Regulation of AKAP-Membrane Interactions by Calcium*

The AKAP gravin is a scaffold for protein kinases, phosphatases, and adaptor molecules obligate for resensitization and recycling of β2-adrenergic receptors. Gravin binds to the receptor through well characterized protein-protein interactions. These interactions are facilitated ∼1000-fold when gravin is anchored to the cytoplasmic leaflet of the plasma membrane. Although the N-terminal region (∼550 residues) is highly negatively charged and probably natively unfolded, it could anchor gravin to the inner leaflet through hydrophobic insertion of its N-terminal myristate and electrostatic binding of three short positively charged domains (PCDs). Loss of the site of N-myristoylation was found to affect neither AKAP macro-scopic localization nor AKAP function. Synthetic peptides corresponding to PCD1-3 bound in vitro to unilamellar phospholipid vesicles with high affinity, a binding reversed by calmodulin in the presence of Ca2+. In vivo gravin localization is regulated by intracellular Ca2+, a function mapping to the N terminus of the protein harboring PCD1, PCD2, and PCD3. Mutation of any two PCDs eliminates membrane association of the non-myristoylated gravin, the sensitivity to Ca2+/calmodulin, and the ability of this scaffold to catalyze receptor resensitization and recycling.

PCDs in gravin are reminiscent of the basic/hydrophobic clusters in the effector domain of the myristoylated alanine-rich C kinase substrate (MARCKS) (7), the juxtamembrane region of ErbB (8), the C-terminal region of the NMDA receptor (9), and N-terminal region of AKAP79 (10,11). Although not identical or homologous in sequence, these basic/hydrophobic regions in the other proteins (e.g. MARCKS) are all capable of binding to membranes and Ca 2ϩ /CaM with significant affinity.
We explore herein how the myristate and three N-terminal PCDs contribute to membrane binding, and thus to the function of gravin. Why is membrane binding important? It localizes gravin to the plasma membrane, where it experiences ϳ1000-fold higher effective concentration of membrane bound ␤ 2 AR because of the well established "local concentration" or "reduction of dimensionality" effect (12). Membrane binding could be important for function, as it is for many other signaling molecules such as Src, protein kinase C (PKC), etc. Using an N-terminal-tagged gravin that cannot be myristoylated, we show that two PCDs are required for functioning of gravin in resensitization and recycling. We also show membrane binding, and thus function, can be reversed by Ca 2ϩ /CaM, which binds to each of the PCD with a sufficient affinity to compete with the membrane and cause translocation of gravin from membrane to cytoplasm.

EXPERIMENTAL PROCEDURES
HA Tag and CFP Fusion Proteins-The HA-tagged fusion proteins of gravin, N-terminal-truncated HA-gravin, and C-terminaltruncated HA-gravin mutants were constructed as previously described (3) and inserted into an expression vector. To generate HA-tagged short, positively charged calmodulin-binding domains (PCD), the sense primers contained 5Ј-NheI site, and nucleotides encoding the HA tag (YPYDVPDYALV) followed by nucleotides corresponding to 171-177 (for PCD1) and amino acids 297-303 (for PCD2), and amino acids 510 -516 (for PCD3). The antisense primers that were synthesized corresponded to amino acids 191-197 (for PCD1) and amino acids 314 -320 (for PCD2), and amino acids 531-537 (for PCD3) of human gravin, with the addition of a BamH1 restriction site, following the stop code. The PCR products were cloned into pcDNA3 (Invitrogen) between NheI and BamHI restriction sites.
To generate HA tag gravin in which PCD1 and PCD2 regions were deleted (Gravin ⌬PCD1,2), the sense primer contained a 5Ј-NheI site and nucleotides encoding the HA tag (YPYDVP-DYALV) in the front of the N terminus of gravin. The antisense primer corresponding to the sequence of gravin from 488 -510 followed by ACC1 site (CCGGTCGACAATATCATTAGC-CTGGGACTCAG). For the second fragment, the sense primer contain the ACC1 site corresponding to the sequence of gravin (CCGGTCGACACAGAAGAAGACGGAAAGGC), the antisense primer contain a BamHI site corresponding to the sequence of gravin (AGCAAGGATCCCGTCTGTCCCC-GTCTC). The PCR products were cloned into pcDNA3 (Invitrogen) between NheI and BamHI sites to generate a mutant form of gravin in which PCD1 and PCD2 are deleted. A similar approach was employed to delete the other pairs of PCDs in gravin. Briefly, to generate HA-tagged gravin with deletions of PCD1 and PCD3, primers were designed to delete amino acids 172-190. A restriction enzyme cutting site XhoI was inserted to delete PCD1. For the deletions of PCD3, amino acids 510 -536, a restriction enzyme cutting site EcoRI was introduced by PCR. To generate HA-tagged gravin in which PCD2 and PCD3 are deleted, primers were designed to delete amino acids 296 -317 by engineering a restriction enzyme cutting site XhoI to delete PCD2, followed by use of primers to delete PCD3, amino acids 510 -536, to introduce a restriction enzyme cutting site EcoRI by PCR. The glycine-to-alanine substitution mutant G2A of gravin-GFP was engineered by standard protocol for PCR-mediated mutagenesis. All of the polymerase chain reactions were performed using Pfu polymerase (Stratagene). The identity of the amplified sequences was confirmed by direct DNA sequencing.
Gravin-CFP and gravin ⌬PCD1,2-CFP were constructed in a technical manner very similar to that employed for HA-tagged gravin and HA-tagged gravin ⌬PCD1,2, except that the vector used in this case was CFP-N1(Clontech) and the C-terminal of gravin (lacking a stop code) was followed by the CFP moiety as a fusion protein. PCD1-CFP and PCD2-CFP were generated similar to that employed for HA-PCD1 and HA-PCD2, except that CFP-N1 (Clontech) was used as vector and the C-terminal of each peptide (without stop code) were followed by CFP moiety to create a fusion protein. All of the polymerase chain reactions were performed using Pfu polymerase (Stratagene). The identity of the amplified sequences was confirmed by direct DNA sequencing.
