Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

Various extracellular signals such as neurotransmitters and hormones influence cellular function by increasing the intracellular cAMP cascade. This second messenger cascade includes activation of cAMP-dependent protein kinase (PKA) and regulation of gene transcription via cAMP response elements (CRE; Ziff, 1990) in the promoter regions of target genes. The CRE binding protein (CREB) mediates the action of the cAMP cascade on gene expression (Hoeffler et al., 1988; Yamamoto et al., 1988; Gonzalez and Montminy, 1989). In addition to CREB, there are several other CRE binding proteins that form the CREB/activating transcription factor (ATF) family (Maekawa et al., 1989; Hai and Curran, 1991).

Phosphorylation of CREB at Ser¹³³ by PKA initiates CREB transactivation. Phosphorylated CREB binds to a coactivator protein, CREB binding protein (Chrivia et al., 1993), which then binds to a transcription factor/RNA polymerase II complex, which directly transactivates the target gene (Kwok et al., 1994). The Ser¹³³ residue can be phosphorylated not only by PKA, but also by other kinases, including CaM kinase II and IV (Dash et al., 1991; Matthews et al., 1994), indicating that intracellular elevation of calcium can also transactivate target genes via CREB. In the CREB/ATF family the basic/leucine zipper (bZIP) motif in the C terminus region is highly conserved. The bZIP motif is essential for DNA binding and heterodimerization of the CREB/ATF proteins (Johnson and McKnight, 1989). In addition, the bZIP domain is necessary for translocation of CREB into the nucleus (Waeber and Habener, 1991).

A study of the mouse CREB gene has revealed that it consists of 11 exons and multiple variants produced by alternative splicing (Cole et al., 1992; Ruppert et al., 1992). Following that report, additional splice variants of mammalian CREB have been identified, some of which exhibit tissue specific expression (Waeber and Habener, 1992; Ellis et al., 1995; Blendy et al., 1996; Girardet et al., 1996; Walker et al., 1996; Youg et al., 1997). The major identified CREB isoforms, , , and  (Ruppert et al., 1992; Blendy et al., 1996) have all of the three domains which are essential for the transactivation function of CREB, including a glutamine rich domain (Q domain), a kinase inducible domain (KID domain), and a bZIP domain.

In contrast, some members of the CREB/ATF family have the bZIP domain but lack the glutamine rich Q and kinase-inducible domains (KIDs), and act as repressors of CREB transactivation (Walker et al., 1996). Examples of these are the I-CREBs and ICER, an isoform of cAMP response element modulator. In addition, CREB variants that lack the bZIP domain also have been reported (Ruppert et al., 1992; Waeber and Habener, 1992; Ellis et al., 1995; Youg et al., 1997; Bartsch et al., 1998), although the functional significance of these proteins is still controversial. These variants are localized primarily to the cytosol due to their lack of a bZIP domain (Waeber et al., 1991; Bartsch et al., 1998).

CREB plays an important role in numerous cellular functions (Lalli and Sassone-Corsi, 1994). In the brain, CREB is reported to play a role in circadian rhythm and formation of learning and memory (Ginty et al., 1993; Milner et al., 1998). Furthermore, the function and expression of CREB in specific brain regions has been implicated in the neuronal adaptations or plasticity due to chronic psychotropic drug treatments (Nibuya et al., 1996); Duman et al., 1997; Nestler and Aghajanian, 1997). To further characterize the expression of CREB variants in the brain, reverse transcriptase-polymerase chain reaction (RT-PCR) cloning of CREB variants was performed. Two novel splice variants, which are predicted to lack the bZIP domain, but conserve the KID domain, were identified.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

