Lateral sequestration of phosphatidylinositol 4,5-bisphosphate by the basic effector domain of myristoylated alanine-rich C kinase substrate is due to nonspecific electrostatic interactions.

A peptide corresponding to the basic (+13), unstructured effector domain of myristoylated alanine-rich C kinase substrate (MARCKS) binds strongly to membranes containing phosphatidylinositol 4,5-bisphosphate (PIP(2)). Although aromatic residues contribute to the binding, three experiments suggest the binding is driven mainly by nonspecific local electrostatic interactions. First, peptides with 13 basic residues, Lys-13 and Arg-13, bind to PIP(2)-containing vesicles with the same high affinity as the effector domain peptide. Second, removing basic residues from the effector domain peptide reduces the binding energy by an amount that correlates with the number of charges removed. Third, peptides corresponding to a basic region in GAP43 and MARCKS effector domain-like regions in other proteins (e.g. MacMARCKS, adducin, Drosophila A kinase anchor protein 200, and N-methyl-d-aspartate receptor) also bind with an energy that correlates with the number of basic residues. Kinetic measurements suggest the effector domain binds to several PIP(2). Theoretical calculations show the effector domain produces a local positive potential, even when bound to a bilayer with 33% monovalent acidic lipids, and should thus sequester PIP(2) laterally. This electrostatic sequestration was observed experimentally using a phospholipase C assay. Our results are consistent with the hypothesis that MARCKS could reversibly sequester much of the PIP(2) in the plasma membrane.

MARCKS (reviewed in Refs. 26 -30), a ubiquitous protein kinase C (PKC) (31) substrate, is present at high concentration in many cell types (e.g. ϳ10 M in brain tissue (27,32)) and is localized to the plasma membrane in quiescent cells. The binding of MARCKS to plasma membranes requires both hydrophobic insertion of its N-terminal myristate into the bilayer and electrostatic interactions between its effector domain and monovalent acidic lipids in the membrane (28). The membranebound basic effector domain produces a significant positive electrostatic potential that can act as a basin of attraction for multivalent acidic lipids such as PIP 2 (8). The electrostatic sequestration of PIP 2 can be reversed either by the binding of calcium/calmodulin to the effector domain or by the PKC phosphorylation of the effector domain, which decreases the positive electrostatic potential (33).
In the work reported here, we investigated whether the binding of MARCKS-(151-175) to PIP 2 is due to nonspecific electrostatic interactions. We used three approaches to determine whether the binding depends on specific residues or just on the number of basic residues. We measured the binding of truncated versions of MARCKS-(151-175) (i.e. peptides missing basic residues from the N-and/or C-terminal regions) to PC/PIP 2 vesicles. We also measured the binding of peptides corresponding to basic regions in other proteins (macrophageenriched myristoylated alanine-rich C kinase substrate (Mac-MARCKS), adducin, Drosophila A kinase anchor protein 200 (DAKAP200), N-methyl-D-aspartate (NMDA) receptor, and growth-associated protein of M r 43,000 (GAP43)) to PC/PIP 2 vesicles to determine whether the binding correlates with the number of basic residues. We compared the binding of peptides with 13 Lys or 13 Arg residues to investigate whether the chemical nature of the basic residues affects the interaction with PIP 2 .
We further tested the hypothesis that several PIP 2 diffuse together to form a binding site for MARCKS-(151-175) (8,25,39) by examining the effect of the mole fraction of PIP 2 in the membrane on the forward rate constant for the binding of the peptides to PC/PIP 2 vesicles. Finally, we determined the relative ability of MARCKS-(151-175) and other basic peptides to sequester PIP 2 laterally in membranes containing physiological concentration of PS by examining the ability of these peptides to decrease the PLC-catalyzed hydrolysis of PIP 2 .
Peptides-Unless specified, all peptides (sequences listed in Table I) were obtained from American Peptide Co., Inc. (Sunnyvale, CA), and were determined to be Ͼ80% pure by reverse phase-high pressure liquid chromatography and MALDI-time-of-flight mass spectroscopy. Each peptide was blocked with an acetyl group at its N terminus and an amide group at its C terminus. A peptide corresponding to a basic region in secretory carrier membrane protein 2 (SCAMP2-(201-211)) was a generous gift from Prof. David Cafiso; its N terminus is unblocked, introducing an extra positive charge.
In the binding and kinetic measurements, we used peptides with an extra Cys at the N terminus, which permitted covalent attachment of either a radioactive ([ 3 H]NEM) or a fluorescent (acrylodan) label. In the potential measurements, we used MARCKS-(151-175) and Lys-13 without the extra N-terminal Cys. In the PLC assays, we used MARCKS-(151-175) and a peptide corresponding to a basic region in phosphatidylcholine-specific phospholipase D 2 (PLD2-(554 -575)) without the extra N-terminal Cys, but GAP43- (30 -56), ⌬N⌬C-MARCKS, Lys-7, and SCAMP2-(201-211) had the extra N-terminal Cys. We added 1 mM dithiothreitol to the buffer to avoid the formation of disulfide bonds when the peptides contain the extra Cys. MARCKS-(151-175) with or without the extra Cys had the same effect on the PLC-catalyzed hydrolysis of PIP 2 .
Peptide Labeling-We used a protocol modified from Molecular Probes (Conjugation with Thiol-reactive Probes) to label peptides with a thiol-reactive fluorescent acrylodan probe, as described in detail elsewhere (44). Briefly, we mixed 1 ml of ϳ1 mM peptide in 10 mM K 2 HPO 4 / KH 2 PO 4 , pH 7.0, with acrylodan probe dissolved in DMF (mole ratio of 1.5:1 acrylodan/peptide) for 1 h, purified the labeled peptide using high pressure liquid chromatography, and checked its purity with MALDI mass spectrometry (CASM, State University of New York, Stony Brook).
We labeled peptides with radioactive [ 3 H]NEM as described previously (45  I Sequences of peptides Basic residues Lys and Arg are in bold. Acidic residues Asp and Glu are in italics. Aromatic residues Phe, Trp, and Tyr are underlined. Unless specified, the N terminus of each peptide is blocked with acetyl and the C terminus is blocked with amide. The exception is that the N terminus of SCAMP2-(201-211) is unblocked, introducing an extra positive charge. The sequences of the basic regions in these proteins are documented in the references indicated. procedure labeled ϳ1% of the peptide. We added to the solution containing the labeled peptide an excess of non-radioactive NEM (mole ratio of 1.5:1 NEM/peptide) to block the unlabeled Cys.
