Dimerization of the Type 4 cAMP-specific Phosphodiesterases Is Mediated by the Upstream Conserved Regions (UCRs)*

The cAMP-specific PDE4 family consists of four genes, each expressed as several splice variants. These variants are termed long and short forms depending on the presence or absence of two unique N-terminal domains called upstream conserved regions 1 and 2 (UCR1 and 2). UCR1 and UCR2 have been shown to form a module necessary for the activation of PDE4 upon phosphorylation by the cAMP-dependent kinase (PKA). Here we have uncovered PDE4 oligomerization as a novel function for the UCR1/UCR2 module. Using several different approaches including gel filtration, sucrose density gradient centrifugation, pull-down of differentially tagged PDE constructs, and yeast two-hybrid assay, we show that the long PDE4 splice variant PDE4D3 behaves as a dimer, whereas the short splice variant PDE4D2 is a monomer. Internal deletions of either the C-terminal portion of UCR1 or the N-terminal portion of UCR2 abolishes dimerization of PDE4D3 indicating that both domains are involved in this intermolecular interaction. The dimerization, however, is structurally distinguishable from a previously described intramolecular interaction involving the same domains. PKA phosphorylation and site-directed mutagenesis shown to ablate the latter do not interfere with dimerization. Therefore, dimerization of the long PDE4 forms may be an additional function of the UCR domains that further explains differences in the regulatory properties between the long and short PDE4 splice variants.

The cAMP-specific PDE4 family consists of four genes, each expressed as several splice variants. These variants are termed long and short forms depending on the presence or absence of two unique N-terminal domains called upstream conserved regions 1 and 2 (UCR1 and 2). UCR1 and UCR2 have been shown to form a module necessary for the activation of PDE4 upon phosphorylation by the cAMP-dependent kinase (PKA). Here we have uncovered PDE4 oligomerization as a novel function for the UCR1/UCR2 module. Using several different approaches including gel filtration, sucrose density gradient centrifugation, pull-down of differentially tagged PDE constructs, and yeast two-hybrid assay, we show that the long PDE4 splice variant PDE4D3 behaves as a dimer, whereas the short splice variant PDE4D2 is a monomer. Internal deletions of either the C-terminal portion of UCR1 or the N-terminal portion of UCR2 abolishes dimerization of PDE4D3 indicating that both domains are involved in this intermolecular interaction. The dimerization, however, is structurally distinguishable from a previously described intramolecular interaction involving the same domains. PKA phosphorylation and site-directed mutagenesis shown to ablate the latter do not interfere with dimerization. Therefore, dimerization of the long PDE4 forms may be an additional function of the UCR domains that further explains differences in the regulatory properties between the long and short PDE4 splice variants.
The second messengers cAMP and cGMP are key molecules for transducing the action of extracellular signals such as hormones, neurotransmitters, and light into the most diverse cell functions, thus playing an important role in a wide array of physiological processes that include vision, cell growth and division, memory, and immune response (1)(2)(3)(4)(5). Intracellular cyclic nucleotide levels are determined by the rates of their synthesis by adenylyl cyclases and of their degradation through phosphodiesterases, enzymes that hydrolyze the phosphodiester bond and generate the corresponding 5Ј-nucleoside monophosphates.
Cyclic nucleotide phosphodiesterases (PDEs) 1 compose a su-perfamily of isoenzymes that are divided into 11 PDE families on the basis of their sequence homology and enzymatic properties (2). They all share a highly conserved catalytic domain of about 270 amino acids fused to additional N-and/or C-terminal sequences that contain distinct domains unique to the members of a PDE family. These terminal domains determine several properties of the PDEs such as the regulation of enzyme activity by post-translational modifications (e.g. phosphorylation by PKA, PKB, ERK2, CaMK, and PKG, Refs. 6 -10) and binding of other messenger molecules (e.g. cGMP, Ca 2ϩ -calmodulin, and phosphatidic acid, Refs. [11][12][13], or by specifying the subcellular localization of the enzymes by protein-proteininteractions and membrane insertion (14 -16). The terminal domains of the PDEs, therefore, provide diverse modules for coordinating PDE activity with the overall signaling network specific to a cell. A structural feature unique to the cAMP-specific PDE4 family is the presence of two domains N-terminal to the catalytic domain called upstream conserved regions 1 and 2 (UCR1 and UCR2). The four genes included in the PDE4 family are each expressed as various splice variants that are distinguished as long and short forms depending on the presence of the UCRs (17)(18)(19)(20). Long forms contain both UCR1 and UCR2, whereas short forms lack UCR1 but still possess all, or at least a portion, of UCR2. In the long PDE4 splice variants where both domains are present, UCR1 and UCR2 have been shown to interact with each other, and this interaction was thought to be intramolecular (21,22). Site-directed mutagenesis further indicated that this interaction may involve positively charged residues in UCR1 (Arg-98 and Arg-101 in PDE4D3) and several negatively charged residues in UCR2 (Glu-146, Glu-147, and Asp-149 in PDE4D3, Ref. 22). In these studies, it was demonstrated that the interaction between the two domains constitutes an important module involved in the regulation of the enzyme activity in several different ways. Phosphorylation of a serine residue at the N terminus of UCR1 (Ser-54 in PDE4D3) by the cAMP-dependent protein kinase (PKA, Ref. 6), which causes enzyme activation, alters the interaction between UCR1 and UCR2. As UCR2 may function as an autoinhibitory domain of the catalytic center, a probable mechanism for PDE4 activation is that the PKA phosphorylation leads to an altered UCR1-UCR2 interaction that removes the autoinhibitory effect of UCR2 on the catalytic site (6,21,22). In addition, phosphorylation by the extracellular signal-regulated kinase 2 (ERK2) at the PDE4 C terminus affects the PDE4 enzyme activity in different ways that depend on the presence of the UCRs. Long PDE4 isoforms containing both a complete UCR1 and UCR2 were inhibited by ERK2 phosphorylation whereas the short splice form PDE4D1, which lacks UCR1, was activated (8,23) thus underlining the role of the UCR1/2 module for the regulation of enzyme activity.
