Molecular determinants of the myristoyl-electrostatic switch of MARCKS.

MARCKS is a protein kinase C (PKC) substrate which binds calcium/calmodulin and actin, and which has been implicated in cell motility, phagocytosis, membrane traffic, and mitogenesis. MARCKS cycles on and off the membrane via a myristoyl electrostatic switch (McLaughlin, S., and Aderem, A. (1995) Trends Biochem. Sci. 20, 272-276). Here we define the molecular determinants of the myristoyl-electrostatic switch. Mutation of the N-terminal glycine results in a nonmyristoylated form of MARCKS which does not bind membranes and is poorly phosphorylated. This indicates that myristic acid targets MARCKS to the membrane, where it is efficiently phosphorylated by PKC. A chimeric protein in which the N terminus of MARCKS is replaced by a sequence, which is doubly palmitoylated, is phosphorylated by PKC but not released from the membrane. Thus two palmitic acid moieties confer sufficient membrane binding energy to render the second, electrostatic membrane binding site superfluous. Mutation of the PKC phosphorylation sites results in a mutant which does not translocate from the membrane to the cytosol. A mutant in which the intervening sequence between the myristoyl moiety and the basic effector domain is deleted, is not displaced from the membrane by PKC dependent phosphorylation, fulfilling a theoretical prediction of the model. In addition to the nonspecific membrane binding interactions conferred by the myristoyl-electrostatic switch, indirect immunofluorescence microscopy demonstrates that specific protein-protein interactions also specify the intracellular localization of MARCKS.

observation confirmed by a high axial ratio (4) and circular dichroic spectral analysis (5). The sequences of bovine, chicken, murine, rat, and human MARCKS have been determined, and they demonstrate a number of interesting features: the amino acid compositions are unusually rich in alanine, glycine, proline, and glutamate; the coding molecular masses (29)(30)(31) differ significantly from the electrophoretic molecular masses (67-87 kDa); there is a highly conserved N-terminal region and a highly conserved effector domain, the latter containing phosphorylation sites and calmodulin-and actin binding sites (3, 6 -11). The interaction of MARCKS with calmodulin and actin is regulated in a complicated manner. First, MARCKS binds to calmodulin only in the presence of calcium, and the phosphorylation of MARCKS prevents this interaction (11). Second, MARCKS binds to the sides of actin filaments and cross-links them, and this cross-linking activity is disrupted by both phosphorylation of MARCKS and by calcium-calmodulin (3).
MARCKS has a punctate distribution in macrophages, and many of the structures containing MARCKS are found at the substrate-adherent surface of pseudopodia and filopodia (12). Immunoelectron microscopy shows MARCKS to be in clusters at points where actin filaments interact with the cytoplasmic surface of the plasma membrane. 2 During phagocytosis MARCKS colocalizes on the forming phagosome with F-actin, PKC␣, talin, and myosin I (13). The data suggests that MARCKS functions as an integrator of PKC and calcium-calmodulin signals in the regulation of the actin cytoskeleton and of actin-membrane interactions.
