Increased Membrane Affinity of the C1 Domain of Protein Kinase Cδ Compensates for the Lack of Involvement of Its C2 Domain in Membrane Recruitment*

Protein kinase C (PKC) family members are allosterically activated following membrane recruitment by specific membrane-targeting modules. Conventional PKC isozymes are recruited to membranes by two such modules: a C1 domain, which binds diacylglycerol (DAG), and a C2 domain, which is a Ca2+-triggered phospholipid-binding module. In contrast, novel PKC isozymes respond only to DAG, despite the presence of a C2 domain. Here, we address the molecular mechanism of membrane recruitment of the novel isozyme PKCδ. We show that PKCδ and a conventional isozyme, PKCβII, bind membranes with comparable affinities. However, dissection of the contribution of individual domains to this binding revealed that, although the C2 domain is a major determinant in driving the interaction of PKCβII with membranes, the C2 domain of PKCδ does not bind membranes. Instead, the C1B domain is the determinant that drives the interaction of PKCδ with membranes. The C2 domain also does not play any detectable role in the activity or subcellular location of PKCδ in cells; in vivo imaging studies revealed that deletion of the C2 domain does not affect the stimulus-dependent translocation or activity of PKCδ. Thus, the increased affinity of the C1 domain of PKCδ allows this isozyme to respond to DAG alone, whereas conventional PKC isozymes require the coordinated action of Ca2+ binding to the C2 domain and DAG binding to the C1 domain for activation.

Many signaling proteins translocate from the cytosol to the membrane to become activated or to reach their substrates. Reversible binding to the membrane is mediated by modules such as the C1, C2, pleckstrin homology, and FYVE domains, with each recognizing specific lipid determinants or membrane properties. The protein kinase C (PKC) 3 family is a classic example of an enzyme whose function is regulated by membrane translocation. For this family of signaling proteins, the C1 and C2 domains direct the enzyme to the membrane following generation of diacylglycerol (DAG) and, for some isoforms, Ca 2ϩ (reviewed in Ref. 1).
The C1 domain is a 5-kDa globular domain composed primarily of two ␤-sheets that are pulled apart to expose a binding pocket for DAG and phorbol. Found in a number of other proteins in addition to PKC (notably, protein kinase D, DAG kinase, and Raf), this domain is an effective DAG sensor when expressed alone (reviewed in Refs. 2 and 3). Moreover, the C1 domain selectively recognizes phosphatidylserine (PS) (4,5). The C2 domain is another small (16 kDa) globular domain that is primarily composed of ␤-strands; in this case, three loops in the ␤-strands at one end of the domain form a Ca 2ϩ -binding site. Conventional C2 domains such as those found in phospholipase C, phospholipase A 2 , phosphatidylinositol 3-kinase, synaptotagmin, and Munc13 bind Ca 2ϩ in this pocket, and the binding drives the membrane interaction (reviewed in Refs. 6 and 7). Novel C2 domains such as that found in PTEN do not bind Ca 2ϩ , yet can still bind phospholipids (8). The coordinated use of a C1 and C2 domain allows sensitivity in the regulation of Ca 2ϩ /DAG-dependent PKC family members.
The PKC family is divided into three groups based on the composition of membrane-targeting modules N-terminal to the kinase domain, which in turns dictates the second messenger dependence (reviewed in Ref. 9). Conventional isozymes (␣, ␤I, ␤II, and ␥) contain a C1 and C2 domain and bind the ligands DAG and Ca 2ϩ , respectively. Novel isozymes (␦, ⑀, , and ) contain a C1 domain that mediates recognition of DAG and a "novel" C2 domain that lacks determinants required for Ca 2ϩ binding. Atypical isozymes ( and /) contain only one membrane-targeting domain, an atypical C1 domain that has an impaired ligand-binding pocket and that does not bind DAG or phorbol ester. Thus, the composition of membrane-targeting modules allows selective activation of isozymes by stimuli that generate only DAG or that elevate both DAG and Ca 2ϩ levels.
The mechanism of membrane recruitment and activation of conventional PKC isozymes has been well studied. In the absence of signals, PKC exists in an inactive state in the cytosol. The enzyme is recruited to membranes following receptor-mediated production of DAG and release of intracellular Ca 2ϩ . PKC first binds Ca 2ϩ in the cytosol via its C2 domain, an event that engages the domain on the membrane via a low affinity interaction (10). The enzyme then diffuses in the two-dimensional plane of the membrane to find DAG, which binds the C1 domain. Although each domain binds with relatively low affinity to the membrane, the coordinated binding of both provides the energy to release an autoinhibitory pseudo-substrate sequence from the substrate-binding site of PKC, allowing binding of substrate and phosphorylation (5). Thus, conventional PKC uses a domain that binds a soluble ligand, the C2 domain, to pre-target it to membranes, thereby allowing the C1 domain to more efficiently find its ligand, which is restricted to the membrane (10). Although conventional isozymes have the benefit of pre-targeting by a Ca 2ϩ -sensitive C2 domain, novel PKC isozymes lack a Ca 2ϩ -interacting domain, raising questions about membrane recruitment.
