Changes in environmental conditions such as the addition of growth factors or irradiation of cells in culture first affect immediate response genes. We have shown previously that short wavelength UV irradiation (UVC) elicits massive activation of several growth factor receptor‐dependent pathways. At the level of the immediate response gene c‐fos, these pathways activate the transcription factor complex serum response factor (SRF)–p62TCF which mediates part of the UV‐induced transcriptional response. These studies have, however, suggested that more that one pathway is required for full UV responsiveness of c‐fos. Using appropriate promoter mutations and dominant‐negative cAMP response element (CRE)‐binding protein (CREB), we now find that UVC‐induced transcriptional activation depends also on the CRE at position −60 of the c‐fos promoter and on the functionality of a CREB. Upon UV irradiation, CREB and ATF‐1 are phosphorylated at serines 133 and 63, respectively, preceded by and dependent on activation of p38/RK/HOG‐1 and of a p38/RK/HOG‐1‐dependent p108 CREB kinase. Although p90RSK1 and MAPKAP kinase 2 are also activated by UV, p90RSK1 does not, at least not decisively, participate in this signalling pathway to CREB and ATF‐1 as it is not p38/RK/HOG‐1 dependent, and CREB is a poor substrate for MAPKAP kinase 2 in vitro. On the basis of resistance to the growth factor receptor inhibitor suramin and of several types of cross‐refractoriness experiments, the UVC‐induced CREB/ATF‐1 phosphorylation represents an as yet unrecognized route of UVC‐induced signal transduction, independent of suramin‐inhibitable growth factor receptors and different from the Erk 1,2–p62TCF pathway.
Many of the cellular reactions to short wavelength radiation (UVC) such as skin erythema and carcinogenesis, UVC‐induced programmed cell death (apoptosis) and cellular survival seem to require UVC‐induced gene expression (Dévary et al., 1992; Friedberg et al., 1995; Schreiber et al., 1995). Two series of experiments support this interpretation. The changes of gene expression after UVC irradiation depend on the activity of protein tyrosine kinases (Dévary et al., 1992; Sachsenmaier et al., 1994). Accordingly, tyrosine kinase inhibitors prevent gene induction and also increase the sensitivity of cells to UVC killing (Dévary et al., 1992). The second type of evidence is based on elimination of a critical transcription factor. Both mutagenic reactions and survival are impaired in the absence of the transcription factor subunit Fos. Antisense c‐fos oligonucleotides suppress UVC‐induced chromosomal aberrations (van den Berg et al., 1991) and embryonic fibroblasts from c‐fos −/− mice show increased sensitivity to UVC (Schreiber et al., 1995; Haas and Kaina, 1995). In addition, UVC influences not only the programme of genes expressed in the irradiated cells, but also the fate of inner organismic cells that are not reached directly by UVC, by inducing the synthesis and secretion of growth factors (Schorpp et al., 1984; Rotem et al., 1987; Krämer et al., 1993).
The fact that adverse agents such as UVC can trigger complex macromolecule synthesis requires the existence of primary sensors which react with the adverse agent or absorb the radiation energy and then translate into cellular language, i.e. activate molecules normally involved in signalling and gene regulation. There has been only limited progress in unravelling the primary target(s) of UVC absorption relevant for signal transduction to the transcription factors. Comparison of dose–response curves for induced gene expression in DNA repair‐proficient and −deficient cells has yielded support for DNA damage as an intermediate of induction (Miskin and Ben‐Ishai, 1981; Schorpp et al., 1984; Stein et al., 1989). Thus DNA seems to be one of the relevant absorbing molecules. A second class of ‘sensors’ has been found recently in the plasma membrane by systematically following the signal pathways in the ‘upstream’ direction starting from transcription factors mediating the UV response (Mai et al., 1989; Dévary et al., 1992, 1993; Radler‐Pohl et al., 1993; Sachsenmaier et al., 1994). On the basis of the inhibition of the UV response by either suramin (a drug that poisons plasma membrane receptors) or by pre‐treatment of cells with growth factors (a procedure which induces, after transient activation, the inactivation and internalization of growth factor receptors), and from the finding that several growth factor receptors are phosphorylated at tyrosine residues within a fraction of a minute after UV irradiation, it has been concluded that growth factor receptors are part of the relevant pathway(s) used by UV to reach, in particular, the immediate response genes (Sachsenmaier et al., 1994; Knebel et al., 1996). According to these data, UVC‐induced transcription of c‐fos in HeLa cells is mediated to a large extent by epidermal growth factor receptor (EGFR), interleukin‐1 receptor (IL‐1R) and basic fibroblast growth factor receptor (bFGFR).
The c‐fos promoter elements required for UVC‐induced transcription have been delimited by deletion mutation (Büscher et al., 1988). The serum response element (SRE) is required and its deletion reduces induction to ∼50%. Indeed, the UV‐induced growth factor‐dependent signal transduction pathways lead to activation of the ternary complex factor p62TCF (Elk‐1; Sap‐1; Sachsenmaier et al., 1994; Price et al., 1996) which, in conjunction with the serum response factor (SRF), binds to the SRE (reviewed by Treisman, 1994). Using new promoter mutants, we now find that the response of the c‐fos promoter depends also on the presence of the cAMP response element (CRE)‐binding protein (CREB) binding site at position −60 of the c‐fos promoter. This not only indicates that promoter elements are complex cooperating assemblies but also shows the existence of still another UVC signalling pathway to c‐fos and possibly to other CRE‐regulated promoters. CREB and ATF‐1 are indeed phosphorylated in response to UVC, at serines 133 and 63 respectively. A p38/RK/HOG‐1‐dependent p108 serine/threonine protein kinase is identified as a UVC‐induced CREB kinase in HeLa cells. The UVC‐induced CREB and ATF‐1 phosphorylation, on the basis of suramin resistance and of various types of cross‐refractoriness experiments, represents a novel UVC signalling pathway independent of activation of the growth factor receptors described so far.
Several promoter elements contribute to the UVC response of the c‐fos promoter: the proximal CRE as a new UV response element
Dissection of the promoter has yielded evidence that the p62TCF/Elk‐1–(SRF)2 complex formed over the SRE (at positions −320/−300), is important for the UVC response of c‐fos (Büscher et al., 1988; Sachsenmaier et al., 1994; Price et al., 1996). When cloned in front of a heterologous UVC‐non‐responsive promoter, the SRE suffices to confer UVC inducibility, and inducibility of the c‐fos promoter in 3T3 cells is inhibited by 50% upon deletion of the SRE. To screen for other UVC targets in the c‐fos promoter, we examined deletion mutants of the proximal CRE (−60 CRE). In HeLa cells transiently transfected with c‐fos promoter–CAT reporter constructs, deletion of the CRE reduced both the UVC and the cAMP inducibility to <50% (Figure 1, compare −711/+45 fos CAT, lanes 1–3, with −711/ΔCRE/+45 fos CAT, lanes 7–9), implicating the CRE in mediating part of the response to both stimuli, and suggesting that factors so far known to be specifically activated by cAMP (Montminy et al., 1986) could also be addressed by UVC.
