A New Phosphatidylinositol 4,5-Bisphosphate-binding Site Located in the C2 Domain of Protein Kinase Cα*

In view of the interest shown in phosphatidylinositol 4,5-bisphosphate (PtdIns(4,5)P2) as a second messenger, we studied the activation of protein kinase Cα by this phosphoinositide. By using two double mutants from two different sites located in the C2 domain of protein kinase Cα, we have determined and characterized the PtdIns(4,5)P2-binding site in the protein, which was found to be important for its activation. Thus, there are two distinct sites in the C2 domain: the first, the lysine-rich cluster located in the β3- and β4-sheets and which activates the enzyme through direct binding of PtdIns(4,5)P2; and the second, the already well described site formed by the Ca2+-binding region, which also binds phosphatidylserine and a result of which the enzyme is activated. The results obtained in this work point to a sequential activation model, in which protein kinase Cα needs Ca2+ before the PtdIns(4,5)P2-dependent activation of the enzyme can occur.

In view of the interest shown in phosphatidylinositol 4,5-bisphosphate (PtdIns(4,5)P 2 ) as a second messenger, we studied the activation of protein kinase C␣ by this phosphoinositide. By using two double mutants from two different sites located in the C2 domain of protein kinase C␣, we have determined and characterized the PtdIns(4,5)P 2 -binding site in the protein, which was found to be important for its activation. Thus, there are two distinct sites in the C2 domain: the first, the lysinerich cluster located in the ␤3and ␤4-sheets and which activates the enzyme through direct binding of PtdIns(4,5)P 2 ; and the second, the already well described site formed by the Ca 2؉ -binding region, which also binds phosphatidylserine and a result of which the enzyme is activated. The results obtained in this work point to a sequential activation model, in which protein kinase C␣ needs Ca 2؉ before the PtdIns(4,5)P 2 -dependent activation of the enzyme can occur.
Phosphatidylinositol 4,5-bisphosphate (PtdIns(4,5)P 2 ) 1 plays a key role in phosphoinositide signaling and regulates a wide range of processes at many subcellular sites. It is primarily detected in the plasma membrane but is also found in secretory vesicles, lysosomes, in the endoplasmic reticulum, the Golgi, and in the nucleus (1)(2)(3)(4)(5). PtdIns(4,5)P 2 can either bind to intracellular proteins and directly modulate their subcellular localization and activity, or it can act as a precursor for the generation of different second messengers. For example, several families of phospholipase C enzymes are responsible for the hydrolysis of PtdIns(4,5)P 2 in cells, leading to the production of diacylglycerol and inositol 1,4,5-trisphosphate (4,6), which may, in turn, lead to the activation of different proteins such as some PKC isotypes.
Protein kinase C (PKC) composes a large family of serine/ threonine kinases, which is activated by many extracellular signals and plays a critical role in many signal-transducing pathways in the cell (7)(8)(9). Based on their enzymatic properties, the mammalian PKC isotypes have been grouped into smaller subfamilies. The first group, which includes the classical isoforms ␣, ␤I, ␤II, and ␥, can be distinguished from the other groups because its activity is regulated by diacylglycerol (DAG) and, cooperatively, by Ca 2ϩ and acidic phospholipids, particularly phosphatidylserine (PS). Members of the second group are the novel mammalian (␦, ⑀, , and ) and yeast PKCs that are not regulated by Ca 2ϩ . The third group comprises the atypical PKC isoforms, , , and , whose regulation has not been clearly established, although it is clear that they are not regulated by DAG or Ca 2ϩ (8,10).
In classical PKC isoenzymes, Ca 2ϩ -dependent binding to membranes shows a high specificity for 1,2-sn-phosphatidyl-Lserine (11)(12)(13)(14). Additionally, this group of isoenzymes is sensitive to other anionic phospholipids, including phosphatidic acid and polyphosphoinositides (15)(16) and to a variety of amphipathic membrane compounds, such as arachidonic acid and free fatty acids (17). Furthermore, in vitro experiments have demonstrated that the presence of other anionic phospholipids in the vesicles decreases the requirement of phosphatidylserine, suggesting that PKC activation, other than the classical activation pathway (activation of phospholipase C and production of diacylglycerol by hydrolysis of PtdIns(4,5)P 2 and increase of intracytosolic Ca 2ϩ ), could take place in vivo (15).
In view of the interest in PtdIns(4,5)P 2 as a second messenger, several studies (11, 18 -24) have addressed the activation of PKC isotypes by this class of lipid. However, the results show little consistency as to which of the different PKC isotypes are activated or as regards their specificity for the different lipids employed. These conflicting results are probably due to the different ways used by investigators to activate the enzyme.