Expression of HA-and Fluorophore-tagged Mutant and Wild-type Versions of Gravin-A431 cells were transfected with expression vectors harboring either gravin-GFP or a truncated version of gravin-CFP using Lipofectamine® (Invitrogen), according to the manufacturer's protocol. The viable clones were selected in the presence of 400 g/ml of the neomycin analogue G418. Resistant colonies were subcloned and screened for GFP or CFP fusion protein expression by fluorescence microscopy. Cells transfected with HA-tagged gravin were selected in 200 g/ml of the hygromycin and screened for expression of an HA-tagged fusion protein by SDS-PAGE of whole cell lysates followed by immunoblotting, using anti-HA antibodies.
Calmodulin Binding Assay-HA-tagged gravin and HAtagged truncated gravin, or the HA-tagged, suspected CaMbinding regions were expressed in A431 cells by transient trans-fection. Cells were harvested and lysed in a lysis buffer (50 mM Tris-HCl, pH 7.5, 2 mM CaCl 2 , 150 mM NaCl, 1% Triton X-100, 0.5% Nonidet-40, 5 g/ml aprotinin, 5 g/ml leupeptin, 100 g/ml bacitracin, 100 g/ml benzamidine). Cell lysates were subject to filtration though a column packed with Sepharose-4B to which CaM was covalently attached. The column chromatography was performed according to the instruction of the manufacturer (Amersham Biosciences). Proteins and/or peptides retained by the immobilized CaM were eluted from the matrix by stripping the Ca 2ϩ from the matrix using a 50 mM Tris-HCl, pH 7.5, 150 mM NaCl buffer containing 2 mM EDTA. The eluted proteins were dialyzed against 10 mM Tris-HCl, pH 7.5 buffer and the resultant-dialyzed proteins were concentrated by a vacuum centrifugation. The purified proteins were subjected to SDS-PAGE on 10 -20% acrylamide gels and the separated proteins were transferred electrophoretically from the gel to the nitrocellulose membrane. The HA-tagged proteins and/or peptides were stained in the blots by making use of an anti-HA antibody. The immune complexes were detected using a horseradish peroxidase-conjugated secondary antibody, in combination with the chemiluminescence reagent, and a brief autoradiography of Kodak X-Omat film.
Binding of PCD Synthetic Peptides to Vesicles and to Calmodulin-Peptides corresponding to the three PCDs shown in Fig. 2 were synthesized with a Cys residue added to the N terminus; the ends were blocked with acetyl and amide groups. The peptides were labeled with radioactive N-[ 3 H]ethylmaleimide (NEM) as described previously (7,13), the labeled peptide purified by HPLC, and the purity checked by mass spectroscopy. We measured the binding of these [ 3 H]NEM-labeled peptides to sucrose-loaded PC/PS 100 nm diameter large unilamellar vesicles (LUVs) using a centrifugation technique described previously (7,13). Briefly, sucrose-loaded PC/PS LUVs were mixed with labeled peptide (2-10 nM) and the mixture centrifuged at 100,000 ϫ g for 1 h to spin down the vesicles and bound peptide. The percentage of bound peptide was calculated from the radioactivity of the peptide in both the supernatant and in the pellet using Equation 1. As considered in detail elsewhere (14 -16), we describe the nonspecific binding or partitioning of the basic/hydrophobic PCD peptides to phospholipid vesicles using Equation 1, where [P] mem /[P] tot is fraction of peptide bound to membranes, [L] acc is the accessible lipid concentration (one-half the total lipid concentration because we add the peptide to preformed vesicles), and K is the molar partition coefficient. (An identical equation results if one assumes, incorrectly, that the peptide forms a 1:1 complex with a lipid.) We used a modified version of Equation 1 to describe the effect of Ca 2ϩ /CaM on the membrane binding of the PCD peptides, incorporating the assumption that Ca 2ϩ /CaM and the membrane compete for the peptide (8)  where K CaCaM is the association constant of the peptide with Ca 2ϩ /CaM. Special conditions were required to estimate the binding of the PCD1 peptide to membranes and to calmodulin. It is apparent from Fig. 2 that the PCD1 region has fewer basic (blue) and more hydrophobic (green) residues than either the PCD2 or PCD3 regions. Thus a peptide corresponding to PCD1 will have the greatest tendency to aggregate in solution. Although the PCD1 peptide, gravin (171-187), appeared to dissolve in solution by visual inspection, it clearly formed aggregates because this peptide (unlike the PCD2 and PCD3 peptides) was significantly removed from solution by centrifugation. Thus measurements corresponding to Fig. 3A with PCD1 peptide (not shown), and those shown in Fig. 3C were conducted with 0.01% Triton X-100 added to the solutions to solubilize the peptide. Addition of detergent will decrease the affinity of the peptide for the membrane (and probably Ca 2ϩ / CaM), so the affinity values reported in Table 1 for gravin (171-187) should be regarded as underestimates. As a control, we studied the effect of Triton X-100 on the membrane binding of the (apparently soluble) gravin (297-317) peptide. In experiments similar to those illustrated in Fig. 3A with 2:1 PC/PS vesicles, we found that the K value decreased exponentially (binding energy decreased linearly) with the % detergent over the range from 0 to 0.025%. Fortunately, the value extrapolated from the measurements with detergent agreed well with the actual value measured in the absence of detergent for gravin (297-317), suggesting the peptide is present as a monomer in the absence of detergent. The observation that a 2-fold increase in peptide concentration did not affect the value of K is also consistent with the assumption the peptides are binding as monomers to the membrane.
Confocal Microscopy-A431 cells stably expressing gravin-GFP, truncated gravin-CFP, or HA-tagged gravin and propagated on glass slides were either left untreated or stimulated with 10 M isoproterenol for 30 min, 10 M A23187 for 30 min, or 20 M BAPTA-AM for 6 h and then washed twice with Hank's balanced salt solution, fixed (2% paraformaldehyde, pH 7.2). For immunostaining studies, cells were permeabilized and stained with anti-HA antigen rabbit antibodies used in combination with anti-rabbit Texas Red antibodies (Molecular Probes). Stained objects were imbedded in ProLong (Molecular Probes) anti-fade reagent. Images were acquired on the Zeiss LCM510 microscope using argon and helium-neon lasers (oilimmersion, ϫ63 objective). Serial sections were acquired as a Z-stack. Z-stacks of images were exported as TIFF files and processed in Adobe Photoshop 5.5.