The importance of the CRE is documented further by experiments overexpressing a dominant‐negative mutant of CREB (CREBM1 which carries a serine→alanine substitution at position 133, Gonzales and Montminy, 1989). This mutant reduced by >50% both the UVC and the cAMP response of the full‐length promoter (Figure 1, compare lanes 1–3 with lanes 4–6). This suggests that a factor with binding ability similar to CREBM1 was involved in the UVC response and that CREBM1 could not be activated by UVC (nor by cAMP) treatment of cells. The specificity of the inhibition by CREBM1 is demonstrated by the fact that it does not interfere with the UVC induction of a NF‐κB‐dependent promoter construct (−105/−79 HIV TATA CAT; Figure 1, compare lanes 14 and 17). Similar roles for the −60 CRE and for a CRE‐binding factor in the UVC response of c‐fos were observed in NIH3T3 mouse fibroblasts and in JEG‐3 human chorionic carcinoma cells (not shown).
The residual cAMP response of the CRE deletion mutant was reduced further by CREBM1 (Figure 1, compare lanes 9 and 12), probably because of the existence of two upstream CRE‐like sequences located at positions −343/−336 and −295/−288 which presumably bind members of the CREB/ATF family (Berkowitz et al., 1989). Whether the upstream CREB/ATF‐binding elements do participate in the UVC response cannot be decided unequivocally, since the contribution of the transcription factor complex binding to the SRE dominates (Figure 1, compare lanes 5, 8 and 11), but they probably do. The remaining cAMP response seen in lane 12 may be due to a non‐homologous intragenic CRE which does not bind CREB (+18/+38, Härtig et al., 1991).
UVC induces the transactivating property of CREB
To show directly that the transactivating property of one of the CRE‐binding factors, CREB (Sassone‐Corsi et al., 1988), can be induced by UVC, we tested whether CREB fused to a different DNA‐binding domain, that of the yeast transcription factor GAL4, could be activated by UVC to initiate transcription of a GAL4‐dependent reporter. HeLa cells were transiently co‐transfected with expression vectors encoding either the DNA‐binding domain (DBD) of GAL4 or a chimeric protein consisting of the GAL4 DBD fused to the full‐length CREB protein (Figure 2A) or to a CREB protein lacking the bZip region (Figure 2B), together with a GAL4 element‐driven CAT reporter gene construct (Chrivia et al., 1993). Both cAMP and UVC promoted GAL4–CREB‐dependent transcription but not that by the GAL4 DBD alone (Figure 2), indicating efficient activation of CREB by both UVC and cAMP. With constructs containing the complete CREB including the bZip region, UVC induction was less efficient than cAMP induction while the reverse was found for the truncated CREB (compare Figure 2A and B). The reason for this difference is unknown.
Activation of CREB by cAMP depends on the phosphorylation of Ser133 by the cAMP‐dependent protein kinase A (PKA; Gonzales and Montminy, 1989; Taylor et al., 1990). Accordingly, a GAL4–CREB gene construct in which Ser133 was mutated to alanine responded less well to cAMP treatment (Figure 2B, lane 11). UVC induction of this hybrid transcription factor was reduced to the same extent (lane 10), indicating that Ser133 is also instrumental in UV‐induced transactivation of CREB. It is not clear how the residual activity in the Ala133 mutant is generated. Perhaps some CBP (see below) can interact with the mutant protein. Residual activation could then be caused by CBP activation (see below). Interestingly, a second phosphorylation site, Ser142, usually found to be inhibitory and the target of Ca2+/calmodulin‐dependent protein kinase II (Sun et al., 1994), does not seem to be addressed by UVC. Mutation of this serine to alanine did not significantly enhance forskolin‐ or UV‐induced transactivation of GAL4–CREB (Figure 2B, compare lanes 7 and 15 and lanes 6 and 14).
The activated CREB communicates with the basal transcription factor TFIIB through the bridging protein CBP (Kwok et al., 1994). A GAL4–CBP fusion protein should be able to activate transcription directly. When fused to the DNA‐binding domain of GAL4, the C‐terminal portion of CBP (amino acids 1678–2441) could indeed activate transcription, strongly suggesting the presence in this region of an autonomous spontaneously active transactivation domain. Interestingly, both cAMP and UVC enhanced transcription ∼2‐fold, suggesting that the cAMP‐ and UVC‐induced signalling pathways also caused an activating modification of CBP (Figure 2A).
In conclusion, our experiments demonstrate that UVC activates a signalling pathway at the end of which CREB (and possibly other CRE‐binding proteins) are located, and that CREB activation contributes to UVC‐induced c‐fos transcription.
UVC is a potent inducer of CREB Ser133 and ATF‐1 Ser63 phosphorylation
At the c‐fos promoter, the transcription factors are DNA bound at most or all times (Herrera et al., 1989). Activation must therefore involve post‐translational modification of the factors in the DNA‐bound state. The experiments with GAL4–CREB wild‐type or with the Ser133 mutation (Figure 2) and with the dominant‐negative CREBM1 (Figure 1) have already pointed to Ser133 as the site of modification after UVC as well as cAMP treatment. With cAMP as the inducer, Ser133 phosphorylation by the catalytic subunit of PKA (Gonzales and Montminy, 1989) has been shown to cause binding of CBP (Chrivia et al., 1993; Kwok et al., 1994). To examine directly the Ser133 phosphorylation state of CREB after UVC stimulation of cells, we used an antibody recognizing only the phosphorylated form of CREB: the amino acid sequence 123–136 (KRREILSRRPS133YRK, Herdegen et al., 1994) with Ser133 carrying the phosphate group, and investigated by immunofluorescence the presence of this epitope in NIH3T3 mouse fibroblasts. Since the phosphorylated epitope is conserved to 100% between CREB and its family member ATF‐1, the antibody used also recognizes ATF‐1 phosphorylated at Ser63 [this residue corresponds functionally to Ser133 of CREB (see sequence in Figure 3) and is phosphorylated by PKA; Lee and Masson, 1993, and references therein]. In untreated cells, the nuclei remained dark after staining with antibody (Figure 3A), consistent with the assumption that CREB and ATF‐1 were in the non‐phosphorylated state. Treatment of the cells with either forskolin or UVC (30 J/m2, a dose which leads to efficient c‐fos transcription in these cells) resulted in strongly positive nuclear staining consistent with the efficient phosphorylation of nuclear CREB and/or ATF‐1 at serines 133 and 63, respectively. A Western blot analysis with this antibody, using whole cell extracts, indeed revealed both cAMP‐ and UVC‐induced phosphorylation of the same proteins (Figure 3B), with the characteristic molecular weights of CREB (43 kDa, Montminy and Bilezikjian, 1987) and ATF‐1 (35 kDa, Hai et al., 1989). The β‐splice form of CREB (Blendy et al., 1996) and the CRE modulator protein CREM (Foulkes et al., 1991) were not detected. A protein of 80 kDa and of unknown nature reacted strongly with the antibody irrespective of the source of the extract. In both untreated and treated cells, this crossreactivity is equally distributed in nucleus and cytoplasm and there is no induced translocation (not shown).