MARCKS also has the capacity to cycle between the membrane and the cytosol (14,15). The nonphosphorylated protein associates tightly with membranes by the cooperative binding energies contributed by the insertion of the myristic acid moiety into the lipid bilayer, and by the electrostatic interaction of the basic effector domain with acidic phospholipids (16,17). PKC-dependent phosphorylation introduces negative charges into the positively charged effector domain, thereby neutralizing the electrostatic interaction, which results in the displacement of MARCKS from the membrane (16,17). The subsequent dephosphorylation of MARCKS is accompanied by its reassociation with the membrane (14). The mechanism by which MARCKS cycles on and off the membrane has been termed the myristoyl-electrostatic switch (18). We report here a site-directed mutagenesis study defining the molecular determinants of the myristoyl-electrostatic switch of MARCKS. EXPERIMENTAL  Preparation of MARCKS Constructs-The mutants were generated using cloned native MARCKS cDNA (8) as a template, and a polymerase chain reaction protocol utilizing VENT DNA polymerase (New England Biolabs). Wild type (wt) MARCKS and the various mutants were engineered into the Invitrogen pcDNA I/NEO vector. All polymerase chain reaction primers contained BamHI restriction sites facilitating ligation into the corresponding vector polylinker site. All constructs were confirmed by restriction digests and DNA sequencing; all mutations are indicated by underlines. The wt construct was generated by engineering the 1.15-kilobase pair XhoII fragment from MM3 into the vector. Myr Ϫ was generated by using coding primer 5Ј-GCTG-CGGATCCGCCAGCATGGCTGCCCAGTTCTCC-3Љ and non-coding primer (3Ј Norm) 5Ј-GGTCTAGAGGATCCCTCTGGAGCTTACTCGGC-3Ј. PHOS Ϫ was generated by engineering a HindIII site in codons 158 and 159 so that the desired molecule could be generated in two pieces. The N-terminal half was generated by using coding primer (5Ј-Norm) 5Ј-CGTCTAGAGGATCCGAAGCCAGCATGGGTGCC-3Ј and non-coding primer 5Ј-GGTGCAAGCTTGAAGGCCTTCTTGAAGGCAAAGCG-CTT-3; the C-terminal half was generated by using coding primer 5Ј-GTCGAAGCTTGGCGGCTTCGCCTTCAAGAAGAGCAAG-3Ј and 3Ј-Norm as the non-coding primer. The two polymerase chain reaction products were subjected to HindIII and BamHI digestion then purified, and the mutant was generated in a three fragment ligation. The G43-M mutant combines the first 7 amino acids of GAP-43 (19) with amino acids 6 -309 of murine MARCKS (8) using coding primer 5Ј-GCG-GGATCCATGCTGTGCTGTATGAGAAGATCCAAGACCGCAGCG-3Ј and 3Ј-Norm as the non-coding primer. ⌬6 -140 was constructed using coding primer 5Ј-GCGGGATCCATGGGAGCACAATTCAGCCAGACC-CCGAAA-3Ј and 3Ј-Norm as the non-coding primer.
Transfection and Culture Conditions-Constructs were stably transfected into Ltk Ϫ cells using calcium-phosphate precipitation (20). Transfectants were selected in Dulbecco's modified Eagle's medium supplemented with 1% L-glutamine, 100 units/ml penicillin G, 100 mg/ml streptomycin (all from JRH Biosciences, Lenexa, KS), 10% heatinactivated fetal bovine serum (HyClone, Logan, UT), and 400 mg/ml G418 (Life Technologies, Inc., Grand Island, NY), in an atmosphere of 5% CO 2 in air. Individual G418-resistant colonies were harvested by localized trypsinization and cloned by limiting dilution. Clones expressing equivalent amounts of wild-type or mutant MARCKS were identified by Western blotting. Where indicated, PMA was used at 200 nM from a 0.5 mM stock solution in Me 2 SO.
Western Blotting-Monolayers were scraped into 100 ml of lysis buffer (10 mM Tris-HCl, pH 7.5, 15 mM EDTA, 50 mM potassium fluoride, 50 mM NaH 2 PO 4 , 10 mM sodium pyrophosphate, and 1% Nonidet P-40) supplemented with the protease inhibitors indicated below. Protein was determined using the Coomassie Plus kit (Pierce). Cell lysates were resolved by 7.5% SDS-PAGE and transferred to Immobilon-P membranes (Millipore, Bedford, MA). Blots were blocked for 1 h at room temperature in milk buffer: 25 mM Tris-HCl, pH 8.0, 0.02% NaN 3 , and 150 mM NaCl (TBS) supplemented with 5% nonfat dry milk. Blots were then incubated with a 1:200 dilution of rabbit anti-murine MARCKS antiserum (21) in milk buffer for 1 h, washed sequentially with milk buffer, and Western blot buffer (TBS supplemented with 0.1% Triton X-100), and then incubated for an additional hour in horseradish peroxidase-conjugated anti-rabbit IgG (Amersham). Bands were visualized using enhanced chemiluminescence (Amersham).
