INTRODUCTION

Proteins of the MARCKS¹ family are protein kinase C (PKC) substrates that are phosphorylated during major cellular events such as neurosecretion, fibroblast mitosis, and macrophage activation (1, 2). The N-terminal glycine residue of MARCKS proteins is myristoylated through a reaction catalyzed by myristoyl CoA:protein N-myristoyltransferase (NMT) (3, 4). Studies with acylated peptides as well as with the intact protein suggest that the myristoyl moiety is involved in membrane binding (5, 6, 7, 8, 9). MARCKS proteins also contain a highly basic domain, the phosphorylation site (PS) domain, which is phosphorylated following activation of cellular PKC with phorbol esters (10), which binds with high affinity to calmodulin (CaM) in the presence of calcium (Ca²⁺) (11) and which cross-links actin filaments in vitro (12).

Phosphorylation of serine residues in the PS domain regulates these interactions by disrupting the MARCKS-calmodulin complex (11), by inhibiting the actin filaments cross-linking activity of MARCKS (12) and by decreasing the affinity of the PS domain for negatively charged membranes such as the plasma membrane (13). These molecular properties have led to the proposal that MARCKS proteins might modulate the function of CaM by decreasing the cytoplasmic concentration of free CaM. MARCKS proteins might also regulate the plasticity of the cytoskeleton by cross-linking actin filaments at the membrane (1). Such models imply dynamic, highly regulated interactions of MARCKS proteins with membranes and other proteins. In order to test these hypotheses a detailed understanding of the thermodynamics and kinetics of these interactions is required.

The MARCKS family comprises two members: MARCKS is an acidic ubiquitous rod-shaped 32-kDa protein; MARCKS-related protein (MRP, also MacMARCKS or F52) is a 20-kDa protein that has 50% homology to MARCKS and is predominantly expressed in brain and reproductive tissues (14, 15). Whereas MARCKS has been relatively well characterized, only a few reports describe the molecular properties of MRP in vitro (8, 11, 16, 17). The recent findings that MRP is essential for brain development (15) and phagocytosis (18) and that MRP has a different subcellular localization than MARCKS (19) further emphasize the need for characterizing MRP.

Recently, the concept of myristoyl switch has emerged. Some myristoylated proteins may sense a signal, such as calcium for recoverin (20) or GTP for the ADP-ribosylation factor (21), which induces a conformational change that exposes the myristoyl moiety and allows binding of these proteins to membranes. McLaughlin and Aderem (22) have proposed that phosphorylation/dephosphorylation events might be the signals that trigger the myristoyl switch in MARCKS proteins. Alternatively, Ca²⁺/CaM might also be a candidate to trigger such a switch in MARCKS proteins.

In order to gain insight into the aforementioned aspects, namely regulation of the interactions of MARCKS proteins with other proteins, characterization of MRP, and characterization of a putative myristoyl switch, we have investigated the structure of MRP as well as its interactions with CaM, with the catalytic subunit of PKC (PKM), and with NMT. How myristoylation affects the behavior of MRP in solution was also explored by comparing the properties of unmyristoylated (unmyr) and myristoylated (myr) MRP.

EXPERIMENTAL PROCEDURES

Recombinant mouse unmyr and myr MRPs were expressed in Escherichia coli and purified, using a mild procedure that does not use heating and acid treatment (8). Rat brain PKM was purchased from Calbiochem. Calf brain PKM was prepared by limited trypsinolysis of purified PKC (23, 24). Bovine brain CaM, 5-dimethylaminonaphthalene-1-sulfonyl-calmodulin (CaM_D), and calf thymus histone (Type III-SS) were purchased from Sigma (Fluka Chemie, Buchs, Switzerland). Purified MRP was stored at 80 °C in 10 mM MOPS, pH 7.4, 0.1 mM EGTA. Unless stated otherwise, all measurements were performed in buffer A (10 mM MOPS, pH 7.4, 100 mM NaCl, 0.001% Triton X-100) containing 0.1 mM EGTA. Triton X-100 was added to minimize the adsorption of MRP on the surfaces of the tubes and cuvettes. This concentration of detergent improved the reproducibility of the experiments without affecting the data.

Sedimentation velocity and sedimentation equilibrium runs of unmyr MRP were performed in an XLA Optima analytical ultracentrifuge (Beckman Instruments) equipped with UV absorption optics and a photoelectric scanning system. The absorbance (A) was measured at 230 nm. Sedimentation equilibrium runs were performed at 20 °C at 22,000 rpm in a 12-mm DS charcoal cell. The molecular weight was calculated using a linear regression program obtaining the best linear fit of ln A versus r², where r is the distance from the rotor center (25).

