A Critical Intramolecular Interaction for Protein Kinase Cϵ Translocation

Disruption of intramolecular interactions, translocation from one intracellular compartment to another, and binding to isozyme-specific anchoring proteins termed RACKs, accompany protein kinase C (PKC) activation. We hypothesized that in inactive ϵPKC, the RACK-binding site is engaged in an intramolecular interaction with a sequence resembling its RACK, termed ϕϵRACK. An amino acid difference between the ϕϵRACK sequence in ϵPKC and its homologous sequence in ϵRACK constitutes a change from a polar non-charged amino acid (asparagine) in ϵRACK to a polar charged amino acid (aspartate) in ϵPKC. Here we show that mutating the aspartate to asparagine in ϵPKC increased intramolecular interaction as indicated by increased resistance to proteolysis, and slower hormone- or PMA-induced translocation in cells. Substituting aspartate for a non-polar amino acid (alanine) resulted in binding to ϵRACK without activators, in vitro, and increased translocation rate upon activation in cells. Mathematical modeling suggests that translocation is at least a two-step process. Together our data suggest that intramolecular interaction between the ϕϵRACK site and RACK-binding site within ϵPKC is critical and rate limiting in the process of PKC translocation.

The protein kinase C (PKC) 1 family of phospholipid (PL) -dependent serine/threonine kinases undergoes a conformational change and translocation, or movement, from the cytosolic to the cell particulate fraction upon activation (1,2). Conformational changes in PKC from an inactive to an active state results in exposure of domains required for PKC anchoring to the particulate fraction and in increased sensitivity of the enzyme to proteases (1)(2)(3)41). Therefore, the inactive state exists in a closed conformation, with the proteolytic sites protected, whereas the active state is in an open conformation with exposed proteolytic sites. Structural alterations from the closed to open states involve disruption of intramolecular interactions within the enzyme.
An intramolecular interaction in inactive PKC between the catalytic site and a site in the regulatory domain that resembles a substrate phosphorylation site but lacks a serine or threonine phosphoacceptor (pseudosubstrate site) has been previously identified (3,6). Deletion of the pseudosubstrate (-substrate) site generated a constitutively active enzyme (6) and mutations of the basic residues in the -substrate site reduced the affinity of the catalytic site to the -substrate site generating a constitutively active enzyme, preferentially localized to the cell particulate fraction (6). Furthermore, conversion of the alanine in the -substrate site to a glutamic acid, mimicking a phosphorylated amino acid, resulted in loss of binding of the -substrate site to the catalytic site, creating a constitutively active enzyme (6). Finally, a peptide corresponding to the -substrate site is a competitive inhibitor of PKC catalytic activity (6).
Since isozyme-specific PKC binding to respective RACKs occurs upon enzyme activation, we predicted that PKC activation involves a conformational change that induces the unmasking of the RACK-binding site in PKC (15)(16)(17). Therefore, we proposed the existence of a second intramolecular interaction in inactive PKC. This intramolecular interaction forms between the RACK-binding site and another site in PKC, termed the pseudo-RACK (RACK), which resembles and mimics a sequence in the RACK (15)(16)(17). It is likely that even in unstimulated cells, PKC exists in at least two conformations: an active open conformation, that has the RACK-and PLbinding sites exposed, and an inactive closed conformation, with these sites unavailable for binding to RACK and PL. In the presence of activators, the open conformation would be stabilized, shifting the equilibrium between the two conformations toward the active state.
RACK sequences were identified by searching for regions of homology between each PKC isozyme and its RACK. In ⑀PKC, the sequence HDAPIGYD (⑀PKC 85-92), named ⑀RACK, has ϳ75% homology with a sequence in ⑀RACK consisting of amino acids NNVALGYD (⑀RACK 285-292). A peptide corresponding to the ⑀RACK sequence functions as an ⑀PKC-selective agonist. We predicted that the agonist activity occurs because the ⑀RACK peptide binds the ⑀RACK-binding site in ⑀PKC, interfering with the intramolecular interaction between the ⑀RACK and the ⑀RACK-binding site, and thus stabilizing the open form. Since, after treating cells with a ⑀RACK peptide, ⑀PKC was co-localized with ⑀RACK (17), we suggested that the RACK peptide is eventually replaced by the RACK, possibly because the affinity of the RACK-binding site to the RACK peptide is lower than the affinity of the RACK-binding site to the RACK (16). A notable difference between the ⑀RACK sequence in ⑀PKC and the homologous sequence in ⑀RACK involves an amino acid change from a polar charged amino acid (aspartate (Asp-86)) in ⑀PKC to a polar non-charged amino acid (asparagine (Asn-286)) in ⑀RACK. Such a change in charge fits the prediction that the proposed intramolecular interaction mediated by the ⑀RACK site and the RACK-binding site in ⑀PKC lacks a crucial interaction and is subsequently competitively replaced by the ⑀RACK upon enzyme activation (18).
The aim of this study was to test the importance of the intramolecular interaction between the ⑀RACK site and the ⑀RACK-binding site and its influence in the process of ⑀PKC translocation and function. If the ⑀RACK site is engaged in an intramolecular interaction with the RACK-binding site, we expected that a mutation in the ⑀RACK site, which increases the resemblance of the ⑀RACK site to the ⑀RACK, should stabilize this intramolecular interaction and yield a more closed conformation. To test this hypothesis, we mutated Asp-86 in the ⑀RACK site to an Asn-86, so that the ⑀RACK site would more closely resemble the ⑀RACK or to a non-polar amino acid, alanine (Ala). We then determined the sensitivity of the mutants and wild-type enzyme to protease digestion and their dependence on activators for binding to ⑀RACK in vitro. These mutants were also expressed in mammalian cells as GFP fusion proteins and the effect of the mutation on translocation rates was examined.