FPLC Calmodulin Binding Assay-HA-tagged PCD1, -PCD2, and -PCD3 fragments, each were expressed via transient transfection into A431 cells. Cells were harvested and lysed in a lysis buffer (50 mM Tris-HCl, pH 7.5, 2 mM CaCl 2 , 150 mM NaCl, 1% Triton X-100, 0.5% Nonidet P-40, 5 g/ml aprotinin, 5 g/ml leupeptin, 100 g/ml bacitracin, 100 g/ml benzamidine). Cell lysates were subjected to column chromatography on CaM-immobilized Sepharose-4B matrix. Bound peptides were eluted from the medium by stripping the matrix of Ca 2ϩ using a gradient of EDTA (0 -3 mM). The eluted fractions were collected at 1-ml intervals. The eluted proteins were dialyzed in 10 mM Tris buffer pH 7.4 to remove salt and concentrated in a vacuum centrifugation. The purified proteins were separated by 10 -20% SDS-PAGE, transferred electrophoretically to nitrocellulose blots, and the resolved proteins stained by immunoblotting using anti-HA antigen primary antibodies.
Knockdown Studies of Gravin-Antisense morpholino oligonucleotides (morpholins) were synthesized and purified to cell culture grade (Gene Tools, LLC). Before addition to A431 cells, morpholinos were mixed at a ratio of 1:1 (w/w) with EPEI special delivery solution (Gene Tools, LLC). Cells were treated with morpholinos (5 g/ml) for 3 days. An additional treatment with morpholinos was performed prior to transfection of the cells with either wild-type or mutant forms of gravin. Following this protocol, cells were analyzed for ␤-adrenergic agonist-induced (i.e. isoproterenol, 10 M) desensitization and recovery (i.e. resensitization) after washout of agonist. The suppression of gravin expression was confirmed by SDS-PAGE of whole cell lysates followed by immunoblotting. Similarly, loading controls were established by immunoblotting of actin. Identical treatment of the cultures with "scrambled" sequence morpholinos (designed by the commercial supplier) were employed as a control.
Protocols for Desensitization and Resensitization of ␤ 2 AR-Two days prior to the analysis of agonist-induced desensitization, the A431 cells were seeded in 96-well microtiter plates at a density of 25,000 -50,000 cells/well. Routinely cells were serum-starved overnight, prior to analysis. Desensitization was accomplished by pretreating the cells with the ␤-adrenergic agonist isoproterenol (10 M) for 30 min. Under these conditions, subsequent ␤-adrenergic stimulation of cyclic AMP accumulation is severely blunted and the number of cellsurface receptors declines precipitously as the receptors are sequestered and internalized (17). Details of the desensitization protocol are described elsewhere (18).
Analysis of ␤ 2 AR Internalization and Recovery of ␤ 2 AR to the Cell Membrane-The internalization of ␤ 2 AR correlates well with the extent of agonist-induced desensitization. Using the binding of a cell-impermeant, radiolabeled antagonist for the determination of cell surface ␤ 2 ARs, the functional status of the recovery of internalized ␤ 2 AR could be assayed with great accuracy. Cultures of A431 cells were treated with isoproterenol for 30 min (i.e."desensitized") or treated with isoproterenol for 30 min then washed free of agonist for 60 min (i.e."resensitized"). The cells then were washed with ice-cold phosphate-buffered saline and resuspended in DMEM containing 20 mM HEPES (pH 7.4) and the hydrophilic, membrane-impermeant ␤ 2 -adrenergic antagonist [ 3 H]CGP-12177 (70 nM). Binding was performed at 4°C for 6 h. The cells were diluted, collected on GF/C membranes at reduced pressure, and washed rapidly. The radioligand bound to the washed cell mass on the filter represents a direct assay of the cell surface complement of receptors. The amount of bound ligand was quantified by liquid scintillation spectrometry (18,19).

N-Myristoylation of Gravin Is Not Required for Function-
Human gravin possesses the N-terminal sequence N-MGAGSSTEQR, a well known signal recognition sequence for N-myristoylation (5,20), a co-translational modification affording weak protein-membrane interactions. Gravin (M r ϳ250,000) was engineered either with the hemagglutinin tag (HA-gravin) at the N terminus (effectively precluding N-myristoylation) or as a fusion protein with the enhanced green fluorescent protein at the C terminus (gravin-GFP, M r ϳ275,000), and stably expressed in human epidermoid carcinoma A431 cells (Fig. 1A). Confocal microscopy of cells shows HA-tagged gravin and gravin-GFP throughout the cell, except in the nucleus, and prominent at the cell membrane (Fig. 1B). Membrane-associated gravin is labeled with white arrows, whereas the cytosolic gravin is labeled with yellow arrowheads. The macroscopic pattern of localization for gravin was unaffected by the loss of the N-myristoylation (Fig. 1B).