Sedimentation velocity runs were performed at 20 °C and 46,000 rpm in a Beckman An60 Ti rotor 316 using a 12-mm DS Epon cell. The relatively low amounts of MRP available, as well as its tendency to oligomerize, precluded experiments in a wide range of protein concentrations. The sedimentation coefficient (s_20,w) does not depend on the concentration of MRP in the low range of concentrations studied (0.2-0.8 mg/ml MRP). We have therefore used an average of seven values, determined in this concentration range, as an approximation for the sedimentation coefficient extrapolated at infinite dilution of protein (_20,w⁰).

The diffusion constant (D) was calculated from the equation,

(Eq. 1)

where R is the gas constant, T is the absolute temperature, M is the molecular weight of MRP calculated from the cDNA sequence of MRP (20,165 g/mol),  is the partial specific volume of MRP calculated from the amino acid composition (0.705 cm³/g), and  is the solvent density (1.003 g/cm³) (26). The parameters describing the geometry of MRP were calculated according to Laue et al. (26) and Byers and Kay (27). Briefly, the experimental Stokes radius (R_S) was calculated from the following,

(Eq. 2)

where N is Avogadro's number and _o is the solvent viscosity (1.02 g·m¹·s¹). The translational frictional ratio, f/f₀, was calculated from the following,

(Eq. 3)

where R₀ is equal to 1.8 × 10⁷ cm and is the Stokes radius of the equivalent unhydrated sphere of molecular weight M and partial specific volume . The axial ratio for a prolate ellipsoid was estimated from the frictional ratio using the power series,

(Eq. 4)

where a is the long axis length (nm), b is the short axis length (nm), and X is equal to (f/f₀  1). This calculation does not take into account hydration effects and gives a maximal axis ratio, (a/b)_p,max. A minimal value for the axis ratio, (a/b)_p,min, can be estimated assuming maximal hydration of the protein and substituting X in Equation 4 as follows,

(Eq. 5)

where ₁ is the degree of hydration estimated from the amino acid composition of MRP (0.539 g of H₂O/g of protein) (28). Finally, the absolute values for a and b were evaluated from the following.

(Eq. 6)

Circular dichroic (CD) spectra were recorded in a Jasco 720 spectropolarimeter equipped with a thermostated quartz cell (1-mm path length) or a double chamber quartz cell (1-cm path length) (Hellma, Mühlheim, Germany). Each spectrum, including the base line, was recorded several times and averaged. The maximal -helical content of the spectra (f) was calculated from the mean residue molar ellipticity,  (deg·cm²·dmol¹·residue¹), at 222 nm according to Ref. 29.

(Eq. 7)

Binding of MRP to CaM_D was performed at 22 °C in a Jasco FP-777 fluorometer (Japan Spectroscopic Co., Ltd., Tokyo, Japan) as described previously (16, 30). The measurements were performed in quartz cuvettes with a 1-cm path length (Hellma). CaM_D fluorescence was excited at 340 nm, and the emission was measured at 480 nm. The signal was corrected for the contribution of buffer. Unless stated otherwise, the measurements were performed in buffer A containing 0.1 mM CaCl₂.

In analogy to the function used to describe the dissociation constant (K_d) for a 1:1 complex between MRP and CaM_D (31),

(Eq. 8)

the apparent dissociation constants (K_d,app) were determined by fitting the data with the empirical equation,

(Eq. 9)

where [CaM_D·MRP] is the concentration of the complex, [CaM_D] and [MRP] are the concentrations of the free proteins, and [CaM_D]₀ and [MRP]₀ are the total concentrations of the proteins. To describe the relationship between K_d,app and K_d, Equation 8 can be transformed into the equation,

(Eq. 10)

using [MRP] = [MRP]₀  [CaM_D·MRP]. A comparison of Equations 9 and 10 shows that the following holds true.

(Eq. 11)

Since the dissociation constant is defined as the concentration of free MRP at which 50% of CaM is saturated, we can replace [CaM_D·MRP] by 1/2 [CaM_D]₀, and Equation 11 becomes the following.

(Eq. 12)

K_d can thus be determined from K_d,app by performing titrations at successively lower CaM_D concentrations and extrapolating to zero [CaM_D]₀.

The kinetics of binding of unmyr MRP to CaM_D were measured at 22 °C in a Durrum stopped-flow fluorometer equipped with a mercury-xenon lamp. The instrument was modified to increase its sensitivity as described (32); its dead time was determined to be 4.5 ms. Excitation was at 334 nm. Emission was detected over 435 nm using a cut-off filter. The average from at least five kinetic traces was fitted by the sum of two exponential functions and analyzed as described for the binding of myosin light chain kinase (MLCK) (33). The experiments were conducted in buffer A containing 0.1 mM CaCl₂.