RT-PCR Cloning of CREB Variants. Total RNA was extracted from mouse brain (striatum) using the RNAquous kit (Ambion, Austin, TX) according to the manufacture's standard protocol. A sense primer (designated primer 8; Fig. 1) and an antisense primer (designated primer 9; Fig. 1), which recognize exons 4 and 9 of the mouse CREB, respectively, were designed for RT-PCR. The sequence of primer 8 was 5'-CAGTCTCCACAAGTCCAAACAGTT-3', and that for primer 9 was 5'-GTAGAATGGTAGTACCCGGCTGA-3'. RT-PCR was carried out with the striatal RNA as template by using the Access RT-PCR system (Promega, Madison, WI), according to the manufacture's recommended method. RT-PCR products of the primer 8-9 set were subjected to agarose gel separation and visualized by ethidium bromide staining. Several RT-PCR products of the predicted size [434 base pairs (bp) and 392 bp] of known variants (CREB and CREB) were observed. In addition, several products of unpredicted size were also observed (data not shown). These RT-PCR products were isolated by using a gel extraction kit (Qiagen, Chatsworth, CA), subcloned into pGEM-T Easy vector (Promega) and verified by sequencing. Two distinct plasmids, encoding novel splice variants of the CREB isoform, were isolated and designated CREB-14/8-9 and CREB-35/8-9 (Fig. 1 and Results).

View larger version (34K): [in this window] [in a new window]   Fig. 1.   Identification, sequence analysis, and predicted structure of the CREB-14 and CREB-35 variants. A schematic of the mouse CREB gene is shown at the top with the exons depicted as the boxed areas. Below are diagrams of CREB, as well as the novel splice variants CREB-14 and CREB-35. In the CREB-14 variant, 14 nucleotides are deleted at the beginning of exon 9 around the boundary of intron 8 and exon 9. For the CREB-35, variant 35 nucleotides are deleted at the beginning of exon 8 around the boundary of intron 7 and exon 8. In both variants, new translation termination codons are predicted near the deleted region that would result in CREB variants lacking the bZIP domain. The shaded boxes indicate the coding regions that variants would lack. Boxed and shaded nucleotides in CREB isoform are the nucleotides that are deleted in either CREB-14 or CREB-35. The nomenclature used here is for the mouse CREB gene according to Ruppert et al. (1992). Note that in each case, the deletion is followed by standard acceptor and donor sites. Q1, glutamine rich region 1; Q2, glutamine rich region 2; KID, kinase-inducible domain; bZIP, basic/leucine zipper domain

RNase Protection Assay (RPA) of Splice Variants. RNase protection analysis was performed by RPAII kit (Ambion) according to the manufacture's standard protocol, with total RNA extracts from various mouse tissues as templates. The riboprobes were generated by linearizing CREB-14 or -35 plasmids with NcoI and ³²P-labeled using SP6 RNA polymerase (Boehringer Manheim, Indianapolis, IN). Protected fragments were loaded onto an 8% Acrylamide-TBE gel and the separated bands were detected by autoradiography. Measurement and quantification of protected band density were carried out by using the Macintosh-based NIH image analysis program (version 1.52).

Subcloning of CREB Variants into Expression Plasmids for Mammalian Cells. To obtain the cDNAs encoding CREB-14 or -35 with initiation sites of CREB isoform, another RT-PCR was performed using primer set 6-7 (Fig. 1). The sense (6) and antisense (7) primers were 5'-CTAAATGACCATGGAATCTGGAGCA-3' and 5'-AGTTACACTATCCACAGACTCCTG-3', respectively. RT-PCR products encoding CREB isoforms were identified based on their predicted size and subcloned into a pGEM-T Easy vector (designated as CREB/6-7). To subclone CREB14 or CREB35 cDNA with the CREB initiation site, ExoRI/PleI fragments of CREB/6-7 and PleI/EcoRI fragments of CREB14/8-9 or CREB35/8-9, respectively, were ligated into the ExoRI sites of the pCI plasmid, an expression plasmid for mammalian cells with a CMV promoter. To subclone CREB-14 and -35 cDNA containing the CREB initiation site into pCI (Promega), an expression plasmid for mammalian cells with a CMV promoter, EcoRI/PleI fragments of CREB/6-7 and PleI/EcoRI fragments of CREB-14/8-9 and CREB-35/8-9, respectively, were ligated into the EcoRI site of pCI plasmids.These plasmids were designated as CREB-14 pCI and CREB-35 pCI, respectively.

