Abstract
The 5-hydroxytryptamine 2A receptor (5-HT2AR) is a member of the class I family of rhodopsin-related G protein-coupled receptors. The receptor is known to activate phospholipase C via the heterotrimeric G proteins Gq/11, but we showed previously that it can also signal through the phospholipase D (PLD) pathway in an ADP-ribosylation factor (ARF)-dependent manner that seems to be independent of Gq/11 (Mitchell et al., 1998). Both coimmunoprecipitation experiments and the effects of negative mutant ARF constructs on 5-HT2AR-induced PLD activation here suggested that ARF1 may play a greater role than ARF6 in the function of this receptor. Furthermore, we demonstrated using glutathione S-transferase (GST)-fusion proteins of receptor domains that ARF1 and ARF6 bind to the third intracellular loop (i3) and the carboxy terminal tail (ct) of the 5-HT2AR. The association of ARF1 with the ct domain of the receptor was stronger than its interaction with i3, or the interactions of ARF6 with either construct. Experiments using ARF mutants that are deficient in GTP loading, and the in vitro addition of GTPγS suggested that GTP loading enhances ARF1 binding to the receptor. The N376PxxY motif in the transmembrane 7 domain of the receptor (rather than a N376DPxxY mutant form) was shown to be essential for ARF-dependent PLD signaling and ARF1 coimmunoprecipitation. In GST-fusion proteins of the 5-HT2AR ct domain, mutation of Asn376 to Asp also markedly reduced ARF1-HA binding, although additional motifs in the Asn376-Asn384 sequence and to a lesser extent elsewhere, seem also to contribute to the interaction.
The 5-hydroxytryptamine (5-HT) receptor superfamily represents a diverse group of receptors encompassing 14 different subtypes. All but one of them (5-HT3) signal through G-protein-linked pathways. The different families of G protein-coupled receptors (GPCRs) for 5-HT transduce their signals via different heterotrimeric G proteins, with 5-HT1 (and 5-HT5) linking to Gi/o effectors, 5-HT2 signaling through Gq/11 to generate inositol trisphosphate and increase intracellular Ca2+ levels, and 5-HT4, 5-HT6, and 5-HT7 all signaling through Gs to increase production of cyclic AMP (Raymond et al., 2001). The 5-HT2AR is of particular interest because it has been implicated in a variety of major psychiatric disorders and in hallucinogenic drug action.
GPCR interactions with heterotrimeric G proteins often (but not exclusively) seem to involve the third intracellular loop (i3) (Wess et al., 1997). The i3 domains of various GPCRs have been shown to provide docking sites for heterotrimeric G protein βγ subunits (Wu et al., 1998) as well as arrestins (Wu et al., 1997; Gelber et al., 1999), GPCR-kinases (Wu et al., 1998), and indeed ARFs (McCulloch et al., 2001; Ronaldson et al., 2002). In a number of GPCRs, other intracellular loops and ct domains have also been implicated in interactions with heterotrimeric G proteins and arrestins (Wu et al., 1997; Oakley et al., 2001). In the 5-HT2AR, the i3 domain has been shown to be important for coupling to Gq/11 and the carboxyl terminal segment of the i3 loop in particular may play a key role in the interaction (Roth et al., 1998). Arrestin isoforms bind to the i3 loop of the 5-HT2AR as they do to i3 domains of the M2 and M3 muscarinic receptors, but show broader specificity in that both nonvisual and visual arrestins are bound (Wu et al., 1997; Gelber et al., 1999).
In addition to Gq/11-mediated phospholipase C (PLC) activation, the 5-HT2AR can activate several other signaling pathways that may involve alternative direct coupling to the receptor. These include activation of phospholipase A2 [which may be mediated by a coupling mechanism other than Gq/11 (Berg et al., 1998)], activation of tyrosine phosphorylation [correlating with evidence for association of the tyrosine kinase Jak2 with the ct domain (Guillet-Deniau et al., 1997)] and ARF-dependent activation of phospholipase D (PLD) (Mitchell et al., 1998).
A specific conserved motif, NPxxY, which is found at the junction of the tm7 and ct domains in a number of rhodopsin family GPCRs, has been implicated as a determinant of ARF-receptor interactions and ARF-mediated signaling, because native receptors with an alternative DPxxY motif, or Asn-to-Asp mutant receptors, show selective defects in this pathway (Mitchell et al., 1998, 2003). However, it has not been clear whether this motif might be accessible as a direct docking site or it instead regulates access to a distinct site.
The two main classes of cellular ARFs are exemplified by ARF1 and ARF6, which are thought to have characteristically distinct subcellular distributions. In many cells, ARF1 is cytosolic or associated with Golgi membranes, whereas ARF6 may be substantially associated with plasma membrane and play a role in regulating endocytosis (D'Souza-Schorey et al., 1995; Peters et al., 1995; Cavenagh et al., 1996). Nevertheless, GTP loading and GPCR activation can cause translocation of ARFs to Golgi and other membranes, and we have shown marked translocation of ARF1 to the plasma membrane after activation of the M3 muscarinic receptor (Mitchell et al., 2003) and the 5-HT2AR (M. Johnson and R. Mitchell, unpublished observations). Thus both ARF1 and ARF6 are potentially available for interaction with plasma membrane GPCRs after agonist stimulation.
In this study, we have addressed the role of ARF1 and ARF6 in 5-HT2AR-mediated PLD signaling, demonstrated coimmunoprecipitation of ARF1 (and to a lesser extent ARF6) with the receptor, and gone on to characterize the docking of ARFs to GST fusion proteins of receptor i3 and ct domains.
