Membrane Localization of Cyclic Nucleotide Phosphodiesterase 3 (PDE3) AND OF, PDE3 WITH ENDOPLASMIC RETICULUM*

Subcellular localization of cyclic nucleotide phos-phodiesterases (PDEs) may be important in compartmentalization of cAMP/cGMP signaling responses. In 3T3-L1 adipocytes, mouse (M) PDE3B was associated with the endoplasmic reticulum (ER) as indicated by its immunofluorescent colocalization with the ER protein BiP and subcellular fractionation studies. In transfected NIH 3006 or COS-7 cells, recombinant wild-type PDE3A and PDE3B isoforms were both found almost exclusively in the ER. The N-terminal portion of PDE3 can be arbitrarily divided into region 1 (aa 1–300), which contains a large hydrophobic domain with six predicted transmembrane helices, followed by region 2 (aa 301–500) containing a smaller hydrophobic domain (of ; 50 aa). To investigate the role of regions 1 and 2 in membrane association,

By hydrolyzing cAMP and cGMP, cyclic nucleotide PDEs 1 are critical in terminating cyclic nucleotide signals and regulating biological processes mediated by these second messengers. Eleven different, but structurally related, PDE gene families (PDE1-11) have been identified (1)(2)(3). Although intracellular location of PDEs is thought to be important in compartmentalization of cyclic nucleotide-mediated processes and regulation of discrete signaling pathways (4), little is known of the mechanisms that target PDEs to their intracellular destinations.
PDE3 isoforms are characterized by their high affinity for both cAMP and cGMP and their sensitivity to inhibition by a number of positive inotropic agents (5,6). The two PDE3 subfamilies, PDE3A and PDE3B, are products of distinct but related genes, with additional diversity generated within the PDE3A subfamily by transcription initiation from alternative sites (5,7). PDE3 isoforms have different subcellular locations, being predominantly membrane-associated in adipocytes, cytosolic in platelets, and both cytosolic and sarcoplasmic reticulum (SR)-associated in myocardium (8 -11). SR-associated and cytosolic PDE3s in myocytes may be functionally distinct (11)(12)(13). Since particulate and soluble forms of PDE3 in cultured vascular smooth muscle cells are differentially regulated by cAMP (12), subcellular location may also be important in defining mechanisms by which PDE3 isoforms are regulated. Studies using specific inhibitors suggest that PDE3 isoforms regulate cAMP and cGMP pools involved in control of lipolysis, glycogenolysis, myocardial contractility, smooth muscle relaxation, mesangial cell proliferation, and insulin and renin secretion (14 -18). Since these processes occur in cells that contain multiple PDE isoforms, compartmentalization could provide a mechanism by which PDE3 isoforms selectively alter specific cAMP pools and regulate distinct signaling pathways.
Isoforms from PDE2 and PDE4 families also differ in their subcellular locations. Three PDE2 (19) splice variants, PDE2A1, PDE2A2, and PDE2A3 (20 -23), have divergent Nterminal sequences; a hydrophobic domain at the N terminus of rat PDE2A2 may be involved in its association with brain membranes (23). Divergent N-terminal regions of PDE4 iso-forms (24,25) also may account for several differences in their catalytic properties and subcellular locations. The N-terminal 25 amino acids of rat PDE4A1A exert an inhibitory effect on catalytic activity (26) and contain information for its targeting to brain membranes (27,28). The N terminus of human PDE4A4B may also allow association with membranes, since the splice variant PDE4A4C, which lacks this region, is entirely soluble (29). The proline-rich N terminus of PDE4A5 interacts with the v-Src-SH3 domain (30).
N-terminal domains may also confer specific properties upon PDE3A and B enzymes. Analysis of the N-terminal regions of PDE3A and B predicts a hydrophobic domain of ϳ200 amino acids containing six transmembrane helices (5), which seems to be important for membrane association (31,32). Just following this region are consensus sites for phosphorylation by protein kinase A (PKA) and protein kinase B (PKB), thought to be important in phosphorylation/activation of PDE3 (33)(34)(35)(36); subcellular location could influence the specificity of protein kinase interactions with PDE3. To analyze some of the structural determinants involved in membrane targeting/association of PDE3, a series of Flag-tagged full-length and N-terminal deletion mutants of PDE3A and PDE3B, as well as a series of N-terminal MPDE3B-EGFP fusion proteins, were constructed and expressed in mammalian cells. The intracellular location of endogenous PDE3B in murine 3T3-L1 adipocytes and of the different PDE3A and PDE3B recombinants was determined by immunofluorescence techniques.
All Flag-tagged constructs were generated by inserting an eight amino acid Flag epitope (Asp-Tyr-Lys-Asp-Asp-Asp-Asp-Lys) immediately upstream from the stop codon of the HPDE3A or MPDE3B cDNA (see Fig. 1). The following specific primers pairs were used: for HPDE3A, sense primer (5Ј-CTTCATCTCTCACATTGTGGGGCCTCTG-TG-3Ј), which corresponds to HPDE3A nt 3009 -3027 and encompasses the unique DraIII site (underlined) in HPDE3A, and antisense primer (5Ј-TTTGCGGCCGCCTCGAGTTATTTATCATCATCATCTTTATAAT-CCTGGTCTGGCTTTTGGGTTGG-3Ј), which corresponds to HPDE3A nt 3423-3403 and encodes an XhoI site, a new stop codon, and the Flag epitope (underlined); for MPDE3B, sense primer (5Ј-TATCACCCACA-TTGTGGGCCCCCTG-3Ј), which corresponds to MPDE3B nt 2901-2925 and contains the unique DraIII site (underlined) in MPDE3B, and antisense primer (5Ј-AAAGCGGCCTCGAGGTCTTAT-TTATCATCATCATCTTTATAATCTTCAAACATTTGTTCTTCCTC -3Ј ), which corresponds to MPDE3B nt 3300 -3280 and encodes an XhoI site, a new stop codon, and the Flag epitope (underlined). Using HPDE3A cDNA (50 ng) as template and HPDE3A primer pairs (3 pmol each) or MPDE3B cDNA (50 ng) as template and MPDE3B primers pairs (3 pmol each), PCR fragments were amplified in a GeneAmp PCR system (PerkinElmer Life Sciences model 9600) with Pfu polymerase (Stratagene). The resulting PCR products contain the unique PDE3 DraIII site at the 5Јend and a stop codon at the 3Јend; the stop codon is flanked upstream by a Flag epitope-coding sequence and downstream by an XhoI site. The PCR products were subcloned into the pCRII vector (Invitrogen) and isolated from this vector as DraIII/XhoI fragments.
