The human cyclic AMP-specific phosphodiesterase PDE-46 (HSPDE4A4B) expressed in transfected COS7 cells occurs as both particulate and cytosolic species that exhibit distinct kinetics of inhibition by the antidepressant rolipram.

Transfection of COS7 cells with a plasmid encoding the human cyclic AMP-specific PDE4A phosphodiesterase PDE-46 (HSPDE4A4B) led to the expression of a rolipram-inhibited PDE4 activity, which contributed ∼96% of the total COS cell PDE activity. A fusion protein was generated which encompassed residues (788-886) at the extreme C terminus of PDE-46 and was used to generate an antiserum that detected PDE-46 in transfected COS7 cells. Immunoblotting studies identified PDE-46 as a ∼125-kDa species that was associated with both the soluble and particulate fractions. The relative Vmax of particulate PDE-46 was ∼56% that of cytosolic PDE-46. Particulate PDE-46 was not solubilized using Triton X-100 or high NaCl concentrations. Immunofluorescence analysis by laser scanning confocal microscopy showed that PDE-46 was located at discrete margins of the cell, indicative of association with membrane cortical regions. The human PDE4A species, h6.1 (HSPDE4A4C), which lacks the N-terminal extension of PDE-46, was found as an entirely soluble species when expressed in COS7 cells. h6.1 was shown to have an ∼11-fold higher Vmax relative to that of PDE-46. In dose-response studies rolipram inhibited particulate PDE-46 at much lower concentrations (IC50 = 0.195 μM) than those needed to inhibit the cytosolic enzyme (IC50 = 1.6 μM). The basis of this difference lay in the fact that rolipram served as a simple competitive inhibitor of the cytosol enzyme (Ki = 1.6 μM) but as a partial competitive inhibitor of the particulate enzyme (Ki = 0.037 μM; Ki′ = 2.3 μM). Particulate PDE-46 thus showed a ∼60-fold higher affinity for rolipram than cytosolic PDE-46.

Cyclic AMP plays a pivotal role in controlling a wide variety of biological processes. Regulation of the synthesis of this second messenger by G protein-controlled adenylate cyclase has been the subject of much investigation. However, it is becoming increasingly apparent that the control of degradation of cAMP is of fundamental importance. Such degradation is achieved by a large and diverse multigene family of cAMP phosphodiesterases (PDEs) 1 (1)(2)(3)(4)(5)(6)(7)(8). The activity and expression of the products of these genes can be variously regulated by signals emanating from all known pathways, thus offering a means through which cellular responses might be integrated (9).
The PDE4 cAMP-specific phosphodiesterase family was first recognized as a distinct entity subsequent to the use of both biochemical separation procedures and compounds, such as the antidepressant rolipram, which serve as selective inhibitors (10). This allowed for the categorization of various cyclic AMPspecific PDE (4) species that had been isolated and purified much earlier (11)(12)(13). Substantiation of the PDE4 enzyme family was achieved definitively by homology cloning, initially, using the Drosophila dunc PDE gene as a probe to isolate rat PDE4 species (14,15). This was, subsequently, extended to the isolation of PDE4 cDNAs from human beings (16 -21). PDE4 enzymes specifically hydrolyze cAMP in a fashion which is insensitive both to the action of cGMP either as a substrate or inhibitor, distinguishing then from PDE2 and PDE3 species, and to Ca 2ϩ /CaM, distinguishing them from PDE1 species (1,3,6,22). In human beings, PDE4 enzymes are encoded by four distinct genes which are located on three chromosomes (3,15,(23)(24)(25). The accepted nomenclature (1) for describing products of the PDE4 genes is based upon (i) the first two letters indicating the source species (e.g.. HS, Homo sapiens), (ii) the designator "PDE" for cyclic nucleotide phosphodiesterase, (iii) an arabic numeral for the gene family, (iv) a single letter for the gene (A/B/C/D for the PDE4 family), (v) an arabic numeral for the splice variant, and (vi) a single letter for each different reported clone.
The central domain of the various PDE4 enzymes is highly homologous (3,15) and is presumed to contain the catalytic site (26). The complexity of the PDE4 family is increased further by the occurrence of alternative mRNA splicing. This, primarily, takes the form of 5Ј-domain swapping to yield alternatively spliced forms with different N-terminal domains (3), although 2EL, a catalytically inactive human PDE4A variant has been reported (23) that exhibits both a 5Ј domain swap together with an insertion toward the 3Ј end. This results in a frameshift inducing premature truncation (23).
It has been suggested (27)(28)(29)(30)(31) that the function of the alternatively spliced N-terminal domains of the rat PDE4A family may be to allow for membrane/cytoskeletal association through protein-protein interaction and also regulation of enzyme activity through manipulation of the V max of the enzyme. Here we analyze the properties of the human PDE4A species PDE-46 (HSPDE4A4B) when expressed in COS-7 cells. We show that this enzyme is expressed in both cytosolic and particulate/ membrane compartments and that these two forms differ markedly as regards their catalytic activity (V max ) and the mechanism of inhibition by the selective PDE4 inhibitor and antidepressant rolipram. We also show that h6.1 (HSPDE4A4C), which lacks the N-terminal extension seen in PDE-46, is expressed in the cytosol where it is ϳ11-fold more active than cytosolic PDE-46.
Generation of a Plasmid Allowing the Expression of PDE-46 in COS Cells-The cloning of PDE-46 and generation of a plasmid containing the entire open reading frame of this PDE4A enzyme has been reported previously by one of us (16). The entire open reading frame for PDE-46 was excised from the p-Bluescript plasmid containing the PDE-46 cDNA as a 3.59-kilobase pair fragment using SpeI and NaeI. This was ligated into the vector pSv-SPORT and then cut using SmaI and SpeI to generate the plasmid pSV-SPORT-pde46.
