PKN Associates and Phosphorylates the Head-Rod Domain of Neurofilament Protein*

PKN is a fatty acid-activated serine/threonine kinase that has a catalytic domain highly homologous to that of protein kinase C in the carboxyl terminus and a unique regulatory region in the amino terminus. Recently, we reported that the small GTP-binding protein Rho binds to the amino-terminal region of PKN and activates PKN in a GTP-dependent manner, and we suggested that PKN is located on the downstream of Rho in the signal transduction pathway (Amano, M., Mukai, H., Ono, Y., Chihara, K., Matsui, T., Hamajima, Y., Okawa, K., Iwamatsu, A., and Kaibuchi, K. (1996) Science 271, 648– 650; Watanabe, G., Saito, Y., Madaule, P., Ishizaki, T., Fujisawa, K., Morii, N., Mukai, H., Ono, Y. Kakizuka, A., and Narumiya, S. (1996) Science 271, 645–648). To identify other components of the PKN pathway such as sub- strates and regulatory proteins of PKN, the yeast two-hybrid strategy was employed. By this screening, a clone encoding the neurofilament L protein, a subunit of neu-ron-specific intermediate filament, was isolated. The amino-terminal regulatory region of PKN was shown to associate with the head-rod domains of other subunits of neurofilament (neurofilament proteins M and H) as well as neurofilament L protein in yeast cells. The direct binding between

PKN is a fatty acid-activated serine/threonine kinase that has a catalytic domain highly homologous to that of protein kinase C in the carboxyl terminus and a unique regulatory region in the amino terminus. Recently, we reported that the small GTP-binding protein Rho binds to the amino-terminal region of PKN and activates PKN in a GTP-dependent manner, and we suggested that PKN is located on the downstream of Rho in the signal transduction pathway ( We have reported a novel serine/threonine protein kinase, designated PKN, having a catalytic domain homologous to the PKC 1 family and unique amino-terminal sequences (1,2). The amino-terminal region of PKN contains repeats of a leucine zipper-like motif, suggesting promotion of protein-protein association through hydrophobic interactions (3), and the basic region adjacent to the first leucine zipper-like motif, which is conserved through evolution in vertebrates (4) and among the various isoforms of PKN (5). In the previous study, we demonstrated that truncation of the amino-terminal region of PKN by limited proteolysis results in the generation of the catalytically active form in vitro, and unsaturated fatty acid and some detergent can remove restriction of the catalytic activity by its amino-terminal region at relatively low concentrations in vitro (1,6). Thus, regions contained in the amino-terminal portion of PKN are presumed to be critical for the regulation of the biological activity of this enzyme.
Recently, we demonstrated that Rho, a small GTP-binding protein, binds to PKN in a GTP-dependent fashion and that this binding leads to the activation of PKN (7,8), suggesting that PKN is one of the targets of Rho. Several reports have been accumulated showing that multiple protein-protein interactions within the regulatory region of serine/threonine protein kinases are important for the regulation of the enzyme activities in vivo, such as Raf-1 and p21-activated protein kinase (9 -19). For example, current models of Raf-1 activation suggest that this kinase is bound in a native complex with 14-3-3 proteins (10, 11) and p50 and hsp90 proteins (12,13). Raf-1 also binds to Ras in a GTP-dependent fashion (14 -17), and active Ras is thought to recruit Raf-1 to the plasma membrane, where it interacts with other modulators of its activity and relevant substrates (18,19). The binding site on Raf-1 for 14-3-3 proteins is distinct from the Ras binding domain, and Raf-1 molecules bound to 14-3-3 proteins may simultaneously bind to Ras (10,11). In our preliminary experiment, PKN immunoreactivity in the crude extract from rat brain was eluted broadly on gel filtration, suggesting that PKN formed complexes. Thus, in the analogy with Raf-1, we attempted to identify other proteins that interact with PKN and potentially regulate its activity by using a yeast two-hybrid system, with the aminoterminal regulatory region of PKN as a target protein. By this screening, a cDNA clone encoding the head-rod domain of NFL, a subunit of neuron-specific intermediate filament protein, was isolated from a human brain cDNA library. In this report we characterize the interaction between PKN and each subunit of NF and vimentin and raise the possibility that PKN can play a role in the regulation of assembly of NF and provide a potential junction from signals generated by growth factor stimulation to intermediate filaments.

