Myristoylation-dependent and Electrostatic Interactions Exert Independent Effects on the Membrane Association of the Myristoylated Alanine-rich Protein Kinase C Substrate Protein in Intact Cells*

The myristoylated alanine-rich protein kinase C sub- strate (MARCKS) is a widely expressed, prominent substrate for protein kinase C. MARCKS is largely as- sociated with membranes in cells, and hydrophobic interactions involving the amino-terminal myristoyl moi- ety are thought to play a role in anchoring MARCKS to cellular membranes. In addition, experiments in cell- free systems have suggested that electrostatic interactions between the positively charged phosphorylation site/calmodulin binding domain (PSD) of MARCKS and negatively charged membrane lipids are also involved in this association. Although it has been inferred from phosphorylation experiments, the electrostatic nature of the interaction between the PSD and membranes has not been demonstrated directly in intact cells. We ex- pressed human MARCKS mutated in the myristoylation site and the PSD in REF52 cells; the cells were then fractionated by ultracentrifugation. Both nonmyristoylatable MARCKS and MARCKS in which the four serines in the PSD were mutated to aspartic acids, mimicking phosphorylation, exhibited decreased membrane affin- ity when compared to the fully myristoylated, wild-type, tetra-Ser protein or a myristoylated, tetra-Asn mutant. A double mutant, nonmyristoylatable protein in which the four serines in the PSD were mutated to aspartic acids exhibited negligible membrane association. Similar results were obtained in 293 cells that stably ex- pressed chicken MARCKS mutated in the same domains. The double mutant, nonmyristoylatable tetra-Asp

Activation of protein kinase C (PKC) 1 and the subsequent phosphorylation of its intracellular substrates result in a wide range of cellular processes including differentiation, mitogenesis, and hormone secretion (1,2). However, the exact role of these PKC substrates in mediating these responses is still not well understood.
The myristoylated alanine-rich protein kinase C substrate (MARCKS) protein is one of the most prominent of these intracellular substrates for PKC (for reviews, see Refs. 3 and 4). Although MARCKS has been well studied, its precise cellular function still has not been determined. Recent gene disruption studies in mice have implicated the protein in the normal development of the central nervous system and postnatal viability (5). MARCKS is characterized by several unusual attributes including amino-terminal myristoylation (6 -9), heat stability, and anomalous migration on SDS-polyacrylamide gels (10 -13). Three highly conserved domains have been identified in MARCKS proteins from all animal species examined to date: A consensus sequence for myristoylation at the amino terminus (14), a conserved sequence at the single intron splice site (12,13,15) and a positively charged phosphorylation site domain (PSD) that contains the serines phosphorylated by PKC (16 -19). The PSD has also been shown to bind calmodulin (CaM) (20 -22) and actin (23).
Both the PSD and the myristoyl moiety of MARCKS have been implicated in anchoring the protein to cellular membranes. The myristoyl group of MARCKS is thought to be involved in attaching the protein to membranes via hydrophobic interactions (7,8,(24)(25)(26). In addition, several studies in cell-free systems have suggested that the positively charged PSD of MARCKS is involved in membrane association through electrostatic interactions (26)(27)(28)(29)(30). Although this interaction has been well characterized in vitro, the only evidence supporting the electrostatic nature of the association of the PSD with membranes in intact cells has been obtained from phosphorylation experiments. In several cell types, phosphorylation of MARCKS by PKC appears to result in the dissociation of the protein from membranes (30)(31)(32)(33)(34)(35). However, because activation of PKC results in the phosphorylation of many intracellular substrates and affects numerous cellular processes, it is possible that the reduced affinity of MARCKS for membranes seen after PKC activation might not be due to electrostatic changes alone.