The initial rates of phosphorylation of MRP were measured at 30 °C in buffer A containing 0-6.3 µM MRP, 0.5 nM PKM, 0.1 mM [-³²P]ATP (0.2 µCi), 0.1 mM CaCl₂, and 6 mM MgCl₂ in a final volume of 20 µl. Under these conditions, the phosphorylation rates were linear during the first 30 min (not shown). The reactions were stopped after 20 min by adding 5 µl of SDS sample buffer (50 mM Tris, pH 6.8, 10% glycerol, 0.2% SDS, 0.02% bromphenol blue) and heating at 95 °C for 3 min. MRP was separated from excess [-³²P]ATP on 12.5% SDS-polyacrylamide gels. The radioactivity associated with MRP was quantified by overnight exposure of the dried gels to a storage phosphor screen and by scanning the screen in a PhosphorImager (Molecular Dynamics). The stoichiometry of phosphorylation was calculated by exposing the gels containing phosphorylated MRP simultaneously with a nitrocellulose membrane on which known amounts of [-³²P]ATP (0-30 pmol in 5 µl) were spotted. The conditions (amount of radioactivity and exposure time) were chosen such that the signal arising from phosphorylated MRP and from the calibration curve was in the linear range. The kinetic data were analyzed with the Hill equation (34),

(Eq. 13)

where v is the velocity of the phosphorylation, V_max is the maximal velocity, [MRP] is the concentration of MRP, n is the Hill coefficient, and S_0.5 is the concentration of MRP that yields half-maximal velocity. The turnover number (k_p) was calculated by dividing V_max by the concentration of PKM.

For phosphopeptide analysis, MRP was phosphorylated as described above except that the PKM concentration was raised to 2.5 nM and the incubation time to 2.5 h. The MRP concentration was 5 µM. Following electrophoretic transfer of phosphorylated MRP to nitrocellulose, MRP was digested with trypsin, and the corresponding phosphopeptides were separated by thin layer chromatography as described (11).

Polyhistidine-tagged human NMT expressed in E. coli was purified as described previously (35). The in vitro incubations were performed using 36 nM NMT, 0.6 µM unmyr MRP, 10 µM myristoyl-CoA (1.2-1.5 × 10⁶ dpm) and 0.5-2.5 µM CaM for 10 min at 30 °C in a final volume of 50 µl of buffer A. The incubations were terminated by the addition of 50 µl of 2 × SDS-sample buffer and heating to 100 °C for 2 min. Samples of each incubation (50 µl) were analyzed by SDS-polyacrylamide gel electrophoresis, and the positions of the bands were identified by fluorography using diphenyloxazole as described (36). Following fluorography, the extent of labeling was determined by volume scanning densitometry using a Computing Densitometer model 300A (Molecular Dynamics). For stoichiometry determinations, the fluorograph was used to localize the radiolabeled bands on the gel. These were excised and rehydrated for 30 min at room temperature, and the diphenyloxazole was removed by two 30-min treatments with dimethylsulfoxide (5 ml). The cleared gel was macerated, transferred to a screw-topped scintillation vial, and incubated with 0.5 ml of Soluene-100 (Packard Instrument Co.) for 24 h at 37 °C. The released radioactivity was determined by scintillation counting using 5 ml of Liquiscint (National Diagnostics).

The data obtained from CD spectroscopy and fluorometry, as well as from scanning of the radioactive phosphorylation and myristoylation experiments, were transferred to the SigmaPlot program (Jandel Scientific, San Rafael, CA) and analyzed using the appropriate functions.

RESULTS AND DISCUSSION

Following expression in E. coli, recombinant mouse MRP was purified to homogeneity (8). The protein has an amino acid composition that correlates well with the sequence deduced from the cDNA (37). The concentrations of myr and unmyr MRP were routinely determined by the Lowry assay using bovine serum albumin as a standard (38). The concentrations obtained with this assay were divided by a factor of 1.6, since amino acid analysis showed that the Lowry assay overestimates the concentration of MRP. Assays based on the method of Bradford (39) are inappropriate, since they overestimate the MRP concentrations by a significantly higher factor of about 10.

In the absence of x-ray and NMR data, our knowledge about the structure of MARCKS proteins is rather limited. To gain insight into the structure of unmyr MRP, we have investigated its hydrodynamic properties in an analytical ultracentrifuge. In the range of concentrations analyzed (0.2-0.8 mg/ml), the majority of the unmyr MRP molecules (>60%) has an estimated molecular mass of 21 kDa. Since the mass calculated from the amino acid composition of the protein is 20.2 kDa, most of the unmyr MRP molecules are monomeric. A significant fraction of protein (10-40%, depending on the experiments) has, however, molecular weights corresponding to dimers and/or higher aggregates. These populations could not be separated in the ultracentrifuge. Hence, the nature as well as the distribution of these oligomers remain to be determined. That unmyr MRP can oligomerize is important with respect to the actin filament cross-linking activity of MARCKS proteins (12). To cross-link actin filaments, MARCKS proteins must either have two binding sites for actin filaments or dimerize. Since only one binding site has been identified so far, namely the PS domain, our results suggest that dimerization of MRP might be physiologically relevant.