Transfection of COS-7 Cells-COS-7 cells were seeded at approximately 1/3 confluency onto 10-cm diameter plates, 18 h before the transfection. Immediately before transfection the culture medium was replaced with 5 ml of Dulbecco's modified Eagle's medium (Life Technologies, Inc. Europe, Glasgow) supplemented with 10% (v/v) Nuserum (Collaborative Biomedical Products) together with 0.1 mM chloroquine. This solution was prepared by diluting the DNA to 250 l in TE buffer (10 mM Tris-HCl, 0.1 mM EDTA, pH 7.6) and adding 200 l of a 10 mg/ml DEAE dextran solution. The mixture was incubated at room temperature for 15 min before addition to the COS-7 cell culture. The cells were incubated for 3-4 h at 37°C in a 5% CO 2 atmosphere before the medium was aspirated and the COS cell culture shocked for 2 min with a 10% dimethyl sulfoxide in a phosphate-buffered saline solution. The culture was then rinsed twice in phosphate-buffered saline before Dulbecco' modified Eagle's medium containing 10% fetal calf serum was added, and the cells were incubated at 37°C in a 5% CO 2 atmosphere for 72 h.
Disruption of COS cells was done as described previously by us in some detail (30). This was done in KHEM buffer (50 mM KCl, 50 mM Hepes-KOH, pH 7.2, 10 mM EGTA, 1.92 mM MgCl 2 ) containing 1 mM dithiothreitol and a mixture of protease inhibitors at a final concentration of 40 g/ml phenylmethylsulfonyl fluoride, 156 g/ml benzamidine, and 1 g/ml each of aprotinin, leupeptin, pepstatin A, and antipain. Membrane pellets were resuspended in this mixture also. The high speed (P2) pellet fraction was generated essentially as described previously by us (27,28,30). This procedure routinely yielded a P1 pellet (1,000 ϫ g av for 10 min) and a P2 pellet (60 min at 100,000 ϫ g av ) as well as a high speed supernatant (S). The homogenization procedure was complete in that there was no detectable latent lactate dehydrogenase activity present in the P1 pellet, indicating an absence of cytosol proteins.
These were designed so as to contain recognition sites for the restriction enzymes BamHI and EcoRI, respectively. The plasmid pSV-SPORT-6.1 (32) formed the template in the following reaction: 1 g of template DNA, 25 pmol of each primer, 0.2 mM each dNTP, 50 mM KCl, 10 mM Tris-HCl (pH 9.0), 0.1% Triton X-100, 1.5 mM MgCl 2 in a final volume of 50 l with a mineral oil overlay. The reaction was subjected to the following conditions in a Techne PHC-3 thermocycler: denaturation for 1Ј at 94°C, annealing for 2 min at 37°C, extension for 3 min at 72°C for one cycle followed by denaturation for 1Ј at 94°C, annealing for 2 min at 60°C, extension for 3 min at 72°C for 30 cycles. Upon completion the reaction was extracted with an equal volume of chloroform and digested as follows: 10 units of BamHI, 10 units of EcoRI, 6 mM Tris-HCl (pH 7.5), 100 mM NaCl, 6 mM MgCl 2 , 1 mM dithiothreitol for 5 h at 37°C. The digested fragment was analyzed by electrophoresis through a 1.75% LGT agarose gel against HaeIII digested phi X174 DNA molecular weight markers, purified directly from the gel by Wizard PCR Prep kit (Promega), and cloned into the BamHI/EcoRI sites of the inducible bacterial expression vector pGEX-3X. This created an in-frame fusion with the GST gene in a plasmid which we have called pGEX-3X-(788 -886)-PDE46.
New Zealand White rabbits were immunized by subcutaneous injection with 120 g of the fusion protein GST-(788 -886)-PDE46 in complete Freund's adjuvant and boosted twice at monthly intervals in incomplete Freund's adjuvant. These antisera were referred to as PAb24/GST-(788 -886)-PDE46 and PAb23/GST-(788 -886)-PDE46. Rabbits were also immunized with native GST, and the antiserum was referred to as GST.
PDE-46 Peptides Used as Immunogens-This was done as described previously (28,33). Briefly, rabbit polyclonal antisera were generated against two peptides. These corresponded to the sequence C 218 KATLSEET 226 of PDE-46 and the sequence I 721 SMAQIPC 728 . Such species were chosen on the basis that they were near the N-and C-terminal residues of h6.1 and thus could be used to identify both h6.1 and PDE-46. Peptides were synthesized with a cysteine at the N terminus to facilitate conjugation to keyhole limpet hemocyanin.
Confocal Analyses-48 h after transfection, COS-7 cells were plated out onto coverslips (18 ϫ 18 mm) at about 60% confluency. After another 24 h the cells were fixed to the coverslips using paraformalde-hyde and labeled with antisera followed by staining with goat anti-(IgG) rhodamine B conjugate (TCS Biologicals) as described previously. Control experiments were done using cells which were treated with (i) preimmune antisera and (ii) goat anti-(IgG) rhodamine B conjugate and no primary antibody, and (iii) by using cells which had not been detergent-permeabilized with TX-100. Experiments were performed on four separate transfection studies in each instance. The data shown are typical of the cells analyzed. For each transfection, single optical sections were taken for ϳ10 cells with full galleries of images obtained for five cells where x-y optical sections were collected in the z plane through the entire cell at 0.2-m intervals.
The microtubule network was visualized using a mouse monoclonal antibody raised against ␣-tubulin. This antibody (34) was a generous gift from Prof. Keith Gull, School of Biological Sciences, University of Manchester, UK.
