Interaction of the Human Prostacyclin Receptor with Rab11

The human prostacyclin receptor (hIP) undergoes agonist-induced internalization and subsequent recyclization in slowly recycling endosomes involving its direct physical interaction with Rab11a. Moreover, interaction with Rab11a localizes to a 22-residue putative Rab11 binding domain (RBD) within the carboxyl-terminal tail of the hIP, proximal to the transmembrane 7 (TM7) domain. Because the proposed RBD contains Cys308 and Cys311, in addition to Cys309, that are known to undergo palmitoylation, we sought to identify the structure/function determinants of the RBD, including the influence of palmitoylation, on agonist-induced trafficking of the hIP. Through complementary approaches in yeast and mammalian cells along with computational structural studies, the RBD was localized to a 14-residue domain, between Val299 and Leu312, and proposed to be organized into an eighth α-helical domain (α-helix 8), comprising Val299–Val307, adjacent to the palmitoylated residues at Cys308–Cys311. From mutational and [3H]palmitate metabolic labeling studies, it is proposed that palmitoylation at Cys311 in addition to agonist-regulated deacylation at Cys309 > Cys308 may dynamically position α-helix 8 in proximity to Rab11a, to regulate agonist-induced intracellular trafficking of the hIP. Moreover, Ala-scanning mutagenesis identified several hydrophobic residues within α-helix 8 as necessary for the interaction with Rab11a. Given the diverse membership of the G protein-coupled receptor superfamily, of which many members are also predicted to contain an α-helical 8 domain proximal to TM7 and, often, adjacent to palmitoylable cysteine(s), the identification of a functional role for α-helix 8, as exemplified as an RBD for the hIP, is likely to have broader significance for certain members of the superfamily.

Rab proteins, the largest subgroup of the Ras superfamily of GTPases, act as key molecular switches to regulate cellular trafficking of various types of protein and non-protein cargo (1,2). Moreover, they may act as scaffolds to integrate membrane trafficking and intracellular signaling both temporally and spatially (2,3). The intracellular trafficking of certain members of the G protein-coupled receptor (GPCR) 3 superfamily is mediated by Rab4, Rab5, and Rab11, which show distinct and overlapping distribution in early and recycling endosomes (4). Whereas Rab5 facilitates receptor internalization, Rab4 and Rab11 mediate their recycling through either "short" or "long or slow" recycling pathways, respectively (4). The Rab11 subfamily of GTPases, consisting of Rab11a, Rab11b, and the epithelial Rab25 (or Rab11c), along with their many specific effectors, have emerged as principal regulators of membrane traffic from the endosomal recycling compartment (ERC) and from apical recycling endosomes in polarized cells (5,6). In addition to interacting with specific effectors, in the case of certain GPCRs, it is increasingly recognized that Rab11 members may interact directly with the receptor cargo to regulate its intracellular trafficking and signaling (7)(8)(9). Moreover, Rab25, a GTPase linked to aggressiveness and metastasis of ovarian and breast cancers (10,11), interacts directly with the ␣5␤1 integrin to regulate invasive tumor migration (12). Hence, Rab11 members may sometimes interact directly with certain of its protein cargo, be it a GPCR or integrin, to greatly influence the actual signaling events associated with that receptor system. The prostanoid prostacyclin, or prostaglandin I 2 , plays a central role in hemostasis, acting as a potent inhibitor of platelet activation and as an endothelium-derived vasodilator (13). As a major product of COX2 (cyclooxygenase 2), prostacyclin is a potent proinflammatory mediator (14,15); it confers cytoprotection during acute myocardial ischemia (16,17) and atherosclerosis (18); and it promotes enhanced endothelial cell survival, supporting angiogenesis and neovascularization (19). The importance of prostacyclin within the vasculature was further highlighted by the finding that certain COXIBs, selective inhibitors of COX2, disproportionately impair prostacyclin synthesis, leaving subjects at increased risk of thrombosis and other adverse cardiovascular episodes (20,21).
Many of the actions of prostacyclin are mediated by the prostacyclin receptor (IP), a member of the GPCR superfamily, and structurally organized into the typical seven-␣-helical trans-membrane (TM) topography (15,22). The IP is primarily coupled to G s /adenylyl cyclase activation but may regulate other effectors in a cell-and/or species-specific manner (23,24). The IP is somewhat unusual among GPCRs in that it undergoes both isoprenylation and palmitoylation within its carboxyl-terminal tail (C-tail) domain, modifications that are critical for its signaling and function (24 -30). More specifically, in the case of the human (hIP), it undergoes farnesylation at Cys 383 within an evolutionarily conserved -CAAX motif (24,25) and is dually palmitoylated at Cys 308 and Cys 311 , whereas an intervening Cys 309 was found not to be palmitoylated, at least under the experimental conditions used (27). Although neither lipidation affected its ligand binding properties, it is proposed that farnesylation in addition to palmitoylation of the hIP may confer a double loop structure within its C-tail domain to provide and/or orientate the critical structural domains for its G protein/effector(s) coupling and, possibly, for its interaction with components of the intracellular trafficking machinery to modulate its internalization after agonist activation (24,25,27). More specifically, although disruption of farnesylation effectively abolishes agonist-induced G s /adenylyl cyclase activation and cAMP generation by the hIP, palmitoylation at either Cys 308 or Cys 311 is sufficient to maintain functional G s coupling, whereas disruption of palmitoylation at both sites abolishes that signaling (24,25,27).
Through recent studies, we have established that the hIP undergoes agonist-induced internalization that occurs through a Rab5a-dependent mechanism (31) rather than through the classic G protein-coupled receptor kinase/␤-arrestin-dependent mechanism typical of many GPCRs. Although deletion of its C-tail domain did not impair its internalization per se, subsequent trafficking and recycling of the hIP was substantially impaired, suggesting that the C-tail contains the structural determinant(s) for hIP sorting after Rab5-mediated endocytosis (31). Through a subsequent yeast two-hybrid (Y2H)-based screen of a human kidney cDNA library, Rab11a was actually identified as a direct and highly specific interactant of the hIP and was found to play a critical role in its subsequent trafficking after agonist-induced internalization (32). Furthermore, the region of interaction with Rab11a was dependent on a 22amino acid hydrophobic sequence within the C-tail domain of the hIP, adjacent to the TM7 ␣-helices, which bears limited overall structural similarity to the Rab11/25 binding domain (RBD) first identified within members of the "family of Rab11interacting proteins" (FIPs) (5).
Herein, we sought to critically define the structural determinants within the hIP required for its interaction with Rab11a. Moreover, bearing in mind that the proposed RBD contains Cys 308 and Cys 311 that are known to undergo palmitoylation, in addition to Cys 309 , we sought to investigate the influence of palmitoylation on that interaction. Through a detailed series of structure/function along with computational studies, the minimal RBD is defined as a 14-amino acid segment that is organized into an ␣-helical domain, comprising Val 299 -Val 307 , adjacent to the palmitoylated residues at Cys 308 -Cys 311 . It is proposed that the ␣-helical domain corresponds to a putative ␣-helix 8 (␣-H8) within the overall structure of the hIP, a feature originally identified by x-ray crystallography data within rhodopsin and, more recently, in other GPCRs, including the human ␤ 1 adrenergic and the A 2A adenosine receptors (33)(34)(35)(36). Data are presented to suggest that agonist-regulated palmitoylation and/or depalmitoylation (deacylation), in particular at Cys 309 Ͼ Cys 308 , in addition to palmitoylation at Cys 311 , each of which lie outside of ␣-H8, may be important in positioning or anchoring the helical region of the RBD in proximity to Rab11a, to regulate the intracellular trafficking of the hIP in response to activation. Because many GPCRs are predicted to contain a similar ␣-H8 domain, in addition to adjacent palmitoylated residue(s), data generated in this study are likely to have broader functional significance for trafficking of other members of the wider receptor superfamily.

Yeast Two-hybrid Screening and Yeast Matings
Y2H screening of a human kidney cDNA library with the C-tail domain, encoding amino acids 299 -386, of the hIP as specific bait identified Rab11a, expressed in the yeast prey plasmid pACT2:Rab11a, as an interactant of the hIP (32). pGBKT7 and pGBKT7:p53, encoding the GAL4 DNA-binding domain alone or as a fusion with p53, were obtained from Clontech. All yeast protocols were standard procedures as described previously (32).
In brief, all pGBKT7-based bait plasmids were transformed into Saccharomyces cerevisiae AH109 (MATa strain) while pACT2-based prey plasmids were transformed into S. cerevisiae Y187 (MAT␣ strain) and were mated with selection of diploids on synthetic double drop-out (DDO) media (SD/ Leu Ϫ , Trp Ϫ ). Positive interactions between bait and prey proteins were identified by expression of the HIS3, ADE2, and lacZ reporter genes on quadruple drop-out (QDO) media (SD/Leu Ϫ , Trp Ϫ , His Ϫ , Ade Ϫ ) and for the ability to cleave X-gal (5-bromo-4-chloro-3-indolyl-␤-D-galactoside), as measured by the filter lift assay of ␤-galactosidase activity. The scoring system used in the latter was based on the ability of three independent colonies selected from respective DDO media to produce blue (ϩ) or white (Ϫ) colonies due to expression of ␤-galactosidase activity, where, as indicated in the figure legends, ϩϩϩ was used to indicate that cells developed blue color within 30 min of assay, and a minus sign indicates that cells remain white over the period of the assay (4 h).
