Interaction of PKN with alpha-actinin.

PKN is a fatty acid- and Rho-activated serine/threonine protein kinase, having a catalytic domain homologous to protein kinase C family. To identify components of the PKN-signaling pathway such as substrates and regulatory proteins of PKN, the yeast two-hybrid strategy was employed. Using the N-terminal region of PKN as a bait, cDNAs encoding actin cross-linking protein α-actinin, which lacked the N-terminal actin-binding domain, were isolated from human brain cDNA library. The responsible region for interaction between PKN and α-actinin was determined by in vitro binding analysis using the various truncated mutants of these proteins. The N-terminal region of PKN outside the RhoA-binding domain was sufficiently shown to associate with α-actinin. PKN bound to the third spectrin-like repeats of both skeletal and non-skeletal muscle type α-actinin. PKN also bound to the region containing EF-hand-like motifs of non-skeletal muscle type α-actinin in a Ca2+-sensitive manner and bound to that of skeletal muscle type α-actinin in a Ca2+-insensitive manner. α-Actinin was co-immunoprecipitated with PKN from the lysate of COS7 cells transfected with both expression constructs for PKN and α-actinin lacking the actin-binding domain. In vitro translated full-length α-actinin containing the actin-binding site hardly bound to PKN, but the addition of phosphatidylinositol 4,5-bisphosphate, which is implicated in actin reorganization, stimulated the binding activity of the full-length α-actinin with PKN. We therefore propose that PKN is linked to the cytoskeletal network via a direct association between PKN and α-actinin.

terminal region of PKN as bait. One of the positive cDNA clones isolated from human brain cDNA library encoded a neurofilament L protein, a neuron-specific intermediate filament protein (6). We have demonstrated that PKN binds to and phosphorylates the head-rod domain of intermediate filament proteins such as each subunit of neurofilament and vimentin in vitro (6) and raised the possibility that PKN plays a role in the assembly of intermediate filament, one of the major components of cytoskeleton. Here we report that the two other groups of positive cDNA clones encoded ␣-actinin, a constituent of the other major component of cytoskeleton.
Preparation of GST Fusion Proteins-cDNAs encoding the spectrinlike repeat 3 (aa 479 -600) and EF-hand-like region (aa 712-834) of HuActNm were amplified by PCR from the human brain cDNA library and were ligated to pGEX4T vector. cDNAs encoding the spectrin repeat 20 and EF-hand-like region of ␣-spectrin were amplified by PCR from rat brain cDNA library and were ligated to pGEX4T vector. Expression and purification of GST or GST fusion proteins were performed according to the manufacturer's instruction (Pharmacia Biotech Inc.) The eluate from glutathione-Sepharose 4B (Pharmacia Biotech Inc.) was dialyzed overnight against 10 mM Tris/HCl at pH 7.5 containing 1 mM EDTA, 1 mM DTT, and 0.1 g/ml leupeptin.
In Vitro Transcription and Translation-A plasmid for in vitro transcription of the full-length coding region of HuActSk1 was constructed as follows. The cDNA encoding the N-terminal region (aa 1-422) of HuActSk1 containing actin-binding domain was amplified by PCR from the human brain cDNA library and was ligated to the C-terminal part (aa 423-894) of clone 4 isolated in the two-hybrid screening. This cDNA for the full-length coding region of HuActSk1 was subcloned into pBluescript II SKϩ. The plasmid was linearized by cutting with XhoI, and the cRNA was transcribed using T3 RNA polymerase. In vitro transcription for the N-terminal region of PKN (aa 1-474, this region is designated as PKNN2 (6)) was performed as described previously (6). For in vitro translation, cRNAs were translated in rabbit reticulocyte lysate (Promega) in the presence of [ 35 S]methionine as described previously (6).
In Vitro Binding Assay--For the in vitro binding experiment, 2 l of in vitro translated PKNN2 was mixed with 5 g of each GST-␣-actinin * This work was supported in part by research grants from the Ministry of Education, Science, Sports and Culture, Japan, the Japan Foundation for Applied Enzymology, and Kirin Brewery Co., Ltd. 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.
