Mechanism for Targeting the A-kinase Anchoring Protein AKAP18δ to the Membrane*

Background: AKAP18δ is anchored to the membrane by a yet unidentified mechanism. Results: Electrostatic attraction to negatively charged lipids is sufficient for membrane localization. Conclusion: The positively charged amino acids of AKAP18δ are spatially arranged into a lipid-binding plane. Significance: Targeting to aquaporin-containing vesicles requires additional protein-protein interactions. A-kinase anchoring proteins (AKAPs) are a family of scaffolding proteins that target PKA and other signaling molecules to cellular compartments and thereby spatiotemporally define cellular signaling events. The AKAP18 family comprises AKAP18α, AKAP18β, AKAP18γ, and AKAP18δ. The δ isoform targets PKA and phosphodiesterase PDE4D to AQP2 (aquaporin-2)-bearing vesicles to orchestrate the acute regulation of body water balance. Therefore, AKAP18δ must adopt a membrane localization that seems at odds with (i) its lack of palmitoylation or myristoylation sites that tailor its isoforms AKAP18α and AKAP18β to membrane compartments and (ii) the high sequence identity to the preferentially cytoplasmic AKAP18γ. Here, we show that the electrostatic attraction of the positively charged amino acids of AKAP18δ to negatively charged lipids explains its membrane targeting. As revealed by fluorescence correlation spectroscopy, the binding constant of purified AKAP18δ fragments to large unilamellar vesicles correlates (i) with the fraction of net negatively charged lipids in the bilayer and (ii) with the total amount of basic residues in the protein. Although distantly located on the sequence, these positively charged residues concentrate in the tertiary structure and form a clear binding surface. Thus, specific recruitment of the AKAP18δ-based signaling module to membranes such as those of AQP2-bearing vesicles must be achieved by additional mechanisms, most likely compartment-specific protein-protein interactions.

A-kinase anchoring proteins (AKAPs) are a family of scaffolding proteins that target PKA and other signaling molecules to cellular compartments and thereby spatiotemporally define cellular signaling events. The AKAP18 family comprises AKAP18␣, AKAP18␤, AKAP18␥, and AKAP18␦. The ␦ isoform targets PKA and phosphodiesterase PDE4D to AQP2 (aquaporin-2)-bearing vesicles to orchestrate the acute regulation of body water balance. Therefore, AKAP18␦ must adopt a membrane localization that seems at odds with (i) its lack of palmitoylation or myristoylation sites that tailor its isoforms AKAP18␣ and AKAP18␤ to membrane compartments and (ii) the high sequence identity to the preferentially cytoplasmic AKAP18␥. Here, we show that the electrostatic attraction of the positively charged amino acids of AKAP18␦ to negatively charged lipids explains its membrane targeting. As revealed by fluorescence correlation spectroscopy, the binding constant of purified AKAP18␦ fragments to large unilamellar vesicles correlates (i) with the fraction of net negatively charged lipids in the bilayer and (ii) with the total amount of basic residues in the protein. Although distantly located on the sequence, these positively charged residues concentrate in the tertiary structure and form a clear binding surface. Thus, specific recruitment of the AKAP18␦-based signaling module to membranes such as those of AQP2-bearing vesicles must be achieved by additional mechanisms, most likely compartment-specific protein-protein interactions.
AKAP18␦ is involved in maintaining body water homeostasis. The rapid increase in water permeability of renal collecting ducts in response to antidiuretic hormone (arginine-vasopressin) is achieved through the fast increase of AQP2 (aquaporin-2) abundance in the plasma membrane, which results from the exocytic insertion of AQP2-containing intracellular vesicles (1). The underlying signal cascade starts with the binding of arginine-vasopressin to the vasopressin V2 receptor of renal principal cells. Stimulation of the V2 receptor leads to a rise in cAMP (2). The subsequent activation of PKA results in AQP2 phosphorylation, which, in turn, initiates an AQP2 redistribution from vesicles distributed throughout the cell to the apical plasma membrane (2).