EXPERIMENTAL PROCEDURES
Materials-Restriction enzymes were from New England Biolabs. Anti-⑀PKC V5 antibodies were from Santa Cruz Biotechnology.
Constructs-The EGFP-⑀PKC construct used in these studies was kindly provided by Bio Image A/S, Soeborg, Denmark. This construct contained two amino acid substitutions (Q34A and S336G), when compared with the human ⑀PKC sequence deposited in GenBank TM , accession number X65293. Site-directed mutagenesis of this clone was performed using the QuikChange site-directed mutagenesis kit (Stratagene) according to the manufacturer's instructions. Primers used for site-directed mutagenesis were: for D86N forward primer (5Ј-GTAGCCTATGGGGGCATTGTGAAAGACAGCCAG-3Ј) and reverse primer (5Ј-CTGGCTGTCTTTCACAATGCCCCCATAGGC TAC-3Ј). For D86A forward primer (5Ј-CTGGCTGTCTTTCACGCTGCCCCCATAG-GCTAC) and reverse primer (5Ј-GTAGCCTATGGGGGCAGCGTGAAA-GACAGCCAG-3Ј). The ⑀PKC mutants were fully sequenced and subcloned by cutting and re-ligating into the XhoI and BamHI (New England Biolabs) sites into pEYFP C1 and pECFP C1 (Clontech). For expression in the baculovirus expression system GFP-⑀PKC was cut with XhoI and BamHI to release the insert, filled in with Klenow (New England Biolabs) and cloned into pvL1392 (BD Biosciences) digested with SmaI (New England Biolabs). The human ⑀RACK construct cloned into PZeoSV used in these studies was kindly provided by Bio Image A/S, Soeborg, Denmark. The ⑀RACK insert was released by cutting with ScaI (New England Biolabs) and XhoI, blunt-ended with Klenow, and re-ligated into the SmaI site of PacG3X (BD Biosciences). ⑀RACK was expressed as a GST fusion protein.
Transient Transfections-Transfections of CHO-Hir, MCF-7, and Sf9 cells was carried out using FuGENE 6 (Roche Applied Science) according to the manufacturer's instruction.
Arg C Proteolysis Assay-ArgC digests were performed as described (19). Crude insect cell lysate containing ϳ70 ng of PKC (determined by Western blot comparison with standardized PKC (Pan Vera)) was digested in 520 l of 20 mM Tris, pH 7.4 with 20 l of endoproteinase ArgC (Roche Applied Science, 25 units/ml) at room temperature. Aliquots were removed at the indicated time points run on 8% SDS-PAGE gels, followed by Western blot using anti-⑀PKC antibodies.
⑀RACK Binding Assay-Insect cells expressing either ⑀PKC or GST-⑀RACK were lysed as described above. GST-⑀RACK, 10 ng was immobilized on glutathione-Sepharose 4B beads (Amersham Biosciences), and washed thoroughly with wash buffer (20 mM Tris-HCl pH 7.5, 2 mM EDTA, 100 mM NaCl, 12 mM ␤-mercaptoethanol, and 0.1% Triton X-100). Immobilized GST-⑀RACK was then incubated with the soluble fraction of insect cell lysates containing 100 ng of wild type, D86A, or D86N ⑀PKC protein, in the presence or absence of phospholipid activators (PL) (phosphatidylserine 12 g/reaction and, sn-1,2 dioleoylglycerol 400 ng/reaction) for 1 h at 4°C. Following a thorough washing with wash buffer, bound proteins were eluted in SDS-PAGE sample buffer and proteins separated on 8% SDS-polyacrylamide gels. The amount of ⑀PKC interacting with ⑀RACK was determined by Western blot probing for ⑀PKC with anti-⑀PKC-V5 antibodies. Blots were then re-probed with anti-GST (Santa Cruz Biotechnology) to verify that the same amount of GST-⑀RACK was present in each binding assay.
Immunoprecipitation-CHO-Hir cells were washed two times in serum-free medium and serum-starved 12-24 h after transfection. Cells were then kept in serum-free medium for ϳ12 h before the experiment. Cells were lysed in homogenization buffer as described above. The supernatant was then used for immunoprecipitation experiments. Immunoprecipitation was performed with anti-GFP monoclonal antibodies 3E6 (Molecular Probes) according to the manufacturer's instruction. Briefly, lysates (ϳ2 mg/ml) were precleared in protein G-agarose beads (25-l packed volume, Invitrogen) for at least 30 min and spun at 1000 ϫ g. The supernatant was then incubated with 1 g of anti-GFP monoclonal antibodies for at least 2 h. Protein G-agarose beads (25-l packed volume) was then added to the lysate-antibodies complex and incubated for at least 2 h (all incubations were done at 4°C). Beads were subsequently washed three times with Buffer 1 (50 mM Tris, 0.15 M NaCl, 1 mM EDTA, 0.1% Triton X-100, pH 7.5), once with Buffer 2 (50 mM Tris, 0.15 M NaCl, 0.1% Triton X-100, pH 7.5) and once with Buffer 3 (50 mM Tris, 0.1% Triton X-100, pH 7.5). For kinase assays, beads were washed an additional time with 20 mM Tris-HCl, pH 7.5.