Analyses were done using a Zeiss laser scanning confocal microscope using an Axiovert 100 microscope with a ϫ63/1.4NA plan apochromat lens. A 543-nm laser line was selected with appropriate filters (LP570; KT 488/543) and a pinhole size of 8 -14. Three-dimensional reconstruction was done using either Zeiss LSM software on an IBM Pentium system or using Imaris (2.2.4) software (Bitplane AG, Zurich; Fairfield Imaging Ltd., East Sussex, UK) on a Silicon Graphics Indigo system.
Treatment with High Salt Concentrations-In a similar fashion to that described before (27,28,30), membranes (0.2 mg) from transfected COS cells were treated with KHEM buffer containing the indicated NaCl concentrations (final pH 7.2). The membranes were left on ice for 30 min at 4°C before centrifugation at 100,000 ϫ g for 1 h at 4°C. The resulting pellet was resuspended in KHEM buffer containing the appropriate NaCl concentration, and the pellet and supernatant fractions were analyzed by Western blotting (22,24).
Solubilization with Triton X-100 -In a similar fashion to that described before (27,28,30), membranes (0.2 mg) from transfected COS cells were treated with KHEM buffer containing the indicated Triton X-100 concentrations. The membranes were left on ice for 30 min at 4°C before centrifugation at 100,000 ϫ g for 1 h at 4°C. The resulting pellet was resuspended in KHEM buffer containing the appropriate Triton X-100 concentration and the pellet and supernatant fractions analyzed by Western blotting.
Relative Activity Determinations-This was done using a modification of the methodology described before by us (28,30,31). Briefly, increasing concentrations of protein (2-50 g) from COS-7 cells transfected with either PDE-46 or h6.1 were analyzed by Western blotting using 125 I-labeled second (anti-rabbit) antibody as a detection system. Quantitative data was obtained by PhosphorImager analysis with plots of intensity against/g of sample protein applied being plotted to gauge the relative concentrations of either h6.1 or PDE-46 in each of the preparations. For the V max determinations, amounts of protein from transfected COS-7 cells, which would provide equal amounts of these two PDE4 species, were taken. They were then assayed for PDE activity at 1 M cAMP substrate concentration and the maximum velocity calculated by substitution in the Michaelis-Menten equation having established the K m for the reaction. In some instances (ϳ25%), however, V max was determined from data obtained from full Michaelis plots using the fitting algorithms described below.
To define K m values, data from PDE assays done over a range of cAMP substrate concentrations were analyzed by computer fitting to the hyperbolic form of the Michaelis-Menten equation using an iterative least squares procedure (Ultrafit; with Marquardt algorithm, robust fit, experimental errors supplied; Biosoft, Cambridge, UK). Relative V max values could be calculated using the Michaelis equation and the experimentally derived K m values as indicated previously (28).
SDS-PAGE and Western Blotting-8 or 10% acrylamide gels were used and the samples boiled for 3 min after being resuspended in Laemmli (35) buffer. Gels were run at 8 mA/gel overnight or 50 mA/gel for 4 -5 h with cooling. For detection of transfected PDE by Western blotting, 2-50-g protein samples were separated by SDS-PAGE and then transferred to nitrocellulose before being immunoblotted using human PDE4A antisera. Labeled bands were identified by using antirabbit peroxidase-linked IgG and either the Amersham ECL Western blotting used as a visualization protocol or 125 I-labeled anti-rabbit antisera used for detection and quantification.
PDE Assay-Cyclic nucleotide phosphodiesterase activity was assayed by a modification of the two-step procedure of Thompson and Appleman (36) and Rutten et al. (37) as described previously by Marchmont and Houslay (12). All assays were conducted at 30°C, and in all experiments a freshly prepared slurry of Dowex:H 2 O:ethanol (1:1:1) was used for determination of activities. Initial rates were taken from linear time courses of activity. For the determination of kinetic param-eters, the PDE assays were conducted using cAMP concentrations over a range from 0.1 to 100 M. As indicated below, the COS7 cell transfection procedure utilized in this study led to very high levels of PDE-46 activity being produced such that they comprised Ͼ94% of the total COS cell PDE activity. Mock transfections, with vector only, as indicated in previous studies (27,28,30,31), did not alter the endogenous COS cell PDE activity. As a routine, however, we subtracted the residual endogenous COS cell PDE activities done in parallel experiments from those activities found in the PDE-46-transfected cells. Protein was routinely measured by the method of Bradford (38) using bovine serum albumin as a standard.

RESULTS AND DISCUSSION
Human PDE4A Species-Three cloned, active human PDE4 enzymes have been reported (16,17,32). Of these, PDE-46, which was cloned by one of us (16), is the largest species ( Fig.  1) and is believed to provide a full-length product of the PDE4A gene, being analogous to the rat PDE4A splice variant RPDE-6 (3,16,30,31). The activity of a truncated form of PDE-46 has been shown to express cyclic AMP specific PDE activity when expressed in Saccharomyces cerevisiae (16). In order to determine the activity and properties of the full-length product of the PDE4A gene when expressed in a mammalian system we engineered the cDNA encoding the entire open reading frame of PDE-46 into the pSV-SPORT vector for expression in COS cells.
h6.1 is a PDE4A species reported on by some of us previously (32). This was generated from two overlapping cDNA clones whose sequence encompassed a region which is nearly identical to that found within PDE-46 (16). This region is from the amino acid residue at position 210 of PDE-46 to the extreme C terminus of PDE-46 ( Fig. 1). It thus encompasses the entire putative catalytic region (16) together with the C-terminal region of PDE-46 but lacks the N-terminal region of PDE-46, which, in the rat PDE4A gene (Fig. 1), has been shown to be alternatively spliced (3). h-PDE1 was the first human PDE4A species to be reported (17), and this contains the region encompassed by the two clones we used to generate h6.1 (32) together with 9 additional amino acids at its N terminus, which were preceded by an initiator methionine (Fig. 1). To generate a PDE that would be expressed, we employed (32) a PCR-based strategy in order to add an initiator methionine together with the N-terminal 9 additional amino acid residues reported in h-PDE1 (17). This generated the reported clone h6.1 which thus bears close similarity to h-PDE1 (17). However, h-PDE1 (17) differs from h6.1 (32) in having certain base changes, which would lead to five differences in amino acids found in and around the putative catalytic region of the protein and which have been suggested (32,39) might account for the differences in rolipram inhibition kinetics exhibited by these two forms. Excluding the engineered region, the nucleotide sequence of h6.1 (32) exactly matches that which can be found in the human PDE4A genomic sequence 2 and, with one base difference, is identical to that reported for the cognate region of PDE46 (16). It is thus possible that the bases encoding the N-terminal 9 residues of h-PDE1 (17) and those that imply other differences may have resulted from cloning/sequencing artifacts.