PKA consists of a dimer of regulatory (R) 2 subunits (RI␣, RI␤, RII␣, and RII␤) and two catalytic subunits, each of which binds to an R subunit. Thereby, the R subunits maintain the catalytic subunits in an inactive state. Upon binding of two molecules of cAMP to each R subunit, the catalytic subunits are released and phosphorylate substrates in close proximity. A-kinase anchoring proteins (AKAPs) direct PKA to defined cellular sites to limit the kinase's access to a subset of substrates (3). This compartmentalization of PKA by AKAPs is crucial for various biological processes, including the arginine-vasopressin-induced redistribution of AQP2 (3)(4)(5). All canonical AKAPs possess an amphipathic helix, which directly binds PKA. The helix interacts with the dimerization and docking (D/D) domain formed by the dimerized R subunits. AKAP18␦ (353 amino acids) is the largest member of the AKAP18 family, which comprises AKAP18␣, AKAP18␤, AKAP18␥, and AKAP18␦ (6). AKAP18␦ tethers PKA to AQP2-bearing vesicles as shown by analysis of immunoisolated vesicles from principal cells of the inner medullary collecting duct (6,7). In addition, AKAP18␦ directly interacts with the phosphodiesterase PDE4D (7). This signaling module, consisting of AKAP18␦, PKA, and PDE4D, controls the local cAMP level and thus PKA activity in the vicinity of AQP2. Thereby, it participates in the control of the localization of AQP2 and thus in the control of arginine-vasopressin-mediated water reabsorption (7). In cardiac myocytes, AKAP18␦ directly interacts with phospholamban. This interaction is involved in Ca 2ϩ reuptake into the sarcoplasmic reticulum and thereby participates in the regulation of cardiac myocyte contractility (8).

* This work was supported by the Upper Austrian Government and Deutsche
The shorter isoforms of the AKAP18 family, AKAP18␣ (81 amino acids) and AKAP18␤ (104 amino acids), are targeted to plasma membranes by (i) myristoylation and palmitoylation and (ii) binding to ion channels such as Na ϩ channels (9,10). In contrast, there is no evidence for specific binding of AKAP18␦ to membrane channels (e.g. AQP2). This agrees with the observation made for another AKAP, gravin. Its myristoylation sites are not required for membrane localization or function (11). Gravin possesses short positively charged domains (the so-called MARCKS (myristoylated alanine-rich C-kinase substrate) effector domain-like regions), which serve to electrostatically link the protein with the membrane. However, similar domains are not present in the AKAP18␦ primary structure (supplemental Fig. 1). The only domain with basic amino acids is the so-called conserved nuclear localization sequence between positions 78 and 86. All of the other positively charged residues are (i) scattered throughout the protein and (ii) well balanced by acidic residues (supplemental Fig. 1). The nuclear localization sequence is unlikely to be responsible for the membrane localization of AKAP18␦. It is also part of AKAP18␥, another splice variant (324 amino acids) that is localized mainly in the cytoplasm (12). AKAP18␥ is 26 amino acids shorter and ϳ75% identical to AKAP18␦ (6). The additional N-terminal sequence (amino acids 1-26) of AKAP18␦ has a net negative charge of Ϫ4, and its electrostatic interactions with the membrane are thus unlikely to be responsible for the difference in cellular location between AKAP18␦ and AKAP18␥.
Although not localized in well defined domains, the tertiary structure of AKAP18␦(87-292) (13) reveals numerous positively charged amino acids that are concentrated into one plane (Fig. 1). We speculate that this arrangement may represent the binding plane, i.e. that the number of positive charges in that plane is large enough to attract the protein to negatively charged membranes. This would explain how AKAP18␦ adopts its membrane localization. Membrane anchoring is mandatory because AKAP18␦ would otherwise be unable to orchestrate the regulation of water balance in renal principal cells and Ca 2ϩ reuptake into the sarcoplasmic reticulum of cardiac myocytes.
To test whether electrostatic interactions of AKAP18␦ with lipids may be sufficient for membrane recruitment, we purified AKAP18␦ fragments that contained different numbers of charged residues and monitored their interaction with lipid bilayers. We found that the binding energy increased with the total length of the fragments, even though no specific targeting domains were identified. The mere increase in the number of positively charged residues was sufficient.

EXPERIMENTAL PROCEDURES
Lipid Vesicle Preparation-Large unilamellar vesicles (LUVs) were prepared from a lipid mixture in chloroform. The lipids were dried on a rotary evaporator, hydrated in a solution containing 300 mM NaCl and 20 mM HEPES buffered at pH 7.3, and put through 21 extrusion cycles stacked with two polycarbonate filters with 400-and 100-nm pore sizes using a miniextruder from Avanti Polar Lipids. The final stock solution subsequently contained 35 mg/ml lipids.
