fig. 6B, GAL4-LFB3(1-559) did not mediate PKA stimulation either with or without coexpression of VP16-CREB or VP16-ATF1. However, when the carboxyl-terminal half of LFB3 [GAL4-LFB3(315-559)] was used, CAT expression was stimulated by the PKA signal and was further enhanced by coexpression of VP16-CREB or VP16-ATF1. A possible reason that the full-length LFB3 did not mediate the interaction is that the accessibility of the LFB3 region responsible for CREB/ATF1 binding was affected by the fusion of the GAL4 DNA-binding domain.

FIG. 6.

Analysis of the LFB3 molecule in a mammalian two-hybrid system for interaction with CREB and ATF1. (A) Schematic representation of Gal4-LFB3 and its deletion derivatives. DD, dimerization domain; POU, POU-specific domain; Pro-Glu, proline- and glutamine-rich activation domain. HA-tagged LFB3Δ(400-450) is diagrammed below. (B) CREB and ATF1 interact with the C terminus of LFB3. The CAT reporter gene construct pG5CAT (1 μg) was cotransfected into LLC-PK₁ cells with vectors expressing different Gal4-LFB3 deletion mutants and with either pVP16, pVP16-CREB, or pVP16-ATF1 (1 μg each). cAMP signaling was triggered by coexpressing the catalytic subunit of PKA (pCEV, 0.5 μg). CAT activity was measured as described in Materials and Methods. (C) Mapping of the LFB3 site interacting with CREB and ATF1. The CAT reporter gene construct pG5CAT was cotransfected into LLC-PK₁ cells with vectors expressing different Gal4-LFB3 deletion mutants and with either pVP16, pVP16-CREB, or pVP16-ATF1. cAMP signaling was triggered by coexpressing the catalytic subunit of PKA (pCEV). CAT activity was measured as described above.

Further dissection of the carboxyl terminus showed that the construct Gal4-LFB3(315-476), which does not contain the transactivation domain of LFB3, is able to mediate PKA-enhanced interaction (Fig. 6C). The shorter construct Gal4-LFB3(315-393) failed to bind CREB or ATF1, suggesting that the interaction region lies between aa 393 and 476 of LFB3. In agreement with this result, Gal4-LFB3(394-559) showed interaction with CREB and ATF1. Interestingly, CAT expression mediated by the construct Gal4-LFB3(477-559), which contained only the transactivation domain, could be induced by PKA (see Discussion).

Phosphorylation of CREB at Ser-133 is not important for interaction with LFB3.

The CREB protein belongs to the basic leucine zipper group of transcriptional regulators. The basic leucine zipper motif is a prerequisite for dimerization and DNA binding (2, 20). It also mediates protein-protein interaction with the oncoprotein Tax, a product of the human T-cell leukemia virus type 1 (1). Further, CREB harbors two distinct activation domains, a glutamine-rich constitutive activator (Q2) and a kinase-inducible domain, which function synergistically to stimulate target gene expression (reviewed in reference 32). PKA induces CREB phosphorylation at Ser-133 in the kinase-inducible domain, which then recruits the RNA polymerase II complex via the coactivator CBP (22).

To examine the role of Ser-133 and the bZIP motif in the interaction with LFB3, we introduced two kinds of mutations into CREB: the first converting Ser-133 to Ala-133 [CREB(S133A)] and the second deleting 71 amino acids at the carboxyl terminus encompassing the basic leucine zipper region (CREBΔ). We tested these mutations in a mammalian two-hybrid system and found that mutant CREB(S133A) behaved similarly to wild-type CREB with respect to PKA-enhanced LFB3 binding, suggesting that Ser-133 is not relevant for this interaction (Fig. 7). Additionally, we found that deletion of the basic leucine zipper region of CREB abolished PKA-enhanced interaction with LFB3 (Fig. 7). This suggests that CREB has to be a dimer to interact with LFB3 or that the basic-leucine zipper region is directly involved in the interaction with LFB3.

