The Regulatory Domain of Protein Kinase C Coordinates Four Atoms of Zinc*

Protein kinase C (PKC) was found to be a zinc me-tallo-enzyme. Atomic absorption measurements on the intact enzyme indicated that four zinc atoms (4.2 f 0.6) were bound per PKC a molecule. Similar stoichiometric ratios were determined for PKC BII and PKC y, other PKC isoforms individually expressed in the baculovirus-insect cell expression system, as well as for purified rat brain PKC. By trypsin treatment of PKC a, a 32-kDa lipid binding regulatory and a 60-kDa catalytic domain were generated that were sub- sequently completely separated by gel filtration in the presence of Triton X- lOO/phosphatidylserine mixed micelles. Zinc was present at levels significantly above background in fractions that contained the 32-kDa fragment and displayed phorbol ester binding activity. Lipid association and phorbol ester binding did not lead to displacement of zinc from the protein. The stoichiometry determined for this fragment (4.7 2 0.9) suggested that zinc was bound exclusively within the lipid binding regulatory domain of intact PKC. Fur- thermore, this stoichiometry is consistent with zinc being coordinated between 6 cysteine residues in a structural motif related to the Zn(II)2Cyss binuclear cluster identified in the GAL4 transcriptional factor (Pan, T., and (1990) Proc. Nutl. U. S. A. 87,2077-2081). staining kit from Bio-Rad. Immunological Methods-Anti-peptide antibodies against PKC a amino acid residues. 19-36 (RFARFGALRQKNVHEVKN) and PKC y residues 102-115 (HKFRLHSYSSPTFC) were raised in rabbits as previously described (24). At dilutions up to 1:10,000 the antisera recognized both their specific peptides and intact PKC? The char- acterization of these antibodies will be detailed elsewhere. A third commercially available antipeptide antibody (GIBCO-Bethesda Re- search Laboratories) against PKC CY amino acid residues 313-326 (AGNKVISPSEDRRQ) was also employed to characterize PKC frag- ments. Electrophoretically separated proteins were transferred to nitrocellulose (25) and probed with these peptide-specific polyclonal antibodies followed by horseradish peroxidase-conjugated goat anti-rabbit IgG second antibody and polypeptide bands visualized by peroxidase staining using 4-chloro-1-naphthol as a substrate.

. Within this region two cysteine-rich repeats exist (Cys 1 and Cys 2) that resemble cysteine-rich DNA-binding zinc finger motifs of transcription factors (9). Each of these regions binds PDBu (Kd 20-60 nM; lo), albeit with substantially lower affinity than intact PKC or the 32-kDa lipid binding regulatory domain (& 2-5 nM; 6). For the entire c1 region (8), as well as individually expressed Cys 1 and Cys 2 regions? PDBu binding was not calcium-dependent, as expected in the absence of the PKC domain that is thought to confer calciumdependent lipid-protein interactions, called the C2 region (11,12).
Until recently PKC was the only known PDBu binding protein (13). n-Chimaerin, a neuron-specific protein containing a single cysteine-rich motif possessing 50% sequence identity with corresponding sequences in the C1 region of PKC (14), was also shown to bind PDBu ( K d 29 k 7 nM; Ref. 13); like Cys 1 and Cys 2, PDBu binding was not calciumdependent.
In both n-Chimaerin sild PKC the cysteine-rich region is essential for phospholipid-dependent PDBu binding. Sitedirected mutagenesis within this region of PKC revealed the crucial role of cysteine residues for this function (8). Very recent experiments employing PKC and n-Chimaerin glutathione S-transferase and TrpE fusion proteins expressed in Escherichia coli, provided qualitative evidence for the ability of these cysteine-rich domains to bind radioactive zinc (15). The results presented here provide the first quantitative measurements demonstrating that PKC contains four atoms of tightly bound zinc within the regulatory domain. Such a structural organization has significant implications in the formation of the PDBu binding site and concerning the mechanism of PKC regulation by DAG and PDBu.
