Interaction Cloning of Protein Kinase C Substrates*

the homologue of the myristoylated alanine-rich C kinase substrate cDNA, whereas is a partial cDNA with homology to the 3‘ of Both cDNAs proteins bind phosphatidyl- serine (PS) and are substrates for PKC. Phosphorylation decreased both PS and PKC binding activities. Both proteins contain high density positive charge domains similar to that found in the major PKC substrate MARCKS. These results demonstrate that PKC inter- actions with certain substrate proteins are ciently high affinity to facilitate their isolation via interaction cloning. Phosphorylation Studies-a-PKC was produced in Sf9 cells in- fected with recombinant a-PKC baculovirus (12). a-PKC was partially purified from cell homogenates following chromatography on DEAE-Sepharose and phenyl-TSK essentially as described (12). Samples were incubated in phosphorylation assay buffer, which was 50 mM Tris-C1 (pH 7.4) containing 1 mM dithiothreitol, 100 pg/ml phosphatidylserine, 5 mM magnesium chloride, 1 mM EGTA, and 1.2 mM calcium in the presence or absence of 25 pM ATP. Phosphatidylserine Overlay Assay-Native or phosphorylated 35A and 35H fusion proteins were blotted to nitrocellulose. The nitrocel- lulose was overlaid with 20 pg/ml [14C]PS (specific activity = 1 pCi/ 75 pg) diluted in 50 mM Tris-CI (pH 7.4) containing 0.5 M sodium chloride, 10 mg/ml bovine serum albumin, 1 mM EGTA, and 1.2 mM calcium for 1 h. Blots were washed briefly in phosphate-buffered saline, dried, and exposed to film for 2-4 days. Calcium requirements for PS binding to 35A and 35H have not been studied in detail. In related experiments, however, we have determined that PS binding to MARCKS does not require calcium.

* This work was supported by Grant 2375 from the Council for Tobacco Research and Grants CA465330, CA37589, and ES05670 from the National Institutes of Health. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
PKC may be targeted to these specific locations via interactions with other proteins, and we have used an overlay assay approach to identify PKC-binding proteins in these cells (6,7). Two lines of evidence demonstrated that the proteins identified by the assay are biologically interesting. First, the two major binding proteins in REF52 cells (Mr = 71 and >200 kDa) are not detected in SV40-transformed REF52 cells (6). SV40-REF52 cells form less stable substrate attachment sites (known as close contacts), which do not contain a-PKC (6). Thus, there is a correlation between loss of a-PKC-binding proteins and loss of targeting to the cytoskeleton. Second, comparison of the properties of REF52 cell PKC-binding proteins and PKC substrates indicated that two of the major binding proteins are also substrates? ( a ) The two major binding proteins comigrated with the major PKC substrates on two-dimensional gels, ( b ) both of the binding proteins and the substrates were heat-soluble, and (c) both of the binding proteins and substrates were not detected in extracts of SV40-REF52 cells. In other work, we have demonstrated that the major PKC substrate, MARCKS, is also detected as a PKCbinding protein in this assay.3 These results indicated that at least some of the binding proteins are also PKC substrates. Based on these data, which demonstrated the potential utility of the assay, we used the overlay assay to screen an expression library in order to identify additional PKC-binding proteins and substrates.