FIG. 7.

Unimportance of CREB Ser-133 in LFB3 binding. Two CREB mutants, a Ser-to-Ala mutant (CREBS133A) and a mutant lacking the leucine zipper (CREBΔ) (0.5 μg each), were cotransfected into LLC-PK₁ cells together with a vector expressing Gal4-LFB3(315-559) (0.5 μg) and the CAT reporter gene construct pG5CAT (1 μg). cAMP signaling was triggered by coexpression of the catalytic subunit of PKA (pCEV, 0.5 μg). CAT activity was measured as described in Materials and Methods.

As PKA enhanced the interaction of CREB with LFB3 without involving Ser-133 of CREB, we examined the possibility that PKA phosphorylates and activates LFB3. For this purpose, we carried out an in vitro PKA assay using HA-tagged LFB3 as substrate. We found that, in contrast to CREB, which was avidly phosphorylated, LFB3 was not phosphorylated in vitro by PKA or by cell extracts from cAMP-stimulated LLC-PK₁ cells (data not shown). It is possible that PKA indirectly activates LFB3 or CREB by phosphorylating a third protein, which in turn activates LFB3 or CREB and thus enhances the interaction between LFB3 and CREB. Cotransfection of vectors expressing HA-tagged LFB3 and either CREB or ATF1 in 293 cells followed by immunoprecipitation and immunoblotting with antiserum to either CREB or ATF1 did not lead to the detection of these proteins in the immunoprecipitates (data not shown). This may suggest the participation of the third protein.

Separation of CREB or ATF1 cooperating activity and basal enhancer activity of LFB3.

In order to verify the role of the region comprising aa 393 to 476 of LFB3 in the cooperation, we generated an internal deletion of aa 400 to 450 to produce LFB3Δ(400-450). We compared the effects of HA-tagged LFB3 and LFB3Δ(400-450) on cAMP induction of the ABC enhancer. In transient transfection assays with LLC-PK₁ cells using pABC-TATA as a luciferase reporter gene construct, we found that, in contrast to wild-type LFB3, increasing amounts of the LFB3Δ(400-450) expression vector did not lead to enhancement of cAMP induction. This result confirms that the region aa 400 to 450 is necessary for LFB3 to cooperate with CREB and ATF1 and mediate the cAMP signal (Fig. 8). To prove that the deletion of aa 400 to 450 affects only the mediation of the cAMP signal and not the transactivation capacity of transcription factor LFB3, we carried out transient transfection assays in 293 cells, where the level of endogenous LFB3 is very low. We compared two luciferase reporter constructs: the cAMP-sensitive construct pABC-TATA and the cAMP-independent pAlb-Luc, which is derived from the albumin promoter and bears a consensus binding site for LFB1 and -3. LFB3 has been shown to act as a basic (i.e., without additional signal) transactivation factor and activate the albumin promoter (42). It is clear that LFB3Δ(400-450) was as active as LFB3 in activating the albumin promoter (Fig. 9A), while LFB3Δ(400-450) failed to mediate cAMP induction of the ABC enhancer (Fig. 9B). We also confirmed the importance of aa 400 to 450 in LFB3 for the interaction with CREB and ATF1 by coprecipitation analysis. As shown in Fig. 9C, CREB and ATF1 bound to His-LFB3 but not to His-LFB3Δ(400-450). These results suggest that CREB and ATF1 cooperating activities and basal transactivation activity are ascribable to distinct regions of LFB3 and that they are separate activities.

FIG. 8.

Inability of LFB3Δ(400-450) to mediate cAMP-dependent induction of the ABC enhancer in LLC-PK₁ cells. LLC-PK₁ cells were transfected with pABC-TATA (0.5 μg) as luciferase reporter gene construct, along with increasing amounts (0, 50, and 100 ng) of empty vector or vectors expressing HALFB3Δ(400-450) or HALFB3. cAMP signaling was triggered by adding 1 mM Br-cAMP (6 h) before luciferase activity was measured.

