Address correspondence and reprint requests to Carrol D'Sa, 34 Park Street, Room S310, New Haven, CT 06508, USA. E-mail: firstname.lastname@example.org
This study examined the regulation of all known phosphodiesterase (PDE) type PDE4A, PDE4B and PDE4D splice variants in cortical neurons by cAMP signaling. Treatment with dibutyryl-cAMP (db-cAMP) caused the induction of two of the known splice variants, PDE4B2 and PDE4D1/PDE4D2. Although the splice variants PDE4A1, PDE4A5/PDE4A10, PDE4B3, PDE4B1, PDE4D3 and PDE4D4 were present in cortical neurons, their mRNA was not regulated at the transcriptional level by db-cAMP. To assess the increase in PDE4B2 and PDE4D1/D2 mRNA expression, the promoters containing these genes were characterized. Transcription from both promoters was stimulated by db-cAMP. Because chronic antidepressant treatment increases PDE4B, and not PDE4D, mRNA expression, we focused on the regulation of the PDE4B2 promoter by cAMP and CREB. Dominant negative mutants of CREB suppressed PDE4B2 promoter activity and a constitutively active form of CREB robustly stimulated it. These data demonstrate that in cortical neurons, a short PDE4B2 intronic promoter is regulated by CREB, confers cAMP responsitivity and directs PDE4B2 mRNA and protein expression.
cyclic AMP response element binding protein, RPA, RNAse protection assay
sodium dodecyl sulfate-polyacrylamide gel electrophoresis
horse radish peroxidase.
Intracellular levels of cAMP within neurons are homeostatically maintained by the adenylyl cyclases, which synthesize and the phosphodiesterases (PDEs), which degrade this second messenger (Conti and Jin 1999). The cAMP-specific phosphodiesterases type IV (PDE4s) belong to the superfamily of PDEs and are distinct from the other cAMP-specific PDEs (e.g. PDE3, PDE7) in that they are highly sensitive to inhibition by the antidepressant, rolipram (Houslay and Milligan 1997; Conti and Jin 1999). Therefore, they serve as pharmacological targets for antidepressant intervention (Duman et al. 1997).
PDE4s are generated from four distinct genes (PDE4A−4D) with multiple splice variants generated from each gene (Houslay et al. 1998; Conti and Jin 1999). By alternative splicing and use of different promoters, each PDE4 gene generates long and short splice variants that form functional proteins, differing from each other in their N-terminal regions. Two conserved domains at the N-terminal of each PDE4 protein called upstream conserved region (UCR)-1 and UCR-2 contribute to a regulatory region between the isoform specific N-terminal region and the conserved catalytic domain (Houslay et al. 1998; Conti and Jin 1999). The long splice variants contain both UCR-1 and UCR-2, whereas the short splice variants lack UCR-1 (Houslay et al. 1998; Conti and Jin 1999). The UCR-1 in the long variants contains the consensus sequences for regulation by protein kinase A-mediated phosphorylation. In one of the long forms, PDE4D3, phosphorylation of Ser54 in UCR-1 results in activation of cAMP hydrolysis (Sette and Conti 1996). The unique N-terminal domain is also involved in membrane targeting of the specific splice variants (Houslay et al. 1998; Yarwood et al. 1999).
This level of complexity is manifested in the vast number of PDE4 splice variants (more than 16) that have been identified to date. For example, in rat, four PDE4A splice variants have been identified. The testis-specific PDE4A8 (Bolger et al. 1996; Naro et al. 1996), the relatively ubiquitous PDE4A5 (McPhee et al. 1995) and a newly identified PDE4A10 comprise the long splice variants (Rena et al. 2001). PDE4A1 is the only short PDE4A splice variant identified so far. It is membrane-associated and present mainly in brain (Shakur et al. 1995). Three PDE4B splice variants have also been identified in rat, i.e. two long isoforms, PDE4B3 (Huston et al. 1997) and PDE4B1 (Swinnen et al. 1989; Bolger et al. 1993) and the short splice variant, PDE4B2 (Swinnen et al. 1989; Lobban et al. 1994). The PDE4D family is the best-characterized PDE4 family with three long splice variants PDE4D5, PDE4D4 and PDE4D3, and the two short splice variants, PDE4D1 and PDE4D2 (Bolger et al. 1997; Conti and Jin 1999).
The large number of splice variants suggests that different regulatory mechanisms impinge on the promoters of these genes to direct tissue-specific expression of the PDE4 isoforms (Swinnen et al. 1989). This is supported by studies in a variety of peripheral systems which indicate that PDE4s can be regulated either at the level of gene transcription (Verghese et al. 1995; Manning et al. 1996) or by a post-translational increase in protein phosphorylation upon activation of the cAMP cascade (Lim et al. 1999; Oki et al. 2000). Most studies to date have focused on the regulation of PDE4s in peripheral systems, with very little information about the regulation of the PDE4s, particularly the splice variants, in brain. Within the brain, PDE4A, PDE4B and PDE4D are highly expressed, whereas PDE4C is absent (Conti and Jin 1999). Previous studies have demonstrated that chronic antidepressant treatment increased PDE4A and PDE4B gene expression in rat frontal cortex, but PDE4D gene expression was unchanged (Ye et al. 1997; Takahashi et al. 1999; Ye et al. 2000). Because chronic antidepressant treatment up-regulates the cAMP cascade in brain (see Duman et al. 1997) and PDE4s are homeostatically regulated by cAMP, the main objective of this study was to study the regulation of all known PDE4 splice variants in cortical neuronal cultures by activation of the cAMP system. We used cortical neuronal cultures because this system closely resembles the physiological milieu of the brain.