Materials and Methods
Preparation of sPrC-Tagged 5-HT2A Receptor and sPrC-Tagged N376D-5-HT2A Mutant Receptor Constructs. The wild-type human 5-HT2AR cDNA (SCS93) and N376D-5-HT2AR mutant cDNA (SCS103), cloned into pcDNA1-amp (Invitrogen, Paisley, UK), were kindly provided by Stuart Sealfon (Mount Sinai School of Medicine, New York, NY). To create an amino terminal epitope-tagged receptor, SCS93 was PCR-amplified with the synthetic oligonucleotide primers [5′-GAAGATCAGGTAGATCCACGGTTAATCGATGGTAAGGCCATGGATATTCTTTGTGAAG-3′] encoding the 12-amino acid Protein C tag (PrC) epitope (EDQVDPRLIDGK) and the reverse primer named 5HT2A.rp, [5′-GAATTCTCACACACAGCTCACCTTTTCATT-3′] using the Herculase-enhanced DNA polymerase (Stratagene Europe, Amsterdam, NL). The resulting PCR product was checked by agarose gel electrophoresis and extracted with Wizard PCR clean-up resin (Promega, Southampton, UK) before amplifying with the primer [5′-GCCACCATGAAGACGATCATCGCCCTGAGCTACATCTTCTGCCTGGTATTCGCCGAAGATCAGGTAGATCCAC-3′] encoding a modified influenza hemagglutinin signal sequence and 5HT2A.rp. The 1.4-kilobase fragment was purified by agarose gel electrophoresis and Qiaex II (QIAGEN Ltd.,Crawley, UK), then subcloned into TOPO TA cloning vector (Invitrogen). The sPrC-5-HT2A insert was checked by sequencing. For expression studies, the sPrC-tagged wild-type receptor was constructed by subcloning the 200-bp EcoRI/SalI fragment from sPrC-5-HT2A cDNA and the 1.5-kilobase SalI/XbaI fragment from SCS93 into the EcoRI/XbaI site of pcDNA3 (Invitrogen). The sPrC-tagged (N376D) mutant receptor was constructed by subcloning the 450-bp EcoRI/PstI fragment from sPrC-5-HT2A cDNA and the 1.3-kilobase PstI/XbaI fragment from SCS103 into the EcoRI/XbaI site of pcDNA3 (Invitrogen).
Construction of 5-HT2A Receptor Intracellular Loop 3 and Carboxy Terminal Tail GST-Fusion Proteins. The human 5-HT2AR intracellular loop 3 (Ile258-Gly326) was PCR-amplified from SCS93 with primer pair [5′-GGGTGATCAAGTCACTTCAGAAAGAAGCTAC-3′] and [5′-CGGAATTCTAGCCCAGCACCTTGCATGCTTTTGCTCATTGCT-3′], and the resulting 200-bp PCR product was purified, digested with BclI/EcoRI, and subcloned into the BamHI/EcoRI sites of pGEX-3X (Amersham Biosciences, Little Chalfont, UK). The 347-bp HincII fragment encoding the human 5-HT2AR carboxy-terminal tail and including the NPLVY motif (Asn376-Val471) was subcloned into the modified (with Mung bean nuclease; Stratagene, La Jolla, CA) BamHI site of pGEX-2T. The sequences encoding the carboxy-terminal tail amino acids (Lys385-Val471) were PCR-amplified from SCS93 using primer pair Bh2ARCT.KTYRS [5′-CGGGATCCAAAGACCTATAGGTCAGCCTTTTCACG-3′] and h2ARCT.rp [5′-CAACTCAATTGTCACACACAGCTCACCTTTTCATT-3′], and the mutant (N376D-Val471) carboxyl tail was PCR-amplified from SCS103 with primer pair h2ARCT.DPLVY [5′-CGGGATCCAGACCCACTAGTCTACACACTGTTCAA-3′] and h2ARCT.rp using Taq polymerase (Promega) and the resulting PCR products purified by Wizard cDNA clean-up resin (Promega), digested with BamHI/MfeI and subcloned into the BamHI/EcoRI sites of pGEX-2T. The BamHI sites then were modified with Mung bean nuclease. Cloned inserts were checked by sequence analysis.
Transient Transfections of Cells. COS-7 cells were grown to 60% confluence in Dulbecco's modified Eagle's medium (DMEM; Invitrogen) containing 10% normal calf serum and 100 units/ml penicillin and 100 μg/ml streptomycin. Before transfection, medium was changed for DMEM containing 2% Ultroser G (USG; Invitrogen) instead of serum. The cells were then transfected with combinations of the cDNAs for sPrC-5-HT2AR, ARF1-V5-His6 (Invitrogen), ARF1 and ARF6 tagged with the hemagglutinin epitope (HA) at the carboxyl terminal, or the dominant-negative mutants T31N-ARF1-HA and T27N-ARF6-HA (Peters et al., 1995) (kindly provided by Julie Donaldson, National Institutes of Health). All transfections (normally at a ratio of 4 μg of receptor construct cDNA to 1 μg of ARF construct cDNA per 75-cm2 flask) were carried out using FuGENE-6 reagent (Roche Diagnostics Ltd, Lewes, UK) according to the manufacturer's guidelines, and appropriate substitutions of empty vectors were made in control samples. Transfected cells were used 72 h after transfection.
Preparation of ARF-Enriched Extracts. COS-7 cells transfected with various ARF-HA constructs were washed with 10 ml of Earle's balanced salt solution (Invitrogen) before being scraped into ice-cold extraction buffer [2 ml/175-cm2 flask, 2 μg/ml aprotinin, 1 mM 4-(2-aminoethyl)-benzenesulfonylfluoride (AEBSF) (Alexis Biochemicals, San Diego, CA), 1 mM dithiothreitol, 1 mM pepstatin, 1 mM sodium orthovanadate, 1 mM NaF, 50 μg/ml soybean trypsin inhibitor (SBTI) in PBS]. All chemicals were from Sigma Chemical Co. (Poole, Dorset, UK) unless otherwise indicated. The cells were then homogenized [Ystral homogenizer (Scientific Industries Intl. UK Ltd, Loughborough, UK) setting 3, 15 s] before being centrifuged at 12,000g for 20 min at 4°C. The supernatant was aliquoted and stored at -40°C.
Purification of ARF1-V5-His6 on Cobalt Affinity Resin. COS-7 cells were transfected with an expression plasmid encoding ARF1-V5-His6, and enriched cytosol was isolated as described above. The cytosolic extracts over-expressing the ARF1-V5-His6 were then incubated with 'Talon' cobalt affinity resin (BD Biosciences Clontech, Palo Alto, CA) that had been pre-equilibrated with wash buffer (10 mM sodium phosphate, 60 mM NaCl, pH 7.0), for 1 h at 4°C. The resin was washed thoroughly with extraction/wash buffer (additionally containing 5 mM imidazole) before being eluted using elution buffer (additionally containing 300 mM imidazole) according to the manufacturer's instructions. The purified ARF1-V5-His6 was then concentrated using Centriplus centrifugal filters (YM-3; Millipore, Bedford, MA) according to the manufacturer's instructions. The amount of protein present was verified by colorimetric measurement using Coomassie protein reagent (Perbio Science UK Ltd., Tattenhall, Cheshire, UK). The percentage purity of the harvested protein purity was then assessed by staining with high sensitivity Colloidal Coomassie (Simply Blue Safe stain; Invitrogen) on an SDS-PAGE gel, identifying the correct band by Western blotting using mouse monoclonal V5-tag antibody (Invitrogen). Standard curves were produced from the purified ARF1-V5-His6 (linear range from 0.4-7.2 ng ARF1-V5-His6 per well). These were coprocessed with experimental samples of cytosolic inputs and of proteins associating with the receptor in either coimmunoprecipitation or GST-fusion protein experiments, enabling us to assess the amounts of ARF1-V5-His6 present.