Two-dimensional Thin Layer Phosphopeptide Mapping-This was performed essentially as described (21).
Protein Kinase C Phosphorylation of Myristoylated and Nonmyristoylated MARCKS-Ltk Ϫ cells stably expressing either wt Ϫ or Myr Ϫ MARCKS were radiolabeled with [ 32 P]orthophosphate and stimulated with 200 nM PMA for the indicated times. The incubations were terminated by the addition of lysis buffer, and MARCKS or Myr Ϫ MARCKS were immunoprecipitated as described (21). Phosphoproteins were resolved by SDS-PAGE, and phosphorylation levels were quantitated using a PhosphorImager (Molecular Dynamics) and standardized for total protein concentration. The fold-stimulation of phosphorylation was plotted as a function of time.
Immunofluorescence Microscopy-Cells were grown on acid-washed round glass coverslips (Propper Manufacturing Co., Inc., Long Island City, NY) for 24 -72 h prior to analysis. After desired treatments, cells were fixed and permeabilized as described previously (22). MARCKS was detected using an affinity purified rabbit anti-murine MARCKS antibody (12,21,22) and a rhodamine-conjugated goat anti-rabbit (FabЈ) 2 secondary antibody (Tago, Burlingame, CA)). Specificity of staining was assessed by omission of primary antibodies, and by staining untransfected Ltk Ϫ cells. Stained cells were photographed on a Zeiss Axiophot microscope using a 100 ϫ Nomarski objective, rhodamine optics, and Kodak professional ASA 200 slide film.
Subcellular Fractionation-Cells in 60-mm dishes at ϳ80% confluency were radiolabeled overnight in medium supplemented with 40 mCi/ml [ 3 H]myristic acid, or starved in lysine-free RPMI (Life Technologies, Inc.) for 2 h and then labeled for 4 h in medium containing 0.1 mCi/ml [ 3 H]lysine (22). Alternatively, cells were labeled with [ 32 P]orthophosphate for 3 h in phosphate-free RPMI (Life Technologies, Inc.). In some cases radiolabeled cells were treated with 200 nM PMA for 30 min immediately prior to fractionation. Cell monolayers were scraped into 1 ml of homogenization buffer (HB, 250 mM sucrose, 20 mM Hepes pH 7.2, 1 mM EDTA) supplemented with protease inhibitors (0.09 TIU/ml aprotinin, 0.5 mg/ml leupeptin, 1 mM phenylmethylsulfonyl fluoride, and 1 mM diisopropyl fluorophosphate) (22). Cells were disrupted at 4°C by nitrogen cavitation and postnuclear supernatants were separated into total membranes and cytosol as described previously (22). MARCKS was immunoprecipitated from each fraction with 6 l of rabbit anti-murine MARCKS antiserum (21) using the RIPA method. Boiled immunoprecipitates were separated by 8% SDS-PAGE. 3 H proteins were visualized by fluorography of En 3 Hanced gels and 32 P proteins were visualized by autoradiography.

Rationale for Mutant Construction-Comparison of known
MARCKS sequences revealed two highly conserved regions: the N-terminal half which contains a myristoylation consensus sequence, and a basic effector domain which contains the PKC phosphorylation sites and a calmodulin-and actin-binding site (1, 2) (Fig. 1). To determine the role of myristoylation, the N-terminal glycine, which serves as the myristoyl acceptor, was mutated to alanine (Myr Ϫ , Fig. 1 (1), and a basic effector domain (ED) which contains the phosphorylation sites (S) and which binds calmodulin and actin. Myr Ϫ , the N-terminal glycine has been replaced with alanine resulting in a nonmyristoylated molecule. G43-M, a MARCKS-GAP-43 chimera which contains the first seven amino acids of GAP-43 (cross-hatched box) and which is doubly palmitoylated, instead of being myristoylated. Phos Ϫ , the serine residues which are phosphorylated by PKC have been mutated to alanine or glycine. ⌬6 -140, the intervening sequence between the myristoylation site and the effector domain (amino acids 6 -140) has been deleted.
