Regulation of activating transcription factor-1 and the cAMP response element-binding protein by Ca2+/calmodulin-dependent protein kinases type I, II, and IV.

The ability of activating transcription factor-1 (ATF1) or the cAMP response element-binding protein (CREB) to enhance transcription can be stimulated by increases in intracellular Ca2+ concentrations. To identify protein kinases which may mediate the ability of Ca2+ to activate these transcription factors, we compared the ability of constitutively active forms of several Ca2+/calmodulin-dependent protein kinases (CaM kinases) to activate ATF1 or CREB. We find that constitutively active CaM kinase I and IV can activate both ATF1 and CREB. In addition, expression vectors for full-length CaM kinase I and IV were able to augment the ability of Ca2+ influx to activate ATF1 or CREB consistent with a role for these kinases in mediating transcriptional responses to Ca2+ signaling. In contrast, CaM kinase II was unable to activate either ATF1 or CREB. These findings provide a potential mechanism that may permit variation in the ability of ATF1 and CREB to respond to changes in intracellular Ca2+ concentrations depending on differences in the relative concentrations of specific CaM kinases.

The ability of activating transcription factor-1 (ATF1) or the cAMP response element-binding protein (CREB) to enhance transcription can be stimulated by increases in intracellular Ca 2؉ concentrations. To identify protein kinases which may mediate the ability of Ca 2؉ to activate these transcription factors, we compared the ability of constitutively active forms of several Ca 2؉ /calmodulin-dependent protein kinases (CaM kinases) to activate ATF1 or CREB. We find that constitutively active CaM kinase I and IV can activate both ATF1 and CREB. In addition, expression vectors for full-length CaM kinase I and IV were able to augment the ability of Ca 2؉ influx to activate ATF1 or CREB consistent with a role for these kinases in mediating transcriptional responses to Ca 2؉ signaling. In contrast, CaM kinase II was unable to activate either ATF1 or CREB. These findings provide a potential mechanism that may permit variation in the ability of ATF1 and CREB to respond to changes in intracellular Ca 2؉ concentrations depending on differences in the relative concentrations of specific CaM kinases.
The cAMP and Ca 2ϩ signaling pathways are used widely to regulate the transcription of specific genes. In at least some cases, cAMP and Ca 2ϩ converge to regulate the activity of a single transcription factor (1,2). A key study which led to this view of convergent regulation by cAMP and Ca 2ϩ involved analysis of the DNA sequences required for Ca 2ϩ -regulated transcription of the c-fos proto-oncogene (1). It was found that the regions of the fos gene that are required for responses to Ca 2ϩ mapped to a previously identified cAMP response element. Other studies provided evidence that cAMP response element-binding protein (CREB) 1 can bind to the fos cAMP response element and that elevations in intracellular Ca 2ϩ result in phosphorylation of CREB at sites similar to those that were phosphorylated in response to elevated cAMP levels. The use of a GAL4-CREB fusion provided strong evidence that the transcription stimulating activity of CREB can be regulated by increases in intracellular Ca 2ϩ (2). It seems likely that convergent regulation of CREB activity by cAMP and Ca 2ϩ permits integration of multiple extracellular signals in the regulation of specific genes.
The ability of cAMP and Ca 2ϩ to regulate CREB activity involves phosphorylation of CREB by the cAMP-dependent protein kinase (PKA) or Ca 2ϩ /calmodulin-dependent protein kinases (CaM kinases). PKA phosphorylates CREB on a single major site at Ser 133 and phosphorylation of this residue is crucial for activation of CREB by PKA (3). CaM kinases I, II, and IV have been shown to phosphorylate CREB in vitro, providing evidence that each of these protein kinases is a candidate for mediating the effects of Ca 2ϩ on CREB activity (2, 4 -6). Interestingly, recent studies have demonstrated that specific CaM kinases can have very different effects on the activation of CREB (7,8). In particular, CaM kinase II can inhibit activation of CREB while CaM kinase IV can activate CREB. The inhibitory effects of CaM kinase II are due to phosphorylation of a negative regulatory site located at Ser 142 of CREB. These observations provide a mechanism that would permit the Ca 2ϩ signaling pathway to be either antagonistic or additive with the cAMP pathway for activation of CREB depending on the relative activity of specific CaM kinases.