Purification of Protein Kinase C-Protein kinase C was purified from rat brain or infected Sf9 cells as described previously (16 Protein Kinase C Is a Zinc Metallo-enzyme purification the enzyme was stored at -70 "C in storage buffer (20 mM Tris-HC1 pH 7.5, 0.5 mM EGTA, 0.5 mM EDTA, 10 mM pmercaptoethanol, 10% glycerol, and 0.05% Triton X-100). Rat brain PKC was purified following a similar procedure as described (6). To obtain enzyme of the quality available for the individual isoforms rat brain PKC from the FPLC HR10/10 phenyl-Superose column (Pharmacia LKB Biotechnology Inc.) was additionally purified on an FPLC HR5/5 Mono Q column (Pharmacia LKB Biotechnology Inc.). Rat brain PKC bound to the Mono Q column equilibrated in storage buffer and eluted as a pure PKC protein peak at 180 mM NaCl in a 34-ml linear salt gradient from 0 to 450 mM NaCl.
Sephadex G-25 Gel Filtration Analysis-To measure the zinc stoichiometry for PKC molecules, purified enzyme (5-10 pg in 100 pl) was analyzed on an FPLC HR5/5 column packed with about 1 ml of swollen Sephadex G-25, equilibrated in Buffer A (20 mM Tris-HCI, pH 7.5, 0.5 mM EGTA, 0.5 mM EDTA, 10 mM 0-mercaptoethanol, 0.03% Triton X-100, 10% ethylene glycol, and 50 mM NaCl) at a flow rate of 1 ml/min. Fractions of 0.2 ml volume were collected. Zinc-free buffers were obtained by extraction with dithizone in chloroform (17), and zinc was determined by atomic absorption spectroscopy. Column void volume and inclusion volume were determined in calibration runs with Triton X-100 and tritiated water, respectively. PKC eluting in void volume fraction 5 was routinely characterized for zinc stoichiometry measurements (Fig. 2B).
Trypsin Digestion of PKC-Purified PKC CY was digested with trypsin essentially as described previously (6) with some modifications. Aliquots of a trypsin stock solution (50 units/ml) in 50 mM Tris-HC1, pH 8.2, with 10% ethylene glycol stored frozen at -70 "C were thawed and allowed to stand at room temperature for 30 min. PKC CY (75 pg) in storage buffer (300 pl) was routinely digested at room temperature with trypsin (0.5 units) in a final volume of 600 pl of storage buffer. Digestion was stopped after 20 min by addition of phenylmethylsulfonyl fluoride dissolved in dimethyl sulfoxide to a final concentration of 1 mM. Digested samples were stored frozen at -70 ' C until further processed.
Separation of Mixed Micelle-bound PKC Regulatory Domain from the Catalytic Domain-Frozen aliquots (600 ~1 ) of trypsin-digested PKC were adjusted to a final concentration of 200 p M calcium, 0.3% (w/v) Triton X-100 mixed micelles containing 20 mol% PS in a final sample volume of 1 ml, and applied to an Ultrogel AcA 44 column (1.7 X 26 cm) equilibrated with 20 mM Tris-HC1 buffer, pH 7.5, containing 10 mM @-mercaptoethanol, 200 p M CaClz , 20 mM NaCl, 0.02% (w/v) Triton X-100, and 10% (w/v) glycerol. Buffers were prepared in plastic containers using the highest quality water to reduce zinc levels. Flow rates were typically 0.22 ml/min. Column fractions (1 ml) were characterized by measuring zinc content, kinase activity, and PDBu binding activity. As a control for background zinc in these experiments, mixed micelles together with trypsin in the absence of PKC were analyzed in the same fashion. In some experiments, as the one indicated in Fig. 38, trypsin-digested PKC a was allowed to associate with mixed micelles as described above, but in the presence of 200 nM [3H]PDBu. Control experiments to determine background zinc were done as described above including PDBu but without PKC (Fig. 3A). In both types of experiments column recovery was greater than 80%.