For analysis of protein expression in S. cerevisiae AH109 (pGBKT7) bait or S. cerevisiae Y187 (pACT2) prey transformants, protein was extracted, resolved by SDS-PAGE, and screened by Western blot analysis using anti-Myc (9B11), with chemiluminescence detection.
For preparation of human platelets, blood was drawn from healthy volunteers, who had not taken any medication for at least 10 days, into 0.15% (v/v) acid/citrate/dextrose anticoagulant (38 mM citric acid, 75 mM sodium citrate, 124 mM D-glucose). Blood was centrifuged at 160 ϫ g for 15 min at room temperature, and platelet-rich plasma was removed and recentrifuged at 160 ϫ g for 10 min to remove contaminating blood cells. Platelets were pelleted from platelet-rich plasma by centrifugation at 720 ϫ g for 10 min at room temperature and resuspended in a modified Tyrode buffer (130 mM NaCl, 10 mM trisodium citrate, 9 mM NaHCO 3 , 6 mM D-glucose, 0.9 mM MgCl 2 , 0.81 mM KH 2 PO 4 , 10 mM Tris, pH 7.4). Platelets were counted using a Sysmex hematology analyzer (TOA Medical Electronics, Kobe, Japan).

Immunoprecipitations
Immunoprecipitations from mammalian HEK 293 cell lines were carried out essentially as described previously (31,32). To assess the effect of 2-bromopalmitate (2-BP) on the interaction between the hIP and Rab11, HEK.hIP WT cells, transiently transfected with pEGFPCI:Rab11a, were incubated ϳ30 h posttransfection with 100 M 2-BP or, as control, with 0.1% DMSO (vehicle) for 16 h at 37°C prior to cicaprost stimulation.
A polyclonal anti-hIP antibody was generated in rabbits immunized with a peptide corresponding to its intracellular domain 2 (residues 121-134) that had been conjugated to keyhole limpet hemocyanin carrier protein. The antiserum was subsequently purified on affinity resin (SulfoLink Coupling gel; Thermo Fisher Scientific) to which the immunizing peptide was coupled. The specificity of the affinity-purified anti-hIP antibody was confirmed by a number of independent approaches, including immunocyto-and immunohistochemistry, immunoprecipitation, and competition of the specific signal in each of those assays using the immunizing peptide. For example, as indicated in supplemental Fig. 1B, the specificity of the anti-hIP antibody was confirmed to specifically immunoprecipitate the hIP from HEK.hIP cells but not from HEK 293 cells; moreover, the preimmune serum did not result in immunoprecipitation of the hIP from HEK.hIP cells (data not shown).
For co-immunoprecipitation of endogenous hIP and Rab11a, washed human platelets (2 ϫ 10 8 platelets/immunoprecipitation) were incubated with 1 M cicaprost or, as control, with drug vehicle. Thereafter, the incubation was stopped by washing the cells or platelets twice in ice-cold PBS followed by lysis in 500 l of radioimmune precipitation buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 1 mM EDTA, 1% Nonidet P-40 (v/v), 0.5% sodium deoxycholate (w/v), 0.1% SDS (w/v), 10 mM sodium fluoride, 25 mM sodium pyrophosphate, 1 mM phenylmethylsulfonyl fluoride, 4 g/ml leupeptin, 2.5 g/ml aprotinin). Lysates were clarified by centrifugation at 14,000 rpm for 5 min, and 50 l was retained for analysis of protein expression in lysates. Following centrifugation at 13,000 ϫ g for 15 min, lysates were subject to immunoprecipitation with the anti-hIP or, as control, with normal rabbit IgG (Santa Cruz Biotechnology, Inc. (Santa Cruz, CA)) at 4°C for 16 h. Thereafter, the lysates were incubated for 1 h with 40 l of a 50% slurry of protein A-Sepharose prior to washing. Immunoprecipitates were resolved by SDS-PAGE on 12.5% gels and immunoblotted with anti-Rab11 (clone 47) and chemiluminescence detection.

Assessment of Agonist-induced Internalization by Confocal Microscopy
Agonist-induced Co-localization of the hIP with Rab11a and Rab5a-In order to monitor changes in cell surface expression of the hIP and/or its co-localization with Rab11a or Rab5a as a function of cicaprost stimulation, HEK.hIP cells were transiently transfected with pADVA (10 g) and pEGFPC1:Rab11a (25 g) or pEGFPC1:Rab5a (25 g), encoding green fluorescent protein (GFP)-tagged forms of the Rab11a or Rab5a protein, respectively, by calcium phosphate/DNA co-precipitation (37). Some 24 h later, cells were seeded onto poly-L-lysine-pretreated coverslips in 6-well plates to achieve 60 -70% confluence following 48 h of incubation at 37°C. Thereafter, cells were washed in serum-free MEM and then prelabeled with anti-HA 101R (1:1000 dilution in MEM) at 4°C for 1 h to label cell surface receptors. Unbound antibody was removed by washing twice with MEM, following which cells were either analyzed immediately (0 h) or were incubated with 1 M cicaprost in MEM for 0 -4 h at 37°C, as indicated in the respective figures. For treatment of cells with acid wash, cells were incubated in low pH acid wash (0.15 M NaCl, HCl, pH 2.5) for 5 min on ice. Thereafter, all cells were washed twice in ice-cold PBS prior to fixation in 3.7% paraformaldehyde, PBS, pH 7.4, for 15 min at room temperature. After washing three times in PBS, cells were permeabilized with 0.2% Triton X-100 in PBS for 10 min on ice, followed by washing three times in TBS. Nonspecific sites were blocked by incubating with Blocking Buffer (1% BSA in TBS), and HA-tagged receptors were immunolabeled with the secondary AlexaFluor594 goat anti-mouse antibody (1:5000, in 1% BSA, TBS). Data presented in the figures are representative images from at least three independent experiments (n Ն 3), from which at least 10 fields were viewed at ϫ63 magnification, where the horizontal bar represents 10 m. For quantification of the co-localization of the hIP or of its various mutants with Rab11a or Rab5a presented in Figs. 3 (A and C), 5 (B and C), and 8C, overlay of pixels in the red and green channels was analyzed using a co-localization analysis plugin available with WCIF ImageJ software (version 1.37c), where quantification is presented as the percentage of co-localized pixels in supplemental Figs. 3, 6, and 10.
Assessment of Agonist-induced Internalization of the hIP by Indirect Immunofluorescence Microscopy-HEK.hIP and HEK. hIP C308S,C309S,C311S cells were grown to 60 -70% confluence on poly-L-lysine-pretreated coverslips in 6-well plates. To monitor the effect of 2-BP on agonist-induced internalization, HEK.hIP cells were treated with 0.1% DMSO (vehicle) or 100 M 2-BP for 16 h at 37°C prior to agonist stimulation. Thereafter, cells were washed in serum-free medium (MEM) and incubated with either vehicle (MEM) or 1 M cicaprost, in MEM, for 2 and 4 h at 37°C. Cells were washed twice in ice-cold PBS prior to fixation in 3.7% paraformaldehyde, PBS, pH 7.4, for 15 min at room temperature. After washing three times in PBS, cells to be permeabilized were incubated with 0.2% Triton X-100 in PBS for 10 min on ice, followed by washing three times in TBS. Nonspecific sites were blocked by incubating cells with Blocking Buffer prior to immunolabeling with anti-HA 101R (1:1000 dilution in Blocking Buffer) for 1 h. Unbound antibody was then removed by washing twice in TBS, followed by detection with the secondary AlexaFluor594 goat anti-mouse antibody (1:5000, in 1% BSA, TBS).
To monitor agonist-induced internalization of endogenous hIP in primary HUVECs, cells were grown to 60 -70% confluence on 0.2% gelatin-pretreated coverslips in 6-well plates and were incubated with 1 M cicaprost for 0 -4 h at 37°C. Thereafter, cells were fixed and were either analyzed directly (nonpermeabilized) or following permeabilization (0.2% Triton X-100 in PBS for 10 min on ice, followed by washing three times in TBS) prior to immunolabeling with anti-hIP antibody (1:500, in 5% nonfat milk, TBS), followed by detection with anti-rabbit AlexaFluor594-conjugated secondary antibody (1:2000, in 5% nonfat milk, TBS). In parallel, nuclei were counterstained with 4Ј,6-diamidino-2-phenylindole (0.5 g/ml, 1 min). To compare agonist-induced internalization of the HA epitope-tagged and non-tagged hIP, previously characterized HEK 293 cell lines stably overexpressing either the HA.hIP or the hIP (27) were plated onto glass coverslips. Some 48 h later, cells were incubated with 1 M cicaprost for 0 -4 h at 37°C and processed for immunofluorescence, as above.
