Transcriptional Cross-talk: Nuclear Factors CREM and CREB Bind to AP-1 Sites and Inhibit Activation by Jun*

The proteins Fos and Jun dimerize to constitute the transcription factor AP-1 which is known to respond to treatment with phorbol esters. AP-1 binds to 12-0-tetradecanoylphorbol-13-acetate-responsive elements (TREs) palindromic sequences. CAMP-responsive elements (CREs) are very similar to TREs and CRE-binding proteins are similar in structure to Fos and Jun. Thus, the two main signal transduction pathways have closely related nuclear effectors which could pos- sibly overlap and/or cross-talk. The gene CRE modu-lator (CREM) encodes both antagonists and an activa- tor of the CAMP transcriptional response by alternative splicing. In this report we show that CREM antagonists are able to block the transcriptional activation elicited by c-Jun. The mechanism by which this repression is obtained does not require heterodimerization between CREM and the Fos and/or Jun proteins. In contrast, we show that both CREM and CRE-binding proteins (CREB) are able to bind TREs and therefore compete with c-Jun for this site. Removal of the phos- phorylation domain in CREM does not affect the down-regulatory function. We also show that c-Fos does not affect the inhibitory function of CREM on c-Jun

The proteins Fos and Jun dimerize to constitute the transcription factor AP-1 which is known to respond to treatment with phorbol esters. AP-1 binds to 12-0tetradecanoylphorbol-13-acetate-responsive elements (TREs) palindromic sequences. CAMP-responsive elements (CREs) are very similar to TREs and CREbinding proteins are similar in structure to Fos and Jun. Thus, the two main signal transduction pathways have closely related nuclear effectors which could possibly overlap and/or cross-talk. The gene CRE modulator (CREM) encodes both antagonists and an activator of the CAMP transcriptional response by alternative splicing. In this report we show that CREM antagonists are able to block the transcriptional activation elicited by c-Jun. The mechanism by which this repression is obtained does not require heterodimerization between CREM and the Fos and/or Jun proteins. In contrast, we show that both CREM and CRE-binding proteins (CREB) are able to bind TREs and therefore compete with c-Jun for this site. Removal of the phosphorylation domain in CREM does not affect the downregulatory function. We also show that c-Fos does not affect the inhibitory function of CREM on c-Jun and that the transcriptional activation elicited by the other members of the jun family (JunB, JunD, and v-dun) is also down-regulated by CREM.
Activation of signal transduction pathways elicits the regulation of gene expression in the nucleus via the modulation of transcription factor activity. In one major pathway, following the binding of ligands to their cognate receptors, inositol phospholipids are hydrolyzed to generate inositol 1,4,5-triphosphate and 1,2-diacylglycerol, which leads to the activation of protein kinase C (Berridge, 1987). In turn, activated protein kinase C phosphorylates several proteins which serve as mediators in the processes of cell proliferation, growth control, and gene regulation.
The second main signaling pathway involves the CAMPdependent protein kinase A and is initiated by the activation of adenylyl cyclase (Nishizuka, 1984;Borrelli et al., 1992). *This work was supported in part by grants from the Centre National de la Recherche Scientifique, the Institut National de la Sante et de la Recherche Medicale, and the Association pour la Recherche contre le Cancer et Rh6ne-Poulenc Rorer. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$ Supported by a fellowship from the Ministere de la Recherche et la Technologie.
The jun and fos genes are members of the immediate early response class, whose expression is rapidly and transiently induced through intracellular pathways activated by extracellular stimuli (Kruijer et al., 1984;Ryder et al., 1988Ryder et al., , 1989Lamph et al., 1988;Quantin and Breathnach, 1988). Functional specificity among the various members of this gene family is likely to be determined by their differential distribution and transcriptional inducibility (Ryder et al., , 1989Hirai et al., 1989;Auwerx et al., 1990;Naranjo et al., 1991;Mellstrom et al., 1991). Therefore, the products of these genes act as cell-specific nuclear third messengers, converting cytoplasmic signals into changes in gene expression.