Cell Transfection and CAT Assay. Human embryonic kidney (HEK) 293 cells were cultured in Dulbecco's modified Eagle medium containing 25 mM glucose, which was buffered with 44 mM NaHCO3 and supplemented with 10% fetal bovine serum, in a humidified atmosphere containing 5% CO₂ at 37°C. Transfection of plasmids was performed by lipofection with TransIT-LT2 reagent (Panvera, Madison, WI), into subconfluent 293 cells (approximately 6 × 10⁶ cells/10 cm dishes). For Western blotting, cells were reseeded into two or three 10-cm dishes 16 h after the transfection and were harvested 48 h later. For CAT assays, 5 µg of CRE-CAT constructs (Montominy et al., 1986) plus 15 µg of either CREB pCI variant or pCI vector alone plus 1 µg pCMV-gal (Promega) were cotransfected into 293 cells. Cells were reseeded into six-well dishes (3.0 cm in diameter) 24 h after the transfection. After the transfection, cells were allowed to settle on the dishes (approximately 8 h after reseeding), then treated with or without forskolin (5 µM; Sigma Chemical Co., St Louis, MO) for 16 h, and then the CAT assay was carried out by CAT enzyme assay system (Promega) as described previously (Widnell et al., 1996a). -galactosidase activity was simultaneously measured by the -galactosidase enzyme assay system (Promega) and CRE-CAT activity was normalized to the -gal activity.

Western Blotting, Immunocytochemistry and In Vitro Transcription/Translation. Transfected or untransfected cells were washed and harvested with isotonic homogenate buffer (250 mM sucrose, 10 mM EGTA, 2 mM EDTA, 50 mM Tris/HCl, 200 µg/ml leupeptin, and 1 mM phenylmethylsulfonyl fluoride, pH 7.4), then lysed with RIPA buffer (10 mM Tris-HCl, 1% NP40, 0.1% sodium deoxycholate, 0.1% SDS, 0.15 M NaCl, 1 mM EDTA, 200 µg/ml leupeptin, and 1 mM phenylmethylsulfonyl fluoride, pH 7.4) and centrifuged at 19,000 × g for 15 min. Supernatants were subjected to SDS-polyacrylamide gel electrophoresis. Western blotting was carried out as described previously (Takahashi et al., 1999). For phospho-CREB immunoblotting, cells were harvested after treatment with 50 µM forskolin for 10 min. Phosphatase inhibitors (1 mM Na₃VO₄, 1 mM NaF and 100 nM calyculin A) were added to homogenates and RIPA buffer. Immunoblotting for N- and C-terminal CREB was performed using polyclonal CREB (Upstate Biotechnology Incorporated, Lake Placid, NY; 1:500 dilution) and monoclonal CREB (X-12; Santa Cruz Biotechnology, Santa Cruz, CA; 1:100 dilution) antibodies, respectively. Immunoblotting for phospho-CREB was performed using polyclonal phospho-CREB antibody (New England Biolabs, Beverly, MA; 1:500 dilution).

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

Sequence Analysis and Predicted Structure of Novel CREB Splicing Variants. By using RT-PCR and total RNA extracted from mouse striatum as a template, we have isolated two cDNAs which encode novel CREB splice variants. The primers used for RT-PCR were derived from the nucleotide sequence of exons 4 and 9 of the mouse CREB gene (Fig. 1). Both variants lack exon 5, suggesting that they are most closely related to the CREB isoform, which also lacks this exon. One splice variant, referred to as CREB-14, also lacks 14 nucleotides at the beginning of exon 9, around the boundary of intron 8 and exon 9 (Fig. 1). Deletion of 14 nucleotides would result in a change of codon usage, including termination of translation in exon 9. Based on this analysis and the initiation site in exon 2, which is the same as that used for CREB and  isoforms, CREB-14 is predicted to have 175 amino acids.