Kinase Assay-The ability of the different ⑀PKC mutants to phosphorylate substrates was assayed by following the incorporation of [␥-32 P]ATP into myelin basic protein according to a method modified from Kikkawa et al. (20) myelin basic protein phosphorylation was measured either by liquid scintillation or for immunoprecipitation experiments, by autoradiography. For the kinase reaction of immunopre-cipitated enzyme, beads were resuspended in 20 l of a kinase reaction buffer composed of 20 mM Tris, pH 7.5, [␥-32 P]ATP (Amersham Biosciences) 0.3 Ci/reaction, ATP (Sigma) 9 M/reaction, myelin basic protein (Sigma) 12 g/reaction, and MgCl 2 50 mM/reaction. 4 l of phospholipids were added per reaction when needed. Phospholipids were prepared as described (20). Kinase reactions were stopped with SDS sample buffer and boiling. Samples were then run on a 12% SDS-PAGE and transferred to nitrocellulose exposed for autoradiography. The same nitrocellulose was then developed with anti ⑀PKC (V5) antibodies (Santa Cruz Biotechnology) to verify the amount of fusion protein immunoprecipitated.
Analysis of PKC Translocation by Western Blot-After 24 h of transfection cells were serum-starved for an additional 24 h and incubated with phorbol 12-myristate 13-acetate (PMA) (LC Laboratories) for the indicated times and concentrations at room temperature and subsequently fractionated as previously described (21). To assess PKC distribution the different cell fractions were run on SDS-PAGE, transferred for Western blot analysis and probed with anti-⑀PKC V5 antibodies. Lysates of overexpressed GFP-⑀PKC were diluted to ϳ20 ng/well. At this concentration endogenous ⑀PKC was not detected.
Microscopy and Analysis-CHO cells were grown on glass cover slips, and serum-starved as described above. For each experiment, cells were transferred to a commercially available metal coverslip holder (Molecular Probes) in which the coverslip formed the bottom of a 1 ml bath. The media was then replaced with extracellular buffer (120 mM NaCl, 5 mM KCl, 1.5 mM CaCl 2 , 1.5 mM MgCl 2 , 20 mM HEPES, and 30 mM glucose). Cells were stimulated by either PMA (100 nM), or with ATP (1 mM) in extracellular buffer. Fluorescence images of GFP-tagged constructs were obtained using the 488 nm excitation line of a laser scanning confocal microscope (Pascal, Zeiss), and emission was collected through a 505-550 nm band pass filter. Cells were imaged on the stage of an inverted microscope (Axiovert 100 M) using a 40ϫ Zeiss Plan-apo oil immersion objective with 1.2 numerical aperture (NA). For dual color imaging of CFP and YFP fusion proteins, a spinning disk Nipkow confocal microscope was used. Cells were viewed using an inverted Olympus IX70 microscope with a 40ϫ oil immersion Olympus objective (1.35 NA) and images were acquired with a CCD camera (Hamamatsu) and 2X2 pixel binning. CFP was excited with the 442 nm laser line of a helium-cadmium laser (Kimmon) whereas YFP was imaged with the 514 nm line of an argon ion laser (Melles-Griot). To selectively photobleach YFP labeled proteins in local regions of the cytoplasm for FRAP experiments, the 514 nm line of an Enterprise laser (Coherent) at maximal power (ϳ400 milliwatts) was used. For these experiments, we used a 60ϫ oil immersion objective (1.4 NA). Under these conditions, and by placing an iris in the beam path, it was possible to bleach 80% of the YFP fluorescence in 500 ms in an area of the cytosol measuring ϳ35 um 2 . Real-time confocal images were acquired every 10 -15 s for 20 min for experiments utilizing PMA, and every 5 s for a total duration of 3 min in the case of cells stimulated with ATP. Reagents were added to the cell chamber after the fifth image in each time series. Control time lapses were acquired using the same imaging conditions used in the experiments to check that the level of general dye photobleaching did not exceed 10 -30%. All images were acquired at room temperature. Images were exported as 12 or 16 bit files and changes in fluorescence intensity were measured using Metamorph® data analysis software (Universal Imaging). To monitor the translocation of PKC, a small region of interest was selected in the cytosol of each cell and fluorescence intensity values graphed against time after subtraction of background values and normalized so that the initial fluorescence was 100%. Averages of 10 -15 cells from three independent experiments were used.
Statistical Analysis and Mathematical Modeling-For quantitative analysis auto radiographs were scanned and quantified using NIH Image software. Analysis of fluorescence data was performed using Metamorph® (Universal Imaging Corporation). To determine statistical significance we used a one-tail type 2 Student's t test (Microsoft Excel). Significance of time course curves was determined using twoway ANOVA test (Stat View®). Mathematical modeling, was obtained from data of cells expressing similar levels of the different ⑀PKC ⑀RACK mutants. Non-linear least squares curve fitting analysis was performed using the EViews software (QMS) and differential equation analysis was performed using Berkeley Madonna™.