When we look at the amino acidic sequence of the PKC␣-C2 domain, we observe that some of the Lys/Arg residues described for synaptotagmin are conserved. Previous studies in our laboratory (32) have suggested that the highly positive charged ␤3-␤4-sheets could interact electrostatically with the negatively charged phospholipids located at the membrane sur-face. Whether or not this is significant in the context of the full-length protein is still not clear.
In this paper, we focus on the characterization of the interaction mechanism between PKC␣-C2 domain and PtdIns(4,5)P 2 and the consequent enzyme activation. For this purpose, we cloned the PKC␣ C2 domain fused to glutathione S-transferase (GST) and full-length PKC␣ fused to a hemagglutinin (HA) tag, which enabled us to perform binding and specific activity studies. Site-directed mutagenesis of key residues located in two areas of the C2 domain shed light on the interaction of the classical isoenzyme with PtdIns(4,5)P 2 and its activation mechanism.

Construction of Expression Plasmids
Rat PKC␣ cDNA was a gift from Drs. Nishizuka and Ono (Kobe University, Kobe, Japan). The cDNA fragment corresponding to residues 158 -285 of the PKC␣-C2 domain and mutants was amplified using PCR (33). Full-length PKC␣ mutants were generated by PCR site-directed mutagenesis (14,34). All constructs, both wild-type and mutant genes, were subcloned into the mammalian expression vector pCGN (a gift from Dr. Tanaka, Cold Spring Harbor Laboratory, Cold Spring Harbor, NY). This vector contains the cytomegalovirus promoter and the multicloning sites that allow expression of the genes fused 3Ј to the HA epitope (35). All constructs were confirmed by DNA sequencing.
The cDNA fragment corresponding to residues 158 -285 of the PKC␣-C2 domain and mutants was amplified using PCR (33).

Expression and Purification of GST-PKC␣-C2
The pGEX-KG plasmid containing the PKC␣-C2 domain was transformed into HB101 Escherichia coli cells. Proteins were expressed and purified as described in a previous work (33).

Cell Culture, Transfection, and Purification of PKC␣
HEK293 cells were grown in Dulbecco's modified Eagle's medium with 10% fetal calf serum. Transfection was performed with the Ca 2ϩ phosphate method described by Wigler et al. (36). Protein purification was performed as described by Conesa-Zamora et al. (14).

Phospholipid Binding Measurements
Standard Assay-The procedure described by Davletov and Sudhof (37) was used with minor modifications. A total of 10 g of PKC␣-C2 domain bound to glutathione-Sepharose beads was used. Lipid vesicles were generated by mixing chloroform lipid solutions in the desired proportions and then dried from the organic solvent under a nitrogen stream and further dried under vacuum for 60 min. 1,2-Dipalmitoyl-L-3-phosphatidyl-N-[methyl-3 H]choline (PerkinElmer Life Sciences; specific activity 56 Ci/mmol) was included in the lipid mixture as a tracer, a concentration of ϳ3000 -6000 cpm/mg phospholipid. Dried phospholipids were resuspended in buffer containing 50 mM Hepes, pH 7.2, 0.1 M NaCl, and 0.5 mM EGTA by vigorous vortexing and subjected to direct probe sonication (four cycles of 1 min). Beads with protein bound to them were prewashed with the respective test solutions and resuspended in 50 l of the corresponding lipid solution. The mixture was incubated at room temperature for 15 min with vigorous shaking and then briefly centrifuged in a tabletop centrifuge. The beads were washed twice with 0.4 ml of the incubation buffer without liposomes. Liposome binding was then quantified by liquid scintillation counting of the beads. The data in the paper and figures are expressed in number of nmol of lipid bound/10 g of protein.
PKC␣ Membrane Binding Assay-PKC␣ was incubated with multilamellar vesicles (500 M total lipid in 0.15 ml) in a buffer containing 20 mM Tris/HCl, pH 7.5, 5 mM MgCl 2 , and 0.5 mM EGTA or 0.2 mM CaCl 2 at room temperature for 15 min. Vesicle-bound enzyme was separated from the free enzyme by centrifuging the mixture at 13,000 ϫ g for 30 min at 20°C. Aliquots from the supernatants and pellets were separated by SDS-PAGE (12.5% separating gel). The proteins were transferred onto a nitrocellulose membrane after electrophoresis. Immunoblot analysis of the epitope tag fused to the protein was performed by using anti-HA antibody 12CA5 and developed with chemiluminescence reagents (PerkinElmer Life Sciences). The proteins were analyzed by densitometry.