Tissue Distribution of CREB-14 and CREB-35 Splice Variants. To characterize CREB-14 and -35 expression, we analyzed levels of mRNA for each variant in different tissues. Expression was determined by RPA with antisense riboprobes derived from CREB-14 or -35 cDNAs, referred to as CREB-14/8-9 and CREB-35/8-9, respectively. The predicted fragment size for each CREB isoform in the RPA is summarized in Fig. 2A. As shown in Fig. 2B, CREB-14 and -35 are expressed in all tissues examined, although at different levels, providing evidence for widespread distribution of both variants. CREB-14 and -35 were found to be enriched in most brain regions, particularly cerebellum, as well as in thymus and testis, relative to other peripheral tissues; the CREB and  isoforms are also expressed at relatively high levels in these tissues. Both variants are expressed at lower levels in the other tissues examined, including lung, heart, liver, spleen, and kidney. Note that larger amounts of total RNA were used for these latter tissues, due to lower levels of expression, to observe clear expression of the variants.

View larger version (51K): [in this window] [in a new window]   Fig. 2.   Schematic of the CREB gene and riboprobes used for analysis of CREB-14 and CREB-35 expression in brain and peripheral tissues. A, structure of the CREB gene is shown at the top and that for the CREB-14/8-9 and CREB-35/8-9 riboprobes is shown below. The protected RNA hybrids for CREB, CREB, CREB-14, and CREB-35 that are predicted to result from RPA with the CREB-14/8-9 and CREB-35/8-9 riboprobes are shown. The sizes (bp) of the protected hybrids are shown on the right. The predicted size of CREB-14 is 378 bp and that for CREB-35 is 357 bp when using the respective riboprobes for each of variant. B, tissue distribution of CREB-14 and CREB-35 mRNA. Expression of CREB-14 and CREB-35 mRNA is enriched in brain, including the cerebellum, cerebral cortex, hippocampus, striatum, and brainstem. Levels of these variants are also enriched in thymus and testis, and are widely expressed in the other tissues examined. Note that larger amounts of total RNA were used for some of the peripheral tissues, including lung, heart, liver, spleen, and kidney, so that a significant signal could be observed.

Expression of CREB-14 and CREB-35 Protein. To characterize the structural and functional properties of the CREB variants, expression vectors for CREB-14 and -35 were prepared. As described in Materials and Methods, RT-PCR was used to isolate the 5' coding region of the CREB isoform, including the translation initiation start site, which was then subcloned separately with each of the original RT-PCR products into a pCI expression plasmid. Sequence analysis confirmed that the expression plasmids contained the predicted 5' coding region, as well as the coding region for each variant. CREB-14 and -35 were first expressed by using an in vitro transcription/translation system under the control of T7 promoter. The CREB-14 and -35 proteins have molecular masses of approximately 27 and 22 kDa, respectively (Fig. 3A).

View larger version (41K): [in this window] [in a new window]   Fig. 3.   Expression of recombinant CREB-14 and CREB-35 proteins. A. Expression of CREB14 and CREB-35 protein using an in vitro translation/transcription system and the expression vectors CREB-14 pCI and CREB-35 pCI. In these vectors, expression is under the control of the CMV promoter. Immunoblot analysis with a CREB amino terminus antibody demonstrates that CREB-14 and CREB-35 have molecular masses of approximately 27 and 22 kDa, respectively. B, expression of CREB-14 and CREB-35 proteins in mammalian cells. The CREB-14 pCI and CREB-35 pCI expression vectors were transfected into HEK293 cells, and immunoblot analysis was conducted with CREB antibodies selective for either the amino- or carboxy terminus. The new stop codon resulting from the nucleotide deletions predict that the CREB-14 and -35 variants will have a truncation of the carboxy terminus and would not be reactive to the antibody selective for this portion of CREB. This is confirmed by the presence of immunolabeled bands of the appropriate size detected with the antibody selective for the amino, but not the carboxy terminus of CREB. C, CREB -14 and CREB-35 are phosphorylated by activation of PKA. The variants retain the consensus phosphorylation site (Ser133). To determine whether these sites undergo phosphorylation, CREB-14- or CREB-35-transfected cells were treated with or without 50 µM forskolin for 10 min and levels of phospho-CREB immunoreactivity were determined. The results demonstrate the presence of phospho-CREB immunoreactive bands at the appropriate molecular weight in response to forskolin treatment, but not under basal conditions. Higher-molecular-weight bands, which most likely represent CREB and ATF-1, also were observed in the presence of forskolin. D, the expression of CREB-14 and -35 protein in native tissues was determined by immunoblot analysis with the antibody selective for the amino terminus of CREB. Shown are representative immuonoblots of recombinant CREB-14 or -35 protein and lysates from various mouse brain regions (100 µg of protein) and peripheral tissues (150 µg of protein). The regions examined included: Cb, cerebellum; Cx, cortex; Hp, hippocampus; St, striatum; Kd, kidney; and Lu, lung. Immunoreactive bands which comigrate with either recombinant CREB-14 or -35 protein were detected.