Although CaM kinase I has been shown to phosphorylate CREB in vitro, the ability of CaM kinase I to activate CREB has not been tested. CaM kinase I was originally identified in brain extracts as an activity that could phosphorylate a specific site of synapsin I, a protein associated with synaptic vesicles. The purified kinase was found to effectively phosphorylate only synapsin I and II (9). This led to the view that CaM kinase I had a very restricted substrate specificity and was probably functionally restricted to the modulation of neurotransmitter release. However, CaM kinase I can be found in tissues other than the brain and it has been shown that CREB appears to be good substrate for CaM kinase I in vitro (2).
In addition to CREB, at least one other member of the CREB/ATF family has been shown to be responsive to both the cAMP and Ca 2ϩ pathways. Transfection studies have shown that ATF1 can be activated by increases in either cAMP or Ca 2ϩ (10,11). Although it is clear that the activity of ATF1 can be regulated by Ca 2ϩ influx, the role of specific CaM kinases in regulating ATF1 activity has not been explored. It is possible that specific CaM kinases might have differential effects on the activity of CREB and ATF1. For instance, although previous studies have shown that CaM kinase II can block activation of CREB (7,8), it is possible that ATF1 can be activated by CaM kinase II. If this were the case, changes in the relative concentration of CREB and ATF1 would permit changes in the interaction between the cAMP and Ca 2ϩ pathways. Thus it is important to examine the ability of specific CaM kinases to modulate ATF1 activity.
In the present study we have compared the ability of CaM kinase I, II, and IV to alter the transcription stimulating activity of ATF1 and CREB. The results indicate that CaM kinase I and IV can activate ATF1 and CREB. Phosphorylation of Ser 133 of CREB or Ser 63 of ATF1 was found to be essential for transcriptional activation by CaM kinase I and IV. Thus, both CaM kinase I and IV may participate in mediating Ca 2ϩstimulated activation of ATF1 and CREB in specific tissues. In contrast, we find that CaM kinase II cannot activate either ATF1 or CREB. As observed previously with CREB, CaM kinase II phosphorylates two sites in ATF1. Thus the negative regulatory effects of CaM kinase II are observed for both ATF1 and CREB.

MATERIALS AND METHODS
Plasmid Constructions-To prepare expression vectors encoding fulllength and constitutively active CaM kinase I, cDNAs were isolated by polymerase chain amplification based on the published sequence of rat CaM kinase I (12). Appropriate primers were used to isolate the complete coding sequence (residues 1-332) or a truncated coding sequence (residues 1-295) using single-stranded cDNA from PC12 cells as a template. The correct sequences of the amplified and cloned coding sequences were confirmed by nucleotide sequence analysis (13) and comparison to the published sequence of rat CaM kinase I. The fulllength or truncated CaM kinase I coding sequences were used to replace the globin coding sequence in the RSV-globin expression vector (14). To prepare a GAL4-ATF1⌬b-zip expression vector, the coding sequence for the transcriptional activation domain of ATF1 (amino acids 1-213) was isolated by polymerase chain reaction amplification from RSV-ATF1 (10) and fused in frame to the 3Ј end of the coding sequence for the DNA binding domain of GAL4 (amino acids 1-147) downstream of the viral cytomegalovirus promoter in plasmid pcDNAI-amp (Invitrogen). Specific mutations within the ATF1 coding sequence were prepared by polymerase chain reaction mutagenesis (15), and the complete coding sequence of the mutants was confirmed by nucleotide sequence analysis (13). The construction of the GAL4-CREB⌬b-zip, CaM kinase II, CaM kinase IV, and PKA expression vectors and the 5xGAL4-TATA-luciferase reporter gene have been described previously (8,16,17).
Cell Culture and Transfection-GH 3 cells were maintained in Dulbecco's modified Eagle's medium with 15% equine serum and 2.5% fetal bovine serum. GH 3 cells were transfected by electroporation using a single pulse at 220 V and 960 microfarads. In most experiments cells received 5 g of the 5xGAL4-TATA-luciferase indicator DNA, 2 g of the GAL4-CREB⌬b-zip or GAL4-ATF1⌬b-zip expression vector, and 5 g of kinase expression vector. In some experiments, cells were treated with 80 mM KCl at 18 h after transfection. Cells were collected and lysates prepared 24 h post-transfection (6 h after KCl treatments). The protein concentration of the lysates was determined (18), and luciferase activities were measured using a constant amount of protein (19). Each experiment included three separate transfections for each experimental group, and the experiments have been repeated three to five times.