RESULTS
Sensitivity to Protease Degradation of the Different ⑀RACK ⑀PKC Mutants-If the Asp-86 in the ⑀RACK site is engaged in an intramolecular interaction with the RACK-binding site, we expected the D86N mutant to reside more in the closed state, and therefore be more resistant to proteolysis. In contrast, the D86A mutant should favor an open conformation and therefore be more susceptible to proteolysis. To test this hypothesis, ⑀PKC Wt and mutants expressed in insect cells were subjected to proteolysis by the endopeptidase, Arg C as previously shown for ␤PKC (19). Degradation by Arg C was monitored by the decrease of fulllength ⑀PKC. Although under these experimental conditions D86A mutation did not alter susceptibility of the enzyme to degradation by Arg C when compared with the wild-type enzyme, the D86N mutant was significantly more resistant to proteolysis than either D86A or wild type (Fig. 1, A and B).
Binding of the ⑀PKC Mutants to ⑀RACK-RACKs bind active PKC (22,23). If the intramolecular interaction between ⑀RACK and the RACK-binding site stabilizes the inactive closed form, increasing or decreasing the affinity of this intramolecular interaction should cause a corresponding decrease or increase in the ability of the enzyme to bind to its RACK. To test this prediction, we determined the binding of insect cell-expressed ⑀PKC (Wt and ⑀RACK mutants) to immobilized GST-⑀RACK in the presence and absence of phospholipid (PL) activators.
Binding of D86A ⑀PKC mutant to ⑀RACK in the absence of PL activators was at least 2-fold greater than binding of either D86N or wild-type enzymes (Fig. 1, C and D). Binding of D86N and of wild-type enzyme to ⑀RACK was significantly increased in the presence of PL, whereas binding of the D86A mutant to ⑀RACK was not increased (Fig. 1, C and D). In the presence of an equal concentration of PL, there was less ⑀RACK binding of D86N ⑀PKC than either D86A or wild-type ⑀PKCs. These results are consistent with the prediction that the ⑀RACK-binding site in the D86A mutant is already available for binding to ⑀RACK, whereas this site is masked in both wild-type or D86N ⑀PKCs and becomes accessible for binding only upon activation.
Rate of Translocation of ⑀PKC and ⑀PKC Mutants in Cells-The above in vitro studies of ⑀RACK mutants and wild-type ⑀PKC support our hypothesis that ⑀RACK is involved in an intramolecular interaction that stabilizes the inactive closed state. To further test this hypothesis, we examined the rate of translocation of ⑀RACK mutants and wild-type ⑀PKC in cells in response to stimulation. We proposed that part of the process of translocation requires disruption of the intramolecular interaction between the ⑀RACK site and the ⑀RACK-binding site. Therefore, we expected that modulation of this intramolecular interaction should affect translocation rates of ⑀PKC. To investigate this hypothesis, ⑀PKC ⑀RACK mutants were fused at their N termini to GFP, CFP, or YFP proteins and translocation was analyzed both by cell fractionation and by real-time imaging. In an in vitro kinase assay, we first confirmed that the GFP fusion proteins had similar catalytic activity. As seen in Fig. 2A, immunoprecipitated GFP-⑀PKC mutants phosphorylated myelin basic protein to a similar extent upon activation. We next confirmed that fusion of GFP to the ⑀PKC mutants does not interfere with the binding of the enzymes to ⑀RACK in vitro. Fig. 2B demonstrated a selective binding of wild type and the two GFP-⑀PKC mutants to GST-⑀RACK.
We also confirmed that the wild-type and ⑀RACK ⑀PKC mutants were sensitive to activation by phorbol ester and that they translocated from the soluble to the particulate fraction of the cell upon activation. Translocation of the ⑀PKC mutants was determined using two different cell lines, MCF-7 and CHO. CHO cells were stimulated for 10 min with 100 nM PMA (Fig.  2C). MCF-7 cells were stimulated with 10 nM PMA for 10 min (Fig. 2D), since stimulation of MCF-7 cells with higher doses of PMA caused a detachment of the cells. After treatment with PMA, cells were fractionated into soluble and particulate fractions, and GFP-⑀PKC was detected and quantified by Western blot analysis using anti-⑀PKC. Only GFP-⑀PKC fusion proteins (ϳ110 kDa) and not the endogenous ⑀PKC (ϳ80 kDa) were detected when 20 ng of protein was loaded per lane, and all the mutants had a similar level of expression (Fig. 2, C and D). On PMA treatment, there was a greater increase in D86A ⑀PKC in the particulate fraction than either D86N or wild-type ⑀PKCs (Fig. 2, C and D). This is not due to PKC degradation; the total amount of ⑀PKC was not changed upon activation. Therefore, GFP-⑀PKC ⑀RACK mutants were catalytically active and responded to activation by PMA by translocating from the soluble to the particulate fraction of the cell. Using two different cell lines, we found that after 10 min of treatment with PMA, more D86A ⑀PKC compared with wild-type or D86N ⑀PKCs translocated to the particulate fraction.
Since cell fractionation experiments are not suitable for dynamic studies, we used the GFP tag to follow the rate of translocation of the different ⑀PKCs by real-time microscopy in CHO cells (Fig. 3). Translocation of ⑀PKC was seen as a decrease of fluorescence intensity in the cytoplasm concomitant with an increase in fluorescence intensity at the cell periphery. All of the ⑀PKC mutants translocated to the same place (cell periphery) when stimulated by PMA (Fig. 3A). Translocation of the D86A mutant to the cell periphery began within 1 min of PMA stimulation, whereas translocation of the wild-type enzyme became apparent only after 2-3 min of PMA stimulation (Fig. 3A, arrows). In contrast, translocation of the D86N mutant became apparent only after 5 min of stimulation. A typical line intensity profile showing the distribution of each ⑀PKC between the cell periphery and cytosol is shown for a representative cell at different time points (Fig. 3B).