To resolve this question, we have directly investigated the electrostatic nature of the interaction between the PSD and membranes in intact cells. To do this, we have mimicked phosphorylation of the protein in intact cells by creating stable lines in two cell types expressing both human and chicken MARCKS in which the four serines in the PSD were mutated to aspartic acids, with equivalent asparagine mutations serving as a control. Although the aspartate mutant would be expected to * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  mimic phosphorylation electrostatically, neither it nor the asparagine mutant should be phosphorylated by PKC. Our data, from both subcellular fractionation experiments and immunoelectron microscopy of MARCKS, indicate that both myristoylation and electrostatic interactions involving the PSD exert independent, essentially additive contributions to the association of MARCKS with membranes in intact cells.

Construction of Expression
Plasmids-For the human MARCKS constructs, a phage clone encompassing the human gene encoding MARCKS (MACS) was isolated as described previously (12). Liquid lysate DNA from the isolated phage clone was subjected to overnight digestion with BglII in the presence of 1 g of RNase IA and 4 M spermidine. A 7.5-kb fragment was found that hybridized to the human MARCKS cDNA; this fragment was isolated by agarose gel electrophoresis and further purified by Gene Clean (Bio 101, Inc., Vista, CA). This 7.5-kb restriction fragment was subcloned into the BamHI site of Bluescribe (Stratagene, La Jolla, CA). Restriction fragment and sequencing analysis of the clone revealed the presence of approximately 3.4 kb of the MACS 5Ј flanking region, as well as the entire mRNA coding sequence, the single intron, and approximately 0.6 kb of the 3Ј-flanking region.
An epitope tag consisting of a 9-amino acid peptide derived from influenza virus hemagglutinin (36) was introduced in-frame between nucleotides 1289 and 1290 (12) in the carboxyl-terminal coding region of the 7.5-kb clone, 5 codons 5Ј of the stop codon, yielding the construct pHG80KWT (Fig. 1A). Insertion of the epitope sequence was achieved by site-directed mutagenesis (Amersham Corp.) according to the manufacturer's protocol; its presence in-frame was confirmed by dideoxy sequencing (37). This epitope is recognized by monoclonal antibody 12CA5 (Berkeley Antibody Co., Richmond, CA).
Constructs mutated in the myristoylation and phosphorylation site domain consensus sequences of the human MARCKS genomic construct pHG80KWT were generated by in vitro mutagenesis (Fig. 1B). The amino-terminal glycine in the MARCKS protein was changed to an alanine (pHG80KA 2 /G 2 ) or the four serines in the phosphorylation site domain (17) were mutated to either aspartic acids (pHG80KD/S) or asparagines (pHG80KN/S) using the Altered Sites oligonucleotide-directed in vitro mutagenesis kit from Promega (Madison, WI). The double mutant pHG80KA 2 /G 2 -D/S was created by changing the aminoterminal glycine of pHG80KD/S to an alanine. Creation of the mutations was confirmed by dideoxy sequencing (37).
Cell Culture-Rat embryo fibroblasts (REF52; a generous gift from Dr. Joseph Nevins, Department of Genetics, Duke University) were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% (v/v) heat-inactivated fetal calf serum, 2 mM glutamine, 100 units/ml penicillin, and 100 g/ml streptomycin. Stable cell lines derived from REF52 cells were maintained in the same medium supplemented with 400 g/ml Geneticin (Life Technologies, Inc.).
Human 293 cells (American Type Culture Collection, Rockville, MD) were maintained in minimal essential medium supplemented with 10% (v/v) heat-inactivated fetal calf serum, 2 mM glutamine, 100 units/ml penicillin, and 100 g/ml streptomycin. Stable cell lines derived from 293 cells were maintained in the same medium supplemented with 400 g/ml Geneticin.
Chick embryo fibroblasts were grown as described elsewhere (30) and used at passage four.