A complex series of splice variants appear to be produced from the rat PDE4A gene (Fig. 1) and these all take the form of N-terminal domain swapping (3,31). The situation for the human PDE4A gene is less well developed although, in addition to PDE46 (16), a splice variant, called 2EL, has been identified ( Fig. 1), which exhibits no apparent PDE activity (23). PDE-46 (16) provides a distinct PDE4A splice variant which is expressed natively in cells. Thus transcripts have been identified in a variety of tissues by RNase protection (16) and in MonoMac 6 cells by Northern blotting (40). Indeed, immunoblotting (40) of MonoMac 6 cells for PDE4A forms identified a single immunoreactive species of ϳ130 kDa, which bears comparison with that found in COS cells transfected to express PDE-46 (see below). In contrast, to our knowledge, h-PDE1 (17) has not been shown to be expressed natively. Nevertheless, at a minimum, h-PDE1 (17) and h6.1, the species we generated (32,39) to mimic it, provide useful representations of what we might presume, from analogy with the rat PDE4A splice variants and the species met 26 RD1 (27) (Fig. 1), as the "core" human PDE4A gene product (16,27,30,31,41). Indeed, in this regard h6.1 (32), as with the rat PDE4A species, which has the engineered deletion of the N-terminal splice regions, met 26 RD1 (27,28,30), is found as a soluble, cytosolic species when transiently expressed in COS-1 cells.
Expression of PDE-46 Activity in COS-7 Cells-Transfection of COS-7 Cells with the plasmid pSV-SPORT-pde46 increased the cAMP phosphodiesterase activity of the homogenate such that Ͻ5% of the PDE activity was attributable to endogenous enzymes. Typically, PDE activities were in the range 8 -12 nmol/min/mg of protein using 1 M cAMP as substrate (n ϭ 5 experiments using different transfections). Mock transfection with the parent plasmid (pSV-SPORT) had no effect (Ͻ3%) on either the total endogenous COS7 cell PDE activity or the fraction of that which was inhibited by rolipram (10 M). The increase in PDE activity seen in pSV-SPORT-pde46-transfected cells was unaffected by the addition of either Ca 2ϩ / calmodulin (100 M; 20 ng ml Ϫ1 ; Ͻ5% change), which would stimulate any PDE1 activity, or by 1 M cGMP (Ͻ5% change), which would alter either PDE2 or PDE3 activities (1,5). This increase in homogenate PDE activity was, however, severely attenuated (Ͼ92%) by the addition of rolipram (10 M rolipram at 1 M cAMP), which serves as a selective inhibitor of PDE4 enzymes.
The increase in PDE activity subsequent to transfection with a plasmid encoding PDE-46 was distributed between both particulate and cytosol compartments, with the majority of the activity being in the cytosolic fraction. Thus we noted that some 88 Ϯ 3% (errors are S.D.; n ϭ 4) of the increase in PDE activity was associated with cytosol and the rest with the particulate/ membrane fraction. Activities in both of these fractions exhibited similar low K m values for cAMP hydrolysis (Table I). These are similar to the values reported for other PDE4 enzymes (3, 6, 7, 13, 14, 16 -19, 31, 32, 39, 42, 43).
Immunological Detection of PDE-46 -We have previously been able to generate specific antisera to rat PDE4A and PDE4B species using dodecapeptides representing the extreme C-terminal sequence of these enzymes (28,33). We were, however, singularly unsuccessful here in trying to generate such antisera useful in either Western blotting or immunoprecipitation. This was done employing both a peptide whose sequence reflected that found at the C-terminal end of both PDE-46 and h6.1 and one which represented an internal sequence of PDE-46 (data not shown; see "Experimental Procedures"). Thus, in order to try and obviate this problem we generated a GST-fusion protein so as to be able to use a larger fragment of PDE-46/h6.1 as an immunogen. This involved using PCR to generate a DNA fragment encoding amino acids at the extreme C terminus (788 -886) of PDE-46 and fusing it, in frame, to GST. The production of this fusion protein was determined by SDS-PAGE (Fig. 2). Using this species as an immunogen we raised antisera which were able to detect both GST and the pde46-GST fusion protein (Fig. 2). The antibodies directed at GST itself could be removed by treatment of the antisera with immobilized GST, leaving a treated antiserum which now recognized only the fusion protein (Fig. 2). This showed that we had been able to generate antibodies which recognized epitopes within the PDE-46/h6.1 region of the fusion protein. Such a treated antiserum was used in subsequent experiments, although similar results were obtained in Western blotting studies using the untreated antiserum to which excess GST was added before use in immunoblotting studies (data not shown). Indeed, as a routine we added excess GST to all immunoblotting studies even when the purified antiserum was employed.