Promoter elements that respond to CAMP, which activates the adenylyl cyclase pathway, are termed CREs (CAMP-responsive elements) (Comb et al., 1986;Delegeane et al., 1987;Sassone-Corsi, 1988). Although the difference between TPAand CAMP-responsive promoter elements is only one nucleotide (TRE = TGACTCA; CRE = TGACGTCA), they seem to mediate induction only by their respective agonists (Sas-sone-Corsi et al., 1990). The factors which recognize CRE sequences are encoded by a large gene family which belong to the leucine zipper (b-Zip) class Hai et al., 1989;Habener, 1990;Foulkes et al., 1991a;Borrelli et al., 1992). CREB, the first member of this family to be characterized, is a protein of 43 kDa which binds to DNA as a homodimer. Stimulation of the cAMP pathway induces phosphorylation of CREB by protein kinase A, and phosphorylated CREB activates transcription from CRE sites (Gonzalez et al., 1991).
We have recently described the CREM (CRE modulator) gene. The most striking feature of the CREM cDNA is the presence of two alternative DNA-binding domains. Various mRNA isoforms were identified that appear to be obtained by differential cell-specific splicing. Alternative usage of the two DNA-binding domains was demonstrated in various tissues and cell types, where quite different patterns of expression were found (Foulkes et al., 1991a. Three major isoforms, a, p, and y, were initially characterized, which revealed alternative usage of the two DNA-binding domains. CREMa encodes a protein with the first DNA-binding domain, whereas CREMP uses the second DNA-binding domain. CREMy is equivalent to CREMP but contains a small deletion of 12 amino acids. The strict cell-and tissue-specific expression of CREM is indicative of a pivotal function in the regulation of the cell-specific cAMP response. The CREM products share extensive homology with CREB, especially in the DNA-binding domains and the kinase-inducible domain (KID) region.
The CREM proteins specifically recognize CREs and show the same binding properties as CREB. CREM proteins containing either DNA-binding domain I or 11, heterodimerize with CREB (Foulkes et al., 1991a), although it appears that CREMa-CREB heterodimer formation is more favored than CREMP-CREB.' These properties suggest that CREM proteins might occupy CRE sites as CREM homodimers or as CREM-CREB heterodimers, thus generating complexes with altered transcriptional functions. In fact CREMa and -6 products act by impairing CRE-mediated transcription and as such are considered as antagonists of CAMP-induced expression (Foulkes et al., 1991a(Foulkes et al., , 1991b. In adult testis a novel CREM isoform (CREMT) has been identified which differs from the previously characterized CREM antagonists by the coordinate insertion of two glutamine-rich domains which confer transcriptional activation function to the protein . Thus, the CREM gene also encodes an activator of CAMP-dependent transcription. During spermatogenesis there is an abrupt switch in CREM expression. In premeiotic germ cells, CREM is expressed at low levels in the antagonist forms. Subsequently, from the pachytene spermatocyte stage onward, a splicing event generates exclusively the CREMT activator . This new transcript accumulates at extremely high levels. This splicing-dependent reversal in CREM function represents an important example of developmental modulation in gene expression and provides further evidence for CREM occupying a pivotal position within the transcriptional response to cAMP .
Here we describe that the CREM antagonists are capable of negatively modulating the transcriptional activity elicited by Jun/AP-1. This phenomenon constitutes a cross-talk between signal transduction pathways at the transcriptional level.

EXPERIMENTAL PROCEDURES
Plasmid Constructs-The reporter plasmid TRE-tk-CAT contains a human metallothionein IIA TRE sequence (TGACTCA) inserted upstream of the herpes simplex thymidine kinase promoter (positions -109 to +57) which is fused to the chloramphenicol acetyltransferase gene (Sassone-Corsi et al., 1988a). pSVCREMa, -0, -7, and -T (sense and antisense) expression vectors are described elsewhere (Foulkes et al., 1991a. pSVCREB contains the rat CREB cDNA (Gonzales et al., 1991) cloned into the expression vector pSG5 (Green et al., 1988). pCaEV (McKnight et al., 1988) is an expression plasmid for the catalytic subunit of mouse protein kinase A and is a gift from G. S. McKnight (University of Washington, Seattle). The c-fos, c-jun, junD, junB, and v-jun expression vectors have been described elsewhere (Sassone-Corsi et al., 1988a;de Groot et aL, 1991). The c-fos, c-jun, CREM, and CREB plasmids which were used for in uitro transcription/translation studies have been described elsewhere (Foulkes et al., 1991a. CREMAa and CREMAP were generated by the deletion of an NcoI fragment which includes the 5' 300 base pairs of the CREM coding sequence (Delmas et al., 1992). We used CREMa, CREMP, and CREB bacterial expression vectors (Delmas et al., 1992), in which NcoI-BamHI full-length cDNA fragments are cloned into p E T l l d (Studier et al., 1990). The ACREB pETlld vector was generated by deletion of an NcoI-KpnI fragment which spans the NHp-terminal portion of the CREB protein, between amino acid positions +1 and +198. The open reading frame was preserved by inserting a synthetic oligonucleotide of 18 base pairs (5' CATGG-CTAACAATGGTAC 3') at the deletion site.