To test the functional role of N-myristoylation itself on gravin function, we compared the functional capabilities of wild-type (WT) gravin to that of a gravin-GFP (G2A) mutant. To assess function, we first treated cells with isoproterenol (10 M, 30 min; ϩISO) to promote agonist-induced desensitization and internalization of ␤ 2 AR (Fig. 1C). Cells made deficient of gravin (knockdown, KD) through treatment with antisense morpholinos fail to display recycling of agonist-stimulated internalized ␤ 2 AR (gravin KD), as measured by radioligand binding experiments with the cell-impermeant ␤-adrenergic antagonist [ 3 H]CGP-12177 that can only access cell surface ␤ 2 AR. At 60-min postwashout of the isoproterenol (W60), knockdown of gravin attenuates the recycling of internalized ␤ 2 AR. ␤ 2 AR internalized in response to agonist treatment do not recycle in the gravin-KD cells (Fig. 1C). We compared the ability of expression of wild-type (myristoylated) gravin versus that of the G2A mutant of gravin (non-myristoylated) to rescue the recycling of internalized ␤ 2 AR in cells in which endogenous gravin was suppressed. Expression of either the wild-type gravin or the G2A gravin mutant reconstituted the ability of the cells to rescue the recycling of the internalized ␤ 2 AR. These observations demonstrate that N-myristoylation itself is not required for gravin function. Similarly, gravin tagged on the N terminus with the HA antigen has been shown to FIGURE 1. Blockade of gravin N terminus and N-myristoylation does not alter cell membrane localization. A431 cells were transfected with expression vector harboring either gravin tagged at the N terminus with the HA antigen (HA-gravin) or a fusion protein of gravin fused C-terminally with enhanced green fluorescent protein (gravin-GFP). A, whole cell lysates were prepared and subjected to SDS-PAGE, transferred to nitrocellulose, and the blots probed with an antibody against gravin, against the HA antigen, or against the GFP moiety. B, A431 clones stably expressing the gravin-GFP and the HA-tagged gravin were analyzed by confocal microscopy in the unstimulated, basal state. The HA-tagged gravin was stained with anti-HA antibodies and a fluorescent secondary antibody. The results are typical of more than five separate experiments performed on as many separate clones. Gravin or gravin mutants associated with the cell membrane are labeled with white arrows; whereas those found in the cytoplasmic compartment are labeled with yellow arrowheads. C, A431 cells were untreated (Control) or treated with antisense morpholinos against gravin to knock down the expression of endogenous gravin (gravin KD). The antisense morpholino-treated cells were employed as such or transiently transfected with an expression vector harboring a gravin G2A mutant deficient (gravin-GFP (G2A)) or a full-length, wildtype gravin (WT gravin). Cells were desensitized with ␤-adrenergic agonist (isoproterenol 10 M, ϩIso) for 30 min and the amount of cell surface-associated ␤ 2 AR measured either using the cell impermeant radiolabeled antagonist CGP as a ligand. Following 30 min of agonist treatment, half of the cells were washed free of the isoproterenol (Wash out) and incubated for a 60-min recovery phase of resensitization when the receptors recycle to the cell membrane. Cell surface-associated CGP binding was measured after the 60 min washout and recovery phase. The results, displayed as mean values Ϯ S.E., are of at least three separate experiments performed with as many separate cultures of A431 cells. function normally (3). Thus wild-type gravin, unlike Src and MARCKS (which use myristate and a single PCD), does not require the myristate for membrane localization and function, although we cannot rule out that N-mysritoylation might facilitate some other subtle albeit important aspect of membrane localization for the AKAP. In view of these data, we adopted the use of the HA-tagged gravin which cannot be myristoylated in some of these studies.
Identification of Three MARCKS Effector Domain-like Regions in Gravin-We and others have identified three positively charged domains (PCDs) in gravin (4,6) that are similar to the positively charged "effector domain" of MARCKS (21). These PCDs all contain both basic (blue: Lys, Arg) and hydro-phobic (green; Phe, Met, Leu, Val, Cys, Trp) residues, but lack acidic residues: PCD1 ( 171 GFKKVFKFVGFKFTVKK 187 ); PCD2 ( 297 KKFFTQGWAGWRKKTSFRKPK 317 ); and, PCD3 ( 510 KV-QGSPLKKLFTSTGLKKLSGKKQKGK 536 ). These PCDs are found in the N terminus of gravin and in the mouse SSeCKS (4) ( Fig. 2A). We hypothesize each of these regions can act in a manner similar to the effector domain of MARCKS protein, i.e. bind electrostatically to acidic lipids on the inner leaflet of the plasma membrane and have this binding reversed by Ca 2ϩ / CaM (21). Analysis of the N-terminal domain of gravin using the programs PONDR (22) and FOLDINDEX (23) (as well as a simple coloring scheme: red, acidic; blue, basic; green, hydrophobic) suggests the entire region is "natively unfolded" (24). A, the sequence of human gravin was scanned for two motifs, the effector domain of the MARCKS protein and for CaM binding sites. Two other landmarks of gravin are displayed: the AKAP sites that constitute the ␤ 2 -adrenergic receptor binding domain; and the binding site for the RII subunit of PKA. These data were obtained using ProSite software. B, full-length HA-tagged, C-terminal and N-terminal truncated gravin mutants were generated for analysis of their ability to bind to CaM immobilized to Sepharose-4B in the presence of Ca 2ϩ . Upper panel C, HA-tagged gravin and mutant forms were expressed in A431 cells, whole cell lysates of the cells were subjected to SDS-PAGE, immunoblotted, and stained with anti-HA antibodies. Lower panel C, fragments of gravin that were retained by immobilized CaM matrix, released from the matrix using buffer containing EDTA, subjected to SDS-PAGE, immunoblotted and stained with anti-HA antibodies. The data presented are representative of at least three separate determinations, each performed with separate cell lysates.
That is, it has a high density of negatively charged residues and a paucity of either basic or hydrophobic residues (except for the three PCDs). This N-terminal region of gravin is similar to the MARCKS protein, which is natively unfolded and is highly negatively charged except for the single basic/hydrophobic effector domain. The three PCD regions of gravin, and the comparable region of AKAP79 ( 31 KASMLCFKRRKKAAKALKPKAG 52 ) (25), all contain both basic and hydrophobic residues, but lack acidic residues. Hence they would be predicted to bind electrostatically to the negatively charged inner leaflet of the plasma membrane and have the potential to bind Ca 2ϩ /CaM with biologically significant affinity. Although others have speculated that CaM may bind to sequences common to gravin and AKAP79, the role of Ca 2ϩ /CaM in regulation of AKAP-based signaling, localization, and function is not fully understood (4). The working hypothesis is that each of the three PCDs helps anchor gravin to the membrane (6), and that each binds Ca 2ϩ / CaM with significant affinity to reverse the membrane binding. PCDs help direct the proteins such as K-Ras4B to the plasma membrane rather than to internal membranes, as the former has a more negative electrostatic surface potential (26,27).