We conclude that CREB and ATF‐1 are indeed phosphorylated at the appropriate serines after UVC irradiation, which can explain the mechanism of UVC‐induced activation of CRE‐binding proteins and their transcriptional role in the UV response of c‐fos.
UV‐induced CREB and ATF‐1 phosphorylation is an immediate early event. It occurs with a lag period of <5 min after irradiation of 3T3 (not shown) or HeLa cells, with initial phosphorylations already seen at 1 min (Figure 4). The kinetics and magnitude of CREB/ATF‐1 phosphorylation and of subsequent dephosphorylation after cAMP and UV resemble each other.
UV‐induced p108 CREB kinase
Which protein kinase(s) could be responsible for UVC‐induced CREB activation? A major criterion for its (their) identification could be that UVC enhanced its (their) activity. We therefore examined CREB‐accepting protein kinases by ‘in‐gel’ kinase assay (see Materials and methods). Nuclear extracts from untreated or from UVC‐irradiated HeLa cells were resolved by SDS–PAGE with specific substrate added prior to acrylamide polymerization: the non‐phosphorylated CREB peptide KRREILSRRPS133YRK (Figure 5, left two panels). In control gels, the peptide was omitted (Figure 5, right panel) or an irrelevant peptide was added to the gel (not shown). As one might expect from the resolution of total nuclear extracts, several protein kinase activities are detectable. Bands of ∼90 and 75 kDa autophosphorylate in the absence of CREB peptide (Figure 5, right panel). The important finding is that only two protein kinase activities are enhanced by UVC irradiation of the cells and that this enhancement depended entirely on the presence of the CREB peptide: a 90 kDa band (thin arrow) was increased after UVC treatment, a band of ∼108 kDa (thick arrow) is only visible in the lanes with extracts from UVC‐irradiated cells (Figure 5, lanes 2 and 7, marked with an asterisk). We did not detect renaturable UVC‐induced kinase activities at lower molecular weights (not shown). The UVC‐induced enhancements of protein kinase activities were transient, raised at 15 min (lane 7) and gone at 4 h (lane 8). Both protein kinases were not activated in cells treated with forskolin plus iso‐butylmethylxanthine (IMX) (Figure 5, lane 4, cAMP).
The protein kinases known to phosphorylate CREB, the activity of the catalytic subunit of PKA and the Ca2+/calmodulin‐dependent kinases, were not enhanced in UVC‐treated cells (not shown, below the molecular weight range of the gel in Figure 5). Experiments addressing the upstream pathway to be described below also speak against an involvement of either PKA or Ca2+/calmodulin‐dependent protein kinases.
CREB phosphorylation upon growth factor stimulation
Interestingly, the 90 and 108 kDa protein kinases were also transiently activated upon EGF treatment of cells (Figure 5, lanes 3 and 9). This finding prompted an examination of the CREB and ATF‐1 phosphorylation states in cells treated with one of several growth stimuli. CREB Ser133 and ATF‐1 Ser63 were indeed phosphorylated in response to phorbol ester, bFGF, IL‐1α and EGF, as well as UVC (Figure 6) and tumour necrosis factor (TNF)‐α (not shown). EGF also causes Ser133‐dependent activation of GAL4–CREB (Figure 2B, lanes 8, 12 and 16), although the enhancement was only slight. Thus CREB and ATF‐1 phosphorylation is not at all exclusive for elevated cAMP and it appears that several signalling pathways address CREB and ATF‐1, and probably involve the same protein kinases. These data match those of other laboratories demonstrating CREB Ser133 phosphorylation in PC12 cells in response to nerve growth factor (NGF) and in melanocytes in response to several growth factors (Ginty et al., 1994; Böhm et al., 1995). A protein kinase induced after NGF migrated with 105 kDa (Ginty et al., 1994) and could be identical with p108 CREB kinase.
Dissection of upstream signalling
Could one of these putative growth factor‐triggered pathways be relevant for the induction by UVC? We have shown previously that several UVC‐initiated signal transduction cascades inducing c‐fos and c‐jun gene transcription depend on the presence on the cell surface of activatable growth factor receptors (Sachsenmaier et al., 1994). UVC strongly stimulates ligand‐independent tyrosine phosphorylation of several receptor tyrosine protein kinases (Sachsenmaier et al., 1994; Coffer et al., 1995; Knebel et al., 1996). To try and identify putative signalling components shared by the UVC response of CREB, specific down‐modulation experiments were performed (Sachsenmaier et al., 1994) and inhibitors were applied.
Specific down‐modulation experiments are based on the following principle: a stimulus, e.g. EGF, activates components of a signal transduction pathway in a transient manner. In the case of EGF, the EGFR undergoes autophosphorylation followed by receptor internalization. Restimulation by the same stimulus, e.g. EGF, is not possible for some period of time. Such down‐modulation has been found for many stimuli. Pre‐treatment of cells with one stimulus makes cells refractory to a second stimulation by the same agent. A second different stimulus can, however, be successful if it utilizes a component other than the down‐modulated one. Refractoriness to UVC would result if UVC required one of the down‐modulated components of the previous stimulus.
One problem in such down‐modulation experiments is the choice of downstream endpoints, because restimulation can only be measured using components with transient induction kinetics. Suitable transiently active components are the transcription factors, e.g. CREB and ATF‐1, and protein kinases, e.g. Erk 1,2. In the experiment shown as Figure 7, we show first the induction by EGF of Erk 1,2 activation (at 15 min, Figure 7A, lane 6), then the decay of activity (the EGF‐induced Erk 1,2 activation has disappeared at 4 h; Figure 7A, lane 7) followed by demonstration of refractoriness to the same stimulus (Figure 7A, lane 8). As described, stimulation with UVC as the second stimulus should then clarify whether the two stimuli share signalling components (Figure 7A, lane 9). This type of analysis has demonstrated that EGF alone (Figure 7A, lane 9) or a cocktail of growth factors (EGF, IL‐1α and bFGF; Figure 7B, lane 9) cannot inhibit UVC‐induced CREB and ATF‐1 phosphorylation, while blocking the UVC‐dependent reactivation of Erk 1,2 in the same cells (Sachsenmaier et al., 1994; and Figure 7A, lane 9). This indicates that the UVC‐induced pathway to CREB and ATF‐1 differs from those used by the growth factor receptors, and from that used by UVC in the activation of Erk 1,2. Accordingly, a specific inhibitor of the EGFR, AG1478 (Levitzki and Gazit, 1995), inhibited UVC‐ and EGF‐induced EGFR phosphorylation (not shown, see Knebel et al., 1996) and Erk 1,2 activation (Figure 7D, lower panel), but not the UVC‐induced phosphorylation of CREB and ATF‐1 (Figure 7D, upper panel). Down‐modulation experiments similar to the one above indicate that the UVC‐dependent pathway to CREB and ATF‐1 does not share components with those of TNF‐α, of IMX plus forskolin (both not shown) and of phorbol ester (Figure 7C).