Protein Expression and Purification-AKAP18␦ versions encoding amino acids 87-292, 1-292, and 76 -353 were generated as N-terminally His 6 -tagged fusion proteins using the vector pET-30 Ek/LIC and Escherichia coli strain Rosetta(DE3). To improve the solubility of AKAP18␦(76 -353), the vector was coexpressed with a plasmid (pET-46) encoding the AKAPbinding domain of the RII␣ subunits of PKA (D/D domain, amino acids 1-44) or full-length RII␣, both lacking the His 6 tag. Precultures were grown overnight in 50 ml of LB medium (25-37°C) and centrifuged, and cell pellets were resuspended and used to inoculate up to four 500-ml cultures in Overnight Express TM Instant TB medium (autoinduction medium; Novagen). Cultures were grown for at least 24 h at 25-37°C and 110 rpm. Cells were harvested by centrifugation, washed with PBS, and stored at Ϫ80°C. Cells were thawed on ice, resuspended in extraction buffer (40 mM phosphate buffer (pH 7.5), 300 mM NaCl, 5-10 mM imidazole, 5 mM ␤-mercaptoethanol, 0.5 mM PMSF, one protease inhibitor tablet (Roche Applied Science), and 5 l/100 ml Benzonase), and lysed using a French press. Cell debris was removed by centrifugation (21,000 rpm, 4°C, 30 min) and filtration through a 0.45-m filter. The cleared supernatant was applied to a Talon cobalt affinity column (Clontech) pre-equilibrated with 40 mM phosphate buffer (pH 7.5), 300 mM NaCl, and 5-10 mM imidazole in an Ä KTA system. The Talon column was washed with equilibration buffer, and protein was eluted with 40 mM phosphate buffer (pH 7.5), 300 mM NaCl, and an imidazole gradient up to 300 mM. Fractions containing AKAP18␦ truncations were pooled (supplemental Fig. 2). If necessary, the pool was dialyzed against 20 mM Tris-HCl and 20 mM NaCl (pH 8.0), loaded onto a Q-Sepharose FF anion exchange column, and eluted with a gradient of up to 1 M NaCl in 20 mM Tris-HCl (pH 8.0). Each final polishing step involved gel filtration with Superdex 75 (GE Healthcare) in 20 mM HEPES and 300 mM NaCl (pH 7.0). Pure fractions were pooled, supplemented with 20% glycerol, and stored at Ϫ80°C.
Fluorescence Correlation Spectroscopy (FCS)-FCS is an accurate and valuable means for quantifying the interaction of proteins with lipid bilayers (16). We first monitored bulk diffusion of AKAP18␦ constructs (ConfoCor 3 attached to an LSM 510 META laser scanning microscope, both from Carl Zeiss, Jena, Germany) (17). tris-NTA(Atto565) was excited at 561 nm. We calibrated the confocal volume by measuring the residence time ( R ) of rhodamine 6G in solution. On the basis of a diffusion coefficient of 426 m 2 s Ϫ1 (18), we arrived at confocal plane radii (r 0 ) of 0.22 and 0.25 m for the two wavelengths 561 and 633 nm.
We obtained the number of particles (N) of each diffusing species from the autocorrelation function (G()) of the fluorescence temporal signal from DilC 18 (5)-DS-labeled lipid vesicles or tris-NTA(Atto565)-labeled proteins (Equation 1), where j and z 0 are the number of different species and the depth of the confocal volume (in the direction of the laser beam), respectively. Recordings (5 ϫ 30 s/measurement) were made in a small self-made measurement chamber (Ͻ100 l) 100 m above the glass slide to estimate the fraction of lipid-bound proteins. The measurement chamber consisted of a plastic cylinder that was glued (Norland optical adhesive NOA63) to a glass slide. To exclude unwanted protein and dye adsorption to the chamber walls at nanomolar protein and dye concentrations, we preincubated the chamber with 1 mg/ml BSA (Sigma) overnight and rinsed it four times with measurement buffer before use. We started the titration experiments with 70 nM AKAP18␦ to be well above the detection limit of FCS. We fitted a standard diffusion model for j ϭ 3 to the FCS autocorrelation curves, where we fixed four out of six parameters. These four parameters, namely the diffusion times of (i) free tris-NTA(Atto565) (45 s), (ii) labeled AKAP18␦ versions (180 -220 s), (iii) labeled AKAP18␦ versions bound to lipid vesicles (2.5 ms), and (iv) the fraction of free tris-NTA(Atto565) compared with protein-bound tris-NTA(Atto565), were determined in 12 independent experiments.