Protein Expression and Gel Retardation Assays-Fos and Jun proteins were synthesized in vitro by using a rabbit reticulocyte translation system (Promega) and labeled with [3sS]methionine. Products were analyzed on 11% polyacrylamide SDS gels. Gels were fixed with acetic acid/methanol and dried before autoradiography. In uitro synthesized unlabeled Fos and Jun proteins were used in gel retardation assays. CREB, ACREB, and CREM proteins were produced in uiuo using the p E T l l d bacterial expression vector (Studier et al., 1990). Proteins were purified as described before (Hoeffler et al., 1991). Gel retardation assays were performed to analyze the binding of the proteins to a synthetic double-stranded 19-base pair oligonucleotide containing the human metallothionein IIA TRE. 2 pl of the in uitro translated proteins and/or 0.2-2 pg of bacterial proteins were incubated with 2 pg of poly(d1-dC) (Boehringer Mannheim) in TM buffer (50 mM Tris-HC1, pH 7.9, 12.5 mM MgCl,, 1 mM EDTA, 1 mM dithiothreitol, 20% glycerol) in a 20-pl final volume. [-y-32P]ATP endlabeled double-stranded oligonucleotide was added, and the reaction was incubated for 20 min at room temperature. DNA-protein complexes were resolved on a 4% polyacrylamide gel containing 0.25 X TBE (1 X TBE = 50 mM Tris borate, pH 8, 3/1 mM EDTA). Transfections and CAT Assays-JEG-3 human choriocarcinoma cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum (Delegeane et al., 1987) and were transfected by the calcium phosphate coprecipitation technique. Cells were plated at a density of lo6 cells/lO-cm plate and transfected with 10 pg of total plasmid DNA. 2 pg of reporter plasmid was included in each transfection sample together with 2 pg of pSVc-jun, pSVjunB, pSVjunD, or pSVv-jun, 1-4 pg of pSVCREM expression plasmids and, when required, with 0.4 pg of pCaEV. CAT activity was assayed as described previously (Sassone-Corsi et al., 1988a) and was quantified by liquid scintillation counting of the TLC plate '*C spots.
Recent results (Foulkes et al., 1991b) indicate that c-fos transcription is down-regulated by CREM proteins via binding to a noncanonical CRE site in the c-fos promoter. The repression by CREM is likely to be responsible for the observed c-fos decreased expression observed after CAMP induction (Bravo et al., 1987). These results further confirm the existence of cross-talk between factors belonging to different signal transduction pathways. On the basis of these observations and considering previous data showing that Fos-Jun dimers can bind to CRE sites (Hoeffler et al., 1989;Sassone-Corsi et al., 1990), we decided to investigate in more detail whether CREB and CREM proteins could bind to a canonical TRE site and possibly modulate Fos and Jun function.