Suramin, a drug that poisons plasma membrane receptors (Betsholtz et al., 1986; Coffey et al., 1987; Fantini et al., 1989), has been used to inhibit UVC‐induced reactions which require the function of growth factor receptors (Sachsenmaier et al., 1994). The UVC‐induced CREB/ATF‐1 phosphorylation turned out to be suramin‐resistant (Figure 8A), while UVC‐induced Erk 1,2 activation was sensitive (Figure 8B, see also the control with growth factor‐rich conditioned medium, EPIF, from UV‐irradiated cells whose effect is suramin‐sensitive, Figure 8 and Krämer et al., 1993). That suramin still inhibits most of the UV‐induced c‐fos transcription (Sachsenmaier et al., 1994) can be explained by assuming that the suramin‐resistant pathway to the CRE does not on its own suffice to induce c‐fos transcription. This is indeed the case for UVC induction (not shown; and Büscher et al., 1988; and see Discussion) as well as for induction by NGF (Ginty et al., 1994).
The possibility remained that UVC triggered the release of calcium from endoplasmic reticulum or the entry from extracellular sources in a growth factor receptor‐independent manner, increasing intracellular calcium and thereby activating Ca2+/calmodulin‐dependent protein kinases which accept CREB and ATF‐1 as substrates. Elevation of the intracellular Ca2+ concentration indeed leads, in numerous cell types, to transcriptional activation of c‐fos both through the SRE (Bading et al., 1993) and through the CRE (Sheng et al., 1990; Bading et al., 1993). Consistently, increased cytosolic calcium was found to cause both Erk 1,2 activation (Bading and Greenberg, 1991; Chao et al., 1992; Fiore et al., 1993; Rosen et al., 1994) and CREB Ser133 phosphorylation (Sheng et al., 1990, 1991). Indeed, disturbing intracellular Ca2+ levels by treatment of cells with the Ca2+ ionophore A23187 (Pressman, 1976) inhibited UV‐induced c‐fos transcription (Büscher et al., 1988) and UV‐induced CREB phosphorylation (not shown). However, this result may be due to a basic need for Ca2+ rather than to UV‐induced activation of Ca2+‐dependent signalling components, since UVC irradiation, in contrast to the Ca2+ ionophore A23187, did not induce a detectable rise in intracellular Ca2+ levels in HeLa cells at doses between 30 and 2000 J/m2. Ca2+ was measured by spectroscopic analysis with the fluorescent‐free Ca2+‐binding dye fluor‐3 and by video‐imaging with the same dye (not shown; Iordanov, 1995). It is unlikely, therefore, that UVC‐induced CREB and ATF‐1 phosphorylations are mediated through a UVC‐induced rise in Ca2+ concentration and through the activation of known Ca2+/calmodulin‐dependent protein kinases.
All these data suggest that the UVC‐induced signalling pathway to CREB and ATF‐1 does not overlap with those examined above. In particular, the receptor tyrosine kinase/Erk 1,2 pathway is clearly different from the pathway to CREB and ATF‐1. Suramin resistance implies that either a receptor at the cell surface was not involved, or that this specific receptor could not be blocked by suramin.
p38/HOG‐1 dependence of UVC‐induced p108 CREB kinase
To define the new UV pathway to CREB, we attempted to link the CREB kinases to known components of signal transduction. p90 CREB kinase is active in control cells prior to UVC or EGF treatment (in contrast to its in vitro substrates CREB and ATF‐1 which are not phosphorylated prior to such treatment, see Figure 5, lanes 1, 5 and 6) and therefore perhaps not the prime candidate for the in vivo response to UVC. Nevertheless, we considered its contribution to CREB and ATF‐1 phosphorylation in vivo. The molecular weight obviously corresponds to that of p90RSK1. p90 CREB kinase migrates in PAGE to a position identical with p90RSK1 as revealed in a Western blot type experiment with antibodies directed against p90RSK1 (compare Figure 9C with D, see legend for experimental details). The activation of p90 CREB kinase by UVC or EGF goes along with the disappearance of the faster migrating species of p90RSK1 (Figure 9D). Thus this modification of p90RSK1 could be the physical basis of p90 CREB kinase activation. p90RSK1 has been shown to be activated by the Ras–Raf–MEK1,2–Erk1,2 pathway (for review, see Treisman, 1996) which is activated by UVC with kinetics preceding the rapid induction of c‐fos transcription. Could UV‐induced CREB and ATF‐1 phosphorylation be mediated through the UV‐induced Ras–Raf–MEK1,2–Erk1,2, p90RSK pathway? This is not likely since under conditions of inhibition of this pathway (pre‐treatment of cells with growth factors, phorbol esters or suramin, specific receptor inhibition) when Erk1,2 is not activatable by UVC, UVC‐induced CREB phosphorylation is not affected.
Thus p90 CREB kinase is either not relevant for CREB and ATF‐1 phosphorylation in vivo or it is part of still another pathway. In addition to the Raf–MEK1,2–Erk 1,2 pathway, UV irradiation strongly activates the MEKK1–MKK4–JNK/SAPK pathway (Hibi et al., 1993; Dérijard et al., 1994). The kinetics are slower than those of CREB phosphorylation (D.Wilhelm and P.Angel, personal communication; Price et al., 1996). Further, UV also activates p38/RK/HOG‐1 (Raingeaud et al., 1995; and Figure 9A; for reviews on proline‐directed kinases, see Cano and Mahadevan, 1995; Treisman, 1996). In order to distinguish between the JNK/SAPK pathway and reactions depending on p38/RK/HOG‐1, we used an inhibitor (SB203580) which specifically inhibits p38/RK/HOG‐1 without affecting Erk1,2 or JNKs/SAPKs (Lee et al., 1994; Cuenda et al., 1995). SB203580 reduced significantly the UVC‐ and EGF‐induced CREB and ATF‐1 phosphorylations (Figure 9B), without interfering with several other UV‐induced reactions such as c‐Jun phosphorylation (Radler‐Pohl et al., 1993), NF‐κB activation (Stein et al., 1989) and IκBα degradation (K.Bender, unpublished; all not shown). Most importantly, SB203580 did not interfere with the UVC‐induced activation of the p90 CREB kinase and the modification of p90RSK1 (Figure 9C and D). As one might have expected, SB203580 did not affect forskolin‐induced CREB phosphorylation either (not shown). These data indicate that the UVC action on CREB and ATF‐1 depends on p38/HOG‐1, and that p90RSK1 (and the p90 CREB kinase) is not part of the pathway.
