Internalization and Recycling of the Human Prostacyclin Receptor Is Modulated through Its Isoprenylation-dependent Interaction with the δ Subunit of cGMP Phosphodiesterase 6*

Prostacyclin, the major cyclooxygenase-derived product of arachidonic acid formed in the vasculature, mediates its potent anti-thrombotic and anti-proliferative effects through its G protein-coupled receptor (GPCR) termed the IP. Unlike many GPCRs, agonist-induced internalization of the IP occurs in an arrestin/GPCR kinase-independent manner. However, deletion of the IP COOH-terminal region prevented internalization suggesting that protein interactions at this region are involved in IP regulation. Using the COOH-terminal region of IP as bait we identified the δ subunit of cGMP phosphodiesterase 6 (PDE6δ) as a novel hIP-interacting protein in two independent yeast two-hybrid screens. Interaction of IP and PDE6δ was confirmed by co-immunoprecipitation in HEK293 cells, and in HEPG2 cells, which endogenously express neither IP nor PDE6δ. IP isoprenylation was critical for this interaction, as PDE6δ was unable to associate with an isoprenylation-deficient mutant IP (IPSSLC). PDE6δ overexpression altered the temporal pattern of agonist-induced internalization of IP, but not IPSSLC, in HEPG2 cells, increasing initial internalization but facilitating the return of IP to the cell surface despite the continued presence of agonist. Depletion of PDE6δ using short interfering RNA abolished cicaprost-induced IP internalization in human aortic smooth muscle cells. Recycling of IP, but not IPSSLC, upon agonist removal was facilitated by overexpression of PDE6δ. Thus PDE6δ interacts specifically with IP to modulate receptor trafficking.

PGI 2 mediates its effects through the membrane-associated receptor IP, a member of the G protein-coupled receptor (GPCR) superfamily. Signaling by this family of receptors is tightly regulated. In the canonical regulatory pathway the activated receptor is phosphorylated by one or more GPCR kinases (GRKs) followed by recruitment of an adapter protein, arrestin. These events culminate in rapid uncoupling of the receptor from its G protein (desensitization) and subsequent receptor internalization (4). Once internalized the receptor may be dephosphorylated and recycled to the cell surface, restoring rapidly agonist activation, or undergo lysosomal degradation resulting in receptor down-regulation and a sustained loss of responsiveness (4). While the regulatory mechanisms employed by many GPCRs adhere, to a reasonable degree, to this pathway, we (5,6) and others (7,8) have reported that, similar to other GPCRs (9,10), regulation of IP deviates from this norm. Thus rapid desensitization of activated IP occurs secondary to it's phosphorylation, not by a GRK but by a second messenger activated kinase, PKC (6). Subsequently the IP is internalized through a phosphorylation-and arrestin-independent pathway (5,7,8). Rapid recycling of the sequestered IP upon agonist withdrawal, which restores the response to agonist without the need for de novo protein synthesis, is evident in platelets (8), fibroblasts (7), and transfected HEK293 cells (5).
The molecular pathways, therefore, that direct agonist-dependent trafficking of IP remain ill-defined but diverge substantially from the classical GPCR regulatory pathway. Deletion of the IP COOH-terminal tail abolished its rapid agonist-dependent receptor internalization (5) suggesting that key regulatory interactions occur along this portion of the protein. Using the COOH-terminal region of the IP as bait, we isolated a specific IP-interacting protein, in vascular smooth muscle cells, which was identified as the ␦ subunit of the phosphodiesterase (PDE) 6 enzyme complex. This protein was originally identified as the fourth subunit of rod-specific cGMP PDE6 (11). In that system the interaction of PDE6␦ with the isoprenylated COOH termini of the enzymes ␣ and ␤ catalytic subunits leads to their solubilization from the plasma membrane and uncoupling of PDE6 from its effector, transducin (12), in addition to potentially regulating intracellular trafficking of newly synthesized PDE6 (13). The functional importance of PDE6␦ appears, however, to extend beyond both retinal tissue and the PDE6 enzymatic complex with which it was first associated. Unlike the other subunits, widespread expression of PDE6␦ in extraretinal tissue, including heart, placenta, lung, and brain (14), has been reported. This, along with its marked conservation throughout evolution (15), suggests a broader role in cellular signaling. Indeed, several groups have reported the interaction of PDE6␦ with a established cell signaling molecules, including members of the Ras superfamily of guanine nucleotide-binding proteins (16) and the ␣ subunit of at least one heterotrimeric G protein G i , leading to their translocation into the cytosol (17). It is apparent that PDE6␦ can act in a manner similar to guanine nucleotide dissociation inhibitors, to inhibit Rap and Ras signaling (16), while con-versely acting as an effector for, or functional component of, signaling via Arl1, Arl2, and Arl3 (17,18) and Rab13 (19), small G proteins involved in vesicular transport.