Antisera, raised against the C-terminal human PDE4A-GST fusion protein, allowed us to detect (Fig. 2) a single immuno- (1) 0.54 Ϯ 0.08 (7) 11.5 Ϯ 3.5 (9) IC 50  When these species were run together on the same gel, they co-migrated with each other and with the single species found using a homogenate extract from pSV-SPORT-pde46-trans-fected COS cells (data not shown). Preincubation of these antisera with the C-terminal PDE4A-GST fusion protein prevented the detection of such an immunoreactive species (data not shown). The size of the immunoreactive species, seen in pSV-SPORT-pde46-transfected cells, was greater than that predicted (99.2 kDa) for PDE-46 on the basis of its primary sequence (16). However, our studies done on the three established rat PDE4A splice variants RD1 (30), RPDE-6 (30), and RPDE-39 (31) showed that all these enzymes exhibited slower migration (i.e. larger apparent size) on SDS-PAGE than might be predicted simply from their sequence. This may be due to folding or to stretches of acidic amino acids reducing the amount of SDS bound to the protein. Furthermore, the sequence of the human PDE-46 shows considerable similarity to that of the rat PDE4A splice variant RPDE-6, a species which when expressed in COS cells migrates on SDS-PAGE with an apparent molecular size of ϳ109 kDa compared to a calculated size of ϳ94 kDa (3,30). As with the increase in PDE4 activity observed in pSV-SPORT-pde46 transfected COS cells, an immunoreactive species was evident in both membrane and cytosol fractions (Fig. 2). Both PDE-46 activity and immunoreactivity from membranes of pSV-SPORT-pde46-transfected COS cell was not released (Ͻ5%) upon treatment with either NaCl concentrations up to 1.5 M or with concentrations of the detergent Triton X-100 up to 5% (data not shown). This bears analogy with the rat PDE-6 and rat PDE-39 PDE4A splice variants (30, 31), but not with the rat RD1 splice variant which, although being exclusively membrane-associated, was readily solubilized by the detergent Triton X-100 (27,30). Such treatments of these particulate fractions did not affect PDE activity or its susceptibility to inhibition by rolipram (data not shown).
The availability of antisera able to detect PDE-46 allowed us to determine the relative amounts of the enzyme in the isolated particulate and cytosol fractions of transfected COS7 cells. From this we were able to determine the relative V max values for these two populations of PDE-46 (Table I). Such analyses showed that the V max of the particulate form of PDE-46 was 56 Ϯ 9% of that of the cytosolic form (errors are S.D.; n ϭ 4 separate experiments).
Expression of h6.1 in COS7 Cells-Previously we have expressed h6.1 in COS1 cells (32) and shown the increases rolipram-inhibited PDE4 activity to be found in the cytosol fraction but not associated with the high speed (P2) membrane fraction. Similarly, here we observed that transfection of COS7 cells with pSV.SPORT-h6.1 led to increased rolipram-inhibited PDE4 activity located in the high speed supernatant, cytosol fraction with no observable increase in the PDE activity of the membrane fraction (Ͻ5% change; n ϭ 4). Cytosolic expressed h6.1 PDE exhibited a specific activity of 1.5 Ϯ 0.4 nmol cAMP hydrolyzed/min/mg of protein (n ϭ 4). All of this increased PDE activity (Ͼ96%) could be inhibited by rolipram (10 M rolipram with 1 M cAMP as substrate). Consistent with such activity studies, immunoblot analysis also showed that, upon transfection with pSV-SPORT-h6.1, then a 99 Ϯ 3 kDa (S.D.; n ϭ 4) immunoreactive species was evident in the cytosol fraction but not in the membrane pellet (Fig. 1). As with PDE-46, h6.1 expressed in COS7 cells exhibited an apparent molecular size, upon SDS-PAGE, which was greater than the value of 76.4 kDa deduced from its sequence (32).
That h6.1 was found as a soluble, cytosolic species when expressed in COS cells, whereas PDE-46 was found in both particulate-associated and soluble forms, indicates that the N-terminal extension of PDE-46 contains information that allows anchoring to particulate/membrane fractions. This situation bears analogy to the protein products of the rat PDE4A

FIG. 2. Generation of antisera to human PDE4A using a fusion protein, formed between the C-terminal region of h6.1 and GST, as the immunogen. In A is shown, in the first panel, a Coomassie
Blue-stained SDS-PAGE analysis of bacterial extracts from cells expressing either glutathione S-transferase (track G) or an N-terminal fusion protein (track F) formed from GST with amino acids 788 -886 of PDE-46. The molecular mass of the fusion protein was determined as 38.5 Ϯ 1.5 kDa experimentally, which agrees well with the size of 37.7 kDa predicted from its primary sequence. GST migrated as a 28.3 Ϯ 1.8-kDa species, compared to a predicted size of 27.5 kDa. 10 g of protein were used in each instance. The second panel shows immunoblots done using antiserum raised against the pde46-GST fusion protein. Lanes were of samples of either GST (G) or pde46-GST fusion protein (F) analyzed using unpurified/native antiserum or antiserum that had been purified using a column of immobilized GST in order to remove selectively antibodies directed against GST. 1 g of protein was used in each instance. The third panel shows immunoblots done on COS cell extracts of the soluble, high speed supernatant fraction (cytosol), and the fourth panel shows immunoblots done on a high speed membrane hP2 pellet (particulate fraction). COS7 cells were transfected with either vector alone (v) or with vectors allowing the expression of either h6.1 or PDE-46 as indicated. 10 g of protein were used in each instance. The experiments shown are typical of those done at least three times. In B is shown the result of a typical experiment of one done over three time of gray scale intensity (arbitray units) from PhosphorImager analysis of immunoblots done using PDE-46 (q) and h6.1 (f) with detection using 125 I-labeled second antibody. This shows that, over the range of applied protein, a linear relationship between intensity and applied protein was observed. Such analyses were used to determine the relative amounts of PDE-46 and h6.1 in fractions from transfected COS cells for the determination of relative V max values in Table I. gene. The PDE4A gene thus appears to encode a "core," highly active, soluble protein to which various N-terminal extensions can be spliced. These N-terminal regions allow interaction with membrane/particulate fractions, generating species which are either totally membrane-associated or distributed between cytosol and membrane compartments (27,28,30,31). In this regard, in transfected COS-7 cells ϳ25% of the cognate species to PDE-46, namely RPDE-6, was found to be associated with the membrane/particulate fraction (30). Thus the N-terminal regions of both the human and rat PDE4A enzymes appear to play a targeting role.