Recently, a new protein kinase has been shown to be activated in conjunction with p38/RK/HOG‐1: the 50 kDa MAPKAP kinase 2 (Stokoe et al., 1992; see Cuenda et al., 1995 for references). MAPKAP kinase 2 is the major enzyme responsible for the phosphorylation of the small heat shock proteins in cells treated with IL‐1, arsenite or sorbitol. Although the ‘in gel’ kinase assay revealed no band of this size, we examined the putative involvement of MAPKAP kinase 2 in UV‐induced CREB and ATF‐1 phosphorylation, a possibility made plausible by its activation in stressed cells. A constitutively active form of MAPKAP kinase 2, GST–MAPKAP kinase 2 Δ3BPC (Engel et al., 1995), renatures only poorly in an ‘in gel’ kinase assay with Hsp25 as substrate (M.Cahill, A.Nordheim and M.Gaestel, unpublished), suggesting that this may also be the case for the UV‐regulated enzyme. In order to assay for the activation of MAPKAP kinase 2 in UV‐irradiated cells and for its activity towards CREB, we immunoprecipitated the kinase from HeLa cell extracts at different time points after UV irradiation and offered His‐tagged CREB or Hsp25 as target proteins. Indeed, UV irradiation activated MAPKAP kinase 2 strongly when tested with its substrate Hsp25, detectable at 2 min after UV irradiation and inhibited by the p38/RK/HOG‐1 inhibitor SB203580 (Figure 10A, lanes 1–6). Also, treatment of cells with EGF activated the kinase through the p38/RK/HOG‐1 pathway (Figure 10A, lanes 7 and 8). In contrast, phosphorylation of His‐tagged CREB protein by the immunoprecipitated MAPKAP kinase 2 was not detectable (not shown). We investigated, therefore, whether constitutively active MAPKAP kinase 2 produced in and purified from Escherichia coli accepted the CREB protein in vitro. While the recombinant enzyme phosphorylated the Hsp25 protein strongly [Figure 10A, lane 9; as shown earlier (Gaestel et al., 1991) Hsp25 is also phosphorylated, to a lower extent, by PKA, lane 10], it phosphorylated the CREB protein only poorly (Figure 10B, lane 1; which was, however, a good substrate for PKA, Figure 10B, lane 2). Table I summarizes the enzymatic parameters for MAPKAP kinase 2 towards the relevant substrate peptides derived from either Hsp25 or CREB, in comparison with PKA. The data show that MAPKAP kinase 2 phosphorylates the CREB peptide much less efficiently than the peptide derived from the Hsp25 protein. We conclude that MAPKAP kinase 2 is strongly activated in UV‐irradiated HeLa cells, but that CREB (and possibly ATF‐1) is only a poor substrate for this enzyme in vitro. Thus MAPKAP kinase 2 may participate but is unlikely to account for the massive phosphorylation of CREB and ATF‐1 after UV irradiation.
Could p108 be the decisive kinase? Several arguments speak for the in vivo significance of p108 in the UVC response of CREB. UVC activation of p108 CREB kinase is also strongly inhibited by SB203580 (Figure 9C), as is UVC‐induced CREB and ATF‐1 phosphorylation (Figure 9B). Also, activation of p108 CREB kinase in response to EGF is inhibited by SB203580 (Figure 9C). Thus p108 must be downstream of p38/RK/HOG‐1. Another argument concerns the strict correlation of induced p108 activity and CREB (and ATF‐1 phosphorylation) after stimuli (EGF, UVC, but not cAMP), and the absence of activity prior to treatment. Thus p108 is likely to contribute the largest portion of the relevant UVC response to CREB.
We show here that the transcription factors CREB and ATF‐1 are phosphorylated at serines 133 and 63, respectively, when cells are treated with one of various growth factors (EGF, bFGF, IL‐1α, TNF‐α, and conditioned medium from UVC‐treated cells, EPIF, Schorpp et al., 1984) or with the phorbol ester 12‐O‐tetradecanoylphorbol‐13‐acetate (TPA). It has been shown previously that CREB is also activated by NGF in PC12 cells (Ginty et al., 1994) and by several growth factors in melanocytes (Böhm et al., 1995). Our central point is the identification in several cell lines of CREB and ATF‐1 as functional targets of a novel signal transduction chain elicited by short wavelength UV irradiation (UVC). CREB is thus not only a target for a limited set of signals (that work through cAMP and Ca2+ as ‘second messengers’) but is addressed by numerous stimuli.
The UVC response pathway to CREB/ATF‐1 involves p38‐HOG‐1
Within 15 min after treatment of HeLa cells with cAMP‐ or Ca2+‐elevating agents, with phorbol esters, with one of several growth factors or with UVC, the transcription factors CREB and ATF‐1 are phosphorylated at serines 133 and 63, respectively. The kinases responsible differ between different inducers: in the case of cAMP‐elevating agents it is PKA (Gonzales and Montminy, 1989); Ca2+‐elevating agents induce the Ca2+/calmodulin‐dependent kinases II and IV (Dash et al., 1991; Sheng et al., 1991; Matthews et al., 1994; Sun et al., 1994). Mitogens and UVC are known to activate the proline‐directed protein kinases Erk 1,2, JNKs/SAPKs and p38/RK/HOG‐1 which accept the substrate sequence (P)‐X‐S*/T*‐P‐X (reviewed by Karin, 1994; Treisman, 1996). The sequence around CREB Ser133 and ATF‐1 Ser63 reads RRPS*YRK and is not likely to be accepted by proline‐directed protein kinases. As shown here, growth factors and UVC activate CREB (and ATF‐1)‐accepting protein kinases in the 90–110 kDa range as well as MAPKAP kinase 2. The p90RSK1 has been implicated previously in Ser133 phosphorylation in response to certain growth factors in human melanocytes (Böhm et al., 1995); and NGF activates, in PC12 phaeochromocytoma cells, a p105 CREB kinase (Ginty et al., 1994). Here, we have identified a UVC‐ and EGF‐induced p108 CREB‐specific protein kinase from HeLa cells and we show that a UV‐ and EGF‐activated protein kinase with the molecular weight of p90RSK1 also seems to accept CREB in vitro. In contrast, MAPKAP kinase 2, which is strongly activated in UV‐irradiated cells, phosphorylates CREB in vitro only poorly, as compared with its preferred substrate, the small heat shock proteins.