For analysis of the effect of phosphatidylinositol 4,5-bisphosphate (PI4,5P2) (Boehringer Mannheim) on the binding between ␣-actinin and PKN, 2 l of in vitro translated full-length HuActSk1 was mixed with 5 g of GST-PKNN1 fusion protein or with 25 g of GST alone in 400 l of buffer P (20 mM Tris/HCl at pH 7.5, 0.5 mM DTT, 120 mM NaCl, 1 mM EDTA) and incubated for 1 h at room temperature with or without PI4,5P2 as indicated in the figure legends. After addition of 25 l of glutathione-Sepharose 4B pretreated with E. coli extract, the binding reaction was continued for an additional 30 min at 4°C. The glutathione-Sepharose 4B was then washed four times in buffer P containing 0.01% Triton X-100. Bound proteins were eluted with GST elution buffer and were subjected to SDS-PAGE. The binding was visualized and quantitated by an imaging analyzer (FUJI BAS1000).
Antiserum-The anti-hemagglutinin (HA) monoclonal antibody 12CA5 was purchased from Boehringer Mannheim. ␣C6, a specific antiserum against PKN, was prepared by immunizing rabbits with the bacterially synthesized fragment of aa 863-946 of rat PKN.
In Vivo Binding Assay-HA-tagged cDNA for HuActSk1 (aa 333-894) was created by fusion of a cDNA encoding 9-aa epitope from the influenza HA to the N terminus of clone 4. A vector pHA-Act was constructed by subcloning this cDNA into pTB701 (8). Empty pHA vector was constructed by subcloning a cDNA encoding only HA epitope into pTB701. A vector pHA-Act or empty pHA vector was cotransfected into COS7 cells with the expression vector pMhPKN3 (2) encoding the full-length human PKN. After 48 h, ϳ10 6 cells were lysed in 500 l of lysis buffer (20 mM Tris/HCl at pH 7.5, 1% Nonidet P-40, 137 mM NaCl, 10% glycerol, 1 mM phenylmethylsulfonyl fluoride, 20 g/ml aprotinin, 10 g/ml leupeptin) for 1 h. Insoluble materials were removed by centrifugation at 15,000 ϫ g for 10 min. Twenty-five-l aliquot of the supernatant was subjected to SDS-PAGE, and the amount of the expression of proteins was assessed by Western blotting using 12CA5 and ␣C6. Four hundred-l aliquot of the supernatants was incubated with 1 g of 12CA5 for 2 h. After addition of 20 l of 50% protein A-Sepharose, the mixtures were further incubated for 1 h. The immunoprecipitates adsorbed to protein A-Sepharose were washed twice with HA wash buffer (100 mM Tris/HCl at pH 7.5, 0.5 M LiCl) and twice with 10 mM Tris/HCl at pH 7.5. The resultant immunoprecipitates were resuspended in 50 l of Laemmli's sample buffer (9), and 25-l aliquot of the extract was subjected to SDS-PAGE, following detection by immunoblotting with ␣C6.
Preparation of Actin Cytoskeletal Proteins-Actin was purified from rabbit skeletal muscle by the method of Mommaerts (10). ␣-Actinin was purified from bovine aorta by the method of Feramisco and Burridge (11). Vinculin was purified from bovine aorta by the method of Kobayashi and Tashima (12). Caldesmon was partially purified from bovine aorta by the method of Abe et al. (13). Filamin, metavinculin, and talin were partially purified from bovine aorta as described (11).