In our previous report, the possibility was raised that disruption of oligomerization may explain some of the properties associated with UCR1/UCR2 (21). Here, we have further investigated the quaternary structure organization of PDE4 splice variants. We demonstrate that long PDE4 splice variants containing both UCR1 and UCR2 form dimers whereas short forms lacking UCR1 are monomers. The domain responsible for enzyme dimerization was mapped to the region between the C terminus of UCR1 (UCR1C) and the N terminus of UCR2 (UCR2N), and engineered deletion of either subdomain ablates dimerization. The dimerization domain involves, therefore, the same domains that were previously shown to interact with each other in an intramolecular interaction (21,22).
Immunoprecipitation-Protein G-Sepharose was washed three times with phosphate-buffered saline and then loaded with the antibody in a reaction mixture containing 20 l of protein G Sepharose beads, 100 l of phosphate-buffered saline, and the respective affinity-purified antibody and was incubated in a rotating mixer for 60 min at 4°C. The unbound antibody was removed by two washes with phosphate-buffered saline, and the immobilized antibody was then incubated with the enzyme extracts for 2 h at 4°C in a rotating mixer. The immunoprecipitate was then separated from the reaction mixture by centrifugation for 2 min at 1,000 ϫ g and was washed three times by resuspension and recentrifugation in the same buffer that was used for the preparation of the cell extracts. After the last centrifugation, the immunoprecipitate was eluted from the Sepharose beads by boiling the samples for 5 min in 1ϫ Laemmli buffer.
SDS/PAGE and Western Blot Analysis-SDS/PAGE was performed by the method of Laemmli (25) using 8 or 10% (w/v) polyacrylamide gels. Protein samples were diluted with 4ϫ concentrated Laemmli buffer, boiled for 5 min, subjected to electrophoresis, and blotted onto Immobilon membranes. Western blot analysis was then performed using the following antibodies: 1) the monoclonal ␣-V5 antibody (Invitro-gen) that recognizes the artificial V5-tag; 2) the monoclonal ␣-Myc antibody that recognizes the artificial Myc-tag; 3) the monoclonal antibody M3S1 raised against the C-terminal region of PDE4D; and 4) the polyclonal antibody K116 that recognizes the N-terminal domain of all PDE4 subtypes. Immunoreactive bands were detected by using either peroxidase-conjugated goat anti-mouse/anti-rabbit antibodies and the ECL detection reagents (Amersham Biosciences) or the alkaline phosphatase-conjugated goat anti-mouse/anti-rabbit antibodies and the corresponding detection system (BioRad, Hercules, CA).
PDE Assay-PDE activity was measured according to the method of Thompson and Appleman (26). In brief, samples were assayed in a reaction mixture of 200 l containing 40 mM Tris-HCl (pH 8.0), 10 mM MgCl 2 , 5 mM ␤-mercaptoethanol, 1 M cAMP, 0.75 mg/ml bovine serum albumin, and 0.1 Ci of [ 3 H]cAMP for 10 min at 33°C. The reaction was terminated by adding 200 l of 10 mM EDTA in 40 mM Tris-HCl (pH 8.0) followed by heat inactivation in a boiling water bath for 1 min. The PDE reaction product 5Ј-AMP was then hydrolyzed by incubation of the assay mixture with 50 g of Crotalus atrox snake venom (Sigma) for 20 min at 33°C, and the resulting adenosine was separated by anion exchange chromatography using 1 ml of AG1-X8 resin and quantitated by scintillation counting.