Vesicle Preparations-We used multilamellar vesicles (MLVs) for potential measurements, 100 nm diameter large unilamellar vesicles (LUVs) for fluorescence measurements, and sucrose-loaded LUVs for centrifugation experiments as described in detail elsewhere (25,46). In our experience, the protocol for preparing MLVs must be followed carefully to ensure a uniform distribution of PIP 2 in PC/PIP 2 (99:1) vesicles (either LUV or MLV). We measure the potential of several individual vesicles to assess the uniformity of the MLV preparations. The protocol described below consistently produces a population of vesicles that move with similar velocity in an electric field, i.e. the potentials due to the negatively charged PIP 2 are similar.
The critical step is to produce a dried lipid film in which PC and PIP 2 are mixed uniformly. We add solutions of PIP 2 (it is important to use the ammonium salt, which is more soluble in chloroform than the sodium salt (47,48)) and PC in chloroform (typically 500 l total volume) to a 50-ml round-bottom flask, which is then well immersed in a 30 -35°C water bath and attached to a rotary evaporator. The flask is rotated without vacuum for 5 min to warm the solution and speed the subsequent evaporation under vacuum. Deleting this warming step can produce a strikingly nonuniform distribution of PIP 2 in the MLVs (as indicated by potential measurements), presumably because PIP 2 is less soluble in chloroform than PC and, if the chloroform does not evaporate rapidly, forms a layer on the bottom of the flask. We next applied the maximum vacuum that does not boil the chloroform to evaporate most of the solvent (ϳ1 min) and then applied full vacuum for 30 min to remove all traces of chloroform. Electrophoresis measurements indicated that most of the MLVs produced using this protocol do indeed contain 1% PIP 2 ; specifically, the potentials of the PC/PIP 2 MLVs made in this way had low standard deviation (see e.g. Fig. 7 for 98:2 PC/PIP 2 MLVs). The PC/PIP 2 LUVs, which are made by extrusion of MLVs, presumably also have a uniform fraction of PIP 2 .
Centrifugation Binding Measurements-We measured the binding of [ 3 H]NEM-labeled peptides (Table I) to sucrose-loaded PC/PIP 2 LUVs using the centrifugation technique described previously (25,46). Briefly, sucrose-loaded PC/PIP 2 LUVs were mixed with trace concentrations of [ 3 H]NEM-labeled peptides (typically 2-10 nM). The mixture was centrifuged at 100,000 ϫ g for 1 h. We calculated the percentage of peptide bound from the radioactivity of the peptide in the supernatant and in the pellet.
We use a molar partition coefficient K (49,50) to describe the binding of the peptide to lipid vesicles without making assumptions about the absorption mechanism. The molar partition coefficient K is defined by the equation where K 1 is the molar partition coefficient for the binding of the first peptide; K 2 is the molar partition coefficient for the binding of the second peptide; R is the gas constant; and T is the absolute temperature. At room temperature (T ϭ 295 K), a 10-fold increase in K corresponds to an increase of ⌬⌬G 0 ϳ1.4 kcal/mol. Several experiments (data not shown) provide important controls for the results we obtained using the centrifugation assay. First, the small standard deviation of the potentials of PC/PIP 2 MLVs shows each vesicle has the same fraction of PIP 2 . Second, binding experiments using both radioactive [ 3 H]PIP 2 and [ 14 C]PC as tracers showed the ratio of PC to PIP 2 did not change during the preparation of LUVs. Third, we obtained similar results for the binding of Lys-10 to PC/PIP 2 vesicles using bovine brain PIP 2 from three different sources; thus, it is unlikely that trace contaminants in one particular PIP 2 preparation are affecting the results significantly. Furthermore, the MALDI mass spectra of these PIP 2 lipids are similar to those reported in literature (55). The concentration of PIP 2 in the chloroform stock solution was routinely determined by drying a known volume of the lipid solution and measuring its weight on a microbalance from Cahn Instruments, Inc. (Cerritos, CA); this procedure was checked occasionally with a phosphate assay. Fourth, electron microscope and light scattering experiments showed that the low concentrations (2-10 nM) of peptide we routinely used in the centrifugation assay did not cause significant vesicle aggregation. Fifth, adding 100 M EDTA to the buffer had no effect on the binding of Lys-10 to PC/PIP 2 (99:1) vesicles, which suggests that PIP 2 was not chelated to a significant degree by cations (e.g. Ca 2ϩ , Mg 2ϩ ). We also obtained the same results for the binding of Lys-10 to PC/PIP 2 vesicles using two different buffers (1 mM MOPS (used routinely) and 10 mM HEPES), which suggests the buffer is not affecting the measurements.
Finally, the molar partition coefficient K should not depend on the peptide concentration when [L] Ͼ Ͼ [P] m ; under these conditions the peptide does not bind a significant fraction of the acidic lipid in the vesicle. K was independent of the peptide concentration for the binding of peptides to 5:1 PC/PS vesicles (data not shown). If the prewarming step noted above was deleted during the preparation of PC/PIP 2 vesicles, however, the MLVs often had widely divergent potentials, and the K value measured with a population of LUVs did depend on the peptide concentration.
Despite all these precautions, the results we obtained for the binding of peptides to PC/PIP 2 vesicles were still less reliable than the results we obtained for the binding of the same peptide to PC/PS vesicles (presumably because of variations in the PIP 2 content of the LUVs). For example, we measured the molar partition coefficient of Lys-10 peptide to PC/PIP 2 vesicles with ϳ20 sets of vesicles. For measurements on each set of vesicles, K has a small standard error. Comparisons of measurements on different sets of vesicles, however, produced as much as 5-fold variability in the molar partition coefficient. K varied by only a factor of 2 in measurements with different sets of PC/PS vesicles. We measured the binding of other peptides using at least three sets of vesicles, and we used a least squares fit of the combined data to Equation 1 to obtain the molar partition coefficients, K, reported below. The error bars in Figs. 1, 3, 5, and 8 represent the standard deviations of the K values calculated from the individual measurements using Equation 1 (e.g. the 21 measurements represented by circles for MARCKS-(151-175) data shown in Fig. 1A).
Fluorescence Binding Measurements-We used a centrifugation-independent fluorescence assay (56) to measure the binding of acrylodanlabeled Lys-7 to PC/PIP 2 (99:1) LUVs. Acrylodan is an environmentsensitive fluorescent probe; when it moves from an aqueous solution (high dielectric) to a lipid bilayer (low dielectric), its emission peak blue shifts from ϳ520 to ϳ460 nm, and its fluorescence increases significantly. Thus we can calculate the binding from the change in the fluorescence.