Kinase Assay-The phosphorylation by PKN was carried out at 30°C in an assay mixture containing 20 mM Tris/HCl at pH 7.5, 4 mM MgCl 2 , 100 M ATP, 185 kBq of [␥-32 P]ATP, phosphate acceptors, 20 ng/ml purified PKN from rat testis (13), and with or without 40 M arachidonic acid as indicated in each experiment. Partial purified protein was boiled for 3 min to destroy endogenous kinase activity before use as a phosphate acceptor. After incubation for 5 min, the reaction was terminated by the addition of an equal volume of Laemmli's sample buffer and separated on SDS-polyacrylamide gels. The gels were dried under vacuum, and the phosphorylation was visualized and quantitated by an imaging analyzer (FUJI BAS1000). When the ␦ protein kinase C peptide (1) was used as a phosphate acceptor, reactions were terminated by spotting a mixture onto a Whatman P81 paper and submerging it in 75 mM phosphate and followed by three 10-min washes. Incorporation of [ 32 P]phosphate into the peptide was assessed by scintillation counting.

RESULTS
PKN Interacts with ␣-Actinin in the Yeast Two-hybrid System-We screened a million yeast colonies transformed with both human brain cDNA library fused to Gal4 transcriptional activation domain and a bait construct encoding PKNN1 fused to Gal4 DNA binding domain. The 82 plasmids were isolated representing 16 different cDNAs as judged by cDNA sequencing. Three positive clones (clones 4, 10, and 25) encoded the skeletal muscle type ␣-actinin (HuActSk1, designated in Ref. 7). The clone 4 encoded HuActSk1 from aa 333 to the C terminus, and both clone 10 and clone 25 encoded HuActSk1 from aa 344 to the C terminus. All these clones contained complete C terminus but lacked the N-terminal actin-binding domain (14) (Fig. 1B). These clones resulted in high ␤-galactosidase levels upon co-transformation with the PKN bait construct in the original yeast host strain YGH1. The specificity of this interaction was tested further by measuring the stability of other combinations of two-hybrid constructs, LexAbd (instead of Gal4bd)-PKN and Gal4ad-␣-actinin to support lacZ expression in L40 cells (MATa trp1 leu2 his3 LYS2::lexA-HIS3 URA3::lexA-lacZ). As shown in Fig. 2, high ␤-galactosidase activity was also developed in this system, suggesting a specific interaction between the N-terminal region of PKN and ␣-actinin. The two-hybrid method was employed to identify the region on PKN that interacted with HuActSk1, and this region was compared with the binding site for RhoA, protein already known to interact with PKN in vitro and in vivo (4,5). The RhoA-binding site has been mapped on the aa 33-111 of PKN that corresponds to the first leucine zipper-like sequence of PKN (15), whereas ␣-actinin very weakly interacted with this region of PKN (Fig. 1A). By contrast, ␣-actinin strongly interacted with aa 136 -189 of PKN, whereas no interaction was detected between RhoA and this region of PKN (data not shown). This region corresponds to the second leucine zipperlike sequence and its immediate N-terminal region, which is conserved through evolution in vertebrates (16). Thus ␣-actinin binds most avidly to the region distinct from that which binds to RhoA. These results raise the possibility that PKN binds simultaneously to RhoA and ␣-actinin.
Binding of PKN to HuActSk1 in Vitro-␣-Actinin is composed of three domains, an N-terminal actin-binding domain, extended rod-shaped domain with four internal 122 aa repeats (spectrin-like repeats), and a C-terminal region containing a pair of presumptive helix-loop-helix Ca 2ϩ -binding motifs, often referred to as EF-hands (reviewed in Ref. 17). To investigate whether PKN binds directly to ␣-actinin and to clarify which part of ␣-actinin is necessary for binding to PKN, various truncated constructs of HuActSk1 were produced as GST fusion proteins in E. coli (Fig. 1B). As shown in Fig. 3, in vitro translated PKNN2 strongly bound to each ␣-actinin fragment (aa 423-653, aa 653-837, and aa 486 -607) but not to the fragment (aa 837-894 and aa 604 -719). The ability of the binding was verified by the demonstration that the complex was resistant to washing with 0.5% Triton X-100, 0.5 M NaCl. These results suggest that the recombinant ␣-actinin interacts directly with recombinant PKN and that two distinct regions of HuActSk1, which correspond to the spectrin-like repeat 3 and the region containing the EF-hand-like motifs, play important roles in the recognition of PKN. As expected from the twohybrid data, the N-terminal region of PKN (aa 136 -474) translated in vitro, which lacked the RhoA binding region, was sufficient for direct binding to ␣-actinin (Fig. 3, lanes 18 -20).