Density Gradient Centrifugation-Density gradients from 5 to 33% sucrose in 50 mM Tris-HCl (pH 8.0), 1 mM EDTA, 0.2 mM EGTA, 10 mM NaF, 150 mM NaCl, and 5 mM ␤-mercaptoethanol were prepared in 14 ϫ 89-mm centrifugation tubes (Beckman, Fullerton, CA) using the Jule gradient former (Jule Inc., New Haven, CT). The gradients were stored overnight at 4°C before they were loaded with 200 l of the soluble cell extracts (100,000 ϫ g supernatant) or marker proteins and centrifuged for 36 h at 28,000 rpm ( Fig. 3) or 24 h at 40,000 rpm ( Fig. 6B) in a Beckman SW41 rotor. Fractions of 200 l were then collected starting from the bottom of the tube using a peristaltic pump and analyzed for PDE activity or protein concentration (molecular mass markers). The following proteins were used as molecular mass markers: bovine thyroglobulin (670 kDa; 19 S), bovine liver catalase (250 kDa; 11.3 S), rabbit aldolase (158 kDa; 7.3 S), bovine serum albumin (67 kDa; 4.6 S), and chicken ovalbumin (44 kDa; 3.5 S).
Yeast Two-hybrid Analysis-The Saccharomyces cerevisae strain Y190 (Clontech, Palo Alto, CA) was co-transformed with a pGAD10 vector construct (encoding the GAL4 activation domain either alone or fused to a PDE4D fragment) and a pAS2.1 vector construct (encoding the GAL4 binding domain either alone or fused to a PDE4D fragment) using the lithium acetate method according to Clontech's yeast twohybrid protocol. The transformants were then streaked on selection medium plates (-S.D. ϭ selection medium), lacking either leucine and trytophan (-2 S.D.), or lacking leucine, tryptophan, and histidine (-3 S.D.) and incubated for 3-5 days at 30°C. For qualitative detection of ␤-galactosidase activity, the colony filter-lift assay was applied according to Clontech's manual. For the photograph shown in Fig. 5B, the yeast transformants were restreaked on one selection medium plate (-2 S.D.) and analyzed using the colony filter-lift assay.
In Vitro Phosphorylation by PKA-For in vitro phosphorylation experiments, cytosolic supernatants of recombinant PDE constructs expressed in MA10 cells were incubated for 10 min at 30°C in a reaction mixture containing 20 mM MgCl 2 , 200 M ATP, and 1.0 unit of the PKA catalytic subunit.
Calculation of Molecular Weights-The molecular weights of the PDE4D constructs were calculated using the Stokes radii and sedimentation coefficients obtained from gel filtrations and density gradient centrifugations, respectively (see Table I), using Equation 1, where S is the sedimentation coefficient in Svedberg units (see Table I); N A is the Avogadro's constant (6.022⅐10 23 mol Ϫ1 ); is the viscosity of medium (0.01 g⅐cm Ϫ1 ⅐s Ϫ1 ); R s ϭ Stokes radius in nm (see Table I); v 2 ϭ partial specific volume of a protein (0.73 cm 3 ⅐g Ϫ1 ); and is the density of medium (1 g⅐cm Ϫ3 ). The theoretical background of the methods and calculations used to determine the mass of proteins is reported by Martin and Ames (27) and Siegel and Monty (28).

Co-immunoprecipitation of Differentially Tagged PDE4D
Splice Forms-As a first approach to investigate PDE4 oligomerization, we performed pull-down experiments utilizing differentially tagged PDE4D constructs. For this purpose, two constructs C-terminally tagged with either a V5/His-or Myc/ His tag were generated for both PDE4D3 and PDE4D2 as representatives of the PDE4 long and short splice variants, respectively. These differentially tagged constructs were then overexpressed in COS7 cells either separately or together. After harvesting the cells, the V5/His-tagged construct was immunoprecipitated from the cell lysates using an ␣-V5 antibody, and the immunoprecipitated pellets were probed for the presence of the Myc/His-tagged proteins. As shown in Fig. 1A, the V5/His-tagged PDE4D3 was specifically immunoprecipitated by the ␣-V5 antibody (lanes 4 and 6) but not by the IgG control (lanes 3 and 5) whereas the PDE4D3-Myc/His construct was not recognized and immunoprecipitated by either the ␣-V5 antibody or by IgG (lanes 1 and 2). Therefore, the pull-down of the Myc/His-tagged PDE4D3, when co-transfected with the PDE4D3-V5/His construct (Fig. 1A, lane 6), is indirect and indicates an intermolecular interaction between the PDE4D3-Myc/His and the co-expressed PDE4D3-V5/His construct.