In the present study, we examined the role of PDE6␦, a specific hIPinteracting protein, in IP signaling internalization and recycling.

MATERIALS AND METHODS
Yeast Two-hybrid Screen-The COOH-terminal (CTER) region of hIP, from Val 299 through Leu 386 was generated by PCR and cloned into the yeast expression plasmid pBDGal4cam (Stratagene). The yeast strain YRG2 (Stratagene) transformed with this hIP-CTER bait was used to screen a Gal4 activation domain fusion cDNA library from human (h) aortic vascular smooth muscle cells (ASMC; Biowhittaker Inc., Walkersville, MD), prepared as described previously (20). Interacting clones were selected on media lacking the amino acids Trp, Leu, and His, according to the manufacturer's instructions. Background His reporter expression was inhibited with 3-aminotriazol (3-AT), to titrate out the number of false positives. LacZ expression in surviving colonies was inferred from ␤-galactosidase activity, assayed, as described (20), in a colony filter lift assay using 5-bromo-4-chloro-3-indolyl ␤-D-galactoside as substrate. The cDNAs from LacZ-positive clones were sequenced across the Gal4/library cDNA boundary and analyzed using the BLAST algorithm at the NCBI (www.ncbi.nlm.nih.gov/BLAST/).
Construct Generation and Site-directed Mutagenesis-Primers were designed to mutate Cys 383 to Ser 383 (sense oligonucleotide: 5Ј GCC AGC GTC GCC AGC TCC CTC TGC TGA TGG ATC C), in the hIP, triple tagged (3x) at its amino terminus with the hemagglutinin (HA) epitope tag (obtained from the University of Missouri-Rolla cDNA Resource Center). Mutagenesis was carried out using the QuikChange II site-specific plasmid DNA mutagenesis kit (Stratagene). The mutated receptor was termed IP SSLC reflecting the change in the amino acid sequence of the last four amino acids of hIP from CSLC to SSLC. Mutagenesis was confirmed by DNA sequencing.
To generate PDE6␦ tagged at its carboxyl terminus with a V5 (GKPIPNPLLGLDST) epitope tag, the stop codon was removed from PDE6␦ by PCR using a sense oligonucleotide containing three miscellaneous bases, followed by the sequence encoding the first five amino acids in the hPDE6␦ sequence (5Ј-ATC ATG TCA GCC AAG GAC). The antisense oligonucleotide was complimentary to the coding sequence for the six amino acids preceding the stop codon (5Ј-AAC ATA GAA AAG TCT CAC). The resulting PCR product, encoding the full-length PDE6␦ absent its stop codon, was cloned directly in pcDNA3.1/V5-His, using the pcDNA3.1/V5-His TOPO TA expression kit (Invitrogen), to allow expression of the PDE6␦-V5 fusion protein.
hIP and hTP␣, tagged at their amino termini with a single copy of HA, and cloned into pCDNA3 (Invitrogen), were generated as described previously (21).
HEK293 and HEPG2 cells grown to 50 -80% confluence in 100-mm dishes were transfected transiently using FuGENE 6 transfection reagent (Roche Applied Science) according to the manufacturer's instructions. Experiments were carried out 48 h post-transfection. In control co-transfections DNA levels were equalized with empty vector.
siRNA Transfection-hASMC grown to 80% confluence in 100-mm dishes were transfected with 100 M control (Ambion, Austin, TX) or PDE6␦-targeting (SMARTpool; Dharmacon RNA Technology, Lafayette, CO) siRNA using the DharmaFECT 1 transfection reagent (Dharmacon RNA Technology) according to the manufacturer's instructions. Cells were incubated with siRNA-containing medium for 48 h, transferred to 12-well plates, and grown for a further 48 h prior to assay.