Expression and Purification of Recombinant ATF1-To facilitate isolation of recombinant ATF1, a poly(histidine)-tagged variant of ATF1 was prepared. DNA sequences encoding either full-length ATF1 or the transcriptional activation domain (amino acids 1-213) were cloned into the pET16b bacterial expression vector (Novagen), downstream and in-frame with the poly(histidine) segment and Factor Xa cleavage site. Recombinant poly(His)-ATF1 was expressed in Escherichia coli and a crude extract prepared as described previously (20), except that 2 ml of 0.5 M isopropyl-D-thiogalactopyranoside was added to each liter of culture to induce the expression of ATF1 and the crude extract was not heat treated. The crude extract from each liter of culture was diluted with one volume of 8 M urea and mixed with 1 ml of Ni ϩ -NTA-agarose (Qiagen) in the presence of 0.8 mM imidazole. The mixture was rotated at 4°C for 1 h before being packed into a column. The column was washed sequentially with 0.8, 8, and 40 mM imidazole in 20 mM Tris-HCl (pH 7.4) and 4 M urea. Poly(His)-ATF1 was eluted with 100 mM imidazole in the same buffer. The recombinant proteins containing ATF1 transcriptional activation domain were expressed and purified essentially in the same way, except that bacterial cells were lysed in 20 mM Tris (pH 7.4) by two passes through a French pressure cell at 10,000 p.s.i., and that urea was not used during the purification process. Purified proteins were dialyzed against 20 mM HEPES (pH 7.9), 50 mM NaCl, 1 mM EDTA, 10 mM ␤-mercaptoethanol, and 10% glycerol and stored at Ϫ80°C until use.
Expression and Purification of Recombinant CaM Kinase I-The full-length CaM kinase I coding sequence was inserted into the baculovirus transfer vector pBlueBac1 (Invitrogen) and used to prepare recombinant baculovirus, which directs the synthesis of CaM kinase I. Recombinant CaM kinase I was expressed in Sf9 insect cells and purified by affinity chromatography on calmodulin-Sepharose 4B (Pharmacia Biotech Inc.), as described previously (21). The eluate from the calmodulin-Sepharose 4B column appeared to contain 37-kDa CaM kinase I as well as a 49-kDa contaminant protein. CaM kinase I was further purified to near homogeneity by size exclusion chromatography on Superdex-75 (Pharmacia). The purified CaM kinase I was dialyzed against 10 mM HEPES (pH 7.6), 1 mM CaCl 2 , 50 mM NaCl, 5 mM ␤-mercaptoethanol, and 10% glycerol and stored at Ϫ80°C until use. Phosphorylation of ATF1 and Phosphopeptide Mapping-Recombinant ATF1 protein (2 g) was phosphorylated in vitro using recombinant CaM kinase I, CaM kinase II␣ (21), CaM kinase IV or the catalytic subunit of PKA in the presence of [␥-32 P]ATP (1000 -5000 cpm/pmol). The preparation of recombinant PKA, CaM kinase II␣, and CaM kinase IV has been described previously (21,22). For phosphorylation by CaM kinase I, II, or IV, the reaction mixtures contained 400 M [␥-32 P]ATP (5000 cpm/pmol), 50 mM HEPES (pH 7.5), 10 mM magnesium acetate, 50 nM kinase, 0.5 mM CaCl 2 , and 1 M calmodulin. Reactions containing PKA were performed in 400 mM [␥-32 P]ATP, 50 nM PKA, 25 mM Tris (pH 7.4), 5 mM magnesium acetate, and 0.5 mM dithiothreitol. Reactions were initiated by addition of protein kinase and incubated for 45 min at 30°C. For two-dimensional phosphopeptide mapping, ATF1 was fractionated by electrophoresis using a 12% polyacrylamide denaturing gel and phospho-ATF1 was identified by autoradiography of the dried gel. The phospho-ATF1 bands were cut from the gel, and two-dimensional tryptic phosphopeptide maps were prepared on the extracted proteins as described (23). Briefly, the gel pieces were rehydrated in 50 mM ammonium bicarbonate. Phosphorylated ATF1 was eluted from the gel, precipitated in 16% trichloroacetic acid, and oxidized with performic acid. Samples were lyophilized, resuspended in 50 mM ammonium bicarbonate, and digested with 30 mg of L-1-tosylamido-2-phenylethylchloromethyl ketone-treated trypsin (U. S. Biochemical Corp.) for three 8-h periods at 37°C. The peptides were lyophilized and resuspended in the pH 1.9 buffer (2.5% of formic acid and 7.8% of acetic acid in water). Equal amounts of radioactivity for each sample was loaded on a cellulose plate and fractionated by high voltage electrophoresis in the pH 1.9 buffer for 25 min at 1 kV using the Hunter thin layer electrophoresis system (HTLE-7000, CBS Scientific Co., Del Mar, California). Samples were separated in the second dimension by thin layer chromatography using n-butanol:pyridine:acetic acid:water in volume ratios of 0.375:0.25:0.075:0.30. The phosphopeptides were visualized by autoradiography.