We first examined HA-gravin and HA-gravin mutants with targeted deletions of the basic putative Ca 2ϩ /CaM binding PCDs. The constructs were designed (Fig. 2B), expressed in A431 clones and tested to determine qualitatively if they were capable of binding Ca 2ϩ /CaM (Fig. 2C, upper panel). Fulllength gravin , the N terminus (1-362) harboring PCD1 and PCD2, as well as the larger gravin (1-652) C-terminal truncate harboring PCD1, PCD2, and PCD3 all displayed binding by immobilized CaM-Sepharose 4B (CaM-matrix). The 554 -938 region of gravin also displayed the capacity to bind to the CaM-matrix. In silico analysis of the primary sequence of the 554 -938 gravin fragment reveals a possible CaM-binding sequence (670 -694). Unlike PCD1, PCD2, and PCD3, the 670 -694 gravin region does not display the structural character of the effector domain of the MARCKS protein, did not display significant binding to lipid vesicles (not shown), and was not investigated further in this study of domains involved in membrane association of gravin. Gravin fragments PCD1, PCD2, and PCD3 that were retained on the CaM-matrix were released by chelation of buffer Ca 2ϩ with molar excess of EDTA (Fig. 2C, lower panel). The C-terminal region of the AKAP (840 -1783), in contrast, failed to bind to the immobilized CaM matrix (Fig. 2C). Similar experiments were performed with the three PCD peptides (minus flanking gravin sequences). Each peptide was successfully expressed in A431 cells and each showed qualitatively the same retention by the CaM matrix and release by exposure to molar excess of EDTA (data not shown). The affinity of these gravin fragments for Ca 2ϩ /CaM cannot be readily deduced from these CaM matrix experiments. A more quantitative test of the hypothesis was conducted by studying directly the binding of synthetic PCD peptides of gravin to phospholipid vesicles.
We synthesized radioloabeled peptides corresponding to gravin PCD1, PCD2, and PCD3 and measured their membrane and Ca 2ϩ /CaM binding affinities. The inner leaflet of a mammalian plasma membrane typically contains between 15-30% acidic lipid, mainly phosphatidylserine. Thus we measured the equilibrium binding to 5:1 and 2:1 phosphatidylcholine (PC)/phosphatidylserine (PS), large unilamellar vesicles (100 nm diameter, LUVs) using a centrifugation technique (8). The results of the membrane binding assay for PCD3 show that this peptide binds strongly to the PC/PS vesicles (Fig. 3A). The affinity increases with the mol fraction of negatively charged PS in the vesicles: it binds 50-fold more strongly to the 2:1 than to the 5:1 PC/PS vesicles. Similar measurements on vesicles containing 10 and 25% acidic lipid (not shown), when combined with the data in Fig. 3A, reveal the binding affinity increases exponentially with the fraction of acidic lipid in the membrane (binding energy increases linearly with fraction of PS), as expected theoretically for nonspecific electrostatic interactions (15) and observed experimentally for the MARCKS effector domain peptide (28) and other basic/hydrophobic peptides (21). PCD1 and PCD2 each bind with qualitatively similar affinities to 5:1 PC/PS vesicles as does PCD3 (Table 1). In other experiments (not shown), we observed that the affinity of PCD1 to bind PC vesicles that contain only 1% PIP2 was even higher than that for PC/PS vesicles that contain 17% PS (Table 1).  AUGUST 18, 2006 • VOLUME 281 • NUMBER 33

JOURNAL OF BIOLOGICAL CHEMISTRY 23937
Ca 2ϩ /CaM, but not apocalmodulin, binds to the PCD3 peptide with sufficient affinity to prevent its association with the membrane (Fig. 3B). In the absence of calmodulin, we chose conditions such that 90% of the peptide is bound to the vesicles or Kϫ [lipid] ϭ 10: note that 100 nM Ca 2ϩ /CaM removes 50% of the peptide from the membrane, which implies that K Ca/CaM ϫ [Ca 2ϩ /CaM] ϭ 10 or that the K d of the Ca 2ϩ /CaM peptide complex is ϳ10 nM. Specifically, the curve is drawn according to Equation 2 (see "Experimental Procedures") with K Ca/CaM ϭ 10 8 M Ϫ1 . The affinity of all three of the PCD peptides for Ca 2ϩ / CaM is very similar (Fig. 3, C and D). The Ca 2ϩ /CaM affinities (K d ϳ10 nM) for all three peptides (see Table 1) are sufficiently high to be of biological significance, i.e. strong enough to displace the protein from the membrane even though most of the Ca 2ϩ /CaM in cells may be bound rather than free in the cytosol (29,30).
Intracellular Calcium Regulates Membrane Localization of Gravin-The ability of Ca 2ϩ /CaM to neutralize the binding of gravin PCD1, PCD2, and PCD3 peptides to vesicles in vitro prompted study of the effects of changing intracellular concentrations of free Ca 2ϩ on localization of the AKAP gravin. To increase intracellular Ca 2ϩ , cells expressing native (i.e. myristoylated) gravin-GFP were treated with the Ca 2ϩ ionophore A23187 in the presence of normal extracellular Ca 2ϩ . In the absence of the Ca 2ϩ ionophore, gravin-GFP is observed throughout the cytoplasm (yellow arrowheads) and in abundance at close proximity to the cell membrane (white arrows, Figs. 1B and 4A, left panel). Thirty minutes after ionophore/ Ca 2ϩ treatment, cells assumed a more rounded morphology and gravin-GFP was found to redistribute uniformly throughout the cytoplasm (Fig. 4A, center panel). Thus, increasing intracellular concentration of Ca 2ϩ decreases the ability of gravin to localize to the cell membrane. Complementary studies in cells preloaded with the Ca 2ϩ -buffering agent BAPTA-AM ester demonstrated gravin localization prominently at the cell membrane (Fig.  4A, right panel). A 1-362 C-terminally truncated gravin harboring PCD1 and PCD2 (but not PCD3) was fused with the cyan fluorescent protein CFP (1-362 gravin-CFP, Fig. 4B). As noted for gravin, the 1-362 gravin-CFP displays localization to the membrane and to the cytoplasm. Treating cells with Ca 2ϩ ionophore provoked a sharp loss in the amount of cell membrane-associated 1-362 gravin-CFP. Buffering the intracellular Ca 2ϩ with BAPTA, in contrast, increased the amount of 1-362 gravin-CFP associated with the cell membrane, mimicking the Ca 2ϩdependent localization of the fulllength AKAP to the membrane (compare panels A and B, Fig. 4). These results show that two PCDs are sufficient to essentially mimic the membrane binding properties of the wild-type gravin protein, in agreement with independent work (6).