Binding Measurements Using Equilibrium Dialysis-We used a third binding assay, equilibrium dialysis (57), to measure the binding of peptide to vesicles. Each Teflon dialysis cell from Harvard Bioscience, Inc. (Holliston, MA) contains two half-cells separated by a piece of polycarbonate dialysis membrane. We loaded the vesicle solution to one half-cell and the radioactively labeled peptide solution to the other one. After a 24-h equilibrium dialysis, we counted the radioactivity of the peptide in both half-cells and calculated the binding. We obtained the same results by loading the mixture of the peptide and the vesicles into one half-cell and loading buffer to the other one.
We routinely used the centrifugation assay for most of our measurements because it requires only a low peptide concentration (2-10 nM) and takes only a short time (ϳ1 h). It is especially useful for measuring the strong binding of peptide (e.g. MARCKS-(151-175) or Lys-13) to PC/PIP 2 (99:1) vesicles where Ͼ10 nM peptide changes the charge of the vesicle. The fluorescence assay using the acrylodan probe typically requires higher peptide concentrations (Ն50 nM) for a good signal/noise ratio, and the probe introduces some binding energy because of its hydrophobic insertion (44). The equilibrium dialysis assay requires a longer time (ϳ24 h), and the loss of peptide to the dialysis membrane and the cell is significant. It requires peptide concentrations Ն200 nM in our hands.
Kinetic Measurements-Kinetic measurements of MARCKS-(151-175) binding were performed on an SLM-Aminco spectrofluorometer with a stopped-flow attachment as described previously (44). Briefly, 100 nM of acrylodan-MARCKS-(151-175) was mixed rapidly with varying concentrations of PC/PIP 2 LUVs (diameter 100 nm) in a stoppedflow chamber. Because the fluorescence of acrylodan-MARCKS-(151-175) increases when it binds to the vesicles, we measured the time trace of the fluorescence and calculated the relaxation time () for this binding process. The relationship between the relaxation time () and the association rate constant (k on ) is described in Equation 3 (44).
where [V] is the vesicle concentration; k off is the dissociation rate constant; [L] is the lipid concentration of the outer leaflet of the vesicle; is the number of lipid molecules on the outer leaflet per vesicle ( ϭ 4R V 2 /A L ϭ 4.5 ϫ 10 4 , where R V , the radius of the LUVs, is 50 nm and A L , the area per lipid molecule, is 0.7 nm 2 (44)). Thus we can calculate the association rate constant (k on ) and (much less accurately) the dissociation rate constant (k off ) from the plot of 1/ versus the lipid concentration. We determined the dissociation rate constant (k off ) more accurately by measuring the relaxation time of peptide transfer from donor to acceptor vesicles as described elsewhere (44).
Potential Measurements-As described previously (25, 58), we measured the electrophoretic mobility (velocity/field) of MLVs with or without the addition of basic peptides and calculated the potential, the electrostatic potential at the shear plane, using the Helmholtz-Smoluchowski Equation 4 (59, 60), where is the potential of a vesicle; u is the velocity of the vesicle in a unit electric field; is the viscosity of the aqueous solution; ⑀ r is the dielectric constant of the aqueous solution; and ⑀ 0 is the permittivity of free space. The potential is proportional to the surface charge density and thus to the number of charged peptides that absorb to the vesicles (59). PLC Monolayer Assay-We measured the effect of peptides on the PLC-catalyzed hydrolysis of PIP 2 in monolayers as described in Refs. 25, 61, and 62. Briefly, we mixed PC, PS, and [ 3 H]PIP 2 in chloroform to form a 55 M lipid stock of PC/PS/[ 3 H]PIP 2 (66:33:1). We then slowly deposited the lipid stock onto the surface of a 15-ml solution in a 5-cm diameter Teflon trough with a stirrer at the bottom. The subphase solution contained 100 mM KCl, 25 mM HEPES, 0.1 mM EGTA, 1 mM dithiothreitol, pH 7.0. Once the chloroform had evaporated (10 min), we measured the surface pressure of the monolayer (typically 25 mN/m with ϳ60 l of the lipid stock) using a square piece of filter paper and a balance from Nima Technology Ltd. (Coventry, UK). We then added peptide (e.g. MARCKS-(151-75) or GAP43- (30 -56)) to the subphase. After 3 min, we added PLC-␦ 1 (final concentration Ϸ0.1 nM), and then after another 3 min, we added 0.1 mM CaCl 2 (ϳ1 M free Ca 2ϩ ) to the subphase to activate PLC-␦ 1 . We collected 200-l aliquots of the subphase at different time after the addition of Ca 2ϩ , measured [ 3 H]IP 3 produced by the hydrolysis of PIP 2 , calculated %PIP 2 hydrolyzed, and plotted it versus time. The data obtained in the first 5 min could be described well with a straight line, and the slope of the line was defined as the initial rate of hydrolysis, as described in detail elsewhere (25).
We could not obtain reliable data with a myristoylated CAP23-(1-13) peptide because it forms micelles or some type of aggregate instead of monomers in solution under our conditions.
Electrostatic Equipotential Calculations-We built an atomic model of MARCKS-(151-175) in an extended conformation using the Insight Biopolymer and Discover modules (Accelrys). The atomic radii and partial charges assigned to the peptide were taken from the CHARMM forcefield (63). Three lipid bilayers, 5:1 PC/PS, 99:1 PC/PIP 2 , and 2:1 PC/PS, were built as described in detail elsewhere (64). The structure of PIP 2 is unknown. A model for the PIP 2 head group was taken from the structure of the pleckstrin homology domain of PLC-␦ 1 bound to IP 3 (Protein Data Bank code 1MAI) (65). We assumed that the head group of PIP 2 is perpendicular to the surface, similar to the orientation of phosphatidylinositol (66, 67). We derived a detailed partial charge distribution for its atoms from similar functional groups in the CHARMM parameterization set.  (68), a program for the visualization of the structural and biophysical properties of biological macromolecules. GRASP, however, is capable of solving only the linearized version of the Poisson-Boltzmann equation. In order to represent accurately the electrostatic properties of the MARCKS-(151-175)/membrane systems, we solved the nonlinear Poisson-Boltzmann (NLPB) equation (33,51,64,69,70) for atomic models of these systems in 100 mM KCl on a finite difference grid of size 65 3 and used focusing boundary conditions (51) to a final resolution of 0.5 grid/Å. The resulting potential maps as well as the atomic coordinates for the MARCKS-(151-175)/membrane models were then read into GRASP and displayed on an SGI Octane Work station. The net charges used in the calculations are 0 for PC, Ϫ1 for PS, and Ϫ4 for PIP 2 . The calculations were performed initially using the partial charge sets for PC and PS described previously (64) and for PIP 2 as described above. Essentially identical results were obtained for the Ϫ25 mV potential profiles Ͼ0.2 nm from the surface of the bilayer when a simpler charge set was assigned to the lipids. Specifically, we assigned a charge of Ϫ0.25 to each of the 4 oxygen atoms forming bonds with the phosphorus atom in PS, a charge of Ϫ0.33 to each of the 12 oxygen atoms forming bonds with the 3 phosphorus atoms in PIP 2 , and a charge of 0 to the remaining atoms in PC, PS, and PIP 2 . This removes the ϩ25 and Ϫ25 mV potential profiles that extend between the positive and negative charges on the zwitterionic PC lipids, which tend to clutter up the diagram.