FIG. 9.

Distinct functions of LFB3 for basal transactivation and mediation of cAMP induction. (A) LFB3Δ(400-450) retains its transactivation capacity. 293 cells were transiently transfected with the cAMP-independent reporter gene expression vector pAlb-Luc (0.5 μg) together with vectors encoding either HALFB3 or HALFB3Δ(400-450) (50 ng). Luciferase activity was measured and expressed as for Fig. 3B. (B) LFB3Δ(400-450) is unable to mediate cAMP induction. 293 cells were transiently transfected with the cAMP-dependent reporter gene expression vector pABC-TATA (0.5 μg) together with vectors encoding either HALFB3 or HALFB3Δ(400-450) (5 ng). cAMP signaling was triggered by coexpressing the catalytic subunit of PKA (pCEV, 0.5 μg). Luciferase activity was measured and expressed as described above. (C) CREB and ATF1 coprecipitate with LFB3 but not with LFB3Δ(400-450). In vitro [³⁵S]methionine-labeled CREB or ATF1 was incubated with empty Ni-chelating Sepharose beads (Pharmacia) (control) or with beads coupled with His-tagged LFB3 or LFB3Δ(400-450) protein. After a wash and elution, samples were subjected to SDS-polyacrylamide gel electrophoresis. The bands were visualized by fluorography.

DISCUSSION

The ABC enhancer of the pig uPA promoter consists of three protein-binding domains and mediates cAMP induction. The presence of all three domains is required for full inducibility. We showed previously that LFB3 binds to the C domain and cooperates with the proteins binding to domains A and B (28, 31). In the present study, we report that ATF1 and CREB bind to the AB domains. However, we cannot exclude the possibility that other CREB/ATF1 family members bind to these domains, since the main protein-DNA complex was not completely supershifted by antibodies against CREB and ATF1 (Fig. 3A). The ABC enhancer was inducible in all cell lines tested when LFB3 was coexpressed, suggesting that the ubiquitous ATF1 and CREB are likely to be the proteins mediating cAMP induction of the ABC enhancer (Fig. 3C).

We characterized the ABC enhancer by creation of several mutations. A 5- or 10-nucleotide insertion between domains AB and the LFB3-binding domain C substantially reduced inducibility. Therefore, both close proximity and a certain angular orientation between LFB3 and AB domain-binding proteins are required for their cooperativity. These requirements suggest a physical interaction; indeed, mammalian two-hybrid systems and coprecipitation studies showed that CREB and ATF1 bind to LFB3 (Fig. 6 and 9).

The consensus CRE sequence differs from the TRE sequence only in 1 nucleotide (34), and the transcription factors binding to TREs, the AP1 factors, are structurally related to CREB and ATF1, having a basic leucine zipper domain as DNA-binding motif (19). However, AP1 factors cannot heterodimerize with CREB (4) or with ATF1 (14, 29). We tested the possible interaction of LFB3 with AP1 factors by converting the imperfect CRE sites into consensus AP1-binding sites. However, no significant induction by TPA was detected on such a construct. Therefore, the cooperative role of LFB3 is specific for the cAMP induction of the ABC enhancer (Fig. 2). Thus, LFB3 interacts most likely with an area of CREB or ATF1 that is not present in AP1 factors.