Creation of Stable Cell Lines in REF52 Cells-REF52 cells were plated at a density of 1 ϫ 10 6 cells/100-mm plate the evening prior to transfection. Approximately 2 h before transfection, the cells were fed with 10 ml of DMEM containing 10% (v/v) heat-inactivated fetal calf serum, 2 mM glutamine, 100 units/ml penicillin, and 100 g/ml streptomycin. The cells were transfected using the overnight calcium phosphate method (38) with 10 g of pHG80KWT (wild-type MARCKS), pHG80KA 2 /G 2 (nonmyristoylatable MARCKS), pHG80KD/S (tetra-Asp MARCKS), pHG80KN/S (tetra-Asn MARCKS), or pHG80KA 2 /G 2 -D/S (nonmyristoylatable tetra-Asp MARCKS) DNA, and 1 g of pSV2Neo (American Type Culture Collection) cDNA per plate. Each plate of transfected cells was divided into eight 100-mm plates 24 h after the calcium phosphate precipitates were removed (39). Selection in media containing 400 g/ml Geneticin was begun 24 h after splitting the cells. Cells were grown in selective media for 5 weeks, after which individual clones were isolated with a pipette tip and transferred to 48-well plates. The cells were consecutively transferred to 24-well plates, 12-well plates, 60-mm dishes, and 100-mm dishes. Clones were screened for expression of MARCKS proteins by Western blot analysis as described below.

FIG. 1. Construction of human MARCKS expression vectors stably transfected in REF52 cells.
In A, a 7.5-kb BglII fragment isolated from the genomic clone encoding human MARCKS (MACS) was subcloned into the BamHI site of Bluescribe (BSϩ; Stratagene). A 27-base pair insert encoding a 9-amino acid peptide epitope from influenza virus hemagglutinin was introduced in-frame between nucleotides 1289 and 1290 (12) to produce pHG80KWT. In B, the construct shown in A was modified as follows. The amino-terminal glycine was changed to an alanine (pHG80KA 2 /G 2 ), or the four serines in the wild-type phosphorylation site domain of human MARCKS were changed to aspartic acids (pHG80KD/S) or asparagines (pHG80KN/S) using oligonucleotide-directed mutagenesis. A double mutant (pHG80KA 2 /G 2 -D/S) in which the four serines were changed to aspartic acids and the amino-terminal glycine was changed to an alanine was also created.
Subcellular Fractionation-REF52 cells stably expressing human MARCKS proteins were grown to confluence and serum-starved overnight in DMEM containing 1% bovine serum albumin (lyophilized and crystallized; Sigma), 100 units/ml penicillin, and 100 g/ml streptomycin. Human 293 cells stably expressing chicken MARCKS proteins were not serum-starved overnight. All cells were separated into soluble and particulate fractions by ultracentrifugation as described previously (30). In general, equivalent cell numbers were used for sample preparation in each experiment; however, preliminary studies with the wildtype protein indicated that the samples could be diluted up to 3-fold with homogenization buffer without significantly affecting the percentage of MARCKS bound to the membranes.
Western Blot Analysis-Proteins from SDS-polyacrylamide gels were transferred to nitrocellulose and blocked in a solution of 4% (w/v) non-fat dry milk in TBS/T (10 mM Tris-HCl (pH 7.6), 154 mM NaCl, 0.3% (v/v) Tween 20) as described elsewhere (30). Blots were incubated with either monoclonal antibody 12CA5, which recognizes the hemagglutinin epitope tag sequence (Berkeley Antibody Co., Richmond, CA), at a 1:800 dilution in TBS/T, or a polyclonal antibody raised against chicken MARCKS (7). As a secondary antibody, 125 I-goat anti-mouse IgG (ICN, Irvine, CA) was used at 0.5 Ci/ml in TBS/T for 1 h at room temperature for experiments with the monoclonal antibody 12CA5. For experiments with the polyclonal antibody to chicken MARCKS, 125 I-protein A (Amersham Corp.) was used at 0.2 Ci/ml in TBS/T for 1 h at room temperature. PhosphorImager analysis was used for quantitation of the blots.