The Binding of Truncated MARCKS-(151-175) Peptides to PC/PIP 2 Vesicles Correlates with the Number of Basic
Residues-We used truncated MARCKS-(151-175) peptides to determine how the binding depends on the number of basic residues. These peptides (sequences listed in the first 4 lines of Table I) include the effector domain peptide (MARCKS-(151-175)), a peptide lacking 3 Lys residues at the C terminus (⌬C-MARCKS), a peptide lacking 5 Lys residues at the N terminus (⌬N-MARCKS), and a peptide lacking all 8 Lys residues (⌬N⌬C-MARCKS). We measured the binding of these radioactively labeled peptides to PC/PIP 2 vesicles using the centrifugation assay, and we calculated the molar partition coefficient K (Equation 1). The molar partition coefficient is equal to the reciprocal of the lipid concentration that produces 50% peptide binding. As shown in Fig. 1A, deleting the 5 Lys residues at the N terminus of MARCKS-(151-175) decreases the binding ϳ40-fold; deleting 8 Lys residues decreases the binding ϳ1,000-fold. Fig. 1B shows that the change in the binding energy (Equation 2) correlates with the change in the number of basic residues in the peptide. Each basic residue we removed appears to contribute about 0.5 kcal/mol to the binding energy.
The Binding Does Not Depend on the Chemical Nature of the Basic Residues-If the binding is driven mainly by electrostatics, it should not depend on the chemical nature of the basic residues. Thus we examined the binding of peptides consisting of either 13 Lys or 13 Arg residues to PC/PIP 2 vesicles. Fig. 2A shows that both Lys-13 (f) and Arg-13 () bind with the same affinity to 99:1 PC/PIP 2 vesicles. Fortuitously, both Lys-13 and Arg-13 bind to PC/PIP 2 vesicles with the same affinity as MARCKS-(151-175) in 100 mM monovalent salt.
The Aromatic Residues and Length of the Peptide Affect the Binding-To investigate the effect of aromatic residues on the binding, we replaced each of the 5 Phe in MARCKS-(151-175) with Ala (peptide defined as FA-MARCKS in Table I); we chose Ala because experiments with model peptides suggest this residue is neither attracted to nor repelled from the interface (71). We measured the binding of the peptides to PC/PIP 2 or PC/PS vesicles using the centrifugation assay. Fig. 2A shows that FA-MARCKS (‚) binds 100-fold less strongly to 99:1 PC/PIP 2 vesicles than MARCKS-(151-175) (E). The observation that the 5 Phe residues contribute significantly (ϳ3 kcal/mol) to the binding energy is consistent with previous observations that the Phe residues of MARCKS-(151-175) penetrate the polar head group region of the membrane (40,45,72). The aromatic residues contribute to the binding of MARCKS-(151-175) not only to PC/PIP 2 vesicles but also to PC/PS vesicles. Fig. 2B shows that FA-MARCKS (‚) binds 100-fold less strongly to 5:1 PC/PS vesicles than MARCKS-(151-175) (E). Our results are consistent with the work of Wimley and White (71), which showed that if a Phe on a neutral peptide inserts completely into the polar head group region of a membrane, it contributes ϳ1 kcal/mol to the binding energy. Thus 5 Phe could maximally contribute 5 kcal/mol, which corresponds to ϳ1,000-fold increase in the molar partition coefficient (Equation 2). There are several possible explanations for the less-than-a-maximal effect observed with our charged peptides. For example, a decrease in free energy due to penetration of aromatics into the bilayer is accompanied by an increase in Born/dehydration free energy required to move the charges closer to the interface, as discussed in detail elsewhere (33,45).
The length of the peptide also appears to affect the binding. Previous experiments have shown that inserting 1 or 2 Ala between each of the basic residues in peptides containing 7 Lys (56), 5 Lys, or 5 Arg (73) residues decreases the binding of the peptides to PC/PS or PC/PG vesicles. The mechanism is not understood; the effect could arise because the local positive potential is reduced or because an extra entropy loss occurs on binding of the longer peptides to the interface (see Ref. 74). For both 99:1 PC/PIP 2 ( Fig. 2A) and 5:1 PC/PS (Fig. 2B) vesicles, FA-MARCKS binds ϳ100-fold less strongly than Arg-13 even though both FA-MARCKS and Arg-13 have 13 basic residues. The simplest explanation is that the length is affecting the binding.
We note in passing that although Lys-13 and Arg-13 bind with similar affinity to PC/PIP 2 vesicles ( Fig. 2A), Lys-13 binds to 5:1 PC/PS vesicles about 10-fold less strongly than Arg-13  Tables I and II. Note that Lys-13 and Arg-13 bind with the same affinity to the PC/PIP 2 vesicles; the affinity is independent of the chemical nature of the basic residues and is identical to that of MARCKS-(151-175). Replacing the 5 aromatic Phe residues with 5 Ala residues (FA-MARCKS, ‚) reduces the binding of the MARCKS effector domain peptide (E) ϳ300-fold. The two curves are drawn with molar partition coefficients K ϭ 2 ϫ 10 6 and 6 ϫ 10 3 M Ϫ1 , respectively. Shortening the peptide increases the binding: although both Lys-13 and FA-MARCKS have 13 basic residues, Lys-13 (13 residues) binds to PC/PIP 2 vesicles ϳ300-fold more strongly than FA-MARCKS (25 residues). B, binding of MARCKS-(151-175) (E), Arg-13 (), and FA-MARCKS (‚) to 5:1 PC/PS vesicles. Note that replacing the aromatics with Ala has the same ϳ300-fold effect on the binding to either PC/PS or PC/PIP 2 vesicles. Shortening the peptide also has the same effect. Binding data in A and B also show that MARCKS-(151-175) (E), Arg-13 (), and FA-MARCKS (‚) bind with similar affinities to vesicles containing either 1% PIP 2 or 17% PS. The curves in B are drawn with molar partition coefficients 2 ϫ 10 6 and 7 ϫ 10 3 M Ϫ1 . The K values for the binding of the peptides to PC vesicles are less than 1 ϫ 10 2 M Ϫ1 (data not shown). Binding measurements were done as described in the legend to Fig. 1. ( Table III). This is consistent with the observation that Lys-5 binds to 4:1 PC/PG vesicles ϳ10-fold less strongly than Arg-5 (75), for reasons not understood. Fig. 2 and Table III 25). All the experimental data we have obtained suggest the driving force for the binding of the peptides we have studied to PC/PIP 2 vesicles is mainly electrostatic in nature; hence, we conclude that the local potential the adsorbed peptide experiences is more negative than the average potential. The simplest explanation is that when a basic peptide (e.g. MARCKS-(151-175)) binds to PC/PIP 2 vesicles, several PIP 2 s must diffuse into the neighborhood of the adsorbed peptide and produce a high local negative potential, as discussed in more detail below.