The reconstituted DraIII sites and all sequences downstream from this site were confirmed by sequencing. For expression studies, Flagtagged HPDE3A and MPDE3B constructs were excised from the pZ vectors with KpnI/XhoI and XhoI, and then ligated to the KpnI/XhoI sites or the XhoI site of pcDNA3, respectively (Invitrogen).
For generation of PCR fragments encoding MPDE3B aa 201-250, The primer pair indicated by the arrows was used to amplify the sequences downstream of the unique DraIII sites in PDE3A and PDE3B. The sense primer corresponds to the region flanking and including the DraIII site. The antisense primer corresponds to the region immediately upstream of the stop codon and contains a 3Ј extension with the Flag epitope (denoted by the flag symbol), a new stop codon, and an XhoI site, in that order. The resulting PCR product was ligated to XhoI/DraIII PDE3 fragments of HPDE3A, MPDE3B and different N-terminal truncant mutants that contained all sequences upstream of the DraIII site, thus generating Flag-tagged PDE3 recombinants. For Flag-tagged M3B-⌬302, an XhoI/AlwnI N-terminal fragment was ligated to an AlwnI/ XhoI Flag-tagged C-terminal fragment from Flag-tagged MPDE3B.
184 -250, 301-450, and 401-500, the sense primers listed in Table I were used. Sense primers for amplification of MPDE3B PCR fragments 201-250 and 184 -250 contained an XhoI site followed by a Kozak sequence and an ATG start site; for fragments 301-450 and 401-500, primers contained a BglII site followed by a Kozak sequence and an ATG start site. Antisense primers are also shown in Table I. All contain a SalI site at the 3Ј end. The PCR products were generated using the conditions outlined above and subcloned into pCRII (Invitrogen). The fragments were isolated from this vector as XhoI/SalI or BglII/SalI fragments and fused, in-frame, to the 5Ј end of EGFP cDNA contained in vector pEGFP-N1. These PDE3EGFP fusion proteins were designated pEGFP-3B-(201-250), EGFP-3B-(184 -250), EGFP-3B-(301-450), and EGFP-3B-(401-500), respectively. All sequences and points of fusion between the 3Ј ends of PDE3B fragments and pEGFP-N1 were confirmed by sequencing.
Construction of R3B-⌬101-266 -RPDE3B pSV/SPORT (full-length rat (R) PDE3B (Ref. 31) cloned into pSV/SPORT vector) was digested with SacII to remove the putative membrane association domain of PDE3B. NarI, which digests one of the short deleted fragments at a unique site, was also used to reduce the numbers of short fragments that would have double-ended SacII sites and therefore be able to religate with the large linearized R3B-⌬101-266/pSV SPORT. To reduce further risks of intermolecular religation, the small digested fragments were removed by gel electrophoresis. The linearized R3B-⌬101-266/pSV SPORT was then recircularized with T4 DNA ligase and the deletion verified by sequencing.
DNA Sequencing-DNA sequences were determined using the ABI Prism dye terminator cycle sequencing ready reaction kit (PerkinElmer Life Sciences) with the following PCR conditions: 25 cycles of 96°C for 10 s, 50°C for 5 s, and 60°C for 4 min. The extension products were purified using Centri-Sep columns (Princeton Separations). Automated sequencing was performed for 16 h on an ABI 373A sequencer or for 7 h on an ABI 377 sequencer.
3T3-L1 fibroblasts were cultured under the conditions described above in DMEM containing sodium pyruvate (110 g/ml), sodium bicarbonate (3 mg/ml), and biotin (8 g/ml). Cells (2 ϫ 10 4 ) were grown to confluence on coverslips, and differentiation was induced by addition of DMEM containing 0.5 mM 3-isobutyl-1-methylxanthine, 1 M dexamethasone, and 5 g/ml insulin (MDI). MDI was added at the same time each day for 3 days. After incubation for 5 days with DMEM containing 5 mM glucose and 5% fetal bovine serum, Ͼ90% of the cells had differentiated into adipocytes. COS-7 cells were homogenized 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, 42 M cytochalasin B and protease inhibitors (initially as a 1000ϫ stock in Me 2 SO) at a final concentration of 23 mM phenylmethanesulfonyl fluoride, 1 mM benzamidine, and 1 g/ml each aprotinin, leupeptin, pepstatin A, and antipain, as described previously (27). Ho-mogenates were centrifuged (100 ϫ g, 5 min, 4°C) and the resulting supernatants centrifuged (100,000 ϫ g, 1 h, 4°C). The final supernatants and resulting pellets were suspended in a volume of complete KHEM buffer equal to that of the supernatant fraction, and stored (Ϫ20°C).
Protein content was determined by the method of Bradford (Bio-Rad) or, where detergent was present, by the bicinchoninic acid protein assay reagent kit (Pierce). Bovine serum albumin was used as a standard in all protein assays.
Transfection-Cells (5 ϫ 10 5 /10-cm plate) were cultured in antibioticfree medium for 24 h. The medium was replaced with 5 ml of serum-free medium, and cells were transfected with 4 g of DNA using Lipo-fectAMINE Plus (Life Technologies, Inc.) following the manufacturer's instructions. After 3 h, the transfection medium was replaced with complete medium. Mock-transfected cells were treated as above, but with empty vector or pcDNA3 vector containing enhanced green fluorescent protein (EGFP) cDNA. Cells were harvested 24 h after transfection. For immunofluorescence, cells were plated on 18 mm ϫ 18-mm coverslips in six-well plates, transfected with 1 g of pcDNA3 or 1 g of EGFP-N1 vector (CLONTECH), and fixed and immunostained 24 -48 h after transfection.