Confirming previous results, using CREM proteins generated in vitro with a reticulocyte lysate, we were unable to detect binding to a TRE, whereas binding to a CRE was readily observed (Fig. 1, compare lanes 1 and 2 with 3 and 4). Similar results were obtained using CREB (not shown). These data either indicate that CREB and CREM are intrinsically unable to bind to a TRE, or, alternatively, that the in vitro generated proteins are not fully functional. Given that the DNA-binding domains of Fos, Jun, CREB, and CREM proteins are highly similar (Habener, 1990;Borrelli et al., 1992), it would seem likely that these factors could overlap in their binding specificities. Thus, we reasoned that CREB and CREM proteins generated in vivo might bind TREs more efficiently than in vitro generated proteins. In Fig. 1, we show that CREMa, CREMP, and CREB generated in bacteria do efficiently bind to a canonical TRE (lanes 6, 8, and 11) and that this binding is about 3-&fold less efficient than to a cognate CRE site (lanes 14 and 16 and not shown). The relative affinities of CREB, CREMa, and CREMP for both CRE and TRE sites are equivalent, as established by several DNA binding experiments using variable amounts of protein ( Fig. 1 and not shown). Truncated forms of CREB and CREM proteins (ACREB, ACREMa, and ACREMp) are described under "Experimental Procedures" and were used to allow an easy visualization of protein complexes. Experiments carried out with the full-length proteins gave equivalent results (not shown).
It is important to stress that, in order to obtain efficient binding of CREB and CREM proteins to a TRE, the addition of small amounts of reticulocyte lysate was necessary (compare lanes 5-7 with 8,9, and 11 ). Reticulocyte lysate has been shown previously to enhance the DNA binding activity of several leucine zipper-containing transcription factors, and its effect appears to be due to stimulating cofactors present in the lysate (Abate et al., 1990;Busch and Sassone-Corsi, 1990b). Interestingly, reticulocyte lysate stimulates CREB and CREM binding to a TRE much more significantly than binding to the cognate CRE site (compare lanes 5 and 6 with  13 and 14 and lanes 9 and 11 with 15 and 16). The molecular basis of this difference is unclear and currently under study.
:' E. Benusiglio, personal communication.  I , 2, and 13-16) or the human metallothionein IIA TRE site (Angel et al., 1987) (lanes [3][4][5][6][7][8][9][10][11][12]. Using 2 pl of in uitro synthesized CREMa and CREMP proteins, binding was detected to a CRE probe (lanes I and 2, respectively) but not when using a TRE probe (lanes 3 and 4, respectively). When 1 pl of rabbit reticulocyte lysate (RCL, Promega) was added to 1 pl of bacterially synthesized CREMa protein (corresponding to 2 gg of protein), an enhanced binding to the TRE probe (lane 6 ) was observed relative to when 1 pl of the protein was used alone (lane 5). The same results were obtained using 1 pl of CREMP (2 pg), CREB, and ACREB (0.2 pg) synthesized in bacteria (lanes 7, 9, and IO, respectively) without rabbit reticulocyte lysate and by adding 1 pl of lysate (lanes 8, I I, and 12, respectively). The modulation of binding of CREMa and CREB protein to a CRE probe by rabbit reticulocyte lysate is shown (lanes 13 and 15, without lysate; and lanes I4 and 16, with lysate). In this assay, it is important to note that the concentration of CREB and ACREB proteins was 1/10 less than CREMn and CREMP. Bands corresponding to specific binding by CREMn, CREMP, CREB, and ACREB are labeled with arrowheads. Presence of a nonspecific complex is due to binding of endogenous lysate proteins and is indicated by a star.
The experiments were performed by using standard procedures, and binding specificity of the proteins generated in bacteria was assessed by the use of unrelated factor-binding sites (Spl, Octl; not shown). Thus, we believe that the reason for not seeing binding of CREB and CREM to a TRE when the proteins are generated in an reticulocyte lysate is related to the amount of protein obtained. The presence of cofactors in cellular extracts which stimulate CREM and CREB binding to a TRE is suggestive that these transcription factors could interact with TREs in vivo. Indeed, as shown later, we demonstrate that CREB and CREM affect c-Jun-mediated transactivation in vivo.