Expression Vectors for Constitutively Active Forms of CaM
Kinase I and CaM Kinase IV Can Activate Both ATF1 and CREB-It has been reported that in vitro, CaM kinase I can phosphorylate both ATF1 and CREB (2). However, the ability of CaM kinase I to activate ATF1 and CREB has not been tested. To examine this issue, we prepared an expression vector for a constitutively active form of CaM kinase I. Previous studies have shown that truncation of CaM kinase II at Leu 290 (24) or CaM kinase IV at Leu 313 (5) removes an autoinhibitoryregulatory region of the enzyme and results in a constitutively active protein kinase, which no longer requires Ca 2ϩ and calmodulin for activity. We aligned the amino acid sequences of CaM kinase I with those of CaM kinase II and CaM kinase IV and found that Lys 295 of CaM kinase I corresponded to Leu 290 of CaM kinase II or Leu 313 of CaM kinase IV. Therefore the coding sequence of CaM kinase I was truncated at Lys 295 and used to prepare a mammalian expression vector. To specifically determine if CaM kinase I could alter the activity of ATF1 and CREB, expression vectors for GAL4-ATF1 or GAL4-CREB fusion proteins (2,11,25,26) were transfected with a luciferase reporter gene containing five copies of a GAL4 binding site (5xGAL4-TATA-luciferase). To avoid possible formation of heterodimers with endogenous members of the CREB/ATF family, the GAL4-ATF1 and GAL4-CREB coding sequences were truncated to remove the carboxyl-terminal basicϩleucine zipper DNA binding region (GAL4-ATF1⌬b-zip, GAL4-CREB⌬b-zip).
Transfection of expression vectors for constitutively active forms of CaM kinase I or CaM kinase IV increased the ability of GAL4-ATF1⌬b-zip or GAL4-CREB⌬b-zip to stimulate expression of the reporter gene (Fig. 1, A and B). The kinase expression vectors did not substantially alter the expression of the luciferase reporter gene driven by the thymidine kinase promoter (Fig. 1C), demonstrating promoter specificity and suggesting that transcriptional activation of GAL4-ATF1 and GAL4-CREB are not due to general effects on the transcription apparatus. These findings provide evidence that both CaM kinase I and CaM kinase IV can increase the transcriptional activity of ATF1 and CREB. Surprisingly, an expression vector for the catalytic subunit of PKA was not able to activate GAL4-ATF1⌬b-zip (Fig. 1A), although PKA strongly activated GAL4-CREB⌬b-zip (Fig. 1B).
Previous studies have shown that phosphorylation of Ser 133 of CREB is crucial for transcriptional activation by PKA and CaM kinase IV (3,8). A large portion of the transcriptional activation domain of ATF1 is similar to the transcriptional activation domain of CREB, including a conserved PKA site in which Ser 63 of ATF1 corresponds to Ser 133 of CREB. Mutation of Ser 63 of ATF1 to alanine (ATF1-S63A) essentially blocked the ability of CaM kinase I and CaM kinase IV to activate GAL4-ATF1. Similarly, mutation of Ser 133 of CREB to alanine (CREB-S133A) blocked activation in response to PKA, CaM kinase I, and CaM kinase IV. These findings demonstrate that CaM kinase I and CaM kinase IV can enhance the ability of ATF1 and CREB to stimulate transcription, likely involving a crucial phosphorylation event at Ser 63 of ATF1 or Ser 133 of CREB.