Antibodies-The IgG fractions of rabbit polyclonal antiserum raised against the first 15 aa of rat PDE3B protein were purified (used at 1:200 dilution for immunostaining) as described previously (31). Flag mouse monoclonal antibody (dilutions of 1:200 for immunostaining and 1:2000 for immunoblotting) was purchased from Kodak IBI; rabbit polyclonal and mouse monoclonal IgG conjugated to rhodamine B or Oregon Green Indirect Immunofluorescence Studies-Cells grown on coverslips were washed three times with PBS. COS-7 cells and NIH 3006 cells were fixed for 20 min in paraformaldehyde (3% (w/v)) in PBS containing 0.5 M Ca 2ϩ and 0.5 M Mg 2ϩ . Cells were washed three times with PBS and quenched for 10 min with 50 mM NH 4 Cl. 3T3-L1 adipocytes were fixed in 70% EtOH, 30 mM glycine, pH 2.5, for 1 h at Ϫ20°C. For plasma membrane staining, 3T3-L1 adipocytes were incubated (1 h) at 4°C with 50 g/ml rhodamine lens culinaris agglutinin (Vector Laboratories), washed six times with ice-cold PBS, and fixed, but not quenched. Cells were permeabilized for 4 min in PBS containing 0.1% (w/v) Triton X-100, washed three times with PBS, three times with PBS/gelatin/ serum (0.2% fish skin gelatin (Sigma) and 0.1% goat serum (Life Technologies, Inc.) in PBS) and incubated for 2 h with primary antibody diluted in PBS/gelatin/serum. The coverslips were washed three times with PBS/gelatin/serum, three times with PBS, stained for 2 h with the secondary antibody, then washed three times with PBS/gelatin/serum and three times with PBS. The coverslips were mounted on glass slides with Mowiol (Calbiochem). Negative controls included staining of nontransfected cells with the primary antibody, staining of transfected cells with the secondary antibody only, peptide inhibition of staining with  primary antibody, staining with pre-immune serum, and staining of transfected cells not treated with detergent. Cells were viewed using a Nikon Diaphot fluorescence microscope fitted with a 60ϫ oil immersion objective. For analysis by confocal microscopy, the cells were examined with a Leica model TCSD/ DMIRBE microscope equipped with argon and argon-krypton lasers for blue (488 nm) and green (568 nm) excitation and a 100ϫ objective. The images shown in the different figures are representative of those obtained from at least three separate immunostaining experiments. Three to five images per experiment were taken for each combination of staining patterns.
Subcellular Fractionation of 3T3-L1 Adipocytes-Subcellular fractions of differentiated 3T3-L1 adipocytes (in 100-mm tissue culture dishes), 13-14 days after initiation of differentiation, were prepared by a modification of Clark (37). The cell monolayers (20 dishes/experiment) were rinsed twice with ice-cold buffer A (20 mM Tris-HCl, pH 7.4, 2.5 mM EDTA, 250 mM sucrose, 0.5 mM Pefabloc, 10 g/ml pepstatin, 10 g/ml, 10 g/ml aprotinin, and 10 g/ml leupeptin), harvested in buffer A, and homogenized (2 ml/dish, 27 strokes with a Teflon pestle, in a 40-ml homogenization vessel). Although 20 dishes were used per experiment, cells from 5 plates were harvested and combined for homogenization. Approximately 2 ml of the whole cell homogenate was removed for later analysis. The homogenate was centrifuged (approximately 100 ϫ g, 10 min) to remove large fragments and whole cells. This supernatant was then centrifuged (13,000 ϫ g, 20 min) using an SS-34 rotor (ϳ10 ml/tube) at Ϫ4°C (at all other times, the samples were handled on ice or at 4°C). The infranatant between the fat cake and the pellet (P1) was removed using a syringe and a pipette. The fat cake was discarded and the infranatant centrifuged (15,000 ϫ g, 20 min) to produce P2. P2 was resuspended, centrifuged (20,000 ϫ g, 20 min), resuspended again in Buffer A (500 l), and then layered upon a sucrose cushion containing 8 -9 ml of Buffer A containing 1.12 M sucrose and centrifuged as described by Clark (37) (77,000 ϫ g, 1 h, SW-41 Ti rotor). P1 was processed the same as P2, except that it was centrifuged twice over the sucrose cushion. The membrane material above the sucrose cushion from the centrifugations was combined, diluted to 0.25 mM sucrose, and centrifuged (35,000 rpm in an SW-41 rotor for 1 h). The resulting pellet (P3) was labeled plasma membrane. The material in P1 that sedimented through the sucrose cushion was labeled mitochondria/ nuclei. P2 was recovered as a pellet between the plasma membrane and the high density microsomal fractions. The supernatant from the initial P2 was centrifuged (30,000 ϫ g, 30 min) to produce P4. The resulting supernatant was then centrifuged (35,000 ϫ g, 30 min) to produce P5. P4 and P5 were resuspended separately and centrifuged (40,000 ϫ g, 30 min) to produce, respectively, the high density microsomal fraction and another fraction, which was recovered as a pellet between the high and low density microsomal fractions. The supernatant from P5 was centrifuged (95,000 rpm, 25 min) in a TLA 100 rotor to produce P6, which was resuspended and recentrifuged to produce low density microsomal fractions. The final supernatant was then filtered as described by Clark (37). All final pellet fractions were resuspended in buffer A containing 1 mM EDTA.