Here, we address the molecular mechanism driving the membrane recruitment of PKC␦, a novel isozyme. We show that PKC␦ relies exclusively on the C1 domain for membrane recruitment and that the intrin-sic affinity of this module for membranes is 2 orders of magnitude higher in PKC␦ compared with PKC␤II, resulting in comparable membrane binding of both isozymes in the presence of saturating ligands. In vivo imaging studies of wild-type PKC␦ and a C2-deleted construct revealed that, even in the context of the cell, the C2 domain does not affect the location or activity of PKC␦. Thus, this isozyme relies on one membrane-targeting domain (tuned to have an appropriately high affinity for membranes) for function.
[␥-32 P]ATP and [ 3 H]dipalmitoylphosphatidylcholine were from PerkinElmer Life Sciences. Restriction enzymes were from New England Biolabs Inc. PCR primers were prepared by Genset Oligos. Thrombin and glutathione-Sepharose 4B were from Amersham Biosciences. Electrophoresis reagents were from Calbiochem and Bio-Rad. Isopropyl ␤-D-thiogalactopyranoside and leupeptin were from Calbiochem. Antibody to PKC␦ was from Cell Signaling Technology. All other chemicals were reagent-grade.

Preparation of Lipid Vesicles
Mixtures of phospholipids in chloroform/methanol (2:1) were dried under a stream of nitrogen followed by vacuum and then resuspended in buffer containing 170 mM sucrose, 5 mM MgCl 2 , and 20 mM HEPES (pH 7.5). They were subjected to five freeze-thaw cycles. Vesicles were extruded through two 0.1-m polycarbonate filters in an extruder (Avestin Inc.). The vesicles were suspended in 100 mM KCl and 20 mM HEPES (pH 7.5) to dilute the sucrose and centrifuged at 100,000 ϫ g for 30 min. The pellet was resuspended in KCl buffer to a final concentration of 1.0 mM lipid as monitored by trace amounts of

Incorporation of PMA into Vesicles
PMA was dissolved in Me 2 SO and added to a solution of vesicles while vortexing. The vesicles were incubated for 30 min prior to use to ensure incorporation of all of the PMA.

Constructs
C2-deleted PKC␦ was made by removing the first 126 residues of full-length PKC␦. Full-length PKC␦ and C2-deleted PKC␦ were cloned into pFastBac (Invitrogen). C2-deleted PKC⑀ was made by removing the first 100 residues of full-length PKC⑀. Full-length and C2-deleted PKC⑀ constructs were cloned into pcDNA3.1(Ϫ) using the XhoI and HindIII sites. The isolated C1 domain of PKC␤II was obtained by cloning residues 91-161 of PKC␤II into pGEX-KG using the BamHI and Hind III sites. The isolated C2 domains of PKC␤II and PKC␦ were obtained by cloning residues 157-298 of PKC␤II and residues 1-133 of PKC␦ into the same sites. The P169G (C1A domain) mutant of PKC␦ was kindly provided by Peter Blumberg. It was cloned into the XhoI and HindIII sites of pcDNA3.1(Ϫ).

Purification of PKC
PKC␤II was purified to homogeneity from the baculovirus expression system as described previously (11). His-tagged wild-type PKC␦ or C2-deleted PKC␦ was expressed in Sf21 cells. Cells were infected at 80% confluency for 1.5 h with virus containing the PKC constructs. The virus was then diluted in EX-CELL (10 ml of EX-CELL with 1% streptomycin for each T75 dish). Cells were grown for 72 h and then pelleted at low speed. 10 ml of lysis buffer (50 mM Tris (pH 8.2), 100 mM KCl, 1% Nonidet P-40, ␤-mercaptoethanol, microcystin, phenylmethylsulfonyl fluoride, benzamidine, and leupeptin) was added. Cells were lysed using a French press and spun at 9000 rpm for 15 min. The supernatant was added to 1.5 ml of nickel-agarose column. Washes were with 11 ml of Buffer A (20 mM Tris (pH 8.2), 500 mM KCl, 20 mM imidazole, 5 mM ␤-mercaptoethanol, and 10% glycerol), 1 ml of Buffer B (20 mM Tris (pH 8.2), 1 M KCl, 5 mM ␤-mercaptoethanol, and 10% glycerol), and 3 ml of Buffer A. Elution was with 7 ml of Buffer C (20 mM Tris (pH 8.2), 100 mM KCl, 1 M imidazole, 5 mM ␤-mercaptoethanol, and 10% glycerol). Fractions were assayed by kinase assay and Western blotting. The His tag was cleaved off with enterokinase. The isolated C1 and C2 domains from PKC␤II and PKC␦ were expressed as glutathione S-transferase fusion proteins in BL21-pLysS cells and purified using glutathione-Sepharose beads as described previously (5). PKC⑀ constructs were transfected into human embryonic kidney cells using the Superfect transfection system (Qiagen Inc.). Cells were lysed in phosphatebuffered saline containing 0.1% Triton after 48 h. Lysates were spun at 4°C for 30 min at 100,000 ϫ g to separate the soluble material from the insoluble pellet. The supernatants were used in PKC binding assays. Binding curves obtained with purified PKC and with PKC from the supernatant of a high speed spin were shown previously to be similar (12).