Although proline‐directed protein kinases will not phosphorylate CREB, CREB kinases may well be downstream of proline‐directed kinases. We therefore considered whether the p42 and p44 mitogen‐activated protein kinases (Erk 1,2; reviewed by Marshall, 1994), the p46 and p54 Jun N‐terminal kinases (JNKs, Hibi et al., 1993; Dérijard et al., 1994; also referred to as stress‐activated protein kinases, SAPKs, Kyriakis et al., 1994) and the p38/RK/HOG‐1‐kinase (also named CSBP kinase, Lee et al., 1994; for review, see Cano and Mahadevan, 1995) addressed one of the CREB kinases detected in UV‐irradiated cells. All three proline‐directed protein kinases are activated in UV‐irradiated cells (Hibi et al., 1993; Radler‐Pohl et al., 1993; Sachsenmaier et al., 1994; Raingeaud et al., 1995; and Figure 9A). While Erk 1,2 and p38 are activated rapidly after UVC (Radler‐Pohl et al., 1993; Sachsenmaier et al., 1994; M.Iordanov, unpublished; Price et al., 1996), JNK activation may be too slow (D.Wilhelm and P.Angel, personal communication; Price et al., 1996) to account for UVC‐induced CREB phosphorylation which is detectable in minutes after treatment (Figure 4). In order to distinguish between the Erk 1,2 and the p38/HOG‐1 pathway, we used various cellular conditions in which the Erk 1,2 pathway was not activatable and asked whether CREB still responds to UVC. Such conditions included treatment with the growth factor receptor poison suramin, pre‐treatment of cells with growth factors and treatment with an inhibitor of the EGF receptor tyrosine kinase, AG1478. In all these cellular conditions, in which Erk 1,2 was not activatable by UVC, UVC was fully effective in inducing CREB/ATF‐1 phosphorylation. These findings strongly suggest that the Erk 1,2 pathway was not or not to a significant extent involved in UVC‐induced CREB/ATF‐1 phosphorylation. Interestingly, p90RSK is activated by Erk 1,2 (Sturgill et al., 1988; for review, see Blenis, 1993; Treisman, 1996), and, therefore, perhaps not part of the UVC‐induced pathway to CREB.
Further support that neither the Erk 1,2 pathway nor the JNK pathway are involved, but that UVC reaches CREB through p38/RK/HOG‐1 came from the use of the inhibitor of p38/RK/HOG‐1, SB203580 (Lee et al., 1994; Cuenda et al., 1995). The inhibitor blocked p108 CREB kinase and MAPKAP kinase 2 activation, as well as CREB and ATF‐1 phosphorylation, but did not affect the UVC‐induced activation of the p90 CREB kinase nor the UVC‐induced modification of p90RSK1. Irrespective of the specificity of the inhibitor, SB203580 dissociates the UV‐induced activation of p90 CREB kinase/p90RSK1 from UV‐induced CREB and ATF‐1 phosphorylation. The inhibitor did not interfere with UV‐induced c‐Jun modification, phorbol ester‐induced c‐Fos modification, UV‐induced degradation of IκBα and forskolin‐induced CREB/ATF‐1 phosphorylation. We conclude that p108 CREB kinase, MAPKAP kinase 2 and CREB/ATF‐1 are downstream of UV‐induced p38/RK/HOG‐1 activation. A dominant‐negative mutant of MAPKAP kinase 2 (M.Müller and M.Gaestel, unpublished) interfered with UV‐induced transactivation of a transiently transfected GAL4–CREB protein (not shown), which would be compatible with a role for MAPKAP kinase 2 in the pathway. However, the mutant seems also to affect other p38‐dependent activities (e.g. p62TCF activation) and the PKA‐dependent phosphorylation of CREB. The experiment therefore does not allow for a definitive conclusion on the order of components and on the role of MAPKAP kinase 2.
We conclude that both enzymes, p108 and MAPKAP kinase 2, may account for UV‐induced CREB/ATF‐1 phosphorylation, with MAPKAP kinase 2 probably less significant because of its low activity on CREB in vitro. Irrespective of which protein kinase carries out UVC‐induced, p38‐dependent CREB phosphorylation (p108 CREB kinase or MAPKAP kinase‐2), our findings allow the important conclusion that mitogen‐activated protein kinases do not always phosphorylate target transcription factors directly. They can act through downstream protein kinases.
What is upstream of p38?
Our data indicate that a new UVC response pathway leads to the activation of p38 and hence to the phosphorylation of CREB/ATF‐1. It is not clear what is upstream of p38. Two UVC pathways have been delimited in the past: DNA damage‐dependent processes (e.g. p53 stabilization; Lu and Lane, 1993; Nelson and Kastan, 1994; Yamaizumi and Sugano, 1994; late responding genes; Miskin and Ben‐Ishai, 1981; Schorpp et al., 1984; Stein et al., 1989) and the activation of growth factor receptor pathways (Dévary et al., 1992; Radler‐Pohl et al., 1993; Sachsenmaier et al., 1994; Knebel et al., 1996). As detailed above, in various conditions in which growth factor receptors were down‐modulated, UVC could still induce CREB/ATF‐1 phosphorylation. This suggests that either suramin‐resistant growth factor receptors do exist, or that UV addresses new, possibly intracellular targets. We have found previously that UVC induces elevated phosphorylation of growth factor receptors at tyrosine residues by inhibiting the dephosphorylation of these receptors, presumably by inactivating the SH group in the catalytic centre of protein tyrosine phosphatases (Knebel et al., 1996). As a result, UVC may prolong not only the lifetime phosphorylated (activated) receptor tyrosine kinases but also that of intracellular protein tyrosine kinases and other tyrosine phosphate‐bearing proteins. Such a general mechanism could explain how so many adverse agents can trigger Erk 1,2, JNK and p38 pathways. For instance, efficient activation of the EGFR–Erk 1,2 pathway can be achieved by oxidants and an SH‐alkylating agent (Knebel et al., 1996), JNK activation has been observed after treatment of cells with N‐methyl‐N′‐nitro‐N‐nitrosoguanidine (MNNG) or methyl methanesulfonate (MMS) (van Dam et al., 1995) and CREB activation has been postulated to be responsible for the transcription of the β–DNA polymerase gene after MNNG, MMS, N‐acetoxy‐acetylaminofluorene and H2O2. (Fornace et al., 1989; Kedar et al., 1991). The β‐DNA polymerase gene could, however, according to these reports, not be activated by UVC.
A difference between single CRE promoters and complex enhancers
Another observation is not yet understood. UVC (as shown here) and NGF (Ginty et al., 1994) cause similarly effective CREB phosphorylation at Ser133 and GAL4–CREB activation as does cAMP. Moreover, CREB and possibly other CRE‐binding proteins contribute to the transcriptional activation of the c‐fos gene by UVC (as shown here) and by NGF (Ginty et al., 1994; Bonni et al., 1995). Strangely, however, both UVC (not shown) and NGF (Ginty et al., 1994; Bonni et al., 1995) induce a CRE‐driven reporter only poorly, in contrast to cAMP. The UVC‐induced contribution of CREB to c‐fos transcription also cannot be measured in promoter constructs lacking the SRE and the +18/+38 element (Büscher et al., 1988). We have not yet been able to find a difference in phosphorylation between CREB activated by cAMP and that activated by UVC. In particular, there is no inhibitory Ser142 phosphorylation (P.Sassone‐Corsi, personal communication) as has been observed as a result of activation of calmodulin‐dependent protein kinase type II (Sun et al., 1994). Accordingly, mutation of Ser142 to alanine does not enhance UV‐induced GAL4–CREB activation (Figure 2B). Also we have no evidence that the activation of CBP differs between UVC and cAMP (Figure 2A). We need to postulate, however, that an as yet undetected difference exists. This difference may distinguish single CRE promoters from complex enhancers where many transcription factors interact. Obligatory collaboration between transcription factors has been reported previously, e.g. Ets with AP‐1 in the polyoma promoter (Wasylyk et al., 1990), and p62TCF with SRF at the c‐fos SRE (Treisman, 1994). UVC‐ or NGF‐activated CREB may represent an example of conditional dependence on collaborating factors. The UVC‐ or NGF‐modified form of CREB must collaborate, the ‘cAMP form’ does not need to. The prominent collaborating partner is the p62TCF–(SRF)2 complex. The c‐fos promoter may represent an exquisite example of such collaborative assembly in that transgenic experiments in mice and derived cell lines have documented that most promoter elements are required for any c‐fos promoter function (Robertson et al., 1995). The complex assembly of the c‐fos enhancer–promoter (as well as of other complex promoters) may be built to contract multiple signalling pathways and to permit nevertheless fine‐tuning in that the relative contributions of individual stimuli depend strongly on the context of the other activated transcriptional elements.