To monitor the co-localization of hIP and hIP C308S,C309S,C311S with the early endosomal marker EEA1 as a function of cicaprost-stimulation, HA-tagged receptors were prelabeled with anti-HA 101R (1:1000 dilution in MEM) at 4°C for 1 h to label cell surface receptors, essentially as described under "Agonist-induced co-localization of the hIP with Rab11a and Rab5a." After removal of unbound antibody, cells were either analyzed immediately (0 h) or incubated with 1 M cicaprost in MEM for 0 -4 h at 37°C, as indicated in the respective figures. Thereafter, cells were fixed and permeabilized prior to immunolabeling with anti-EEA1 antibody (1:500, in 1% BSA, TBS), followed by detection with anti-rabbit AlexaFluor488conjugated secondary antibody (1:2500, in 1% BSA, TBS). HAtagged hIPs were detected with anti-mouse AlexaFluor594conjugated secondary antibody (1:5000, in 1% BSA, TBS), as described under "Agonist-induced co-localization of the hIP with Rab11a and Rab5a." All confocal microscope images were captured as single slices using Carl Zeiss laser scanning system LSM 510 and Zeiss LSM imaging software or using the Zeiss Axioplan 2 microscope and Axioplan Version 4.4 imaging software. Data presented in the figures are representative images from at least three independent experiments (n Ն 3) from which at least 10 fields were viewed at ϫ63 magnification, where the horizontal bar represents 10 m.

Palmitoylation of the hIP
Palmitoylation was carried out essentially as described previously (27,38). Briefly, cells were plated 48 h in advance such that they would achieve a density of ϳ3 ϫ 10 5 cells/10-cm dish (ϳ80% confluence) on the day of metabolic labeling. Cells were washed once in PBS and then metabolically labeled in serumfree minimal essential medium (1.5 ml) containing 1.5 mCi of [ 3 H]palmitic acid (60 Ci/mmol; PerkinElmer Life Sciences) for 4 h in the absence or presence of 1 M cicaprost for 0 -4 h at 37°C. To assess the effect of 2-BP on palmitoylation, cells were preincubated with 100 M 2-BP or, as control, with 0.1% DMSO (vehicle) for 16 h at 37°C prior to metabolic labeling and agonist stimulation. Labeling was terminated by washing the cells in ice-cold PBS, and HA-tagged hIPs were immunoprecipitated as described (27,38). Immunoprecipitates were resolved by SDS-PAGE and electroblotted onto polyvinylidene difluoride membrane. Blots were soaked in Amplify TM for 30 min followed by fluorography using Eastman Kodak Co. X-Omat XAR film for 60 days at Ϫ70°C. In addition, the 3 H fluorograms were subjected to PhosphorImager analysis, and intensities of cicaprost-induced palmitoylation of the hIP were determined and expressed, in arbitrary units, as mean percentage palmitoylation Ϯ S.E. of hIP WT relative to basal levels detected at 0 h. Following fluorographic exposure, polyvinylidene difluoride membranes were screened by immunoblot analysis using the anti-HA 3F10 peroxidase-conjugated antibody followed by chemiluminescence detection.

Computational Structure Predictions
Structure prediction of the hIP and its C-tail domain (hIP 299 -386 ) were generated by online submission to the iterative TASSER (I-TASSER) algorithm, three-dimensional protein structure prediction software that builds models based on multiplethreading consensus target-to-template alignments by LOM-ETS (Local Meta-Threading-Server) and I-TASSER simulations (40,41). I-TASSER predictions were in agreement those generated from independent predictions using PHYRE (available on the World Wide Web) and were further analyzed for helical content using the AGADIR prediction algorithm (available on the World Wide Web). Jpred 3 (42), a web system of different secondary structure prediction algorithms was used to confirm that mutations generated, such as the Ala-scanning mutagenesis, per se did not affect the formation of the ␣-helical domain within the RBD.

Data Analyses
Statistical analyses were carried out using the unpaired Student's t test throughout or, where relevant and specifically indicated, using two-way analysis of variance, employing the GraphPad Prism, version 4.00 package. p values less than or equal to 0.05 were considered to indicate a statistically significant difference. JUNE 11, 2010 • VOLUME 285 • NUMBER 24

Investigation of the Role of Cysteines within the RBD of the hIP in Mediating Its
Interaction with Rab11a-Rab11a was recently identified as a direct interactant of the hIP (32). Consistent with this, immunoprecipitation of endogenous hIP expressed in human platelets confirmed the constitutive association of endogenous Rab11a with the hIP and that this association was increased in response to stimulation with cicaprost, a selective IP agonist (supplemental Fig. 1, A and B). The previous studies suggested that Val 299 -Gln 320 (Fig. 1A), adjacent to TM7 of the hIP, may be critical for its interaction with Rab11a (32). Notably, this 22-residue putative "RBD" contains three Cys residues, where Cys 308 and Cys 311 , but not Cys 309 , have been established to undergo palmitoylation to regulate both hIP-G protein coupling and effector signaling (27). Hence, in the current study, it was first sought to investigate the role of those Cys residues and/or the requirement for their palmitoylation in influencing the interaction between the hIP and Rab11a.
Initially, exploiting the Y2H-based approach to interrogate such protein-protein interactions, the effect of mutation of Cys 308 , Cys 309 , and Cys 311 to corresponding Ser residues, either singly or in combination, on the interaction between the C-tail domain of either the wild type (hIP 299 -386,WT ) or isoprenylation-deficient (hIP 299 -386,SSLC ) hIP was investigated. In agreement with previous studies (32), both the hIP 299 -386,WT and hIP 299 -386,SSLC showed a strong interaction with Rab11a (Fig.  1B). Although each of the bait strains (S. cerevisiae AH109 (pGBKT7)-based) mated successfully with the prey strain (S. cerevisiae Y187 (pACT2:Rab11a)), control diploids either harboring the empty vector pGBKT7 or encoding p53 failed to show any interaction with Rab11a (Fig. 1B). Mutation of Cys 308 and Cys 311 , alone or in combination, did not affect the interaction between either the hIP 299 -386,WT or hIP 299 -386,SSLC and Rab11a (Fig. 1B). Conversely, there was no specific interaction between either hIP 299 -386,WT,C309S or hIP 299 -386,SSLC,C309S and Rab11a, suggesting that Cys 309 is specifically required for such interaction. Furthermore, both wild type and non-isoprenylated forms of hIP 299 -386,C308S,C309S , hIP 299 -386,C309S,C311S , and hIP 299 -386,C308S,C309S,C311S failed to interact with Rab11a (Fig. 1B). Failure to detect an interaction was not due to altered/reduced expression of the various hIP 299 -386,WT or hIP 299 -386,SSLC bait proteins, as confirmed by immunoblot analysis (supplemental Fig. 1D). Collectively, these data suggest that Cys 309 within the C-tail domain of the hIP is critical for its interaction with Rab11a in yeast and that the interaction is largely independent of the isoprenylation status of the hIP.
Effect of Palmitoylation on Co-immunoprecipitation of Rab11a with the hIP-The interaction between the hIP and Rab11 was then examined in the previously characterized mammalian HEK.hIP and HEK.hIP C308S,C309S,C311S cell lines that overexpress HA-tagged forms of the hIP WT and the nonpalmitoylated hIP C308S,C309S,C311S and at nearly equivalent levels (27). Consistent with previous studies (32), Rab11a was detected in the anti-HA immunoprecipitates from HEK.hIP WT cells even in the absence of agonist ( Fig. 2A, top). The IP agonist cicaprost increased Rab11a association with the hIP (Fig. 2A,  top), an effect that was not due to differences in Rab11a expression ( Fig. 2A, bottom) or in hIP immunoprecipitation or expression (supplemental Fig. 2, A and C, respectively). In contrast, Rab11a was not co-immunoprecipitated with the palmitoylation-deficient hIP C308S,C309S,C311S , either in the absence or presence of agonist ( Fig. 2A, top), despite equivalent expression of Rab11a ( Fig. 2A, bottom) and efficient immunoprecipitation and expression of the hIP C308S,C309S,C311S (supplemental Fig. 2, A and C) (27). Moreover, Rab11a was not immunoprecipitated from the control HEK.␤-Gal cells ( Fig. 2A, top).