Because monomeric unmyr MRP could be separated from the oligomers, hydrodynamic and structural parameters could be determined for the monomeric population. The results of this analysis are shown in Table I. An unusually high frictional ratio can be calculated from the sedimentation coefficient of unmyr MRP, indicating that the protein has an elongated structure. Assuming the form of a prolate ellipsoid, we can calculate that the axis ratio (a/b)_p ranges between 6.8 and 12.1, depending on the hydration of the protein. Since unmyr MRP is elongated and contains a large proportion of polar residues, the protein is probably highly hydrated, and the axis ratio of 6.8 better describes its geometry. Using this ratio, we can estimate that the molecular dimensions of unmyr MRP monomers are 1.9 nm × 12.9 nm. For comparison, electron micrographs of the 30-kDa MARCKS reveal rods with dimensions of 3.7-5.1 nm × 35.6 nm (12).

Table I. Hydrodynamic and structural parameters for monomeric unmyr MRP Mr and s20,w were estimated from sedimentation equilibrium and sedimentation velocity runs, respectively. The other parameters were calculated according to equations 1-6. A similar Stokes radius was obtained by dynamic light scattering (E. Schleiff, unpublished results). Parameter Value Mr 21 ± 1 kDa s20,w 1.70 ± 0.04 S D 7.0 × 107 cm2/s RS 3.0 nm f/f0 1.67 (a/b)p,max 12.1 amax 19.0 nm bmax 1.6 nm (a/b)p,min 6.8 amin 12.9 nm bmin 1.9 nm

To obtain further information on the structure of unmyr and myr MRP we have measured its CD spectrum. Fig. 1 shows that the secondary structure of MRP is not dominated by classical secondary structure elements such as -helices or -sheets and is reminiscent of a random coil structure. A maximal -helical content of 15% can be calculated at 222 nm from Equation 7, a value similar to that obtained for MARCKS (40). The additional 100 amino acid residues present in MARCKS are therefore unlikely to add important elements of secondary structure to the protein. This conclusion is in agreement with the observation that these additional residues are scattered through the entire MARCKS primary sequence.

Binding to other proteins and to membranes, as well as post-translational modifications such as phosphorylation might stabilize or induce the formation of elements of secondary structure in MRP. The question of whether MRP has a potential for such conformational changes was addressed by measuring CD spectra of the protein in the presence of the -helical inducers trifluoroethanol (TFE) and hexafluoroisopropyl alcohol (HFIP) (41, 42). Fig. 2A shows that TFE has a significant effect on the CD spectra of myr MRP. A difference spectrum between solutions containing 69% and 0% TFE clearly demonstrates that TFE induces the formation of -helices (minima at 208 and 222 nm) (Fig. 2B). To further characterize this effect, MRP was titrated with increasing concentrations of TFE and HFIP (Fig. 2C). The ratio _222 nm/_208 nm is an index for the -helicity of proteins and is 1.1 for an ideal -helix (43). HFIP increases the -helicity of MRP at significantly lower concentrations than TFE. Interestingly, the curve reaches a plateau at about 30% HFIP. This observation suggests that an equilibrium exists between the random coil conformation of MRP in aqueous solution and an -helical conformation, which is stabilized by HFIP. Is such an -helical conformation physiologically relevant? HFIP and TFE have been proposed to stabilize secondary structures in peptides and proteins by modifying the physicochemical properties of their surrounding (41). Although this concept has recently been questioned (44), these solvents might mimic situations in which the surface properties of proteins are significantly modified. With this hypothesis in mind, we would predict that binding of MRP to other proteins or to membranes might induce a drastic increase in the -helical content of the protein. This hypothesis is attractive, since we have recently found kinetic evidence that MRP binds at the surface of planar membrane bilayers with different conformations.² To further investigate this phenomenon, the binding of MRP to lipid vesicles should therefore be measured by CD spectroscopy. However, since MRP has a significantly lower affinity for lipid vesicles than the peptide corresponding to the PS domain of MARCKS (8, 45), such an analysis might be complicated by the scattering of the vesicular suspensions.

TFE and HFIP-induced -helices in MRP.

The solubility of MARCKS proteins at high temperature and low pH has been widely used as a tool for their purification (15, 46). Whether these treatments alter the secondary structure of these proteins has not yet been determined. To investigate this aspect we have heated MRP solutions to 95 °C or decreased their pH to 4.0. These treatments have small but reversible effects on the CD spectra, suggesting that MRP not only remains soluble but also retains its structure when exposed to these harsh conditions (not shown).