Determination of PKC Activity-Protein kinase C activity was determined using histone 111-S as a substrate (18). A standard kinase activity assay contained 0.3% Triton X-100/phospholipid mixed micelles (10 mol% PS, 2 mol% dioctanoyl glycerol), 20 p M ATP, 5 mM MgClz , 200 p~ CaClZ, and 200 pg/ml histone 111-S unless otherwise stated. A unit of PKC activity is defined as the incorporation of 1 pmol of phosphate into histone 111-S/min. PHIPDBu-binding Assay-Binding of [3H]PDBu was measured by the vesicular liposome assay (19). The 100 pl of assay mix contained 40 pg/ml PS, 2 mM CaClZ, 20 mM HEPES, pH 7.4, and 125 nM [3H] PDBu. Unspecific binding was determined in the presence of 40 p M unlabeled PDBu. After 10 min at room temperature samples were stored on ice until applied to GF/C Whatman filters, that subsequently were washed with 5 X 1 ml of ice-cold buffer (20 mM Tris, pH 7.5,200 p M CaClZ in 20% methanol) and then mixed with 5 ml of scintillation fluid and counted in a LKB liquid scintillation counter.
Zinc Measurements-Zinc was determined quantitatively by atomic absorption spectroscopy (17) using a Perkin-Elmer Cetus Zeeman 3030 atomic absorption spectrophotometer. Standard curves for the analysis were generated in an automated procedure using a zinc standard solution diluted into the linear assay range (0.5-2.5 ng of zinc/ml) in 0.2% HNO,. For stoichiometry determinations, Sephadex G-25 column void volume fractions (Fig. 2B), containing bound zinc, were diluted 20 to 60-fold in 0.2% HNOa before analysis. Fractions from the Ultrogel AcA 44 gel filtration column (Fig. 3) were diluted 50-fold. Typically, zinc values for stoichiometry represent the average of four measurements with standard error. Background zinc present in the column buffers, determined either in fractions beyond the Sephadex G-25 column inclusion volume (fractions 17-19, Fig.  2B) or in parallel runs in the absence of protein on the Ultrogel AcA 44 column (Fig. 3A), were subtracted from sample values before calculating zinc stoichiometry. For such calculations a molecular mass of 80 kDa for intact PKC, 32 kDa for the lipid binding fragment, and an atomic mass of 65 for zinc were assumed. The error indicated for zinc stoichiometries is equivalent to the standard error calculated for the values averaged.
Protein Quantitation-Protein concentrations were routinely determined using the Amidoschwarz dye binding assay (20) with fatty acid-free bovine serum albumin (Sigma) as the standard. Other methods were used for comparison purposes. The Micro BCA" (Pierce Chemical Co.) determination was done following the specifications of the manufacturer using their own bovine serum albumin standard. For the determination of protein concentrations of pure PKC a by quantitative amino acid analysis, the purified enzyme was dialyzed against 10 mM ammonium bicarbonate after the final purification step. Prior to total hydrolysis in 6 M HCl for 24 h at 110 "C, the enzyme preparation was extensively lyophilized. The hydrolyzed sample was dried in a SpeedVac concentrator and then analyzed in a Beckman model 6300 amino acid analyzer using the sodium citrate buffers provided by the manufacturer. To determine the protein concentration amino acid analysis results of 9 different amino acids were compared with the cDNA-derived (21)  Gel Electrophoresis-Routinely, protein samples precipitated prior to electrophoresis in chloroform/methanol (22) were analyzed by SDS-PAGE (23) on 10% minigels (Bio-Rad) and protein bands visualized by staining with Coomassie Blue (Sigma) and when necessary subsequently silver-stained with a commercially available staining kit from Bio-Rad.
Immunological Methods-Anti-peptide antibodies against PKC a amino acid residues. 19-36 (RFARFGALRQKNVHEVKN) and PKC y residues 102-115 (HKFRLHSYSSPTFC) were raised in rabbits as previously described (24). At dilutions up to 1:10,000 the antisera recognized both their specific peptides and intact PKC? The characterization of these antibodies will be detailed elsewhere. A third commercially available antipeptide antibody (GIBCO-Bethesda Research Laboratories) against PKC CY amino acid residues 313-326 (AGNKVISPSEDRRQ) was also employed to characterize PKC fragments. Electrophoretically separated proteins were transferred to nitrocellulose (25) and probed with these peptide-specific polyclonal antibodies followed by horseradish peroxidase-conjugated goat antirabbit IgG second antibody and polypeptide bands visualized by peroxidase staining using 4-chloro-1-naphthol as a substrate.