Binding Competition of Fos-Jun by CREB and CREM-Next we wanted to establish whether the binding of CREB, CREMa, and CREMP to a TRE would be sufficient to compete with the formation of the Fos-Jun/TRE complex. To do so, we first generated Fos and Jun in vitro using a reticulocyte lysate system, and we observed the expected binding of the heterodimer to a labeled TRE ( Fig. 2A, lanes 1 and 6). In order to study the effect of CREB on Fos-Jun binding to a TRE, we generated in bacteria a truncated form of CREB, ACREB (see "Experimental Procedures"), since the CREB dimer migrates at a similar position as the Fos-Jun heterodi- mer in a nondenaturing gel shift assay. Importantly, ACREB binding is also stimulated by the presence of reticulocyte lysate (Fig. 1, compare lane 10 with 12). Since ACREB consists of the 134 carboxyl-terminal amino acids, it is evident that whatever in the lysate elicits the stimulatory effect, it has as its target this part of the CREB protein. Addition of ACREB in increasing amounts with Fos-Jun causes a significant decrease in Fos-Jun binding to the TRE ( Fig. 2A, lanes  2-5). This shows that recognition of the TRE by CREB is specific and that the 194 amino-terminal residues in CREB are dispensable to obtain competition. We performed similar experiments with CREMa and CREMP proteins and obtained the same results (Fig. 2 A , lanes 6-10 and not shown). Experiments with truncated forms of CREMa and CREMP (see "Experimental Procedures") gave the same results as ACREB (not shown).
In order to establish the precise amount of CREM protein required to displace the Jun protein from binding, we per-formed experiments with bacterially generated Jun, CREMa, and CREMP proteins (Fig. 2B). The concentrations of the proteins were carefully established prior the experiment. As shown in the Fig. 2B (compare lane 1 with 4-11), there is a decrease in Jun binding already at sub-stoichiometric concentrations of CREMa and CREMB (lanes 4,5,8, and 9), whereas at a higher concentration, Jun binding is totally abolished (lanes 7 and 11).
It is noteworthy that there are no additional retarded bands on the gel shift, indicating that no new heterodimeric complexes are formed. This further strengthens the idea that the observed phenomenon is based on direct competition for the same binding site and that no heterodimerization occurs between Fos or Jun and CREB or CREM.
CREM Does Not Dimerize with Fos or Jun-In order to establish whether CREMa and CREMP could dimerize with Fos or Jun, we performed immunoprecipitation experiments. We tested both CREMa and CREMP proteins, since they differ in the DNA-binding and dimerization domain (Foulkes et al., 1991a). 35S-Labeled proteins were generated by in vitro transcription and translation, and complexes were analyzed after immunoprecipitation with specific anti-Fos (generated against the "peptide; Curran et al., 1985) and anti-Jun antibodies (generated against the Pep-1 peptide; Bohmann et al., 1987). Because of similarities in protein sizes on polyacrylamide gel electrophoresis gels, we decided to use truncated forms of CREMa (CREMAa) and CREMP (CREMAB), where the NH2-terminal 100 amino acids have been removed by an in-frame Ne01 deletion' (see also "Experimental Procedures"). As shown in Fig. 3A, although, as expected, Fos and Jun proteins co-immunoprecipitate using the anti-Fos antibody (lane lo), both CREMAa and CREMAP do not co-immunoprecipitate with either Fos or Jun (Fig. 3A, lanes 8 and 9; Fig.  3B, lanes 6 and 7 ) using anti-Fos or anti-Jun antibodies. Similar results were obtained with full-length CREMa and CREMP in vitro synthesized proteins (not shown). Thus, CREM proteins behave similarly to CREB, which has already been shown not to dimerize with Fos or Jun (Lamph et al., 1990).
CREM and CREB Block Jun-mediated Tramactiuation-In order to study the in vivo consequences of the competition for TRE binding by Fos-Jun and CREB or CREM, we performed transfection experiments in cultured cells. We used TRE-tk-CAT, a reporter plasmid which contains a copy of the metallothionein AP-1 site inserted upstream of the herpes thymidine kinase promoter (-109/+57) fused to the bacterial chloramphenicol acetyltransferase gene (CAT; Angel et al., 1987). This reporter is efficiently transactivated upon cotransfection with a c-jun expression vector (Sassone-Corsi et al., 1988a;see Fig. 4, first two bars of the histogram). Coexpression of CREMa, -P,--y, and -7 (Foulkes et al., 1991a;1992) proteins dramatically represses the c-jun-mediated transactivation (Fig. 4). Cotransfection with CREM antisense vectors had no effect upon c-jun transactivation (Fig. 4). Similar effect was obtained by co-expression of CREB. It is notable that some down-regulation is obtained already when transfecting a 1:2 ratio of CREM or CREB: c-jun expression vector.