When the experiment was carried out under the same conditions using the differentially tagged constructs of the short splice form PDE4D2 (Fig. 1B), a co-IP of the Myc/His-tagged PDE construct with the co-transfected PDE4D2-V5/His (Fig.  1B, lane 6) was not observed. Moreover, the Myc-tagged PDE4D2 could not be co-immunoprecipitated when co-transfected with the long splice variant PDE4D3-V5/His (data not shown). These results provided a first indication that the PDE4D3 forms oligomers whereas the short splice form PDE4D2 does not, and that the oligomerization domain is located within the N-terminal region of PDE4D3, which is absent in PDE4D2.
In order to exclude the possibility that the co-IP of the differentially tagged PDE4D3 constructs is caused by the fusion peptides (i.e. an interaction between the N-terminal domain of PDE4D3 with either the Myc/His-or the V5/His tag), the PDE4D3 co-IP was repeated with two untagged PDE forms, PDE4D3, and PDE4D3-⌬CAT. In this PDE4D3-⌬CAT construct, the C-terminal epitope recognized by the M3S1 antibody (which was used for the pull-down of the full-length PDE4D3) has been removed. As shown in Fig. 1C, the full-length PDE4D3 could be selectively immunoprecipitated by the M3S1 antibody (lanes 2 and 6) whereas the PDE4D3-⌬CAT construct was not immunoprecipitated (lane 4). Conversely, PDE4D3-⌬CAT could be co-immunoprecipitated when co-expressed with the full-length PDE4D3 (lane 6). This finding, therefore, excludes the possibility of an artificial co-immunoprecipitation caused by the tags used and confirms that the potential intermolecular interaction between the PDE4D3 constructs is mediated by the domains present in the PDE sequence.
Determination of Apparent Molecular Weights of PDE4D3 and PDE4D2 by Size Exclusion Chromatography and Density Gradient Centrifugation-In order to determine the oligomerization of the PDE4D3 splice form by an independent approach, both native and recombinant PDE4D3 were analyzed by size exclusion chromatography. FRTL5 cells were chosen as the source of endogenous PDE, as PDE4D3 is the only PDE4D splice form in these cells and the most abundant PDE expressed. When fractionated on an analytical gel filtration column, the full-length recombinant PDE4D3 expressed in either MA-10 cells (Fig. 2B), Sf9 insect cells, or E. coli (data not shown), as well as the endogenous PDE4D3 from FRTL5 cells ( Fig. 2A) all eluted in the same fractions that correspond to an apparent molecular mass of ϳ388 kDa when compared with molecular size standards. The PDE4D3 (M theor. ϭ 76,000), therefore, behaved as an oligomer. The short splice variant PDE4D2 (M theor. ϭ 58,000) overexpressed in either MA-10 cells (Fig. 2C), insect cells, or E. coli (data not shown) eluted from the column in fractions corresponding to an apparent molecular weight of ϳ167,000. Thus, the PDE4D2 eluted at a higher apparent molecular weight than that predicted by the co-IP experiments. This behavior was dependent on the C terminus of the protein because the deletion mutants PDE4D2-⌬CAT and PDE4D3-⌬CAT, each lacking 74 amino acids at the PDE4D C terminus (see Fig. 1C for a scheme of the PDE4D3-⌬CAT), eluted in fractions corresponding to apparent molecular weights of ϳ65,000 (4D2-⌬CAT, Fig. 2C) and ϳ152,000 (4D3-⌬CAT, Fig. 2B), respectively.
The behavior of PDE4D3 and PDE4D2 in gel filtration may be explained by either an asymmetry of the PDE protein, the presence of the highly negatively charged C-terminal residues, or an additional oligomerization domain at the PDE4D C terminus. To distinguish among these possibilities, density gradient centrifugations of both proteins were performed as an independent assessment of molecular weights. As shown in Fig.  3, the migration of full-length recombinant PDE4D3, as well as endogenous PDE4D3 from FRTL5 cells (data not shown), corresponded to an S value of 6.1 (apparent molecular weight ϭ 108,000) whereas PDE4D2 had an S value of 3.7 (apparent molecular weight ϭ 50,000). These values are consistent with dimer and monomer, respectively, as the apparent molecular weight of PDE4D3 is twice the apparent molecular weight of PDE4D2. The removal of the C-terminal residues in the PDE4D3-⌬CAT construct did not appreciably change the mobility of the protein in comparison to the full-length PDE4D3 (Fig. 3). This finding lends support to the hypothesis that the elution of full-length PDE4D3 and PDE4D2 at a much higher apparent molecular weight than the corresponding ⌬CAT proteins is caused by an anomalous behavior of the full-length proteins on the gel filtration column and not due to an additional oligomerization domain at the C terminus. Finally, the molecular weights of PDE4D3 and PDE4D2 correspond to dimer and monomer, respectively, based on calculations according to Siegel and Monthy (28) shown in Table I. These calculations utilize Stokes radii obtained from gel filtrations and sedimentation coefficients obtained from density gradient centrifugations in order to take both size and shape of the proteins into account.