The amount of fluorescence decrease in the cytoplasm was proportional to the amount of fluorescence increase at the cell periphery (24), since the total fluorescence did not change. To quantitatively monitor translocation of the different ⑀PKC enzymes, we measured the decrease of fluorescence intensity in a region of the cytoplasm upon PMA stimulation using Metamorph software (Universal Imaging). A graphical comparison of ⑀PKC translocation obtained by the decrease of fluorescence in the cytoplasm indicated that the rates of translocation were significantly different between the two mutants and wild-type ⑀PKC (p Ͻ 0.0001; Fig. 3C). The D86A mutant translocated at Average translocation rates were expressed by normalizing the initial fluorescence intensity to 100% (average of at least three independent experiments, with at least three cells analyzed for each experiment). The time courses for the mutants and wild-type enzymes were statistically different from each other by two-way ANOVA with p Ͻ 0.001. a faster rate than either the wild-type or D86N ⑀PKC mutant. The wild-type enzyme reaches a similar steady state level (the level at which there was no further accumulation of ⑀PKC at the cell periphery) as the D86A mutant, but did so at a slower rate. In contrast, the D86N mutant achieved a steady state at a higher level than either the wild type or D86A mutant. Therefore, the D86A mutant translocated at a faster rate than either the D86N or wild-type ⑀PKCs, and the amount of D86N ⑀PKC mutant that reached the cell periphery was lower than the amounts of either D86A or wild-type ⑀PKCs.

Mathematical Modeling of ⑀PKC Translocation Suggests That ⑀PKC Translocation Is at Least a Two
Step Process-To further characterize the initial process of translocation of the different ⑀PKC ⑀RACK mutants, we fitted the decrease of fluorescence in the cytoplasm to a mathematical model. Fluorescence time courses were analyzed by a non-linear regression analysis with single and bi-exponential equations (1 and 2) as previously done by Nalefski and Newton (25), where I(t) is the concentration of observed molecules in the cytosol at time (t), and C 1-5 are constants in Equations 1 and 2.
Residual error graphs obtained using single exponential equations (Fig. 4A) and using bi-exponential equations (Fig. 4B) for the D86A ⑀PKC mutant translocation process is shown in Fig. 4. Residual error analysis for D86N and wild-type ⑀PKCs showed similar results (data not shown). Since a superior fit, with a smaller and more equally distributed error was obtained with a bi-exponential equation, we adopted a two-step model to illustrate the initial process of ⑀PKC translocation where the first step is the opening of ⑀PKC and disruption of the intramolecular interaction between the ⑀RACK and the ⑀RACK-binding site, and the second step is ⑀PKC binding to the membrane (binding to the membrane in this case may be binding to either lipids or proteins). This process can be described as follows in Scheme 1, Closed ⑀PKC in the cytosol -| 0 Open ⑀PKC in the cytosol Open ⑀PKC in the cytosol -| 0 where k 1 is the rate of ⑀PKC opening in the cytosol, k Ϫ1 is the rate of closing in the cytosol, k 2 is the rate of ⑀PKC binding to the membrane, and k Ϫ2 is the rate of detachment from the membrane. Using the bi-exponential equation, we estimated the values for the constants C1, C3, and C5 (Table I). The value of C1 corresponds to the level of ⑀PKC in the cytosol at steady state. We next determined whether the hypothesis that the values obtained for C1 and C3 of D86A and wild-type ⑀PKCs are statistically equivalent. Using WALD's parameter test, and using F-Test and i -Squared test, we obtained p values of 0.9 and 0.7, respectively. We therefore assumed that the steady state level for D86A and Wt, C1, is the same. We also assumed that the second step (binding to the membrane) should not differ between the RACK ⑀PKC mutants, since C3 was equal for both wild-type and D86A ⑀PKC. Therefore, if the steady state level (C1) was the same for D86A and wild-type ⑀PKCs and the second step (binding to the membrane) was also the same, it is plausible that k Ϫ1 (rate of closing of ⑀PKC) for both D86A and D86N ⑀PKCs would be negligible. When k Ϫ1 is negligible, the following equations (3-5) hold and can be used to calculate k Ϫ2 (Equation 6) and K 2 (Table I).
FIG. 4. Mathematical modeling analysis of D86A is represented in the figure; similar results were obtained with D86N and wildtype ⑀PKC. A, non-linear regression analysis using the single or B, bi-exponential equation. C, fit between curves of the raw data for D86A. The curves were obtained by nonlinear regression with a bi-exponential equation. D, fit between curves of the raw data for D86A. The curves were obtained by a differential equation using the values for k 1 , k Ϫ1 , k 2 , and k Ϫ2 provided in Table II. The residual error for all curves fitted data was similar to the one obtained with a non-linear regression using a bi-exponential equation.  I 1 (t)) we can then use the following differential Equations (7-9) to describe our model. dI 0 (t)/dt ϭ Ϫ(K1)(I 0 (t)) ϩ (KϪ1)(I 1 (t)) (Eq. 7) dI 1 (t)/dt ϭ Ϫ(K Ϫ 1 ϩ K2)I 1 (t) ϩ (K1)(I 0 (t)) ϩ (K Ϫ 2)(I 2 (t)) (Eq. 8) We next solved the differential equations for I(t) using the Runge-Kutta algorithm (Berkeley-Madonna) and compared the solutions to the experimental data (loss of fluorescence in the cytoplasm). We solved for I(t) and used the calculated param-eters for k 1 , k Ϫ1 , k 2 , and k Ϫ2 (Table I) assuming that the initial amount of closed ⑀PKC in the cytoplasm is equal to 100% [I o (0) ϭ 100]. Fig. 4C, shows the fit between curves of the raw data for D86A ⑀PKC, and the curve obtained by nonlinear regression with a bi-exponential equation. Fig. 4D shows the fit between curves of the raw data for D86A and the curve obtained by the differential equations solving for I(t). Similar fitting results were obtained for D86N and wild-type ⑀PKCs (data not shown). The residual error for all curve fitting data was similar to the one obtained with a non-linear regression, using a bi-exponential equation (Fig. 4B; data not shown).