Citrate synthase activity was assayed as described by Sere (38) and modified by Doenst (39). 5Ј-Nucleotidase activity was determined by a modification of Simpson (40). The assay mixture contained (in 150 l): 50 mM Tris-HCl, pH 8.0, 10 mM MgCl 2 , 1 mM 2Ј-AMP and 3Ј-AMP (to inhibit nonspecific phosphatase activity), and 60 M 5Ј-AMP containing approximately 100,000 cpm [ 3 H]5Ј-AMP/assay. The assay was started by the addition of the substrate, incubated for 30 min at 37°C, boiled for 2-3 min to stop the reaction, cooled on ice, and then applied to a QAE-Sephadex (Amersham Pharmacia Biotech) column (1 ml) and eluted with 3.6 ml of H 2 O to separate 5Ј-AMP from adenosine. Radioactivity was measured in the presence of 10 ml of scintillation mixture. PDE activity was determined by a modification of Manganiello and Vaughan (41). The assay mixture contained (in 100 l): 37.5 mM Hepes, pH 7.4, 15 mM Tris-HCl, pH 7.4, .075 mM EGTA, 22.5 mM MgCl 2, 0.1 M cAMP containing approximately 30,000 cpm [ 3 H]cAMP/assay, and plus or minus 1 M cilostamide to differentiate between total activity and PDE3 activity. NAPDH-cytochrome c reductase activity was determined by the method of Dallner (42), except that the assay (1.5 ml) contained 0.3 mM KCN and was performed at 37°C. Galactosyltransferase activity was determined by a modification of the methods of Sichel (43) and Fleischer (44). Briefly, the assay mixture contained (in 150 l): 35 mM Hepes, pH 7.2, 45 mM NaCl, 40 mM MnCl 2 , 40 mM 2-mercaptoethanol, 40 mM N-acetylglucosamine (or 0 for con-trols), 0.67% Triton X-100, and 400 M UDP-[ 14 C]galactose (Gal) containing approximately 200,000 cpm/assay. Incubation without UDP-Gal was performed at 37°C for 15 min, at which time substrate was added and the incubation continued for an additional 30 min. The reaction was stopped by the method of Sichel (43) and then 5/6 of the reaction mix was applied to a QAE-Sephadex (Amersham Pharmacia Biotech) column (1 ml) and eluted with 3.6 ml of H 2 O. Radioactivity was measured in the presence of 10 ml of scintillation mixture.
To determine the localization of native MPDE3B, fully differentiated 3T3-L1 adipocytes were also costained with a rabbit polyclonal antibody against the N terminus of RPDE3B (31) and with mouse monoclonal antibodies against the ER protein BiP and the Golgi marker p58 (Fig. 3, panels 1 and 2). To visualize the plasma membrane, cells were stained with rhodamine-labeled lens culinaris agglutinin (Fig. 3, panel 3). To prevent excessive agglutinin staining of internal membranes, cells were chilled and treated with agglutinin prior to fixation and permeabilization. Although agglutinin staining clearly demarcated plasma membranes, some internal agglutinin staining was also observed, but not to the extent of that observed with PDE3B, BiP, and p58. Similar to the subcellular fractionation studies (Fig. 2), this study (Fig. 3) also revealed that native MPDE3B is primarily associated with the ER, as shown by the similarity of its distribution to that of BiP, the ER marker, with less co-localization with Golgi or plasma membrane markers. No specific staining of 3T3-L1 fibroblasts was observed with preimmune serum or anti-PDE3B antibody, or of adipocytes with preimmune serum (data not shown).
Subcellular Localization of RPDE3B in Stably Transfected NIH 3006 Cells-NIH 3006 cells stably expressing RPDE3B were costained with a rabbit polyclonal antibody against the N terminus of RPDE3B (31) and with mouse monoclonal antibodies against BiP, p58, or the microtubular protein, ␤-tubulin (Fig. 4). Examination of the costained cells using confocal mi-croscopy indicated the presence of a reticular network in NIH 3006 cells, as in 3T3-L1 adipocytes. The pattern of staining for RPDE3B in NIH 3006 cells was virtually identical to that for BiP, the ER marker. A composite image of the PDE and ER staining (Fig. 4, image 1c) demonstrates superimposition of the two staining patterns. In contrast to 3T3-L1 adipocytes (Fig. 3), where some endogenous MPDE3B associated with Golgi, there was little evidence for co-localization of RPDE3B with the Golgi marker in stably transfected NIH-3006 cells (Fig. 4). PDE3 staining was distinct from that for ␤-tubulin as well as p58. The costaining patterns for PDE/p58 and PDE/␤-tubulin were identical to those obtained for BiP/p58 and BiP/␤-tubulin, providing further evidence that recombinant PDE3 reacts like an ER protein, as does endogenous PDE3B in murine 3T3-L1 adipocytes.
The specificity of the anti-PDE3B antibody was established by the absence of PDE immunostaining after incubation of the antibody with its antigenic peptide, after incubation of cells with second antibody only, or after incubation of nonpermeabilized cells with anti-PDE3B antibody. Incubation of cells with preimmune serum gave weak nonspecific nuclear staining, not abolished by prior incubation of antiserum with the RPDE3B antigenic peptide. Additionally, the anti-PDE3B antibody did not immunostain HPDE3A expressed in NIH 3006 cells (31). Incubation of anti-PDE3B antibody with peptide antigen also abolished staining on Western blots of the 123-kDa band normally present in lysates from MPDE3B-transfected cells (data not shown).
Construction of Truncated HPDE3A and MPDE3B Recombinants-In order to investigate which regions of PDE3 are in- volved in membrane association, we first constructed a series of catalytically active N-terminal deletion mutants and studied their localization in COS cells and Sf9 cells (32).
From analysis of secondary structure, the N-terminal portion of PDE3A and PDE3B can be arbitrarily divided into region 1 (aa 1-300) and region 2 (aa 301-500). Hydrophobicity plots (Fig. 5) reveal the presence of a large hydrophobic domain (spanning aa 60 -255 and 68 -250 in HPDE3A and MPDE3B, respectively) in region 1, followed by region 2, largely hydrophilic but containing a small hydrophobic domain.