Determination of Binding Constants and Binding Energies from Experimental Measurements-Assuming Langmuir adsorption, we quantified protein binding to lipid bilayer membranes using the molar partition coefficient (K) (19). K is the proportionality factor between the mole fraction of AKAP18␦ versions bound to the membrane () and the concentration of AKAP18␦ versions free in bulk solution ( where N A is Avogadro's number, k B is the Boltzmann constant, and T is the temperature. Equation 6 can be used to convert the molar partition coefficients gained by Equation 5 into energy units, the binding free energy (⌬G).

RESULTS
Diffusion of single fluorescent particles in and out of the focus of a laser beam gives rise to intensity fluctuations of the emitted light. We first recorded these fluctuations from the free tris-NTA(Atto565) dye in an aqueous solution using the FCS extension of a laser scanning microscope. Fluctuation analysis in terms of autocorrelation functions (see Equation 1) allowed determination of the residence time of the dye in the focus to be equal to R,1 ϭ 45 s. Linkage of the small dye to the much larger AKAP18␦(1-292) fragment was expected to increase the residence time. In line with these expectations, the experimentally determined residence time ( R,2 ) of the labeled protein was equal to 180 -220 s (Fig. 2, black line). The binding of the AKAP18␦(1-292)⅐tris-NTA(Atto565) complex to the even larger liposomes should result in a further decrease in diffusion, which the residence time of R,3 ϭ 2.5 ms confirmed (Fig. 2, cyan line). In control experiments, we demonstrated that the free tris-NTA(Atto565) dye did not bind to the vesicles (supplemental Fig. 4).
At intermediate lipid concentrations (Fig. 2, red, green, blue, and pink lines), protein binding was incomplete, i.e. the resulting autocorrelation curves were a superposition of (i) protein diffusing free in solution, (ii) protein bound to vesicles, and (iii) free dye diffusing in solution. Setting j ϭ 3 in Equation 1 and fitting it to the autocorrelation curves allowed determination of the number of particles N 1 , N 2 , and N 3 of tris-NTA(Atto565), AKAP18␦(1-292)-bound tris-NTA(Atto565), and lipid vesiclebound AKAP18␦(1-292)⅐tris-NTA(Atto565), respectively. To reduce the error in the assessment of the lipid-bound AKAP18␦ fraction, N 3 /(N 3 ϩ N 2 ) ϭ [A] m /[A] tot , we fixed R,1 , R,2 , R,3 , and N 1 / (N 1 ϩ N 2 ϩ N 3 ). N 1 / (N 1 ϩ N 2 ϩ N 3 ) is the fraction of the free dye. It was determined at the beginning of each lipid titration experiment (i.e. in the absence of the lipid). The above procedure assumes that the diffusing species do not vastly differ in their brightness. This requirement is fulfilled (i) if all of them are holding just 1 dye molecule, which (ii) changes neither its adsorption spectra nor its quantum yield upon binding to the membrane.
To verify that this must have been the case, we analyze the worst case, i.e. the situation with the highest number of dyes per vesicle. The plot of [A] m /[A] tot against [L] acc (Fig. 3) shows that the lowest lipid concentration corresponds to this situation because it indicates the highest protein-to-lipid ratio of 7 ϫ 10 Ϫ8 /1.5 ϫ 10 Ϫ4 ϭ 4.7 ϫ 10 Ϫ4 . Assuming that one lipid occupies ϳ68 Å 2 , we arrive at 4.6 ϫ 10 4 accessible lipids per vesicle and at an outer leaflet area of ϳ3.1 ϫ 10 6 Å 2 for 100-nm lipid vesicles, i.e. we added 21.6 (4.6 ϫ 10 4 ϫ 4.7 ϫ 10 Ϫ4 ) proteins per vesicle into the aqueous solution. 30% of these AKAPs (compare Fig. 3) were bound to the vesicular membrane, so the average vesicle held seven AKAPs. This transforms into 0.7 dyes per vesicle because only 10% of these proteins were labeled. Thus, it is safe to conclude that the diffusing species, which entered the autocorrelation analysis, all had just one fluorescent label.