Expression of GST Proteins. GST-5-HT2Ai3 and GST-5-HT2Act constructs, in the pGEX-3X and pGEX-2T vectors, as well as a control GST fusion protein of the STREX insert of the large conductance Ca2+- and voltage-activated K+ (BK) channel, in the pGEX-5 × 1 vector (kindly supplied by Mike Shipston, University of Edinburgh) were expressed in BL21-RIL bacterial cells, grown in standard 2× yeast extract/tryptone/NaCl medium with 2% glucose. When the cells had reached an A600 of 0.6 to 0.8 units/ml, expression of the fusion proteins was induced by the addition of 0.1 mM isopropyI-β-d-thiogalactoside for 3 h at 37°C. Cells were harvested by centrifugation then lysed with BugBuster reagent (Merck Biosciences, Beeston, Nottingham, UK) for 10 min and again centrifuged. The supernatant, containing the GST fusion proteins, was added to glutathione-Sepharose beads (Amersham Biosciences). The beads were incubated with the bacterial supernatant for 20 min at room temperature to allow binding of the GST fusion proteins to the beads. The matrix formed was then washed extensively with PBS and used immediately.
In Vitro Protein Interaction Assays. Cellular extracts enriched in the various HA-tagged ARF constructs were incubated with GST fusion proteins bound to glutathione-Sepharose beads (Amersham Biosciences) in 250 μl of buffer A (20 mM Tris-HCl, pH 7.5, 0.6 mM EDTA, 1 mM dithiothreitol, 70 mM NaCl, and 0.05% Tween 80) for 90 min at 4°C, with rolling. In some experiments, GTPγS (100 μM) was added to the incubations. The beads were washed four times in buffer A and then the retained proteins were removed from the beads with 2× Laemmli buffer (2% SDS, 5% mercaptoethanol, 20 mM Tris-HCl, pH 7.4) before SDS-PAGE and immunoblotting (see below). Input levels of fusion proteins (monitored by GST immunoreactivity) were carefully balanced to ensure comparability between samples. Protein interaction assays using ARF1-V5-His6 were carried out in an identical manner, except that increased amounts of bacterial cytosol containing GST construct was added to the glutathione-Sepharose beads (∼2×).
Coimmunoprecipitation of sPrC-5-HT2A Receptor and ARF1/6-HA. Transfected COS-7 cells were incubated in medium alone (no serum or USG) for 4 h before being exposed to 5-HT (10 nM-10 μM) and/or brefeldin A (BFA, 100 μM). Cells were washed once in Hank's balanced salt solution (HBSS; Invitrogen) before the addition of 1 ml/75-cm2 flask of immunoprecipitation buffer (HEPES 20 mM, pH 7.5, NaCl 150 mM, 1% CHAPS, 0.5% sodium deoxycholate, 2 μg/ml aprotinin, 4 μg/ml leupeptin, 1 mM AEBSF, 2 μg/ml pepstatin, 1 mM sodium orthovanadate, 1 mM sodium fluoride, 5 mM sodium molybdate, and 50 μg/ml SBTI) with 5-HT if it had been used in the initial stimulation. Extraction was carried out on ice for 40 min with occasional agitation. Solubilized cellular extracts were centrifuged at 12,000g for 15 min at 4°C to remove particulate material. One ml of supernatant was precleared with 20 μl of protein G-Sepharose 4B fast flow (50% suspension in immunoprecipitation buffer) for 45 min at 4°C. After centrifugation (to pellet the beads), the supernatant was added to a tube containing either mouse monoclonal (Ca2+-dependent) PrC-tag antibody (clone HPC4, 4 μg/ml; Roche) or control, nonimmune mouse IgG, 4 μg/ml; Sigma) with 40 μl/ml protein G-Sepharose suspension and 1 mM CaCl2. Samples were incubated with rolling, at 4°C overnight. In the standard procedure, the beads were pelleted, washed twice in immunoprecipitation buffer before 40 μlof2× Laemmli buffer containing 5 mM EGTA was added per sample equivalent to 1 ml of original supernatant. In some experiments, the receptor complex was removed from the PrC antibody-protein G-Sepharose beads by incubation in 50 mM Tris and 2 mM EGTA, pH 7.5, for 30 min at room temperature. The supernatant was removed and SDS and mercaptoethanol added equivalent to 2× Laemmli buffer.
Western Blots. Western blots were carried out on samples from immunoprecipitation and GST fusion protein interaction assays. Either 20% or 12.5% precast homogeneous PhastGels (Amersham Biosciences) were used. SDS-PAGE and electroblotting onto polyvinylidene difluoride membranes (Immobilon-PSQ, Millipore, Watford, UK) were performed on a PhastSystem apparatus (Amersham Biosciences). The detection antibodies were rabbit polyclonal anti-HA (Autogen Bioclear, Calne, Wilts, UK), goat polyclonal anti-GST (Amersham Biosciences) and mouse monoclonal anti-PrC (Roche) followed by preabsorbed secondary antibodies conjugated to horseradish peroxidase (Chemicon Intl. Ltd., Harrow, UK). Visualization of antibody bands was by Luminol (New England Biolabs, Hitchin, UK), and blots were exposed to ECL film (Amersham Biosciences). Densitometric analysis of ECL images from Western blots was carried out using the ScanAnalysis program (Biosoft, Cambridge, UK). Immunoprecipitation samples were also run on larger gels on the NuPAGE SureLock mini-cell system (Invitrogen) under reducing conditions, with 10% homogenous Bis-Tris gels and then blotted on the same system according to the manufacturer's instructions.