The results indicate that certain PDE4 splice variants are constitutively expressed in cortical neurons. Following activating of the cAMP cascade, certain PDE4 splice variants accumulate in neurons by a regulatory process involving increased gene expression. This study will provide insights into how the various PDE4 splice variants are regulated by cAMP-CREB cascade and the molecular mechanisms by which chronic antidepressant treatment regulates this pathway in brain.
Methods and materials
All cell culture reagents were from Life Technologies (Rockville, MD, USA). Dibutyryl-cAMP (db-cAMP) and Actinomycin D were purchased from Sigma (St Louis, MO, USA).
Cortical neuronal cultures and cell cultures
All experiments were performed using cortical neurons dissected from E-17 Sprague–Dawley rats. Neurons were grown in Neurobasal media containing B27 supplement, l-glutamine and penicillin/streptomycin (Life Technologies). The cells were seeded onto culture dishes at a density of 4 × 106 cells per well in a 6-well plate or 8 × 106 cells per 60-mm dish. The dishes were pre-treated overnight with 0.2% PEI in 50 mm sodium borate, pH 8.4. CATH.a and SHSY-5Y cells were grown according to published procedures (Boundy et al. 1998).
RT-PCR cloning of PDE4A, PDE4B and PDE4D splice variant-specific probes
Total RNA was extracted from rat olfactory tubercle, cerebellum and hippocampus using the RNAqueous kit according to manufacturer's standard protocol (Ambion, Austin, TX, USA). A partial PDE4A1 cDNA (203 bp) was obtained by RT-PCR using the olfactory tubercle as template and oligonucleotides based on the rat PDE4A1 cDNA sequence (Accession #L27062). The forward (Primer 1) and reverse oligonucleotides (Primer 2) were 5′-GCTTCCATCATG CCTCTGGTTGAC-3′ and 5′-TGTGGGTGAGGGGATCTCCACT TC-3′. A partial PDE4B3 cDNA (402 bp) was obtained by RT-PCR using cerebellar RNA as template and oligonucleotides based on the rat PDE4B3 sequence (Accession #U95748). The forward (Primer 3) and reverse (Primer 4) oligonucleotides were 5′-ACCCTCAAGC CAGCTTCTTCAG-3′ and 5′-TGACAGAGCGGTAGGTCTGGATG-3′. A partial PDE4D4 cDNA (519 bp) was obtained by RT-PCR using hippocampal RNA as template and oligonucleotides based on the rat PDE4D4 sequence (Accession #AF031373). The forward (Primer 5) and reverse (Primer 6) oligonucleotides were 5′-CACCTCCTACGCGGTGGAGACTG-3′ and 5′-GTCTGCAGGGTCTCTAGCTGGTC-3′. All RT-PCR reactions were performed using the Access RT-PCR kit (Promega, Madison, WI, USA) according to manufacturer's standard recommendations. The RT-PCR products of predicted size were isolated from an agarose gel, subcloned into the pGEM-T Easy vector (Promega, San Diego, CA, USA) and verified by sequencing.
RNAse protection assays of splice variants
At 5–6 days in vitro (DIV), cortical neurons were treated with db-cAMP or pre-treated with actinomycin D, 30 min before cAMP. Three hours later, total RNA was isolated from cortical neuronal cultures using the RNAqueous kit (Ambion). Briefly, [α-32P] CTP-labeled antisense PDE4A, PDE4B or PDE4D riboprobes were generated using 1 µg of the plasmids described above as templates (linearized with Nco1) and SP6 RNA polymerase (MAXIscript SP6 Transcription kit, Ambion). The RNAse protection assay (RPA) was performed according to manufacturer's standard protocol with the RPA III kit (Ambion) using 6 µg of total RNA to detect PDE4A, PDE4B and PDE4D mRNA, respectively. Total RNA was incubated with 1 × 105 cpm of synthesized [α-32P] CTP-labeled antisense PDE4A, PDE4B or PDE4D riboprobes for 16 h at 58°C in hybridization buffer (RPA III kit). Samples were treated with RNAse A/RNAse T1 mix (1 : 100 dilution) for 30 min at 37°C followed by sodium dodecyl sulfate (SDS)/Proteinase K digestion (RPA III kit). After ethanol precipitation, samples were resuspended in gel loading buffer (RPA III kit). Protected fragments were loaded onto an 5% acrylamide/8 m urea denaturing gel according to manufacturer's recommendations. Gels were exposed to X-ray film using an intensifier screen and autoradiograms were quantified as described below.