Binding of PKN to Non-skeletal Muscle Type ␣-Actinin (Hu-ActNm) in Vitro and Ca 2ϩ Dependence of Its Interaction-A number of distinct isoforms of ␣-actinin have been character- ized, including skeletal, smooth, and non-muscle ␣-actinins, from various kinds of cells and tissues. The only recorded functional difference among these ␣-actinins is that binding of the non-muscle isoform to F-actin is inhibited by Ca 2ϩ , whereas binding of the muscle isoform is Ca 2ϩ -insensitive (18 -21). In human, only one clone of the non-muscle cytoskeletal isoform (HuActNm, designated in Ref. 7), having strong sequence homology with HuActSk1 (89% similarity and 80% identity for pairwise comparison), has been isolated (22,23). Then we tested whether PKN could bind to the region of HuActNm corresponding to the PKN-binding site of HuActSk1. As shown in Fig. 4A, PKN could also bind to spectrin-like repeat 3 domain of HuActNm, whereas the binding to the EF-hand-like domain of HuActNm was not detected in the absence of Ca 2ϩ . However, PKN could effectively bind to the EF-hand-like region of Hu-ActNm in the presence of 1 mM Ca 2ϩ (Fig. 4B). Although it is uncertain at present whether this Ca 2ϩ dependence is retained in the binding between PKN and the full-length HuActNm, the Ca 2ϩ dependence may be one of the reasons why the cDNA clone encoding the non-muscle type ␣-actinin was not isolated in the two-hybrid screening of "brain" cDNA library. Beggs et al. (7) compared the sequences of EF-hand-like regions of Hu-ActSk1 with the EF-hand consensus of Kretsinger (24) and indicated that the first EF-hand-like region of HuActSk1 has only 11/16 matches with either an arginine or lysine at the Y position and that these peptides would probably not be able to coordinate Ca 2ϩ binding properly. Our results support this estimation from a different point of view.
Specificity of the Interaction between PKN and ␣-Actinin-␣-Actinin is a member of spectrin superfamily, including spectrin, dystrophin, and so on (17,25,26). Family members are characterized by the N-terminal actin-binding domain, central rod-shaped spectrin-like repeats, and the C-terminal EF-handlike domain. ␣-Spectrin contains 21 rod-shaped repeats in the N-terminal to the EF-hand-like domain. The C terminus of ␣-spectrin is clearly related to ␣-actinin, and especially the repeat 20 of ␣-spectrin has extensive homology to the repeat 3 of ␣-actinin (27,28), and the position of the repeat in each HuActSk1 (skeletal muscle type ␣-actinin) (aa 423-894) was expressed as a fusion protein with VP16 transcription activation domain, and its interaction with the various deletion mutants of PKN expressed as fusion proteins with the LexA DNA binding domain was examined in the two-hybrid system. LZ indicates the leucine zipper-like motif. BR indicates the region rich in basic aa. Solid box indicates the bait construct. B, HuActSk1. The numbers (4, 10, and 25) on the right indicate the clone numbers isolated in the screening. The N-terminal region of PKN (aa 1-540; this region was designated as PKNN1) was expressed as a fusion protein with Gal4 DNA binding domain, and its interaction with the various deletion mutants of HuActSk1 expressed as a fusion protein with Gal4 activation domain or VP16 activation domain was examined in the two-hybrid system. SR indicates spectrin-like repeats. C, HuActNm (non-skeletal muscle type ␣-actinin). SR indicates spectrin-like repeats. protein seems to be related to each other. Since PKN bound to the repeat 3 of ␣-actinin, we examined whether PKN can bind to the repeat 20 of ␣-spectrin. As shown in Fig. 5, in vitro binding between PKN and the repeat 20 of rat ␣-spectrin was not detected in the same condition in which PKN bound to the repeat 3 of ␣-actinin. These results indicate that PKN specifically binds to the spectrin-like repeat of ␣-actinin.