If membrane binding of gravin is required for its interaction with receptor, decreasing the membrane binding by increasing the level of [Ca 2ϩ ] i should inhibit the ability of gravin to recycle the receptor to the plasma membrane. Thus we tested the effects of increased [Ca 2ϩ ] i on the ability of cells that were desensitized with isoproterenol (10 M, 30 min) to display "resensitization" of cyclic AMP accumulation in response to a second stimulus of isoproterenol (10 M) following a washout of agonist at 30 (W30), 45 (W45), and 60 (W60) min of duration post washout (Fig. 4C, upper panel). In the control cells, the resensitization of the cyclic AMP response to a second stimulus of isoproterenol is ϳ80% complete within 60 min of washout of the agonist. For the cells treated with the Ca 2ϩ ionophore A23187, however, the resensitization is largely attenuated. Treatment with the ␣ 1 -adrenergic agonist phenylephrine (100 M), known to increase intracellular concentrations of Ca 2ϩ , also attenuates the resensitization of the cyclic AMP response (Fig. 4C, lower panel).
In wild-type A431 cells, ␤ 2 AR are internalized within 30 min of treatment with agonist (isoproterenol 10 M, ϩ ISO). The amount of cell surface ␤ 2 AR was established by binding assays using [ 3 H]CGP-12177. The amount of receptor on the cell surface declines sharply because of agonist-induced internalization of the receptor (Fig. 4D). The CGP binding studies comparing either ionophore-treated (A23187) or phenylephrinetreated cells to untreated, control cells demonstrate that increases in intracellular [Ca 2ϩ ] i stimulated by treatment with either Ca 2ϩ ionophore or by stimulating the cells with the ␣ 1 -agonist phenylephrine (100 M) do not affect the amount of agonist-induced internalization of the ␤ 2 AR at 30 min poststimulation with isoproterenol (Fig. 4D, 30 min). The recycling of the internalized receptor to the cell surface observed 60 min following agonist washout, in sharp contrast, was markedly attenuated by increasing intracellular [Ca 2ϩ ] i with ionophore or by stimulation of the cells with phenylephrine (Fig. 4D,  W60).
Gravin Requires Two PCDs to Function in Recycling of ␤ 2 -Adrenergic Receptors-The functional role of AKAP's membrane association and of these small, positively charged PCDs was explored. Gravin acts as a "tool box" for a variety of docking proteins, including protein kinases and phosphoprotein phosphatases, that are necessary for normal biology of the ␤ 2 AR (2). In cells expressing endogenous gravin, ␤ 2 AR are internalized within 30 min of treatment with agonist (isoproterenol 10 M, ϩ ISO). Binding of the cell-impermeant [ 3 H]CGP-12177 antagonist ligand again was used to quantify the amount of cell sur-face ␤ 2 AR (Fig. 5A). After a washout of agonist and a recovery period, CGP binding in control (Control) cells returns to normal within 60 min, as the ␤ 2 ARs are recycled back to the cell membrane. In cells made deficient of gravin by knockdown with antisense morpholinos (gravin-KD), ␤-adrenergic agonists stimulate desensitization/internalization of ␤ 2 ARs normally (Fig. 5A), whereas resensitization/recycling of the receptor to the membrane is lost (Fig. 5A) (3, 31, 32).
To probe functional roles of PCDs in this AKAP, we deleted pairs of PCDs from HA-tagged gravin, either PCD1 and PCD2 (⌬PCD1,2), or PCD2 and PCD3 (⌬PCD2,3), or PCD1 and PCD3 (⌬PCD1,3). The deletion mutants of gravin were expressed in gravin-KD cells and ␤ 2 AR recycling to the cell membrane measured following agonist-induced internalization (Fig. 5,  A-C). Depletion in endogenous gravin as well as expression of exogenous gravin were demonstrated by SDS-PAGE of whole cell lysates and subsequent immunoblotting (Fig. 5, A-C). ␤ 2 AR internalized in response to agonist recycles to the cell membrane in the control cells expressing endogenous gravin (wild-type, WT; control scrambled sequence morpholino, Control), but not in the cells in which gravin is made deficient (antisense morpholinos to gravin, gravin KD). Expression of the HA-tagged gravin⌬PCD1,2 in the gravin-KD cells (gravin KD ϩ ⌬PCD1,2) fails to rescue the inability of the gravin-KD cells to recycle the agonist-treated, internalized receptors back to the cell membrane. Expression of a wild-type gravin in the gravin KD cells (gravin KD ϩ WT gravin), however, enables ␤ 2 AR recycling (Fig. 5A). Expression of the deletion mutants HA-tagged gravin(⌬PCD2,3; Fig. 5B) and HA-tagged gravin(⌬PCD1,3; Fig. 5C), like the (⌬PCD1,2; Fig. 5A) gravin mutant, displayed the same inability to recycle internalized ␤ 2 AR, as measured by CGP binding. Thus, the ability of gravin to function in receptor resensitization and recycling is lost when any pair of the PCDs is deleted in non-myristoylated gravin, suggesting that the localization of the gravin scaffold to the inner leaflet of the cell membrane is essential to the role of gravin in receptor recycling.