Immunoelectron Microscopy of Stable 293 Cell Lines-Three 100-mm dishes each of 293 cells stably expressing chicken MARCKS, in which the four serines in the phosphorylation site domain were mutated to aspartic acids (60KD/S), asparagines (60KN/S), and nonmyristoylatable chicken MARCKS in which the four serines in the phosphorylation site domain were mutated to aspartic acids (60KA 2 /G 2 -D/S), were prepared for frozen section immunoelectron microscopy essentially as described previously (30). In the current experiments, the sections were incubated with either a 1:40 or a 1:400 dilution of the primary antibody to chicken MARCKS (7).

Membrane Association of Human MARCKS Mutated in the Myristoylation and Phosphorylation
Site Domains-In order to determine if both the phosphorylation site domain and the myristoyl moiety of MARCKS are involved in membrane association of the protein in intact cells, we stably transfected REF52 cells with genomic constructs expressing wild-type human MARCKS and human MARCKS mutated in the myristoylation and phosphorylation site consensus sequences. These constructs all contained and were driven by approximately 3.4 kb of promoter from the human MARCKS gene (MACS), and contained the single intron from this gene (Fig. 1). This portion of the MARCKS promoter contained all of the elements necessary to drive high level expression of MARCKS in REF52 cells (Fig. 3). In studies performed in transgenic mice, these upstream sequences contained all of the elements necessary for normal spatial and temporal expression of MARCKS in the intact animal (50). In addition, these human genomic constructs also contained a 9-amino acid hemagglutinin epitope tag sequence five codons 5Ј of the stop codon; this addition does not interfere with the ability of a human MARCKS transgene to completely complement MARCKS deficiency in mice (50).
To investigate the importance of myristoylation in the membrane association of human MARCKS in REF52 cells, the amino-terminal glycine of MARCKS was mutated to an alanine, resulting in the synthesis of the nonmyristoylated protein.
Upon fractionation of these cells by ultracentrifugation, only 8 Ϯ 4% (S.D.) of the nonmyristoylated protein was found in the membrane fraction, compared to 82 Ϯ 13% for the wild-type, fully myristoylated protein (Fig. 4). These data represent the means from five separate experiments. These results indicate that the presence of the myristoyl moiety is required for most, but not all, of MARCKS membrane association in this cell type.
Because phosphorylation of MARCKS appears to result in a decreased affinity for cell membranes (30 -35), and because the phosphorylation site domain of MARCKS has been shown to be important for its membrane association in cell-free systems (26 -30), we investigated the importance of this region to membrane affinity in intact cells. When the four serines in the PSD of MARCKS were mutated to aspartic acids, a mutation that can mimic the electrostatic properties of phosphoserine (40), 33 Ϯ 12% (S.D.) of the tetra-Asp protein was associated with the membrane fraction, compared to 82 Ϯ 13% for the wild- In A, the 1.5-kb HindIII/EcoRI fragment of the chicken MARCKS cDNA (10) was subcloned into an expression vector containing the CMV immediate early promoter to produce pCMV/60KWT. In B, the aminoterminal glycine was changed to an alanine (pCMV/60KA 2 /G 2 ), or the four serines in the wild-type phosphorylation site domain were changed to aspartic acids (pCMV/60KD/S) or asparagines (pCMV/60KN/S) using oligonucleotidedirected mutagenesis. A double mutant (pCMV/60KA 2 /G 2 -D/S) in which the four serines were changed to aspartic acids and the amino-terminal glycine was changed to an alanine was also created. type, tetra-Ser protein (Fig. 4). This decrease in binding could potentially result from an electrostatic repulsion between the tetra-Asp domain and negatively charged membrane lipids. To address this possibility, we also made a mutation in which the four serines in the PSD were changed to asparagines. This amino acid is similar in structure to aspartic acid; however, it is neutral in charge at physiological pH and, therefore, should not inhibit membrane binding of the protein if it is due to electrostatic interactions. The tetra-Asn protein exhibited a high affinity for the membranes; 95 Ϯ 3% of the protein was associated with the membrane fraction, compared to 33 Ϯ 12% for the tetra-Asp protein (Fig. 4). This difference was highly statistically significant (p Ͻ 0.0005) using Student's t test. The modest increase in the membrane localization of the tetra-Asn mutant when compared to the wild-type protein can probably be accounted for by the partial phosphorylation of the wild-type protein in cells, even those that had been serum-deprived.