Localization of CREB-14 and -35 in Transfected Cells. To examine the localization of the CREB variants in transfected HEK293 cells, we constructed plasmids for expression of FLAG-tagged versions of the CREB variants (referred to as CREB-14FL and CREB-35FL; see Materials and Methods). We also transfected FLAG-tagged CREB cDNA (a gift from Dr. M.Greenberg, Harvard Medical School, Cambridge, MA) as a control. FLAG immunocytochemistry was performed in cells transfected with FLAG-CREB-14, FLAG-CREB-35 and FLAG-CREB. In cells expressing either FLAG-CREB-14 or FLAG-CREB-35, FLAG-immunoreactivity was seen predominately in the cytosol and rarely in the nucleus (Fig. 4). These observations are consistent with the prediction that the CREB variants are not translocated to the nucleus because they lack the bZIP domain. In contrast, in cells transfected with FLAG-CREB, FLAG immunoreactivity was observed largely in the nucleus, although immunoreactivity in the cytosol was also observed (Fig. 4).

View larger version (45K): [in this window] [in a new window]   Fig. 4.   Cellular localization of CREB-14 and CREB-35 variants in transfected HEK293 cells. FLAG-tagged CREB-14 and CREB-35, as well as FLAG-tagged CREB, were transfected into HEK293 cells. The FLAG-tagged CREB variants were visualized by FLAG-immunoreactivity by using Cy3-conjugated secondary antibody and fluorescence microscopy. The nucleus (n) of each labeled cell is indicated. The results demonstrate that CREB-14 and CREB-35 are localized primarily in the cytosol. In contrast, CREB is localized predominately in the cell nucleus .

Effects of CREB-14 and CREB-35 on CRE-Reporter activity. To study the functional characteristics of the CREB variants, we cotransfected each with a CRE-CAT reporter construct into HEK293 cells. Basal and forskolin-stimulated CRE-CAT activity was determined in the cotransfected cells. Forskolin incubation (5 µM) increased levels of CRE-CAT activity by more than 4-fold, relative to basal reporter activity (Fig. 5). Overexpression of either CREB-14 or -35 repressed CRE-CAT activity (Fig. 5). The inhibitory effects of the CREB variants were more prominent on basal CRE-CAT than on forskolin-stimulated activity (basal CRE activity: 100 ± 5.7; -14, 43 ± 3.3; -35, 31 ± 3.4; forskolin-stimulated CRE-CAT activity: control, 444 ± 21; -14, 341 ± 20; 35, 319 ± 12; mean ± S.E.M. percentage of control, basal CRE-CAT activity) (Fig. 5).

View larger version (60K): [in this window] [in a new window]   Fig. 5.   Effects of CREB-14 and CREB-35 expression on CRE-CAT activity. A, HEK293 cells were transfected with either CREB-14 pCI or CREB-35 pCI plus the CRE-CAT reporter construct. Cells were also transfected with pSV-gal as a marker for transfection efficiency. For the control, the empty pCI vector was transfected instead of constructs containing the CREB variants. Either basal or forskolin (5 µM) stimulated CRE-CAT activity was measured in transfected cells. Expression of either CREB-14 or CREB-35 suppressed CRE-CAT activity under both basal and forskolin-stimulated conditions. Data were normalized to the level of -gal activity. B, the influence of the CREB variants on the regulation of CRE-CAT activity in response to recombinant CREB was also determined. Expression constructs for CREB and CREB-14 or -35 were cotransfected into HEK293 cells as indicated. The cells also were transfected with CRE-CAT, different amounts of the empty pCI vector to serve as a control and to maintain equivalent amounts of DNA under each condition, and with pSV-gal as a marker for transfection efficacy. Either basal or forskolin (5 µM)-stimulated CRE-CAT activity was measured in transfected cells. The results are the mean ± S.E.M.; n = 6 per group. *P < .01 or **P < .05 compared with control (Student's t test).