Although the non-metabolizable palmitate analogue 2-BP, an inhibitor of palmitoylation (43), did not prevent co-immunoprecipitation of Rab11a with the hIP in the absence of agonist, it prevented the cicaprost-induced increase in Rab11a where residues 299 -320, corresponding to the putative RBD are highlighted in gray; palmitoylable Cys residues (Cys 308 , Cys 309 , and Cys 311 ) and the isoprenylated Cys (Cys 383 ) are labeled. B, the S. cerevisiae Y187 (pACT2:Rab11a) prey strain was mated with the respective S. cerevisiae AH109 (pGBKT7: hIP 299 -386,WT )/(pGBKT7:hIP 299 -386,SSLC ) transformed bait strains, as listed, or as controls with S. cerevisiae AH109 (pGBKT7:p53) or S. cerevisiae AH109 (pGBKT7). Diploids were selected on DDO media, whereas interactants were selected on QDO media and by their ability to express ␤-galactosidase due to GAL4-dependent transcriptional activation of the HIS3, ADE2, and lacZ reporter genes, respectively, due to positive interaction between bait and prey proteins. Data presented are representative of at least three independent experiments (n Ն 3). association with the hIP (Fig. 2A, top). In fact, in the presence of 2-BP, there was a decrease in the level of Rab11a in the anti-HA immunoprecipitates, suggesting that agonist stimulation modulates the palmitoylation status of the hIP and its interaction with Rab11a.
Another GPCR shown to have a direct physical interaction with Rab11 is the TP␤ isoform of human TXA 2 receptor (TP), where amino acids 335-344 within its C-tail domain were proposed to be the putative RBD in its case (7). Due to the presence of an established palmitoylatable Cys residue (Cys 347 ) (38) lying in close proximity to the RBD within TP␤, somewhat similar to that found within the hIP, the influence of palmitoylation on the interaction between TP␤ and Rab11a was also examined. Consistent with previous reports (7), Rab11a was co-immunoprecipitated with TP␤ WT in the absence of agonist and following stimulation with the TXA 2 mimetic U46619 ( Fig. 2A, top). Likewise, Rab11a was co-immunoprecipitated with the palmitoylation-deficient TP␤ C347S,C373S,C377S , both in the absence and presence of agonist. Collectively, these data suggest that, in contrast to the hIP, the palmitoylation status of TP␤, such as at Cys 347 adjacent to its RBD, does not modulate the interaction between TP␤ and Rab11a.
To further investigate the influence of palmitoylation on the interaction between the hIP and Rab11a, co-immunoprecipitations were also carried out in the previously characterized HEK. hIP C308S , HEK.hIP C309S , HEK.hIP C311S , HEK.hIP C308S,C309S , HEK.hIP C308S,C311S , and HEK.hIP C309S,C311S cell lines stably expressing HA-tagged forms of the single or double Cys 3 Ser mutations (27). The level of Rab11a in the hIP C308S , hIP C309S , and hIP C311S immunoprecipitates was substantially reduced relative to that of the hIP WT , and there was no cicaprost-in-duced association of Rab11a with either the hIP C308S or hIP C309S , but in the case of the hIP C311S , there was a modest agonist-dependent increase (Fig. 2B, top). Furthermore, Rab11a was not detected in the hIP C308S,C309S , hIP C308S,C311S , and hIP C309S,C311S immunoprecipitates from either non-stimulated or cicaprost-stimulated cells. Lack of Rab11a in the latter immunoprecipitates was not due to its reduced expression in any of the respective HEK 293 cell lines (Fig. 2B, lower  panels), whereas efficient immunoprecipitation and expression of the hIP Cys 3 Ser variants was confirmed in all cases (supplemental Fig. 2, B and C, respectively) (27). It was notable that although each of the hIP variants carrying the Cys 311 3 Ser mutation (e.g. hIP C311S , hIP C308S,C311S , hIP C309S,C311S , and hIP C308S,C309S,C311S ) showed altered patterns of glycosylation (supplemental Fig. 2), this did not lead to their altered maturation or expression at the plasma membrane, as determined by radioligand binding (27) or image analyses (e.g. Figs. 3 and 4), or account for the reduced ability of the palmitoylation-defective hIP C308S,C311S and hIP C308S,C309S,C311S mutants to signal per se, as previously reported by us (27).
Influence of Palmitoylation on Co-localization of the hIP with Rab11a-To further examine the influence of palmitoylation, co-localization between either wild type or palmitoylation-defective hIPs with Rab11a in the respective HEK 293 clonal cell lines was investigated by confocal microscopy. First, only HAtagged hIPs actually located at the cell surface, as opposed to on intracellular membranes, were prelabeled with anti-HA 101R antibody at 4°C, and thereafter, HA-hIP localization was either analyzed directly (0 h) or following stimulation with cicaprost for 2 or 4 h at 37°C. In the absence of agonist, the prelabeled hIP WT was detected exclusively at the cell surface as expected (Fig. 3A, Anti-HA, 0 hr), whereas GFP-Rab11a was predominantly located in the ERC with no co-localization between the hIP WT and Rab11a (Fig. 3A, Overlay, 0 hr). Following 2-h cicaprost treatment, there was substantial relocalization of the hIP WT away from the cell surface to punctate intracellular vesicles concomitantly with relocalization of Rab11a to more discrete endocytic vesicles (Fig. 3A, 2 hr). Furthermore, there was co-localization between the hIP WT and Rab11a (Fig. 3A, Overlay, 2 hr). At 4 h after agonist stimulation, much of the hIP WT had recycled to the cell surface (Fig. 3A, Anti-HA, 4 hr).
Moreover, the Rab11a continued to exhibit ERC and diffuse cytosolic staining, and there was no co-localization of the hIP C308S,C309S,C311S with Rab11a (Fig. 3A, Overlay, 4 hr).
The agonist-induced trafficking of the hIP and of the hIP C308S,C309S,C311S was also examined following acid stripping in order to further distinguish cell surface from internalized receptor. Both the prelabeled hIP and the hIP C308S,C309S,C311S were confirmed to be expressed at the cell surface prior to agonist stimulation, as evidenced by the sensitivity of the fluorescent signal to acid stripping (Fig. 3B, 0 hr). At 2 h after cicaprost stimulation, the prelabeled hIP and hIP C308S,C309S,C311S internalized, being insensitive to acid treatment. However, at 4 h, although much of the internalized wild type hIP had recycled to the plasma membrane, little if any of the hIP C308S,C309S,C311S recycled, as indicated by its sensitivity and resistance to acid stripping, respectively (Fig. 3B). Moreover, consistent with previous data, the hIP, but not the hIP C308S,C309S,C311S , co-localized with GFP-Rab11a in response to agonist stimulation, and that co-localization per se (e.g. at 2 and 4 h) was not sensitive to acid stripping.
Co-localization of GFP-Rab11a was also investigated in HEK.hIP C308S , HEK.hIP C309S , HEK.hIP C311S , HEK. hIP C308S,C309S , HEK.hIP C308S,C311S , and HEK.hIP C309S,C311S cell lines (Fig.  3C). As expected, the prelabeled hIP Cys 3 Ser variants were exclusively detected at the plasma membrane with no co-localization with Rab11a in resting cells (Fig. 3C, 0 hr). Although each of the individual hIP Cys 3 Ser variants underwent cica- prost-induced internalization, albeit to varying extents (Fig. 3C,  2 hr), they displayed distinct differences in terms of their subcellular relocalizations. For example, hIP C308S localized to punctate vesicles to the perinuclear region of the cell reminiscent of the ERC, with co-localization with Rab11 evident (Fig.  3C, hIP C308S , 2 hr). Similarly, hIP C308S,C309S exhibited some relocalization to the perinuclear region (Fig. 3C, hIP C308,309S , 2  hr). After a 4-h cicaprost stimulation, most of the internalized hIP C308S recycled back to the cell surface, whereas a significant amount of hIP C308S,C309S did not recycle but rather remained localized to the ERC (Fig. 3C, 4 hr). The pattern of internalization of the hIP C309S was similar to that of hIP WT , in that it co-localized with Rab11a in punctate vesicles after a 2-h cicaprost stimulation, and after 4 h, it mostly recycled to the cell surface with no evidence of co-localization with Rab11a ( Fig. 3C, hIP C309S , 4 hr). In contrast, the hIP C311S , hIP C308S,C311S , and hIP C309S,C311S behaved similarly to the hIP C308S,C309S,C311S , whereby they each internalized (2 h) but did not recycle to the cell surface, remaining in punctate vesicles (Fig. 3C, 2 and  4 hr). Moreover, similar to the hIP C308S,C309S,C311S , there was no evidence of co-localization between hIP C311S , hIP C308S,C311S , or hIP C309S,C311S and Rab11a. Taken together, these data suggest that mutating Cys 308 , Cys 309 , or Cys 311 , either singly or in combination, did not impair the agonist-induced internalization per se. However, it substantially and differentially altered trafficking of the internalized receptors, reducing their ability to co-localize with Rab11a and, in turn, their recycling to the plasma membrane, such as at 4 h after agonist activation ( Fig. 3 and supplemental Fig. 3). Most notably, mutation of Cys 311 , either alone or in any of the three other possible combinations, rendered forms of the hIP that undergo agonistdependent internalization to discrete endosomes but do not co-localize with Rab11a or recycle to the plasma membrane.