The availability of antisera able to detect both PDE-46 and h6.1 has allowed us to determine their relative concentrations in extracts from transfected COS cells. This was done using 125 I-labeled second antisera (Fig. 2B). From this we were able to determine the relative V max values for h6.1 compared to cytosolic PDE-46, which expresses a large N-terminal splice domain (Table I). This showed that h6.1 had a considerably higher (ϳ11.5-fold) maximal activity compared to that of PDE-46.
Laser Scanning Confocal Microscopy Analyses of Transfected COS Cells-COS cells were transfected so as to express either h6.1 or PDE-46, fixed, permeabilized, and challenged with specific antiserum and rhodamine-labeled anti-rabbit antiserum before analysis of their immunofluorescence using laser scanning confocal microscopy as described under "Experimental Procedures." No fluorescent signal was evident in cells which had either not been permeabilized, indicating that the PDE signal was indeed intracellular, or where no primary antibody had been added or if an excess of blocking GST-PDE4A fusion protein had been added (see "Experimental Procedures"; data not shown). On the basis of immunoblotting and activity studies, levels of any endogenous PDE4A would be extremely low (Ͻ1%) compared to the levels of PDE-46 and h6.1 seen in transfected COS cells. Consistent with this, transfected cells were clearly distinguishable as highly fluorescent species set among the presumed nontransfected cells, which showed an extremely low and poorly resolved signal (data not shown).
In COS cells transfected so as to express h6.1 we observed a clear fluorescent signal which extended throughout the cytosol of these cells. This was evident from optical sections viewed in the x-y plane cutting across the middle of the COS cells (Fig. 3). Obtaining multiple images in the z plane allowed for the reconstruction of cellular fluorescence in three dimensions and, from this, z-x and z-y slices down through the cell could be generated (Fig. 3). These showed that the fluorescent signal due to h6.1 permeated through the entire cytosol of the transfected cells but was clearly excluded from the nucleus (Fig. 3). However, the intensity of immunofluorescence due to h6.1 appeared to be somewhat asymmetrically distributed through the cytosol, with increased intensity levels seen in the cytosol region that was closest to the nucleus (Fig. 3). In this regard, such a region around the nucleus was shown to be highly enriched in bundles of microtubules (Fig. 3). Interestingly, the pattern of fluorescence observed with h6.1 differed from that seen (28) for the engineered (27) core-soluble rat PDE4A form, met 26 RD1 which was distributed evenly throughout the cytosol of transfected COS cells. This might indicate that h6.1 can associate with localized structures inside the cell but in a reversible fashion, which is disrupted upon cell breakage to release soluble h6.1.
A strikingly different pattern of immunofluorescence was obtained using COS cells that had been transfected so as to express PDE-46. In x-y sections taken horizontally through the middle of the cells then concentrated areas of fluorescence were found at the cell periphery (Fig. 4). This was supported from both z-x and z-y slices down through the cell which, again, highlighted a concentration of fluorescence at the cell margin as well as an evident fluorescent signal permeating through the cell cytosol (Fig. 4). That the highest immunofluorescence was observed at the cell margin indicates that the relative concentration of PDE-46 at such a location must be considerably higher than the concentration of PDE-46 found within the cytosol. However, from biochemical analyses done on disrupted cells, by far the greatest amount of PDE-46 was found as a soluble species. One explanation may be that particulate-associated PDE-46 achieves a much higher local concentration than that of the pool of soluble PDE-46, which is distributed throughout the cytosol. The fluorescence at the cell periphery was, however, not evenly distributed. Rather, it appeared to occur at distinct margins of the cell that are suggestive of association with cortical structures, including lamellae and pseudopods (44,45). In this regard, such immunofluorescence patterns are similar to those described for fodrin (nonerythroid spectrin) and the actin binding protein, cortactin (44,45). In -FIG. 3. Immunofluorescent laser scanning confocal microscopy studies of h6.1 transfected COS cells. These were performed as described under "Experimental Procedures." COS7 cells are typically from 4.5-6.5-m thick from top to bottom. In panel A, a series of optical sections were taken every 0.2 m and used in a three-dimensional reconstruction with the use of Imaris software. This allowed z-x and z-y vertical sections to be generated. These sections are shown at the top and right-hand side, respectively, of panel A. In the main panel is an x-y horizontal section done through the center of the cell with the transverse white lines, indicating the points through which the z-x and z-y sections were made. Transfection of COS cells often leads to multinucleate structures as indicated. White indicates the highest level of fluorescence. Note from these data that fluorescence is highest in the area of the cytosol which surrounds the nucleus but, as is evident from the z-x and z-y sections, extends out through the cytoplasm. Panel B shows an x-y transverse section through the center of another COS cell transfected so as to express h6.1. Again, fluorescence (white) extends throughout the cytosol but with an increased intensity in the area surrounding the nucleus. In panel C is shown a x-y section through the middle of COS cells that had been labeled with anti-tubulin antibody. This shows microtubules extending throughout the cell from the nucleus, which is evident as clear zone in the center of the cell. Note that there is a ring of microtubule bundles around the nucleus. The horizontal white bar shown in various panels is the 25-m scale marker. deed, if PDE-46 was to interact with cytoskeletal or cytoskeletal-associated proteins this might explain why particulate PDE-46 was not solubilized by detergent or high salt treatment. This distribution of fluorescence due to PDE-46 was very different from that seen from the rat PDE4A RD1 when expressed in COS cells (28). RD1 has a very different N-terminal region to PDE-46 and is found as an exclusively membraneassociated species (30) in transfected COS cells, where it appears to be associated with both the Golgi apparatus, and the cytosol surface of the plasma membrane, where a distinct punc-tate pattern of fluorescence was observed (28). This suggests that alternative splicing may provide N-terminal regions that allow for distinct targeting of PDE4 enzymes within the cell.