Sucrose-loaded Vesicle Binding Assay for Full-length PKC␤II, Full-length PKC␦, and C2-deleted PKC␦
Binding of PKC to sucrose-loaded vesicles was measured as described previously (13). PKC was incubated for 10 min with 100 M sucroseloaded vesicles in buffer containing 100 mM KCl, 5 mM MgCl 2 , 20 mM HEPES (pH 7.5), 200 M CaCl 2 , and 0.3 mg/ml bovine serum albumin, followed by centrifugation at 100,000 ϫ g for 30 min at 25°C to separate free PKC from vesicle-bound PKC. The percentage of PKC bound to vesicles in the pellet was determined by assaying both the supernatant and pellet fractions for PKC activity under identical conditions using protamine sulfate as a substrate. Vesicles were added to the supernatant fraction to keep the total amount of lipid in both fractions equal.
The percentage of membrane-bound enzyme was determined using Equations 1 and 2, where A v is the vesicle-associated kinase activity, A b is the activity of the bottom fraction (in cpm), A t is the activity of the top fraction (in cpm), ␣ is the fraction of sedimented vesicles, and ␤ is the fraction of kinase activity found in the bottom fraction in the absence of vesicles.
Binding of isolated domains was determined similarly, but the top and bottom fractions were run on a gel and silver-stained, and bands were quantified by densitometry. Densitometry results were used in place of activity in Equations 1 and 2. Binding of PKC⑀ was assayed using a Western blot in place of silver-stained gels.
The data were fit to Equation 3,

PKC Translocation in Whole Cells
Cell Culture-COS-7 cells were plated onto glass coverslips in 35-mm dishes and grown to 60 -80% confluency. Cells were transfected with FuGENE 6 (Roche Applied Science) and grown 12 h post-transfection.
Imaging-Cells were imaged in the dark at room temperature with phorbol 12,13-dibutyrate (PDBu) or peroxide added as noted. Images were acquired on a Zeiss Axiovert microscope (Carl Zeiss MicroImaging, Inc.) using a MicroMax digital camera (Roper Princeton Instruments) controlled by MetaFluor Version 3.0 software (Universal Imaging, West Chester, PA). Data were collected through a 10% neutral density filter. Yellow fluorescent protein images were acquired through a 495/10-nm excitation filter, a 505-nm dichroic mirror, and a 535/ 25-nm emission filter. Red fluorescent protein (monomeric; RFP) images were obtained through a 568/25-nm excitation filter, a 600-nm dichroic mirror, and a 653/95-nm emission filter. Cyan fluorescent protein (CFP) images were obtained through a 420/10-nm excitation filter, a 450-nm dichroic mirror, and a 475/40-nm emission filter. Fluorescence resonance energy transfer (FRET) images were obtained through the same excitation filter and dichroic mirror as CFP images, but the emission filter was 535/25 nm. Integration times were varied between 50 and 200 ms to optimize signals and to minimize photobleaching. Fig. 1A shows that PKC␦ bound to model membranes in a manner that was sigmoidally dependent on the mol fraction of PS; analysis of binding data revealed half-maximal binding with 26.0 Ϯ 0.8 mol % PS and a Hill coefficient of 4.2 Ϯ 0.5. Deletion of the C2 domain resulted in a modest enhancement of membrane binding, with half-maximal binding occurring at 20 Ϯ 2 mol % PS and a Hill coefficient of 4 Ϯ 1. This modest enhancement of membrane binding of the C2-deleted construct was even less apparent when binding was measured in the presence of Ca 2ϩ (Fig. 1B); the construct lacking the C2 domain was half-maximally bound to membranes containing 21 Ϯ 2 mol % PS compared with 21.6 Ϯ 0.3 mol % for the full-length protein. Activity curves were similar to binding curves in that the C2-deleted construct reached half-maximal activation at 2-3% lower PS compared with full-length PKC␦ (data not shown), whereas the Hill coefficient remained unchanged. Deletion of the C2 domain of PKC␦ did not decrease the enzyme-specific activity. These data reveal that deletion of the C2 domain slightly increases the affinity of PKC␦ for membranes, without significantly altering the degree of apparent cooperativity. Thus, the C2 domain of PKC␦ does not contribute to the PS dependence for membrane binding or activation of PKC␦ or to the apparent cooperativity in these processes.