Binding of PKN to ␣-Actinin in Vivo-The interaction of ␣-actinin with PKN in vivo was examined by cotransfection experiment in COS7 cells (Fig. 6). An epitope-tagged ␣-actinin was generated by fusion of a 9-aa epitope from the influenza HA to the N terminus of clone 4 protein, enabling the selective immunoprecipitation of the tagged ␣-actinin polypeptide with anti-HA monoclonal antibody 12CA5 (29). This HA-tagged ␣-actinin contains the complete C-terminal region of ␣-actinin, whereas it lacks the N-terminal actin-binding domain. After co-expression of HA-tagged ␣-actinin with the full-coding region of PKN in COS7 cells, anti-HA immunoprecipitates contained substantially immunoreactive PKN. These results suggest that the C-terminal region of ␣-actinin can associate in vivo with PKN.
PI4,5P2-dependent Binding between PKN and ␣-Actinin-␣-Actinin in vivo bound to various amounts of endogenous PI4,5P2, and the specific interaction between ␣-actinin and PI4,5P2 regulates the F-actin-gelating activity of ␣-actinin (30). This indicates that PI4,5P2 causes a conformational change in ␣-actinin. Exogenously added PI4,5P2 can bind to ␣-actinin strongly, and the binding is tight and stable (30). Then we examined the binding activity of PKN with ␣-actinin in the presence or absence of PI4,5P2. Since PI4,5P2 binding region resides in the actin-binding domain of ␣-actinin (14), in vitro translated full-length ␣-actinin containing actin-binding domain was used in this in vitro binding experiment (Fig. 1B). FIG. 5. Specific interaction of PKN with spectrin-like repeats of ␣-actinin. 35 S-Labeled in vitro translated PKNN2 was incubated with the bacterially synthesized GST or GST-fused spectrin-like repeats of ␣-actinin or spectrin repeat of ␣-spectrin. Aliquots of the initial binding reaction mixtures (10 l) were removed before precipitation and applied to electrophoresis, which were indicated as Input. GST or GST-fused proteins were collected with glutathione-Sepharose beads, analyzed by 10% SDS-PAGE, and followed by autoradiography. Interestingly, the full-length ␣-actinin very weakly but specifically bound to PKN in the absence of PI4,5P2. However, addition of 10 M PI4,5P2 stimulated the binding of the fulllength ␣-actinin to PKN (Fig. 7A). Therefore PI4,5P2 appears to influence the conformation of ␣-actinin and discloses the partially cryptic binding region for PKN, although the other possibility cannot be ruled out that PI4,5P2 functions as a bridge between ␣-actinin and PKN. This binding activity was elevated with increased PI4,5P2 concentration up to 2.5-10 M and then was lowered to 100 M (Fig. 7B). This two-phase pattern of PI4,5P2 dependence was also reported in the binding of ␣-actinin with PI3-kinase (31). Fukami et al. (30) reported that the effect of PI4,5P2 on gelating activity of ␣-actinin is increased up to 5-10 M of PI4,5P2 and that a further increase in concentration of PI4,5P2 gives a reduction in gelating activity to the basal level due to the formation of large PI4,5P2 micelles. The two-phase pattern of PI4,5P2-dependent binding of ␣-actinin with PKN also may be explained by the same reason.