Binding of MRP to CaM_D induces a shift (500-490 nm) as well as an increase (1.8-2.5-fold) in the fluorescence emission maximum of the dansyl group covalently bound to the N terminus of CaM (8, 11, 16). These properties were used to determine the dissociation constant of the complex by titrating CaM_D with MRP, as shown in Fig. 3. Typically, half-maximum saturation of the fluorescence signal is reached at a ratio [MRP]/[CaM_D] slightly above 0.5, indicating that MRP binds to CaM in a 1:1 ratio. The data were thus fitted with an hyperbolic function, assuming a noncooperative interaction between MRP and CaM_D and a 1:1 stoichiometry (see ``Experimental Procedures''). In Fig. 4A, the apparent dissociation constants calculated from these curves are plotted as a function of [CaM<sub>D</sub>]. Fitting the data with a linear regression and extrapolating the apparent dissociation constants to [CaM<sub>D</sub>] equal to zero give the dissociation constants of the complexes with myr MRP (k<sub>d</sub> = 4 ± 1 nM) and unmyr MRP (K<sub>d</sub> = 7 ± 2 nM). The slopes of the linear regressions are close to the theoretical value of 0.5 described in Equation 12 (0.55 for unmyr MRP and 0.65 for myr MRP), indicating that, in the CaM<sub>D</sub> concentration range shown in Fig. 4A, MRP binds to CaM<sub>D</sub> as a 1:1 complex. The dissociation constant for unmyr MRP is close to the value published for the same protein purified by heat and acid treatment (K<sub>d</sub> = 9.5 nM) (11). Hence, heating or acidifying MRP solutions does not alter the affinity of the protein for CaM<sub>D</sub>. Interestingly, if [CaM<sub>D</sub>] is increased above 40 nM, the apparent dissociation constants deviate from the linear relationship observed at lower concentrations (Fig. 4B). This behavior might result from the tendency of the protein to oligomerize (see ``Hydrodynamic Properties of MRP''). The apparent dissociation constants seen at higher concentrations may therefore reflect the affinity of MRP oligomers for CaM_D. Alternatively, MRP oligomers may not bind to CaM_D; the concentrations of monomers, and consequently the apparent dissociation constants, would then be overestimated. In this respect, evidence was found for the binding of two molecules of MLCK on CaM (47). Since the intracellular concentrations of MARCKS proteins (12 µM) and CaM (60 µM) (2) are well above the concentrations presented in Fig. 4B, these oligomerization effects might be physiologically relevant and merit further analysis.

In order to understand their mechanism of recognition we have investigated the kinetics of binding of unmyr MRP to CaM_D using stopped-flow fluorescence spectroscopy. A typical kinetic trace is shown in Fig. 5A. Although the binding is apparently monophasic, a careful analysis of the kinetic trace indicates the presence of a second phase, whose amplitude is about 10% of the total fluorescence signal. Consequently, the data were fitted with the sum of two exponential functions. The validity of this analysis can better be appreciated in Fig. 5B, which shows that the residuals of the fitting do not significantly deviate from the zero line. Fig. 5C plots the observed association rate constants k_obs1 and k_obs2 as a function of [unmyr MRP]. While k_obs1 is a second order rate constant, which depends on [unmyr MRP], k_obs2 is a first order rate constant, which is independent of [unmyr MRP]. Based on these observations, we propose the following scheme for the binding of unmyr MRP to CaM, where (CaM·unmyr MRP)* is an initial transient complex and (CaM·unmyr MRP) is the final stable complex. This model is reminiscent of the biphasic model proposed for the binding of MLCK to CaM (33). In this model, the mathematical relation between the above rate constants and the observed rate constants is given by the equations,

(Eq. 14)

and

(Eq. 15)

where f_k is a function of k₁, k₊₂, and k₂. From the slope of the upper line shown in Fig. 5C, we can determine the second order rate constant k₊₁. This value (k₊₁ = 1.6 × 10⁸ M¹ s¹) is similar to the value estimated for MLCK (k₁ = 1.1 × 10⁸ M¹ s¹). k_obs2 is a constant equal to (11 ± 4) s¹, which is the sum of the first order rate constants k₊₂ and k₂. For comparison, an 8-fold slower rate constant was estimated for the isomerization of the complex between MLCK and CaM (k₂ + k₂ = 1.3 s¹) (33). In analogy to the model of Török and Trentham, k_obs1 reflects the formation of the initial complex between MRP and CaM_D, while k_obs2 describes reorganization of the MRP·CaM_D complex. Keeping in mind that MRP oligomerizes and that this oligomerization might affect the affinity of the protein for CaM, k_obs2 might also describe a structural rearrangement involving MRP oligomers at the CaM binding site.