Co-immunoprecipitation-Cells were treated with 2 mM dithiobis(succinimidyl propionate) (Pierce) for 30 min, to cross-link covalently surface proteins, and lysed with radioimmune precipitation buffer (50 mM Tris, 5 mM EDTA, pH 8.0, 150 mM NaCl, 1% Nonidet P-40, 0.1% SDS, 0.5% deoxycholic acid, and a mixture of protease inhibitors) for 10 min at 4°C prior to centrifugation at 14,000 rpm for 10 min. The resulting supernatants were precleared by 1-h rotation with 50 l of 10% (w/v) protein G-Sepharose (Amersham Biosciences) to each tube. Anti-HA-protein G-Sepharose was prepared by adding 5 g anti-HA ascites per lysate to 10% protein G-Sepharose followed by 1-h rotation. HAhIP or HAhTP␣ was immunoprecipitated, from precleared lysates, by adding 100 l of anti-HA-protein G-Sepharose to each lysate and rotating for 16 h at 4°C. Protein G was precipitated at 14,000 rpm for 1 min, washed three times with radioimmune precipitation buffer, and resuspended in 10 l of sample buffer (NuPAGE). Immunobloting for co-immunoprecipitated V5-PDE6␦ was carried out, as described above.
Measurement of Cell Surface HAhIP-Surface expression was measured by ELISA, as described previously (5). Briefly, 48 h after transfection, cells seeded on 24-well dishes were treated with the agent of interest at 37°C and reactions stopped by aspiration and fixation (4% paraformaldehyde in PBS, 4°C, 10 -15 min). HA expression was quantified by incubation with monoclonal anti-HA antibody (1:1000 dilution in PBS) followed by alkaline peroxidase-conjugated anti-mouse IgG (Jackson ImmunoResearch, 1:10,000 dilution in PBS) for 30 min. Cell surface alkaline phosphatase was detected, after four washes with PBS, following conversion of 4-nitrophenyl phosphate by measurement of absorbance at 405 nm.
Binding Studies-Cell surface IP was examined in hASMC by radioligand binding to intact cells. Indomethacin (3 M)-pretreated hASMC were incubated with a saturating concentration (50 nM) of [ 3 H]iloprost, an IP agonist, in binding buffer (Hanks' balanced salt solution containing 2 mg/ml bovine serum albumin) in the presence or absence of excess (10 M) unlabeled iloprost for 18 h at 4°C. Cells were washed three times in ice-cold buffer, treated with 1 M NaOH for 30 min at 37°C and cell-associated radioactivity quantified.
cAMP Measurements-HEPG2 cells grown to confluence in 12-well plates were pretreated with isobutylmethylxanthine (0.01 M) 30 min prior to agonist addition. Cells were treated and reactions were terminated by aspiration. cAMP was extracted with ice-cold 65% ethanol for 30 min. Samples were dried under vacuum and reconstituted in assay buffer, and cAMP was quantified by radioimmunoassay as described previously (22).

RESULTS
Yeast Two-hybrid Screen-The COOH terminus of hIP (hIP-CTER) was used as bait in a hASMC cDNA library yeast two-hybrid screen. Eleven ␤-galactosidase-positive clones, which grew in the absence of His/Leu/Tryp and presence of 5 mM 3-AT, were identified. Analysis of the DNA sequences revealed that each encoded the same 150 amino acid protein; two independent clones were identified, 9 were identical. BLAST analysis revealed an exact match with a 17-kDa protein identified as the ␦ subunit of PDE6, a cGMP PDE (23). These findings were confirmed in a second independent screen, carried out at lower stringency (2.5 mM 3-AT). The coding sequence of PDE6␦ was cloned, minus its stop codon, into the expression vector pcDNA3.1/V5-His to generate the PDE6␦-V5 fusion protein.
The interaction of hIP with PDE6␦ was confirmed using HEK293 transected transiently with PDE6␦-V5 and co-transfected with or without HAhIP. Lysates, subjected to immunoprecipitation with anti-HA, were examined for the presence of PDE6␦ by immunoblotting with anti-V5. PDE6␦ was co-immunoprecipitated only when both proteins were co-expressed (Fig. 1C, lane 5). In contrast, PDE6␦-V5 was not evident in anti-HA precipitated lysates prepared from HAhTP␣/ PDE6␦-V5 co-transfected cells (Fig. 1C, lane 6) confirming that the interaction was specific to hIP and was not due to the presence of HA or V5 epitope tags.