Materials and methods
The −711/+45 fos CAT and −711/Δ−65/−52/+45 fos CAT constructs (Härtig et al., 1991) and the −105/−79 HIV TATA CAT construct (Stein et al., 1989) have been described. The CREBM1 expression vector has been described by Gonzales and Montminy, (1989). Rous sarcoma virus (RSV) GAL4 1–147 and RSV GAL4–CBP (1678–2441), as well as the reporter GAL4–CAT were as reported (Chrivia et al., 1993; Kwok et al., 1994). RSV GAL4–CREB is described in Sun et al. (1994), and GAL4–CREBΔbZip and the respective mutants in Ser133 and 142 in Sun and Maurer (1995). Constitutively active MAPKAP kinase 2 was expressed in E.coli using the vector pGEX‐5X‐3‐MK2 Δ3BPC (Engel et al., 1995).
Cell culture and transient transfections
HeLa tk− and NIH3T3 cells were grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal calf serum (FCS), penicillin (100 U/ml), and streptomycin (100 μg/ml). Transfection of HeLa tk− cells, using the DEAE‐dextran method, followed by chloramphenicol acetyltransferase (CAT) assays were performed as described (Angel et al., 1987) except that the chloroquine step was omitted. CAT activity was quantified using a Fuji Bio‐Imaging Analyzer.
Chemicals, growth factors and treatment of cells
TPA (Sigma) was stored as a 200 μg/ml stock solution in dimethylsulfoxide (DMSO, Fluka) at −20°C and was applied at a final concentration (f.c.) of 200 ng/ml. EGF (human, recombinant, from Saccharomyces cerevisiae; Sigma), bFGF (human, recombinant, from E.coli; British Biotechnology), IL‐1α (human, recombinant, from E.coli; British Biotechnology) and TNF‐α (Saxon Biochemicals) were stored as 10 μg/ml stock solutions in phosphate‐buffered saline (PBS)/1% bovine serum albumin (BSA) at −20°C. Forskolin and IMX (Sigma) were stored as 10 mM and 0.5 M stock solutions in DMSO, respectively, at −20°C. In order to stimulate cAMP‐dependent signal transduction, the cells were treated with a combination of forskolin and IMX (10 μM and 0.5 mM f.c.). Suramin (Germanin, Bayer) (0.3 mM f.c. in DMEM) was freshly prepared and used only once. The EGFR inhibitor AG1478 was stored in DMSO at 4°C at a concentration of 100 μM. EPIF medium was prepared as described previously (Krämer et al., 1993). UVC irradiation of cells was performed as described previously (Radler‐Pohl et al., 1993), except that the PBS washing step was omitted. The p38/RK/HOG‐1 inhibitor SB203580 was a generous gift of Dr J.C.Lee, King of Prussia, PA.
Antibodies and Western blotting procedure
The antibody used to detect phosphorylated CREB is an affinity‐purified rabbit polyclonal serum raised against a CREB peptide phosphorylated at Ser133 (peptide 123‐KRREILSRRS133YRK‐136; Herdegen et al., 1994). It recognizes only CREB phosphorylated at Ser133 and ATF‐1 phosphorylated at Ser63 but does not cross‐react with the non‐phosphorylated forms of these proteins. The anti‐p42 MAP kinase‐specific antibody αIcp42 was a generous gift of Dr C.Marshall, Institute of Cancer Research, London, UK. The anti‐Erk 1,2 (K‐23)‐, the anti‐p38/RK/HOG‐1 (#sc‐535)‐ and the anti‐p90RSK1 (# sc‐231‐G)‐specific antibodies were from Santa Cruz Biotechnology. The anti‐phosphotyrosine‐specific antibody was given by Dr P.Druker, Portland. The anti‐MAPKAP kinase 2 antibody was raised in rabbits against recombinant GST–MAPKAP kinase 2 Δ3B (Plath et al., 1994). Anti‐rabbit immunoglobulins from goat and anti‐mouse immunoglobulins from rabbits, both peroxidase‐coupled, were from Dako (Glostrup, Denmark) and detected using the enhanced chemiluminescence method (Amersham). Whole cell lysates from treated or untreated cells for Western blot analyses were prepared by lysing the cells directly in 400 μl of boiling Laemmli buffer per 10 cm Petri dish. Fifty μl of the lysate were resolved by 10% SDS–PAGE. Equal loading was ascertained by staining with Coomassie blue. Electrophoretic transfer and detection with specific antibodies were performed as described previously (Radler‐Pohl et al., 1993; Sachsenmaier et al., 1994).
Immune complex kinase assays of MAPKAP kinase
A total of 1×106 HeLa tk− cells were harvested, lysed in 100 μl of lysis buffer [20 mM Tris acetate, pH 7.0, 0.1 mM EDTA, 1 mM EGTA, 1 mM Na3VO4, 10 mM β‐glycerophosphate, 50 mM NaF, 5 mM pyrophosphate, 1% Triton X‐100, 1 mM benzamidine, 0.27 M sucrose, 0.1% β‐mercaptoethanol, 0.2 mM phenylmethylsulfonyl fluoride (PMSF)] on ice for 15 min and centrifuged at 15 000 g for 15 min. Then 25 μl of the supernatant was diluted in 500 μl of immunoprecipitation buffer [IP buffer plus BSA; 20 mM Tris–HCl, pH 7.4, 150 mM NaCl, 50 mM NaF, 1% Triton X‐100, 1% (w/v) BSA)] and incubated with 5 μl of a rabbit antiserum against MAPKAP kinase 2 at 4°C overnight. Immunocomplexes were precipitated by incubation with 25 μl of 50% slurry of protein A–Sepharose in PBS (Pharmacia, Uppsala, Sweden) for 1 h. The IP pellet was washed four times with 500 μl of IP buffer minus BSA and redissolved in 25 μl of assay mixture (50 mM β–glycerophosphate, 0.1 mM EDTA, 0.1 mM ATP, 10 mM magnesium acetate, 1.5 μCi of [γ‐33P]ATP) containing 2 μg of recombinant Hsp25 (Gaestel et al., 1989). The kinase reaction was incubated for 15 min at 30°C. The reaction was terminated by addition of 8 μl of 4× SDS loading buffer. 33P labelling of the substrate was detected and quantified after SDS–PAGE using the Bio Imaging Analyser BAS 2000 (Fuji, Tokyo, Japan).