Characterization of the Mutant MARCKS Molecules-The coding regions of native murine MARCKS and the mutants described above were inserted into the BamHI site of pcDNAI/neo and transfected into Ltk Ϫ cells which lack endogenous MARCKS (8). Because transient transfectants yielded extraordinarily high levels of MARCKS with cytoarchitectural distortions (data not shown), this system was not an acceptable model in which to study the biology of MARCKS; therefore stable transfectants were obtained. Clonal populations expressing similar levels of wild type (wt) or mutant MARCKS were compared ( Fig. 2A). The cells were radiolabeled with myristic acid, and the various myristoylated MARCKS molecules were visualized by fluorography after immunoprecipitation and SDS-PAGE (Fig. 2B). As expected the G43-M chimera was labeled with palmitic acid but not with myristic acid, while Myr Ϫ did not incorporate myristic acid (Fig. 2B). In parallel experiments, we found that all the MARCKS mutants, except for Phos Ϫ , were phosphorylated in phorbol ester-stimulated cells (Fig. 2C). In order to determine whether the transfected wild-type MARCKS was phosphorylated on the correct residues, two-dimensional phosphopeptide mapping was performed (Fig. 2D). Phosphopeptide 1, representing Ser-152 and -163 (both are cleaved to FS), has twice the intensity of phosphopeptide 2 (cleaved to FKKS), which contains Ser-156 (21). The pattern is identical to MARCKS isolated from murine peritoneal macrophages (21). Two-dimensional thermolytic phosphopeptide mapping of all the mutants, except Phos Ϫ , revealed an identical pattern (data not shown).
Subcellular Localization and Stimulus-dependent Redistribution of MARCKS Mutants-In macrophages, neutrophils, neurons, and primary fibroblasts, the phosphorylation of MARCKS is accompanied by its release from the membrane (12,14,15,22). In unstimulated cells, ϳ95% of wt MARCKS partitioned with the membrane fraction (Fig. 3). Treatment of the cells with PMA, which stimulates PKC, resulted in the phosphorylation of MARCKS (Fig. 2B), and the translocation of Ϸ50% of the protein from the membrane to the cytosol (Fig. 3). The translocated cytosolic protein was phosphorylated with approximately twice the stoichiometry of the membrane bound protein (Fig. 3), consistent with our previous observation that MARCKS is displaced from membranes upon phosphorylation of all its PKC phosphorylation sites (14). Experiments with the Phos Ϫ mutant confirmed this observation: non-phosphorylatable MARCKS is membrane-bound and does not translocate from membrane to cytosol upon PMA treatment (Fig. 3).
Myristoylation is absolutely required for membrane binding of MARCKS since the Myr Ϫ mutant was present only in the cytosol (Fig. 3). In addition, the rate of phosphorylation of the Myr Ϫ mutant was substantially lower than that of wt MARCKS (Fig. 4), suggesting that MARCKS must be membrane bound, and in close apposition to activated PKC, for efficient phosphorylation.
The role of fatty acid acylation in targeting MARCKS to the membrane was further probed using the G43-M chimera, which contains two palmitic acid residues instead of one myristate at the N terminus. Interestingly, the palmitoylated chimera associated tightly with the membrane, but in contrast to myristoylated MARCKS, was not released from the membrane upon PMA-induced phosphorylation (Figs. 3 and 2B). The small amount of cytosolic G43-M was only labeled with [ 3 H]lysine, and not with [ 3 H]palmitic acid, suggesting that it had been depalmitoylated (Fig. 3).
The contribution of the conserved N terminus to membrane binding and phosphorylation-dependent translocation was examined using the ⌬6-140 deletion mutant. This molecule associated with the membrane in unstimulated cells, and did not translocate to the cytosol upon PMA-induced phosphorylation (Fig. 3).