Screening for Endogenous PDE6␦ Expression-Widespread expression of PDE6␦ has been reported (14). To elucidate the role of PDE6␦ in modulating IP function, we first sought to identify a suitable cell model for ectopic PDE6␦ expression. Of the several commonly used cell types screened by Western blot for endogenous PDE6␦ expression, HEPG2 cells were found to express negligible levels of protein ( Fig. 2A). Furthermore IP expression also was absent in HEPG2 cells, as determined by cAMP accumulation in response to treatment with the IP agonist cicaprost (data not shown). As such, HEPG2 cells represent an ideal cell model to determine the functional consequence of IP-PDE6␦ interaction. HEPG2 cells were transfected with PDE6␦-V5 with or without 3xHAhIP and their interaction confirmed by co-immmunoprecipitation (Fig. 2D, lane 1).
Isoprenylation as a Requirement for IP-PDE6␦ Interaction-Interaction between PDE6␦ and the majority of its binding partners is dependent on their isoprenylation (11,16). The IP is one of the very few GPCRs that contains a conserved "CAAX" motif, Cys 383 -Ser-Leu-Cys 386 , for isoprenylation in its COOH-terminal tail. Indeed addition of C15-farnesyl isoprenoid groups to the IP has been reported. Moreover this modification was appeared critical for both IP signaling and efficient agonist-dependent sequestration (24). To determine whether the IP-PDE6␦ interaction was dependent on IP isoprenylation, we generated an isoprenylation-deficient IP receptor (IP SSLC ) in which Cys 383 , a critical residue within the isoprenylation motif, was mutated to Ser (24).
Interaction of IP SSLC with PDE6␦ was examined in HEPG2 cells transfected transiently with PDE6␦-V5 and co-transfected with 3xHAhIP or IP SSLC . PDE6␦ was co-immunoprecipitated from HEPG2 cells co-expressing 3xHAhIP but not from cells co-expressing the isoprenylationdeficient IP mutant or those expressing PDE6␦ alone (Fig. 2D), demonstrating that similar to many other PDE6␦ interactors, IP isoprenylation was required for its interaction with PDE6␦. Concordantly, treatment of cells with lovastatin (20 M), which inhibits IP isoprenylation (25), also prevented IP-PDE6␦ interaction (data not shown). Effect of PDE6␦ Overexpression on IP-mediated cAMP Generation-Impaired effector activation via IP SSLC has been demonstrated (24) leading those authors to conclude that isoprenylation of IP was critical for its coupling to downstream G proteins. The dependence of the IP-PDE6␦ interaction on IP isoprenylation led us to examine whether PDE6␦ modulates IP signaling. HEPG2 cells, transfected with 3xHAhIP with or without PDE6␦, were stimulated with the prostacyclin analogue cicaprost for 5 min. A robust concentration-dependent increase in cAMP was observed in response to IP activation (EC 50 ϭ 0.11 Ϯ 0.06 nM, n ϭ 3). Overexpression of PDE6␦ did not alter significantly IP-mediated cAMP generation (EC 50 ϭ 0.1 Ϯ 0.06 nM, n ϭ 3). In a similar manner IP-cAMP signaling was unaltered when IP-transfected HEK293 cells were co-transfected with PDE6␦-V5 (data not shown). Thus, the interaction with PDE6␦ was not responsible for the signaling deficit associated ablation of IP isoprenylation (24).