Kinase Activity Assay
The lipids used for the reaction were previously dried under a stream of N 2 , and the last traces of organic solvent were removed by keeping the samples under vacuum for 1 h. Lipids were suspended in 20 mM Tris-HCl, pH 7.5, 0.05 mM EGTA and vortexed vigorously to form multilamellar vesicles. A 20-l sample of the lipids was added to the reaction mixture (final volume, 150 l), which contained 20 mM Tris-HCl, pH 7.5, 0.2 mg/ml histone III-S, 20 M [␥-32 P]ATP (300,000 cpm/ nmol), 5 mM MgCl 2 , and 200 M CaCl 2 . The reaction was started by addition of 5 l of the PKC␣ purified from transfected HEK293 cells. After 10 or 30 min, the reaction was stopped with 1 ml of ice-cold 25% trichloroacetic acid and 1 ml of ice-cold 0.05% bovine serum albumin. After precipitation on ice for 30 min, the protein precipitate was collected on a 2.5-cm glass fiber filter (Sartorius) and washed with 10 ml of ice-cold 10% trichloroacetic acid. The amount of 32 P i incorporated into histone was measured by liquid scintillation counting. The linearity of the assay was confirmed from the time course of histone phosphorylation during 30 min. Additional control experiments were performed with mock cell lysates to estimate the endogenous PKC␣ and nonspecific activities, which represented less than 1% of the total enzyme activity measured.

Structural Models
The PDB identifiers for the experimentally determined C2 and ENTH domain structures used in the calculations were 1DSY (38) and 1HFA (39). The Swiss-PDB Viewer 3.7 program by GlaxoSmithKline R&D Geneva (57) was used to visualize the structures.

Characterization of the PtdIns(4,5)P 2 and PKC␣-C2 Domain
Interaction-Our first aim was to characterize the biochemical properties of the interaction between PtdIns(4,5)P 2 and the C2 domain of PKC␣. For this, a recombinant fusion protein was used, in which the C2 domain of PKC␣ was NH 2 -terminally fused to GST (PKC␣-C2 domain) (33). The dependence of phospholipid binding on the PtdIns(4,5)P 2 concentration was studied by using phospholipid vesicles containing POPC/POPS (4:1, mol/mol) in the presence and in the absence of 100 M CaCl 2 (Fig. 1A). The results show that in the presence of Ca 2ϩ , PtdIns(4,5)P 2 slightly inhibited the PKC␣-C2 domain phospholipid binding activity, which amounted to 8 nmol regardless of the PtdIns(4,5)P 2 concentration in the lipid vesicles. More interesting were the results obtained in the absence of Ca 2ϩ because, in this case, the PKC␣-C2 domain bound to a similar phospholipid vesicle composition in a PtdIns(4,5)P 2 -dependent manner, showing a maximal binding of 7 nmol and a [PIP 2 ]1 ⁄2 value of 1 mol % (Fig. 1A). These results suggest that the C2 domain of PKC␣ may interact with PtdIns(4,5)P 2 independently of Ca 2ϩ . To explore whether this mechanism is specific to PtdIns(4,5)P 2 or is related to the net negative charges present in the membrane, binding assays were performed by substituting PtdIns(4,5)P 2 as supplier of net negative charges to the phospholipid vesicles (POPC/POPS/PIP 2 , 75:20:5 mol/mol) for POPS (POPC/POPS, 70:30). It was observed that 7 nmol of lipid were bound in the former case and only 1.3 nmol in the latter case (Fig. 1A, inset), suggesting that there is a certain specificity for PtdIns(4,5)P 2 and that the increased binding observed under these conditions is not only dependent on the net negative charges present at the membrane vesicles.
In order to study whether Ca 2ϩ can affect the PKC␣-C2 domain binding affinity in the absence of POPS in the lipid vesicles, binding assays were performed with POPC and increasing PtdIns(4,5)P 2 concentrations in the absence and in the presence of 100 M CaCl 2 (Fig. 1B). In the absence of Ca 2ϩ , lipid binding reached a maximum of 7 nmol and [PIP 2 ]1 ⁄2 ϭ 1.2 mol %, whereas in the presence of Ca 2ϩ , binding reached 9 nmol and [PIP 2 ]1 ⁄2 ϭ 0.3 mol %. This demonstrates that in the absence of POPS, Ca 2ϩ slightly increases both the affinity of the protein for PtdIns(4,5)P 2 -containing vesicles and the maximal binding capacity.