The C2 Domain of PKC␦ Does Not Contribute to Membrane Binding-
In contrast to PKC␦, the binding of PKC␤II to membranes was strongly dependent on Ca 2ϩ . Fig. 1C shows that the binding of PKC␤II to membranes displayed a steep sigmoidal dependence on PS in the presence of Ca 2ϩ (closed circles; half-maximal binding at 14.52 Ϯ 0.06 mol % PS, n ϭ 12 Ϯ 1) and was barely measurable in the absence of Ca 2ϩ (open squares; only 20% maximal binding to vesicles containing 60 mol % PS). In the absence of Ca 2ϩ , the C2 domain of PKC␤II was not engaged on the membrane, and membrane binding was driven by the C1 domain alone. Thus, the relative contributions of the C1 domains to membrane binding of PKC␤II and PKC␦ can be compared by determin- ing the membrane affinity of PKC␤II in the absence of Ca 2ϩ with that of PKC␦. From Fig. 1 (A and C), calculation of apparent binding affinities by quantifying the amount of bound and free PKCs at a fixed PS concentration (25 mol %) reveals that the affinity of PKC␦ for membranes was 100-fold greater than the affinity of PKC␤II under conditions in which binding was mediated only by the C1 domain. These data reveal that PKC␦ compensates for its non-membrane-binding C2 domain by having a C1 domain that binds membranes with 2 orders of magnitude higher affinity compared with the C1 domain of a conventional PKC.
To confirm that the C2 domain of PKC␦ does not contribute to membrane binding, we tested the ability of the isolated C2 domain to bind anionic membranes. Fig. 2 shows that the C2 domain of PKC␦ did not bind membranes even at 0.5 mM total lipid. In contrast, the C2 domain of the conventional isozyme PKC␤II bound vesicles containing 40 mol % PS with an apparent K A of 3.4 ϫ 10 3 M Ϫ1 , consistent with previously published data (5). Similarly, the PMA dependence for membrane bind-ing and activity were similar for both full-length PKC␦ and C2-deleted PKC␦, with the C2-deleted construct again binding to membranes slightly better than the full-length enzyme (data not shown). Deletion of the C2 domain also did not affect the binding of PKC␦ to vesicles containing phosphatidylglycerol, cardiolipin, or ceramide (data not shown). These data reveal that the C2 domain does not contribute to membrane binding regardless of the mechanism used to tether PKC␦ to membranes.
The C2 Domain of PKC␦ Is Predicted to Repel Anionic Membranes, whereas That of PKC␤II Is Predicted to Bind Membranes-To probe the physical basis for the inability of the PKC␦ C2 domain to bind membranes, we performed potential of mean force calculations to determine the free energy change as the C2 domain approaches anionic membranes. The construction of the 1-palmitoyl-2-oleoylphosphatidyl-Lserine/1-palmitoyl-2-oleoylphosphatidylcholine vesicle model with atomic details has been described elsewhere (14). The electrostatic solvation free energies were calculated by solving the Poisson-Boltzmann equation at an ionic strength of 0.15 M using the adaptive Poisson-Boltzmann solver (15). The Protein Data Bank codes for the atomic structures of the PKC␤II C2 domain and the PKC␦ C2 domain are 1A25 and 1BDY, respectively. The orientation of the PKC␤II C2 domain was determined according to Monte Carlo simulations with the above free energy calculation, whereas the orientation of the PKC␦ C2 domain was determined by structural alignment with the PKC␤II C2 domain at its optimal orientation with the lowest binding free energy. Fig. 3 shows that the C2 domain of PKC␤II is predicted to be attracted to membranes containing 28.8 mol % PS in phosphatidylcholine (PC) as its center of mass approaches a radial distance of 135 Å from the vesicle membrane. At this distance, where the PKC␤II C2 domain is actually anchored right on top of the PS lipid, the free energy of binding is on the order of Ϫ10 kcal/mol. In striking contrast, the free energy of binding is predicted to increase dramatically as the C2 domain of PKC␦ approaches anionic membranes, with a free energy on the order of ϩ150 kcal/mol describing the interaction of the PKC␦ C2 domain with membranes at a distance of 135 Å. It should be noted that this implies that the PKC␦ C2 domain will not approach this location in reality. This result suggests that the C2 domain of PKC␦ is strongly repelled from anionic membranes if it is forced to very close proximity and therefore will not contribute to the binding of PKC␦ to the anionic membrane.   The C2 Domain of PKC␦ Does Not Affect the Translocation of PKC in COS-7 Cells-We next addressed whether the C2 domain regulates the membrane translocation of PKC␦ in the context of the cell. COS-7 cells were transfected either with the cDNA encoding full-length PKC␦ tagged with RFP or with the cDNA encoding C2-deleted PKC␦ tagged with CFP. Cells were imaged before and after addition of 200 nM PDBu. Deletion of the C2 domain of PKC␦ did not significantly alter the fluorescence distribution pattern of tagged PKC␦ or the kinetics of redistribution (Fig. 4A). Specifically, PKC␦ redistributed from the cytosol to the Golgi following addition of PDBu. The rate of translocation of C2-deleted PKC to the Golgi was not significantly different from that of wildtype PKC␦ (compare images at 30 s), indicating that the C2 domain does not affect the redistribution of PKC␦. Slightly more C2-deleted construct was observed in the nucleus before translocation and at the nuclear membrane after translocation. A fluorescently tagged construct of the isolated C2 domain of PKC␦ was primarily cytosolic, and this distribution was unaffected by PDBu (data not shown). Taken together, these data suggest that the C2 domain in isolation or within the context of the full-length PKC protein does not contain determinants required for targeting or translocation of PKC to membranes. To test the effect of the C1B domain on translocation, full-length and C2-deleted constructs with a mutation in the C1A domain that impairs function were used. Fig. 4B shows that both constructs translocated to the Golgi, indicating that the C1B domain drives the interaction of the full-length protein.