Thereafter, agonist-induced trafficking and co-localization of the HA epitope-tagged hIP and of the non-tagged hIP stably expressed in the previously characterized HEK 293 cell lines (27) was compared in order to exclude the possibility that the presence of the HA tag might impair or alter the ability of the hIP or of its many variants used in this study to undergo internalization or recycling (supplemental Fig. 4A). Cicaprost-induced internalization of the HA-hIP and of the hIP was examined by indirect immunofluorescence microscopy and confocal imaging (single slices) under permeabilizing conditions using a previously characterized antibody directed to the intracellular loop (IC) 2 domain of the hIP (39). Both the HA-hIP and the nontagged hIP underwent similar cicaprost-induced trafficking, with maximal internalization at 2 h, whereas at 4 h, they each recycled to the plasma membrane (supplemental Fig. 4A). Moreover, agonist-induced trafficking of the hIP endogenously expressed in primary HUVECs was also examined and was confirmed to display a pattern somewhat similar to that of the hIP ectopically expressed in the HEK 293 cell lines. More specifically, in primary HUVECs, in the absence of agonist, the hIP  was predominantly expressed at the plasma membrane. However, in response to cicaprost simulation, it underwent a timedependent internalization with loss of cell surface expression coinciding with increased intracellular accumulation, with maximal internalization at 1-2 h after agonist stimulation, whereas at 4 h, much of it recycled to the plasma membrane (supplemental Fig. 4B). In keeping with the fact that the anti-hIP antibody was directed to an intracellular epitope (IC 2 ), no immunostaining was observed in non-permeabilized cells (supplemental Fig. 4B).
Influence of Palmitoylation on the Internalization and Association of the hIP with Rab5a-To further confirm that the lack of association and co-localization of the palmitoylation-deficient hIP variants with Rab11a is not due to impaired agonistinduced internalization per se or indeed an artifact of overexpression of GFP-Rab11a, their cellular localization/trafficking was also examined by confocal imaging under both non-permeabilizing and permeabilizing conditions in the presence of endogenous Rab proteins only. Initially, the hIP WT and hIP C308S,C309S,C311S were predominantly located at the cell surface (Fig. 4A, 0 hr, Non-Permeabilized) with some evidence of intracellular localization, albeit at much lower levels (Fig. 4, 0 hr, Permeabilized). Upon agonist stimulation, cell surface expression of both the hIP WT and hIP C308S,C309S,C311S was lost (Fig. 4A, 2 hr, Non-Permeabilized), concomitant with their increased intracellular localization (Fig. 4A, 2 hr, Permeabilized) and co-localization with the endosomal marker EEA1 (Fig. 4B). However, although much of the hIP WT recycled to the cell surface following prolonged (4-h) cicaprost stimulation, most of the hIP C308S,C309S,C311S remained internalized with no evidence of recycling to the plasma membrane (Fig. 4A, 4 hr, Non-Permeabilized) and showed prolonged co-localization with EEA1 (Fig. 4B). Moreover, in the presence of 2-BP, the hIP WT internalized to punctate vesicles in response to cicaprost but failed to efficiently recycle to the cell surface following prolonged agonist stimulation (Fig. 4A, 2 hr and 4 hr, Non-Permeabilized). Similarly, and consistent with data in Fig. 3C, although each of the individual hIP Cys 3 Ser variants underwent cicaprost-induced internalization, albeit to varying extents, all variants carrying the Cys 311 mutation failed to recycle to the plasma membrane (supplemental Fig. 5).
As stated, previous studies have established that the hIP undergoes agonist-induced internalization through a Rab5adependent mechanism (31). Therefore, in view of findings herein with Rab11a, we also sought to examine whether palmitoylation may affect the actual association, be it direct or indirect, between the hIP and Rab5. Consistent with previous data (31), Rab5a was detected in the anti-HA immunoprecipitates from HEK.hIP WT cells, and levels were increased in response to cicaprost (Fig. 5A). In contrast to that observed with Rab11a, Rab5a co-immunoprecipitated with the palmitoylation-deficient hIP C308S,C309S,C311S and in an agonist-augmented manner (Fig. 5A).
Thereafter, the possible influence of palmitoylation on the actual agonist-induced co-localization of the hIP with Rab5a was investigated by confocal imaging (Fig. 5, B and C, and supplemental Fig. 6A). In the absence of agonist, the prelabeled hIP WT was detected exclusively at the cell surface, whereas GFP-Rab5a exhibited diffuse staining within the cytosol with some evidence of association with preformed endosomes (Fig.  5B, 0 hr). Following 2 h of cicaprost-treatment, there was substantial relocalization of the hIP WT to intracellular vesicles reminiscent of early endosomes (Fig. 5B, Anti-HA, 2 hr). Concomitant with this, Rab5a associated with more discrete endosomes (Fig. 5B, GFP, 2 hr), and there was significant co-localization between the hIP WT and Rab5a (Fig. 5B, Overlay, 2 hr). At 4 h, much of the hIP WT recycled to the cell surface, with the remainder co-localizing with GFP-Rab5a in endosomes at or near the cell periphery (Fig. 5B, 4 hr). Similar to the hIP WT , the prelabeled hIP C308S,C309S,C311S was exclusively detected at the plasma membrane with no co-localization between Rab5a in resting cells (Fig. 5C, 0 hr). At 2 h after cicaprost stimulation, much of the hIP C308S,C309S,C311S underwent internalization to Rab5a-positive endosomes, but at 4 h, most of it failed to recycle to the plasma membrane, remaining in those endosomes (Fig. 5C, 2 hr and 4 hr). Consistent with this, the prelabeled cell surface hIP C308S , hIP C309S , hIP C311S , hIP C308S,C309S , hIP C308S,C311S , and hIP C309S,C311S also co-localized with Rab5 in response to cicaprost (supplemental Fig. 6, B and C, 2 hr), which, in cells expressing the Cys 311 3 Ser mutation either alone or in combination (hIP C311S , hIP C308S,C311S , and hIP C309S,C311S ) failed to recycle, remaining in Rab5-positive endosomes (supplemental Fig. 6, B and C, 4 hr). Taken together, these data indicate that although palmitoylation of the hIP has a role in regulating its interaction with Rab11a, it does not affect its agonist-induced internalization or association with Rab5a. Effect of Agonist Stimulation on hIP Palmitoylation-Although the hIP was established to undergo palmitoylation at Cys 308 and Cys 311 (27), that study did not investigate possible agonist-induced changes in its palmitoylation status per se. Furthermore, although 2-BP did not prevent the interaction between the hIP and Rab11 in the absence of agonist, it significantly impaired the cicaprost-enhanced effect ( Fig. 2A). This suggests that palmitoylation of the hIP is modulated by agonist and that such changes in its palmitoylation status may alter its ability to interact with Rab11. Therefore, to determine whether palmitoylation of the hIP may actually be modulated by cicaprost, HEK.hIP cells were metabolically labeled with [ 3 H]palmitate in the absence or presence of cicaprost (1 M) for 0 -4 h (Fig. 6A). Consistent with previous reports (27), palmitoylation of the hIP was readily detected in the absence of agonist, as evidenced by the presence of a broad radiolabeled band between 47.5 and 62 kDa in the immunoprecipitates from HEK.hIP cells (Fig. 6A, 0 hr). Although there were slight reductions in palmitoylation of the hIP in response to cicaprost at 30 min and 1 h (p ϭ 0.001 and p ϭ 0.05, respectively), there was no net change in the overall level of palmitoylation over the labeling period (Fig. 6, A and B, 0 -4 h; analysis of variance, p ϭ 0.23). In contrast, no metabolic labeling was associated with the immunoprecipitates from HEK.hIP C308S,C309S,C311S or from HEK.hIP WT cells preincubated with 2-BP (Fig. 6A), confirming that the hIP C308S,C309S,C311S does not undergo palmitoylation and that palmitoylation of the hIP is a dynamic process being efficiently inhibited by 2-BP. In all cases, confirmation of equivalent protein expression and efficient recovery of the hIP and hIP C308S,C309S,C311S immunoprecipitates was obtained by subsequent screening of the 3 H fluorogram by immunoblot analysis (anti-HA-hIP; supplemental Fig. 7), which also indicated that the lack of detection of palmitoylation of the hIP C308S,C309S,C311S was not due to failure of its immunoprecipitation.