The steady state level for D86N was higher than for either D86A or wild type (Table I), and therefore it is not possible to assume that k -1 for this mutant was negligible. In this case, we also used the Runge-Kutta algorithm to solve differential equations. Because the rate of the second step, binding to the membrane, should not be altered for either of the ⑀PKC mutants, we assumed that the values of k 2 and k -2 should be in the range of a For D86N the relation k 1 ϭ C 5 is not valid because the corresponding values for k 1 , k Ϫ1 , k 2 , and k Ϫ2 were obtained by simulations using differential equations.
FIG. 5. D86A translocates faster than Wt upon PMA activation and faster than D86N upon stimulation with ATP. A, in cells transfected with both wild-type and the D86A ⑀PKCs, the D86A ⑀PKC mutant translocated faster than wild-type ⑀PKC. Confocal image of a CHO cell transfected with both YFP-⑀PKC D86A and CFP-⑀PKC wild type (in pseudo-color) at different times upon stimulation with 100 nM PMA. B, quantitative analysis (CFP and YFP) of the decrease in fluorescence CFP-⑀PKC wild type (f) and YFP-⑀PKC D86A (‚) in a region in the cytoplasm of the cell. Levels of the YFP-⑀PKC D86A mutant in the cytoplasm decrease faster than levels of CFP-⑀PKC wild type. B, when compared with wild type, the D86A mutant had a similar translocation rate and the RACK D86N mutant had a slower translocation rate upon stimulation with 1 mM ATP. Confocal images of translocation of ⑀PKC mutants at different time points after the addition of 1 mM ATP. The arrows indicate the time at which translocation to the cell periphery began to be apparent for each ⑀PKC enzyme. C, translocation rates were analyzed by measuring the loss of fluorescence in the cytoplasm relative to time after addition of 1 mM ATP: Wt, f; D86A, ‚; D86N, E. Data are averages of at least three independent experiments with at least three cells in each experiment. The time course for the D86N mutant was statistically different from either D86A or wild-type ⑀PKCs using a two-way ANOVA test with p Ͻ 0.001. the ones obtained for D86A and wild type.
Together, our experimental data and mathematical modeling suggest a two-step translocation process of ⑀PKC to the cell membrane upon activation. Whereas the second step was independent of intramolecular interactions, the first step, which we predicted involves opening of the enzyme, was greatly dependent on these intramolecular interactions.
Co-transfection of D86A ⑀RACK Mutant and Wild-type ⑀PKC in the Same Cell-To further investigate the behavior of the ⑀PKC mutants within a single cell, we co-transfected two different ⑀PKC constructs into CHO cells, fused to either YFP or CFP. A representative cell that co-expresses a YFP-D86A and a CFP-Wt ⑀PKC at similar levels is shown in Fig. 5A. Upon stimulation with PMA, the D86A mutant translocated faster than the wild-type enzyme (Fig. 5A). Quantitative analysis, expressing the decrease of fluorescence intensity in the cytoplasm is in Fig. 5A (right panel). The D86A mutant was found at the cell periphery at 1.5 min after PMA stimulation whereas translocation of wild-type ⑀PKC was still minimal even after 10 min (Fig. 5A). Therefore, even when the D86A ⑀RACK mutant and wild-type ⑀PKC were in the same cell, D86A translocated faster than wild-type ⑀PKC.
Translocation Rates of the ⑀RACK ⑀PKC Mutants upon Cell Stimulation by a G Protein-coupled Receptor-We next determined whether differences in translocation rates of the different ⑀PKCs were also observed when translocation was stimulated via receptor signaling rather than by PMA. ATP has been previously used to activate PKC in CHO cells by stimulating purinergic G protein-coupled receptors (26). We found that translocation of GFP-⑀PKC upon stimulation with ATP to the cell periphery was faster than PMA-induced translocation (Fig.  5B versus Fig. 3). Translocation of the D86A and wild-type ⑀PKCs was already apparent 10 s after stimulation, whereas translocation of the D86N ⑀PKC enzyme occurred after 40 s of stimulation reaching a steady state at significantly higher levels than either D86A or wild-type ⑀PKCs (Fig. 5B).