Western blot analysis
Cortical neurons were treated with db-cAMP at 4–5 DIV. After 5 h, neurons were washed twice with ice-cold phosphate-buffered saline (PBS), scraped from dishes and harvested in homogenization buffer containing 1% SDS with protease inhibitors. Samples were boiled, proteins (25–30 µg) separated on a 7.5% SDS–PAGE (polyacrylamide gel electrophoresis) gels and transferred to nitrocellulose membranes. The membranes were blocked in PBS–0.05% Tween 20 (PBST) buffer and incubated overnight with monoclonal antibodies directed against the PDE4A and PDE4D protein at a dilution of 1 µg/mL (Takahashi et al. 1999). These monoclonal PDE4A and 4D antibodies were made against the C-terminus of recombinant human PDE4A and PDE4D proteins and were generously provided for research purposes by Pfizer Corporation (Groton, CT, USA). Each antibody has been shown to be specific for its respective PDE4 isozyme by our laboratory previously (Takahashi et al. 1999). The PDE4B monoclonal antibody was used at a dilution of 0.5 µg/mL. This monoclonal PDE4B antibody (96G7A) was made against the C-terminus of recombinant human PDE4B2 protein, purified from mouse ascites by Protein-A affinity chromatography and was generously provided for research purposes by ICOS Corporation (Bothell, WA, USA). The filters were incubated with goat anti-mouse secondary antibody (1 : 2000) conjugated to horseradish peroxidase (HRP; Vector Laboratories, Burlingame, CA, USA). The nitrocellulose filters were developed using the ECL system and exposed to Hyperfilm (Amersham Life Science, Cleveland, OH, USA).
pGL2-basic, a promotorless luciferase vector was used to test the activity of the PDE4B and PDE4D promoters. The isolation of the 4B and 4D promoters and their sequences has been described previously (Monaco et al. 1994; Vicini and Conti 1997). To make the PDE4B deletion mutants, the 4B promoter (0.98 kb) was excised from pGL2 with EcoR1 and Hind3 and subcloned into EcoR1/Hind3 sites of pBS vector to create 0.98 4BpBS. To make the 0.55 kb 4B deletion mutant (−550 to + 1), 4BpBS was digested with Xba1, re-ligated and the 0.55 kb 4B fragment was subcloned into pGL2. The 0.1 PDE4BLuc (−105 to + 1) was made by PCR using 0.98 4BpBS as template and synthetic oligonucleotides containing EcoR1 and Hind3 restriction sites, respectively. The PCR fragment was purified, digested with EcoR1 and Hind3 and subcloned into pGL2. The construction of the 1.5, 1.1 and 0.2 kb PDE4DLuc constructs have been described previously (Vicini and Conti 1997). All sequences were confirmed by sequencing at the Keck facility (Yale University). The mutant CREB expression plasmids, A-CREB, M1-CREB and CREB-VP16 have been described in detail previously (Shieh et al. 1998; Tao et al. 1998). Briefly, M1-CREB is a mutant in which Ser-133 of CREB is mutated to an alanine residue. This mutant can bind to the CRE site within genes; however, it cannot be activated via phosphorylation thus inhibiting CRE-mediated gene transcription (Shieh et al. 1998; Tao et al. 1998). The A-CREB mutant dimerizes with CREB family members and prevents them from binding to CRE sites (Shieh et al. 1998; Tao et al. 1998). CREB-VP16 is a constitutively active CREB mutant in which CREB is fused at its N-terminus to the transactivation domain of a herpes virus activator (Shieh et al. 1998; Tao et al. 1998).
Transfections and luciferase activity assay
Transfections were performed on DIV 5 neurons in vitro using 2 µg of 4B promoter DNA and 6 µL of Fugene 6 Transfection reagent according to manufacturer's instructions (Roche Molecular Biochemicals, Indianapolis, IN, USA). After 4 h, media was replaced with conditioned media. For the CATH.a and SHSY-5Y cells, media was replaced after 6–7 h. Db-cAMP was added 24 h post-transfection and luciferase assay performed 48 h post-transfection. For the co-transfections, we used 2 µg of 4B promoter DNA, along with a threefold excess of A-CREB, M1-CREB, CREB-VP16 or control vector to equalize the amounts of DNA added per well. For the luciferase assay, individual dishes were washed twice with PBS and lysed in 200 µL lysis buffer. Twenty microlitres of the cell extract were used for a luciferase reporter assay (Roche), which was measured on a luminometer. Luciferase activity was normalized relative to the protein content in each well to correct for individual differences.
Site-directed mutagenesis of the PDE4B promoter
Site-directed mutagenesis was performed with the Promega site-directed mutagenesis kit using manufacturer's standard instructions (Promega). The CRE site TGACATCA in the 0.98 kb PDE4BpGL2 was mutated to CTCGAGTA containing an Xho1 site. The mutation was confirmed by restriction digestion and sequence analysis (Keck Facility, Yale University).
Levels of PDE4A, PDE4B or PDE4D splice variant mRNA were determined by outlining the bands on the RPA films and quantified on the Macintosh-based NIH Image analysis program. 14C step standards were used to verify the linearity of densitometry. All results were subjected to statistical analysis. Experiments were subjected to Student's t-tests and anova, with post hoc significance determined at the p < 0.05 level.
Effect of db-cAMP on mRNA expression of PDE4A, PDE4B and PDE4D splice variants in cortical neuronal cultures
Several different PDE4A, PDE4B and PDE4D splice variants with diverging 5′ ends have been identified (Houslay et al. 1998; Conti and Jin 1999). To determine which PDE4A, PDE4B or PDE4D splice variants are present in cortical neurons, we used a sensitive RNAse protection assay that distinguishes between all known PDE4A, PDE4B and PDE4D long and short splice variants, respectively. The construction of these antisense cRNA riboprobes has been described in Methods and materials. RNA extracted from cortical neurons was hybridized to a riboprobe that recognizes PDE4A splice variants (Fig. 1).