Confocal microscopy imaging of eGFP-tagged ␤ 2 AR confirmed the data obtained with the [ 3 H]CGP-12177 binding (Fig.  6A), i.e. loss of two PCDs virtually abolishes the ability of the gravin to function in receptor recycling. ␤ 2 AR-eGFP located in the cell membrane is labeled with white arrows, whereas internalized ␤ 2 AR-eGFP is labeled with yellow arrowheads. In gravin-deficient cells, expression of wild-type gravin, but not ⌬PCD1,2-gravin, rescues the ability of these cells to recycle internalized receptors to the cell membrane. Deletion of a single PCD alone (i.e. ⌬PCD3) from gravin, in contrast, did not significantly alter the localization or the function of gravin (data not shown). The simplest interpretation is that two positively charges clusters (i.e. PCD1, PCD2, or PCD3) are required in tandem for membrane association and function of the HA-tagged gravin.

Binding of gravin PCDs to artificial lipid vesicles
Binding of peptides corresponding to the indicated sequences of gravin either to membranes or to Ca 2ϩ /CaM. K is the molar partition coefficient for the binding of the peptides to 5:1 PC/PS vesicles, determined from Equation 1 as illustrated in Fig.  3A for gravin(510 -536), and similar experiments (not shown) for the other peptides. Gravin-(171-187) binds to 99:1 PC/PIP 2 vesicles (K ϭ 1 ϫ 10 5 M Ϫ1 ) with 10-fold higher affinity than to 5:1 PC/PS vesicles. Other basic peptides exhibit a similar high affinity for polyvalent PIP 2 vs monovalent PS (13), as discussed elsewhere (21). K Ca/CaM is the affinity of the peptide for Ca 2ϩ /CaM determined from Equation 2 (see "Experimental Procedures") using the competition assays illustrated in Fig. 3, B-D. The sequences of the peptides are shown in Fig. 2.

DISCUSSION
In the current work we explored how gravin PCDs bind lipid vesicles, whether or not Ca 2ϩ /calmodulin alters gravin binding to membranes, and how membrane association impacts gravinmediated ␤ 2 ARs resensitization and recycling after agonist-induced desensitization/internalization. Binding of gravin to the receptor under unstimulated conditions would appear too weak to occur unless gravin is bound to the inner leaflet of the cell membrane (Fig. 7). When gravin is bound to the cell membrane, the receptor-gravin interaction would benefit by a ϳ1000-fold increase in the local concentration of the scaffold with the membrane-embedded ␤ 2 AR (Fig. 7A). Agonist binding to the ␤ 2 AR activates the stimulatory adenylyl cyclase pathway and the PKA tethered to gravin (3). Activated PKA phosphorylates both the scaffold and receptor, dramatically enhancing their protein-protein interactions (Fig. 7B). Although not required itself for gravin function, myristate may facilitate further the interactions between the membrane and two PCDs required for binding. We cannot rule out the potential importance of the myristate in either membrane anchoring or lateral distribution, although the ␤and ␥-isoforms of gravin, lacking N-terminal myristoylation, appear to localize as does gravin (6). FIGURE 6. Recycling of ␤ 2 AR requires gravin with two intact PCDs. A, A431 cells were untreated (control) or treated with antisense morpholinos to knockdown the expression of endogenous gravin (gravin KD). The antisense morpholino-treated cells were employed as such (gravin KD) or transiently transfected with an expression vector harboring either an HA-tagged gravin mutant deficient of PCD1, PCD2 (⌬PCD1,2) or an HA-tagged full-length, wild-type gravin (WT gravin). Cells were desensitized with ␤-adrenergic agonist (isoproterenol 10 M, ϩIso) for 30 min and localization of ␤ 2 AR-eGFP determined by confocal microscopy. ␤ 2 AR-eGFP located in the cell membrane is labeled with white arrows, whereas internalized ␤ 2 AR-eGFP is labeled with yellow arrowheads. Full-length and the ⌬PCD1,2 mutants of HA-tagged gravin were generated as schematic representation of topological organization of the human gravin fragment (Fig. 2). B, to detect ␤ 2 -adrenergic receptor-gravin association, we incubated whole cell lysates of A431 cells either expressing HA-tagged full length, wild-type gravin (HA-gravin), or expressing the ⌬PCD1,2 mutant version of HA-tagged gravin (HA-gravin ⌬PCD1,2) with antibodies against either the HA tag (anti-HA tag) or the ␤ 2 -adrenergic receptor (anti-␤ 2 AR, CM-4) covalently linked to protein A/G-agarose beads. The immune complexes were subjected to SDS-polyacrylamide gel electrophoresis, transferred to nitrocellulose, and probed with an antibody specific for either the ␤ 2 -adrenergic receptor or the HA tag. The data presented are representative of at least three separate determinations performed on separate occasions. IP, immune precipitation; IB, immunoblotting.
Our localization studies with gravin-GFP, compare well with those of gravin-␤-GFP (6). We measured quantitatively the binding affinities of peptides corresponding to each PCD in gravin to phospholipid vesicles and Ca 2ϩ /CaM. (As with the basic region of MARCKS, the PCDs in gravin probably bind to membranes with only about half the energy of the PCD peptides because flanking acidic residues in the protein reduce electrostatic interactions.) By functional studies we were able to show that the presence of two PCDs is essential to the ability of the non-myristoylated gravin to provoke ␤ 2 AR resensitization/recycling following classical agonist-induced desensitization. As with earlier studies (6), the current study does not reveal how many PCDs are essential to anchor myristoylated isoforms of gravin to the membrane, in enabling ␤ 2 AR recycling. Myristate itself does not provide enough energy to anchor a protein to the plasma membrane (14): both Src and MARCKS require a PCD in addition to myristate. Thus, we are confident that at least one, and possibly two, PCD might be required for membrane anchoring of the myristoylated forms of gravin. In this regard, we did demonstrate that Ca 2ϩ /CaM produce translocation of native myristoylated gravin from membrane to the cytosol and block ␤ 2 AR recycling.