Effects of ␣-Actinin on the PKN Kinase Activity-We investigated whether the binding of ␣-actinin to PKN directly altered PKN regulation or catalytic function. The purified ␣-actinin from bovine aorta neither activated PKN autophosphorylation nor affected PKN-catalyzed protein kinase C pseudosubstrate peptide phosphorylation when added at Ͼ100 molar excess to PKN (data not shown). When assayed in the presence of 10 M PI4,5P2, peptide phosphorylation activity of PKN was stimulated ϳ1.5-fold; however, addition of ␣-actinin to this assay mixture slightly inhibited the phosphorylation activity toward the basal level. Thus, ␣-actinin does not seem to be a direct activator of PKN purified from the soluble fraction of rat testis in vitro.
Phosphorylation of ␣-Actinin and Other Actin Cytoskeletalassociated Proteins by PKN-Since PKN bound to ␣-actinin, we tested whether ␣-actinin itself could be a substrate for PKN. In the absence of modifiers, PKN purified from rat testis did not phosphorylate ␣-actinin purified from bovine aorta. However, in the presence of 40 M arachidonic acid, PKN phosphorylated purified ␣-actinin with a stoichiometry of ϳ0.02 mol of P i per protein monomer by image quantitation (Fig. 8A). The bacterially expressed C-terminal region of ␣-actinin (amino acid 333-894) was not phosphorylated at all by PKN, but PKN phosphorylated the bacterially expressed N-terminal region of ␣-actinin (amino acid 1-332) that was lacking in the originally isolated clone 4 in the presence of arachidonic acid (data not shown), suggesting that phosphorylation of ␣-actinin by PKN occurred in the N-terminal region. We searched for PKN substrates among other actin cytoskeletal proteins, including filamin, metavinculin, vinculin, talin, caldesmon, and actin. Among them, caldesmon and G-actin were relatively preferred substrates for PKN. (The maximal phosphorylation by PKN per mol of protein subunit was estimated by image quantitation to be ϳ0.3 mol of P i per mol of caldesmon and ϳ0.05 mol of P i per mol of G-actin, respectively.) As shown in Fig. 8B, phosphorylation of G-actin and caldesmon was stimulated up to ϳ2-fold and Ͼ6-fold in the presence of arachidonic acid, respectively.

DISCUSSION
The actin cytoskeleton plays a critical role in a number of cellular processes including motility, chemotaxis, and cell division (32)(33)(34)(35). Members of Rho family of small GTP-binding proteins have been implicated in the regulation. Rho promotes the formation of actin stress fibers and focal adhesions (36,37), although the mechanism by which Rho mediates the effect on the actin cytoskeleton is unclear. Recently it has shown that Rho interacts physically and regulates the activity of PI(4)P5kinase in mouse fibroblasts and thereby regulates the cellular levels of PI4,5P2 (38,39). PI4,5P2 can regulate in vitro the interactions of a number of actin-binding proteins, such as ␣-actinin (30), profilin (40), gelsolin (41), cofilin (42), and p39 capZ (43). It has also been shown that the decrease in PI4,5P2 bound to ␣-actinin and vinculin by treatment with platelet-derived growth factor correlates with the depolymerization of actin (44). Recently, Glimore and Burridge (45) have reported that microinjection of antibodies against PI4,5P2 into Balb/c 3T3 cells inhibits assembly of stress fibers and focal adhesions by serum stimulation. One possibility is that PI4,5P2 synthesis could mediate some of the effects of Rho on the actin cytoskeleton (45). In our experiment, Rho directly interacts with and activates PKN, and PKN could directly associate with ␣-actinin in PI4,5P2-dependent manner in vitro. On the other hand, phosphoinositides such as PI4,5P2 have been reported to affect directly the kinase activity of PKN in vitro (46). Thus, the possibility is raised that PKN is also implicated in mediating some effects of Rho on the actin cytoskeleton. ␣-Actinin itself and some actin-based cytoskeletal proteins such as actin and caldesmon serve as the relatively preferred substrates for PKN, although the stoichiometries were low in our in vitro assay condition. Thus one might further speculate that PKN mediates the effects of Rho and phosphoinositides by phosphorylating these proteins, although it is not known whether these proteins are physiologically relevant substrates of PKN. Further investigation will be required to clarify the role of PKN in the cytoskeletal network.