Role of PDE6␦ in IP Sequestration-Although the precise function of PDE6␦ remains unknown, its interaction with proteins such as the ␣ and ␤ subunits of PDE6, and the small G protein Rab13, resulted in their translocation from the membrane to the cytosol. We (5) and others (7) have reported that sequestration of activated IP does not operate through the classical phosphorylation/arrestin-dependent pathway associated with GPCRs. Indeed, while the pathways that direct IP trafficking remain ill-defined, a COOH-terminal truncated IP was not internalized following activation with agonist (5). Moreover, in contrast to a phosphorylation-deficient IP mutant (6), sequestration of the isoprenylation deficient IP SSLC mutant markedly was impaired (24), suggesting a role for isoprenylation in IP trafficking. The involvement of PDE6␦ in IP internalization was assessed in HEPG2 cells transfected with 3xHAhIP with or without PDE6␦-V5. IP internalization in response to cicaprost (1 M) occurred more rapidly in HEPG2 cells overexpressing PDE6␦; 15 min after agonist activation cell surface expression of 3xHAhIP was reduced to only 87.6 Ϯ 6.6% of control in 3xHAhIP transfected cells compared with 64.2 Ϯ 7.5% in cells co-expressing PDE6␦-V5 ( Fig. 3A; p Ͻ 0.01). Furthermore, PDE6␦ overexpression altered the temporal pattern of IP internalization. In 3xHAhIP transfected HEPG2 cells, surface expression of IP declined steadily over a 2-h treatment period. In contrast, and similar to IP transfected HEK 293 cells (5), which express endogenously PDE6␦ ( Fig. 2A), substantial IP internalization was evident after only 15 min of cicaprost treatment in HEPG2 cells co-transfected with PDE6␦. Indeed, in stark contrast to the PDE6␦-null HEPG2 cells, sequestration of IP was maximal at the 15-min time point and appeared reversible, with a significant return of surface IP as agonist exposure was extended ( Fig. 3A; 82.9 Ϯ 4.9% after 2 h versus 64.2 Ϯ 7.5% after 15 min; p Ͻ 0.05).
PDE6␦ Has No Effect on IP SSLC Internalization-We investigated next whether the effect of PDE6␦ on IP internalization was mediated FIGURE 2. Isoprenylation as a requirement for IP-PDE6␦ interaction. A, cell lysates derived from NIH3T3, HeLa, COS-7, and HEK293 and HEPG2 cells were resolved by 10% reducing SDS-PAGE and endogenous PDE6␦ detected using an anti-PDE6␦ antibody, as described under "Experimental procedures." Cell lysates derived from HEPG2 cells cotransfected with V5-PDE6␦ plus either empty vector (pcDNA3), 3xHAhIP, or the isoprenylation-deficient 3xHAhIP SSLC mutant were resolved by 10% reducing SDS-PAGE and expression of 3xHAhIP and 3xHAhIP SSLC (B) or V5-PDE6␦ (C) confirmed. D, lysates were subjected to immunoprecipitation with anti-HA antibody and co-immunoprecipitated V5-PDE6␦ detected using an anti-V5-HRP antibody. Co-immunoprecipitation of V5-PDE6␦ was only observed when 3xHAhIP and V5-PDE6␦ were co-expressed (arrow). Western blots are representative of three independent experiments. through its direct interaction with the isoprenylated receptor. Cicaprost-induced internalization of the IP SSLC mutant was observed in HEPG2 cells and occurred more rapidly than previously reported in HEK293 cells (24), likely reflecting cell type-specific differences in the relative importance of receptor trafficking pathways present in these cells. Importantly, however, and in direct contrast to the wild type IP, co-expression of PDE6␦ did not modify internalization of the isoprenylation-, and PDE6␦ interacting-, deficient IP mutant (IP SSLC ; Fig. 3B). Moreover, cicaprost-induced sequestration of IP SSLC was indistinguishable from the wild type IP in the absence (open bars; Fig. 3, A and B) but not in the presence of PDE6␦ (closed bars). Thus, divergence in the sequestration of the wild type and mutant receptors was evident only when PDE6␦ was co-expressed, either endogenously (24) or ectopically (closed bars; Fig. 3, A and B). Taken together these data suggest strongly that that interaction with PDE6␦ was responsible for the defect in sequestration reported previously for the isoprenlyation deficient mutant IP (24).
Effect of siRNA-mediated Knockdown of PDE6␦ on IP Internalization in hASMC-We next wished to determine whether PDE6␦ modulated IP sequestration in hASMC, which endogenously express both proteins ( Fig. 4A and Ref. 21). siRNA targeted against PDE6␦ (PDE6␦-siRNA) resulted in almost complete abolition of PDE6␦ protein expression in hASMC, when compared with either untransfected or control siRNA transfected cells (Fig. 4A). Substantial IP internalization was observed in control siRNA-transfected hASMC following a 1-h treatment with 100 nM cicaprost ( Fig. 4B; 79.3 Ϯ 2.6%; p Ͻ 0.001 compared with cells without cicaprost treatment), with a similar response observed in untransfected cells (data not shown). Strikingly, cicaprost-induced IP internalization was abolished in hASMC transfected with PDE6␦-siRNA (Fig.  4B), demonstrating the importance of this novel pathway in facilitating IP sequestration in its native setting.