One hallmark of most CaM-binding domains is their tendency to bind to CaM as amphipathic -helices. Whether these domains are already -helical in the unbound state or whether binding induces the formation of these -helices, as shown with many peptides, is still a matter of debate (48, 49, 50). Since unbound MRP has only a low -helical content in solution (see Fig. 1) and since the amino acid sequence of the effector domain of MRP (the PS domain) can potentially form a basic amphipathic -helix (1), binding of MRP to CaM might induce the formation of an -helix in the PS domain. We have therefore investigated the effect of complex formation on the secondary structure of CaM and MRP. In contrast to the organic solvents HFIP and TFE (see Fig. 2), CD spectroscopy shows that the binding of CaM to MRP has no observable effect on the overall secondary structure of the proteins (Fig. 6). Since studies with spin-labeled peptides indicate that the PS domain has an extended structure in solution (51), our results suggest that this PS domain is not in an -helical conformation in the MRP·CaM complex. Hence, MRP might belong to a family of CaM ligands that do not follow the -helical model. In this respect, proteins believed to contain a sequentially discontinuous CaM-binding domain, such as phosphorylase kinase, Ca²⁺-ATPase, and caldesmon, do not seem to form -helices when bound to CaM (49).

Phosphorylation is central to the regulation of the function of MARCKS proteins. This event is generally associated with a reorganization of the subcellular localization of MARCKS proteins (19). Phosphorylation of MARCKS proteins has also been proposed to regulate the structure of the actin cytoskeleton (12) as well as the concentration of free cytosolic CaM (2, 52).

Phosphorylation of myr MRP has not yet been investigated. In contrast to PKC, PKM does not require lipids for its activation (53). This property makes it a suitable enzyme to study the solution properties of MRP. Fig. 7 shows the kinetics of phosphorylation of unmyr (Fig. 7A) and myr (Fig. 7B) MRP by PKM. Since the data could not be fitted assuming a noncooperative Michaelis-Menten model, we have used the Hill equation in which cooperativity is expressed by the Hill coefficient (see Equation 13). The kinetic parameters resulting from this analysis are shown in Table II. myr MRP is phosphorylated with high affinity (S_0.5 = 3.5 µM) and positive cooperativity. This cooperative behavior might result from the phosphorylation of two serine residues in the PS domain (11). A thin layer chromatography of phosphopeptides generated by trypsinolysis shows two labeled peptides of equal intensities, confirming that two serines are indeed phosphorylated in MRP. Alternatively, our data with ultracentrifugation and with CaM_D binding (see ``Hydrodynamic Properties of MRP'' and ``Binding of MRP to CaM'') indicate that cooperativity might result from oligomerization of MRP. Note that the turnover number (k_p = 130 min¹) is relatively large compared with other substrates, such as histones H1 for which PKM has a turnover number of about 40 min¹ (according to the Calbiochem product data sheet). MRP can thus be considered as an efficient substrate for PKM.

Table II. Kinetic parameters for the PKM-dependent phosphorylation of MRP The MRP concentration that yields half-maximum velocity (S0.5) and the Hill coefficient (nH) were calculated according to Equation 13. The turnover number (kp) is equal to Vmax/[PKM]. unmyr MRP myr MRP S0.5 3.2  ± 0.4 µM 3.5  ± 0.2 µM nH 2.1  ± 0.3 2.5  ± 0.3 kp 118  ± 18 min1 130  ± 6 min1

A comparison of our data with the results previously published for unmyr MRP (S_0.5 = 238 nM) (11) shows that our S_0.5 value is significantly larger. Since the concentration of MRP was determined by the Lowry assay in that study, the value of 238 nM is overestimated by a factor of 1.6 (see ``Determination of the Concentration and Amino Acid Composition of MRP''). The affinity of our protein for PKM is therefore 16-fold lower. This discrepancy does not result from the fact that Verghese and colleagues have purified unmyr MRP by heat and acid treatment, since these treatments do not significantly change the kinetic parameters of our protein (not shown). We have also compared two preparations of PKM (see ``Experimental Procedures'') and found no significant differences in the phosphorylation of MRP (not shown). Having excluded these possibilities, the reason for this difference remains unknown.

MARCKS proteins have been proposed to serve as a reservoir for CaM (2). Since CaM does not bind to phosphorylated MARCKS proteins, it has also been suggested that PKC could regulate the level of free CaM by phosphorylating MARCKS, thus leading to the release of CaM (2). Several reports show that CaM inhibits the phosphorylation of MARCKS (54, 55) and MRP (15) in vitro. Experimental evidence that CaM inhibits MARCKS phosphorylation in vivo was also recently obtained in keratinocytes and glioma cells (56, 57). MARCKS proteins would consequently allow a cross-talk between PKC and Ca²⁺/CaM-dependent signal transduction pathways (2, 52). An understanding of the regulation of these interactions requires an analysis of the stoichiometry of the CaM-dependent inhibition of MRP phosphorylation. To investigate this aspect we have quantified the effect of CaM on the phosphorylation rates of MRP. Fig. 8 shows that the rate of phosphorylation of myr MRP can be efficiently inhibited by CaM with a half-maximum CaM/MRP ratio of 0.7 at 1 µM MRP. Since the subcellular concentration of CaM is in sufficient excess over MRP (2), our results indicate that CaM should also efficiently inhibit the phosphorylation of MRP in vivo.