Decreasing the fraction of acidic lipids in the vesicular membrane also decreased the fraction of the protein, which was bound to the lipids (Fig. 3). For a quantitative analysis, we fitted Equation 5 to the data shown in Fig. 3. The only fitting parameter was the molar partition coefficient (K). For 58, 20, and 0 mol % DPhPG, K was equal to 2520 Ϯ 450, 350 Ϯ 60, and 20 Ϯ 13 M Ϫ1 , respectively (Fig. 4), indicating that binding is driven by electrostatic attraction. Besides electrostatics, other factors such as lipid packing density may affect binding. We tested this assumption by substituting DPhPG for natural brain PLE. The resulting K ϭ 2160 Ϯ 570 M Ϫ1 was close to the K value for 58 mol % DPhPG (Fig. 4) even though only 23.6% (w/w) (18.5% (w/w) phosphatidylserine and 4.1% (w/w) phosphatidylinositol; manufacturer's information) of the lipids are charged.
We verified the acidic lipid content of extruded brain PLE LUVs by comparing the electrophoretic mobility of these LUVs with that of LUVs from synthetic lipids of known composition. Because particle velocity depends on the electric field strength, the so-called -potential served as the output parameter. indi-  cates the electric potential at a distance of 2-4 Å from the vesicle surface depending on whether one or two layers of adsorbed water molecules and ions move along with the vesicle. At Ϫ40.2 Ϯ 2 mV, of brain PLE vesicles was smaller than the value of Ϫ42 Ϯ 2 mV for LUVs containing 30% DPhPG and 70% DPhPC. Extruded DPhPC/DPhPG LUVs with 10 and 60 mol % DPhPG exhibited values of Ϫ23.5 Ϯ 2 and Ϫ52.5 Ϯ 3 mV, respectively, in 50 mM KCl.
We also performed binding experiments with two more AKAP18␦ fragments, namely AKAP18␦(87-292) and AKAP18␦(76 -353). AKAPs interact with dimers of RII subunits, which are formed through interactions of the N-terminal 45 amino acids of each protomer. The dimerized N termini form the so-called D/D domain of RII␣. Either the D/D domain or full-length RII␣ was added to AKAP18␦(76 -353) to exclude the possibility that we mistook basic residues involved in protein-protein interaction for those residues that are responsible for lipid binding. These experiments revealed a linear dependence of ln K on the number of positively charged residues per construct for AKAP18␦(87-292), AKAP18␦(1-292), and AKAP18␦(76 -353)/DD (Fig. 5).
This is in agreement with the observation made for oligopeptides that ⌬G (or ln K; see Equation 6) scales with the amount of positively charged residues. The surprising observation is that part of a soluble protein, i.e. the D/D domain, increases membrane affinity. The D/D domain offers a pocket that accommodates the hydrophobic interface of an amphipathic helix of AKAP18␦(76 -353). The outer edges of the binding crevice are acidic and thus could contribute to the electrostatic interaction with the membrane (21,22). The remainder of RII␣ adds little to the binding affinity, as would be expected for a water-soluble protein (Fig. 5).
To further prove the electrostatic nature of the AKAP18␦membrane association, we screened the surface charges of both membrane and protein by increasing the salt concentration from 300 to 800 mM NaCl. The decreased surface potentials of membrane and protein resulted in significantly decreased membrane affinity of AKAP18␦ fragments, thereby confirming the crucial role of electrostatics in the binding process. For instance, the fraction of vesicle-bound AKAP18␦(76 -353)/ RII␣ was halved (Fig. 6).
Structural investigations of AKAP18␦ revealed that the central domain can bind 5Ј-AMP (13). The binding of 5Ј-AMP to AKAP18␦ neutralizes a region of positive charge at the base of the binding groove, and both hydroxyl groups of the 5Ј-AMP become available for interaction. The 5Ј-AMP-mediated change in the charge and shape of the molecule could potentially be involved in regulating AKAP18␦-membrane affinity. However, the addition of 5 mM 5Ј-AMP to AKAP18␦(76 -353)/ RII␣ did not induce a significant change in K (supplemental Fig.  3). In addition, 5Ј-AMP apparently did not change the location of full-length AKAP18␦ overexpressed in HEK293 cells (data not shown).

DISCUSSION
We have defined a hitherto unobserved mechanism through which a scaffolding protein can anchor to membranes. The affinity is provided solely by electrostatic attraction of amino acid residues, which, although distantly located on the sequence, concentrate in the tertiary structure to form a binding surface. Neither lipid anchors nor interactions with other proteins are required.