CREB Variants Are Regulated in CREB-Overexpressing Transgenic Mice. The CREB gene contains a CRE, and previous studies have demonstrated that activation of the cAMP system regulates the expression of CREB isoforms (Widnell et al., 1994). To determine whether CREB-14 and -35 are also regulated by the cAMP system, we examined the expression of these variants in CREB-overexpressing mice. We recently have developed inducible CREB-overexpressing transgenic mice by using the tetracycline-regulated system (Chen et al., 1998). In these mice, the tetracycline transactivator is under the control of the neuron-specific enolase promoter and expression of CREB is under the control of the tetracycline-responsive promoter. Addition of tetracycline induces a conformational change in tetracycline transactivator and thereby blocks its ability to activate tetracycline-responsive promoter-CREB expression (tetracycline-off system). For these studies, we used one of the lines (6-B) that exhibits high levels of CREB overexpression in the striatum and cerebellum (Chen et al., 1998).

View larger version (67K): [in this window] [in a new window]   Fig. 6.   Regulation of CREB-14 or CREB35 expression in CREB transgenic mice. Levels of CREB-14 (top) and -35 (bottom) mRNA were examined by RPA in the striatum of mice overexpressing CREB. Representative autoradiograms are shown on the left and quantitation of the results are shown on the right. The results demonstrate that there is a small but significant down-regulation of CREB-14 in the CREB-overexpressing mice. In contrast, levels of CREB-35 were up-regulated in the CREB transgenic mice. In both cases, levels of the CREB, , and  isoforms were significantly increased in the CREB transgenic mice, as a result of overexpression of the CREB transgene and up-regulation of CREB and  (see text for discussion). The results are expressed as mean ± S.E.M.; n = 10 per group. *P < .01 or **P < .05 compared with control mice (nontransgenic mice; Student's t test).

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

In the present study, we have identified two novel CREB splice variants, CREB-14 and CREB-35. Previous studies have demonstrated the presence of several CREB isoforms as a result of alternative splicing (Waeber et al., 1991; Ruppert et al., 1992; Ellis et al., 1995; Blendy et al., 1996; Girardet et al., 1996; Walker et al., 1996; Youg et al., 1997; Bartsch et al., 1998). In most of these previous reports, the CREB variants were generated by splicing out an entire exon (Ruppert et al., 1992; Youg et al., 1997) or the insertion of a novel exon (Waeber et al., 1991; Ruppert et al., 1992; Girardet et al., 1996; Bartsch et al., 1998). The CREB-14 and CREB-35 variants described in this study have a deletion of exon 5, similar to that in the CREB variant. In addition, CREB-14 and -35 also lack several nucleotides at the beginning of exon 9 and 8, respectively, indicating that they are novel CREB splice variants. This indicates that CREB-14 and -35 are generated by recognition of alternative intron-exon boundaries within these regions. In this regard, CREB-14 and -35 are similar to the CREB1 variant (Ellis et al., 1995), in which an insertion of several nucleotides was found at the beginning of exon 7.

The sequence deletions in CREB-14 and -35 suggest an alternative codon usage that would result in an early stop codon. This would result in the formation of proteins with reduced molecular weight relative to CREB, and loss of the carboxy bZIP domain that is required for dimerization and DNA binding. Expression by either a transcription/translation system or transfection of HEK293 cells demonstrates that CREB-14 and -35 have molecular masses of approximately 27 and 22 kDa, respectively. In addition, deletion of the bZIP domain in CREB-14 and -35 was directly demonstrated by immunoblot analysis by using antibodies that selectively recognize the amino or carboxy terminus of CREB. In cells overexpressing the CREB variants, immunoreactivity against the amino, but not the carboxy, terminus was observed, consistent with the sequence prediction. In addition, immunocytochemical analysis of cells expressing FLAG-tagged recombinant CREB-14 and -35 demonstrated that the variants were localized primarily to the cytosol, with very low levels in the nucleus, as would be predicted from the loss of nuclear localization signals and the DNA binding domain.

