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
). 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. A). 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. C).
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. and ).
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
). 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. ). 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. ). 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. ). 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. ). The influence of PKA on gene expression is mediated by phosphorylation of various transcription factors, including CREB, ATF1, and CREM (13
). 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
). Our observation that Ser-133 is not relevant for CREB/LFB3 interaction (Fig. ) 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. and B). Interestingly, the capacities of the two proteins LFB3 and LFB3Δ(400-450) to enhance basal transcription activity were comparable (Fig. A), 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.