Binding Mechanism of PtdIns(4,5)P 2 to PKC␣-C2 Domain-The data described above might be explained by the existence of two types of binding mechanism, one Ca 2ϩ /PS-dependent and the other mostly Ca 2ϩ -independent and PtdIns(4,5)P 2 -dependent. Whether these two mechanisms corresponded to two different sites in the domain or to just one site with different biochemical behaviors could still not be answered at this point. To address this question, we made use of two PKC␣-C2 domain mutants. One of them (PKC␣-C2-D246N/D248N) has been demonstrated to be affected in the Ca 2ϩ -binding site (33, 34) ( Fig. 2A) and does not bind Ca 2ϩ or PS. The second (PKC␣-C2-K209A/K211A) has been demonstrated to be affected in an area corresponding to a lysine-rich cluster located in the ␤3-␤4sheets ( Fig. 2C) and partially lacks its ability to bind PS or phosphatidic acid in the absence of Ca 2ϩ (32). Fig. 2B shows the results obtained when PKC␣-C2-D246N/ D248N mutant was employed in the binding assays using ves-icles containing POPC and increasing concentrations of PtdIns(4,5)P 2 in the absence of Ca 2ϩ . In this case, maximal binding activity was 8 nmol of lipid and [PIP 2 ]1 ⁄2 ϭ 0.5 mol %. These results are slightly higher than those obtained when the wild-type C2 domain was used in the assay (Figs. 1B and 2B, dotted line), suggesting that the substitution of these residues located at the Ca 2ϩ -binding site has no effect on the ability of the domain to bind to PtdIns(4,5)P 2 -containing vesicles in a Ca 2ϩ -independent manner. Furthermore, the results obtained suggest that the neutralization of the Ca 2ϩ -binding sites performed by the mutagenesis strategy probably helps to reduce the negative electrostatic potential exhibited by this area, in a similar way to the effect brought about by Ca 2ϩ when it binds to this region (40), thus facilitating the slight increase in the binding affinity observed in both cases.
When the second mutant, PKC␣-C2-K209A/K211A, was included in the binding assay under the same conditions as described above (Fig. 2D), only 1.3 nmol of lipids were bound to the protein, indicating that these two residues are directly involved in the Ca 2ϩ -independent and PtdIns(4,5)P 2 -dependent binding activity of the domain. Furthermore, when this experiment was performed in the presence of 100 M CaCl 2 , the PKC␣-C2-K209A/K211A mutant was able to bind only 5.3 nmol of lipids at high concentrations of PtdIns(4,5)P 2 , which amounted to only 58% of the lipid bound with wild-type protein under these conditions (Fig. 2D). In summary, these data support a double-site model for PtdIns(4,5)P 2 binding. The first and most important would be the lysine-rich cluster, which binds PtdIns(4,5)P 2 with no need for Ca 2ϩ , and the second would be located in the Ca 2ϩ binding region. However, this last site exhibits a relatively low affinity for PtdIns(4,5)P 2 even in the presence of Ca 2ϩ , suggesting that this inositide fits better in the lysine-rich region than in the Ca 2ϩ binding region.
Characterization of the PtdIns(4,5)P 2 -PKC␣ (Full-length) Interaction-As described in the Introduction, PKC␣ activation is regulated by multiple factors (10). Moreover, several of the domains included in the protein are involved in the full activation of the enzyme, which further complicates the situation. To investigate the role of this new PtdIns(4,5)P 2 -binding site in PKC␣ membrane translocation, the binding of the protein to POPC multilamellar vesicles was studied by including 5 mol % PtdIns(4,5)P 2 or 20 mol % POPS in the presence of Ca 2ϩ and, in some cases, using a phorbol ester (PMA) as a C1 domain-dependent activator (Fig. 3A). Both PtdIns(4,5)P 2 and POPS were able to produce 50% protein translocation to the membranes in the absence of PMA. This translocation increased to 75% when both phospholipids were included in the same vesicles (5 mol % PIP 2 and 20 mol % POPS). As expected, the inclusion of a saturating concentration of PMA (0.3 mol %) cooperated and increased the proportion of translocated protein to 100%, independently of the negatively charged phospholipids present in the membrane vesicles.
Strikingly, very different results were obtained when these experiments were performed in the absence of Ca 2ϩ (Fig. 3B). No PKC␣ translocation was detected in the absence of PMA. Furthermore, when 0.3 mol % PMA was included in the assay, only 10% of protein translocation was detected in vesicles containing either POPS or PtdIns(4,5)P 2 /POPS, suggesting that these binding sites were not functional or accessible to the membrane vesicles under these conditions. To test this further, a PKC␣ substrate (histone III-S) was included in the binding reaction. In this case, protein translocation was 51, 42, and 45% when PtdIns(4,5)P 2 , POPS, or POPS/PtdIns(4,5)P 2 , respectively, were included in the vesicles, indicating that the presence of substrate enhances the ability of the PtdIns(4,5)P 2 and PS sites to interact with the lipid vesicles. When PMA and histone were included in the model membranes, protein translocation to them was 100%, suggesting that under these conditions either PtdIns(4,5)P 2 or POPS with PMA access their corresponding sites (C2 and C1 domain, respectively), producing full translocation of the enzyme to the membrane vesicles.