Immunolocalization of MARCKS Mutants-The subcellular localization of the wild type and mutant MARCKS molecules was studied using indirect immunofluorescence microscopy. In quiescent cells, wild type MARCKS had a diffused distribution throughout the plasma membrane (Fig. 5B). This distribution is similar to the staining of endogenous MARCKS in unstimulated mouse embryo fibroblasts (22), suggesting that the transfected cells are a good model system in which to study this protein. No staining of MARCKS was seen when the primary antibody was omitted (data not shown). As with endogenous MARCKS in mouse embryo fibroblasts (22), activation of PKC is accompanied by a redistribution of MARCKS to perinuclear vesicles (Fig. 5D), which were previously identified as lysosomes (22). This translocation of MARCKS occurs via a soluble intermediate (22). Myr Ϫ stained diffusely throughout the cytoplasm in both control and PMA-treated cells (Fig. 5, I and J), confirming the biochemical fractionation data which demonstrated that myristoylation is required for membrane binding (Fig. 3). The nuclear staining seen in Fig. 5, I and J (and Fig. 5,  G and H), is an artifact which becomes apparent as a result of increased exposure times required to visualize MARCKS when it is not associated with the plasma membrane. The nuclear staining is also seen in mock transfected (pI/NEOi) Ltk Ϫ cells which do not express MARCKS (data not shown). MARCKS is undetectable by immunoblotting in nuclei purified from Myr Ϫ and mock transfected cells (data not shown). Phos Ϫ -stained the plasma membrane diffusely, was particularly concentrated in membrane extensions, and was not displaced upon PMA treatment (Fig. 5, E and F). This finding is also consistent with the fractionation data which demonstrate that nonphosphorylatable MARCKS remained associated with the plasma membrane and was not released upon PMA treatment of the cells (Fig. 3). G43-M had a similar distribution to Phos Ϫ , and while PMA induced phosphorylation of the chimera (Fig. 2B), it did not significantly alter its intracellular distribution (Fig. 5, K  and L). Once again, this reinforces the fractionation data which demonstrated that activation of PKC was not accompanied by significant translocation of G43-M from the membrane to the cytosol (Fig. 3). ⌬6 -140 demonstrated a reticular staining throughout the cell (Fig. 5, G and H). Since this mutant fractionates quantitatively with the membrane fraction (Fig. 3), it is likely that ⌬6 -140 is associated with a variety of intracellular membranes. PMA treatment did not influence the subcellular distribution of this molecule (Fig. 5, G and H), although it did increase its phosphorylation (Fig. 2B). DISCUSSION MARCKS associates with the plasma membrane of quiescent cells and is released from the membrane to the cytosol upon PKC-mediated phosphorylation (14,15). Studies with artificial membranes suggest that binding of MARCKS to membranes requires both hydrophobic insertion of its myristoyl chain into the lipid bilayer and electrostatic interaction of its basic domain with acidic lipids (16,17). PKC-mediated phosphorylation introduces negative charges into the basic effector domain, thereby decreasing the electrostatic interaction, resulting in the release of the protein from the membrane (16,17). This cycle of membrane binding and release has been termed the myristoyl-electrostatic switch (18). In this report we use a mutational analysis to confirm and extend this model.