To further characterize the electrostatic nature of the interaction of MARCKS with membranes, we investigated the effect of ionic strength on the membrane association of wild-type MARCKS expressed in chick embryo fibroblast cells. After the cells were homogenized in the buffer described in Swierczynski and Blackshear (30), the cell extracts were incubated for 45 min at 0°C with frequent mixing in homogenization buffer containing 0, 50, 150, 300, and 500 mM NaCl. At NaCl concentrations of 0, 50, and 150 mM, approximately 90% of immunoreactive MARCKS protein was associated with the membrane fraction (data not shown). At concentrations of 300 and 500 mM NaCl, approximately 66 and 55% of the protein, respectively, associated with the membrane fraction (data not shown). Thus, as shown previously (24,26), high ionic strength resulted in a decreased affinity of MARCKS for membranes.
In order to determine if the myristoyl moiety and the phosphorylation site domain exerted an additive effect on the membrane affinity of MARCKS, we created a stable cell line expressing a double mutant human MARCKS protein in which the amino-terminal glycine was mutated to an alanine and the four serines in the phosphorylation site domain were mutated to aspartic acids. This double mutant exhibited negligible membrane affinity; only 0.8 Ϯ 0.2% of the protein associated with the membrane fraction (Fig. 4), compared to 8 Ϯ 4% for nonmyristoylated MARCKS and 33 Ϯ 12% for tetra-Asp MARCKS. The difference between the means of the double mutant compared to the nonmyristoylated mutant was highly significant (p Ͻ 0.005).
Analysis of the partition coefficients for the mutants also provided evidence for independent, additive contributions of myristoylation and electrostatic interactions involving the PSD to MARCKS membrane association in REF52 cells. Since variation in membrane concentration did not have a significant effect on MARCKS membrane association, we can assume for this calculation that the concentration of membranes was constant for each mutant; therefore, the partition coefficient (K p ) ϭ % MARCKS membrane /% MARCKS cytosol . Using this simple equation, the partition coefficient for each of the human MARCKS mutants in REF52 cells was calculated. In comparison to the partition coefficient of the wild-type, fully myristoylated protein, the partition coefficient of nonmyristoylated MARCKS was 52-fold lower. Similarly, the partition coefficient of the tetra-Asp mutant was 9-fold lower than the partition coefficient of the wild-type, tetra-Ser protein. The partition coefficient of the nonmyristoylatable tetra-Asp double mutant was decreased 565-fold in comparison to that of the fully myristoylated, wild-type, tetra-Ser protein. This decrease in the partition coefficient for the double mutant was similar to the product of the decrease in the partition coefficients for the nonmyristoylated and tetra-Asp proteins (52 ϫ 9 ϭ 468), suggesting that the two mutations exerted essentially additive effects on MARCKS membrane association.
Taken together, the data derived from the subcellular fractionation experiments on these stable fibroblast lines expressing human MARCKS indicate that both the myristoyl moiety and the phosphorylation site domain exert independent and essentially additive effects on the membrane association of MARCKS in intact cells.