RESULTS AND DISCUSSION
PKC isoforms were expressed in Sf9 insect cells using the baculovirus expression system and subsequently purified (16). PKC is essentially pure in peak fractions obtained from an FPLC HR10/10 phenyl-Superose column. To examine whether PKC may have bound zinc, fractions from this column were characterized by histone 111-S phosphorylation activity and by atomic absorption spectroscopy (Fig. 1). For all three individually expressed isoforms (PKC a, BII, and y) as well as PKC purified from rat brain extracts, the peak of kinase activity coincided with elevated levels of zinc. Results obtained for PKC a are shown as a representative example. In the following experiments we focussed on the PKC a isoform that was available in larger quantities than other individual isoforms.
To verify the initial observation, purified PKC a (Fig. 2C,  lanes 1 and 2 )  calibration of a 1-ml Sephadex G-25 gel filtration column with Triton X-100 micelles (void volume) and tritiated water (inclusion volume) is illustrated ( A ) . Absorbance at 280 nm (0) and counts/min/fraction (0) are shown. In B a representative experiment with purified PKC a is shown. Kinase activity (0) co-eluted in void volume fractions with a major peak of zinc (0). However, zinc was also present above background levels in the column inclusion volume where a similar peak was observed when buffer alone was analyzed (not shown). Protein concentrations of PKC a determined in peak fractions, together with the illustrated zinc measurements provided the basis for stoichiometry determinations. A typical example of PKC a protein (0.5 pg) characterized in this fashion is shown in C after staining with Coomassie Blue (lune 1) and subsequently silver staining ( l u n e 2). present in PKC a (Fig. 2). Void and inclusion volume of the column were determined in trial runs with Triton X-100 and tritiated water respectively ( Fig. 2A). PKC a eluted with maximal kinase activity in fraction five corresponding to the column void volume (Fig. 2B). As observed previously with PKC eluting from the phenyl-Superose column (Fig. l), a peak of zinc was measured in fractions displaying PKC activity. In this experiment, however, protein-bound (fractions 5 and 6) and free (fractions [7][8][9][10][11][12] zinc were separated, thus allowing the determination of zinc bound to PKC. To determine the stoichiometric ratio between PKC and zinc, a simple, accurate, and sensitive method for determining protein at the low concentrations of protein present in such peak fractions was essential. The Amidoschwarz dye binding assay (20), with bovine serum albumin as a standard, was found to be the most reliable method. The protein concentration obtained by this method with a standard PKC a preparation (91 k 8 pg/ml) was compared with those measured by two independent methods described in greater detail under "Experimental Procedures.'' Both methods gave values for the protein concentration that lay within 20% of the value measured by the Amidoschwarz dye binding assay: Micro BCA (74.5 k 6 pg/ml) and quantitative amino acid analysis (81 & 5 pg/ml). Thus, in the experiments described all stoichiometries were calculated based on protein determinations by the Amidoschwarz dye binding assay. For PKC a an average ratio of 4.2 k 0.5 zinc atoms were found per molecule (Table I). Following the same procedure similar stoichiometries of zinc were found in PKC PII, PKC y, and rat brain PKC preparations (Table I).
Protein kinase C contains two cysteine-rich regions (Cys 1 and Cys 2) within the C1 domain. Both the Cys 1 and Cys 2 regions independently exhibit phorbol ester binding activity (10) and show homology to cysteine-rich regions present in transcription factors. Based on sequence comparisons to DNA binding motifs, two possible structures have been inferred for the cysteine-rich phorbol ester binding regions of PKC (lo), a Cys4 "zinc finger" domain similar to those of the glucocorticoid receptor (26) and a Zn(II)&ys6 binuclear cluster as identified in the GAL4 transcriptional factor (27). If a Cys4 zinc finger-like structure is involved in coordinating zinc, a stoichioinetry of two per PKC molecule should result (one zinc per finger); alternatively, if a structure similar to a zinc binuclear cluster is involved, the expected stoichiometry should be four (two zinc per cluster). The results presented here favor the later possibility.