The myristoyl-electrostatic switch makes the following predictions, all of which are confirmed by the data presented here. First, myristic acid is necessary but not sufficient for membrane binding since mutation of the N-terminal glycine to alanine prevents myristoylation and abrogates membrane binding (23,24,38). Second, phosphorylation introduces phosphate groups into the effector domain which partially neutralizes the electrostatic interaction of MARCKS with the membrane, resulting in its release to the cytosol. Mutation of serines 152, 156, 160, and 163 which are phosphorylated by PKC results in a MARCKS mutant which is neither phosphorylated nor released from the membrane into the cytosol in phorbol ester-stimulated cells. Third, the model predicts that decreasing the distance between the myristic acid moiety and the basic effector domain would increase the membrane binding affinity of the protein (16,18). This occurs because the overall association constant is the product of the hydrophobic interaction of myristate with the bilayer (K a ), and the electrostatic interaction of the basic effector domain with acidic lipids (K b ), scaled by the distance, R, between them. Thus the overall apparent association constant, K, is: where a ϭ 4 nm/M Ϫ1 is a constant that depends only on the area/phospholipid (25). Therefore, the insertion of myristate into the bilayer confines the basic effector domain to a hemisphere of radius r, which greatly increases the probability that this basic domain will associate electrostatically with acidic lipids (18). Thus decreasing the distance, r, between the myristoyl moiety and the basic domain will increase the affinity of the protein for the membrane. This prediction is supported by the observation that ⌬6 -140, a mutant in which the intervening sequence between the myristoyl moiety and the basic effector domain is deleted, is not displaced from the membrane by PKC dependent phosphorylation. There is also a second reason why ⌬6 -140 associates with membranes more tightly than its wild-type counterpart; amino acids 6 -140 are highly acidic and this would have the tendency to both repel the protein from the membrane and neutralize some of the positive charges of the effector domain. This view is supported by the observation that the partition coefficient of the basic domain of the intact pro- were left untreated (Con) or exposed to 200 nM PMA for 30 min prior to fractionation. Cells were broken using nitrogen cavitation, and separated into total membranes (M) and cytosol (C) as described under "Experimental Procedures." MARCKS was immunoprecipitated from each fraction prior to SDS-PAGE. [ 3 H]MARCKS and mutants were visualized by fluorography of En 3 Hanced gels. In the lower panels cells were labeled with 32 P and then stimulated and fractionated as described above. tein onto vesicles containing 20% acidic lipid is about 10,000fold lower than that obtained with a synthetic peptide derived from the effector domain of MARCKS (16,26). The final prediction follows from the observation that myristic acid provides barely enough energy to anchor the protein to the bilayer (16,(27)(28)(29), allowing for the reversible interaction of the myristoylated protein with the membrane. This appears to be confirmed by experiments demonstrating that the G43-M chimera, which bears two palmitic acids at its N terminus in place of one myristic acid moiety, is not released from the membrane upon phosphorylation. The small amount of cytosolic G43-M appears to have been depalmitoylated since we could not detect any radiolabeled palmitate associated with it. This implies that the membrane binding energy contributed by two palmitic acid moieties is sufficient to anchor the molecule to the membrane without the participation of the basic effector domain. This hypothesis is further supported by the observation that spontaneous desorption of doubly acylated peptides from artificial membranes is much slower than that of their singly acylated counterparts (30).
Why is membrane binding important for the function of MARCKS? MARCKS is phosphorylated by PKC, a kinase which is known to be active at the membrane (31). Thus, membrane binding would place MARCKS in close apposition to activated PKC, thereby facilitating the efficient phosphorylation of the substrate. This appears to be the case. First, MARCKS colocalizes with PKC␣ on phagosomes of macrophages, as well as in the transient adhesion zones known as podosomes (12,13). Second, nonmyristoylated, cytosolic, MARCKS is phosphorylated much more slowly than its myristoylated, membrane-bound, counterpart. Since myristoylated and nonmyristoylated MARCKS are similarly phosphorylated in vitro, 3 it appears that MARCKS is more efficiently phosphorylated when membrane bound.