The topology for HPDE3A and MPDE3B was predicted using the TMBASE data base for membrane-spanning protein segments (46). As illustrated in Fig. 6, one algorithm predicts the existence of six transmembrane helices within the hydrophobic domain in region 1. In this model, the N terminus begins near the cytoplasmic surface; subsequent sequence traverses the membrane six times, forming three transmembrane loops of different sizes, and ends on the cytoplasmic side. This model further predicts that succeeding amino acid sequence, including the catalytic domain, is situated on the cytosolic face of the membrane. In region 2, smaller hydrophobic domains (aa 340 -390 and 320 -370 in HPDE3A and MPDE3B, respectively) could also potentially be involved in membrane association.
To facilitate immunofluorescence detection of PDE3, fulllength HPDE3A and MPDE3B and the N-terminal mutants were tagged with the Flag epitope, which was inserted at the extreme end of the hydrophilic C terminus to maximize accessibility of the epitope for antibody binding (Fig. 7). PDE activities observed after expression of untagged and Flag-tagged PDE constructs in COS-7 cells (data not shown) were similar, indicating that the Flag tag did not disrupt enzyme activity.
We expressed wild-type HPDE3A, MPDE3B, and N-terminal truncation mutants in Sf9 insect cells and determined the subcellular distribution of their activities (32). HPDE3A, MPDE3B, and H3A-⌬189 activities were predominantly recovered in particulate fractions of Sf9 cells. These recombinants, which contained all or part of the hydrophobic domain in region 1, were like integral membrane proteins, in that they could be solubilized with salt plus detergent (32). The activity of M3B-⌬302, which lacked region 1, was recovered in both cytosolic (ϳ60% of total) and particulate fractions. The association of M3B-⌬302 with particulate structures reflected that of a peripheral, non-integral membrane protein, and was disrupted by salt extraction (32). The small hydrophobic domain in region 2 could be responsible for this salt-sensitive association, since the activity of H3A-⌬397 (which lacks this domain) was predominantly cytosolic and its residual particulate activity was not salt-sensitive. Activities of M3B-⌬604 and H3A-⌬511, which lacked both regions 1 and 2, were almost completely cytosolic (32).
Expression of Flag-tagged Full-length and Truncated HPDE3A and MPDE3B Isoforms in COS-7 Cells-The Flagtagged PDE constructs were expressed transiently in COS-7 cells; Western blots ( Fig. 8) indicated that the subcellular distribution of the expressed proteins in these cells was similar to that in Sf9 cells (32). Flag-tagged MPDE3B (Fig. 8a) and HPDE3A (Fig. 8b) were present in particulate but not in soluble fractions; H3A-⌬189 (Fig. 8b) was also predominantly in particulate fractions. M3B-⌬302 (Fig. 8a) was present in both fractions, but predominantly in the soluble fraction, as was H3A-⌬397 (Fig. 8b), which lacked region 1 and part of region 2. M3B-⌬604 (Fig. 8a) and H3A-⌬510 (Fig. 8b), were exclusively cytosolic. The immunoreactive bands of lower M r in the MPDE3B, HPDE3A, H3A-⌬189, and H3A-⌬397 lysates could represent proteolytic fragments of PDE3 or a soluble translation product generated by utilization of a downstream initiation site. HPDE3A contains several internal methionines flanked by potential Kozak sequences, e.g. initiation at codons for methionine 484 or 485 may explain the lower M r band seen in the HPDE3A, H3A-⌬189, and H3A-⌬397 Western blots. Initiation at any of the several potential alternative sites in MPDE3B does not appear to account for the size of the lower M r band in the supernatant. To visualize the intracellular localization of full-length and truncated PDE3 constructs, transfected COS-7 cells were immunostained with anti-Flag antibody (Fig. 9). This antibody did not stain COS-7 cells transfected with empty vector or nonpermeabilized, transfected COS-7 cells (data not shown). As seen in Fig. 9, both HPDE3A and MPDE3B showed strong reticular staining, with intense staining in the perinuclear region, similar to that observed for the ER marker BiP (data not shown). Staining for H3A-⌬189 was identical to that of HPDE3A and MPDE3B.
Distribution of M3B-⌬302 was similar to that of full-length MPDE3B (Fig. 9). In addition to reticular staining, in some images, M3B-⌬302 exhibited diffuse staining at the cell periphery. Localization of M3B-⌬302 in the extended ER is consistent with a significant portion (ϳ40%) of the activity of expressed M3B-⌬302 in particulate fractions of Sf9 cells (32), and suggests that the small hydrophobic domain or other sequences in region 2 allow for interaction of M3B-⌬302 with the ER in COS-7 cells (Fig. 9). On the other hand, removal of region 1 in M3B-⌬302 does result in cytosolic localization of M3B-⌬302 in homogenates of COS-7 (Fig. 8) and Sf9 cells (32), and, in Sf9 cells, particulate M3B-⌬302 was almost completely solubilized by salt alone without detergent (32). Taken together, these results indicate that although the hydrophobic domain in region 2 might promote targeting of M3B-⌬302 to the extended ER ( Fig. 9), the molecules associated as non-integral membrane proteins and could be readily dissociated or released by homogenization (Fig. 8). The transmembrane helices in region 1 are most likely responsible for the strong association of MPDE3B, HPDE3A, and H3A-⌬189 with, and/or their insertion into, the ER (Fig. 9) as integral membrane proteins (32). In a related study, immunostaining (using the anti-N-terminal PDE3B antibody; Ref. 31) of R3B⌬101-266, a RPDE3B mutant in which most of the hydrophobic domain in region 1 was deleted but which retained the smaller hydrophobic domain in region 2, demonstrated that the mutant, like M3B-⌬302, colocalized with BiP (data not shown). As also seen in Fig. 9, H3A-⌬397 staining was mainly diffuse, except in the perinuclear region, where some semblance of the reticular staining pattern observed with the larger constructs was evident. Staining for H3A-⌬510 and M3B-⌬604 was diffuse with no apparent reticular localization. The smallest construct, M3B-⌬604, exhibited nuclear staining, not seen with the other constructs. In COS-7 cells expressing the soluble EGFP, the staining pattern of EGFP was similar to that for M3B-⌬604 and H3A-⌬510, suggesting that these proteins, like EGFP, were present exclusively in the cytosol and, in some cases, the nucleus.