To further examine the role of palmitoylation of the hIP in regulating its interaction with Rab11, [ 3 H]palmitate labeling of the hIP C308S , hIP C309S , hIP C311S , hIP C308S,C309S , hIP C308S,C311S , and hIP C309S,C311S was also investigated (Fig. 6, C and D, 0 hr  and 2 hr). In the absence of agonist, palmitoylation of the hIP C308S , hIP C309S , and hIP C308S,C309S was readily detected, albeit at reduced levels compared with the hIP WT (Fig. 6, C and  D, 0 hr), whereas levels of palmitoylation of the hIP C311S , hIP C308S,C311S , and hIP C309S,C311S cells were even lower (Fig. 6,  C and D, 0 hr). The fact that the hIP C308S,C311S , but not hIP C308S,C309S,C311S , underwent palmitoylation suggests that Cys 309 is subject to palmitoylation, albeit at low levels, under resting conditions. Cicaprost did not mediate a significant change in palmitoylation of the hIP C308S but led to slight reductions in palmitoylation of the hIP C309S and hIP C308S,C309S (Fig.  6, C and D, 2 hr). In contrast, agonist stimulation resulted in an 80% reduction in palmitoylation of the hIP C311S (Fig. 6, C and D, 2 hr). Furthermore, some 50 and 85% reductions in palmitoylation of the hIP C309S,C311S and hIP C308S,C311S were observed, respectively (Fig. 6, C and D, 2 hr). These data suggest that Cys 309 more than Cys 308 and, to a much lesser extent, Cys 311 undergo depalmitoylation/deacylation as a consequence of agonist stimulation. As previously, confirmation of equivalent protein expression and efficient immunoprecipitation of the hIP and its Cys 3 Ser variants was established by screening of the 3 H fluorogram by immunoblot analysis (anti-HA-hIP; supplemental Fig. 7).
Taken together, these data further confirm that the hIP WT is constitutively palmitoylated and does not appear to undergo overall net changes in its palmitoylation status in response to agonist stimulation (Fig. 6, A and B). However, more detailed analysis of individual Cys residues reveals that palmitoylation of the hIP is indeed regulated by agonist and to varying degrees depending on the particular Cys residue (Fig. 6, C and D). Cys 311 is the residue subject to the majority of palmitoylation, which is not largely affected by agonist, but both Cys 308 and Cys 309 are subject to a low level of palmitoylation, which is substantially reduced in response to agonist, in particular in the case of Cys 309 . The finding that palmitoylation of Cys 309 was only detected in the absence of agonist and that, following cicaprost-treatment, it is fully deacylated is in fact consistent with previous studies where palmitoylation of the hIP was only examined following 2 h of cicaprost stimulation and it was found that Cys 309 was not palmitoylated under those conditions (27). Although data herein cannot fully exclude the possibility that Cys 309 is only palmitoylated when the other two palmitoylation sites at Cys 308 and Cys 311 are mutated to the equivalent Cys 308 3 Ser and Cys 311 3 Ser, respectively, the fact that the hIP C308S,C311S mutant undergoes agonist-regulated palmitoylation, being palmitoylated in the absence of agonist but not following 2-h stimulation with cicaprost, suggests that it is unlikely that palmitoylation at Cys 309 is an artifact of disruption of palmitoylation at the two other sites (Cys 308 and/or Cys 311 ). Additionally, although not reaching statistical significance over four individual metabolic labeling experiments, the level of palmitoylation of the hIP C309S was consistently lower than in the wild type hIP, further suggesting that Cys 309 is palmitoylated independently of Cys 308 or Cys 311 .
Identification of the Domain(s) within the hIP Mediating the Interaction with Rab11a-As stated, the putative RBD region within the hIP required for its interaction with Rab11a was localized to a 22-amino acid segment, corresponding to Val 299 -Gln 320 , adjacent to TM7 (Fig. 1A) (32). I-TASSER (40,41) was used to generate a three-dimensional model of the hIP (Fig. 7A) and was found to have greatest structural similarity with that recently determined for the human A 2A adenosine receptor (Protein Data Bank entry 3EML (34)). In addition to its expected 7TM ␣-helical bundle, the hIP was also predicted to contain an eighth ␣-helical domain, where residues 299 -307 of the putative RBD are actually engaged in the ␣-H8 domain adjacent to the palmitoylation residues between Cys 308 and Cys 311 (Fig. 7A). In light of this, we sought to further define the minimal structural domain(s) within the hIP 299 -320 region that mediates the interaction with Rab11a through additional deletional analysis and Y2H-based protein interactions. Consistent with the previous study (32), the hIP 299 -386 and hIP 299 -320 subfragments supported the interaction (Fig. 7B). Conversely, removal of the 299 -320 region or successive amino-terminal deletion of residues within the ␣-H8 or palmitoylation domain, as in the case of the hIP 320 -386 , hIP 303-386 , hIP 307-386 , or hIP 312-386 , respectively, failed to support the interaction with Rab11a (Fig. 7B). Amino-terminal removal of any residues within the ␣-H8 domain, as in the cases of hIP 303-316 and hIP 303-320 , impaired the hIP-Rab11 interaction, whereas carboxyl-terminal deletions, as exemplified by hIP 299 -316 and hIP 299 -312 , established that residues between Val 299 and Leu 312 are sufficient for the interaction, whereas residues Gly 313 -Gln 320 are not actually required. Hence, the minimum region required for interaction with Rab11a is the 14-residue hIP 299 -312 , predicted to be engaged in the ␣-H8 domain adjacent to the palmitoylated residues at Cys 308 -Cys 311 (Fig. 7A).
Disruption of hIP and Rab11a Interaction through Ala-scanning Mutagenesis-Ala-scanning mutagenesis, in combination with the Y2H screening approach, was used to further define the critical residue(s) within the minimum hIP 299 -312 RBD required for interaction with Rab11a, where the mutations introduced were confirmed not to affect the formation of the ␣-H8 structure per se (42). Consistent with previous data involving the Cys 3 Ser mutations (Fig. 1B), both the hIP 299 -386,C308A and hIP 299 -386,C311A , but not hIP 299 -386,C309A , showed positive interaction with Rab11a, further confirming a specific requirement for Cys 309 for that interaction (Fig. 8A). Moreover, Ala-scanning mutagenesis of residues Phe 300 , Arg 302 , Leu 303 , Lys 304 , Leu 305 , and Val 307 disrupted the interaction between the hIP 299 -386 with Rab11a, whereas mutation of Val 299 , Gln 301 , Trp 306 , Leu 310 , and Leu 312 had no measurable affect (Fig. 8A). Hence, it appears that there is a structural requirement for numerous hydrophobic in addition to positively charged Lys and Arg residues between Val 299 and Leu 312 rather than one or more residue(s) being responsible for the interaction between the hIP and Rab11a. This is consistent with the specific requirement for the predicted ␣-H8 domain adjacent to the palmitoylated residues at Cys 308 -Cys 311 within the hIP RBD as necessary for Rab11a binding.
To confirm these data in mammalian cells, Ala-scanning mutagenesis of Val 299 -Leu 312 was also performed on HA-tagged forms of the hIP to allow for assessment of their interaction with Rab11a through co-immunoprecipitation (Fig. 8B) and immunolocalization (Fig. 8C) studies. Initially, it was necessary to compare expression and signaling of the wild type hIP with that of each of its Ala-scanning variants (Val 299 -Leu 312 ) and indeed FIGURE 7. Structure and function analysis of the RBD of the hIP. A, I-TASSER three-dimensional structural analysis of the hIP predicts that it contains seven TM domains, typical of its GPCR structure, and that the Val 299 -Gln 320 region (inset) contains an eighth ␣helical domain between residues Val 299 and Val 307 adjacent to the palmitoylation residues Cys 308 -Cys 311 . B, the S. cerevisiae Y187 (pACT2:Rab11a) prey strain was mated with the respective S. cerevisiae AH109 (pGBKT7:hIP 299 -386,WT ) bait strains, as listed, or as controls with S. cerevisiae AH109 (pGBKT7:p53) or S. cerevisiae AH109 (pGBKT7). Diploids were selected on DDO media, whereas interactants were selected on QDO media and by their ability to express ␤-galactosidase due to positive interaction between the bait and prey proteins. n Ն 3. of the Cys 3 Ser variants following their transient expression in HEK 293 cells (supplemental Fig. 8, A-C). The HA-tagged hIP and each of its variants (14 Val 299 -Leu 312 Ala-scanning and seven Cys 3 Ser mutants) showed equivalent expression following transient transfection in HEK 293 cells (supplemental Fig. 8A). Moreover, with the exception of hIP R302A and hIP K304A , each of the variant hIPs showed similar patterns of expression when analyzed by immunofluorescence microscopy under non-permeabilizing and permeabilizing conditions. In the case of the hIP R302A and hIP K304A harboring the R302A and K304A mutations, respectively, despite their equivalent overall expression (supplemental Fig. 8, A and B (Permeabilized)), these receptors failed to traffic to the plasma membrane (supplemental Fig. 8B, Non-Permeabilized). The reason for the failure of the latter variants to mature and traffic to the plasma membrane is unclear and will be the subject of a separate investigation, being beyond the scope of this study. In terms of signaling, as indicated in supplemental Fig. 8C, each of the Cys 3 Ser and Cys 3 Ala variants showed levels of cicaprost-induced cAMP generation similar to that of the hIP with the exception of the hIP C308S,C311S and the hIP C308S,C309S,C311S , which, consistent with our previous report (27), failed to generate substantial increases in cAMP in response to agonist. Moreover, in keeping with previous imaging data (supplemental Fig. 8B), with the exception of the hIP R302A and hIP K304A , which failed to mature/traffic to the plasma membrane, each of the remaining Ala-scanning hIP variants showed cicaprost-induced cAMP generation similar to that of the wild type hIP (supplemental Fig. 8C).