As indicated above, it was surprising that expression vectors for constitutively active forms of CaM kinase I and CaM kinase IV were able to activate GAL4-ATF1⌬b-zip, but that a PKA expression vector was not able to substantially increase the activity of this fusion protein. To examine this issue further, we performed a titration experiment using increasing concentrations of expression vectors for each of three protein kinases (Fig. 2). We found that at all of the tested concentrations of DNA, expression vectors for CaM kinase I and CaM kinase IV enhanced the ability of GAL4-ATF1⌬b-zip to stimulate re-porter gene expression. At lower concentrations of the expression vectors, CaM kinase I consistently resulted in greater activation of ATF1 than CaM kinase IV. In contrast, PKA had little effect on activation of GAL4-ATF1⌬b-zip at any of the tested concentrations of expression vector. The failure of PKA to substantially activate GAL4-ATF1 might be due to inhibitory effects of the kinase. To address this issue, a PKA expression vector was co-transfected with a vector for a constitutively active form of CaMKI (Fig. 3). As observed above, the CaMKI expression vector enhanced the ability of GAL4-ATF1⌬b-zip to stimulate reporter gene activity while the PKA vector by itself had little effect. Co-transfection of the two kinase expression vectors demonstrated that PKA does not block activation by CaMKI. Overall these studies provide evidence that the transcription stimulating activity of ATF1 appears to be responsive to CaM kinases I and IV and relatively unresponsive to PKA.
We also examined the ability of endogenous kinases to activate GAL4-ATF1⌬b-zip (Fig. 4). To stimulate endogenous CaM kinases, transfected GH 3 cells were depolarized by KCl treatment, which results in Ca 2ϩ influx through voltage-dependent Ca 2ϩ channels. We previously found that in GH 3 cells, KClinduced activation of GAL4-CREB is dependent on the presence of extracellular Ca 2ϩ (8). KCl treatment enhanced the ability of GAL4-ATF1 to stimulate reporter gene expression. These findings provide evidence that endogenous Ca 2ϩ -responsive enzymes can activate GAL4-ATF1. Treatment of transfected cells with chlorophenylthio-cAMP had a modest effect on the ability of GAL4-ATF1⌬b-zip to stimulate reporter gene expression.
CaM Kinase I Phosphorylates CREB on Ser 133 and ATF1 on Ser 63 in Vitro-To explore the biochemical events that are likely important in mediating the ability of CaM kinase I to activate ATF1, we characterized the in vitro phosphorylation of ATF1 and CREB using two-dimensional tryptic phosphopeptide mapping (Fig. 5). As has been observed previously (3), phosphorylation of CREB by PKA results in a major phosphopeptide (Fig. 5A). We frequently observed a minor phosphopeptide after phosphorylation by PKA. Mutation of Ser 133 of CREB to alanine eliminated the major phosphopeptide as well as the minor phosphopeptide (data not shown). This observation supports previous conclusions that PKA phosphorylates a single site, Ser 133 , on CREB. Presumably, the minor phosphopeptide that we frequently detect is the result of partial digestion with trypsin. A very similar phosphopeptide map was obtained when CREB was phosphorylated with CaM kinase I, consistent with studies by Sheng et al. (2) showing that CaM kinase I can phosphorylate CREB on Ser 133 . Phosphorylation of ATF1 by PKA or CaM kinase I (Fig. 5, C and D) yielded phosphopeptide maps that were very similar to those obtained after phosphorylation of CREB with PKA. The region surrounding Ser 63 of ATF1 is identical to the region surrounding Ser 133 of CREB, and phosphorylation of these sites would result in identical tryptic phosphopeptides (Fig. 6). To further examine the phosphopeptides obtained from ATF1, a mixing experiment was performed. Phosphopeptides obtained from ATF1 which had been phosphorylated by PKA