If indeed the zinc is coordinated in the suggested fashion, the lipid binding domain of PKC should contain all the zinc bound to intact PKC. To test this prediction, PKC a was TABLE I Summary of zinc stoichiometries determined for different PKC isozymes and domains The experimental data shown summarize the zinc stoichiometries (moles of zinc/mol of PKC) determined for PKC purified from rat brain and individual isozymes from baculovirus-infected Sf9 cells (PKC a, PII, y) by gel filtration analysis with a 1-ml Sephadex G-25 column. Stoichiometry values for the regulatory domain of PKC a (a regulatory) were measured after association with mixed micelles (Triton X-lOO/PS) and separation on an Ultrogel AcA 44. Zinc concentrations were also measured in fractions from this column containing the catalytic domain (a catalytic). The results were obtained with a number of different enzyme preparations: rat brain PKC, 2; PKC a, 4; PKC PII, 1; PKC y, 1; PKC a regulatory or catalytic domain, 2. All PKC preparations characterized possessed kinase and [3H]PDBu binding activities comparable with those described previously (16). [3H]PDBu binding activity of the isolated PKC a lipid binding domain was as reported (6). For each stoichiometry determination sample concentrations of zinc were measured four times and protein concentrations were measured in duplicate. The errors indicated for zinc stoichiometries are equivalent to the standard error calculated for the averaged stoichiometry determinations, except for the PI1 isozyme where the standard error of the respective measurements is indicated in brackets. digested with trypsin to generate regulatory and catalytic subunit fragments and analyzed by gel filtration as described previously (6). Under the conditions chosen for PKC a digestion with trypsin, 65-90% of initial enzyme activity, quantitated both as kinase and phorbol ester binding activity, were typically maintained. Some 75-100% of residual kinase activity became lipid independent, indicating essentially complete cleavage of the PKC (data not shown). This was corroborated by gel analysis (Fig. 4A, lane 2). Two major breakdown products were routinely observed, migrating as expected at M , values of 45,000-50,000 and 32,000; these most likely correspond to catalytic and lipid binding domains, respectively (6). Trypsin-treated PKC a incubated with mixed micelles containing Triton X-100 and PS in the presence of [3H]PDBu was analyzed by gel filtration in zinc-free buffers (Fig. 3). A typical elution profile showing zinc background levels measured in trial runs with mixed micelles, trypsin, and ["HIPDBu is illustrated (Fig. 3A). When trypsin-digested PKC a was analyzed a peak of zinc was consistently found associated with fractions containing mixed micelles (Fig. 3B). Otherwise, zinc levels significantly above background were only detected in column inclusion volume fractions. Column fractions were also characterized in terms of their phorbol ester binding and zinc concentrations (0) were measured as described. Protein associated zinc is essentially only present at levels significantly above background in the mixed micelle peak where PDBu binding was measured. High levels of zinc were consistently found in column inclusion volume fractions coinciding with free, non-protein-associated zinc (fractions 42-50). Error bars indicate the standard error of four or more measurements per fraction. kinase activity properties. Like zinc, phorbol ester binding activity was only observed in mixed micelle fractions, where the smaller 32-kDa proteolytic fragment was the most prominent (>go%) protein species (Fig. 4 A , lane 3 ) . The larger proteolytic fragments eluted in fractions as predicted from the M , values determined by SDS-PAGE. Coincident with the presence of 45-50-kDa fragments in these fractions (Fig.  4A, lane 4 ) , kinase activity was measured. Minor amounts (about 10%) of this activity were also found in mixed micelleassociated fractions, presumably due to the presence of the intact enzyme (Fig. 4A, lane 3 ) . When conversion of lipiddependent to lipid-independent kinase activity was complete, no such peak of kinase activity was observed in mixed micelle fractions, but zinc was still present. These observations indicate that contamination by intact PKC does not account for the presence of the majority of zinc in these fractions (data not shown).