Consistent with the yeast data, there was significant impairment in the ability of the hIP to interact with Rab11a when certain residues between Val 299 and Leu 312 were mutated. For example, despite showing efficient and equivalent expression of Rab11 (Fig. 8B, bottom) and hIP expression and immunoprecipitation (supplemental Fig. 9) in all cells, there were substantially altered or varying levels of Rab11a in the anti-HA immunoprecipitates of many of the mutated hIP 299 -312 Val 3 Ala variants in the absence of agonist (Fig. 8B). More specifically and more revealing, mutation of certain residues disrupted the increased agonist-induced association of Rab11a with certain hIP variants (hIP F300A , hIP R302A , hIP L303A , hIP K304A , hIP L305A , hIP V307A , hIP C308A , hIP C309A , and hIP L312A ), whereas other mutations (hIP V299A , hIP Q301A , hIP W306A , hIP L310A , and hIP C311A ) maintained the cicaprost-induced immunoprecipitation of Rab11a somewhat similar to the wild type hIP (Fig. 8B, upper panels). Moreover and consistent with the previous co- FIGURE 8. Alanine-scanning mutagenesis of hIP 299 -312 . A, the S. cerevisiae Y187 (pACT2:Rab11a) prey strain was mated with the respective S. cerevisiae AH109 (pGBKT7: hIP 299 -386,WT ) bait strains harboring the various Ala-scanning mutations of residues Val 299 -Leu 312 , as listed, or as controls with S. cerevisiae AH109 (pGBKT7:p53) or S. cerevisiae AH109 (pGBKT7). Diploids were selected on DDO medium, whereas interactants were selected on QDO medium and by their ability to express ␤-galactosidase due to positive interaction between the bait and prey proteins (n Ն 3). B, HEK 293 cells, co-transfected with plasmids encoding GFP-tagged Rab11a and HA-tagged hIP WT , hIP V299A , hIP F300A , hIP Q301A , hIP R302A , hIP L303A , hIP K304A , hIP L305A , hIP W306A , hIP V307A , hIP C308A , hIP C309A , hIP L310A , hIP C311A , or hIP L312A were incubated with either vehicle (Ϫ) or with 1 M cicaprost for 2 h (ϩ), as indicated, prior to immunoprecipitation with anti-HA 101R antibody. Immunoprecipitates (IP) were resolved by SDS-PAGE and were immunoblotted (IB) with anti-Rab11 antibody (top), or with anti-HA 3F10-horseradish peroxidase antibody (supplemental Fig. 5). To verify expression of the Rab11 proteins, aliquots of whole cell lysates (ϳ50 g/lane) were also immunoblotted with anti-Rab11 antibody (bottom). The relative positions of the molecular size markers (kDa) are indicated to the left of the panels. n Ն 3. C, HEK.hIP WT , HEK.hIP V299A , HEK.hIP F300A , HEK.hIP Q301A , and HEK.hIP L303A cells, each transiently transfected with pEGFPCI:Rab11a, were prelabeled with anti-HA 101R antibody for 1 h at 4°C; thereafter, cells were either analyzed directly (0 h) or were incubated with 1 M cicaprost at 37°C for 2 or 4 h. Cells were fixed and permeabilized prior to detection of HA-tagged hIPs with anti-mouse AlexaFluor594conjugated secondary antibody and enhanced GFP-Rab11 or both (overlay). Only the overlay images are presented where n Ն 3.
immunoprecipitations carried out with the Cys 3 Ser mutations (hIP C308S , hIP C309S , and hIP C311S ; Fig. 2B), the hIP C308A , hIP C309A , and hIP C311A also showed impaired ability to co-immunoprecipitate Rab11a in the absence of agonist (Fig. 8B, top), but similar to the hIP C311S , the hIP C311A retained the agonistinduced responsiveness, whereas the Cys 308 (hIP C308S and hIP C308A ) and Cys 309 (hIP C309S and hIP C309A ) equivalents did not.
In agreement with the co-immunoprecipitation data, confocal imaging established that in general, mutation of many of the residues within the Val 299 -Leu 312 RBD domain resulted in impaired co-localization of the internalized receptor to Rab11a-positive vesicles and altered intracellular trafficking relative to the wild type hIP (Fig. 8C and supplemental Fig. 10). For example, although the hIP F300A internalized normally to discrete intracellular vesicles, there was reduced co-localization to Rab11a-positive vesicles (Fig. 8C, F300A, 0 hr and 2 hr). Moreover, following 4 h of agonist stimulation, the hIP F300A did not efficiently recycle to the cell surface compared with the hIP WT (Fig. 8C, F300A, 4 hr). On the other hand, both the hIP V299A and the hIP Q301A behaved similarly to the hIP WT , whereby they internalized to discrete vesicles co-localizing with Rab11a after 2 h of cicaprost stimulation. Furthermore, they each efficiently recycled to the cell surface 4 h poststimulation (Fig. 8C, V299A and Q301A, 0, 2, and 4 hr). In contrast, although the prelabeled hIP L303A internalized in response to agonist (2 h), it remained in distinct vesicles, failing to recycle following prolonged exposure to agonist (Fig. 8C, L303A, 0, 2,  and 4 hr). Hence, although there is evidence to support an association between the hIP L303A with Rab11, the co-localization data ( Fig. 8C and supplemental Fig. 10) suggest that there is a significant impairment in its intracellular trafficking.
Taken together, the hIP contains a RBD, located between residues Val 299 and Leu 312 and adjacent to TM7, that plays a critical role in mediating the direct interaction with Rab11a and in regulating its intracellular trafficking after agonist stimulation. Computational structural predictions along with complementary biochemical approaches in yeast and mammalian cells support the proposal that the 14-amino acid RBD domain is organized into an ␣-H8 domain, comprising residues Val 299 -Val 307 , adjacent to the palmitoylated residues at Cys 308 -Cys 311 . Palmitoylation at Cys 311 in addition to agonist-regulated deacylation at Cys 309 Ͼ Cys 308 , outside of the ␣-helical domain, are likely to be important in positioning or anchoring the helical RBD of the hIP in proximity to Rab11a, to regulate its intracellular traffic after agonist stimulation.

DISCUSSION
Agonist-induced trafficking plays a critical role in regulating the signaling responses by members of the GPCR superfamily (2,4). A number of independent studies have established that the IP undergoes desensitization and internalization in human platelets and other cell types, fine tuning the responses to prostacyclin in vivo, but that it does not involve the classic G protein-coupled receptor kinase/␤-arrestin-mediated mechanism (31,44). We have recently established that agonist-induced internalization of the hIP occurs through a Rab5a-dependent mechanism (31). Through a subsequent Y2H screen, Rab11a was identified as a direct and specific interactant of the hIP and was found to play a critical role in its agonist-induced trafficking, mainly through the slow or late endosome recycling pathway (32). Furthermore, the interaction with Rab11a was dependent on a 22-amino acid hydrophobic sequence within the C-tail domain of the hIP, adjacent to TM7 and bearing certain structural similarity to the RBD first identified within members of the FIPs/Rab11 effector proteins (5). The current study confirmed that the interaction between the hIP and Rab11a occurs in human platelets and is regulated in response to agonist.
Palmitoylation or, more correctly, thio(S)-acylation, is a post-translational modification that occurs through attachment of C16 palmitate to Cys residue(s) via labile thioester bonds and is thought to regulate the association of proteins or domains thereof with membranes in a dynamic or transient manner (27,45,46). A large number of GPCRs contain at least one palmitoylable Cys typically located some 10 -14 amino acids downstream of the ␣-helical TM7 domain within the C-tail region (47). It has been suggested that regulated cycles of palmitoylation/depalmitoylation, such as in response to agonist, may result in the formation of a transient fourth intracellular loop (IC 4 ) to regulate receptor-G protein coupling, downregulation, and/or internalization, to list but a few processes. More specifically, in the case of the hIP, in addition to its established isoprenylation (24,25), it undergoes palmitoylation at Cys 308 and Cys 311 in response to stimulation with the selective IP agonist cicaprost (2 h), whereas Cys 309 was found not to be palmitoylated under those conditions (27). Because the proposed RBD contains Cys 308 and Cys 311 that are known to undergo palmitoylation, in addition to Cys 309 , the current study sought to investigate the role of those Cys residue(s) and the possible influence of palmitoylation on the interaction between the hIP and Rab11a and to identify the key structural determinants of the RBD.