Membrane Association of Chicken MARCKS Mutated in the Myristoylation and Phosphorylation Site Domains-We have
shown previously by subcellular fractionation experiments and immunoelectron microscopy that approximately 44% of nonmyristoylated chicken MARCKS stably expressed in human 293 cells was associated with membranes, compared to 82% of the wild-type protein (30). In order to determine whether electrostatic interactions between the PSD of MARCKS and the membrane could account for the substantial remaining membrane association of the nonmyristoylated protein in this cell type, we created stable lines in 293 cells expressing a double mutant form of MARCKS; this mutant contained four aspartic acids instead of four serines in the PSD and was nonmyristoylatable (60KA 2 /G 2 -D/S). Upon fractionation of the cells by ultracentrifugation and immunoblotting analysis (Fig.   FIG. 3. Subcellular fractionation of human MARCKS stably expressed in REF52 cells. REF52 cells were stably transfected with 10 g of wild-type (WT), nonmyristoylatable (A 2 /G 2 ), tetra-Asp (D/S), tetra-Asn (N/S), nonmyristoylatable tetra-Asp (A 2 /G 2 -D/S) human MARCKS DNA, and 1 g of pSV2Neo cDNA (Neo) using the calcium phosphate method. The cells were homogenized and separated into membrane (M) and cytosolic (C) fractions by ultracentrifugation, as described previously (30). Cells expressing Neo alone were included as a control. The membranes were then resuspended in the original volume of homogenization buffer, and then equal volumes (approximately 20% of each sample) were subjected to SDS-polyacrylamide gel electrophoresis, transferred to nitrocellulose, and probed with the 12CA5 antibody raised against the hemagglutinin epitope tag. Shown is the resulting Western blot, which is overexposed to show small amounts of protein in some fractions. The difference between the means from the A 2 /G 2 cells and the A 2 /G 2 -D/S cells was highly significant using Student's t test (p Ͻ 0.005), as was the difference between the means from the N/S and D/S cells (p Ͻ 0.0005). 5), virtually none (2 Ϯ 0.2% (S.D.)) of the double mutant protein was found associated with the cell membrane fraction (Fig. 6). This is markedly lower (p Ͻ 0.05) than the 18 Ϯ 12% membrane association exhibited by the nonmyristoylated wild-type protein (Fig. 6).
We also analyzed stable cell lines expressing chicken MARCKS in which the four serines in the PSD were mutated to aspartic acids (60KD/S) to examine electrostatic effects on the fully myristoylated protein. When the cells expressing 60KD/S were fractionated by ultracentrifugation, 59 Ϯ 9% (S.D.) of the MARCKS protein was associated with the membrane fraction (Fig. 6). We also analyzed a stable cell line expressing chicken MARCKS in which the four serines in the PSD were mutated to asparagines (60KN/S) as a control for the electrostatic effects. When the cells expressing 60KN/S were fractionated in the same way, the majority (94 Ϯ 5%) of the MARCKS protein was associated with the membrane fraction (Fig. 6). This difference was highly statistically significant (p Ͻ 0.0001) using Student's t test.
To confirm the membrane affinities observed with ultracentrifugation followed by immunoblotting, we performed immunoelectron microscopy on these stable cell lines. The polyclonal antibody raised against chicken MARCKS (7) did not crossreact with any endogenous human proteins expressed in the 293 cells expressing Neo alone (Fig. 7). In the 293 cells expressing tetra-Asn MARCKS, gold particles were localized to the plasma membrane and to the membranes of large cytoplasmic vesicles with clear contents (Figs. 7 and 8), as described previously for the wild-type protein (30). In cells expressing tetra-Asp MARCKS, the gold particles appeared to be distributed in both the cytoplasm and the plasma membrane (Fig. 7). In contrast to both the cells expressing the tetra-Asn and tetra-Asp mutants, the gold particles in the cells expressing the double mutant, 60KA 2 /G 2 -D/S, were found almost entirely in the cytosol (Fig. 8).