Inhibition of Particulate and Cytosol Forms of PDE-46 by Rolipram-Rolipram caused the dose-dependent inhibition of both particulate and soluble PDE-46 forms expressed in pSV-SPORT-pde46-transfected COS-7 cells (Fig. 5). However, distinct differences in these dose-effect curves were evident in that the particulate activity appeared to be more sensitive to inhibition by rolipram, exhibiting an IC 50 value of 0.195 Ϯ 0.035 M rolipram, compared with the cytosolic form which exhibited an IC 50 value of 1.6 Ϯ 0.3 M rolipram (n ϭ 4 separate experiments; errors as S.D.). In view of this we undertook a detailed study of the inhibition of the particulate and cytosolic forms of PDE-46 by rolipram. From such studies we were able to determine that rolipram served as a simple competitive inhibitor of the cytosolic enzyme (Fig. 6). Thus, double reciprocal activity plots, performed at different rolipram concentrations, showed a common intersection on the y axis (Fig. 6) and produced linear slope replots against rolipram concentration (Fig. 6) and linear Dixon plots (data not shown) (46). From these we were able to derive K i values of ϳ1.6 M for the association of rolipram with cytosolic PDE-46 (Table I). Such a value is slightly greater than that recorded for h6.1 where, similarly, rolipram acts as a simple competitive inhibitor (32,39).
Double reciprocal plots of particulate PDE activity, done at different rolipram concentrations, exhibited a common intersection on the y axis (Fig. 7A). However, in marked contrast to the action of rolipram on the soluble enzyme, in this instance, for the particulate enzyme, we noted that both the slope replots (Fig. 7A) and the Dixon plots (46) were nonlinear (Fig. 7B). The parabolic nature of these two plots is consistent with rolipram serving as a partial competitive inhibitor of the particulate enzyme (47) (see Scheme 1). In this situation inhibitor can bind not only to free enzyme (E), as with a simple competitive inhibitor, but it can also bind to the enzyme-substrate complex (ES). For apparent competitive inhibition to be realized, however, the EIS complex has to hydrolyze cAMP at a rate which is identical (experimentally indistinguishable) from that at which the ES complex hydrolyzes cAMP, i.e. the rate constants k 3 ϭ k 4 . This is because a partial competitive inhibitor does not change the V max of the reaction (47). Partial competitive inhibitors thus yield double reciprocal plots of initial rates of substrate utilization against substrate concentration done at different inhibitor concentrations, which are identical to those seen for full competitive inhibitors: that is, they are linear and show no effect on V max but an increase in apparent K m . This is self-evident from the rate equation (Equation 1) for partial competitive inhibition which is given by where v obs is the initial rate, V max is the maximum velocity of the reaction, K s is the affinity of the enzyme for substrate (cAMP), K i is the affinity constant which reflects binding of inhibitor (rolipram) to the free enzyme and K i Ј is the affinity constant which reflects binding of inhibitor (rolipram) to the enzyme-substrate complex. As can be seen from Equation 2, a prediction of this form of mechanism (47) is that the slope replots from double reciprocal plots of the initial rates of PDE activity against [cAMP] done at different rolipram concentrations will be nonlinear.
Similarly, as can be deduced from Equation 3, the Dixon (46) replots of the reciprocal of the initial rates of PDE activity against rolipram concentration will be nonlinear.
Such graphical analyses of the kinetic data for the particulate enzyme (Fig. 7) show quite clearly that these predictions hold true with both such plots being markedly nonlinear. A further prediction of such a kinetic mechanism is that double reciprocal plots of the change in slope (⌬ slope of the double reciprocal plot data) against inhibitor (rolipram) concentration should be linear, as implied by Equation 4.
Graphical analysis shows that this, indeed, appears to be the case for the particulate enzyme (Fig. 7C). Indeed, from such a plot the affinity constant (K i Ј) for rolipram binding to the enzyme-substrate complex (ES) can be determined from the intercept on the x axis ( Fig. 7; Table I). Using these data it is then possible to employ curve fitting routines applied to the slope replot data (Fig. 7) in order to determine values for the association of rolipram to the free particulate enzyme (K i ) and also the constant for substrate (cAMP) association with enzyme (K s ) ( Table I). Thus the inhibitor affinity constant for the free particulate enzyme (K i ) was seen to be (Table I) considerably lower than that for the substrate-bound particulate enzyme (K i Ј). Furthermore, the K i value for the particulate form of PDE-46 was considerably lower than that observed for the cytosolic form (Table I), indicating that rolipram bound more tightly to particulate PDE-46 than it did to the cytosolic form of this enzyme. This explains why, in dose-effect studies, that the particulate enzyme began to be inhibited at much lower concentrations of rolipram than those which exerted actions on the cytosolic enzyme (Fig. 5). Indeed, using the kinetic constants derived for the cytosolic and particulate forms of PDE-46 (Table I), initial rate data for the dose-dependent inhibition of these enzymes at 1 M cyclic AMP substrate concentration can be calculated using Equation 1 in order to derive IC 50 values for rolipram inhibition. Doing this we derived IC 50 values of 0.125 and 2.2 M for the particulate and cytosolic forms of PDE-46, respectively. These values compared well with the experimentally determined IC 50 values of 0.195 and 1.6 M for the particulate and cytosolic forms of PDE-46, respectively (n ϭ 4; see above).