PLD Assays. Transfected cells in 12-well plates were deprived of USG by transferring to DMEM for 18 h, during which time they were labeled with [3H]palmitate (1.5 μCi/well; PerkinElmer Biosciences, Hounslow, UK). After washing with minimal essential medium containing HEPES (25 mM, pH 7.5) and 0.5% fatty acid-free bovine serum albumin, cells were preincubated for 20 min in similar medium with or without BFA, before addition of butan-1-ol (30 mM) and various concentrations of 5-HT for a further 30 min. The vehicle for BFA (dimethylformamide at 0.2%) was added to control wells and has been shown previously to have no detectable effect on signaling responses. Reactions were terminated by removal of medium and addition of 0.5 ml ice-cold methanol to each well. Phospholipids were extracted and [3H]phosphatidyl butanol was separated on Whatman LK5D thin-layer chromatography plates (Whatman, Maidstone, Kent, UK) as described previously (Mitchell et al., 2003).
PLC Assays. Transfected cells in 12-well plates were deprived of USG by transferring to Earle's balanced salt solution containing 10 mM HEPES, pH 7.5, and 0.18% glucose for 18 h, during which time they were labeled with [3H]inositol (0.75 μCi/well; PerkinElmer Biosciences). The medium was changed for Earle's balanced salt solution containing 10 mM HEPES, 0.18% glucose, and 0.2% bovine serum albumin and washed once. Cells were preincubated for 20 min with 10 mM LiCl (and BFA where required) before being incubated for 30 min with various concentrations of 5-HT with or without BFA. Reactions were terminated by the removal of medium and the addition of 1 ml of ice-cold 10 mM formic acid (Sealfon et al., 1995) and the cells were left on ice for at least an hour to ensure lysis. Inositol phosphates ([3H]InsP) were separated by anion exchange chromatography as described previously (Mitchell et al., 2003).
[3H]Ketanserin Binding. Assessment of PrC-5-HT2A receptor expression in the transfected COS-7 cells was by homologous displacement of [3H]ketanserin binding (specific activity, 88.0 Ci/mmol; PerkinElmer Life Sciences). After removal of USG from the culture medium for the last 4 h, cell membranes were prepared by harvesting the cells into ice-cold ketanserin binding buffer (50 mM Tris, 5 mM MgCl2, and 1 mM EGTA, pH 7.2) and disrupted with an Ystral homogenizer. Membranes were washed twice in buffer, centrifuging each time at 12,000g, 30 min, 4°C. Finally the membranes were suspended in buffer at ∼200 μg/ml protein. Membranes were incubated for 60 min at 37°C with 0.8 nM [3H]ketanserin and either no other drug, increasing concentrations of unlabeled ketanserin, or 10 μM mianserin to determine nonspecific binding. At the end of the assay, binding was stopped by addition of ice-cold buffer, the membranes were pelleted by centrifugation (10 min at 12,000g) and the supernatant was aspirated. Bound radioactivity was determined by liquid scintillation counting. The homologous displacement curves were fitted to a Hill model using nonlinear curve fitting program, Fig-P (Biosoft, Cambridge, UK). This allowed measurement of KD and number of binding sites. For experiments to determine receptor binding where cell membranes were solubilized in immunoprecipitation buffer, the lysate was first treated with polyethylene glycol-8000 (PEG; final concentration, 15%) to precipitate solubilized proteins, including PrC-5-HT2A receptor, and washed, again precipitating with PEG before trituration of the pellet in ketanserin binding buffer for the assay. Additionally, PEG was used to terminate the assay, and nonsolubilized membranes included in the experiments as controls were treated in the same way.
When membranes were treated with solubilization buffer, the total amount of specific [3H]ketanserin binding that was recovered from the supernatant plus residual membrane fractions was much less than the input levels of binding in untreated membranes (approximately 15%). It is not clear to what extent this was caused by deleterious effects of detergent/PEG exposure on the ligand binding site or by inefficient capture of solubilized receptor. However, some loss of ligand binding capacity is likely to result during these procedures. It therefore seems likely that the amount of receptor still detectable by ligand binding after extraction represents an underestimate of the amount of receptor protein actually solubilized. Nevertheless, some 97% of the [3H]ketanserin binding that could be recovered was found in the solubilized fraction. When this fraction was immunoprecipitated with PrC-tag antibody, the proportion of solubilized specific binding that became associated with the protein G beads rather than remaining in the supernatant was 70 ± 18%.
[3H]Ketanserin Binding to Coimmunoprecipitates. Cells transfected with wild-type or N376D mutant sPrC-5-HT2A constructs and wild-type ARF1-HA were transferred to USG-free DMEM for 16 h before assay. After washing in PBS at 37°C, and incubation with 5-HT (1 μM, 5 min) in some cases, cells were scraped into ice-cold buffer (PBS, 10% glycerol, 2 μg/ml aprotinin, 4 μg/ml leupeptin, 1 mM AEBSF, 2 μg/ml pepstatin A, 50 μg/ml SBTI, 1 mM sodium orthovanadate, and 1 mM sodium fluoride) and homogenized using an Ystral homogenizer. Nuclei and debris were removed by centrifugation at 1000g for 8 min before the supernatant was centrifuged (12,000g for 30 min) to obtain a membrane fragment pellet. The pellet was washed once before trituration in the above buffer containing 1% CHAPS and 0.75% deoxycholate and rolling for 30 min at 4°C. Samples were precleared with protein G-Sepharose and then centrifuged (12,000g; 30 min) to obtain the solubilized extracts used for immunoprecipitation. Samples were incubated with 2 μg/ml mouse monoclonal anti-HA tag (clone 12CA5) or nonimmune mouse IgG for 90 min at 4°C and then immune complexes were collected by incubation with protein G-Sepharose for 45 min at 4°C. The pellets from centrifugation at 12,000g for 10 min were washed once in solubilizing buffer and then resuspended in [3H]ketanserin binding buffer (see above) that additionally contained 10% glycerol. Ligand binding was assayed as described above. Incubations (250 μl) were stopped by the addition at 4°Cof100 μl of 0.05% bovine γ-globulin in PBS and 1 ml of 30% PEG in PBS. After 15 min on ice, samples were centrifuged at 12,000g for 20 min, the supernatant was aspirated, and the tube tips were removed for liquid scintillation counting.