Subcellular Localization of MPDE3B, M3B-⌬604, and H3A-⌬397 in Transfected COS-7 Cells-With subcellular fractionation and immunoblotting, three localization patterns were observed. MPDE3B, HPDE3A, and H3A-⌬189 were predomi- nantly membrane-bound; M3B-⌬604 and H3A-⌬510 were predominantly cytosolic; and M3B-⌬302 and H3A-⌬397 were present in both membrane and cytosolic fractions. To evaluate the subcellular localization in more detail, one construct from each category, i.e. MPDE3B, M3B-⌬604, and H3A-⌬397, was expressed in COS-7 cells, alone or with soluble EGFP. MPDE3Btransfected cells were costained for PDE3 and either an ER or a Golgi marker. Cells cotransfected with MPDE3B and EGFP cDNA were stained for PDE3. Since mouse monoclonal anti-Flag antibody was used to stain Flag-tagged PDE, it was not possible to co-stain the ER or Golgi with the mouse monoclonal anti-BiP or anti-p58 antibodies used in previous experiments. Rabbit polyclonal BiP or p58 antibodies gave poor results. Therefore, rabbit polyclonal anti-calreticulin and anti-␤-COP antibodies, which gave staining patterns similar to BiP and p58, respectively, were used to stain the ER and the Golgi, respectively.
As shown in Fig. 10, in COS-7 cells, as in NIH 3006 cells, the MPDE3B staining pattern was identical to that of the ER marker calreticulin (panel 1), but not that of ␤-COP (data not shown) or EGFP (panel 2). The staining pattern for M3B-⌬604, on the other hand, was not like that for calreticulin (panel 3) or ␤-COP (data not shown), but was identical to that observed for EGFP (panel 4). The staining pattern for H3A-⌬397 differed from that for ␤-COP (data not shown), but was in some ways similar to those for calreticulin (panel 5) and EGFP (panel 6), which reinforces data from Western blotting experiments suggesting that this protein is present in the ER as well as the cytosol. The similarity of costaining patterns of MPDE3B/ EGFP and M3B-⌬604/ER to each other and to that of EGFPtransfected cells stained for the ER marker (data not shown) provides further proof that MPDE3B and M3B-⌬604 behaved like reticular and soluble proteins, respectively.
Expression of MPDE3B-EGFP Fusions in COS-7 Cells-Green fluorescent proteins have been widely used as a tool for identifying molecular targeting signals. Fusions of the putative transmembrane spanning domain of type I inositol 1,4,5triphosphate receptor to GFP identified this region as the domain involved in association with the ER and in homotetramer formation (47). The first 35 aa of endothelial nitric-oxide synthase, which directed GFP from the cytosol to the Golgi network (48), were identified as a Golgi-targeting signal.
To more closely examine the roles of regions 1 and 2 in the membrane association of MPDE3B, a series of cDNAs, encoding PDE3B N-terminal fragments fused to the N terminus of EGFP, were constructed (Fig. 11). COS-7 cells expressing EGFP or the PDE3-EGFP fusion proteins were stained for the ER marker BiP and examined using confocal microscopy.
Interestingly, unlike M3B-⌬302, EGFP-3B-(301-450), which contains the small hydrophobic domain between aa 328 and 370 (Fig. 11), did not display reticular staining, but was observed in the cytoplasm and nucleus (Fig. 12). This result suggests that additional structural information is needed for efficient targeting to the ER. It is also consistent with our results in COS-7 cells (Figs. 8 and 9) and Sf9 cells (32), and indicates that, although structural elements in the hydrophobic domain in region 2 might be important in targeting to the ER (Fig. 9), this domain supports weak membrane association of M3B-⌬302 (which is readily disrupted by homogenization (Fig. 8) and salt extraction (Ref. 32), but does not allow for movement of EGFP-3B-(301-450) to, and/or its retention by, the ER (Fig. 12). EGFP-3B-(401-500), which is predominantly hydrophilic (Fig. 11), did not co-localize with BiP, but exhibited diffuse cytoplasmic and nuclear staining (Fig. 12). DISCUSSION Earlier studies demonstrated that native PDE3 is membrane-associated in both adipocytes and liver cells (8,49). In rat adipocytes, PDE3, which is activated by insulin and plays a key role in mediating its antilipolytic action (14), constitutes FIG. 9. Intracellular localization of Flag-tagged PDE3 constructs in transfected COS-7 cells. Forty-eight hours after transfection with the Flagtagged PDE3 constructs, COS-7 cells were fixed and permeabilized and the transfected proteins localized using a mouse monoclonal anti-Flag antibody (diluted 1:200), followed by goat anti-mouse IgG conjugated to rhodamine as described under "Materials and Methods." Cells transfected with EGFP were fixed and visualized directly using confocal microscopy. Each panel shows the staining pattern obtained for COS-7 cells expressing the indicated recombinant protein.
Ͼ80% of the PDE activity in the particulate fraction (50). As assessed by sucrose density gradient centrifugation, adipocyte PDE3B was recovered in a microsomal fraction enriched in the ER marker NADH-dehydrogenase but not in 5Ј-AMPase, adenylate cyclase, and insulin binding activity, which are usually associated with the plasma membrane fraction (8). In rat hepatocytes, PDE3 associated with a "dense vesicle" fraction, whereas PDE4 was associated with the plasma membrane (49). These earlier findings support our subcellular fractionation studies and immunofluorescence observations demonstrating native PDE3B in the ER of 3T3-L1 adipocytes and recombinant PDE3 associated with the ER in NIH-3006 and COS-7 cells. These results, however, do not exclude the possibility of an association of PDE3 with other cellular membranes.