Effect of PDE6␦ Overexpression on IP Recycling-A substantial number of sequestered GPCRs, including the IP (5,7,8), are recycled to the plasma membrane following agonist withdrawal. Having observed a biphasic change in surface expression of IP, in which the initial rapid decline was reversed with extended treatment in PDE6␦ co-expressing HEPG2 cells, we wished to determine its contribution to IP recycling. As the rate of IP internalization differed significantly between PDE6␦-null and PDE6␦ transfected cells, we employed those treatment conditions that evoked maximal and equivalent levels of sequestration in the two settings, namely 1 M cicaprost for 1 h or 15 min, respectively, prior to agonist removal. Surface expression of IP was restored to control levels within 60 min of agonist withdrawal only in cells co-expressing IP with PDE6␦ (Fig. 5), with levels remaining significantly depressed in PDE6␦null cells (84.8 Ϯ 6.2% of control, p Ͻ 0.05). In contrast, co-expression of PDE6␦ did not induce recycling of IP SSLC (Fig. 5), indicating that the effect of PDE6␦ on IP recycling is a consequence of its interaction with the isoprenylated receptor.

DISCUSSION
We (5, 6) and others (7,8) demonstrated that, similar to other GPCRs, IP undergoes rapid densensitization and internalization in response to agonist activation. However, IP does not utilize the classical GRK/arrestindependent pathways that are implicated commonly in regulation of this superfamily (4). Instead, desensitization is mediated through PKC-dependent phosphorylation (6), while IP sequestration occurred, in part, through a dynamin-dependent clathrin-coated vesicular pathway that did not employ arrestin as an adapter molecule (5). Determinants for both regulatory steps are located in the COOH-terminal tail region of the IP, although only in the case of PKC-dependent phosphorylation has the critical residue, Ser 328 , been identified (6). Although this particular site was not required for agonist-dependent sequestration of the IP, the COOH-terminal tail of the receptor proved indispensable for this regulatory step to proceed (5), suggesting that this region may provide a docking site for proteins involved in IP regulation (5). Indeed, novel GPCR interacting proteins, including some that contribute to receptor trafficking, have been reported previously (26,27). Thus, we sought to identify proteins interacting with the IP COOH-terminal region and characterize their role in IP regulation.  Using the COOH-terminal region of IP as bait in a yeast two-hybrid screen of a hASMC cDNA library, we identified the ␦ subunit of PDE6 enzyme as a novel IP-interacting protein. The identity of this IP interactor was verified in a second independent screen, and the interaction was confirmed in two distinct mammalian cell lines, HEK293 and HEPG2 cells, transfected with HA-and V5-epitope-tagged versions of the hIP and PDE6␦, respectively. Co-immunoprecipitation PDE6␦-V5 was specific to the IP receptor and not an artifact either of its forced overexpression or the presence of the HA-or V5-tag, since under similar transfection conditions, in HEK293 cells, PDE6␦-V5 was not coimmunoprecipitated with HAhTP␣ (Fig. 1C).
Although first identified in the retina as a subunit of PDE6, a cGMP phosphodiesterase (28), its broad expression (14), marked conservation (19), and the relative promiscuity of its protein interactions (17)(18)(19) have led to recharacterization of PDE6␦, not as an PDE6 subunit but rather an associated protein with a more general role in cell signaling (29). Having no catalytic activity of its own, PDE6␦ interacts with the isoprenylated, COOH termini of PDE6 ␣ and ␤ subunits, resulting in the solubilization of the holoenzyme from the membrane to the cytosol. Similarly, PDE6␦ interacts members of the Ras, Rap, Rho, and Rab families of small G proteins resulting in their membrane extraction and fulfilling a role reminiscent of a guanine nucleotide dissociation inhibitor (16,17). In contrast to these inhibitory roles, PDE6␦ reportedly acted as an effector for two other small G proteins, Arl2 and Arl3 (17), both of which appear to play a role in vesicular transport, interacting specifically with their GTP bound form.