were mixed with peptides derived from CREB that had been phosphorylated with PKA (Fig. 5E). The mixing experiment demonstrates that after phosphorylation by PKA, the resulting ATF1 and CREB phosphopeptides migrate indistinguishably in the two-dimensional map. A similar mixing experiment demonstrated that the phosphopeptides obtained after phosphorylation of ATF1 with CaM kinase I co-migrated with CREB PKA-derived peptides (Fig. 5F). As phosphorylation of Ser 63 of ATF1 would produce these identically migrating peptides, these findings support the view that both PKA and CaM kinase I phosphorylate a single major site in ATF1 at Ser 63 . Similar findings were obtained for ATF1 after phosphorylation by CaM kinase IV (data not shown  tutively active form of CaM kinase I was able to activate both ATF1 and CREB. Of course, it is possible that although the truncated form of the enzyme can activate these transcription factors, the full-length enzyme might not be able to function in this manner. We previously found that an expression vector for full-length CaM kinase IV was able to augment the ability of Ca 2ϩ influx to stimulate the activity of GAL4-CREB (8). We transfected GH 3 cells with expression vectors for full-length CaM kinase I or CaM kinase IV and examined the ability of KCl to enhance GAL4-CREB⌬b-zip mediated activation of the GAL4-dependent luciferase reporter gene (Fig. 7). We found that both CaM kinase I and CaM kinase IV had effects to augment activation of CREB by KCl-induced Ca 2ϩ influx in a concentration-dependent manner. Interestingly, CaM kinase I was found to stimulate reporter gene activity in the absence of KCl treatment. This effect was not observed with the CaM kinase IV expression vector. At similar concentrations of expression vector, CaM kinase I and CaM kinase IV resulted in similar levels of KCl-stimulated reporter gene activity. Because CaM kinase I increased both basal and KCl stimulated reporter gene activity, the calculated -fold inductions did not increase. In contrast, the CaM kinase IV vector resulted in a substantial increase in KCl-induced -fold activation. The reason for CaM kinase I-induced CREB activation in non KCl-treated cultures is not clear. None the less, the finding that full-length CaM kinase I can alter maximal KCl-induced activity of GAL4-CREB suggests that this enzyme as well as CaM kinase IV is a candidate for mediating transcriptional activity in response to Ca 2ϩ influx. Neither kinase expression vector had effects on transcriptional activation mediated by the GAL4 DNA binding domain alone (Fig. 7, C and D), demonstrating that the observed effects are specific for CREB.
An Expression Vector for a Constitutively Active Form of CaM Kinase II Is Unable to Activate ATF1-We previously found that although CaM kinase II can phosphorylate CREB, it is not able to activate CREB (8). These studies provided evidence that CaM kinase II is not able to activate CREB because it phosphorylates Ser 142 as well as Ser 133 . Phosphorylation of Ser 142 was found to block the activation of CREB, which would otherwise occur following phosphorylation of Ser 133 . As a serine residue is found in ATF1 at a location corresponding to Ser 142 of CREB (Fig. 6), it seemed possible that CaM kinase II might also fail to activate ATF1. GH 3 cells were transfected with expression vectors for constitutively active CaM kinases, GAL4-ATF1⌬b-zip or GAL4-CREB⌬b-zip and the GAL4-dependent luciferase reporter gene (Fig. 8). These studies revealed that the CaM kinase II expression vector was unable to activate GAL4-ATF1 or GAL4-CREB. In the same experiment, expression vectors for constitutively active CaM kinase I or CaM kinase IV substantially activated GAL4-ATF1 or GAL4-CREB. These findings extend our previous observation that CaM kinase II cannot active CREB and demonstrate that CaM kinase II also cannot activate ATF1.