Assessment of the stoichiometry of zinc bound to the 32-kDa fragment in mixed micelle peak fractions of the Ultrogel AcA 44 gel filtration column revealed similar values to those found for the intact PKC isoform (4.7 k 0.9; see Table I).
Essentially the same results were obtained when trypsintreated PKC a associated with Triton X-1OO/PS mixed micelles in the absence of PDBu was analyzed (data not shown), suggesting that residues involved in the coordination of zinc need not directly interact with PDBu, since this event does not displace zinc from the lipid binding domain.
Previous work (6) and the data presented here strongly suggest that the 32-kDa fragment is indeed the lipid binding regulatory domain of PKC. To confirm this and assess precisely where cleavage occurred within PKC, fragments were probed with polyclonal, antipeptide antisera directed against different segments of the regulatory PKC domain (Fig. 4B).
Both the antisera against the pseudosubstrate domain (PKC a residues 19-36) and a peptide within the Cys 2 region (PKC y residues 102-115) gave the same staining pattern and results for the latter antibody are illustrated (lanes 1-4). As expected the antisera recognized intact PKC (lanes 1 and 2; M, 80,000) and 32-kDa fragment (lanes 2 and 3) but not the protein bands at M, 45,000-50,000 (lanes 2 and 4 ) . An antibody against the hinge region (PKC a residues 350-370) recognized intact PKC a (MI 80,000), two of the three bands at M, 50,000 (lane 5 ) , and to a lesser extent the 32-kDa fragment (not shown). A similar array of protein bands ( M , 45,000-50,000) representing the PKC catalytic domain, has also been observed previously after limited proteolysis of PKC a within the hinge region by calcium-dependent neutral proteases (28). Thus, all zinc bound to PKC is shown to be coordinated within the lipid binding PKC regulatory domain.
Attempts to demonstrate a physiological role for zinc in PKC by dialysis against chelators under nondenaturing conditions failed to remove the metal ion, indicating tight association between zinc and the protein (data not shown). The presence of urea and 1,lO-phenanthroline, a high-affinity zinc chelator with low affinity for both calcium and magnesium, was equally ineffective, because PDBu binding activity could never be reconstituted after such exposures. In most cases the enzyme was lost during the dialysis of such dilute solutions containing 50-100 pg/ml protein. Experiments designed to unequivocally demonstrate a functional role for zinc coordinated within the regulatory domain of PKC will require far larger quantities of pure enzyme than currently available.
In summary, the data presented show that protein kinase C is a metallo-enzyme in which zinc is bound within the lipid binding domain. Stoichiometry measurements of intact enzyme indicated that four zinc atoms are coordinated per PKC molecule. The zinc bound is associated within the segment of the protein possessing PDBu binding activity rather than in the segment encoding the kinase activity of the enzyme. Cysteine-rich motifs in both PDBu binding proteins PKC and n-Chimaerin are essential for that activity (8,13). Similar or related sequence motifs present in a variety of transcription factors such as GAL4, TFIIIA, and the glucocorticoid receptors are involved in the coordination of zinc in distinct DNAbinding structural motifs: (i) zinc clusters, (ii) zinc fingers, and (iii) zinc twists (29). Thus, the cysteine residues in the cysteine-rich PDBu binding motifs may be critically involved as ligands for zinc. The observed zinc stoichiometry for PKC corresponds to a value expected if it were coordinated by cysteine residues within the Cys 1 and Cys 2 regions (10) in a structural motif related to the Zn(II)$2ys6 binuclear cluster of the GAL4 transcriptional factor (27). If this suggestion is supported by subsequent structural studies, then each of these zinc binuclear clusters may function as structural domains in the physiological activation mechanism of PKC by DAG (9). Furthermore, since PDBu binding did not lead to displacement of zinc from the enzyme, one may predict that residues other than the essential cysteines of the Cys 1 and Cys 2 regions interact directly with PDBu and also DAG, in contrast to previous suggestions (30).