Results
Figure 1 illustrates functional signaling responses of the sPrC-5-HT2AR expressed in COS-7 cells. [3H]Ketanserin binding experiments indicated that the sPrC-5-HT2AR was expressed in COS-7 cell membranes at a mean level of 0.84 ± 0.04 pmol/mg of total protein, with an IC50 for homologous displacement of 4.8 ± 0.4 nM. The receptor produced robust PLC and PLD activation responses to 5-HT stimulation that were similar to the untagged receptor (Sealfon et al., 1995; Mitchell et al., 1998) but with slightly greater potency; the EC50 values for PLC and PLD responses were 5.0 ± 2.2 and 6.8 ± 1.9 nM, respectively (Fig. 1, A and B). The EC50 value for PLC activation by the untagged receptor in similar experiments was found to be 28 ± 2 nM (R. Mitchell and M. Johnson, unpublished observations) and 22 ± 5 nM (Sealfon et al., 1995). The effects of cotransfection of ARF mutants were investigated on 5-HT-induced signaling events mediated by the sPrC-5-HT2AR expressed in COS-7 cells. PLC activation was unaffected by cotransfection of either T31NARF1-HA or T27N-ARF6-HA; mutant constructs of the ARF isoforms that have a dysfunctional GTP binding domain (Peters et al., 1995) (Fig. 1A). However, T31N-ARF1-HA, but not T27N-ARF6-HA, significantly inhibited 5-HT2AR-mediated PLD activation (Fig. 1B). Figure 1C shows the inhibitory effect of BFA [a blocker of the BIG1/2 class of ARF GTP-exchange factor (GEF) (Morinaga et al., 1999)] on the PLD response of the 5-HT2A receptor. When only the receptor was expressed, BFA caused a concentration-dependent inhibition of 5-HT-induced PLD activation, with significant inhibition at 50 μM and above. In cells additionally expressing T27NARF6-HA, BFA was also inhibitory throughout a similar concentration range. However, in cells expressing T31NARF1-HA, the residual 5-HT-induced PLD response became insensitive to BFA. The PLC response of the 5-HT2AR was unaffected by BFA (data not shown).
To investigate whether a physical interaction of either ARF1 or ARF6 with the receptor occurs upon activation, we looked for coimmunoprecipitation of HA-tagged ARF isoforms with the sPrC-5-HT2AR (Fig. 2). When nonimmune mouse IgG (NI IgG) was substituted for the PrC-tag antibody in the immunoprecipitation, minimal levels of either ARF1-HA or ARF6-HA immunoreactivity could be detected in the pulldown assays. Significant levels of ARF1-HA (and to a much lesser extent, ARF6-HA) seemed to be specifically associated with the PrC-tag antibody pulldown assays of the sPrC-5-HT2AR, even under basal conditions (Fig. 2A). Densitometric analysis indicated that basal levels of ARF1-HA coimmunoprecipitated were increased on average 3.32 ± 1.58-fold (n = 5) over nonspecific, as determined with NI IgG. The corresponding value for ARF6-HA was lower, a 0.32 ± 0.05-fold (n = 3) increase over nonspecific. Relative densitometric values of anti-HA tag blots for coimmunoprecipitated ARF1-HA as a percentage of the input of total expressed ARF1-HA were 0.13 ± 0.03% in unstimulated samples compared with 0.21 ± 0.03% after 5-HT (1 μM, 5 min). Equivalent values for ARF6-HA were 0.053 ± 0.010% and 0.067 ± 0.008%, respectively. Although coimmunoprecipitation of both isoforms of ARF with the receptor seemed to be increased by addition of 5-HT (mean increase to 1.61- and 1.26-fold of control, respectively), only the effect on ARF1-HA was statistically significant (p < 0.05, Wilcoxon test, n = 5 in each case) These values are useful for relative comparisons but do not faithfully reflect the fraction of ARF1/6-HA within an individual cell that is associated with sPrC-5-HT2A receptor, because we know from dual label confocal immunofluorescence experiments (data not shown) that a subpopulation of the COS-7 cells expressing ARF constructs fail to express detectable levels of receptor. When the input level of sPrC-5-HT2AR was varied by substituting pcDNA3 for part of the receptor cDNA in the transfection (while keeping the ARF1-HA plasmid concentration constant), less ARF1-HA was coimmunoprecipitated, in proportion to the level of receptor expression as monitored by specific [3H]ketanserin binding (Fig. 2B).
An attempt to quantify the amount of ARF1 specifically coimmunoprecipitating with the receptor was made by cotransfection of an ARF1-V5-His6 construct, the binding of which was then monitored as anti-V5 tag immunoreactivity. Densitometric measurements of ARF1-V5-His6 coimmunoprecipitated with the receptor by PrC tag antibody (or NI IgG control), as well as input levels of ARF1-V5-His6 in cell lysates, were compared with a coprocessed standard curve of anti-V5 tag immunoreactivity prepared using ARF1-V5-His6 protein, purified on a Co2+ affinity column. The protein content of the purified ARF1-V5-His6 standard was determined, allowing estimation of the absolute amount of ARF1-V5-His6 that was specifically coimmunoprecipitated in the PrC-tag antibody pulldown assays. Under basal conditions, this was calculated as 11.1 ± 1.0 ng of ARF1-V5-His6 protein from one 175-cm2 flask. Because essentially all transfected COS-7 cells expressing detectable levels of sPrC-5-HT2AR also express ARF constructs, comparison of the amount of coimmunoprecipitated ARF with the Bmax for [3H]ketanserin binding in equivalent samples allows an estimate of the mean proportion of sPrC-5-HT2ARs that were bound to ARF1-V5-His6. Under basal conditions, this was calculated as 23% of the total; under 5-HT-stimulated conditions, this would be predicted to increase to 37% (1.61-fold of basal association). However, as pointed out above, it is not possible to use this approach to reliably estimate the proportion of expressed ARF construct that is associated with receptor.
The effect of 5-HT on coimmunoprecipitation of ARF1-HA with the sPrC-5-HT2AR was investigated further (Fig. 3). Nonimmune mouse IgG controls for 5-HT-stimulated samples gave results similar to those of unstimulated NI IgG controls (data not shown). The time course of 1 μM 5-HT-induced changes in coimmunoprecipitation of ARF1-HA with sPrC-5-HT2AR showed a clear increase in association from 2 min, a peak around 5 min, and then a gradual decline (although levels seemed to remain above basal for as long as 2 h) (Fig. 3A). The concentration dependence of the effect of 5-HT (10-min incubation) is shown by the insert in Fig. 3A, reaching a maximum by approximately 1 μM 5-HT. Figure 3B shows that the 5-HT-induced increase in ARF1-HA coimmunoprecipitation with the sPrC-5-HT2AR was reduced to basal levels by the ARF-GEF inhibitor, BFA (100 μM) (Fig. 3B). As before, the concentration of ARF1-HA present in the lysate was very similar between the different conditions (as shown in Fig. 3B, middle), and the amount of receptor immunoprecipitated (Fig. 3B, top) was also even. These observations using immunoprecipitation of the sPrC-5-HT2AR indicate that ARF1, and to a lesser extent ARF6, shows a physical association with the receptor that can be increased in the presence of 5-HT, correlating with the functional studies on PLD activation.