Diffusion Rates of the ⑀PKC ⑀RACK Mutants-Different translocation rates may reflect differences in overall mobility (diffusion) of the ⑀PKCs in cells. Therefore, we measured fluorescence recovery after photo bleaching (FRAP) of the cytoplasmic enzyme; mobility of ⑀PKCs was measured by monitoring the time required for the fluorescence to recover in a bleached region. Fifty percent FRAP was reached at similar times for all ⑀PKCs. (The average of at least 10 cells/each ⑀PKC was: Wt ϭ 9.0 Ϯ 1.2, D86A ϭ 9.2 Ϯ 1.1 and D86N ϭ 9.1 Ϯ 1.6 s.) By comparing the fluorescence in the bleached region after full recovery (F ϱ ) with that observed before bleaching (F i ) and just after bleaching (F 0 ), we determined the mobile fraction ϭ (F ϱ Ϫ F 0 )/(F i Ϫ F 0 ). This was important to determine, since the mobile fraction may be affected by differences in interactions of the wild type and the mutant ⑀PKCs with other proteins and membranes. We found that the mobile fraction was the same for all ⑀PKCs (63 Ϯ 2% for Wt, 63 Ϯ 4 for D86A and 64 Ϯ 2.5% for D86N, averages of at least 10 cells/each). Therefore, differences in translocation rates were not due to differences in mobility of the inactive enzyme, but rather, to modulation of the intramolecular interaction between the ⑀RACK and RACK-binding site. DISCUSSION A key component in signal transduction is the inherent mechanism by which the enzymes remain inactive in the absence of extracellular stimuli. This mechanism involves the use of intramolecular interactions that stabilize a closed conformation with unexposed active site. Upon stimulation, the enzyme adopts an open conformation whereby intramolecular interactions are interrupted and binding sites for intermolecular in-teractions that stabilize the open form are exposed, resulting in a catalytically active enzyme. In the case of PKC, these intermolecular interaction sites are the binding sites for phospholipids and anchoring proteins (27,28). We have previously suggested that one intramolecular interaction in ⑀PKC that maintains the enzyme in a closed form is between an ⑀RACKbinding site and a ⑀RACK site (17). Here, we demonstrated that alterations in this intramolecular interaction affected the translocation rate of the enzyme, further supporting a role of this intramolecular interaction in ⑀PKC translocation and signaling.
As noted earlier, the ⑀RACK sequence in ⑀PKC is ϳ25% different from the sequence in ⑀RACK we suggested that the charge change (Asn-Asp) contributed to the difference in strength of the intramolecular interaction within ⑀PKC as compared with the intermolecular interaction between ⑀PKC and its RACK, ⑀RACK (16,17). By mutating the Asp-86 in the ⑀RACK sequence of ⑀PKC to an Asn, we have created an enzyme that translocates slower than the wild-type enzyme, presumably because we have increased the intramolecular interaction between the ⑀RACK and the RACK-binding site in ⑀PKC. Mutating Asp-86 to an Ala in ⑀PKC abolished the intramolecular interaction between the ⑀RACK and the RACKbinding site, and resulted in an enzyme that translocated at a faster rate than the wild-type enzyme.
Mutations in the ⑀RACK Site in ⑀PKC Affect Intrinsic Properties of ⑀PKC-We used three criteria to demonstrate the role of the ⑀RACK site in the intramolecular interaction within ⑀PKC and the role of Asp-86 in this interaction. The first criterion was the sensitivity of the enzyme to proteases; an open enzyme should be more susceptible to protease degradation after activation (19). We showed here that the D86N mutant was more resistant to proteolysis; D86N mutant required twice the time for the same extent of degradation of either the D86A or wild-type enzymes (Fig. 1). Therefore, D86N mutant is a more closed or inactive enzyme. Because sensitivity to proteolysis of the A mutant was the same as wild-type Asp-86 ⑀PKC, we could not determine whether it was conformationally different from the wild-type enzyme using this method.
The second criterion examined PL-dependent binding of wild-type and mutant enzymes to ⑀RACK. If Asp-86 is critical for an intramolecular interaction, we predicted that D86A would be less dependent on lipid activation for ⑀RACK binding. Indeed, the single amino acid substitution modulated the intramolecular interaction between the ⑀RACK and the ⑀RACKbinding site; the D86N mutant had a greater similarity to the ⑀RACK sequence, and a reduced ability to bind to its ⑀RACK, indicating that the D86N mutant is in a more closed conformation. In contrast, the D86A ⑀PKC mutant was less dependent on activators for RACK binding, indicating that it is in a more open conformation.
Mutations in the ⑀RACK Site and Translocation Rates of ⑀PKC in Cells-A third criterion demonstrating a critical role of the ⑀RACK site in intramolecular interactions examined the rate of translocation of the enzyme upon activation in cells. The D86A ⑀PKC mutant translocated significantly faster than either D86N or wild-type ⑀PKCs, as measured by cell fractionation studies. Using real-time confocal microscopy we demonstrated that the D86A mutant translocated at a faster rate than wild-type ⑀PKC, which in turn translocated faster than the D86N mutant. Together, it appears likely that the ⑀RACK site mediates a critical intramolecular interaction that stabilizes the closed conformation in ⑀PKC in the absence of stimulation.
A scheme for the mechanism of translocation of the different ⑀PKC mutants is in Fig. 6. Mathematical modeling of our data further elucidated the molecular events leading to translocation (see Fig. 4). Using non-linear regression analysis, an equation with two exponents gave a better fit than a single exponential equation, indicating that PMA-induced translocation involves at least two steps (Figs. 3 and 4). We proposed that the first step represents the opening and closing processes of the enzyme and the second step represents binding of the open enzyme to the cell membrane. Importantly, the steady state level of the D86N mutant was higher than that of either D86A or of wild-type ⑀PKC, indicating that the amount of D86N that reached the cell periphery was lower than the amount of either wild-type or D86A ⑀PKCs. The D86A mutant translocated significantly faster than either the D86N or wild-type ⑀PKCs. Since the steady state level of D86A in the particulate fraction was similar to the steady state level of the wild-type enzyme, and the second step of translocation (binding to the membrane) was the same for all mutants, the rate of closing of an enzyme, once it was open, could be considered negligible. Ochoa et al. (29) demonstrated that the binding of the V1 domain to PL is not altered by D86A mutation, supporting our hypothesis that the second step of translocation (binding to the membrane) is not altered. However, for the N mutant, the k 1 (rate of ⑀PKC opening) was slower than that of D86A and similar to wild-type ⑀PKC, and the k Ϫ1 (rate of ⑀PKC closing) was no longer negligible.