The CREM KID Domain Is Dispensable for Down-regulation of c-Jun Function-Since it has been shown that the negative effect of CREB upon transcription from the c-jun promoter is dependent on the degree of phosphorylation of CREB (Lamph et al., 1990), we decided to test whether the effect of CREM on c-Jun activation from a TRE could also be modulated by protein kinase A. To do so, we cotransfected an expression vector for the catalytic subunit of protein kinase A with CREM (Fig. 5). It has been shown that co-expression of the catalytic subunit of the protein kinase A strongly enhances CRE-mediated transcription (Mellon et al., 1989;Foulkes et al., 1991a). In this way, we scored for the effect of protein kinase A-induced phosphorylation on the Jun-mediated transactivation. The results showed that the co-expression of protein kinase A does not affect the negative function of CREM on transactivation by Jun (compare Figs. 4 and  5B). Furthermore, we studied the contribution of the phosphorylation domain in CREM to the down-regulation of c-Jun activity. To do so, we used expression vectors for the truncated versions of CREMa and CREMP (CREMAa and CREMAP) which lack the phosphorylation domain (KID) that contains sites for protein kinase A, protein kinase C, and other kinases, also conserved in CREB (Foulkes et al., 1991a;  Foulkes et ul., 1991aFoulkes et ul., , 1992 or pSVCREB. For each expression plasmid, 1 or 4 pg were transfected (left to right). TRE-tk-CAT reporter transcription was induced by cotransfection with 2 pg of the c-jun expression vector pSVc-jun. Hoeffler et al., 1988). The expression of the CREMAa and CREMAP proteins is comparable with CREMa and CREMP, as shown by immunoprecipitation with specific CREM antibodies after transfection.* The truncated forms of CREM also negatively regulate transactivation elicited by c-Jun, with an efficiency comparable with CREMa and CREMP. Thus, the presence of the KID region of CREM is not required for the repression function.
Regulatory Function of v-Jun, JunB, and JunD Is Also Blocked by CREM-The Jun family of proteins includes JunB and JunD, which are extremely similar to c-Jun but have different tissue and cell distributions (Wilkinson et al., 1989;Ryder et al., 1988;Auwerx et al., 1990;Hirai et al., 1989: Mellstrom et al., 1991. In transfection experiments with TRE-tk-CAT, we have been able to observe transactivation by JunD (about 60% with respect to c-Jun) and a weaker activation by JunB (about 20% with respect to c-Jun). Transactivation is also detected with the viral protein v-Jun (Baichwal and Tjian, 1990). Thus, we tested whether CREM and CREB could also block the transregulatory function of these three Jun proteins. As shown here, the function of vJun, JunB, and JunD is down-regulated by CREM (Fig. 6) and by CREB (not shown).
We have also tested whether co-expression of c-Fos could affect the negative regulation of CREM on Jun. As shown in Fig. 6 even when Fos was coexpressed with all Jun proteins, both CREM and CREB proteins still efficiently down-regulate trans-activation through the AP-1 site.

DISCUSSION
The structure of the DNA-binding domain of all the transcription factors which belong to the b-Zip family is highly similar (Landschulz et al., 1988; for review see Busch and Sassone-Corsi, 1990a). Despite this similarity, not all of these proteins interact with homologous DNA sequences. However, a large group of b-Zip factors which include at least 16 distinct proteins, appear to recognize promoter elements which have a related palindromic structure. These are the products of the CAT reporter plasmid in combination with pCaEV, pSVCREB, and the sense ( S ) and antisense (AS) pSVCREMa, -/3, -7, and -7 expression plasmids. cJun-induced transcription was obtained by cotransfection with pSVc-jun. c-Jun-mediated transactivation is enhanced by co-expression of protein kinase A (not shown; see also deGroot and Sassone-Corsi, (1992)). B, histograms summarizing CAT assays results. CAT activity is expressed as percentage of chloramphenicol acetylated (see legend Fig. 4 and "Experimental Procedures"). The TRE-tk-CAT reporter plasmid was cotransfected with 0. 4 pg of pCnEV, the expression vector encoding the catalytic subunit of protein kinase A (Mellon et al., 1989). These plasmids were cotransfected with CREM and CREB sense ( S ) and antisense ( A ) expression plasmids (see also legend to Fig. 4). pSVCREMAa and AB are expression plasmids for the deleted forms of CREMa and $2 c-Jun-induced transcription was obtained by cotransfection with 2 pg of pSVc-jun. . CAT activity is expressed as percentage of chloramphenicol acetylated. TRE-tk-CAT reporter plasmid was cotransfected with 1 and 4 pg of the pSVCREMn and pSVCREMP. Transcription was induced by cotransfection with pSVv-jun, pSVjunB, pSVjunD, and pSVc-fos which encode v-Jun, JunB, JunD, and c-Fos, respectively. The protein kinase A expression vector pCnEV was included in some experiments as indicated.