Reversibility of PDE4D3 Dimerization-In order to analyze the stability of PDE4D3 dimerization, the pull-down of differentially tagged PDE4D3 constructs, either co-expressed or expressed separately, was compared. Differentially tagged PDE4D3 constructs fused C-terminally to either a V5/His tag or a Myc/His tag (see Fig. 1A for the design of these constructs) were transfected in COS7 cells either on the same or separate plates. In the latter case, the cell pellets from both plates were mixed together before the cells were lysed. After homogenization, cytosolic fractions of the cell lysates were prepared as described under "Experimental Procedures" and were incubated for up to 24 h at 4°C under continuous end-to-end rotation of the tubes. They were then used for pull-down experiments using the ␣-V5 mAb, and the immunoprecipitates were analyzed for the presence of the differentially tagged PDE constructs in Western blotting using the ␣-Myc mAb and the ␣-V5 mAb. Whereas the co-transfected constructs behaved as shown in Fig. 1A, co-immunoprecipitation of the separately expressed constructs was not detectable even when both constructs were incubated for 24 h at 4°C (data not shown) indi-FIG. 1. Oligomerization of PDE4D3, but not PDE4D2, is detected in pull-down assays. A, vectors encoding human PDE4D3 (see scheme) fused to either a C-terminal V5/His tag (lanes 3-6) or a Cterminal Myc/His tag (lanes 1, 2, 5, and 6) were transfected in COS7 cells either separately (lanes 1-4) or co-transfected together (lanes 5 and 6). Two days later, cells were harvested, and the cytosolic fractions of the cell lysates were used for pull-down assays with either mIgG as a control (lanes 1, 3, and 5) or the ␣-V5 mAb, which recognizes the V5 tag of the 4D3-V5/His construct. Immunoprecipitates (shown in the two lower panels), as well as the cell lysates used as input for the pull-down experiments (shown in the two upper panels), were analyzed for the presence of the differentially tagged PDE constructs. B, the experiments were performed as described under A except that differentially tagged constructs of the short splice variant PDE4D2 were used. C, two vectors encoding either full-length human PDE4D3 or a C-terminally truncated PDE4D3 that is no longer recognized by the M3S1 mAb (see scheme) were transfected in COS7 cells either separately (lanes 1-4) or co-transfected together (lanes 5 and 6). Pull-down assays using the corresponding cell lysates were then performed using either the M3S1 mAb generated against the PDE4D C terminus (lanes 2, 4, and 6) or mIgG as a control (lanes 1, 3, and 5). Both PDE4D3 constructs are recognized by the K116 pAb generated against an N-terminal region of PDE4 that was used to detect these constructs in the cell lysates (upper panel) and the immunoprecipitates (lower panel). All data are representative of experiments performed at least three times. Equal amounts of protein were loaded in each lane.
cating that the dimerization is stable and not subject to dissociation/reassociation. This result is consistent with gel filtration and density gradient centrifugation profiles that did not show the presence of a monomeric PDE4D3 species. It also demonstrates that the co-IPs of the differentially tagged constructs shown in Fig. 1, A and C are not due to aggregation of the PDEs during the immunoprecipitation procedure.
Mapping of the PDE4D3 Dimerization Domain-To confirm the dimerization of PDE4D3 and to further narrow the location of the dimerization domain within the N terminus, the co-IP of several N-terminal-truncated constructs with the full-length enzyme was investigated. As shown in Fig. 4, the deletion of 79 residues at the N terminus of the protein, which removes the N-terminal PDE4D3-specific sequences as well as the N-terminal part of the UCR1 (termed UCR1N, aa 50 -80 of PDE4D3), did not affect the co-IP. Further truncation of the UCR1 domain, however, prevented co-immunoprecipitation indicating that these sequences are part of the dimerization domain. , each expressed in MA-10 cells, were applied to density gradient centrifugation using 5-33% linear sucrose gradients and centrifuged for 36 h at 28,000 rpm. After centrifugation, 200-l fractions were collected and analyzed for PDE activity. Several marker proteins were separated on parallel gradients at the same time. 14 C-methylated bovine serum albumin (Sigma) was used as an internal standard. All data are representative of experiments performed at least three times.
induce quiescence. Under these conditions, PDE4D3 is the only PDE4D splice form in these cells and the most abundant PDE expressed. The cells were then harvested and lysed as described under "Experimental Procedures," and the high-speed supernatant was separated on a TSK-3000 gel filtration column. The resulting fractions were analyzed for PDE-activity (q). Three hundred microliters of the fractions containing enzyme activity were used for pull-down assays using the PDE4Dspecific M3S1 mAb. After several washings, the final IP-pellets were resuspended in 300 l of homogenization buffer and used for the determination of PDE-activity (E). B and C, MA-10 cells were transfected with vectors containing either wild type PDE4D3 or PDE4D2 cDNAs (q) or the corresponding -⌬CAT constructs encoding for C-terminally truncated PDE4D splice variants (E). After cell harvest and lysis, the corresponding high speed supernatants were applied to size exclusion chromatography. All data are representative of experiments performed at least three times. The elution profiles of the PDE constructs were confirmed by Western blotting. The marker proteins reported were separated on the gel filtration column under identical conditions. could be co-immunoprecipitated in pull-down experiments, whereas the PDE4D2 (encoding aa 167-672 of PDE4D3) could not, the region between amino acids 80 and 167 including the subdomains UCR1C, LR1, and UCR2N (see Fig. 5A for domain organization) probably contains the putative dimerization domain.