Since PKC phosphorylates its substrates at the membrane surface (53) and since CaM most probably binds to cytosolic, but not to membrane-bound, MARCKS proteins (58),² CaM might act as a chaperone that can protect newly synthesized MRP from undesired phosphorylation by cytosolic kinases, therefore allowing the protein to reach its correct subcellular localization. In this respect, it is interesting to note that MARCKS proteins can serve as substrates for proline-directed protein kinases (59, 60).

N-terminal myristoylation of proteins is normally a co-translational event (61). Post-translational myristoylation of MARCKS proteins might, however, be relevant in vivo, since a pool of unmyristoylated MARCKS is present in brain (62, 63) and since a demyristoylation activity was detected in brain synaptosomes (64). In order to investigate whether MRP can be myristoylated post-translationally, we have reconstituted an in vitro myristoylation system consisting of purified unmyr MRP and NMT as well as of [³H]myristoyl-CoA. Fig. 9 shows that unmyr MRP is myristoylated in the presence of NMT. No labeling is present in the absence of NMT, demonstrating that labeling is due to a covalent modification catalyzed by NMT. Also, this modification is specific for the N terminus of MRP, since myr MRP cannot be labeled (not shown). Similar results were obtained with human MARCKS expressed in E. coli.³ These results suggest that, in vivo, nonmyristoylated pools of MARCKS proteins can potentially be myristoylated post-translationally. Since the activity present in brain requires hours to significantly demyristoylate MARCKS proteins (8, 64), post-translational myristoylation/demyristoylation cycles might regulate long term cellular events in which MARCKS proteins are involved.

In vitro myristoylation can also be used as a tool to investigate the structure of MRP. A quantification of the incorporation of the label shows that MRP is modified to near stoichiometry (0.5-0.7 nmol of myristate/nmol of MRP). Thus, the N terminus of MRP is accessible to NMT and exposed to the surface of the protein. A similar extent of labeling was obtained with MARCKS.³ Fig. 9 also demonstrates that binding of CaM to the PS domain of MRP does not change the ability of NMT to myristoylate MRP. A kinetic analysis shows that the rate of myristoylation is also not significantly altered by CaM (not shown). We can therefore exclude the possibility that the N terminus of MRP is at the interface of the MRP·CaM complex. Since PKC-dependent phosphorylation might change the conformation of MRP, we have also investigated the ability of the phosphorylated protein to serve as a substrate for NMT. As for the CaM·MRP complex, phosphorylation does not alter the myristoylation of MRP. Overall, these studies suggest that the N terminus is exposed to the surface of unmyr MRP and spatially separated from the PS domain. Whether this arrangement is still present in myr MRP or whether myristoylation induces a conformational change in MRP cannot be investigated with this technique. With respect to the myristoyl switch model (see Introduction), it would be interesting to determine whether myristoylation alters the relative positioning of the N terminus and the PS domain.

George and Blackshear (65) have observed that millimolar concentrations of Ca²⁺ and Mg²⁺ inhibit the binding of MARCKS, expressed in a cell-free system, to membranes isolated from fibroblasts. Since MARCKS proteins are acidic (pI < 4.5) (2), it would not be surprising that these proteins bind divalent cations. These cations could consequently modulate the interactions of MARCKS proteins with membranes or with other proteins. To investigate these aspects, we have measured the effects of Ca²⁺ and Mg²⁺ on the binding of MRP to CaM as well as on its phosphorylation by PKM.

Calmodulin contains four high affinity binding sites for Ca²⁺ (K_d = 10⁶ to 10⁷ M). Ca²⁺ induces a large conformational change that exposes hydrophobic residues on CaM and allows its binding to target molecules (49, 50). In agreement with these properties, half-maximum binding of a peptide corresponding to the PS domain of MARCKS to CaM occurs with 0.4 µM Ca²⁺ (66). Fig. 10A shows titration curves of 30 nM CaM_D with MRP in the presence of various concentrations of Ca²⁺. Whereas MRP has only a weak effect on the fluorescence of CaM_D in the absence of Ca²⁺, MRP binds to CaM_D with high affinity (k_d,app = 24 nM) in the presence of 0.1 mM CaCl₂. Interestingly, the affinity of the complex is decreased 6-fold as the concentration of Ca²⁺ is increased to 1 mM (K_d,app = 141 nM) and 17-fold as the Ca²⁺ concentration is raised to 10 mM (K_d,app = 400 nM). In an attempt to further characterize this effect, we have titrated the CaM_D·MRP complex with increasing concentrations of Ca²⁺ and Mg²⁺. Since the fluorescence of CaM_D already depends on Ca²⁺ (not shown), the data were corrected for this effect by dividing by the fluorescence of CaM_D in the absence of MRP. Fig. 10B shows that Ca²⁺ and Mg²⁺ decrease the fluorescence signal in a concentration-dependent manner. Both curves are identical, demonstrating that the inhibitory effect of Ca²⁺ on the complex formation between MRP and CaM is not specific to this cation. Fig. 10C shows that Mg²⁺ has the same effect as Ca²⁺ on the fluorescence of CaM_D. The curves measured in Fig. 10B must therefore reflect the properties of the CaM_D·MRP complex rather than of CaM_D alone.