To assess the biological function of IP-PDE6␦ interaction we identified a cell model suitable for ectopic expression of both proteins. Most cell types examined expressed PDE6␦ at a high level ( Fig. 2A), concurrent with the widespread expression of the protein (14); however, HEPG2 cells demonstrated negligible expression ( Fig. 2A). Thus HEPG2 cells, which also lack functional IP expression (data not shown), were selected as an ideal model to study the impact of PDE6␦ on IP function.
The majority of the protein partners of PDE6␦ (11,17), including the IP (24), identified to date are modified post-translationally by the addition of an isoprenoid group at the so-called "CAAX" motif where A stands for an aliphatic amino acid and X stands for any amino acid. Mutation of IPs "CAAX" motif, from Cys 383 -Ser-Leu-Cys 386 to Ser 383 Ser-Leu-C 386 rendered the receptor insensitive to isoprenylation (24) and ablated its interaction with PDE6␦ (Fig. 2D), linking the two processes. It should be noted that although a frequent requirement, isoprenylation is not always necessary for interaction with PDE6␦; interaction with the Arl proteins can proceed without their isoprenylation (17), while, though isoprenylated, neither Rab4 nor Rab6 are capable of interaction with PDE6␦ (16). Thus PDE6␦, though promiscuous, is not indiscriminate in its associations and likely realizes consequences relevant to the function of its specific partners including the IP.
Isoprenylation appears to be a critical post-translational modification in the biosynthesis of IP; the isoprenylation-deficient IP SSLC mutant receptor does not couple efficiently to its requisite G protein (24), making the IP unusual among GPCRs. At least one heterotrimeric G protein subunit, G i (17), has been added to PDE6␦'s repertoire of associated proteins, raising the possibility of its involvement in potentially complex interactions within the microenvironment of a GPCR. Signaling of IP to activation of adenylyl cyclase was unaltered, however, by co-expression of PDE6␦ in either HEK293 cells (data not shown) or HEPG2 cells, suggesting that the interaction with PDE6␦ does not contributed directly to signaling events immediately downstream of IP activation, at least in relation to the G s -adenylyl cyclase cascade. Thus, although IP isoprenylation is necessary for its interaction with both G s (24) and PDE6␦ (Fig. 2D), the two processes are not functionally linked. We have not examined the role PDE6␦ in modulating IP signal transduction through additional signaling pathways with which it has been associated, namely G q (22) and, in the case of the mouse receptor, G i (30). However, the biological relevance of IP signaling via pathways other than cAMP generation is not apparent.
Although the precise role of PDE6␦ remains unclear, it's ability to solubilize proteins from the membrane into the cytosol has led to the suggestion that it is involved principally in the regulation of protein transport. Indeed, many of PDE6␦'s protein partners are themselves involved in vesicular transport. Thus, Arl family proteins have been implicated in transport to and from the trans-Golgi network (31), while Rab13 mediates the continuous recycling of occludin to the cell surface (32). The isoprenylation-deficient IP SSLC mutant, which we have determined is incapable of interacting with PDE6␦ (Fig. 2D), is also defective in terms of agonist-induced sequestration (24). As such we examined whether IP-PDE6␦ might modulate the agonist-induced internalization of IP, a process that remains ill-defined.
Ectopic expression of PDE6␦ in HEPG2 cells facilitated the rapid, and at least partially reversible, internalization of IP in the presence of agonist (Fig. 3A). These observations were not related to extraneous changes in the expression level of functional IP due to co-expression of PDE6␦, since the maximal cAMP generation in both settings was identical. It may be argued that, since PDE6␦ interacts with a number of proteins involved in vesicular transport, modulation of IP internalization occurred secondary to a general PDE6␦ trafficking effect. However, interference with the IP-PDE6␦ interaction, by mutation of Cys 383 , would not prevent these additional PDE6␦ interactions, yet trafficking of the IP SSLC mutant was unaltered by co-expression of PDE6␦ (Fig. 3B). This finding argues against a generalized vesicular transport effect of PDE6␦ and for a shift in IP trafficking directed specifically by its interaction with the isoprenylated COOH terminus of IP. Moreover, the similarity with which the wild type and mutant IP internalized in PDE6␦-null HEPG2 cells (open bars; Fig. 3, A and B), together with the inability of PDE6␦ to modify sequestration of the mutant (Fig. 3B), suggests strongly that the inability of the mutant to interact with PDE6␦ in HEK293 cells was responsible for the trafficking impairment reported in that model (24).