Cells stained with the Golgi marker BODIPY TR ceramide (Molecular Probes) displayed fluorescence patterns identical to those displayed by cells transfected with PKC␦ and treated with PDBu (Fig. 4C), indicating that PKC␦ does in fact translocate to the Golgi. Translocation of PKC␤II to the plasma membrane has been shown previously (16).
The C2 Domain of PKC␦ Does Not Affect the Activation of PKC in COS-7 Cells-We used the genetically encoded PKC activity reporter CKAR (C-kinase activity reporter) to measure the activation of PKC␦ in live cells (19). Phosphorylation of CKAR by PKC leads to a decrease in FRET (19). CKAR and either full-length or C2-deleted PKC␦ fused to RFP were transfected into COS-7 cells, and the change in the FRET ratio (plotted as CFP emission/FRET emission) was measured after addition of stimulus. The rate and magnitude of activation in response to PDBu were similar for both constructs (Fig. 4D). The initial activation rates were 0.12 Ϯ 0.03 FRET units/min for full-length PKC and 0.11 Ϯ 0.05 FRET units/min for C2-deleted PKC␦. More than 80% of the PKC activity was from the transfected constructs as assessed by measuring the PKC activity in lysates of control and transfected cells. The average rates for endogenous PKC were only 0.0140 Ϯ 0.0008 FRET units/min (data not shown). The RFP tag did not interfere with CKAR emission properties; experiments performed with glutathione S-transferase-tagged PKC␦ instead of the RFP-tagged construct yielded similar results. Similarly, both full-length and C2-deleted constructs were activated by H 2 O 2 (Fig. 4E), which has been shown to activate PKC␦ indirectly through phosphorylation of Tyr 311 (possibly by Src family kinases) (17) and which has also been shown to activate PKC directly by oxidation of cysteines in the C1 domain (reviewed in Ref. 18). There was some cellto-cell variation in activation rates and magnitudes with peroxide, but both constructs were activated to similar extents. Representative curves are shown. Control experiments using the non-phosphorylatable mutant of CKAR revealed no FRET change upon either PDBu or peroxide treatment in cells transfected with PKC␦ constructs or in cells without PKC transfection (19) (data not shown).
PKC␦ Compensates for Lack of Membrane Binding of the C2 Domain by Using the C1 Domain to a Greater Extent- Fig. 5 shows that fulllength PKC␤II reached half-maximal binding at lower PS compared with full-length PKC␦. Specifically, half-maximal binding to membranes was achieved with 22.2 Ϯ 0.2 mol % PS for PKC␤II compared with 45 Ϯ 4 mol % PS for PKC␦ in the presence of Ca 2ϩ . The Hill coefficients were 6.0 Ϯ 0.4 for PKC␤II and 4.6 Ϯ 1.7 and 4.2 Ϯ 1.0 for PKC␦ in the presence and absence of Ca 2ϩ , respectively. As observed in the presence of DAG (Fig. 1), the PS dependence in the presence of PMA displayed higher cooperativity for PKC␤II compared with PKC␦. Thus, full-length PKC␤II, which uses both the C1 and C2 domains, binds membranes with higher affinity and higher apparent cooperativity compared with full-length PKC␦, which uses only the C1 domain.
We next examined the contribution to membrane affinity of just the C1 domain of each isozyme. For these experiments, vesicles contained PMA instead of DAG, which was used in the experiments for Figs. 1 and 2, because the higher affinity of PKC for phorbol over DAG was needed to achieve adequate binding of isolated domains. Fig. 6 shows that fulllength PKC␦ (open squares) reached half-maximal binding at 13.3 Ϯ 0.9 mol % PS. To specifically look at the contribution of only the C1B domain, membrane binding of a construct of PKC␦ with a mutation in its C1A domain that impairs C1A domain function (closed squares) was examined; half-maximal binding was achieved at a similar PS concentration (closed squares; 17 Ϯ 3 mol % PS). The results were similar when the isolated C1B domain of PKC␦ was used, but the construct had some solubility problems; therefore, it was easier to assay the binding of fulllength mutant PKC␦. Full-length PKC␤II incubated with EGTA (open circles) reached half-maximal binding at 27 Ϯ 2 mol % PS. Under these conditions, only the C1A ϩ C1B domains were used. The isolated C1B domain of PKC␤II (closed circles) reached half-maximal binding at 30 Ϯ 3 mol % PS. As reported previously (5), the membrane affinity of this C1B domain is comparable for membranes as an isolated module or in the context of the full-length protein. Quantification of the binding data at 20 mol % PS revealed that the apparent membrane affinity for fulllength mutant PKC␦ was ϳ3-fold higher than that for the isolated C1B domain of PKC␤II. These data reveal the C1B domain of PKC␦ has an ϳ3-fold higher intrinsic affinity for membranes compared with the C1B domain of PKC␤II at this lipid composition. Assays with isolated domains and mutant proteins (Fig. 6) were performed with 1% PMA in the membrane versus 0.03% PMA for full-length proteins (Fig. 5). The Hill coefficients were 2.6 Ϯ 0.8, 3.1 Ϯ 0.5, 3 Ϯ 1, and 3 Ϯ 0.5 for constructs that use the C1B domain or C1A ϩ C1B domains of PKC␤II and the C1B domain or C1A ϩ C1B domains of PKC␦, respectively.