To examine the receptor-ARF interaction in more detail, we generated GST fusion protein constructs of the intracellular loop 3 (i3; Ile258-Gly326) and carboxy-terminal tail (ct; Asn376-Val471) of the 5-HT2AR, and investigated their ability to bind ARF1-HA and ARF6-HA in vitro. Figure 4 is a schematic representation of the 5-HT2AR showing the amino acid sequences used for these GST constructs. GST-fusion protein constructs of the i3 and ct domains of the 5-HT2AR, or the STREX exon of the BK channel as a control, were attached to glutathione Sepharose beads and used in in vitro interaction assays at equivalent input levels [as estimated by GST immunoreactivity (Fig. 5A)]. The input levels of ARF1-HA and ARF6-HA or the negative mutant constructs, deficient in GTP binding, T31N-ARF1-HA and T27N-ARF6-HA, were also balanced for HA-immunoreactivity (Fig. 5A). Figure 5B compares ARF1-HA and ARF6-HA interaction with the constructs. ARF1-HA showed much greater relative binding to the ct domain of the 5-HT2AR than to the BK channel construct or the i3 domain of the 5-HT2AR. ARF6-HA showed a much lower level of binding to the 5-HT2Act construct and little interaction with the 5-HT2Ai3 or BK constructs. Figure 5C shows the binding profiles for the functionally negative mutants of ARF1 and ARF6, T31N-ARF1-HA and T27N-ARF6-HA. The binding of T31N-ARF1-HA to the 5-HT2Act construct was greatly reduced compared with that of the ARF1-HA wild type, whereas the low background levels of T31N-ARF1-HA binding to 5-HT2Ai3 and BK channel constructs were similar to those seen with wild-type ARF1-HA. T27N-ARF6-HA showed a low level of binding to the 5-HT2Act construct that seemed to be similar to that of the ARF6-HA wild type. The apparently slightly higher levels of T27N-ARF6-HA binding compared with the wild type in the experiment displayed correspond to its somewhat higher input level. GTPγS facilitated the interaction of submaximal levels of ARF1-HA with the 5-HT2Act construct and also seemed to strengthen a weak interaction with the 5-HT2Ai3 construct, which had been minimal in the absence of added nucleotide (Fig. 5D). GTPγS did not facilitate T31NARF1-HA binding to the receptor constructs. Similar results were obtained in three different experiments.
Further to this work, a range of concentrations of both ARF1 and ARF6 were added to the GST-5-HT2Ai3 and GST-5-HT2Act constructs to assess the concentration-dependence of binding. Figure 6 shows the relative proportion of ARF isoform bound to the constructs at increasing ARF concentrations, expressed as a ratio of densitometric values for HA-immunoreactivity of the bound ARF versus GST immunoreactivity in the construct. Increasing the concentration of ARF1-HA present with each construct caused corresponding increases in the amount of ARF bound. ARF1-HA bound to the GST-5-HT2Act construct to a much greater extent than to the GST-5-HT2Ai3 construct. ARF6-HA also bound to both the GST-5-HT2Act and the GST-5-HT2Ai3 constructs, but to a much lesser extent than the ARF1-HA bound to either construct, requiring higher levels of added ARF6-HA to obtain detectable binding. To obtain an estimate of the affinity of ARF1-V5-His6 (added in cytosolic extracts) for the GST-5-HT2A ct construct, we quantified the amount of ARF binding by comparison of densitometric values for V5-immunoreactivity with a coprocessed standard curve of known amounts of ARF1-V5-His6 that had been purified by Co2+ affinity column (as above). The ARF-V5-His6 ligand was provided for the interaction assay as an enriched cytosolic extract rather than affinity-purified material because binding was more robust and consistent using this approach. It is not clear whether this is a result of deleterious effects of the purification on ARF1-V5-His6 conformation and function or whether unknown cytosolic factors are additionally required for optimal ARF binding. Working within the linear range of the standard curve for V5-immunoreactive band density against purified ARF1-V5-His6 concentration, we added a range of cytosolic ARF1-V5-His6 concentrations (1-25 ng) to interaction assays and derived values for the amounts of ligand bound from the corresponding V5-immunoreactive band densities. Nonlinear curve fitting of the saturation curve gave a value for affinity (50% saturation) of 1.7 ± 0.4 nM, with greater than 90% occupancy of available sites by 4 to 5 nM.
The differential signaling properties of the wild-type 5-HT2AR and the N376D mutant 5-HT2AR (Mitchell et al., 1998, 2003) suggest that the NPxxY motif, at the junction of the ct and the 7th transmembrane domain, may participate in the binding of ARF1 to the receptor. To test this theory, we carried out signaling experiments, coimmunoprecipitation, and GST-fusion protein studies with both the wild-type and N376D mutant form of the receptor. Figure 7A shows that 1 μM 5-HT-induced PLD activation by the wild-type sPrC-5-HT2AR was significantly reduced by BFA (100 μM) or by coexpression of T31N-ARF1-HA but not T27N-ARF6-HA. Corresponding responses of the N376D-sPrC-5-HT2AR examined in the same experiments showed no significant inhibition. Figure 7B shows the results of coimmunoprecipitation experiments in COS-7 cells cotransfected with ARF1-HA and either the wild-type sPrC-5-HT2AR or its N376D mutant form. After stimulation with 5-HT (1 μM, 5 min) or control, solubilized extracts were immunoprecipitated with HA-tagged antibody, and 5-HT2ARs associated with the immunoprecipitate were assayed as specific [3H]ketanserin binding. Low levels of nonspecific [3H]ketanserin binding were present in each case, and these showed no discernible differences between samples. In cells transfected with the wild-type sPrC-5-HT2AR but not those expressing the N376D mutant, significant levels of specific [3H]ketanserin binding became associated with the HA tag immunoprecipitate after 5-HT stimulation. When NI IgG was substituted for the HA tag antibody during immunoprecipitation, no specific [3H]ketanserin binding was captured. Using the [3H]ketanserin binding protocol there was no evidence for significant basal association between ARF1-HA and sPrC-5-HT2AR (6 ± 8% of specific binding), unlike the data shown in Figs. 2 and 3. There were a number of methodological differences (such as shorter time for immunoprecipitation and the addition of glycerol) between this procedure and that used previously, which may result in the differences in basal association. 5-HT stimulation clearly caused an increased interaction between the sPrC-5-HT2AR and ARF1-HA. Comparison of the amount of specific [3H]ketanserin binding associated with ARF1-HA immunoprecipitates and the input levels of binding in solubilized extracts gave an estimate that 30 ± 5% of sPrC-5-HT2aRs were associated with ARF1-HA after 5-HT stimulation. This value is in reasonably close agreement with values estimated from the immunoprecipitation/immunoblot experiments (above).