Peptide assay and in vitro phosphorylation of proteins
Assays using the peptides KKKALNRQLSVAAamide and KRREILSRRPSYRKamide and recombinant MAPKAP kinase 2 Δ3BPC (Engel et al., 1995) or the catalytic subunit of PKA (Sigma) were performed as phosphocellulose filter binding assays (Stokoe et al., 1992). Enzymatic parameters were calculated according to Lineweaver–Burk analysis.
Two μg of recombinant Hsp25 (Gaestel et al., 1989) or His‐CREB were incubated with either 0.5 μg of MAPKAP kinase 2 or 0.1 μg of PKA catalytic subunit (Sigma) in the phosphorylation assay mixture for 15 min at 30°C. The reaction was stopped by addition of 8 μl of 4× SDS loading buffer. Phosphorylation products were analysed by SDS–PAGE and subsequent Phospho‐Imaging.
Preparation of nuclear extracts, ‘in gel’ kinase assays and ‘on membrane’ kinase assays
Nuclear extracts were prepared according to Dignam et al. (1983) with modifications. HeLa tk− cells were grown to confluency in 10 cm Petri dishes and then serum starved (0.5% FCS) for 24 h. After treatment, the medium was removed, the cells were washed twice with ice‐cold PBS (137 mM NaCl, 2.7 mM KCl, 0.7 mM CaCl2 , 6.5 mM NaH2PO4, 1.5 mM KH2PO4, pH 7.6) and were harvested in 5 ml of ice‐cold PBS. All subsequent steps were carried out at 4°C. The cells were centrifuged at 250 g. The volume of the cell pellet was determined (packed cell volume = PCV). The cells were resuspended in 3× PCV buffer A [10 mM HEPES pH 7.9, 1.5 mM MgCl2, 10 mM KCl, 0.5 mM dithiothreitol (DTT), 0.5 mM PMSF, 10 μg/ml of each aprotinin, leupeptin and pepstatin] and incubated on ice for 15 min. The cells were lysed by pushing them five times through a 26 gauge needle and the nuclei were pelleted by centrifugation at 5600 g for 45 s. The nuclei were resuspended in 1.5× PCV buffer C (20 mM HEPES pH 7.9, 25% glycerol, 1.5 mM MgCl2, 420 mM NaCl, 200 mM EDTA, 0.5 mM DTT, 0.5 mM PMSF, 10 μg/ml of each aprotinin, leupeptin and pepstatin) and rotated for 30 min. The nuclear membranes were pelleted by centrifugation at 5600 g for 5 min. The supernatant (nuclear extract) was diluted 1:1 in 2× Laemmli buffer. Five to 10 μl of nuclear extract were used for ‘in gel’ kinase assays which were performed as described previously (van Dam et al., 1995) except that the proteins were resolved by 7.5% SDS–PAGE with or without co‐polymerized substrate (400 μg/ml separating gel solution). For ‘on membrane’ kinase assays, CREBtide was co‐polymerized in a 5 mm thick 1% agarose gel (prepared in 1× electrotransfer buffer for SDS–PAGE gels) at a concentration of 400 μg/ml, and then electrotransferred onto Immobilon™ membrane. Whole cell lysates (prepared as for Western analysis) from untreated and treated HeLa tk− cells were separated in a 7.5% SDS–PAGE and then electrotransferred onto the membrane with immobilized CREBtide. The membrane was then subjected to the same experimental procedure as for the ‘in gel’ kinase assay.
NIH3T3 cells were plated out at low density onto round cover slips (diameter 10 mm) in 6 cm Petri dishes, grown for 24 h and then serum starved for an additional 24 h. At 15 min post‐treatment with forskolin or UVC, the cells were washed with PBS and were fixed to the cover slips with 4% p‐formaldehyde at room temperature for 8 min. After washing three times with PBS, the cells were permeabilized with 0.2% Triton X‐100 for 10 min, and washed again four times with PBS. In order to prevent unspecific binding of the antibody, the cells were incubated with 10% FCS in PBS for 10 min. The cover slips were washed three times with PBS and transferred into 24‐well tissue culture plates (one slip per well). The cells were then incubated with 30 μl of 1:1000 dilution of the first antibody in PBS/10% FCS at room temperature for 30 min. After washing three times with PBS, the cells were incubated with a 1:50 dilution of the rhodamine‐coupled secondary antibody (anti‐rabbit IgG, from goat, Dianova) at room temperature for 30 min. The cover slips were washed three times with PBS and once with H2O, and then mounted onto microscopic slides using pre‐warmed (50°C) glycergel (Dako). The microscopic analysis of the slides was carried out using a Diaplan fluorescent microscope (Zeiss) and photographs were taken using Kodak Ektachrome P800/1600 film.
After submission of this manuscript, two reports characterizing CREB‐phosphorylating protein kinases appeared. Xing et al. (Science, 273, 959) identified p90RSK family member RSK2 as a functional CREB kinase in phorbol ester‐treated K562 cells and in NGF‐treated PC12 cells. Tan et al. (EMBO J., 15, 4629–4642) demonstrated that in FGF‐ or arsenite‐treated human neuroblastoma SK‐N‐MC‐cells, MAPKAP kinase 2 is the only detectable SB203580‐sensitive CREB kinase.
We thank Dr R.Goodman, Portland, for the gene constructs pRSVGAL4 1–147, pRSVGAL4 CBP and pGAL4CAT; Dr R.Maurer, Portland, for the GAL4–CREB gene constructs and for the recombinant His‐CREB protein; Dr B.Druker, Portland, for the anti‐phosphotyrosine specific antibody; Dr C.Marshall, London, for the anti‐MAP kinase antibody; Dr A.Ullrich, München, for the inhibitor of EGFR tyrosine kinase AG1478; Dr J.C.Lee, King of Prussia, PA, for giving us the p38/RK/HOG‐1‐specific inhibitor SB203580; and Dr R.Benndorf, Berlin, for the Hsp25 peptide. We thank the members of our laboratories, in particular Axel Knebel, for discussions and Dr P.Sassone‐Corsi, Strasbourg, for communicating results prior to publication. M.Litfin and G.Schwederski contributed excellent technical assistance. We also thank Ingrid Kammerer and Diane Nichol for help with the text. H.J.R. was supported by the DFG (Ra 247/6‐1) and the Dr Mildred Scheel‐Stiftung (W86/94 Ra I), M.G. by the DFG (Ga 453/2‐4 and SFB273/YE4) and P.H. by Fonds der Chemischen Industrie.
- Copyright © 1997 European Molecular Biology Organization