CaM Kinase II Phosphorylates ATF1 at Two Sites in Vitro-Two-dimensional phosphopeptide mapping experiments were performed to examine sites on ATF1 that are phosphorylated by CaM kinase II in vitro (Fig. 9). After phosphorylation of ATF1 by CaM kinase II, two major phosphopeptides and several minor phosphopeptides were observed (Fig. 9A). One of the major phosphopeptides migrated similarly to the major peptide that is obtained after phosphorylation of ATF1 or CREB with PKA (data not shown), suggesting that this peptide likely represents phosphorylation of Ser 63 . The identity of this site as Ser 63 was further supported by mutation of Ser 63 to alanine, which eliminated this major phosphopeptide (Fig. 9B). We suspected that the other major phosphopeptide might result from phosphorylation of Ser 72 of ATF1 which appears to correspond to Ser 142 of CREB (Fig. 6). To further examine this possibility, we mutated Ser 72 as well as the adjacent Ser 73 to alanine and used the mutant coding sequence to produce recombinant ATF1 in E. coli. Unfortunately, despite several attempts, we were unable to obtain sufficient quantities of highly purified ATF1-S72A to permit phosphopeptide mapping experiments. The reason for this difficulty is not clear, although we have found it much more difficult to express and purify ATF1 than CREB. We were able to express and purify ATF1-S73A and ATF1-S72A-S73A and use these proteins for phosphopeptide mapping experiments. While mutation of Ser 73 to alanine did not substantially alter the phosphopeptide maps (Fig. 9C), mutation of both Ser 72 and Ser 73 eliminated one of the major ATF1 phosphopeptides (Fig. 9D). These findings suggest that CaM kinase II probably phosphorylates Ser 72 of ATF1. However, we cannot exclude the possibility that either Ser 72 or Ser 73 of ATF1 can be phosphorylated by CaM kinase II.
Mutation of Ser 72 of ATF1 Permits CaM Kinase II to Activate ATF1-The phosphopeptide mapping experiments identified two sites within ATF1 that are phosphorylated by CaM kinase II. This finding raised the possibility that the inability of CaMKII to activate ATF1 might involve phosphorylation of a site that inhibits transcriptional activation, similar to our previous findings concerning CaMKII effects on activation of CREB. As Ser 63 of ATF1 is crucial for phosphorylation mediating transcriptional activation, it seemed likely that phosphorylation of the second site, presumably Ser 72 , mediates the inhibitory effects of CaM kinase II on ATF1 activation. To test this possibility, a mutant ATF1 coding sequence was prepared in which Ser 72 was replaced with an alanine. Mutation of Ser 72 to alanine greatly enhanced activation of GAL4-ATF1⌬b-zip as compared to the wild type construct (Fig. 10). This finding strongly supports a model in which phosphorylation of Ser 72 serves to inhibit activation of ATF1. DISCUSSION We have compared the ability of CaM kinases I, II, and IV to alter the activity of ATF1 and CREB. We find that CaM kinase I or IV can activate either ATF1 or CREB. In contrast, CaM kinase II cannot activate either ATF1 or CREB. These findings provide further insight into the signal transduction pathways that are important for mediating the ability of Ca 2ϩ to activate CREB and ATF1. The results provide strong evidence that CaM kinase I as well as CaM kinase IV can activate CREB and ATF1. In titration experiments we found that low concentra- tions of the CaM kinase I vector were considerably more effective than the CaM kinase IV vector in activating CREB. While we have not determined that equal concentrations of expression vector produce equivalent amounts of CaM kinase I and IV, these studies at least raise the possibility that CaM kinase I may be more effective than CaM kinase IV in activating CREB. The finding that full-length CaM kinase I or IV can augment the activation of ATF1 or CREB in response to Ca 2ϩ influx supports the view that both of these enzymes are good candidates as physiological regulators of CREB and ATF1 activity.