The PDE4A probe completely protects 203 bp of the short splice variant, PDE4A1 (Accession #L27062), and partially protects 131 bp of the long splice variants, PDE4A5 (Accession #L27057)/PDE4A10 (Accession #AF110461). Because the PDE4A riboprobe was designed at the splice junction of the long and short splice variants, it cannot distinguish between the long PDE4A5 and PDE4A10 splice variants. RNAse protection assays (RPA) revealed that PDE4A1 and PDE4A5/PDE4A10 were present in neurons under basal conditions. We next examined the effect of perturbations in cAMP signaling on PDE4A mRNA expression by treating neurons with the cell-permeable cAMP analog, db-cAMP. RPA analysis revealed that db-cAMP did not influence PDE4A1 or PDE4A5/PDE4A10 splice variant mRNA levels in the cortical neurons.
We next examined the effect of db-cAMP on expression of the PDE4B splice variants (Fig. 1). The PDE4B antisense riboprobe was designed to completely protect 402 bp of the long splice variant PDE4B3 (Accession #U95748), partially protect 205 bp of the long splice variant PDE4B1 (Accession #J04563) and partially protect 95 bp of the short splice variant, PDE4B2 (Accession #L27058). In addition, this probe also picked up an additional band of approximately 275 bp, which most likely corresponds to a novel long splice variant, designated PDE4BX1. This has not been previously described, possibly because the probe was designed to overlap the splice junction of the long and short variants. Results indicate that PDE4B3, PDE4B1 and PDE4BX1 were present in neurons under basal conditions. However, their mRNA expression was unaffected by db-cAMP treatment. The short splice variant, PDE4B2 was barely detectable in neurons under basal conditions. However, treatment with db-cAMP caused a dramatic increase of about 3.6 fold compared to unstimulated neurons (Fig. 1).
Next the expression of PDE4D splice variants was determined (Fig. 1). The PDE4D probe was designed to completely protect 519 bp of the long splice variant PDE4D4 (Accession #AF031373), partially protect 433 bp of the long splice variants PDE4D3 (Accession #U09457)/PDE4D5 (rat clone has not yet been identified) and partially protect 80 bp of the short splice variants PDE4D1 (Accession #U09455)/PDE4D2 (Accession #U09456). In addition, this probe also picked up an additional band of approximately 250 bp. Because the PDE4D riboprobe was designed at the splice junction of the long and short splice variants, it cannot distinguish between the long PDE4D3 and PDE4D5 and the short PDE4D1 and PDE4D2 splice variants. Treatment with db-cAMP did not affect the long splice variants PDE4D4 and PDE4D3/PDED5. PDE4D1/PDE4D2 mRNA was not evident in unstimulated neurons but was strongly induced by 2.7 fold following cAMP treatment (Fig. 1).
Effect of actinomycin D on PDE4A, PDE4B and PDE4D spice variants expression
Increases in PDE4B and PDE4D splice variant mRNA expression could be due to increased mRNA stability or increased gene transcription. To address this issue, cortical neurons were exposed to actinomycin D, an inhibitor of gene transcription, 30 min prior to db-cAMP treatment and PDE4B and 4D mRNA levels were assessed 3 h later. Exposure to actinomycin D significantly decreased cAMP-stimulated PDE4B2 mRNA levels (Fig. 2) and completely abolished the cAMP-induced PDE4D1/PDE4D2 mRNA levels. There was no change in the long PDE4B and PDE4D splice variants or the short and long PDE4A splice variants (Fig. 2). Actinomycin D alone had no effect on the basal levels of PDE4 variants (data not shown). These results are consistent with the conclusion that cAMP modulates PDE4B2 and PDE4D1/D2 production in cortical neurons at the transcriptional level.
Effect of cAMP on PDE4A, PDE4B and PDE4D immunoreactivity in primary cultures
We next examined the influence of db-cAMP on PDE4A, PDE4B and PDE4D immunoreactivity. Using a monoclonal antibody directed against PDE4A, we detected several immunoreactive bands in cortical neuronal homogenates. Previously, we have demonstrated the specificity of these bands as by pre-incubation of the antibody with recombinant PDE4A protein (Takahashi et al. 1999). The major immunoreactive bands at 75 and 110 kDa correspond to the short and long splice variants, PDE4A1 and PDE4A5, respectively. Immunoblot analysis revealed that although both these variants were present in neurons under basal conditions, their immunoreactivity was not changed by db-cAMP treatment. This is consistent with the lack of effect of db-cAMP on PDE4A mRNA levels.
Using a monoclonal antibody against PDE4B, we also detected several imumunoreactive species. Under basal conditions, there were three major bands. Two of the bands at 92 and 110 kDa corresponded to the long splice variants, PDE4B3 and PDE4B1 already reported (Lobban et al. 1994; Huston et al. 1997). None of these bands were regulated by db-cAMP. In contrast, a 5-h treatment with db-cAMP caused a significant induction of an immunoreactive species of 65 kDa (Fig. 3).