PKC⑀ Uses Its C1 Domain to an Intermediate Extent Compared with PKC␤II and PKC␦-Finally, we examined the effect of deleting the C2 domain on the membrane binding of another novel PKC isozyme, PKC⑀. Fig. 7A shows that deletion of the C2 domain of PKC⑀ caused a slight decrease in binding to PS/DAG membranes in contrast to deletion of the C2 domain of PKC␦, which modestly enhanced binding. Fig.  7B presents a comparison of the binding data from Figs. 1 (A and C) and 7A for PKC␦, PKC⑀, and PKC␤II under conditions in which only the C1 domain drives the membrane interaction. The membrane binding affinity of the C1 domain of PKC⑀ was slightly weaker than that of PKC␦, reaching half-maximal binding at 34 versus 19% PS, but much stronger than that of PKC␤II, which barely bound under the conditions used (5 mol % DAG with increasing mol % PS). These data reveal that the membrane binding affinity of the C1 domain of novel PKCs is higher than that of conventional isozymes to compensate for the lack of contribution of the C2 domain and that, within subclasses, the membrane affinity is further fine-tuned.

DISCUSSION
In this study, we have shown that the C2 domain of PKC␦ has no intrinsic affinity for membranes and does not contribute to either the membrane translocation or cellular activity of this isozyme. Furthermore, we have shown that this isozyme compensates for the nonfunctional C2 domain by having a C1 domain that binds membranes 2 orders of magnitude more tightly than the C1 domain of conventional PKC isozymes, which have two membrane tethers. Thus, our results reveal a model in which determinants on the C1 domain tune the affinity with which PKC isozymes bind membranes. Most important, isozymes whose membrane translocation is driven by a single membrane-targeting module and those whose membrane translocation is driven by two membrane-targeting modules adjust the affinities of individual modules so that the net binding is comparable.
The C2 Domain of PKC␦ Is Not a Membrane-targeting Module-The C2 domain is present in a variety of proteins and serves the function of membrane-targeting module as well as protein-interacting module. Most C2 domains bind Ca 2ϩ in an aspartic acid-lined cavity formed between loops in the ␤-strand-rich domain. Binding of Ca 2ϩ neutralizes the negative electrostatic potential surrounding the Ca 2ϩ -binding site and promotes binding to either neutral (as occurs for the C2 domain of phospholipase A 2 ) or anionic (as occurs for conventional PKCs) phos-  A, binding of full-length PKC⑀ and C2-deleted PKC⑀ to membranes containing 5 mol % DAG with increasing amounts of PS. Binding was measured as described under "Experimental Procedures." Curves were fit to Equation 3 under "Experimental Procedures." Data represent the average of two independent experiments performed in duplicate. B, replot of C2-deleted PKC⑀, C2-deleted PKC␦, and full-length PKC␤II from A and Fig. 1 (A and C), respectively.
pholipids. The C2 domain of novel PKCs lacks key aspartic acid residues required for Ca 2ϩ coordination and is thus unable to bind Ca 2ϩ . Whether this module constitutively binds anionic membranes has not been established. Elucidation of the crystal structure of the C2 domain of PKC␦ suggested several possible membrane-interacting sites, either clusters of positively charged residues or regions of neutral/hydrophobic residues (15). In addition, electrostatic calculations predicted a very weak affinity of the C2 domain of PKC␦ for PS that also involved hydrophobic interactions (20).
Here, we have show that, under all conditions tested, the C2 domain of PKC␦ does not interact with membranes. In fact, deletion of the C2 domain modestly facilitates the membrane binding of PKC␦. Such autoinhibition by the C2 domain has also been reported for the Aplysia novel PKC (21). One possibility is that the non-membrane-binding C2 domain sterically hinders access of the C1 domain to membranes. A second possibility is that electrostatic repulsion by the novel C2 domain hinders the access of PKC␦ to membranes. This latter possibility is supported by the thermodynamic calculations showing electrostatic repulsion of the C2 domain of PKC␦ from anionic membranes (Fig. 3). In this case, repulsion from membranes can be overcome by engagement of the C1 domain on membranes.
Comparison of the structure of the C2 domain of PKC␦ with C2 domains that do not bind Ca 2ϩ but do bind lipids reveals that several important lipid-binding residues/regions are not present in PKC␦. The C2 domains of PKC␦ and PKC⑀ share only 19% sequence identity and are proposed to form two distinct subclasses of novel C2 domains (22). The C2 domain of PKC⑀ has recently been shown to bind weakly to PS, phosphatidic acid, and phosphatidylglycerol (23). The lipid-binding regions were localized to residues in loops 1 and 3 of the C2 domain of PKC⑀ that are not conserved in the C2 domain of PKC␦. There is also an extra ␣-helix in PKC⑀ that is not found in PKC␦ (8). The C2 domain of PKC␦ is also similar in structure to that of PTEN, which also does not bind Ca 2ϩ , but does bind PS/PC. However, the C2 domain of PTEN has a lysine-rich sequence in the CBR3-like region, which has been implicated in membrane binding (8). This sequence is absent in the C2 domain of PKC␦, which contains acidic residues instead (8).