Binding of Poly-Lys Peptides to PC/PIP 2 Vesicles Also Correlates with the Number of Basic Residues in Each Peptide-MARCKS-(151-175) contains both basic and aromatic residues. Its binding to PC/PIP 2 vesicles depends on a complicated interplay of at least two factors that drive adsorption, electrostatic attraction of basic residues for the surface and hydrophobic insertion of aromatics into the polar head group region, and two factors that oppose adsorption, Born/dehydration repulsion and the entropy price required to pull several PIP 2 lipids together. These factors (except the entropy price) have been discussed elsewhere (33,45). We investigated a simpler system, measuring the binding of peptides that contain only basic residues. Fig. 3A shows the binding of Lys-7, Lys-10, and Lys-13 to PC/PIP 2 vesicles. Fig. 3B shows that the change in the binding energy (Equation 2) correlates with the change in the number of basic residues. Specifically, adding 6 Lys residues to Lys-7 increases the molar partition coefficient 10 3 -fold, i.e. increases the binding energy ϳ4 kcal/mol (Equation 2). (Or each basic residue contributes about 0.7 kcal/mol to the binding energy.) The simplest interpretation is that the binding energy increases because the peptides with more basic residues can bind more PIP 2 molecules (probably 3 or 4 for Lys-13 and MARCKS-(151-175)); EPR and electrokinetic data suggest MARCKS-(151-175) forms an electroneutral complex with several PIP 2 s (25,39).
Different Assays Produce Similar Binding Data-We also used a centrifugation-independent fluorescent assay to measure the binding of acrylodan-labeled Lys-7 to PC/PIP 2 vesicles. Fig. 4 shows that acrylodan-labeled Lys-7 binds weakly to PC vesicles but binds significantly to PC/PIP 2 (99:1) vesicles. This binding is ϳ10-fold stronger than that of radioactive NEMlabeled Lys-7 (Fig. 3), probably due to the effect of hydrophobic acrylodan enhancing the binding of Lys-7 to vesicles as noted previously with PC/PS (56). The third assay we used is equilibrium dialysis. The results (data not shown) agree with those obtained using the centrifugation assay (Fig. 3); radioactive NEM-labeled Lys-10 binds weakly to PC vesicles but binds with significant affinity to 99:1 PC/PIP 2 vesicles. Fig. 5A shows that the binding of both MARCKS-(151-175) (25) and Lys-13 to 99:1 PC/PIP 2 vesicles decreases as the salt concentration increases. This result is expected if electrostatic interactions contribute significantly to the binding. Increasing the ionic strength also decreases the binding of basic peptides to PC/PS (5:1) vesicles (Fig. 5B). A theoretical model explaining why the binding of basic peptides to membranes containing monovalent acidic lipids depends on salt concentration is given elsewhere (33,51). Briefly, increasing the salt concentration screens the charges on the membrane interface and thus decreases the magnitude of the negative electrostatic potential experienced by the basic peptide at the membrane surface. ficient arises from an increase in the dissociation rate constant (29,44). This result is consistent with theoretical (77) and experimental (78,79) evidence that monovalent acidic lipids do not significantly redistribute when a basic peptide binds. There is, however, evidence that PIP 2 redistributes when MARCKS-(151-175) binds to PC/PIP 2 vesicles. Specifically, both potential (25) and more direct EPR experiments (39) suggest that one MARCKS-(151-175) binds several (three or four) PIP 2 to form an electroneutral complex. If one MARCKS-(151-175) pulls together several PIP 2 when it binds to a PC/PIP 2 membrane (Ͻ1% PIP 2 ), the association rate constant, k on , should decrease as the mole fraction of PIP 2 on the membrane decreases because the PIP 2 must diffuse further to associate with the peptide. We did indeed observe this effect. Fig. 6 shows that k on for the interaction between MARCKS-(151-175) and 99:1 PC/PIP 2 vesicles is 1 ϫ 10 11 M Ϫ1 s Ϫ1 , which is close to the diffusion-limited rate (assuming the diffusion coefficient of the peptide in the aqueous solution is 3 ϫ 10 Ϫ6 cm 2 s Ϫ1 , see Ref. 44). When we reduce the mole fraction of PIP 2 from 1 to 0.1%, k on (proportional to the slope) decreases ϳ10-fold and is no longer diffusion-limited (Fig. 6).

The Binding of Basic Peptides to PC/PIP 2 Vesicles Decreases as the Ionic Strength Increases-
The molar partition coefficient K also decreases ϳ10-fold (25). As expected from the measurements of the forward rate constant and equilibrium partition coefficient, the dissociation rate k off (from independent measurements described above) does not depend on the percentage of PIP 2 in the vesicles and is ϳ1 s Ϫ1 (data not shown). Thus the change in K is due only to the reduced association rate k on . (The data we obtained are overdefined and self-consistent; the molar partition coefficient K for the interaction between MARCKS-(151-175) and 99:1 PC/PIP 2 vesicles deduced from kinetic measurements of k on and k off is 2 ϫ 10 6 M Ϫ1 (K ϭ k on /( k off ), see Ref. 44), which is consistent with the value of K from direct binding measurements (Fig. 1).) If one MARCKS-(151-175) pulls together several PIP 2 when it binds to a PC/PIP 2 membrane, k on should decrease not only if the distance the PIP 2 lipids must diffuse increases (Fig. 6) but also if the viscosity of the membrane increases. As expected, increasing the viscosity by incorporating 30% cholesterol into 99:1 PC/PIP 2 vesicles decreases the association rate constant significantly (ϳ5-fold; data not shown).