The PDE4D-specific monoclonal antibody recognizes three major immunoreactive bands corresponding to 105, 98 and 93 kDa. These bands correspond to the long PDE4D splice variants, PDE4D4, PDE4D5 and PDE4D3, respectively. The identity of these bands has been verified previously by pre-incubation of the antibody with recombinant PDE4D protein (Takahashi et al. 1999). We did not detect any bands at 68 kDa corresponding to the short splice variants, PDE4D1 or PDE4D2 under basal conditions or in the presence of db-cAMP. The absence of these bands under basal conditions is similar to what has been reported previously in brain leading to the conclusion that these short PDE4D isoforms are absent or expressed at very low levels in brain (Iona et al. 1998; Takahashi et al. 1999). Because PDE4D1/PDE4D2 mRNA is detected after db-cAMP treatment, the results suggest that either the protein is not translated or that it turns over very rapidly. We detected a very faint band at 98 kDa corresponding to the long splice variant, PDE4D5 and a strong band at 105 kDa corresponding to PDE4D4. None of these bands were regulated by db-cAMP. However, we observed a db-cAMP-induced shift in the electrophoretic mobility of a long PDE4D isoform (93 kDa) corresponding to the PDE4D3 spice variant, suggesting that PDE4D3 may be modified post-translationally by protein phosphorylation in response to cAMP.
Functional characterization of the PDE4B2 promoter and effect of db-cAMP on promoter activation in neurons. Having obtained evidence that cAMP triggers the activation of PDE4B2 at the transcriptional level, we wanted to characterize the 5′-flanking promoter region that contributed to cAMP responsiveness. The PDE4B2 promoter is a short intronic promoter containing 983 bp of the PDE4B 5′-flanking region upstream of the translation start site for PDE4B2 (−983 to + 1). It contains two putative TATA boxes, several AP-1 binding sites and two non-canonical CREs. The first CRE (TGACATCA) is located downstream of the putative transcription start site, in the 5′-untranslated region (−256 bp upstream of the translation start site). This sequence differs from the canonical CRE palidrome (5-TGACGTCA-3) at position 5 (Montminy et al. 1990). A second non-canonical CRE (TCACGTAA) is located − 574 bp upstream of the +1 ATG. The isolation of this promoter, its sequence analysis and transcription start sites has been described previously (Monaco et al. 1994).
To examine the activity of the PDE4B promoter in neurons, DNA containing the 0.98 kb 5′-flanking region construct was transiently transfected into cortical neurons and compared to transient transfections of two standard neural-derived cell lines, CATH.a cells (a mouse catecholaminergic cell line) and SHSY-5Y cells (a human neuroblastoma cell line). These studies showed that basal transcription from the 0.98 kb PDE4B gene was highest in CATH.a cells, followed by SHSY-5Y cells and then neurons. To identify regulatory elements, we performed deletional analysis of the PDE4B2 promoter. In the 0.55 kb mutant (−550 to +1), we deleted the region containing the CCAAT box, CRE-2, AP-1 sites and the distal cap site but retained the proximal cap site, the TATA box and the non-canonical CRE-1 (0.55 kb4BLuc). Interestingly, deletion of 433 bp of the 5′-flanking sequence showed a trend towards an increase in basal transcriptional activity in neurons and SHSY-5Y cells, although it was not statistically significant (Fig. 4). We also deleted up to −105 bp (0.1kb4BLuc; −105 to + 1) of the 4B promoter. This fragment does not contain the cap sites, TATA box or the CRE element. Basal levels of promoter activity were dramatically reduced in this construct.
We next determined cAMP inducibility of the PDE4B promoter. Cortical neurons transfected with full-length 0.98 kb PDE4B promoter responded to db-cAMP treatment with a four- to fivefold increase in activity. On the other hand, activity of the 0.98 kb promoter was stimulated 1.8-fold in CATH.a cells and sevenfold in SHSY-5Y. Activity of the 0.55 kb PDE4BLuc fragment was also robustly stimulated by db-cAMP, suggesting that the 453 bp upstream of this region were not required for cAMP inducibility. The 0.1 kb PDE4BLuc fragment was not significantly stimulated by cAMP (Fig. 4).
Functional characterization of the PDE4D1/D2 promoter and effect of db-cAMP on promoter activation in neurons. We next examined the activity of the PDE4D1/D2 promoter in cortical neurons and compared it to CATH.a and SHSY-5Y cells. This promoter directs the expression of both short splice variants, PDE4D1 and PDE4D2. The isolation of this promoter, the construction of 1.1 and 0.2 kb PDE4D1/D2 deletion mutants and its functional characterization in response to cAMP and FSH in Sertoli and MA10 cells has been reported (Monaco et al. 1994; Vicini and Conti 1997). However, the regulation of this promoter in neurons has not been examined. Unlike the PDE4B promoter, the PDE4D1/D2 promoter lacks TATA and CAAT boxes and contains several AP-2, SP-1 and GC-rich regions. It also contains at least three distinct mRNA cap sites. Interestingly, like the PDE4B promoter, the PDE4D promoter contains a CRE located in the 5′-untranslated region (5′-UT; − 83 bp upstream of the ATG) and downstream of the transcription start site (Vicini and Conti 1997).
We initially examined basal activity of the PDE4D1/D2 promoter by transiently transfecting neurons, CATH.a and SHSY-5Y cells. Basal activity of the PDE4D promoter was highest in CATH.a, followed by SHSY-5Y and then neurons (data not shown). Interestingly, basal activity of the PDE4D promoter was significantly lower than the PDE4B promoter in all cell lines tested (data not shown). The 1.1 kb PDE4DLuc (−1218 to −121) contains all the regulatory elements present in the full-length 1.5 kb PDE4DLuc (−1540 to + 2) but lacks the 5′-UT that contains the putative CRE element (−83). The 0.2 kb PDE4Dluc (−299 to −121) contains 1 NF-κB site and two SP-1 sites. It, however, lacks the proximal cap sites and the 5′-UT that contains the putative CRE element. Highest basal activity was observed with the full-length 1.5 kb fragment, whereas very low activity was observed with the 0.2 kb fragment.