PtdIns(4,5)P 2 -dependent Activation of PKC␣-To study whether protein translocation to the phospholipid vesicles correlates with its activation, PKC␣ specific activity was measured under conditions similar to that used in the binding assays. Fig. 4A shows the catalytic activity of the enzyme in the presence of 100 M CaCl 2 , 0.3 mol % PMA, and increasing concentrations of POPS or PtdIns(4,5)P 2 in the phospholipid vesicles. Under these conditions, 20 mol % PtdIns(4,5)P 2 increased enzyme activity more than 5 times that obtained when using vesicles containing 20 mol % POPS. These data correlate well with the binding assays, because 5 mol % PtdIns(4,5)P 2 exhibited a similar binding and activation capacity to 20 mol % POPS.
In the absence of Ca 2ϩ , vesicles containing PtdIns(4,5)P 2 increased the enzyme activity more than 6 times that obtained with POPS-containing vesicles, which also suggested an important role for the PtdIns(4,5)P 2 -binding site in enzyme activation under these conditions (Fig. 4B). These data correlated well with those obtained in the binding assay, because full translocation of the enzyme to the phospholipid vesicles occurred in the presence of histone and PMA.
Note also that the PtdIns(4,5)P 2 -dependent specific activity of PKC␣ in the presence of Ca 2ϩ is almost 3 times the specific activity obtained in the absence of Ca 2ϩ , indicating again that, directly or indirectly, Ca 2ϩ plays a role in the PtdIns(4,5)P 2dependent activation of the enzyme.
Demonstrating the Existence of the New PtdIns(4,5)P 2 Site in Full-length PKC␣-To address this question the same C2 domain mutants as above were used in the context of full-length PKC␣, namely PKC␣-D246N/D248N and PKC␣-K209A/K211A to measure the effect of increasing concentrations of PtdIns(4,5)P 2 and POPS on the specific enzyme activity. The results were compared with those obtained with wild-type protein. Fig. 5A shows the activation of PKC␣-D246N/D248N and PKC␣-K209A/K211A in the presence of 100 M CaCl 2 , 0.3 mol % PMA, and increasing concentrations of PtdIns(4,5)P 2 . It can be clearly observed that PKC␣-K209A/K211A activation represents only 34% and PKC␣-D246N/D248N 60% of the total activity exhibited by wild-type protein (Table I).
In contrast, when POPS was included in the lipid vesicles instead of PtdIns(4,5)P 2 (Fig. 5B), PKC␣-D246N/D248N activation was only 28%, whereas PKC␣-K209A/K211A activation was very similar to the maximum activity reached by wild-type protein under the same conditions. These data suggest that in the activation driven by PtdIns(4,5)P 2 , lysines 209 and 211 play a more important role than the aspartate residues located in the Ca 2ϩ -binding site. On the other hand, when both mutants were tested for activation in the absence of Ca 2ϩ (Fig. 6), no differences were found with wild-type PKC␣, whether vesicles containing PtdIns(4,5)P 2 or POPS were used. Nevertheless, the enzyme was increasingly activated with increasing concentrations of PtdIns(4,5)P 2 although the maximum activation in this case decreased 3-fold. This might be explained by the fact that in full-length protein there is also a partial contribution of the C1 domain to the PtdIns(4,5)P 2 activation. It is important to note that the maximal activation under these conditions for wild-type protein only represents 30% of the maximal enzyme activation obtained in the presence of Ca 2ϩ , suggesting that the greatest contribution to PtdIns(4,5)P 2 -dependent activation of the enzyme resides in the C2 domain.
It is also interesting to note that the maximal specific activity obtained when the PKC␣-K209A/K211A mutant was assayed in the presence and in the absence of Ca 2ϩ was 279 and 235 nmol P i /min/mg, respectively (Table I), representing small differences in the PMA-dependent activation of the mutant.

PtdIns(4,5)P 2 -dependent binding of the PKC␣-C2 domain mutants.
A, three-dimensional model of the PKC␣-C2 domain showing the location of the two aspartate residues mutated to asparagine (D246 and D248) in the Ca 2ϩ binding region. The two Ca 2ϩ ions are shown in orange. The structure corresponds to the PDB 1DSY described by Verdaguer et al. (38). B, binding of PKC␣-C2D246N/D248N domain to vesicles containing increasing concentrations of PtdIns(4,5)P 2 in the absence of Ca 2ϩ (q). The dotted line represents the results obtained for wild-type C2 domain in Fig. 1B. Importantly, the Ca 2ϩ /PtdIns(4,5)P 2 -dependent activation was completely abolished in this case, suggesting once again that the lysine-rich cluster located in the C2 domain plays a very important role in the activation of the enzyme under these conditions. Strikingly, when the PKC␣-D246N/D248N mutant was assayed, 240 and 485 P i /min/mg were obtained in the absence and in the presence of Ca 2ϩ , respectively. Note that, even in the presence of Ca 2ϩ , this mutant did not recover full activation although the lysine-rich cluster remained intact, suggesting again that Ca 2ϩ access to the Ca 2ϩ binding region is probably a key event in the PtdIns(4,5)P 2 -dependent activation process.