For cytosolic PDE-46, which obeys simple competitive kinetics, then saturating concentrations of rolipram (Ն10 M) will completely (Ͼ99%) inhibit PDE activity, and such an effect was observed experimentally (Ͼ98% inhibition). However, for particulate PDE-46, which obeys kinetics of partial inhibition, then the inhibitor-bound enzyme is able to hydrolyze cAMP, and thus rolipram will be unable to obliterate the entire particulate PDE-46 activity. From the experimentally determined kinetic constants (Table I), it is possible to calculate, using Equation 1, the magnitude of the residual activity PDE expected in the presence of 10 M rolipram while using 1 M cAMP as substrate. This activity would be expected to form ϳ6% of that observed in the absence of added rolipram. Such a value is consistent with experimental data showing that 10 M rolipram can inhibit PDE activity assayed with 1 M cAMP by ϳ92% (Fig. 5). Such an altered kinetic mechanism may result as a consequence of PDE-46 becoming associated with membrane/particulate fractions. Indeed, there is a precedent for this where the solubilization of monoamine oxidase led to an alteration in kinetic mechanism (48).
It is intriguing, however, that the particulate form of PDE-46 showed a markedly higher affinity for rolipram (ϳ62-fold) than the soluble enzyme, as indicated from the differences in their K i values (Table I). In this regard, various investigators have reported on the existence of high affinity binding sites for rolipram with association constants between one and two orders of magnitude lower than the IC 50 values reported for the rolipram inhibition of soluble PDE4 preparations (49 -52). One possible contributor to this is likely to be PDE4D3, which shows a markedly enhanced susceptibility to inhibition by rolipram when it is phosphorylated on a unique site in its Nterminal splice region by protein kinase A (20). However, it is possible that particulate-associated PDE-46 may provide a contribution to the population of high affinity rolipram binding sites by virtue of the alteration in the kinetics of rolipram inhibition of this PDE4A splice variant. It is, however, important to appreciate that because of the partial competitive nature of rolipram inhibition of particulate PDE4A, the magnitude of the affinity of this fraction of the enzyme for rolipram will be severely underestimated in studies done trying to make inferences about this by determining IC 50 values at a single concentration of cAMP. This is because such a value will reflect In C is shown double reciprocal plots of the change in slope (delta slope) from the inset of A against the inhibitor (rolipram) concentration. A linear regression line could be fitted to these data with r 2 Ͼ 0.995, and this was used to determine the value of K i Ј as per Equation 4 (47). PDE activity is expressed in nanomoles/min/mg of protein. These are typical sets of data from experiments done four times using different transfections. not only the high affinity binding of rolipram to the free enzyme but also the lower affinity binding of rolipram to the cAMPbound enzyme.
In studies done using h-PDE1, expressed in S. cerevisiae and in baculovirus-infected insect cells, anomalous rolipram binding kinetics were observed together with high affinity rolipram binding (53,54). Intriguingly, however, using h6.1 expressed in S. cerevisiae, rigorous kinetic analyses served only to show that simple competitive kinetics of inhibition by rolipram ensued (39). It has been suggested (39) that the differences in inhibition between these two forms might reflect differences in the sequence of these forms at residues in and around the catalytic region. Indeed, it has been shown (55) that single amino acid changes in a PDE4 enzyme can lead to profound changes in sensitivity to rolipram inhibition. As the sequence of h6.1, other than the 9 residues at its N terminus, reflects that of PDE-46, it is possible that residue changes seen in h-PDE1 might have elicited a conformation change which mimics that adopted by the particulate form of PDE-46. It is, however, possible that, in contrast to h6.1, when h-PDE1 was expressed in S. cerevisiae it became modified in a manner which triggered a similar change in conformation to that exhibited by the particulate form of PDE-46, resulting in altered kinetics of rolipram inhibition.
The molecular cloning of the dunc PDE from Drosophila has served as a paradigm for the PDE4 PDE family. However, an interesting anomaly is that this enzyme was apparently not inhibited by rolipram (14,16,31,56,57). This may be due to structural differences in the binding site and, indeed, in this regard it has been demonstrated (55) that single amino acid changes in PDE4 can dramatically affect the ability of rolipram to serve as an inhibitor. However, it is also possible that rolipram might in fact bind to the Drosophila dunc enzyme without inhibiting it if the dunc enzyme obeyed kinetics of partial competitive inhibition. For, under conditions where the affinity of both free enzyme and the enzyme-substrate complex for rolipram was identical (K i ϭ K i Ј) then, as was deduced by Dixon (47) from Equation 1, no apparent inhibition of enzyme activity would result. CONCLUSIONS Transfection of COS7 cells with a plasmid encoding the PDE4A splice variant PDE46 demonstrates, as seen for the rat homologue RPDE-6 (30), both cytosolic and particulate forms. Here we show that particulate PDE-46 is highly localized to cortical areas of the cell. Intriguingly, these two populations of PDE-46 demonstrate very different susceptibilities to inhibition by rolipram. This appears to reflect changes in their kinetic mechanisms of inhibition and the conformation of the active site. In this regard the maximum activity of the particulate form is only about half that of the cytosol form. We suggest that PDE-46 can exist in two conformationally distinct states which here are reflected in the cytosolic and particulate populations. The PDE4 selective inhibitor rolipram thus appears to serve as an effective detector of these two states, which can be characterized by changes in both the kinetics of inhibition and the affinity of interaction with rolipram. While the functional significance of these two states and whether PDE46 can dynamically switch between them remains to be determined, their existence has implications for the design and application of inhibitors for therapeutic use. The biological significance of the targeting of PDE-46 remains to be elucidated. It may relate to the anchoring of protein kinase II isoforms (58,59), so as to effect the compartmentalization of cAMP signaling within the cell, or it could be so as to confer specific regulatory properties upon the immobilized PDE fraction. Thus modification of the N-terminal region of PDE-46 and alterations in its interactions with anchoring/binding proteins may well have profound regulatory consequences.