We next determined cAMP inducibility of the 4D promoter. Cortical neurons transfected with full-length 1.5 kb PDE4D promoter responded to db-cAMP treatment with a threefold increase in activity (Fig. 5). On the other hand, activity of the 1.5 kb promoter was stimulated 2.1-fold in CATH.a cells and 6.2-fold in SHSY-5Y cells. cAMP-stimulated activity of the 1.1 kb 4DLuc fragment was significantly lower than the 1.5 kb fragment in neurons and SHSY-5Ys, suggesting the CRE present in the 5′-UTR is required for cAMP inducibility in these cells (Fig. 5).
Mutagenesis of CRE-1 inhibits PDE4B gene promoter activity
Previously, we had examined the regulation of PDE4B and PDE4D in rat brain and found that chronic antidepressant administration increased PDE4B, and not PDE4D, mRNA expression and protein immunoreactivity in rat frontal cortex (Takahashi et al. 1999). Because chronic antidepressant treatment up-regulates the cAMP-CREB cascade, we examined whether the non-canonical CRE-1 site in the 5′-flanking region directly contributes to cAMP-dependent PDE4B promoter activity. A mutation was made by site-directed mutagenesis rendering this cis-acting DNA sequence incapable of binding CREB or its family members. The sequence TGACATCA within the full-length PDE4B promoter construct (0.98 PDE4B) was changed to CTCGAGTA (altered nucleotides are underlined; Fig. 6).
Mutation of the CRE-1 significantly suppressed basal and cAMP-stimulated PDE4B promoter activity by 37% and 52%, respectively, suggesting that this site contributes to cAMP-dependent regulation of PDE4B gene promoter activity.
Influence of CREB dominant negative mutants on PDE4B gene promoter activity
To examine whether PDE4B promoter activation by cAMP is dependent on CREB, we used two different dominant negative CREB mutants, a Ser133 to Ala phosphorylation mutant (M1-CREB), and a bZIP domain mutant (A-CREB) (Tao et al. 1998). Cortical neuronal cultures were cotransfected with the 0.98 kb PDE4B promoter along with a control vector plasmid, M1-CREB or A-CREB. One day after transfection, cells were left untreated or treated with db-cAMP and analyzed for luciferase activity 12 h later. We found that M1-CREB and A-CREB significantly blocked db-cAMP-stimulated PDE4B promoter activity (Fig. 7).
M1-CREB decreased cAMP-stimulated PDE4B promoter activity by 54%, whereas A-CREB decreased PDE4B promoter activity by 48%. In addition, M1-CREB also significantly suppressed basal transcription from the 4B promoter by 40%. The inability of A-CREB to suppress basal transcription may be due to the different vector backbones present in these two constructs. These results suggest that CREB or one of its dimerization partners transactivates the PDE4B promoter in response to cAMP stimulation.
Influence of constitutively active CREB protein on PDE4B gene promoter activity
To further investigate the role of CREB in activating the PDE4B promoter, we co-transfected a constitutionally active form of CREB (CREB-VP16) and examined its effect on PDE4B promoter activity. CREB-VP16 is a chimeric protein in which the N-terminus of CREB is fused to the transactivation domain of a herpes virus transcriptional activator (Tao et al. 1998); CREB-VP16 lead to an accumulation of PDE4B-luciferase reporter gene in cortical neurons (251%), even in the absence of db-cAMP stimulation (Fig. 7). Although this increase was not to the same extent as seen with treatment with db-cAMP in the same experiment (423%), it was significantly higher than cells transfected with control vector, suggesting that CREB-VP16 activates the PDE4B gene independent of cAMP stimulation. Addition of db-cAMP did not further enhance CREB-VP16-stimulated PDE4B activity.
The main objective of the present study was to examine which PDE4 splice variants are present in cortical neurons and determine how they are regulated by perturbations in intracellular cAMP signaling. We also examined regulation of the gene promoters of the splice variants responsive to cAMP signaling. Using a sensitive RNAse protection assay, we demonstrated that neurons express PDE4A5/PDE4A10, PDE4A1, PDE4B3, PDE4B1, PDE4D4 and PDE4D3 splice variants under basal conditions. Cortical neurons also express low levels of the short PDE4B splice variant, PDE4B2, although the short variant (s) of PDE4D, PDE4D1/D2, are undetectable. The results also demonstrated differential regulation of PDE4A, PDE4B, and PDE4D gene expression in response to activation of the cAMP-CREB cascade.
Incubation of cortical neurons with db-cAMP did not influence either the short or long splice variants of PDE4A. These results are interesting because in other systems, such as monocytes, treatment with β-adrenergic receptor agonists, in combination with the specific PDE4 inhibitor rolipram, increased PDE4A and PDE4B mRNA expression (Torphy et al. 1995). In rat brain, increased noradrenergic stimulation increased PDE4A immunoreactivity (Ye et al. 1997), whereas in the human neuroblastoma cell line, SHSY-5Y, no PDE4 isoforms were regulated by activation of the cAMP system (Engels et al. 1994). In human Jurkat T-cells, treatment with forskolin, an activator of adenylyl cyclase, decreased the expression of a novel long PDE4A isoform with no change in PDE4B or PDE4C expression (Erdogan and Houslay 1997). Taken together, these studies showed that PDE4A gene is selectively regulated in different cell types to meet the demand for homeostatic control of cAMP levels.