In contrast to other AKAPs (11,23), the charged amino acids are not bundled in domains. This conclusion stems from the investigation of four different AKAP18␦ versions (AKAP18␦(87-292), AKAP18␦(1-292), AKAP18␦(76 -353)/ DD, and AKAP18␦(76 -353)/RII␣). AKAP18␦(87-292) lacks the highly positively charged sequence between positions 76 and 86 that is present in AKAP18␦(1-292). To identify a possible role for PKA in modifying binding affinity, we co-purified AKAP18␦(76 -353), which comprises the PKA-binding domain, with the D/D domain of RII␣ or full-length RII␣ subunits of PKA. Only the acidic residues of the RII␣-binding crevice (21) increased membrane affinity. The remainder of the water-soluble RII␣ had little effect on K. K of the other three AKAP versions is an exponential function of their total number (m) of positively charged amino acids (Table 1).  AKAP18␦(1-292). Measurements of 0 (white bar), 20 (gray bar), and 58 (black bar) mol % DPhPG are compared with brain PLE (cross-hatched bar) as a natural mixture. The acidic lipid content of brain PLE was calculated from the manufacturer's information and cross-checked via -potential control measurements of lipid vesicles. This result is in line with the expected dependence of ⌬G on m. However, the interaction energy per residue is smaller, as has been observed in experiments with short positively charged peptides (containing three, five, and seven lysine residues) (19). Steric reasons dictate that only every second lysine may bind to acidic lipids in the case of oligopeptides (24). With increasing chain length, steric obstacles may become even more important. Moreover, theoretical predictions of the interaction energy are subject to large errors because the usual assumption of point charges no longer applies. In any case, model calculations would require knowledge about the exact position of every positive and negative amino acid with respect to the membrane. Lack of structural information for three of the four constructs thus renders that task impossible.
However, we have tried to identify the number of membraneinteracting residues for the one AKAP18␦ fragment with known crystal structure (13). This was facilitated by the spatial arrangement of the charges. We identified a protein surface that was not only flat but that also exhibited a high concentration of positively charged amino acids. It contained as many as 12 positively charged amino acid residues (Lys-101, Lys-102, Lys-108, Arg-116, Arg-219, Lys-223, Lys-232, Lys-238, Lys-239, Lys-272, Lys-273, and Lys-274) and only two negatively charged residues (Asp-128 and Glu-283) of AKAP18␦(87-292). Because of steric reasons, we assume that of Lys-272, Lys-273, and Lys-274, only two positive charges can interact with the membrane. One of the negative charges is in a flexible loop that is likely to bend away from a negatively charge surface. This very special arrangement of charges suggests that this is the lipid-interacting interface (Fig. 1). The hypothesis is supported by the calculations of ⌬G. If ⌬G was derived from the interaction of these positively amino acids with negatively charged lipids, we arrive at a binding free energy per positive charge of about Ϫ0.44 kcal/mol (Table 1), which is in the order of the interaction energy derived from short polylysine model peptides (19).
The pure electrostatic targeting mechanism does not impose any selectivity to the location of the scaffolding protein. This seems at odds with its task of regulating water homeostasis. One would expect that AKAP18␦ binds specifically to AQP2containing vesicles instead of binding randomly to any negatively charged membrane. Hence, we have to assume that this specificity is achieved through other mechanisms such as protein-protein interactions. AKAP18␦ may directly interact with a transmembrane-or membrane-associated protein on AQP2bearing vesicles in a similar manner as with phospholamban in the membrane of the sarcoplasmic reticulum in cardiac myocytes (8). Alternatively, AKAP18␦ may be specifically targeted to AQP2-bearing vesicles through interactions of one of its binding partners, PKA or PDE4D (7), with a protein on AQP2bearing vesicles. A similar concept most likely applies to many other scaffolding proteins, where myristoylation or palmitoylation of the protein does not accomplish targeting to a specific membrane compartment (25,26).
In summary, we conclude that AKAP18␦ is anchored to membrane lipids by electrostatic interactions. Membrane affinity stems from the special spatial arrangement of positively charged amino acids into a binding plane (Fig. 1).

TABLE 1 Comparison of AKAP18␦ versions with pentalysine
The number of residues, the positive and negative charges, the molar partition coefficient, the binding free energy, and the binding free energy per positive charge are shown for each AKAP18␦ protein.

Construct
Residues