EXPERIMENTAL PROCEDURES
Materials-The rat kidney Xgtll library was purchased from Clontech (Palo Alto, CA). pBluescript I1 SK was from Stratagene (La Jolla, CA). pQEIV was from Qiagen (Chatsworth, CA). PKC antibodies were from Upstate Biotechnologies (Lake Placid, NY). ["C]PS (50 mCi/mmol) was from Amersham Corp. Sequence homologies were identified in the GenBank data bank using software from Intelligenetics.
Library Screening and Isolation of Clones-cDNA clones for PKCbinding proteins were isolated by screening a rat kidney Xgtll library using a PKC overlay assay similar to the method of Wolf and Sayhoun (9) with modifications (7). IPTG-induced proteins were immobilized on nitrocellulose lifts. Briefly, the nitrocellulose was blocked with 5% nonfat dry milk, and then incubated with 10 pg/ml PKC partially purified from rabbit brain (10). Incubations were at room temperature in 50 mM Tris-C1 (pH 7.4) containing 0.5 M sodium chloride (TBS) with the following additions: 10 mg/ml bovine serum albumin, 20 pg/ ml PS, 1 mM EGTA, 1.2 mM calcium, 10 pg/ml leupeptin, and 1 pg/ ml aprotinin. After washing, bound PKC was fixed by incubating with 0.5% formaldehyde. Excess formaldehyde was inactivated by washing with 2% glycine. After washing with phosphate-buffered saline, the nitrocellulose sheets were incubated with anti-a-PKC monoclonal antibody M6 (11) diluted in TBS containing 10 mg/ml bovine serum albumin. Positive clones were identified after incubating with alkaline phosphatase-conjugated second antibody and color development. Positive clones were plaque-purified, and the X DNA was recovered. The inserts were excised with EcoRI and gel-purified.
The clones were inserted into pBluescript I1 SK for sequencing.
Sequencing was performed on a model 370A automated sequenator (Applied Biosystems). Inserts were excised from SK and subcloned into pQE Type IV in the appropriate reading frames to produce poly(His) fusion proteins. Where indicated, the poly(His) fusion proteins were purified by nickel affinity chromatography in 8 M urea buffered with 10 mM Tris-C1 and 100 mM phosphate (pH 8.0) according to the manufacturer's instructions (Qiagen, Chatsworth, CAI. Phosphorylation Studies-a-PKC was produced in Sf9 cells infected with recombinant a-PKC baculovirus (12). a-PKC was partially purified from cell homogenates following chromatography on DEAE-Sepharose and phenyl-TSK essentially as described (12). Samples were incubated in phosphorylation assay buffer, which was 50 mM Tris-C1 (pH 7.4) containing 1 mM dithiothreitol, 100 pg/ml phosphatidylserine, 5 mM magnesium chloride, 1 mM EGTA, and 1.2 mM calcium in the presence or absence of 25 p M ATP.
Phosphatidylserine Overlay Assay-Native or phosphorylated 35A and 35H fusion proteins were blotted to nitrocellulose. The nitrocellulose was overlaid with 20 pg/ml [14C]PS (specific activity = 1 pCi/ 75 pg) diluted in 50 mM Tris-CI (pH 7.4) containing 0.5 M sodium chloride, 10 mg/ml bovine serum albumin, 1 mM EGTA, and 1.2 mM calcium for 1 h. Blots were washed briefly in phosphate-buffered saline, dried, and exposed to film for 2-4 days. Calcium requirements for PS binding to 35A and 35H have not been studied in detail. In related experiments, however, we have determined that PS binding to MARCKS does not require calcium.

RESULTS
Nitrocellulose lifts from a rat kidney Xgtll expression library were incubated with partially purified PKC in the presence of the PKC cofactors phosphatidylserine (PS) and calcium. After washing, bound PKC was detected with anti-CY-PKC specific monoclonal antibodies. Out of 150,000 colonies screened, nine positive clones were detected. Upon subsequent screening, four were found to directly interact with the PKC antibodies and, therefore, do not represent binding proteins. Two apparent binding protein clones, 35A and 35H, were chosen for further analysis.
To demonstrate that the clones isolated actually code for PKC-binding proteins, clones 35A and 35H were expressed as bacterial fusion proteins. In the absence of IPTG induction, PKC binding to a small number of bacterial cell proteins was noted (Fig. 1). In the presence of IPTG, binding protein activities were induced. The induced proteins did not directly react with the anti-PKC antibodies (Fig. 1). Thus, although there is some background due to PKC interactions with endogenous bacterial cell proteins, this did not prevent identification and isolation of PKC-binding protein clones.
Clone 35A is a 1. Clone 35A was subcloned into pBluescript SK for sequencing. We scanned the GenBank data bank for sequence similarities (8) and found one sequence, F52 (131, with significant similarity. Alignment of the deduced amino acid sequences is shown.  homologous to MARCKS at the NH2-terminal myristoylation site and the internal PKC phosphorylation sites (16)(17)(18)(19). The homology between mouse F52 and rat 35A included the entire F52 coding region and continued into the 3' non-coding region (data not shown). Sequence homology in the first 400 bases of the 3' non-coding region was 92%. The strong homology in both coding and non-coding regions indicates that clone 35A is the rat homologue of mouse F52. Clone 35H is a 767-bp sequence containing an open reading frame of 612 hp. DNA sequence analysis of clone 3SH demonstrated two regions with substantial homology to the 3' coding region of human P-adducin (Fig. 3A). Thus, 35H appears to be a partial clone of a P-adducin homologue. The 5' end of 35H was 72% homologous to the human P-adducin sequence (20) from 1757 to 1895 bp. The translated sequences were 76% homologous in this region (amino acids 462-521, Fig. 3, B and C). The 3' end of' 35H was 79% homologous over the last 98 bp of the P-adducin coding region. The translated sequences were 84% homologous in this region (amino acids 693-726, Fig. 3, B and D). Homology  quence of 35H (329 bp) between the two regions of homology was not homologous to the intervening P-adducin sequence (504 bp). Since P-adducin message was reported to be undetectable in rat kidney (20), we presume that 35H represents a unique sequence, possibly with functional properties similar to adducin. Mouse F52 has been shown recently to be phosphorylated by PKC at a site corresponding to the phosphorylation sequence found in MARCKS (14). This sequence was conserved in clone 35A (Fig. 4). Adducin has also been reported to be a good PKC substrate, and incorporates 2-3 mol of phosphate/ mol (21,22). Two of the putative adducin PKC phosphorylation sites are within the carboxyl-terminal domain (20) (see Fig. 323). Closer comparison of the translated /3-adducin and 35H sequences in this region demonstrated a change from a potential phosphorylatable serine in adducin to an asparagine in 35H (Fig. 4B). To determine if 35H encoded a PKC substrate, the bacterially expressed protein was incubated with PKC and cofactors as described in the legend to Fig. 5. Preliminary phosphopeptide mapping studies indicate serine 177, threonine 191, and serine 193 as potential sites of phosphorylation. Thus, despite the change from serine 718 of adducin, 35H protein is a PKC substrate. In related experiments, antisera to 35H were used to characterize the protein in cultured renal proximal tubule epithelial cells. The antibody recognized an 80-kDa protein, which was rapidly phosphorylated in phorbol ester-treated cells. These results indicate that a protein antigenically related to 35H is a PKC substrate in The 3' region of 35H (Fig. 3 0 ) encodes a highly basic domain that is similar to phosphorylation sequences in F52 and MARCKS (Fig. 4). Although the linear sequences are not homologous, both have very high positive charge densities (+13/25 amino acids for MARCKS, F52, and 35A and +12/ 32 amino acids for /3-adducin and 35H). The presence of high positive charge density domains in MARCKS, 35A, and 35H suggests that these regions are important in mediating PKC binding activity.