Similar to Shirai et al. (30), we found that ATP-induced translocation of ⑀PKC was much faster than PMA-induced translocation (Figs. 5B versus 3). Since ATP-mediated translocation was a fast process, differences between the wild type and D86A mutants were not observed at the time intervals analyzed. However, the translocation of the D86N mutant was significantly slower than either the wild type or A mutant, further supporting the importance of the RACK site and the disruption of intramolecular interaction for PKC translocation.
Schaefer et al. (31,32) suggested that differences in translocation between classical and novel PKCs are due to differences in diffusion rates, and collision efficiencies with the membrane. Although diffusion and collision with the membrane are likely factors in the translocation rate, our data demonstrate that conformational changes in the enzyme also occur, leading to at least a two-step process.
There is evidence for a two-step process in the translocation of classical PKC isozymes. Nalefski and Newton (25) demonstrated that binding of ␤PKC to the membrane is a two step process, in which one of the steps involves a conformational change (25). Bolsover et al. (33) recently demonstrated that ␣PKC undergoes a calcium-dependent conformational change exposing the lipid-binding sites, which precedes membrane binding. In the case of the novel PKC (calcium-independent) isozymes, it is still not clear what triggers the opening of the enzyme. Here, we showed that disrupting the intramolecular interaction between the RACK and RACK-binding site is a critical step in activation that precedes translocation and anchoring to the cell periphery. Whether this anchoring is mainly mediated by binding to membranes, whether it involves anchoring proteins, and whether binding to lipids precedes binding to proteins could not be determined in cells overexpressing ⑀PKC. However, we have previously demonstrated that translocation of endogenous ⑀PKC results in its co-localization with its ⑀RACK (34) and that disruption of binding to ⑀RACK in cells with a peptide corresponding to one of the RACK-binding sites in ⑀PKC (⑀V1-1 peptide) inhibits ⑀PKC translocation and colocalization with ⑀RACK (35,36). Importantly, we showed that a peptide corresponding to the ⑀RACK sequence induces ⑀PKC translocation and co-localization with ⑀RACK and triggers ⑀PKC function (17).
Our data suggest that the process of ⑀PKC translocation to the membrane that we followed in this study involves at least two steps and may occur independently of binding to RACK. An additional step involving RACK binding could not be detected in this system, since when the enzyme is overexpressed, as it is in this study, it is likely that the binding proteins, including RACKs, are no longer present in stoichiometric amounts relative to the enzyme (37). Indeed, the overexpressed wild-type ⑀PKC and the endogenous ⑀RACK did not co-localize in the cells, even after activation with phorbol ester. In contrast, the endogenous ⑀PKC co-localized with the ⑀RACK following activation in non-transfected cells (not shown). Because many attempts to co-express ⑀RACK with ⑀PKC have failed, the translocation experiments in this study reflect mainly the interaction of the GFP enzyme with lipids in the cell membrane.
Where in ⑀PKC is this RACK-binding site? The V1 domain of ⑀PKC is homologous to the C2 domain of the ␤PKC (1). However, there is an additional RACK-binding site in the V5 region of ␤PKC (38) and molecular dynamics studies with the C2 region of ␤PKC showed that an intramolecular interaction between the RACK and the RACK-binding site in the C2 region is not possible (39). Instead we suggest that the intramolecular interaction between the RACK and the RACK-binding site in ␤PKC is likely to occur between the C2 and V5 regions in ␤PKC. This may also be the case for ⑀PKC. Recently Stubbs and co-workers (40) have demonstrated that in ␣PKC there is also an additional intramolecular interaction between the C1 and C2 domains that maintains the enzyme in its inactive state. In addition, they have suggested that ␣PKC forms dimers through an intermolecular interaction between the C1 and C2 domains, we cannot reject the hypothesis that this is also the case for the ⑀RACK-binding site and the ⑀RACK (40).

CONCLUSIONS
By mutating the ⑀RACK site in the intact ⑀PKC, we have demonstrated the importance of interrupting the intramolecular interaction between the ⑀RACK and the ⑀RACK-binding site in the process of ⑀PKC translocation. To our knowledge, this is the first time that a single charge change made outside of the (-substrate) site or the catalytic site, or of calciumbinding sites evoked a change in translocation rate. Together, we conclude that disruption of the interaction between the ⑀RACK and the RACK-binding site is a critical and ratelimiting step in PKC translocation. FIG. 6. A scheme showing the first steps in ⑀PKC translocation to the membrane upon activation; a two-step process. An additional step in the process of translocation includes binding to the RACK (not shown in the scheme). For simplicity, we show only the intramolecular interaction between the ⑀RACK and the ⑀RACK-binding site. Disruption of an intramolecular interaction between the ⑀RACK site and the ⑀RACK-binding site must precede binding to the membrane and is a rate-limiting step in the process. Modulating this intramolecular interaction by mutating Asp-86 altered the translocation rates of ⑀PKC, by affecting the first step of the translocation process.