Our results are consistent with a recent study that compared the binding abilities of the PKC␦ C1A and C1B domains and also looked briefly at a C2-deleted construct of PKC␦ that bound similarly to fulllength PKC␦ (24). However, our studies go further than these conclusions. Our results reveal that the C2 domain of PKC␦ also does not bind phosphatidylglycerol, ceramide, and cardiolipin in vitro, lipids that have been proposed to modulate the function of this isozyme. Specifically, PKC␦ has been show to bind ceramide in vitro (25) and to translocate to the Golgi after ceramide treatment, where it is tyrosine-phosphorylated and activated (26). PKC␦ has also been shown to translocate to mitochondria in response to oxidative stress (27). Cardiolipin is found in the inner mitochondrial membrane and could possibly interact with the C2 domain. However, our data indicate that binding to cardiolipin is through the C1 domain as well (data not shown).
The C2 domain of PKC type II from Aplysia, which has the highest degree of homology to mammalian PKC⑀, has been reported to have no significant affinity for membranes unless it is phosphorylated at Ser 37 . It was concluded that the phosphate at position 37 causes two loops between strands to pull apart, thus exposing a new lipid-binding site (21). Mutation of a similarly positioned Ser to Glu (S16E and S24E) in the C2 domain of PKC␦ did not confer lipid binding (data not shown). Phosphorylation by protein kinase A, which is predicted to phosphorylate Thr 50 (28), did not promote membrane binding. The above results indicate that the PKC␦ C2 domain does not bind to lipids in its native state or following phosphorylation at Ser/Thr residues. Rather, this module likely serves as a protein-interacting domain, and some studies have found binding partners for it such as actin and GAP-43 (29,30). A recent study reported that the C2 domain of PKC␦ is in fact a phosphotyrosine-binding domain and that the C2 domain binds directly to a glycoprotein, CDCP1, following phosphorylation by Src or H 2 O 2 treatment, the latter of which broadly promotes tyrosine phosphorylation (31). We wondered if binding of the C2 domain to phosphoproteins could cause translocation or activation of PKC␦ in cells. In our studies, activation in response to peroxide (as measured by CKAR) was observed for both full-length PKC␦ and C2-deleted PKC␦, suggesting that the C2 domain is not crucial for this. Very little, if any, translocation was observed in response to peroxide with either construct. This is in agreement with a previous study that showed a 3-fold increase in activation of PKC␦ in response to H 2 O 2 , but no translocation (32).
Phorbol ester-induced translocation of full-length PKC␦ and C2-deleted PKC␦ in COS-7 cells is not significantly different. Both constructs translocate to the Golgi in ϳ2-3 min. Very subtle differences can be seen in the translocation patterns of full-length versus C2-deleted PKC␦ (Fig. 4A). The overlay shows more nuclear staining with C2-deleted PKC␦ than with full-length PKC␦. It is possible that the C2 domain recognizes protein binding partners that are localized to the nucleus. Moreover, the destination of PKC␦ in the cell depends on the hydrophobicity of the phorbol ester used (33,34). In fact, PKC␦ has been shown to translocate to the plasma membrane, nuclear membrane, Golgi, endoplasmic reticulum, and mitochondria in various cell types and in response to different stimuli (27,35,36). However, the C2 domain is not needed for translocation of PKC␦ in COS-7 cells. It has been shown that the V5 region (specifically the last few residues of the catalytic domain) of PKC␦ is responsible for some of its localization, translocation, and downstream effects (37). The above literature agrees with our observations that the C2 domain of PKC␦ does not contain determinants for localization of PKC␦, and in fact, PKC␦ is also activated by tyrosine phosphorylation and in this way does not even require membrane binding (17).
Mechanism of Apparent Cooperativity in Lipid Binding-The binding of conventional PKCs to membranes has long been known to show steep sigmoidal dependence on the mol fraction of PS (38). This steep sigmoidal dependence is observed only in membranes containing the C1 domain ligands (DAG or phorbol esters) and only when the anionic lipid is sn-1,2-phosphatidyl-L-serine. Kinetic analyses have revealed that conventional PKCs bind eight PS molecules in large unilamellar vesicles with Hill coefficients of 4 -8 describing this binding (39,40). Based on these observations, we have hypothesized that PKCs bind multiple anionic phospholipids, with stereospecific binding to at least one PS molecule (41). The isolated C1 (but not C2) maintains domain stereospecific binding to PS, revealing that the C1 domain contains determinants that specifically recognize sn-1,2-phosphatidyl-L-serine (5). Some of this high cooperativity results from the reduction in dimensionality that occurs when PKC translocates from the cytosol to the membrane upon engaging the first lipid. Beyond this contribution to cooperativity, the remaining cooperativity has yet to be accounted for.