To gain further insight into the effect of divalent cations on the properties of MRP we have investigated whether Ca²⁺ can also modulate the phosphorylation of MRP. Since PKM does not require Ca²⁺ for its activation (53), an effect of Ca²⁺ on the phosphorylation of MRP should reflect a property of MRP rather than PKM. Fig. 11 shows that the initial rate of phosphorylation of myr MRP is inhibited by Ca²⁺ with a half-maximum concentration of 2.7 mM, further supporting the hypothesis that MRP binds Ca²⁺. However, this conclusion is complicated by the observation that Ca²⁺ also inhibits the phosphorylation of histone in the millimolar range (67). Whether Ca²⁺ directly inhibits the activity of PKC or rather binds to lipids or to the substrate was not investigated. In a control experiment, we have also observed that PKM-dependent phosphorylation of histone is inhibited by Ca²⁺ in the same concentration range as for MRP (not shown). Finally, in contrast to Ca²⁺, the phosphorylation rate of MRP increases as the Mg²⁺ concentration is increased with an estimated half-maximum concentration of 0.75 mM. This behavior simply reflects the requirement of the catalytic site of PKM for Mg²⁺ (68).

In the absence of appropriate controls, we cannot determine whether the inhibitory effects seen in Figs. 10 and 11 result from the binding of divalent cations to MRP or to its binding partners CaM and PKM. The above studies therefore suggest, but do not demonstrate, that divalent cations bind to MARCKS proteins with millimolar affinities. What might be the biological significance of these observations? Although the putative affinity of Ca²⁺ for MRP is well above physiological concentrations, high Ca²⁺ intracellular concentrations (>0.1 mM) can be transiently reached close to Ca²⁺ channels (69). This property is used by proteins with low affinities for Ca²⁺ during synaptic vesicles exocytosis. For example, the binding of syntaxin to synaptotagmin-I requires more than 0.2 mM Ca²⁺ for half-maximum binding (70). The intracellular concentration of free Mg²⁺, on the other hand, is significantly higher than Ca²⁺ (0.5 mM) (71); binding of Mg²⁺ to MRP might therefore be physiologically relevant. To get definitive answers to these hypotheses, other techniques should be used that can directly detect and quantify the binding of such cations to MRP.

Since the hydrophobic myristoyl moiety of MRP is unlikely to be exposed to the aqueous solvent, myristoylation could alter the conformation of the protein in solution. To investigate these aspects, most of the experiments described in this report were performed with both unmyr and myr MRP. None of these properties, i.e. the secondary structure of MRP, its phosphorylation by PKM, its binding to CaM, and the inhibitory effect of divalent cations, were significantly changed by myristoylation.

Sufficient quantities of myr MRP could not be purified to perform the analytical ultracentrifugation studies. We do not know, therefore, whether myristoylation changes the molecular dimensions of monomeric MRP and/or its oligomerization state. Also, the kinetics of binding of myr MRP to CaM were not investigated for the same reason. Although myr and unmyr MRP have the same affinity for CaM, myristoylation might modulate the mechanism of recognition of MRP by CaM and consequently alter the kinetics of complex formation. These experiments await the development of a better expression/purification system for myr MRP.

Taken together, our results show that myristoylation does not modulate the properties of MRP in solution. Since the myristoyl moiety of MRP is embedded in the membrane (8), myristoylation is certainly involved in the interactions of MRP with membranes but might also influence the interactions of MRP with other proteins at the membrane surface. In this respect, we have recently obtained evidence that myristoylation has an effect on the conformation of membrane-bound MRP.²

In this study we report a thorough characterization of the structure of MRP as well as of its interactions with CaM, PKM, NMT, and divalent cations in solution. Whether MRP binds to actin filaments has not yet been reported. A complete characterization of the in vitro properties of MRP will also require an understanding of its interactions with membranes (8, 9).² Ultimately, protein-protein interactions involving MRP should be investigated at the membrane surface. Such studies will allow an analysis of the putative functions of MRP by answering the following questions: 1) can MRP sequestrate CaM at the membrane surface? 2) can MRP anchor actin filaments to the membrane? 3) what are the regulatory functions of PKC and Ca²⁺/CaM? 4) how is myristoylation involved in these interactions?