Based on the experimental and computational data presented herein, an overall model is presented in Fig. 9 to explain the direct and specific interaction of the hIP with Rab11a in regulating its intracellular trafficking following agonist stimulation. It is proposed that a major part of the RBD (Val 299 -Leu 307 ) overlaps with the putative ␣-H8 domain located within the cytoplasmic C-tail region, in proximity to TM7, and adjacent to the palmitoylated residues at Cys 308 -Cys 311 . It is noteworthy that the ␣-H8 equivalents of several other GPCRs, including that of rhodopsin and the A 2A adenosine receptor, have been proposed to lie perpendicular to the TM bundle in the presence of an interface, such as when associated in a lipid or micelle environment, but may adopt a looplike or disordered structure in an aqueous environment or in the conformationally active receptor (35,48,49). Moreover, the putative ␣-H8 domains in many of those receptors, as exemplified by rhodopsin or the ␤ 2 adrenergic receptor, contain adjacent palmitoylated Cys residues (35,36). As outlined in Fig. 9C, it is proposed that Cys 311 is the major site of palmitoylation either in the absence or presence of agonist and is required to anchor or orientate the ␣-H8 domain of the hIP for interaction with Rab11a. Moreover, cicaprost-regulated palmitoylation, particularly deacylation at Cys 309 Ͼ Cys 308 , may allow for agonist-induced conformational flexibility within ␣-H8, as proposed for other GPCR equivalents, to allow for optimal agonist-dependent interaction with Rab11a. Consistent with the proposed requirement and role for palmitoylation of Cys 311 , it is the most abundantly palmitoylated residue, and all variants carrying that hIP C311S mutation, either alone or in combination, internalize but remain in endocytic vesicles and do not recycle or co-localize with Rab11a. Consistent with the proposed requirement and role for palmitoylation of Cys 308 and Cys 309 to regulate the agonist-induced association with Rab11a, these are the residues that are most sensitive to agonist-induced deacylation and, in the case of the hIP C311S or hIP C311A variants, which retain both Cys 308 and Cys 309 , still show a limited agonist-induced association/immunoprecipitation with Rab11a. Notably also, both the single hIP C308S and hIP C309S undergo recycling, somewhat similar to that of the hIP WT , whereas the double hIP C308S,C309S does not, suggesting that palmitoylation at either residue at any given time may be required. Moreover, studies in yeast involving the hIP C309S and hIP C309A highlight a critical role for Cys 309 for the interaction with Rab11a. It is proposed that failure of the Y2H studies to identify the critical role for Cys 308 , as in the case of the single hIP C308S or hIP C308A , for interaction with Rab11a is likely to be due to compensatory palmitoylation at the adjacent Cys 309 residue and/or due to the preferred palmitoylation of FIGURE 9. Proposed model of the structure of the RBD within the hIP and of its interaction with Rab11. A, I-TASSER modeling of the hIP depicting the ␣-H8 domain (Lys 297 -Val 207 ) and neighboring sequence, including TM2, -5, and -7 domains. B, side view of region after 180°rotation about the vertical axis. C, model of the putative RBD of the hIP, where residues Val 299 -Val 207 form an essential part of the ␣-H8 domain adjacent to the palmitoylated Cys 308 , Cys 309 , and Cys 311 . Amino acid side chains are not shown. The transparent blue arch superimposed on the structure depicts Rab11a, or part thereof, interacting with essential residues of the RBD. It is proposed that Cys 311 , the major site of palmitoylation, both in the absence or presence of agonist, anchors or orientates the ␣-H8 domain to the plasma membrane, whereas cicaprost-regulated palmitoylation, particularly at Cys 309 Ͼ Cys 308 , may allow for agonist-induced conformational flexibility within the ␣-H8 domain, allowing for optimal agonist-dependent interaction with Rab11a. Residues found to be important for interaction with Rab11a predominantly lie on one face of the ␣-H8 domain, as depicted by the orientation of the side chains in A and B.
Cys 309 , as opposed to Cys 308 , by the yeast palmitoyl transferase. Moreover, failure of the Y2H studies to identify the critical role for Cys 311 , as in the case of the single hIP C311S or hIP C311A , is likely to be due to fact that the proposed membrane association role of the palmitoyl-Cys 311 would not be required, or indeed be permissible, for the Y2H "bait" versus "prey" interaction and resulting GAL4-dependent transcription to occur (supplemental Fig. 1C). Despite showing normal maturation/expression at the plasma membrane, as evidenced by both radioligand binding data (27) and extensive immunolocalization data herein, each of the hIP variants harboring the C311S mutation stably expressed in HEK 293 cells showed altered or delayed patterns of glycosylation. Hence, because the hIP C308S,C311S and hIP C308S,C309S,C311S , either stably or transiently expressed (27) (this work), did not display normal agonist-induced signaling, despite normal radioligand binding, expression at the plasma membrane, and agonist-induced internalization, the possibility that their altered recyclization/co-localization with Rab11a may be due to non-functional receptors, as opposed to altered palmitoylation, cannot be fully excluded. However, the fact that the hIP C311S and hIP C309S,C311S variants actually displayed normal signaling but, similar to hIP C308S,C311S and hIP C308S,C309S,C311S , did not show co-localization with Rab11a or recyclization in response to agonist stimulation makes this less likely. Hence, these data strongly suggest that the impaired Rab11a-mediated recyclization of all of the C311S variants is due to impaired palmitoylation as opposed to altered/delayed glycosylation. It will be of interest to determine why the hIP C308S,C309S,C311S undergoes agonist-induced internalization and association with Rab5a ( Fig. 5) but does not undergo subsequent recyclization or association with Rab11a or agonistmediated intracellular signaling through G s or G q (27) (this work).
Thus, although residues within the palmitoylation domain are critical for the agonist-induced Rab11a-mediated trafficking of the hIP, such as by orientating and/or providing possible flexibility to the RBD, it is proposed that hydrophobic residues mainly orientated on one face of the ␣-H8 domain provide the binding surface to which Rab11a itself may actually bind (Fig. 9,  A and B). Although studies in yeast and co-immunoprecipitations in HEK 293 cells suggest that positively charged Arg 302 and Lys 304 residues may also be important in influencing the interaction between the ␣-H8 with Rab11a, this is inconclusive due to the fact that the variant hIP R302A and hIP K304A did not display normal maturation/expression at the plasma membrane (supplemental Fig. 8B). The reason for the failure of the latter hIP R302A and hIP K304A variants to mature and traffic to the plasma membrane is unclear and will be the subject of a separate investigation, being beyond the scope of the current study.
The putative ␣-H8 has been proposed to act as a conformational switch between the active and inactive states of certain other GPCRs (35,36). Moreover, peptides based on that domain from several GPCRs have been proposed both to be conformationally flexible, depending on the solvent environment (35,36), and to inhibit signaling by rhodopsin (50), the ␤ 2 adrenergic receptor (51), and the cannabinoid 1 receptor (52), and to activate G i and G o proteins (53). To our knowledge, data presented in this study represent the first demonstration of a direct interaction between the ␣-H8 of any GPCR with Rab11a. Given the conservation of the ␣-H8 along with the presence of adjacent palmitoylatable Cys residue(s), it is indeed formally possible that such an interaction may also occur between other GPCRs and/or other members of the Rab11 or wider Rab superfamily. Indeed, the first GPCR proposed to have a direct interaction with Rab11 is the TP␤ isoform of human TXA 2 receptor, where amino acids 335-344 within its C-tail were proposed to be the critical RBD therein (7). Due to the presence of Cys 347 , a residue established to be palmitoylated (38), in close proximity to the RBD within TP␤, somewhat similar to that found within the RBD of the hIP, the role of palmitoylation in influencing the interaction between TP␤ and Rab11a was investigated herein but was found not to be significant in modulating that interaction. However, a notable difference between the RBD within the hIP and that of the TP␤ is that the region of interaction in the latter receptor, namely residues 335-344 (7), does not actually map to the ␣-H8 but rather to an additional putative ␣-helical 9 domain within its more distal C-tail region.
The identification of the hIP adds to a limited list of proteins that directly interact with Rab11 members, which also includes certain other members of the GPCR superfamily and the ␣5␤1 member of the integrin family (7-9, 12, 32, 54, 55). The actual region within Rab11a itself that mediates the interaction with the hIP remains to be identified and will be the subject of a separate detailed investigation. Moreover, the identity of the specific Rab11 effector(s), be they members of the FIPs, Rabphilin-11/RabllBP, myosin Vb, phosphoinositide 4-kinase ␤, or Sec15 (6, 56 -60), that direct the trafficking of the hIP after Rab11 activation remains to be investigated. Given the exquisite relationship that exists between certain Rab11 members and their effectors, as exemplified by Rab25 and Rab-coupling protein in coordinating the agonist-dependent recycling of the ␣5␤1 integrin (54), or Rip11 in regulating insulin-dependent GLUT 4 transport in adipocytes (61), the identification of the Rab11 effector(s) involved in the trafficking of the hIP is likely to add significantly to the understanding of the true physiologic significance of the direct interaction between the hIP and Rab11.