The deviation of protein-binding sequences in the ABC enhancer from the respective consensus sequences is instrumental in ensuring the tight regulation of uPA gene expression in a hormone-dependent and tissue-specific manner. The AB domains alone do not allow cAMP induction, and the C domain alone does not lead to enhanced basal activity. However, if each binding sequence is converted to the consensus element, the A*B* domains without the C domain can mediate cAMP induction and the ABHNF shows considerable basal activity, thus weakening both tissue specificity and hormone dependency (Fig. 1). The most likely explanation for the loss of tissue specificity with A*B*C is that, while CREB/ATF1 binds only weakly to the imperfect CRE sequence, it binds to the consensus CRE with a higher affinity. It is known that PKA-induced phosphorylation influences two aspects of CREB regulation: DNA binding and transactivation. CREB binds to the consensus CRE with a high affinity without being phosphorylated (37), and the role of its phosphorylation lies mainly in recruitment of the cofactor CBP (7). In fact, we could show that the binding of CREB or ATF1 to the A and B domains is increased upon phosphorylation by PKA but that consensus CREs are bound by CREB and ATF1 independently of PKA treatment (Fig. 4). Similar results with purified CREB were described earlier (37). Binding of LFB3 to the consensus LFB1/3 sequence is stronger than that to the domain C sequence (data not shown). Cooperation between LFB3 and CREB or ATF1 may lead to stronger binding of the whole complex to the ABC enhancer than binding of the individual factors to their binding sites. Therefore, the imperfect consensus sequences in the ABC enhancer are necessary for the weak binding of the individual transcription factors to their binding sites, ensuring the coupling of hormonal and tissue-specific gene regulation. Only upon formation of the whole complex with LFB3 and CREB or ATF1 after phosphorylation by PKA, is the binding of the transcription factors strong and transcription is initiated.

Experiments using the mammalian two-hybrid system showed that PKA enhances the interaction between LFB3 and CREB or ATF1 (Fig. 6). The influence of PKA on gene expression is mediated by phosphorylation of various transcription factors, including CREB, ATF1, and CREM (13, 23). Phosphorylation of CREB at Ser-133 by PKA results in increased transactivation activity, paralleled by induced association with CBP. CBP interacts with TFIIB, thus connecting CREB to the basal transcriptional machinery (12, 22). Our observation that Ser-133 is not relevant for CREB/LFB3 interaction (Fig. 7) supports the notion that this residue has to remain accessible for CBP. In addition, cell-free phosphorylation experiments showed that LFB3 is not the direct target of PKA. It remains to be seen whether LFB3 is phosphorylated in vivo upon stimulation with agents leading to PKA activation. In the two-hybrid system, we observed that Gal4-LFB3(477-559), which is composed of the transactivation domain of LFB3, could activate the reporter gene when the PKA catalytic subunit was coexpressed. Currently, we are examining the possibility that the transactivation domain of LFB3 is indirectly phosphorylated in vivo by PKA or interacts with a protein whose activity is modulated by PKA. The potential involvement of another protein in the hormonal or tissue-specific regulation of the uPA gene might be supported by the fact that an attempt to coimmunoprecipitate LFB3 with either CREB or ATF1 was not successful (data not shown).

The mammalian two-hybrid system showed that the region aa 393 to 476 of LFB3 is involved in the interaction with CREB or ATF1. The importance of this region for mediating cAMP induction was confirmed in transient transfection assays. In contrast to full-length LFB3, LFB3 lacking aa 400 to 450 failed to mediate cAMP induction of the ABC enhancer (Fig. 8 and 9B). Interestingly, the capacities of the two proteins LFB3 and LFB3Δ(400-450) to enhance basal transcription activity were comparable (Fig. 9A), indicating that cooperation with CREB or ATF1 and basal transactivation activity are two distinct functions of LFB3 mapping to two different regions of the protein. A corresponding region (aa 280 to 440) of HNF1α (LFB1), the LFB3-related, liver-enriched transcription factor, has recently been reported to be involved in protein-protein interaction with another liver-enriched transcription factor, HNF4 (21). In contrast to our observation, where LFB3 cooperates with CREB or ATF1 in positively mediating cAMP signals on the ABC enhancer, this region of HNF1α mediates suppression of HNF4-dependent genes without directly binding to DNA. HNF4 is a liver-enriched transcription factor of the steroid hormone receptor superfamily (45), distinct from basic leucine zipper type transcription factors such as CREB. It would be interesting to see whether the region aa 400 to 450 of LFB3 cooperates with transcription factors other than CREB and ATF1.