Our data suggest that wt MARCKS must be phosphorylated to high stochiometry in order to be released to the cytosol. This is consistent with the observation that the related protein, MacMARCKS, which has one less phosphorylation site within its effector domain than MARCKS, is not released from the membrane upon phosphorylation (39). It also explains the apparent anomaly in which phagocytosis in macrophages is accompanied by the phosphorylation of MARCKS and its recruitment to the phagosomal membrane (13). If MARCKS was partially phosphorylated under these conditions, it would be expected to associate with membranes. 3 J. Ahn and A. Aderem, manuscript in preparation. In addition to the nonspecific hydrophobic and electrostatic interactions with lipids, there is good evidence that the binding of MARCKS to biological membranes involves specific proteinprotein interactions. For example, MARCKS has a punctate distribution in macrophages, and the structures containing MARCKS resemble podosomes, the transient complexes formed at the substrate adherent membrane during locomotion (12). During phagocytosis, MARCKS is concentrated under the forming phagosome, where it colocalizes with PKC␣, F-actin, talin, and myosin I (13). In mouse embryo fibroblasts, phosphorylation of MARCKS induces its translocation from the plasma membrane to lysosomes, and this redistribution appears to occur via a cytosolic intermediate (22). We previously demonstrated that phosphorylation of MARCKS is required for translocation to lysosomes (22), and this is confirmed here since Phos Ϫ remains on the plasma membrane. The observation that the G43-M chimera does not translocate from the plasma membrane to lysosomes, despite its phosphorylation by PKC, implies that membrane release is also a prerequisite for translocation to lysosomes. Experiments with staurosporine suggest that PKC must be continuously active in order for MARCKS to translocate to lysosomes (22), further reinforcing the notion that MARCKS translocates to lysosomes in a phosphorylationdependent manner. It is not clear whether MARCKS is partially dephosphorylated prior to binding to lysosomal membranes. Regardless of the mechanism of translocation, the very specific, regulated, association of MARCKS with a variety of internal membranes implies highly specific protein-protein interactions. The observation that ⌬6 -140 is localized differently, despite its tight association with membranes, suggests that targeting motifs may reside within amino acids 6 -140.
Thus, membrane binding of MARCKS is likely due to nonspecific hydrophobic and electrostatic interactions, as well as specific protein-protein interactions. As discussed elsewhere (18), nonspecific membrane binding reduces the dimensionality of protein diffusion from three-dimensional to two-dimensional. This results in an increase in apparent concentration of the protein, which facilitates the interaction of MARCKS with other membrane-bound proteins (receptors).
The actin cross-linking activity of MARCKS is regulated by calcium/calmodulin and by PKC-dependent phosphorylation of MARCKS (3). MARCKS is phosphorylated during chemotaxis (14) and phagocytosis (13), events which are accompanied by regulated rearrangement of the actin cytoskeleton. The phosphorylation of MARCKS activates the myristoyl-electrostatic switch, which in turn promotes the cycling of MARCKS between the membrane and cytosol. This, in turn, would regulate actin-membrane interactions as well as actin structure at the membrane. This hypothesis is supported by the observation that the Ltk Ϫ cells expressing the Phos Ϫ or G43-M MARCKS mutants show defects in cell migration and cell spreading (32).
The association of other myristoylated proteins with membranes may also be modulated by similar types of switches. ADP-ribosylation factors are myristoylated, regulatory, GTPbinding proteins involved in vesicular transport (reviewed in Ref. 33). The functions of ADP-ribosylation factor in membrane traffic and organelle integrity depend on their reversible association with specific membranes (33). The binding of GTP produces a conformational change that promotes membrane binding of ADP-ribosylation factors, possibly because an occluded myristoyl moiety now becomes available for membrane insertion (reviewed in Ref. 33). Hydrolysis of GTP results in a conformational change which promotes the release of ADP-ribosylation factor from the membrane. Recoverin, a myristoylated member of the EF-hand superfamily, serves as a calcium sensor in cells (34). Upon binding calcium ions, recoverin translocates from the cytosol to the membrane (34,35). X-ray crystallography and NMR data suggest that in the absence of calcium the myristoyl moiety is occluded within a hydrophobic groove (36,37). Calcium induces a conformational change in recoverin which results in the extrusion of the myristoyl moiety from the hydrophobic pocket and renders it available to participate in membrane binding. Thus, an array of myristoyl switch mechanisms appear to facilitate signal transduction by mediating reversible membrane binding.