Our studies demonstrate that recombinant HPDE3A and MPDE3B contain domains that enable these proteins to associate with the ER in COS-7 cells and NIH-3006 fibroblasts and, for native MPDE3B, with ER in 3T3-L1 adipocytes. Although the primary structures of the N-terminal portions of HPDE3A and MPDE3B are quite different, the secondary structures appear to be highly conserved, and the predicted topology for the two isoforms is almost identical (see Figs. 5 and 6). The absence of the first two predicted transmembrane loops in region 1 (aa 1-300) of H3A-⌬189 (Fig. 6) did not appear to have a substantial effect on membrane-targeting/association. In Sf9 cells, MPDE3B, HPDE3A, and H3A-⌬189, which contained all or some of the transmembrane helical segments in the hydrophobic domain of region 1, exhibited characteristics of integral membrane proteins and were released from particulate fractions with salt and detergent (32). Thus, either the first 189 amino acids are not essential, or any part of the hydrophobic domain in region 1 may be sufficient for strong association with, or insertion into, ER membranes.
Unlike that of the other constructs, the membrane localization of M3B-⌬302 observed by immunofluorescence was not reflected in the immunoblots of subcellular fractions from COS-7 cells in which a considerable portion of M3B-⌬302 was cytosolic. In Sf9 cells, the M3B-⌬302 that associated with particulate fractions exhibited characteristics of peripheral, nonintegral membrane proteins, and was solubilized with salt alone (32). Thus, although the small hydrophobic domain lying between aa 328 and 370 in region 2 of M3B-⌬302 may be adequate for targeting to, and some interactions with, ER membranes in intact cells, the weak association that is preserved during immunostaining may not be robust enough to withstand homogenization and subcellular fractionation in COS-7 cells, and extraction of particulate fractions with salt in Sf9 cells (32).
In Sf9 cells (32) and COS-7 cells, H3A-⌬397 is predominantly cytosolic. Although most of H3A-⌬397 was located in the cytoplasm, the fraction that associated with the ER seemed to be present predominantly in the perinuclear region. M3B-⌬302 lacks region 1, but contains the small hydrophobic domain between aa 328 and 370 in region 2 of MPDE3B. H3A-⌬397 lacks the small hydrophobic domain between aa 340 and 390 in region 2 of HPD3A. The localization of M3B-⌬302 and H3A-⌬397 suggests that the small hydrophobic domain between aa 340 and 390 in region 2 of HPDE3A (absent in H3A-⌬397) contains information sufficient for targeting to the extended ER membrane network (as does the domain between aa 328 and 370 in M3B-⌬302) and that PDE3 membrane interactions may not be uniform throughout the ER. Although H3A-⌬397 retains information for targeting to the perinuclear ER (which might be a site of synthesis), in the absence of region 1 and the hydrophobic domain in region 2, this mutant lacks the ability to move to and/or anchor efficiently to the extended ER, and is easily dissociated from membranes. It is also possible that PDE3 staining is more intense in the perinuclear region than in other parts of the ER network because sites of specific PDE3 interaction are more numerous in this region, or because the sequence between aa 397 and 510 targets H3A-⌬397 to this region. Whether the greater intensity of staining in the perinuclear region is an artifact of overexpression remains to be determined, although perinuclear staining for the endogenous ER markers BiP and calreticulin was also more intense.
Evidence does exist for non-uniform distribution of ER proteins. In rat vas deferens smooth muscle, the patterns of dis- tribution of the Ca 2ϩ buffering molecules calsequestrin and calreticulin are distinct (51). The ratio of calsequestrin:calreticulin in the perinuclear SR is ϳ1:1, whereas in the peripheral ER calsequestrin is more abundant, with a ratio of 5:1. The ER retention sequence KDEL may dictate the uniform distribution of calreticulin, while the non-uniform distribution of calsequestrin (which lacks the KDEL motif) may depend on interactions with other anchoring proteins (52). Several different domains may cooperate in targeting certain proteins to the ER (53). For example, although microsomal aldehyde dehydrogenase contains a C-terminal ER retention signal (54), the hydrophilic domains flanking the hydrophobic region are also involved in targeting (55). Hydroxysteroid dehydrogenase type II is anchored to the ER via its N terminus, but deletion of the Nterminal hydrophobic region does not alter its localization, perhaps because hydroxysteroid dehydrogenase type II contains other hydrophobic segments (56). The first N-terminal transmembrane domain is sufficient to permit retention of the SR ATPase-Ca 2ϩ pump in the ER. However, as with PDE3, other structural determinants are also involved, since SR ATPase-Ca 2ϩ pump lacking the first two transmembrane domains was retained in the ER, albeit not as efficiently as the wild type protein (57).
In general, our studies of the localization of MPDE3B-EGFP fusion proteins were consistent with those using truncated, catalytically active recombinants. EGFP-3B-(1-250) and EGFP-3B-(1-200), containing all six and the first four transmembrane helical segments of region 1, respectively, did localize to the ER, as did EGFP-3B-(201-250) and EGFP-3B-(184 -250). Since EGFP-3B-(201-250) contains only one complete (the sixth) transmembrane helical segment, perhaps one segment contains sufficient information for efficient association with the ER. Whether the other transmembrane segments in region 1, if synthesized individually as EGFP-3B fusion proteins, could readily associate with the ER is not known. EGFP-3B-(1-100), which contains one transmembrane helical segment, also associated with the ER, but with a perinuclear distribution similar to that of H3A-⌬397. Consistent with the weaker ER-interactions and cytosolic distribution of MB-⌬302 and H3A-⌬397, when separated from downstream sequences, EGFP-3B-(301-450) and EGFP-(401-500) fusion proteins did not associate with ER.