Here, we have shown that a novel PKC isozyme that uses only the C1 domain for membrane binding displays markedly reduced cooperativity compared with isozymes that can also use a C2 domain. Specifically, a Hill coefficient of 12 describes the apparent cooperativity in the binding of PKC␤II to PS-containing membranes compared with 4 for the binding of PKC␦ to equivalent PS-containing membranes. These data suggest that the coordinated recruitment of the C1 and C2 domains to membranes accounts for some of the high apparent cooperativity in PS binding for PKC␤II. This high apparent cooperativity is reduced when only one domain engages on the membranes. Our current data support a model in which the C1 domain specifically binds at least one PS molecule, in which the binding of this first lipid increases the probability of binding additional lipids through reduction in dimensionality, and in which the involvement of the C2 domain at least doubles the number of anionic lipids that retain PKC at membranes.
The C1 Domain of PKC␦ Has Higher Membrane Affinity than the C1 Domain of Conventional PKCs-Analysis of membrane binding of PKC␤II in the presence or absence of Ca 2ϩ revealed that the C2 domain is the major determinant that recruits this isozyme to membranes even in the presence of high concentrations of DAG. For example, the C2 domain contributes Ͼ99% of the binding energy to membranes containing 5 mol % DAG and 20 mol % PS (K A ϭ 10 6 M Ϫ1 when the C2 domain is engaged with 200 M Ca 2ϩ and the C1 domain is engaged with 5 mol % DAG, and K A ϭ 10 2 M Ϫ1 when only the C1 domain is engaged) (Fig. 1C). In marked contrast, the C1 domain alone drives the membrane interaction of PKC␦ (K A ϭ 10 4 M Ϫ1 at 20 mol % PS and 5 mol % DAG) (Fig. 1A). Therefore, the C1 domain of PKC␦ binds ϳ2 orders of magnitude more tightly to membranes than that of PKC␤II, thus compensating for the lack of a membrane-binding C2 domain. We have previously shown that the C1 and C2 domains of PKC␤II function as independent membrane-targeting modules; the energy released by binding both modules is just slightly less than the sum of binding each individual module (5). Consistent with this, it was shown that both the C1 and C2 domains of PKC␤II are required for efficient translocation of the enzyme to the membrane in cells following agonist stimulation; deletion of either domain of PKC␤II impairs translocation in response to physiological stimuli, an impairment that can be rescued by elevating Ca 2ϩ or adding phorbol esters to tighten the membrane affinity of the functional domain (42).
The C2 Domain of PKC⑀ Has Modest Membrane Affinity-Deletion of the C2 domain of PKC⑀ caused a modest decrease in the PS dependence of membrane binding. This is consistent with a report that the isolated domain has a weak affinity for anionic membranes (23). Similarly, a recent report showed that the isolated C2 domain of PKC⑀ has a weak affinity for membranes, although it was also reported that the affinity of full-length PKC⑀ for membranes increases when the C2 domain is deleted (43).
Comparison of the membrane affinity of C2-deleted constructs of PKC␦ and PKC⑀ reveals that C2-deleted PKC␦ binds membranes with ϳ5-fold higher affinity (for example, K A ϭ 10 4 M Ϫ1 (Fig. 1A) and 2 ϫ 10 3 M Ϫ1 (Fig. 7A) at 20 mol % PS for PKC␦ and PKC⑀, respectively). For the three PKC isozymes tested, the affinity of the C1 domain for membranes is tuned according to the affinity of the C2 domain. Thus, the affinity of the C1 domain of PKC␦ (C2 domain with relatively no affinity for membranes) is the strongest, followed by that of PKC⑀ (C2 domain with relatively low affinity) and then PKC␤ (C2 domain with relatively high affinity). However, the K d for the binding of the C2 domain of PKC␤II is 7.3 ϫ 10 Ϫ10 M with vesicles containing 40 mol % PS (10), whereas the K d for binding of the C2 domain of PKC⑀ is 5.4 ϫ 10 Ϫ5 M with the same vesicle composition (23), and the C2 domain of PKC␦ does not bind at all. Although the K d values were obtained from different assays, they are still an indication of the relative binding abilities of each domain. The Hill coefficient for the C2 domain of PKC⑀ is 2.52 (23), which fits in well with the estimated two binding sites for PS on the C2 domain of PKC␤II. Therefore, it is possible that the C2 domains of PKC␤II and PKC⑀ bind to the same number of PS molecules, even though their binding strengths vary greatly.
Conclusion-Here, we have shown that the C2 domain of PKC␦ does not function as a membrane-targeting module. For this isozyme, high affinity membrane binding is achieved exclusively by the use of the C1 domain, which binds membranes 2 orders of magnitude more tightly than the C1 domain of conventional PKC isozymes. Our data reveal that the affinity of the C1 domain is tuned to allow comparable membrane affinity of both Ca 2ϩ -regulated and Ca 2ϩ -independent isozymes. The kinetics of membrane translocation are faster for conventional PKC isozymes that can take advantage of pre-targeting to membranes by Ca 2ϩ , but the affinities with which these isozymes bind membranes are comparable at saturating ligand concentrations.