In contrast to previous studies of ATF1 (10,11), we found that PKA was unable to enhance the ability of GAL4-ATF1⌬bzip to stimulate reporter gene activity. The earlier studies utilized either full-length ATF1 or full-length ATF1 fused to the DNA binding domain of GAL4. As ATF1 can form heterodimers with CREB (11), the activation observed with fulllength ATF1 may actually reflect phosphorylation and activation of CREB as part of an ATF1⅐CREB heterodimer. Studies utilizing leucine zipper variants of CREB which are engineered to form specific heterodimers have shown that a hemiphosphorylated CREB dimer or a CREB⅐CREM␣ heterodimer can mediate PKA-induced transcriptional activation (27,28). Our use of a GAL4-ATF1⌬b-zip construct would eliminate formation of CREB⅐ATF1 heterodimers, and this might account for the failure to respond to PKA. The finding that ATF1 has only a limited ability to mediate transcriptional responses to PKA would be consistent with recent studies, which concluded that ATF1 is a specific antagonist of CREB-mediated transcriptional activation (29). On the other hand, we found that both CaM kinase I and CaM kinase IV were able to activate GAL4-ATF1⌬b-zip. This finding is quite surprising as phosphopeptide mapping experiments demonstrated that PKA, CaM kinase I, and CaM kinase IV all appear to phosphorylate ATF1 at the same site in vitro, Ser 63 . How is it that all three kinases appear to phosphorylate ATF1 on Ser 63 in vitro, yet only CaM kinase I or IV can activate GAL4-ATF1⌬b-zip in a transfection experiment? We have used co-transfection experiments to demonstrate that PKA cannot block the ability of CaMKI to activate ATF1. Thus, it seems unlikely that PKA phosphorylates a factor that inhibits responses to ATF1. It may be that CaM kinase I and IV phosphorylate other transcription factors or coactivators, which are required to mediate responses to phospho-ATF1. Microinjection experiments have shown that phosphorylation of CREB on Ser 133 is sufficient for transcriptional activation and thus other PKA-dependent phosphorylations are not required for a response to phospho-CREB. Similar microinjection experiments might yield insight into the differential activation of ATF1 by PKA and CaM kinase I or IV.
The use of expression vectors for full-length CaM kinase I and IV demonstrated that both of these kinases can participate in Ca 2ϩ -regulated activation of CREB. Most of our studies relied upon the use of truncated, constitutively active forms of these enzymes. It is possible that the truncated forms of the kinase may have enhanced access to the nucleus and therefore might alter nuclear events which are not normally regulated by the holoenzyme. The finding that full-length CaM kinase I and IV can activate CREB supports a role for these enzymes in the physiological regulation of transcription. It is interesting that transfection of an expression vector for full-length CaM kinase I stimulated basal, CREB-mediated transcriptional activation in the absence of depolarization-induced Ca 2ϩ influx. On the other hand, full-length CaM kinase IV appeared to be much more Ca 2ϩ -dependent, suggesting that there may be differences in Ca 2ϩ -mediated activation of the two enzymes. There is evidence that activation of both CaM kinase I and IV requires activator protein, which appears to be another protein kinase (30 -34). Our findings suggest that some step in the activation of CaM kinase I may be more sensitive to Ca 2ϩ than occurs for activation of CaM kinase IV. It seems likely that this altered Ca 2ϩ sensitivity occurs at the level of the CaM kinase activator kinase. Thus our findings suggest possible differences in the enzymes which mediate activation of CaM kinase I and IV.
Our studies have shown that CaM kinase II is unable to activate ATF1. This finding extends our previous observation that CaM kinase II cannot activate CREB (8) and demonstrates that the failure to respond to CaM kinase is conserved for both proteins. Unlike CaM kinase I and IV, which phosphorylate ATF1 or CREB at a single major site, CaM kinase II phosphorylates both ATF1 and CREB at two sites. One of the sites, Ser 133 of CREB or Ser 63 of ATF1, is the same site that is recognized by PKA, CaM kinase I, and CaM kinase IV. Phosphorylation of Ser 133 of CREB or Ser 63 of ATF1 is crucial for transcriptional activation. The second site in ATF1 that is phosphorylated by CaM kinase II appears to be Ser 72 , which corresponds to Ser 142 , the inhibitory CaM kinase II site in CREB. Neither Ser 142 of CREB nor Ser 72 of ATF1 match a consensus CaM kinase II phosphorylation site, which is Arg-X-X-Ser/Thr (35). Phosphorylation at non-consensus sites by CaM kinase II has been reported for several proteins (36,37). It will likely require experiments exploring the phosphorylation of an extensive set of peptides based on these proteins to further define the alternative consensus sites recognized by CaM kinase II. In any case, the present findings and our previous studies (8,17) suggest that the ability of CaM kinase II to phosphorylate a second, non-consensus site in ATF1, and CREB appears to inhibit activation of these transcription factors.
In conclusion, we have demonstrated differential activation of ATF1 and CREB by specific CaM kinases. Both CaM kinase I and IV can activate ATF1 and CREB. CaM kinase II is unable to activate either transcription factor. Based on these studies, it seems likely that tissue-specific or developmental changes in the concentrations of individual CaM kinases permits variations in the ability of the Ca 2ϩ signaling pathway to regulate the activity of specific genes.