Several sections for each of the cell lines were analyzed by counting the gold particles and classifying them as either "membrane" or "cytosolic." As described previously (30), the membrane was designated as the plasma membrane and the membranes of large, cytoplasmic vesicles with clear contents. The rest of the cell was considered to be cytosolic. In five sections of cells expressing tetra-Asn MARCKS, 79 Ϯ 3% (S.D.) of the gold particles (total of 1098 counted) was associated with membranes. In four sections of cells expressing tetra-Asp MARCKS, 48 Ϯ 17% of the gold particles (total of 1266 counted) was associated with membranes. In five sections of cells expressing nonmyristoylatable, tetra-Asp MARCKS, 13 Ϯ 5% (S.D.) of the gold particles (total of 1572 counted) was associated with membranes. These results support the data obtained from the subcellular fractionation experiments in that tetra-Asn MARCKS was associated predominantly with membrane structures, tetra-Asp MARCKS was distributed almost equally between membranes and cytosol, and nonmyristoylatable tetra-Asp MARCKS was almost exclusively cytosolic. DISCUSSION These studies examined the effect of mutations in the myristoylation and PSD consensus sequences of MARCKS on the membrane association of the protein in intact cells. It was necessary to make stable cell lines expressing each type of mutant protein, since transient transfection experiments resulted in an incompletely myristoylated protein (30). The present results, from subcellular fractionation of human MARCKS mutants stably expressed in REF52 cells, and both subcellular fractionation and immunoelectron microscopy of chicken MARCKS mutants stably expressed in 293 cells, fully support a two-component model (26,29,30) for MARCKS membrane association in vivo. According to this model, hydrophobic interactions involving the myristoyl tail and electrostatic interactions involving the phosphorylation site domain are responsible for the membrane affinity of MARCKS. In addition, our results suggest that the effects of these two regions are additive and account for essentially all of the membrane binding potential of MARCKS in these cell types.
Several previous experiments in cell-free systems have suggested that the membrane association of the positively charged PSD is mediated by electrostatic interactions. For example, the affinity of a positively charged, wild-type MARCKS PSD peptide for synthetic lipid vesicles increased concomitantly with the percentage of negatively charged phosphatidylserine (PS) in the vesicles (29). Binding to the vesicles decreased markedly when the peptide was either phosphorylated by PKC or the four serines were mutated to aspartic acids, mimicking phosphorylation (29). Similarly, murine MARCKS expressed in a baculovirus system (26), MARCKS purified from bovine brain (27), and bacterially expressed, MARCKS-glutathione S-transferase fusion proteins (28) exhibited high affinity for negatively charged lipids. These interactions were also disrupted by PKCdependent phosphorylation of MARCKS (26 -28). Consistent with electrostatic interactions, phosphorylation of MARCKS purified from bovine brain did not result in a decreased affinity for neutral phosphatidylcholine (PC) vesicles (27). Finally, in vitro translated MARCKS in which the four serines in the PSD were mutated to aspartic acids, mimicking phosphorylation (40), demonstrated a decreased affinity for cellular membranes isolated from LM/TK Ϫ fibroblasts when compared to wild-type MARCKS or a corresponding tetra-asparagine mutant (30).
As these data indicate, the potential electrostatic association between the PSD and membranes has been thoroughly examined in vitro; however, this interaction has not been well characterized in intact cells. To our knowledge, the only data describing the involvement of the PSD in membrane association in cells have been indirect data obtained from phosphorylation experiments. Treatment of C6 glioma cells (31), neutrophils (33), and chick embryo fibroblasts (30) with the phorbol ester phorbol 12-myristate 13-acetate resulted in a decreased affin-ity of MARCKS for cellular particulate fractions. In addition, immunofluorescence experiments in mouse embryo fibroblasts demonstrated that phosphorylation of MARCKS resulted in the translocation of the protein from the plasma membrane to lysosomes (35). While the introduction of the negatively charged phosphate groups could cause an electrostatic repulsion between the PSD of MARCKS and the membrane, the observed decreases in membrane binding upon phosphorylation might not be due to electrostatic effects alone. Other proteins and processes modified by activation of PKC could possibly influence the membrane affinity of MARCKS.
To clarify these uncertainties, we created stable lines in REF52 cells expressing human MARCKS proteins, and in 293 cells expressing chicken MARCKS proteins, in which the four serines in the wild-type PSD were mutated to either aspartic acids or asparagines. Aspartic acid is negatively charged at physiological pH; therefore, the tetra-Asp mutant should electrostatically mimic the fully phosphorylated MARCKS protein (40). However, this mutant is nonphosphorylatable by PKC and should not be affected by the activation of PKC. Our results show that the tetra-Asp MARCKS mutant exhibited a marked decrease in affinity for cellular membranes when compared to both the wild-type, tetra-Ser protein and the corresponding tetra-Asn protein. The tetra-Asn protein was chosen as a specific control for the electrostatic interaction studies because it mimics aspartic acid structurally, but not electrostatically, and also cannot be phosphorylated by PKC. These results allow us to conclude that the decrease in membrane binding observed with the tetra-Asp mutant was mediated entirely by electrostatic interactions involving the PSD.