The decrease in the forward rate constant illustrated in Fig.  6 probably is not due to a significant change in peptide conformation upon binding to the PC/PIP 2 surface. Measurements with spin-labeled MARCKS-(151-175) peptides show that the peptide remains in an extended conformation and exhibits little ␣-helical structure when it binds to the PC/PIP 2 membrane (39), a result consistent with our CD measurements (data not shown).
Potential Measurements Suggest That Both MARCKS- (151-175) and Lys-13 Bind Strongly to PC/PIP 2 Vesicles-The potential is proportional to the surface charge density, and thus to the number of charged peptides that absorb to the vesicles (59). This allows us to study peptide binding by measuring the effect of peptides on the potential of PC/PIP 2 (98:2) vesicles. (The MLVs were present at sufficiently low concentration to bind only an insignificant fraction of the peptide.) Fig. 7 shows that adding 10 Ϫ7 M MARCKS-(151-175) or Lys-13 changes the potentials of 98:2 PC/PIP 2 vesicles from Ϫ14 to about Ϫ7 mV. Thus, Lys-13 and MARCKS-(151-175) bind with similar affinity when present at low concentrations; this result is consistent with the results obtained in Fig. 2A using a different technique. The calculation of binding constants from these potential data is model-dependent and is not explored further in this communication.
Because we measured the potential of single vesicles, the small standard deviations shown for the potentials of PC/PIP 2 vesicles in the absence of peptides (left of the break in Fig. 7  to the effector domain of MARCKS, as illustrated in Table I. MacMARCKS (also known as MARCKS-related protein or F52) is a MARCKS family member highly expressed in macrophages (26,27,30). Adducin is a membrane skeletal protein that binds to F-actin and assists the recruitment of spectrin to F-actin (reviewed in Ref. 80). DAKAP200 is a Drosophila scaffolding protein that binds regulatory subunits of protein kinase AII (81). NMDA receptors act as glutamate-gated ion channels in the central nervous system (reviewed in Ref. 82). MARCKS (83)(84)(85), MacMARCKS (86,87), DAKAP200 (81), adducin (88), and the NMDA receptor (89,90) all bind with high affinity to calcium/calmodulin and can be phosphorylated by PKC. These proteins all contain basic regions that constitute calcium/calmodulin binding domains and also have serine residues that can be phosphorylated by PKC (Table I). GAP43 is present at high concentrations in neural tissue and plays important roles in nerve growth (reviewed in Refs. [91][92][93]. Even though the basic region in GAP43 has little sequence homology to the MARCKS effector domain (see Table I), Laux et al. (24) have postulated that GAP43, like MARCKS, may bind a significant fraction of PIP 2 in neural tissues. Fig. 8 summarizes data (not shown) similar to those illustrated in Fig. 1A, Fig. 2A, and Fig. 3A for peptides corresponding to the basic regions in these proteins. All the peptides bind significantly to PC/PIP 2 vesicles (K values also listed in Table  II), and the binding correlates qualitatively with the number of basic (and aromatic) residues in the peptides, consistent with the hypothesis that the binding is mainly driven by local electrostatic interactions.
As summarized elsewhere, there is good evidence that a cluster of basic residues in the neuronal Wiskott-Aldrich syndrome protein (N-WASP) binds PIP 2 (94). Clusters of basic residues in WASP (94), cortical cytoskeleton-associated protein of approximate molecular mass 23 kDa (CAP23) (24), and Syndecan-4 (95) may also interact with PIP 2 . Peptides corresponding to the basic regions in these proteins, however, bind only weakly to PC/PIP 2 vesicles (sequences and binding data in Tables I and II). The result was expected because these peptides have Յ8 basic and 0 or 1 aromatic residues. More basic/ aromatic residues appear to be required to bind an unstructured peptide strongly to a PC/PIP 2 vesicle (see Fig. 1B, Fig.  3B, and Fig. 8). There are several reasons why a cluster of basic residues in a structured protein such as N-WASP might inter-  Tables I and II. We measured the molar partition coefficients K using the centrifugation technique as described in the legend to Fig. 1. Each molar partition coefficient is obtained by fitting binding data of at least three sets of vesicles with Equation 1.  1 Ͻ10 2 act more strongly with PIP 2 than the corresponding unstructured peptide. On the other hand, the lack of a significant PIP 2 /CAP23-(1-13) interaction noted in Tables II and IV probably can be extrapolated to the intact protein, because CAP23 is "natively unstructured." As discussed in the reviews cited in the Introduction, many proteins are capable of binding PIP 2 . For example, Janmey and co-workers (see Ref. 96 and references therein) have shown that gelsolin and a fluorescently labeled peptide corresponding to a basic region in the protein bind PIP 2 . Dell'Acqua and co-workers (97,98) have shown that clusters of basic residues in AKAP79 interact with acidic lipids, including PIP 2 . For the remainder of this paper we focus on MARCKS, MacMARCKS, and GAP43 because these proteins (and potentially DAKAP200, AKAP79) are present at sufficiently high concentrations in at least some cell types to act as PIP 2 buffers.
Electrostatic Equipotential Profiles of Membranes with or without MARCKS-(151-175)-To illustrate the role of electrostatic interactions in the binding of MARCKS to bilayers, we calculated the electrostatic equipotential profiles adjacent to 5:1 PC/PS, 99:1 PC/PIP 2 , and 2:1 PC/PS membranes with or without a single membrane-associated MARCKS-(151-175) (Fig. 9). The electrostatic equipotential profiles (Ϫ25 and ϩ25 mV) were calculated from finite difference solutions to the NLPB equation (33,51,64,69,70) for atomic models of the membrane and peptide, and assuming a solution containing 100 mM monovalent salt. Fig. 9A shows the Ϫ25-mV equipotential profile (red) adjacent to a 5:1 PC/PS membrane. When MARCKS-(151-175) binds to the membrane (Fig. 9B), a strong positive potential (ϩ25 mV, blue) is produced in its vicinity. Theory (77) and experiments (78,79) suggest that the acidic lipid PS is not significantly concentrated in the plane of the membrane when the basic peptide binds to the membrane surface. Fig. 9C shows a model with 4 PIP 2 lipids in a patch of membrane containing ϳ400 lipids, i.e. a bilayer containing 1% PIP 2 . The equipotential profiles (Ϫ25 mV, in red) of this 99:1 PC/PIP 2 membrane are discrete and centered around each PIP 2 ; to a first approximation they are hemispheres with a value about twice that calculated using Debye-Hü ckel theory (the potential doubles when charges are confined to the interface because of the image charge effect and another factor, as discussed in Refs. 59 and 99). Both experimentally (25,76) and computationally, the average surface potential adjacent to a 99:1 PC/PIP 2 membrane (Ϫ8 mV; Fig. 9C) is significantly less negative than that adjacent to a 5:1 PC/PS membrane (Ϫ30 mV; Fig. 9A). Nevertheless, MARCKS-(151-175) and other peptides bind with similar affinity to these two membranes (Table III and Fig. 2). The binding is driven mainly by electrostatic interactions, implying that MARCKS-(151-175) experiences similar local electrostatic potentials in both membranes. This could happen if the peptide sequesters several PIP 2 , as illustrated in Fig. 9D. EPR (39) and kinetic (see Fig. 6 above) data provide more direct evidence that MARCKS-(151-175) sequesters several PIP 2 lipids in its immediate neighborhood when it binds to a PC/PIP 2 membrane.