A recent study has reported the promoter isolation and characterization of a new human PDE4A long splice variant, PDE4A10 (Rena et al. 2001). This splice variant is present in brain and the promoter region of the human splice variant has a CRE element, besides an SP1 and GC-rich region. Interestingly, forskolin and the PDE inhibitor, IBMX did not influence promoter activity of a CRE-containing PDE4A10 luciferase construct, indicating that its CRE site is unresponsive in HeLa cells (Rena et al. 2001). Although the promoter of the long PDE4A5 variant has yet to be functionally characterized, the murine and human PDE4A1 splice variants have very short 5′-untranslated regions (131 bp) that do not contain any transcription start sites or regulatory elements (Sullivan et al. 1998; Olsen and Bolger 2000). Therefore, it is conceivable that the short PDE4A1, unlike the short PDE4B and PDE4D splice variants, is not transcriptionally modulated by cAMP. Previous studies have shown that chronic antidepressant treatment increased PDE4A isozyme mRNA and protein immunoreactivity in rat frontal cortex (Ye et al. 1997; Takahashi et al. 1999) and hippocampus (Ye et al. 2000). The mechanism underlying this regulation remains unclear. Our present results suggest that a direct regulation of PDE4A1 or PDE4A5 mRNA expression by acute activation of the cAMP system cannot account for antidepressant-induced up-regulation of PDE4A expression in cerebral cortex. However, it is possible that long-term activation of the cAMP cascade (e.g. days or weeks) could lead to altered expression of PDE4A.
Our results demonstrated that long forms of PDE4D are present in cortical primary cultures. The antisense riboprobe that we used to determine the presence of the long PDE4D splice variants in neurons could not distinguish between PDE4D3 and PDE4D5. From our immunoblot analysis, it appeared that PDE4D3, and not PDE4D5, is present in cortical neurons. Therefore the 433 bp band detected in the RPA analysis most likely corresponded to PDE4D3. This result is consistent with previous studies on splice variant expression of PDE4D3, as well as PDE4D4, in rat brain (Bolger et al. 1994, 1997; Sette et al. 1994; Iona et al. 1998). The 98 kDa fragment corresponding to PDE4D5 was very faint. We have found that this splice variant is present at very low levels in the brain, a finding consistent with other reports (Bolger et al. 1994, 1997; Iona et al. 1998). Levels of PDE4D3 mRNA were not altered by incubation with db-cAMP. The apparent molecular weight of PDE4D3 was, however, increased by incubation with db-cAMP, suggesting that this isozyme is phosphorylated by activation of the cAMP cascade. Additional studies including in vivo32P-incorporation would be needed to verify this suggestion. However, this conclusion is consistent with previous reports in thyroid cells where phosphorylation of Ser54 within the regulatory domain of PDE4D3 increased enzyme activity (Jin et al. 1998). It is possible that phosporylation-induced activation of PDE4D3 in neurons could contribute to short-term regulation of cAMP hydrolysis.
In contrast to the long splice variants, the short PDE4D splice variant(s) was induced in neurons by incubation with db-cAMP. The antisense PDE4D riboprobe used for these studies did not distinguish between the short PDE4D1 and PDE4D2 variant. Previous studies using RT-PCR have showed that both transcripts were present in adult rat brain (Bolger et al. 1997), although PDE4D2 immunoreactivity was present only in Sertoli cells and absent in brain (Conti et al. 1995; Iona et al. 1998). We did not observe any induction of PDE4D1/PDE4D2 immunoreactivity following cAMP treatment. One possible explanation for this result is that the short splice variant(s) accumulated at low levels in neurons and was not easily detectable. Therefore, PDE4D1/D2 may be regulated only in a restricted population of cortical neurons. This possibility is supported by previous studies that reported PDE4D1/PDE4D2 proteins were undetectable in brain. These investigators concluded that this promoter is silent in brain tissue (Iona et al. 1998; Takahashi et al. 1999). The detection and regulation of these mRNAs in the present study using the sensitive RPA analysis confirmed that the PDE4D1/PDE4D2 splice variant(s) was present in neurons and was transcriptionally regulated by cAMP.
Promoter analysis also demonstrated that the short PDE4D1/D2 intronic promoter was induced in cerebral cortical neuronal cultures and SHSY-5Y cells by activation of the cAMP system. This is similar to what was seen in thyroid and Sertoli cells where the PDE4D promoter directed the expression of PDE4D2, and not PDE4D1, protein (Vicini and Conti 1997; Jin et al. 1998). The mechanism underlying the regulation of the short PDE4D1/PDE4D2 promoter by cAMP in neurons and other cell lines remains unclear. The CRE site is located in the 5′-UT region that is included in the PDE4D2 mRNA in Sertoli cells. In our study, deletion experiments confirmed that this element contributed to cAMP induction in cerebral cortical neurons and SHSY-5Y cells. Deletion constructs lacking the CRE retained cAMP resposiveness. The CRE in the 5′-UT did not contribute to either the basal or cAMP induction of the PDE4D promoter in neurons. Identification of the exact role of this 5′-UT region will require further investigation. In Sertoli cells, this element is not necessary for the follicle stimulating hormone (FSH) induction of promoter activity, but constructs containing it showed the highest responsiveness (Vicini and Conti 1997). The GC regions, along with the AP-2 sites in PDE4D were also important for cAMP responsiveness in Sertoli cells (Vicini and Conti 1997) and these elements may play a role in cAMP-dependent activity of the PDE4D gene in neurons.