Elevating intracellular Ca 2ϩ decreases the association of gravin with the cell membrane and blocks the ability of gravin to resensi-tize and recycle the ␤ 2 AR. Responses integrating signals provoked by changes in intracellular concentrations of cyclic nucleotides and free Ca 2ϩ often are triggered in vivo by neurotransmitters such as norepinephrine, operating through both ␣ 1 -and ␤ 2 -adrenergic receptors. Catecholamines epinephrine and norepinephrine stimulate both ␤ 2 -adrenergic receptors (elevating cyclic AMP levels and activating PKA) and ␣ 1 -adrenergic receptors (increasing intracellular Ca 2ϩ concentrations transiently). Gravin is shown to be a scaffold integrating signaling via cyclic nucleotides and intracellular Ca 2ϩ concentrations, ultimately favoring trafficking of the scaffold away from the cell membrane during a Ca 2ϩ transient (Fig.  7B), modulating a process that appears essential for receptor recycling (3,32). The fall of intracellular [Ca 2ϩ ] (after the Ca 2ϩ transient) would result in a reduction in the neutralization of PCDs by Ca 2ϩ / CaM (Fig. 7C), restoring the movement of gravin from the cytoplasmic compartment to the cell membrane. A recent model for NMDA receptor regulation of MAGUK-AKAP79/ 150-based signaling likewise suggests the operation of a Ca 2ϩ -sensitive association of the scaffold with the cell membrane (33,34). These two studies highlight the central role of short, positively charged domains and Ca 2ϩ in regulating the localization of AKAPs, such as gravin and AKAP79/150, at the cell membrane and provide further information on the broader topic of how multivalent scaffolds like gravin and AKAP79/150 integrate signals emanating from different signaling cascades.
Earlier mutagenic studies of the non-myristoylated ␤and ␥-isoforms of AKAP12 suggested that two of the three PCDs in the gravin-␥ N-terminal region are necessary for plasma membrane association, as determined by confocal microscopy (6). Deleting any one cluster decreased membrane association modestly, deleting any two clusters decreased membrane association further (6). Substitution of acidic residues for Ser and Thr residues in these basic regions of gravin-␥, mimicking PKC phosphorylation, also reduced apparent association with the membrane (6). The current work extends these earlier findings (6) by an investigation of the membrane binding and role of PCDs in the function of this AKAP scaffold.
For MARCKS, either phosphorylation by PKC or Ca 2ϩ /CaM binding to the effector domain stimulates translocation of MARCKS from the plasma membrane to the cytosol (12,21). The N-terminal region of gravin and the entire MARCKS protein dis- FIGURE 7. Schematic of gravin association with the cell membrane and ␤ 2 -adrenergic receptor. A, fulllength gravin association with the inner leaflet of the cell membrane in the absence of activation of the ␤-adrenergic receptor by agonist. Note that at resting concentrations of intracellular Ca 2ϩ small, positively charged domains PCD1, PCD2, and PCD3 bind avidly to the inner leaflet of the cell membrane rich in PS. Under these conditions, some association of the scaffold with the receptor via the RBD can be measured directly. B, upon stimulation of the cell with an agent such as catecholamine there is an increase in the intracellular concentration of cyclic AMP, activation of PKA and phosphorylation of PKA sites on the receptor and the scaffold in the RBD (3). The increase in intracellular Ca 2ϩ concentrations neutralizes the membrane association of these three domains enabling the binding of CaM, repelling the scaffold from the cell membrane as the phosphorylated receptor and RBD of gravin enhance the interaction of the scaffold harboring PKA, PKC, and the phosphatase PP2B. C, following the transient rise and fall of intracellular Ca 2ϩ levels, the small, positively charged domains drive membrane association of the scaffold again and enable the resensitization (dephosphorylation) and recycling of the receptor as well as dephosphorylation of the AKAP. play a highly acidic, apparently unfolded region with one (MARCKS protein) or three (gravin) basic/hydrophobic domains that target the protein to the plasma membrane through electrostatic interactions with acidic lipids. We show (Fig. 3) that Ca 2ϩ /CaM can bind with sufficiently high affinity (K d ϳ10 nM) to each of the PCDs to release these peptide sequences from the membrane. Similarly, inserting negatively charged residues into PCDs to mimic PKC phosphorylation decrease interactions between PCDs and the cell membrane, producing translocation of AKAP from the membrane to the cytosol (6). In summary, our measurements show the peptides corresponding to the PCD1, PCD2, and PCD3 regions of gravin bind reversibly to phospholipid vesicles, in a Ca 2ϩ /CaM-sensitive manner. These biophysical measurements and the confocal microscopy results support our hypothesis that local concentrations of Ca 2ϩ (and thus Ca 2ϩ / CaM) can regulate association of the AKAP gravin (via these basic regions) with the inner leaflet of the plasma membrane.
The first 150 residues of AKAP79 include three clusters of basic residues proposed to function as a "polybasic membrane targeting domain" (11). In AKAP79, the sequence 31 KASMLCFKRRK-KAAKALKPKAG 52 , also has been directly implicated as a CaM binding domain (25). Previous studies of the regulation of interactions of AKAP79 with protein kinase C, phosphatidylinositol 4,5bisphosphate, F-actin, and cadherins by protein phosphorylation and by Ca 2ϩ /CaM support the tenet that the N-terminal, three PCDs in the membrane targeting domain of this AKAP are negatively regulated by Ca 2ϩ /CaM binding (10,25,33,35). Gravin uses three types of interactions to anchor itself to the plasma membrane: (i) the comparatively weak hydrophobic insertion of the N-terminal myristate for some isoforms, (ii) electrostatic interaction of the three PCDs with acidic lipids, and (iii) protein-protein interactions with the GPCR ␤ 2 AR. The protein-protein interaction of scaffold with the receptor (i.e. mediated by RBD binding) is dynamic and regulated by protein phosphorylation/dephosphorylation (3). The PCD-membrane interaction is dynamic and regulated by Ca 2ϩ /CaM binding to the PCDs. The PCDs in gravin function to increase the local concentration of the AKAP at the cell membrane and thereby enhancing the protein-protein interactions between the scaffold and this GPCR (36). Our observations demonstrate that gravin functions as a scaffold integrating signals from two major pathways, ␤ 2 AR/G s /AC/cyclicAMP and ␣ 1 AR/G q /PLC␤/Ca 2ϩ .