Is the electrostatic interaction sufficient to offset the entropy price required to sequester (i.e. concentrate laterally) the PIP 2 ? Calculations (as outlined in Ref. 33) using solutions to the NLPB equation suggest that the partitioning of a single PIP 2 from bulk membrane to a position adjacent to membrane-absorbed MARCKS-(151-175) produces a decrease of ϳ3 kcal/mol in the electrostatic free energy. This is sufficient to overcome the decrease in mixing entropy.
The phospholipids in the inner leaflet of the plasma membrane of a typical mammalian cell contain not only ϳ1% PIP 2 but also ϳ30% PS. Fig. 9E shows the essentially flat Ϫ25-mV equipotential profile (red) adjacent to a 2:1 PC/PS membrane; this profile is located ϳ10 Å above the membrane. (The simpler Gouy-Chapman theory, which assumes the charges are smeared uniformly over the surface, predicts an essentially identical profile (64) that agrees well with experimental data (59) Fig. 2, A and B and similar experiments. We tested this prediction in a PLC assay as described below. MARCKS and GAP43 Laterally Sequester PIP 2 -We reported previously (41) that both native MARCKS and MARCKS-(151-175) inhibit the PLC-catalyzed hydrolysis of PIP 2 in vesicles containing 33% PS and 1% PIP 2 . Similar effects were seen in monolayer experiments, which avoid potential artifacts due to vesicle aggregation (25). We measured the ability of other peptides to inhibit the PLC-catalyzed hydrolysis of PIP 2 on PC/PS/PIP 2 (66:33:1) monolayers; Table IV summarizes the results. 10 M GAP43- (30 -56) inhibited the PLCcatalyzed hydrolysis of PIP 2 by Ͼ90%, but 10 M non-myristoylated CAP23-(1-13) had no effect on the PLC-catalyzed hydrolysis of PIP 2 . Because both MARCKS and GAP43 are present at high concentration in at least some cell types, the PLC inhibition experiments suggest that they could bind and sequester much of the PIP 2 in the plasma membrane of these cells.
We also tested the ability of ⌬N⌬C-MARCKS, a truncated MARCKS peptide with 5 basic residues and 5 aromatic residues (Table I), to inhibit the PLC-catalyzed hydrolysis of PIP 2 (Table IV). ⌬N⌬C-MARCKS binds to PC/PS (5:1) vesicles 100fold less strongly than MARCKS-(151-175) (Table III), and ϳ100-fold higher concentration of ⌬N⌬C-MARCKS than MARCKS-(151-175) was required to inhibit the PLC-catalyzed hydrolysis of PIP 2 in PC/PS/PIP 2 (66:33:1) monolayers by Ͼ90% (Table IV). Because ⌬N⌬C-MARCKS has only 5 basic residues, it probably sequesters only 1 PIP 2 in the monolayer. Other small peptides with both basic and aromatic residues, e.g. 10 M PLD2-(554 -575) (6 positive charges and 4 aromatics) and 1 M SCAMP2-(201-211) (4 positive charges and 4 aromatics) also inhibited the PLC-catalyzed hydrolysis of PIP 2 by Ͼ90% (Table IV). In contrast, 100 M Lys-7, which binds significantly to 2:1 PC/PS vesicles, had no effect on PIP 2 hydrolysis in PC/PS/PIP 2 (66:33:1) monolayers (Table IV). Thus aromatic residues appear to be important not only in the binding of a peptide to PC/PIP 2 vesicles (Fig. 2) but also in lateral sequestration of PIP 2 in PC/PS/PIP 2 membrane. Aromatics could act in several different ways when they insert into the bilayer, e.g. position the charges on the basic peptide closer to the PIP 2 or increase the potential produced by the basic residues as revealed by calculations using the NLPB equation (not shown). The increase in potential produced as a single charge approaches a more idealized low dielectric surface as described elsewhere (99). Additional computational and experimental work is required to tease out the role that aromatics play in increasing the sequestration of PIP 2 by clusters of basic residues in peptides and proteins. DISCUSSION Several experimental results support our conclusion that the binding of MARCKS-(151-175) to PC/PIP 2 membranes is driven mainly by local, nonspecific, electrostatic interactions between the 13 basic residues on the peptide and the multivalent PIP 2 lipids. First, the binding of truncated MARCKS-(151-175) peptides to PC/PIP 2 vesicles correlates with the number of basic residues in the peptides (Fig. 1). Second, the binding of peptides corresponding to MARCKS-like domains in other proteins also correlates with the number of basic (and aromatic) residues (Fig. 8). Third, the binding does not depend on the chemical nature of the basic residues; Lys-13 and Arg-13 bind with the same affinity to PC/PIP 2 vesicles (Fig. 2). Fourth, MARCKS-(151-175) binds equally well to membranes containing PI(4,5)P 2 and PI(3,4)P 2 (25). Fifth, the binding decreases as the ionic strength increases (Fig. 6). The atomic model in Fig.  9D, where the potentials have been calculated from the NLPB equation, illustrates the electrostatic component of the binding.
Aromatic residues and the length of the peptide also affect the binding to PC/PIP 2 membranes. Replacing 5 Phe with Ala in MARCKS-(151-175) decreased the binding 100-fold (Fig. 2), and shortening the peptide (25 to 13 residues) while keeping the same number of basic residues increased the binding 100fold (Fig. 2).
Several experiments provide strong evidence that one MARCKS-(151-175) binds to several PIP 2 in a PC/PIP 2 bilayer. EPR measurements on spin-labeled PIP 2 (39) provide perhaps the most direct evidence, supported by the kinetic data reported here (Fig. 6). potential (Fig. 7) and competition experiments (25) also support this conclusion.