promoter elements as homo-or heterodimers. Heterodimeric combinations dramatically increase the combinatorial possibilities and, thus, the potential functionality of these factors. The Fos and Jun products associate to constitute the transcription factor AP-1, which is a target of the TPA-induced signal transduction pathway. The ATFand CRE-binding proteins are nuclear targets of a distinct pathway, which employs intracellular CAMP as a second messenger. Thus, proteins which belong to distinct signaling systems have similar structure and bind to similar DNA elements. Importantly, only some specific dimer combinations are possible among the members of the ATF/CRE-binding proteins, indicating that strict rules are operating in the cell to allow finely tuned regulation.
In this work we have presented data showing that a canonical TRE can be recognized by two CRE-binding proteins, CREB and CREM (Figs. 1 and 2). CREM is a gene which encodes multiple transcriptional regulators by differential splicing (Foulkes et al., 1991a. Two alternative DNAbinding domains are used which confer different binding specificity to the encoded products? CREM proteins bearing either DNA-binding domain are able to bind to TRE sequences (Fig. 1). In addition, we observed a dramatic downregulation of the transactivation potential elicited by all Jun proteins (c-Jun, JunB, JunD, and v-Jun, see Figs. 4-6). We have presented evidence that the down-regulation is likely to be obtained by occupation of the TRE by CREM dimers, since CREM proteins do not heterodimerize with Jun or Fos (see Fig. 3). Importantly, down-regulation is obtained already at a 1:2 ratio of the transfected expression vectors for CREM and Jun, suggesting that CREM binding for the TRE is comparable with Jun. We have found that there are no significant differences between the various CREM proteins in the negative regulation of Jun-mediated transactivation (see Fig. 4), paralleling the observation that both CREM DNAbinding domains confer similar binding activity to the respective CREM proteins.
An important conclusion of this study is that the phosphorylation domain (KID) of CREM is dispensable for the down-regulatory function. In fact, CREM truncated proteins which lack the KID region (CREMAa and CREMAP, see Fig.  4) efficiently repress Jun-mediated transactivation. In addition, co-expression of the catalytic subunit of the protein kinase A does not affect the negative regulatory effect of both CREMa and CREMP (Fig. 5). Taking into account these results, and considering that the truncated CREM proteins efficiently bind DNA: it seems clear that the negative function of CREM over Jun is almost exclusively due to the occupation of the TRE by the CREM dimer (see also Fig.  2B).
One interesting aspect of the results reported here is the fact that transcription factors which could be considered as targets of different signal transduction pathways could affect each others function. It was already shown that Fos-Jun heterodimers bind and activate transcription from CREs (Hoeffler et al., 1989;Sassone-Corsi et al., 1990). Here, we have shown that CRE-binding proteins can negatively act upon the transactivation exerted by Jun on TREs. This mechanism of interference between signal transduction pathways at the nuclear level is paralleled by similar events in the cytoplasm (Deli et al., 1988;Bell et al., 1985;Kelleher et al., 1984;Cambier et al., 1987;Yoshimasa et al., 1987). Cross-talk between signaling systems in the nucleus has been described in several cases (Imagawa et al., 1987;Tratner et al., 1992; Auwerx and Sassone-Corsi, 1991; de Groot and Sassone-Corsi, 1992). The observations described in this report demonstrate the complexity of the molecular mechanisms of gene regulation and their links with intracellular signal transduction.