This region has been the object of a comprehensive yeast two-hybrid analysis that identified an intramolecular but not an intermolecular interaction. The yeast two-hybrid system was, therefore, also employed in the present study to further characterize the putative PDE dimerization domain. Confirming previous reports (21,22), UCR1C (aa 80 -116), and UCR2N (aa 132-185) specifically interact with each other (Fig. 5B; see co-transformant AD-UCR1C:BD-UCR2N) whereas neither of them (AD-UCR1C and AD-UCR2N) interacts with a construct containing both the UCR1C and the UCR2N domain (BD-UCR1C-UCR2N). However, in further experiments it could be demonstrated that although not interacting with the separate UCR2N or UCR1C, the fragment spanning UCR1C-LR1-UCR2N did interact with itself (co-transformant AD-UCR1C-UCR2N: BD-UCR1C-UCR2N). Therefore, an intermolecular interaction between amino acid positions 80 -185 also could be detected using the yeast two-hybrid system, and both UCR1C and UCR2N are essential for this interaction.
Design and Characterization of Monomeric PDE4D3 Constructs-To confirm the conclusions drawn from the yeast twohybrid experiments, PDE4D3 constructs carrying nested deletions of UCR1C and/or UCR2N were generated as ⌬CATconstructs and their apparent molecular weights were determined by gel filtration. In addition, a PDE4D3-⌬CAT construct carrying single point mutations of several charged residues within the UCR1C and UCR2N domains, which were previously shown to disrupt the putative intramolecular interaction between the two domains (22), was analyzed.
As shown in Fig. 6A, the nested deletion of either UCR1C or UCR2N leads to abrogation of the enzyme dimerization and the proteins eluted at apparent molecular weights of ϳ72,000 from the gel filtration column. However, the 4D3-⌬CAT construct containing the point mutations eluted in a manner indistinguishable from the wild type 4D3-⌬CAT (at ϳ152,000). The nested deletion of either UCR1C or UCR2N in the full-length PDE4D3 caused the elution of both proteins from gel filtration columns at an apparent molecular weight of ϳ183,000 in comparison to ϳ388,000 of the full-length enzyme (Table I). When analyzed by density gradient centrifugation, deletion of either UCR1C and/or UCR2N in the full-length PDE4D3 or the 4D3-⌬CAT construct also caused a shift of the enzyme migration to 3.6 S (M app. ϭ 47,000; see Table I) and 3.8 S (M app. ϭ 51,000;  UCR2N (aa 133-166), or the entire sequence between UCR1C and UCR2N (aa 80 -166), respectively. *These data represent the average and range of two measurements. All other data shown represent mean Ϯ S.E. for n Ն 3 experiments. M theor. ϭ theoretical molecular weight for a monomer calculated from the amino acid composition; M app. ϭ apparent molecular weight; M calc. ϭ molecular weight calculated according to Siegel and Monthy ((28); see "Experimental Procedures" for formula).  1 and 2) as well as three Myc-tagged PDE4D3 constructs with increasing N-terminal deletions (lanes [3][4][5][6][7][8] were expressed in COS7 cells either alone (lanes 1, 3, 5, and 7) or co-expressed with a V5/His-tagged PDE4D3 construct (lanes 2, 4, 6, and 8; see scheme for design of constructs). Cell cultures were transfected, harvested, and lysed as described under "Experimental Procedures," and the cell lysates were then applied to pulldown experiments using the ␣-V5 mAb that recognizes the V5 tag of the PDE4D3-V5/His construct. Both the immunoprecipitates (two lower panels) as well as the cell lysates used as input for pull-down experiments (two upper panels) were analyzed for the tagged PDE4D constructs by Western blotting using the ␣-V5 and the ␣-Myc mAbs. Equal amounts of protein were loaded in all lanes. All data are representative of experiments performed at least three times. Fig. 6B), respectively.
In agreement with the above results, the nested deletion of UCR1C and UCR2N ablated the oligomerization detected in co-IP assays whereas the mutation of the charged residues within these domains did not (Fig. 7). A summary of the properties of all the constructs is reported in Table I.