We also examined the effect of the tetra-Asp mutation on membrane association of the nonmyristoylated protein. In REF52 cells, the nonmyristoylated, tetra-Ser protein was 8% membrane associated; this was markedly greater (p Ͻ 0.005) than the 0.8% membrane association exhibited by the nonmyristoylated, tetra-Asp protein. Similarly, when the analogous mutant chicken proteins were expressed in 293 cells, the nonmyristoylated, tetra-Ser protein exhibited 18% membrane association compared to the 2% seen with the nonmyristoylated, tetra-Asp protein (p Ͻ 0.05). These data were reinforced by the immunoelectron microscopic analysis showing that most of the immunoreactive nonmyristoylated, tetra-Asp MARCKS was localized to the cytosol, whereas most of the fully myristoylated, tetra-Asn mutant was localized to the plasma membrane. Thus, both types of nonmyristoylatable, tetra-Asp proteins, expressed in two different cell types, exhibited negligible membrane association. We can conclude from these data that the membrane association of wild-type MARCKS in these cells is almost entirely due to the combined effects of hydrophobicity and electrostatics. It therefore seems unlikely that MARCKS membrane binding is mediated by a cytoplasmic-face membrane protein receptor, as has been postulated for pp60 v-src (41)(42)(43)(44)(45). This conclusion is supported by our previous findings that the association of in vitro translated MARCKS with membranes isolated from LM/TK Ϫ cells was not affected when the membranes were boiled or pretreated with trypsin (24).
Taken together, our results obtained in both the REF52 and 293 cell experiments indicate that both the myristoyl moiety and the PSD contribute independently to MARCKS membrane association in intact cells. In both cell types, the further decrease in membrane affinity of the nonmyristoylatable, tetra-Asp MARCKS protein when compared to the nonmyristoylatable protein alone suggest that these two mutations exert an additive inhibitory effect on membrane binding in cells, and that the decrease in membrane binding seen with MARCKS phosphorylation can be accounted for by phosphorylation-induced changes in the electrostatic properties of the PSD.
These results demonstrate conclusively that changes in the electrostatic potential of the MARCKS PSD decrease the affinity of the protein for cellular membranes in intact cells. The significance of both myristoylation and electrostatic interactions with the PSD being involved in anchoring MARCKS to membranes has been described in terms of a "myristoyl-electrostatic switch" (46). According to this theory, MARCKS is anchored to the membrane via two biophysically weak interactions in order to promote reversible membrane association. Hydrophobic and electrostatic interactions combine in order for MARCKS to stably associate with membranes; however, when one of these interactions is disrupted, as apparently occurs when MARCKS is phosphorylated by PKC, the protein's affinity for the membrane decreases, and a significant proportion can be "released" into the cytosol. The phosphorylation-dependent, reversible nature of MARCKS membrane association most likely plays an important role in the still undefined physiological functions of the protein.
Protein sequence domains closely resembling the MARCKS PSD have been identified recently in other proteins. These include the MARCKS homologue, MRP (19); the spectrin and actin linking protein adducin (47); the recently cloned human diacylglycerol kinase (48); and the NR1 subunit of the Nmethyl-D-aspartate receptor (49). We and others have focused on these so-called MARCKS homology domains as PKC phosphorylation-reversible, calmodulin-binding domains. 2 However, since none of these MARCKS homology domains contain acidic residues and all are bounded by blocks of basic residues, we anticipate that all could potentially be involved in mediating PKC phosphorylation-reversible association with negatively charged cellular membranes. This prediction awaits experimental confirmation.