Another way in which we attempted to distinguish between the contributions of the PMA and Ca 2ϩ to the full activation of the enzyme was by measuring the specific activity of the wildtype protein and of the two mutants in the presence of Ca 2ϩ and in the absence of PMA. Under such conditions, the activation of PKC␣ was 1.3-fold less than that obtained in the presence of PMA, suggesting that the main contribution for this type of activation is supplied by the C2 domain. Furthermore, when increasing concentrations of PtdIns(4,5)P 2 were used, PKC␣-D246N/D248N mutant did not activate more than 45% of the total activity exhibited by wild-type protein ( Fig. 7 and Table I). The activation capacity of PKC␣-K209A/K211A mutant was even more limited under these conditions (only 18% of total activation) suggesting that this site is primarily involved in the PtdIns(4,5)P 2 -dependent activation process.
Parallel binding experiments were performed under the same conditions used to measure the specific activity of the enzyme. Model membranes containing POPC and 10 mol % PtdIns(4,5)P 2 were generated, and vesicle membrane binding was measured in the presence and in the absence of 0.3 mol % PMA. In the case of POPC membranes, only 10% of wild-type and mutants were bound to the vesicles (Fig. 8). In a similar way to what happened in the activation assays, when PtdIns(4,5)P 2 was included in the membranes, 78% of wildtype PKC␣, 45% of PKC␣-D246N/D248N, and 18% of PKC␣-K209A/K211A bound to them. When PMA was included in the membrane vesicles, up to 100% of wild-type protein was bound, and only 70 and 28% of PKC␣-D246N/D248N and PKC␣-K209A/K211A, respectively, were bound under the same conditions. These results correlate very well with the specific activity assays, suggesting that the lack of activity in each case corresponds to a lack of membrane binding.
Taken together, these results suggest that the lysine-rich cluster of the C2 domain is the most important site in the PtdIns(4,5)P 2 -dependent activation of PKC␣ and, although contributions of the C2 and C1 domains are necessary for full activation of the enzyme, the C2 domain plays a more specific role in this activation mechanism.
Specificity of Phosphoinositides for PKC␣ Activation-10 and 20 mol % of PtdIns(3)P and PtdIns(3,4,5)P 3 were included in multilamellar vesicles containing POPC and 0.3 mol % PMA, and the specific activity of PKC␣ was measured in the presence of 100 M CaCl 2 . The effect of each phosphoinositide on specific activity was compared with the activity observed in the presence of 10 and 20 mol % PtdIns(4,5)P 2 , respectively (Fig. 9).
Under these conditions, PtdIns(3,4,5)P 3 activated only 57% of the total activity displayed by PtdIns(4,5)P 2 suggesting that the latter is more efficient in activating the protein, probably because it docks better to the binding site, whereas some degree of steric impedance makes PtdIns(3,4,5)P 3 less capable of activating the enzyme. In the case of PtdIns(3)P, the maximum activation reached was only 25% that obtained with PtdIns(4,5)P 2 , demonstrating that PKC␣ exhibits a higher specificity for PtdIns(4,5)P 2 than for other phosphoinositides. DISCUSSION Classically, PtdIns(4,5)P 2 has been defined as the precursor of the mediators, diacylglycerol and inositol (1,4,5)P 3 , after hydrolysis by hormone-sensitive phospholipase C enzymes (41). However, new experiments are now revealing another signaling mode, which is controlled by intact PtdIns(4,5)P 2 rather than its hydrolysis products. For example, this phosphoinositide has been implicated in the control of membrane dynamics, cell shape, and cytoskeleton rearrangement (42)(43). Thus, PtdIns(4,5)P 2 has emerged as a highly versatile signaling molecule in its own right, although it remains to be clarified  how the functions of the proteins that interact with these lipids are coordinated.
In this work, we have defined a new PtdIns(4,5)P 2 -binding site located in a crevice of the C2 domain of PKC␣ formed by the ␤3and ␤4-sheets that is different from the Ca 2ϩ binding region where Ca 2ϩ and PS interact (38,32). Furthermore, a different activation mechanism of the enzyme, specifically triggered by PtdIns(4,5)P 2 , has been defined.
Experiments performed with the isolated domain revealed that in the absence of Ca 2ϩ this site is more specific for PtdIns(4,5)P 2 than other negatively charged phospholipids. Thus, two major and different phospholipid-binding sites can be defined in the isolated domain as follows: first, the Ca 2ϩ binding region located on top of the domain, which binds Ca 2ϩ and phosphatidylserine and probably PtdIns(4,5)P 2 but with relatively low affinity; and second located in the crevice formed by ␤3and ␤4-sheets, which is highly enriched in lysine residues and which binds PtdIns(4,5)P 2 in a Ca 2ϩ -independent manner. Recent structural studies in our laboratory have shown that this lysine-rich cluster can bind negatively charged phospholipids, such as diacetoyl-sn-glycero-3-phosphoserine and dicaproyl-sn-glycero-3-phosphatidic acid, mainly through electrostatic interactions (32). Strikingly, binding inhibitions of about 50 -70% were only found when the experiments were performed in the absence of Ca 2ϩ , suggesting that these residues might somehow be involved in the membrane translocation of the protein independently of Ca 2ϩ . The results obtained in the present work clearly explain the preliminary data obtained, because the lysine-rich cluster has higher affinity for PtdIns(4,5)P 2 than PS, and this interaction is Ca 2ϩ -independent.