Taken together, our results suggest that structural elements in both regions 1 and 2 within the N-terminal portion are involved in the interaction of recombinant PDE3 with ER membranes (Fig. 6). Although the transmembrane helices in the hydrophobic domain of region 1 contain information that allows for strong interactions with, or insertion into, membranes and may serve as an ER retention signal, this information may not be confined to any one portion of this domain. Efficient targeting seems to also involve additional weaker (but important) interactions driven by the shorter downstream hydrophobic areas (aa 340 -390 and 320 -370 in HPDE3A and MPDE3B, respectively) and the adjacent hydrophilic sequences (i.e. sequences N-terminal to aa 510 and 604 in MPDE3B and HPDE3A, respectively) in region 2. The exact mechanism by which these domains enable recombinant PDE3, which lacks a KDEL sequence, to be directed to and associate with the ER remains to be elucidated. It is clear, however, that the ability of PDE3 to associate with the ER was completely absent in PDE3A or PDE3B mutants lacking both region 1 and region 2, since H3A-⌬510 and M3B-⌬604 were located exclusively in the cytosol and, in the latter case, also in the nucleus. The reason for nuclear localization of M3B-⌬604 is unclear. Since analysis of PDE3A and PDE3B sequences using the PSORT II program (58) does predict the presence of weak nuclear localization signals (NLS) in H3A-⌬510 and M3B-⌬604, the slightly larger size of H3A-⌬510 may prevent its entry into the nucleus. Alternatively, truncated M3B-⌬604 may contain stronger NLS or an as yet unidentified NLS which promotes nuclear entry, or it may gain access to the nucleus in association with a protein that contains its own strong NLS.
The membrane association domains of PDE3 could potentially be involved in interactions either with molecular chaperones, which assist in targeting and membrane insertion (59,60) or with anchoring proteins analogous to the cAMP protein FIG. 11. N-terminal fragments of PDE3B used to construct EGFP-3B fusion proteins. A series of cDNA fragments coding for various portions of the N-terminal region of PDE3B were used in-frame to the 5Ј end of EGFP cDNA in vector pEGFP-N1 as described under "Materials and Methods." For reference, each fragment is shown in alignment with its predicted secondary structure in MPDE3B. kinase anchoring proteins (AKAPs) which target PKA to the plasma membrane, ER, Golgi, mitochondria, peroxisomes, and microtubules (61). AKAPs provide an effective mechanism for compartmentalization of PKAs, providing sites where PKAs can respond optimally to fluctuations in cAMP levels and are in proximity to specific substrates such as PDE3B (5). As with AKAPs and PKA, the subcellular distribution of PDE3 isoforms differs in different cell types. Such isoforms are found in both the SR and the cytoplasm in myocardium (10 -13). Since many of the components that regulate cardiac contractility are present in the SR, including PKA-RII and an anchoring protein AKAP 100 in rat cardiac myocytes (62, 63), it is not surprising that PDE3 should also exist in this location, where it could effectively regulate cAMP involved in contractility and other responses mediated by PKA.
Recent studies strongly support the notion that components of particular signaling pathways may be organized into signaling complexes to achieve specificity of signaling in cells where there is considerable overlap in pathways activated by different receptors. In adipocytes, for example, activation of PI3-kinase by insulin and PDGF initiates signaling via distinct pathways; activation of PI3-kinase by insulin, but not by PDGF, results in increased glucose transport. The selection of these two pathways may be related to the specific subcellular pool of PI3kinase that is activated. Insulin preferentially stimulates microsomal PI3-kinase, whereas PDGF activates PI3-kinase in plasma membranes (64). A pool of PI-3-kinase activated by insulin may exist in a complex with IRS-1 that is associated with the actin cytoskeleton (65); cytochalasin D, an inhibitor of actin filament assembly, reduces glucose transport by decreasing the association of PI3-kinase and IRS-1 with vesicles containing glucose-transporters, suggesting that the actin cytoskeleton is involved in the insulin-induced relocalization of the PI3-kinase and IRS-1 (66).
Compartmentalization may also be important in the regulation of components of kinase cascades, such as PKB, which are activated by PI3-kinase-dependent mechanisms. Both PI3-kinase and PKB activities are required for induction of meiotic maturation in Xenopus oocytes (67), and injection of mRNA encoding a constitutively active form of PKB, but not a mutant PKB lacking its membrane-targeting sequence, induced meiotic maturation, demonstrating that alterations in the location of signaling components produce dramatic effects on these responses (67). In rat adipocytes, insulin induced translocation of PKB to plasma membrane fractions (68) and of one specific PKB isoform, PKB␤, to a microsomal vesicle fraction enriched in glucose transporters (69). Since insulin, IRS-1, PI3-kinase, and PKB are all upstream mediators of PDE3B activation, it will be of interest to determine whether translocation of IRS-1, PI3-kinase, and PKB is involved in activation of PDE3.
Involvement of PDE isoforms in the compartmentalization of cAMP signaling responses has been documented in several cell types. In patch clamp studies conducted on isolated frog ventricular myocytes, Jurevicius and Fischmeister (70) found that the ␤-adrenergic agonist, isoprenaline, induced an increase in cAMP near the sarcolemma, 40-fold greater than that in the rest of the cell. The PDE inhibitor 3-isobutyl-1-methylxanthine dramatically reduced this localized rise in cAMP concentration, with diffusion of the cAMP signal to distal regions of the myocyte, suggesting that PDE activity is an absolute requirement for the maintenance of compartmentalized cAMP responses. Although the specific PDE activities involved were not characterized in this study, others have demonstrated that compartmentalization of cAMP signaling can apparently be regulated by specific PDE isoforms. Chini et al. (16) showed that PDE4 in mesangial cells regulated a cAMP pool that activates PKA involved in inhibition of the production of reactive oxygen metabolites, while PDE3 regulated a cAMP pool that suppresses cell proliferation. Since PDEs provide the only known mechanism for terminating cellular effects induced by cyclic nucleotides, knowledge of the processes that drive membrane-targeting/association and compartmentalization of these enzymes will be vital in understanding PDE function and regulation of cAMP signaling responses (4).
FIG. 12. Immunolocalization of EGFP-3B fusions. COS-7 cells expressing EGFP-3B fusion proteins were fixed. EGFP-3B fusion proteins were visualized directly using confocal microscopy; the ER was identified using mouse monoclonal anti-BiP antibodies, followed by anti-mouse IgG conjugated to rhodamine, as described under "Materials and Methods." In the first two columns are pairs of images of the same cells showing EGFP-3B fusions (a) and ER staining (b). Images in the third column (c) are composites of the superimposed images of differently stained cells in the two preceding frames.