In contrast to these deletions, both variants contain the kinase-inducible domain located in exon 7, indicating that CREB-14 and -35 are substrates for phosphorylation by CREB kinases, including PKA and CaMKII and IV. This was examined by immunoblot analysis by using phospho-CREB specific antibodies. In the basal or untreated cells, levels of phospho-CREB immunoreactivity were undetectable. Stimulation of the cAMP cascade by incubation of cells with forskolin resulted in high levels of a phospho-CREB bands of the appropriate size in cells expressing either CREB-14 or -35 (i.e., 27 or 22 kDa, respectively). These findings confirm that CREB-14 and -35 are substrates of PKA, and suggest that they are likely to be substrates for other CREB kinases.

The functional characteristics of CREB-14 and -35 were also examined by using a CRE-CAT reporter assay. Expression of either CREB-14 or CREB-35 repressed both basal and forskolin-stimulated CRE-CAT activity in transfected cells. In addition, the function of recombinant CREB was also repressed by expression of the CREB variants. Previous studies have reported two other CREB variants that, like CREB-14 or CREB-35, lack the bZIP domain but retain the kinase inducible, phosphorylation domain. These are the CREBw isoform in rat (Waeber et al., 1991) and the CREB1c isoform in Aplysia (Bartsch et al., 1998). The cellular localization of these isoforms in the cytosol is similar to that found in the present study for CREB-14 and -35. However, one of these variants (CREBw) but not the other (CREB1c) was reported to repress CRE-reporter activity (Girardet et al., 1996; Walker et al., 1996; Bartsch et al., 1998). This discrepancy could be explained by different experimental conditions or expression efficiency.

The mechanism by which CREB-14, CREB-35, and CREBw inhibit endogenous CREB function is not clear. Because these variants lack the bZIP domain they can not dimerize or bind to DNA to inhibit CREB function. The most likely mechanism for the inhibition is that these variants may act as pseudosubstrates of PKA in the cytosol and thereby decrease the phosphorylation of endogenous CREB in the nucleus. The question remains as to whether endogenous expression of these isoforms is at a level that is sufficient to repress CREB function under physiological conditions. CREB-14 and CREB-35 are widely expressed in all tissues that were examined, and were particularly enriched in brain, thymus, and testes. Moreover, CREB-35 mRNA appears to be abundant and is expressed at high levels relative to the other major CREB isoforms, CREB , , and . Expression of immunoreactive bands of the appropriate molecular weight were also observed, indicating that CREB-14 and -35 are expressed at the protein levels. In addition, the finding that CREB-35 is up-regulated in CREB transgenic mice suggests that expression of this isoform may serve a negative feedback role for the cAMP system. These findings raise the possibility that CREB-35, as well as CREB-14, could influence CREB function in vivo. Alternatively, these isoforms may have other, unidentified functions in the cytosol that are regulated by phosphorylation. This possibility is supported by studies demonstrating that CREB is present in developing dendrites (Crino et al., 1998) and in mitochondria (Cammarota et al., 1999).

Yet another possibility is that CREB-14 and CREB-35 may be aberrant splice variants that are repressed under physiological conditions. This has been demonstrated for other aberrant splice variants that are associated with human disease (Nakai and Sakamoto, 1994). For example, the pathogenesis of sporadic amyotrophic lateral sclerosis recently was determined to result, in part, from aberrant RNA processing of a glutamate transporter (EAAT2; Lin et al., 1998). In this regard, the CREB-14 and CREB-35 variants could be involved in the regulation of the cAMP-CREB signaling pathway under pathological conditions. Further examination will be needed to determine the function of the CREB-14 and -35 variants in normal physiological signaling, and possibly in aberrant conditions.