However, the scenario changed when binding and activity assays were performed with full-length protein, in which case Ca 2ϩ and PMA were also necessary to obtain full activation of the enzyme when PtdIns(4,5)P 2 was used as activator. The binding experiments carried out in the absence of Ca 2ϩ showed that the protein needs to be in a particular conformation/state to be accessible to negatively charged phospholipids and PMA. This agrees with previous reports that have described how, in the absence of Ca 2ϩ , the pseudosubstrate is clamped to the catalytic domain, and the binding of PS/Ca 2ϩ and diacylglycerol is necessary to promote the liberation of the pseudosubstrate, leading to an "open conformation" and full activation of the enzyme (10,44) The use of PKC␣-D246N/D248N and PKC␣-K209A/K211A mutants has enabled us to distinguish between the two different sites and to discriminate their role in the PtdIns(4,5)P 2dependent activation. In the absence of PMA, the catalytic activity of the PKC␣-K209A/K211A mutant was completely abolished, and by taking into account that the Ca 2ϩ binding region was intact in this case, it is clear that Lys-209 and Lys-211 are key residues in the PtdIns(4,5)P 2 -driven activation process. However, the experiments performed with the PKC␣-D246N/D248N mutant also revealed that some contribution from this site is needed to obtain full activation of the enzyme because the lysine-rich cluster is intact in this case and the catalytic activity recovers to reach only 45% (Fig. 7). Note that this mutant (Asp-246 substituted by Asn) has been described as the most important residue for coordinating the second Ca 2ϩ bound to the domain and probably for producing a conformational change in the protein (38,46). Thus, the results obtained strongly suggest that Ca 2ϩ binding to its own region is a preliminary requirement for the PtdIns(4,5)P 2 interaction with the lysine-rich cluster in the C2 domain.
Further support for a double and distinct phospholipid-binding site in the C2 domain was obtained when the mutants were activated with POPS-or PtdIns(4,5)P 2 -containing vesicles and compared with the wild-type protein (Fig. 5). The D246N/ D248N substitution was more critical in the PS-dependent activation, whereas the K209A/K211A substitution was more critical in the PtdIns(4,5)P 2 -dependent activation.
Concerning the role of the C1 domain in the PtdIns(4,5)P 2dependent activation of the enzyme, the above experiments showed that this domain makes a small contribution to full binding and activation of the enzyme, which would take place in a third step in the sequential activation mechanism. Recent structure-function studies have demonstrated that several positively charged residues located on the domain surface interact non-specifically with anionic phospholipids prior to membrane penetration of hydrophobic residues (47), and this could be the reason for the additional 25-30% binding and activation obtained in a PMA-dependent process.
It is also interesting to note that PtdIns(4,5)P 2 is able to activate the enzyme to a greater extent than POPS, which has always been defined as the main activator of PKC␣. These findings have to be considered carefully because the exact lateral organization and effective concentrations of these phospholipids in the plasma membrane are still not well defined, especially in the case of PtdIns(4,5)P 2 , which several hypotheses suggest may accumulate in different metabolic pools existing in the cells, for example in rafts, nascent phagosomes of macrophages, or membrane ruffles (see Ref. 31 for a review).
In relation to phosphoinositide specificity, it seems that PtdIns(4,5)P 2 is the best of the activators studied in this work, probably because the docking site in full-length PKC␣ adjusts better to this inositol phosphate than to others. The lesser degree of activation produced by PtdIns(3,4,5)P 3 could be explained by steric clashes in the area that cannot accept three phosphates. Similar results have been described recently in the literature, and it is clear that certain regions are more suitable for binding PtdIns(4,5)P 2 than PtdIns(3,4,5)P 3 , for example a new PtdIns(4,5)P 2 binding region described in plant phospholipase D␤ (48), the ENTH domains of epsin (49), and clathrin assembly lymphoid myeloid leukemia protein (CALM) (39).