Finally, our PLC inhibition measurements suggest that peptides corresponding to the basic domains in MARCKS and GAP43 can laterally sequester PIP 2 , even when the monovalent acidic lipid PS is present at a 30-fold excess. These experiments provide support for the hypothesis discussed below that these domains can reversibly sequester PIP 2 in the plasma membrane of a cell, where PS is typically present at a much higher concentration than PIP 2 .
Comparing the binding of MARCKS effector domain peptide and the well characterized PH domain of PLC-␦ 1 (PLC-PH) to PIP 2 reveals several important differences. First, MARCKS-(151-175) lacks structure and is in an extended conformation in solution or when bound to a membrane (39,40), and its binding to PIP 2 is driven by nonspecific local electrostatic interactions. In contrast, PLC-PH has a well defined structure and its binding to PIP 2 is mediated by 12 specific "lock and key" hydrogen bonds (65). This explains why MARCKS-(151-175) exhibits no specificity for PI(4,5)P 2 over PI(3,4)P 2 , whereas PLC-PH binds more strongly to the former lipid, a specificity that is well understood from the crystal structure of its complex with IP 3 (65). Second, one MARCKS-(151-175) binds to 3-4 PIP 2 , whereas PLC-PH forms a 1:1 complex with PIP 2 ; this explains why MARCKS-(151-175) binds 100-fold more strongly to 99:1 PC/PIP 2 vesicles than does PLC-PH (8). Third, MARCKS-(151-175) binds strongly to both PC/PS and PC/PIP 2 membranes (Fig. 2), whereas PLC-PH binds with significant affinity only to membranes containing PIP 2 (100). A corollary is that the targeting of MARCKS-(151-175) to membranes does not require PIP 2 , whereas the targeting of PLC-PH to membranes does require PIP 2 . Fourth, the functions of the two domains are different. The putative function of MARCKS is to reversibly sequester PIP 2 in the plane of the membrane by nonspecific local electrostatic interactions, whereas the well established functions of PLC-PH include targeting the enzyme to the plasma membrane and increasing the local concentration of substrate PIP 2 that the catalytic domain experiences, which  2 [Peptide] indicates the concentration of peptide required to produce Ͼ90% inhibition of the PLC-catalyzed hydrolysis of PIP 2 ; Ͼ indicates no effect was observed at the concentration indicated. In the PLC assay experiments, peptides PLC-␦ 1 (final concentration Ϸ0.1 nM) and CaCl 2 (ϳ1 M free Ca 2ϩ ) were added sequentially to the subphase of monolayers containing 66 allows the enzyme to act precessively (9,10). The mechanism by which MARCKS binds to a phospholipid bilayer or to a plasma membrane is well understood and requires the electrostatic interaction of its basic effector domain with acidic lipids (28,30). Our working hypothesis is that the basic effector domain sequesters much of the PIP 2 in the membrane through nonspecific local electrostatic interactions. Specifically, the local potential adjacent to the effector domain of MARCKS is positive (Fig. 9, B and F). Because PIP 2 has higher valence (Ϫ3 or Ϫ4) than PS (Ϫ1), it will be more strongly attracted to the positive potential adjacent to the cluster of basic residues (8). Our PLC experiments show that both MARCKS and its effector domain inhibit the PLC-catalyzed hydrolysis of PIP 2 in vesicles or monolayers containing ϳ1% PIP 2 and 33% PS (25,41); earlier studies showed that calcium/ calmodulin binding or PKC phosphorylation releases the effector domain of MARCKS from the membrane, restoring the rate of PLC-catalyzed hydrolysis of PIP 2 (33,41). Our results showing that peptides corresponding to the basic regions of other proteins (e.g. GAP43) inhibit the PLC-catalyzed hydrolysis of PIP 2 support the hypothesis that they too may sequester PIP 2 in the plasma membrane.
One important caveat is that, in most cell types, the concentration of MARCKS and other putative PIP 2 buffers (e.g. GAP43) is not well established and may not be sufficiently high to buffer a significant fraction of the PIP 2 . For example, although the concentration of MARCKS in brain tissue is sufficiently high (10 M, see Refs. 27 and 32) to sequester most of the PIP 2 , the concentration of MARCKS/MacMARCKS in quiescent macrophages is probably too low to perform this function. Upon activation of the macrophage, however, the concentration of MacMARCKS increases 20-fold to about 5-10 M (calculated from the data in Ref. 101) 2 and the concentration of MARCKS also increases (102).
Biological Corollaries-Caroni and colleagues (24) have reported evidence from cell biology experiments that supports the hypothesis MARCKS sequesters a significant fraction of PIP 2 in the plasma membrane. For example, if the hypothesis is correct, overexpression of MARCKS might be expected to increase the synthesis of PIP 2 to maintain a constant free level of PIP 2 . Indeed, overexpression of MARCKS in PC12 cells does increase the production of PIP 2 (24). Furthermore, MARCKS is not uniformly distributed in the plasma membrane of some cell types (103); it is concentrated in the membrane ruffles of fibroblasts (104) and, along with MacMARCKS, in the nascent phagosomes of macrophages (105,106). If MARCKS sequesters PIP 2 , this lipid should colocalize with MARCKS in these regions, and it does (107)(108)(109). However, elucidating the mechanism of colocalization is complicated by the fact that PIP 2producing kinases are also concentrated in ruffles and nascent phagosomes. Thus both local synthesis and sequestration could contribute to the accumulation of PIP 2 , as discussed elsewhere (8). If only local synthesis were important, the concentration of PIP 2 should decrease gradually from the membrane ruffle or nascent phagosome to the bulk plasma membrane because of diffusion (110). If local electrostatic sequestration by MARCKS is also important, sharp demarcation lines between regions of high PIP 2 concentration and the bulk of the plasma membrane could exist. In fact, such demarcations (i.e. steep gradients) for polyphosphoinositides are indeed observed at the borders of membrane ruffles 3 and nascent phagosomes (109,110). The interpretation of these experiments, however, is not straightforward because the probe used to detect the lateral organiza-tion of PIP 2 in the plasma membrane, the PH domain of PLC-␦ 1 , is itself negatively charged when bound to PIP 2 , and a green fluorescent protein construct of this PH domain, like PIP 2 , is also electrostatically sequestered by the basic residues in the effector domain of MARCKS. 4 In summary, experiments from cell biology (24) and model systems (reviewed in Ref 8) provide complementary evidence that MARCKS (and possibly GAP43, see Ref. 24) could act to reversibly buffer PIP 2 in the plasma membrane of many cell types.