~K A E~S K V S S G T P I K l E O P N Q F V P L N~N~E V l E K R N K l R E Q N R Y D L K T A G~S Q l L
Previous work has demonstrated that PKC-binding proteins are also phosphatidylserine (PS)-binding proteins; however, not all PS-binding proteins bind PKC efficiently (7,9,23). These results emphasize the potential importance of phospholipid bridging in mediating PKC interactions with other proteins but also suggest that protein-protein interactions may provide additional stabilization.
[14C]PS overlays were used to determine if the proteins expressed by clones 35A and 35H bound PS. PS binding was detectable with as little as 30 ng (-1 pmol) of protein (Fig. 5A). Phosphorylation decreased both PS binding and PKC binding (Fig. 5, A 35A and 35H proteins were expressed as poly(His) fusion proteins and purified by nickel affinity chromatography. An aliquot (2.5 pg) was diluted in phosphorylation assay buffer in a total volume of 100 pl. Samples were incubated in the presence or absence of 1 pg of partially purified a-PKC for 2 h a t 30 "C. T o ensure complete phosphorylation, additional PKC and ATP were added to aliquots of each sample a t the end of 2 h and the incubations were continued for 2 more h. Samples (100 ng) were diluted in electrophoresis sample buffer, run on 10% denaturing polyacrylamide gels, and blotted to nitrocellulose. In A, samples were overlaid with [14C]PS, In B, the nitrocellulose was overlaid with PKC to detect PKC-binding proteins.
phosphorylation. These results also suggest that the PS binding and PKC phosphorylation sites are in close proximity.

DISCUSSION
Two PKC substrates have been cloned using an overlay assay to screen an expression library for PKC-binding proteins. Thus, the overlay assay represents a novel approach for cloning and identifying PKC substrates. Comparison of the sequences of the isolated clones and MARCKS, which is also a PKC-binding protein: suggests that highly basic domains are common among PKC-binding proteins. In MARCKS, this basic domain has been associated with several functions including actin, calmodulin, and phosphatidylserine binding (19, [24][25][26]. It is likely that these domains function in electrostatic interactions with PS, and that PS bridging to PKC is an important component of the interactions detected by this method, as originally suggested (9). However, since the assay is performed in the presence of 0.5 M sodium chloride and since not all PS-binding proteins are good PKC-binding proteins (7), protein-protein interactions may also play a role. It is relevant to point out that several PKC substrates have been shown to interact directly with PS (27,28). Overall, the data suggest that PS-dependent protein-protein interactions are important in determining PKC phosphorylation targets. Using a similar assay system, others have described a distinct domain in annexin I and p65 that appears to interact with activated PKCs (29,30). The substrates that we have cloned with this overlay assay approach are distinct from the low molecular mass (-30 kDa) receptors for activated C kinase (RACKS) recently described (23). Thus, it is possible that several types of PKC-binding proteins with distinct functional domains will eventually be identified.
An important factor in considering what may be the primary targets for PKC phosphorylation in vivo is the subcellular localization of PKCs. In principle, location could regulate accessibility to substrate proteins (discussed in Ref. 31).
Although it is interesting to note that a-PKC colocalizes with MARCKS in close contacts of macrophages (32), with vinculin in focal contacts of REF52 cells (4, 6), and with 35H in cell-cell junctions of renal proximal tubule epithelial cells,4 we do not yet know if a-PKC interactions with these proteins target a-PKC to these locations. Additional studies describing the interactions between the phosphorylated and unphosphorylated forms of 35A and 35H with PKC in vitro and in vivo are required to fully address this problem.