Impact of PKA Phosphorylation on PDE4D3 Dimerization-Because the intramolecular UCR1C-UCR2N interaction is ablated by PKA phosphorylation at the PDE4D3 N terminus (Ser-54; Ref. 22), the impact of PKA phosphorylation on enzyme dimerization was determined. However, in vitro phosphorylation of PDE4D3 by PKA did not shift the elution of the enzyme to a lower apparent molecular weight indicating that phosphorylation does not abrogate PDE4D3 dimerization. On the contrary, the PKA-phosphorylated enzyme consistently eluted approximately one fraction prior to the unphosphorylated form from the gel filtration column. Similar shifts in elution were observed whether native enzyme (FRTL5 cells treated with forskolin to induce PKA phosphorylation; data not shown), recombinant PDE4D3 (data not shown), or the PDE4D3-⌬CAT construct was used (Fig. 8A). The shift in elution of the phosphorylated PDE4D3 underscores the major conformational changes of the holoenzyme induced by this post-translational modification. Moreover, it is likely that the UCR domains are required for mediating the PKA-dependent regulation of enzyme activity as the internal deletion of either UCR1C or UCR2N prevents the activation of PDE4D3 upon PKA phosphorylation (Fig. 8B). DISCUSSION Two decades of investigation on the role of PDEs in cyclic nucleotide signaling have established that the function of these enzymes extends well beyond termination of signals from the G protein-coupled receptor/adenylate cyclase system. Indeed, PDEs play an important role in fine-tuning cAMP levels in the cell, establishing cross-talk between different signaling pathways (7,8,13,29) and controlling cAMP compartmentalization (30 -35). Given the recent elucidation of the crystal structure of PDE4B (36), much is known about the atomic organization and function of the catalytic domain of PDEs. Though important, considerably less is known about the structural organization of the N-and C-terminal PDE domains. This is particularly true for PDE4 as the function of the conserved N-terminal domains (UCR1 and UCR2) unique to these enzymes is far from established. Using several different experimental approaches, we provide evidence here that these conserved regions constitute the dimerization domain of these proteins.
Four distinct lines of evidence indicate that the UCR1/UCR2 module is responsible for PDE4 oligomerization. Firstly, co-IP assays indicate that PDE4D3 is able to form oligomers whereas PDE4D2, which does not contain UCR1 and UCR2N, behaves as a monomer. Removal of the C terminus does not affect oligomerization of PDE4D3 assessed in the co-IP assay. Secondly, studies with a combination of gel filtration and sucrose density gradient centrifugation support the hypothesis that PDE4D3 behaves as a dimer whereas PDE4D2 does not. Thirdly, internal deletion of the UCR domains converts the full-length PDE4D3 from an oligomer to one with the properties of a monomer. Finally, yeast two-hybrid experiments further confirm that a domain encompassing the C terminus of UCR1 and the N terminus of UCR2 is capable of oligomerization. All these findings strongly support the hypothesis that the N terminus is critical for oligomerization of PDE4D3. We should emphasize that our data do not exclude the possibility that other domains, perhaps at the C terminus or within the catalytic domain, contribute to the stabilization of the quaternary structure of PDE4.
It is generally accepted that most PDEs exist as dimers or oligomers. This property has been most extensively studied for PDE5 and PDE6, and both physical characterization and threedimensional molecular organization have been reported (37,38). The dimerization domain is located at the N terminus of the protein, possibly overlapping with the cGMP binding sites/ GAF domains (37,38 A and AϩD), a PDE4D3-⌬CAT construct with an additional nested deletion of the UCR1C-LR1-UCR2N domain (lanes B and BϩD; deletion of aa 80 -166), or a construct carrying single point mutations of the four charged residues previously shown to ablate the intramolecular interaction between UCR1C and UCR2N (lanes C and CϩD), were transfected in COS7 cells either alone (lanes A, B, and C) or co-transfected with a vector encoding PDE4D3-V5/His (lanes AϩD, BϩD, and CϩD). After cell harvest, the resulting homogenates were used for pull-down assays using the M3S1 mAb and the IP-pellets (lower panel) as well as the cell lysates (upper panels) were analyzed for the presence of the three different PDE4D3-⌬CAT constructs by Western blotting using the K116 pAb. These data are representative of experiments performed two times. iments are, however, required to confirm this possibility as PDE3A/B catalytic domains appear to behave as dimers (39).
At odds with the data reported on PDE5, it was thought that PDE4 oligomerization is mediated by several different domains localized within the C terminus or the catalytic domain (40 -44). On the basis of deletion mutants expressed in bacteria, a previous report had reached the conclusion that the C terminus mediates dimerization of a short PDE4D splice form (41). It should be pointed out that these conclusions were derived only from gel filtration chromatography data. We believe that the elution of PDE4 from gel filtration columns does not closely follow the mass of the native enzyme because of an anomalous migration of the PDE4 due to either an asymmetric shape and/or the highly charged C-terminal domain. Indeed, when the C-terminal portion was removed, PDE4D3 and PDE4D2 behave as dimers and monomers, respectively. Sucrose density gradient centrifugation does not appear to be affected by the presence of the C terminus as no major change in mobility was observed in the truncated constructs. Indeed, several studies have cautioned the exclusive use of gel filtration as the means to determine the molecular weight of a protein (28,45). Anomalous behavior of PDE4 also has been observed by SDS-PAGE with the protein migrating with an apparent molecular weight higher than the theoretical molecular weight. In addition, several other reports have attempted to investigate the state of oligomerization of PDE4 with mixed results due to the extensive aggregation experienced with purified recombinant proteins (40,42).