It is important to note that although there are no sequence or structural homologies between the ENTH and C2 domains, they share important functional traits. Basically, the ENTH domains consist of nine ␣-helices forming a solenoid structure (39,49), whereas C2 domains consist of an eight-stranded ␤-sandwich (50). However, recent studies (39) on the ENTH domain of CALM have shown an unusual PtdIns(4,5)P 2 -binding site, which is located in an exposed cluster of lysines and appears to be different from the PtdIns(4,5)P 2 -binding sites described previously for epsin, another ENTH domain-containing protein. Interestingly, when the tri-dimensional distribution of the side chains of the lysines involved in PtdIns(4,5)P 2 binding in the ENTH domain of CALM was compared with the distribution of the side chains of the lysines involved in the PtdIns(4,5)P 2 -binding site in the C2 domain of PKC␣, they exhibited a very high degree of similarity (Fig. 10, a and b). Additionally, when a simulation was performed using Swiss PDB viewer 3.7, the PtdIns(4,5)P 2 molecule found in CALM was seen to fit in the lysine-rich cluster of the C2 domain, and the distances observed between the lysine residues and the 4and 5-phosphates of the inositol ring were compatible with hydrogen bond interactions like those observed in the ENTH domain of CALM (Fig. 10c).
Biological Implications-Recent studies (4, 41) have indicated that there are many sources of PtdIns(4,5)P 2 in the cell, which appear to involve different enzymes. The recent development of techniques that permit direct visualization of PtdIns(4,5)P 2 and PtdIns(3,4,5)P 3 has provided important in- FIG. 11. Schematic model of the PtdIns(4,5)P 2 -dependent activation mechanism of PKC␣. a represents a state in the absence of Ca 2ϩ , similar to that obtained in resting cells. In this case, the protein adopts a closed conformation where the PtdIns(4,5)P 2 site is not accessible to the enzyme. b shows an intermediate state occurring when the Ca 2ϩ concentration increases in the cytosol or when high Ca 2ϩ concentrations are used in an in vitro model system. Under these conditions, the protein adopts an open conformation state, and if PtdIns(4,5)P 2 is present in the membrane, it binds to the lysine-rich cluster (LRC) located in the ␤3-, and ␤4-sheets. c represents a fully active state when the C1 domain can access the DAG generated in the membrane, which increases the catalytic activity of the enzyme through a more stable anchorage to the plasma membrane.
formation on the spatio-temporal localization and function of these compounds in living cells (41,43). By taking into account that the total level of PtdIns(4,5)P 2 in a cell does not vary in response to agonist stimulation, it has been postulated that local increases of PtdIns(4,5)P 2 at discrete sites, such as caveolae, nuclei, membrane ruffles, and focal contacts, probably mediate the diverse signaling functions of this phosphoinositide. However, the mechanism that could concentrate PtdIns(4,5)P 2 in the plane of the plasma membrane is still under debate (3,5,31).
At the same time, it has been described that PKC can colocalize in many of these membrane domains, and based on the results obtained in this work, it could be postulated that PKC-PtdIns(4,5)P 2 interaction could be an alternative pathway for translocating and activating PKC to these particular areas or compartments of the cell. For example, it has been demonstrated recently (24,51,52) that the formation of a ternary complex of PtdIns(4,5)P 2 , sydecan-4, and PKC␣ could be the key event in the regulation of focal adhesion and stress fiber formation. There is another nice example that shows that PKC␣ delivery to the endosome is a caveolae-mediated process (54), and it is possible that PtdIns(4,5)P 2 might constitute an anchoring point for PKC here. Similarly, it has been demonstrated that PKC needs to translocate to membrane ruffles in order to phosphorylate substrate protein such as ␣ 6 ␤ 4 integrin and myristoylated alanine-rich C kinase substrate among others (55,56). Additionally, many important lipid signaling pathways also occur within the nucleus, the first components identified being PtdIns(4,5)P 2 and its precursors (45). Several isoforms of PKC are found in the nucleus or, at least, can translocate there. For example, PKC␣ translocates to the nucleus and perinuclear region of NIH-3T3 cells following PMA treatment (53). This localization could be compatible with a different, as yet undefined, mechanism of enzyme activation.
In summary, the present work clearly defines a new site in the C2 domain of PKC␣ totally distinct from the PS/Ca 2ϩ binding region (Fig. 10), located in the lysine-rich cluster, where specifically binds PtdIns(4,5)P 2 , leading to the enzyme activation. A sequential mechanism for this particular PtdIns(4,5)P 2 -dependent activation would fit the puzzle posed by the results: in the absence of Ca 2ϩ , the protein is in a "closed conformation" and neither PtdIns(4,5)P 2 nor DAG can access the corresponding sites. However, Ca 2ϩ , when present, binds to the Ca 2ϩ binding region of the C2 domain, presumably leading to a conformational change in the full-length protein that now enables PtdIns(4,5)P 2 to access the lysine-rich cluster and activate the enzyme (Fig. 11).
In general, differential engagement by the two lipid binding pockets of the PKC␣-C2 domain could form part of distinct signaling complexes that may result in the selective activation, inhibition, or translocation to unique subcellular compartments, thus converting the enzyme into a multifunctional kinase.