Figure 7C shows an experiment to investigate whether the N376PLVY motif in the 5-HT2AR ct domain may form part of the binding site for ARF1-HA. GST-fusion protein constructs of the wild type (Asn376-Val471) 5-HT2Act, the mutant (N376D-Val471) 5-HT2Act and the truncated (Lys385-Val471) 5-HT2Act were prepared. Equal amounts of these constructs (and GST alone) were determined by Coomassie blue staining and by GST immunoreactivity at the predicted molecular mass before interaction assays with ARF1-HA. The ratios of the densitometric values for bound ARF1-HA immunoreactivity to fusion protein input were then calculated on an arbitrary scale relative to that for the wild-type construct. Both individual images and the mean densitometry ratios for bound ARF1-HA:construct input showed a clear reduction in binding (to around 40% of wild type) by the N376D mutation and a further loss (to around 20%) by deletion of the Asn376-Asn384 sequence (Fig. 7C).
Discussion
The present findings demonstrate that a negative mutant construct of ARF1, but not ARF6, inhibits the activation of PLD, but not PLC, by the 5-HT2AR. This indicates a selective functional role for the ARF1 isoform in the PLD signaling pathway of the 5-HT2AR. PLD but not PLC responses of the 5-HT2AR were correspondingly reduced in a concentration-dependent manner by BFA, an inhibitor of the ARF-GEFs, BIG1/2, which are reported to show selectivity for ARF1 rather than ARF6 (Morinaga et al., 1999). Further evidence consistent with a functional role for ARF1 in BFA-sensitive PLD responses came from experiments assessing the BFA sensitivity of 5-HT2AR PLD responses in cells coexpressing negative-mutant ARFs, T31N-ARF1-HA or T27N-ARF6-HA. The inhibitory effect of the negative-mutant ARF1 construct pre-empted any further inhibition by BFA, suggesting that they both acted within the same pathway, whereas negative mutant ARF6 was without effect. Other GPCRs may show different selectivity for ARF isoforms. In COS-7 cells expressing the M3 muscarinic receptor and in A10 smooth muscle cells, PLD responses to carbachol and to angiotensin II or ET-1, respectively were attenuated by T31N-ARF1 and by T27N-ARF6 (Shome et al., 2000; Mitchell et al., 2003), whereas some other GPCRs such as P2u and PAC1-hop1 receptors may show selectivity for ARF6 over ARF1 (Ronaldson et al., 2002; Mitchell et al., 2003).
Coimmunoprecipitation experiments demonstrated that the sPrC-5-HT2AR, under basal conditions, could specifically bind ARF1-HA and to a lesser extent ARF6-HA. The binding of ARF1-HA was increased by 5-HT stimulation of the receptor in a concentration- and time-dependent manner. Only minor increases in the association of ARF6-HA with the sPrC-5-HT2AR could be seen. The presence of BFA during the stimulation with 5-HT reversed the 5-HT-induced association of ARF1-HA, although we know from confocal immunofluorescence imaging that BFA does not prevent agonist-induced translocation of ARF1-HA to the plasma membrane (data not shown). This evidence that ARF1 is the isoform predominantly docking to the receptor in a BFA-sensitive manner is consistent with the data on PLD activation. Estimates of the proportion of 5-HT2A receptors that was associated with ARF1-HA after 5-HT stimulation ranged between 30 and 37% depending on the approach taken, with corresponding estimates of basal association between 6 and 23%. Interaction of other proteins, such as Gq/11 and perhaps arrestins, with the 5-HT2AR may well mean that relevant binding sites were not accessible to ARF1 in part of the receptor population.
The GST-fusion protein experiments suggested that the ct domain of the 5-HT2AR provides a binding site for ARFs at which ARF1 shows a higher affinity than ARF6. Comparison of the immunoreactivity for ARF1-V5-His6 bound to the GST-5-HT2Act construct with known amounts of purified ARF1-V5-His6 allowed an estimate of the affinity of interaction. This was in the low nanomolar range (1.7 ± 0.4 nM), which is of lower affinity than that for arrestin interaction with the M3R i3 domain (Wu et al., 1997) but higher than the corresponding interaction of Gβγ (Wu et al., 1998). The 5-HT2A receptor i3 domain shows only low affinity for ARF in vitro but may still represent an auxiliary binding site in vivo. In contrast, the i3 domain effectively binds arrestins in similar experiments (Gelber et al., 1999). The interaction of ARF1-HA with the ct or i3 domain of the 5-HT2AR seemed to be facilitated by GTPγS, suggesting that occupancy of its nucleotide recognition site by GTP rather than GDP promotes the interaction. Correspondingly, the GTP-binding-defective mutant ARF1 construct (T31N-ARF1-HA) showed an almost complete lack of specific binding to the ct or i3 domain GST fusion proteins that was unmodified by GTPγS. The lower level of ARF6-HA binding seemed to be little affected by T27N mutation of ARF6-HA, but this was not investigated further. The means by which agonist induces increased (BFA-sensitive and GTP status-sensitive) binding of ARF1 to the 5-HT2AR is not clear. Involvement of BIG1/2 is implicated by the BFA sensitivity, but it is not known whether agonist activation of the receptor might facilitate GTP loading of ARF1 by direct protein-protein interaction, by regulation of BIG1/2, or by other means. However, GTP binding operates a conformational switch in ARFs that might contribute to additional protein-protein interactions (Goldberg, 1998).