Although our data strongly support the view that the major dimerization domain coincides with UCR1/UCR2, it is possible that additional inter-subunit interactions stabilize the PDE4 dimer. Previous studies, including our own, showed by gel filtration and density gradient centrifugation that the catalytic domains of PDE4A and PDE4B behave as oligomers (40,(42)(43)(44). As the catalytic domains of the PDE4 subtypes are highly conserved, it is unlikely that this discrepancy is due to sequence differences between PDE4A/B and the PDE4D used in our present study. It is more likely that, although the UCR domains are the crucial dimerization domains and essential for dimerization in vivo, several other contacts between the two molecules that stabilize the dimer are present, for example, within the catalytic domain. Whereas these interactions might not be sufficient for dimerization in vivo, they might be sufficient for oligomerization in purified and/or concentrated preparations of overexpressed PDE4 constructs. This conclusion is consistent with data reported on the molecular organization of PDE5 and PDE6 (38) where the dimerization and tightest interaction between the two interacting monomers is dependent on the N-terminal domain. Nevertheless, the three-dimensional organization of those proteins also indicates that the catalytic domains of both monomers are in close proximity. These contacts are not sufficient to allow dimerization of a purified N-terminally truncated PDE5 construct (37). Similar contacts in the catalytic domain of PDE4 cannot be detected by co-IP or yeast two-hybrid assay, but may cause recombinant purified PDE4 constructs to form oligomers. Indeed, a point of contact within the catalytic domain has been implied by the analysis of the crystal structure of PDE4B (36).
Previous reports have shown that UCR1 and UCR2 are involved in protein interactions (21,22) whereas other interacting domains within the PDE4 were not observed in yeast twohybrid assays. Our present findings are consistent with, and expand, these previous reports. Several positively charged residues within the C-terminal part of UCR1 and several negatively charged residues within the N-terminal part of UCR2 were previously identified as critical for the UCR1/UCR2 in-teraction as it could be ablated by the mutagenesis of these residues. The UCR1C-UCR2N interaction was thought to be intramolecular, as it could only be detected when UCR1C and UCR2N were expressed as separate constructs. Although enzyme dimerization is mediated by the same domains, we found that mutation of these charged residues had no effect on the state of oligomerization of PDE4D3. Furthermore, PKA phosphorylation of the N terminus of PDE4D3 (Ser-54), previously shown to abrogate the intramolecular UCR1-UCR2-interaction, does not affect enzyme dimerization. There are at least two possible explanations for this discrepancy that cannot be distinguished at the present time. The first possibility is that enzyme dimerization is the only interaction mediated by the UCR domains, but it involves more residues than the previously identified charged amino acids. Therefore, dimerization of the full-length enzyme cannot be ablated by site-directed mutagenesis of only these residues. Conversely, in experiments where UCR1C and UCR2N are encoded by two separate truncated constructs, this interaction is weakened and site-directed mutagenesis of only these charged residues is sufficient to abrogate the interaction. On the other hand, the UCR domains may be required for both intermolecular dimer formation and for an intramolecular interaction. In the latter case, PDE4 dimerization represents a new, additional feature of the UCR domains and may explain some of the differences between the regulatory properties of long and short PDE4 splice variants.
In preliminary studies, we show that the activation of PDE4D3 upon phosphorylation by PKA is ablated in constructs lacking either UCR1C or UCR2N (Fig. 8B), thus indicating that these domains are necessary for translating the conformational changes at the PDE4 N terminus into altered catalytic functions. These issues, relating enzyme dimerization and functions (i.e. dependence of kinetic properties, inhibitor sensitivity, and metal ion dependence on the oligomerization state), will be addressed in future studies.
In summary, using several independent approaches, the present study demonstrates that PDE4D2, representative of the short splice variants, behaves as a monomer, whereas PDE4D3, a prototype of the long splice variants with complete UCR domains, is a dimer. The long PDE4 splice variants, therefore, might have a molecular organization similar to that of PDE5 and PDE6. This raises the possibility that, although different in their primary structure, the property of enzyme dimerization at the N terminus may be evolutionarily conserved among the PDE families. Thus, the possibility also should be entertained that the molecular mechanisms that regulate enzyme activity upon modulation of the N-terminal domains, via post-translational modifications or the binding of signaling molecules, may be conserved and similar in all PDEs.