We found that the long PDE4B splice variants, PDE4B3 and PDE4B1, were also constitutively expressed in primary cerebral cortical neuronal cultures, whereas the short variant, PDE4B2, was expressed at very low levels. These low levels could explain why previous studies could not detect PDE4B2 mRNA and immunoreactivity in brain (Monaco et al. 1994; Iona et al. 1998). Interestingly, RPA analysis using a PDE4B2-specific probe had detected this transcript in rat brain (Bolger et al. 1994). Up-regulation of the short variant PDE4B2 by cAMP and prevention of this induction by actinomycin D indicated that this variant is transcriptionally regulated in neurons. Our results demonstrated that PDE4B2 protein was not detectable under basal conditions but was induced in cortical neurons by incubation with db-cAMP.
Promoter analysis also demonstrated that PDE4B is stimulated at least four- to fivefold by db-cAMP in cerebral cortical neurons as well as in two neuronal cell lines. These results are consistent with the induction of PDE4B2 mRNA and protein by activation of the cAMP cascade. Unlike the short form of PDE4D1/D2 promoter, the PDE4B2 promoter contains a TATA box and directs transcription more efficiently (Monaco et al. 1994). The PDE4B gene also contains a CRE (CRE-1) in the 5′-UT region that could be responsible for the induction of gene expression by activation of the cAMP cascade. Deletion analysis supports this possibility. Deletion mutants (−983,-1 LUC and −550,-1 LUC) containing this CRE were completely responsive to db-cAMP, whereas constructs lacking this site (−105,-1 LUC) were not induced by incubation with the cAMP analog. There is also a partial CRE (CRE-2) upstream of the TATA box (−574) but the complete responsiveness observed with the deletion construct lacking this site (−550,-1 LUC) indicated that it was not required for cAMP induction. Additionally, mutations in the CRE-1 element significantly inhibited PDE4B promoter activity, suggesting that this site contributed to cAMP-dependent regulation of PDE4B gene expression. There was also a decrease in the fold-stimulation by cAMP between the wild-type and mutated CRE-1 PDE4B. These results suggested that additional elements are involved in cAMP regulation of PDE4B promoter activity. Additional mutation studies of CRE-2 alone, as well as both sites together, will be required to directly examine their role in cAMP-induction of PDE4B gene expression.
The role of CREB in the regulation of PDE4B gene expression was further examined by expressing either a dominant negative CREB mutant or a constitutively active form of CREB. We used two different dominant negative CREB mutants, a Ser133 to Ala phosphorylation mutant (M1-CREB) and a leucine zipper mutant (A-CREB) that dimerizes with wild-type CREB but does not bind to DNA. Expression of these dominant negative CREB mutants significantly decreased the PDE4B promoter activity in cortical neurons in response to db-cAMP. In contrast, expression of the constitutively active form of CREB (CREB-VP16) significantly increased PDE4B promoter activity. These findings indicate that a CREB family member is essential for cAMP regulation of PDE4B2 promoter activity. An important next step will be to determine the identity of these CREB family members and the cis-elements that they bind to in the PDE4B2 5′ regulatory region.
Our findings, along with previous studies that antidepressant treatment up-regulates the cAMP-CREB cascade, raise the possibility that the PDE4B2 gene is a downstream target of antidepressant treatment. Activation of the cAMP-mediated protein kinase A cascade in cortical neurons ultimately leads to phosphorylation of CREB at Ser133 and activation of gene transcription (Shaywitz and Greenberg 1999). We have previously demonstrated that chronic antidepressant treatment increased CREB phosphorylation at Ser133 in rat brain (Thome et al. 2000). We have also reported that chronic antidepressant administration increased expression of PDE4B, although the specific splice variants that contributed to this regulation were not identified (Takahashi et al. 1999). The regulation of PDE4B2 by cAMP signaling in the present study suggests that this splice variant may contribute to the increase in PDE4B isozyme expression observed in rat frontal cortex following chronic antidepressant treatment. We are currently testing this hypothesis by studying the regulation of the PDE4B2 splice variant in rat frontal cortex by antidepressants.
In summary, our results demonstrated a rich and complex level of regulatory diversity for the three major PDE4 genes in the brain. We found that cortical neurons utilized different PDE4 isozymes depending on whether they are in a basal or cAMP-activated state. One functional consequence of long-term, cAMP-CREB-induction of PDE4B2 gene expression may be an attenuation of the antidepressant response. Therefore, a selective inhibitor of PDE4B may have antidepressant efficacy and/or could enhance the response to other classes of antidepressants.
We thank Michael Greenberg for kindly providing us with the A-CREB and CREB-VP16 plasmids. We thank Sharon Wolda at ICOS Corporation for generously providing us with the PDE4B monoclonal antibody. This work is supported by USPHS grants MH45481 and 2 PO1 MH25642, a Veterans Administration National Center Grant for PTSD, and by the Connecticut Mental Health Center.