Although the tm7 NPxxY motif has been implicated as a critical determinant of ARF coimmunoprecipitation and ARF-dependent signaling in rhodopsin family GPCRs, the precise site of ARF binding to the ct of the 5-HT2AR remains to be elucidated. Mutation of this motif to DPxxY strongly inhibits BFA-sensitive, ARF-mediated activation of PLD (Fig. 7A; Mitchell et al., 1998) and sPrC-5-HT2AR coimmunoprecipitation with ARF1-HA (Fig. 7B). Using GST-fusion proteins of the 5-HT2AR ct domain, mutation of Asn376 to Asp causes a marked (60%) reduction in ARF1-HA association and removal of the Asn376-Asn384 segment reduces association further to around 20% of the wild-type values. This suggests that the majority of the key elements involved in 5-HT2AR-ARF1 interaction, at least under these circumstances, may lie within the Asn376-Asn384 segment. Structural modeling based on rhodopsin and secondary structure predictions (PHD predict; http://cubic.bioc.columbia.edu/predictprotein/) suggest that the Pro377 residue is likely to form a pronounced kink in the tm7 helix and that Thr381-Lys385 may form a flexible hinge to an eighth helical segment that runs in the plane of the membrane until a palmitoylation anchor at Cys397 (Konvicka et al., 1998; Palczewski et al., 2000; Yeagle et al., 2000; Visiers et al., 2002). In the case of rhodopsin, activation of the receptor newly exposes to the intracellular surface an epitope that includes residues equivalent to Leu378-Tyr380 here (Abdulaev and Ridge, 1998), consistent with the idea that receptor activation may reveal residues involved in ARF association. The predicted fourth intracellular loop of rhodopsin, in particular residues equivalent to Asn384-Gln386 here, is involved in interaction with the α and γ subunits of transducin (Ernst et al., 2000; Marin et al., 2000). Interactions between amino acids in the NPxxY motif and the subsequent seven residues are thought to influence heterotrimeric G protein activation by both rhodopsin and the 5-HT2CR (Prioleau et al., 2002; Fritze et al., 2003). Elements of this surface might also contribute to ARF docking. The interaction of 5-HT2AR with Gαq/11, however, is also thought to involve the carboxyl portion of the i3 loop (Roth et al., 1998).
Additional functional roles have been proposed for the N/DPxxY motif. The most consistent evidence is for a role linking the tm2 and tm7 helices (Sealfon et al., 1995). The NPxxY motif and the Y residue in particular have been proposed to constitute an internalization motif in some but by no means all GPCRs (Hunyady et al., 1995). Mutation of the Asn or Asp residue to Ala generally causes massive disruption of signaling pathways and of internalization, whereas reciprocal mutation of Asn or Asp seems to have relatively minor effects other than on ARF-dependent PLD activation (Sealfon et al., 1995; Le Gouill et al., 1997; Mitchell et al., 1998). In the case of the 5-HT2A receptor, we confirmed that the NPxxY motif, rather than the DPxxY mutant motif, was necessary for functional BFA-sensitive and T31N-ARF1-HA-sensitive PLD responses from the receptor for 5-HT-induced coimmunoprecipitation of the receptor with ARF1-HA and for the major part of in vitro binding of ARF1-HA to the ct domain of the receptor.
There is increasing evidence that particular GPCRs can interact with diverse scaffolding and signaling proteins other than their conventional partners, the heterotrimeric G-proteins (Brady and Limbird, 2002; Premont and Hall, 2002). Receptor ct segments may dock bivalent adapter proteins containing PDZ or other domains, signaling proteins, and modulators of signaling functions (Dev et al., 2001; Oakley et al., 2001; Brady and Limbird, 2002). In the 5-HT2 receptor family, the distal ct residues are targeted by the PDZ-domain proteins PSD-95 and MUPP-1 (Backstrom et al., 2000; Becamel et al., 2001; Xia et al., 2003), interactions that may modify the signaling function of the receptors. A novel PDZ domain protein, tamalin, has been shown to bind to both mGluR1/5 receptors and the ARF-GEF ARNO (Kitano et al., 2002). It is conceivable that an analogous arrangement might occur in the case of the 5-HT2A receptor, locating an ARFGEF in the proximity of ARF.
ARF may not be the only small G protein that can interact with GPCRs. We showed that Rho A can be coimmunoprecipitated in a complex with NPxxY GPCRs (Mitchell et al., 1998), and there is evidence that both Gα13 and Gαq may interact with Rho-GEFs to facilitate Rho function (Sagi et al., 2001). Other small G proteins of unknown identity have also been found to associate with the formyl-Met-Leu-Phe receptor (Polakis et al., 1989).
In summary, these experiments provide intracellular signaling, coimmunoprecipitation and in vitro domain interaction evidence for ARF association with the 5-HT2AR, corresponding to its functional activation of PLD. Furthermore, ARF1 rather than ARF6 seems to participate in this mechanism through GTP-dependent interaction with a ct domain of the receptor.
Acknowledgments
We are grateful to Stuart Sealfon and Julie Donaldson for their kind gifts of reagents.
Footnotes
-
This work was supported by grants from the Medical Research Council (to R.M.) and the Wellcome Trust (to E.M.L.).
-
ABBREVIATIONS: 5-HT, 5-hydroxytryptamine, serotonin; 5-HT2AR, 5-hydroxytryptamine 2A receptor; GPCR, G protein-coupled receptor; PLD, phospholipase D; ARF, ADP-ribosylation factor; i3, third intracellular loop domain; ct, carboxy-terminal tail domain; tm7, 7th transmembrane domain; sPrC, protein C epitope tag with signal sequence; HA, hemagglutinin epitope tag; BFA, brefeldin A; GST, glutathione S-transferase; NI IgG, nonimmune mouse IgG; PLC, phospholipase C; PCR, polymerase chain reaction; PrC, Protein C; bp, base pair(s); DMEM, Dulbecco's modified Eagle's medium; USG, Ultroser G; AEBSF, 4-(2-aminoethyl)-benzenesulfonylfluoride; PAGE, polyacrylamide gel electrophoresis; GTPγS, guanosine 5′-O-(3-thio)triphosphate; CHAPS, 3-[(3-cholamidopropyl)dimethylammonio]propanesulfonate; SBTI, soybean trypsin inhibitor; ECL, enhanced chemiluminescence; MEM, minimum essential medium; PEG, polyethylene glycol; GEF, GTP-exchange factor; BK, large-conductance potassium channel; PDZ, postsynaptic density 95/disc-large/ZO-1.
- Received August 6, 2003.
- Accepted August 13, 2003.
- The American Society for Pharmacology and Experimental Therapeutics