While rapid ligand-dependent IP sequestration has been demonstrated previously in platelets (33), neither the endocytotic pathways utilized during, nor the regulatory elements involved in IP internalization, have been identified in cells, which like platelets, endogenously express the receptor. We examined the contribution of PDE6␦ in facilitating IP internalization in hASMC, a cell type relevant to IP-mediated cardiovascular protection (1, 34), using siRNA-mediated knockdown. Strikingly, while significant cicaprost-induced IP internalization was observed in hASMC transfected with control siRNA, the response to cicaprost was abolished in hASMC depleted of PDE6␦ (Fig. 4B), supporting a specialized role for PDE6␦ in IP trafficking in endogenously expressing cells.
In addition to modulation of receptor sequestration, recycling of IP upon removal of agonist, a phenomenon that restores responsiveness in both transfected (5) and native cells (7,8), was enhanced by PDE6␦ (Fig.  5), secondary to its specific interaction with the isoprenylated IP. We did not extend this time course beyond 2 h and thus have not considered the potential impact of PDE6␦ on restoration of surface IP expression by longer term mechanisms including de novo protein synthesis (7,35). However, it appears that the rapid return of IP to the cell surface was directed, at least in part, by PDE6␦.
The apparent reversibility of IP internalization in PDE6␦-overexpressing HEPG2 cells may similarly be explained by enhanced IP recycling. Receptor trafficking is a dynamic process, with ligand-induced internalization and constitutive recycling occurring in parallel (36). The absolute density of IP present at the plasma membrane at any given point following its activation is determined by a combination of these processes. In HEK293 cells, IP utilizes multiple internalization pathways (5), contributing to post-endocytic receptor sorting and the rapidity with which IP recycles to the surface. Indeed, PDE6␦ appears to direct the IP along a pathway characterized by rapid internalization (Fig. 3A) and recycling (Fig. 5). Overexpression of PDE6␦ may force this pathway to predominate committing the IP to rapidly recycle to the cell surface following its sequestration, rendering ligand-dependent internalization reversible.
We have not examined directly the consequence of this altered trafficking pathway for IP signaling; such investigations are not generally informative because the absolute cellular level of a second messenger, in this case cAMP, measured so long after initial activating event, reflects multiple cellular mechanisms, such as hydrolysis by cAMP-specific phosphodiesterases, and indirect kinase-driven effects on receptor and/or effector activity and is therefore not a faithful readout of receptor function. Instead we have focused on events that modulate cell surface expression of the receptor, a critical determinant of both termination and restoration of receptor activation. The marked shift in trafficking of IP evoked by co-expression, native or ectopic, of PDE6␦ suggests that a novel endocytotic pathway is utilized by this receptor as a result of its unique post-translational modification.
Loss of receptor function in response to agonist activation is a hallmark of GPCRs reported in all but a few family members. The regulatory events that occur immediately following agonist activation vary mechanistically from receptor to receptor but qualitatively are similar; a combination of receptor phosphorylation, uncoupling from its cognate G protein, and sequestration from the plasma membrane culminate in diminished responsiveness to agonist (4). The relative predictability of these regulatory steps contrasts with the divergent post-endocytotic pathways that the sequestered GPCR may take, which determine, to a large extent, restoration of receptor activation. The determinants of post-endocytotic receptor trafficking remain unclear, especially for those receptors that are regulated in an arrestin-independent manner. Certainly, this is the case for the IP where there is substantially more information regarding the pathways that do not contribute to its trafficking (5) compared with those that do. In our quest to redress that paucity of knowledge, we have, in the present study, added IP to the list of GPCRs that associate with proteins unrelated to classical GPCR signaling and demonstrated the relevance of this association for IP regulation in cells that endogenously express both proteins. The requirement that IP be isoprenylated, for it to associate with PDE6␦, is highly unusual for a GPCR, raising intriguing questions as the specialized function of PDE6␦ in context-specific regulation of IP. Indeed it is apparent that the cellular function of IP extends beyond its traditional G s -cAMP signaling cascade to include cAMP-independent interruption of the cell cycle (37), heterodimerization with at least one other prostanoid receptor, namely the thromboxane receptor (21). The specialized isoprenylationdependent interaction of IP with PDE6␦ may contribute to this complex biology.