EXPERIMENTAL PROCEDURES
Two-hybrid Screens and Constructs for the Two-hybrid System-A scheme of the fusion constructs for each subunit of NF used in this study is represented in Fig. 1. An EcoRI/BamHI fragment encoding amino acids 1-540 of human PKN, designated as PKNN1, was inserted into the vector pGBT9 (20). This plasmid, which contains a TRP1 marker, was cotransfected with the human brain cDNA library constructed using pGAD10 (Clontech Laboratories, Inc.; each plasmid contains a LEU2 marker) into the yeast strain YGH1 (a, ura3-52, his3-200, ade2-101, lys2-801, trp1-901, leu2-3, Can r , gal4 -542, gal80 -538, LYS2::gal1 uas -gal1 tata -HIS3, URA3::gal1-lacZ). Screening was performed as described by Hannon et al. (21). About 1 ϫ 10 6 transformants were analyzed. Primary positive clones were recovered and retransfected into the original yeast host strain in combination with the Gal4bd-PKN or the Gal4bd-p53 tumor suppressor. Library plasmids that activated marker expression only in the presence of PKN were analyzed further.
Other yeast expression vectors were prepared as follows. The yeast expression vector pVP/NFL#21, for VP16ad fused to the head-rod domain of human NFL, was made by subcloning an EcoRI insert (amino acids 1-349) of the library plasmid 21 originally isolated from twohybrid screening into pVP16 vector. Vectors for VP16ad fused to the tail domain of NFL, VP16ad fused to the tail domain of NFH, VP16ad fused to the head-rod domain of NFM, and VP16ad fused to the tail domain of NFM were constructed by subcloning each BamHI/NotI insert of pGST/ NFLt, pGST/NFHt, pGST/NFMhr, and pGST/NFMt (described later under "Preparation of GST Fusion Proteins"), respectively, into pVP16. pVP/NFHf, for VP16ad fused to the full length of NFH, was constructed by digesting pBH, a plasmid containing the entire coding region of NFH in pBluescript II SK vector (kindly provided by Dr. N. Hirokawa, University of Tokyo), with BglII, and the cDNA insert was subcloned into pVP16. pVP/NFHhr for VP16ad fused to the head-rod domain of NFH was made by digesting pVP/NFHf with Tth111I/EcoRI and filling in the ends with T4 DNA polymerase and ligating them together with T4 DNA ligase. This removed the DNA sequences 3Ј-terminal to the Tth111I site, which encodes the tail domain of NFH. pBTM/NFL#21 and pBTM/ NFLt for LexAbd fused to the head-rod domain of NFL and LexAbd fused to the tail domain of NFL, respectively, were constructed as described above except for using pBTM116 instead of pVP16. pBTM/ PKNN1 for LexAbd fused to PKNN1 and pVP/PKNN1 for VP16ad fused to PKNN1 were constructed by subcloning the EcoRI/BamHI fragment of human PKN into pBTM116 and pVP16, respectively. pBTM/PKNC1 for LexAbd fused to the carboxyl-terminal region of PKN (amino acids 511-942, this region was designated as PKNC1) and pVP/PKNC1 for VP16ad fused to PKNC1 were constructed by subcloning the ClaI/EcoRI fragment of human PKN into pBTM116 and pVP16, respectively.
In Vitro Transcription and Translation-For in vitro transcription, truncated human PKN was made as follows. pPKNN2 for the aminoterminal region of PKN (amino acids 1-474, this region was designated as PKNN2) was made by digesting phPKN-H4 (human PKN cDNA in pBluescript II SK; see Ref. 2) with BstEII, filling in the ends with T4 DNA polymerase, and self-ligating. This removed the original amino acid sequences carboxyl-terminal of the BstEII site (1425 nucleotides from the initiating ATG), and a stop codon was created in the plasmid sequence. The fragment encoding amino acids 614 -942 (this region was designated as PKNC2) containing the conserved catalytic domain of human PKN was made by polymerase chain reaction amplification. pPKNC2 was made by subcloning this fragment into pRc/CMV (Invitrogen Corp.). These plasmids were linealized by cutting with XbaI, and cRNAs were transcribed using T7 RNA polymerase. For NFL, pBL (22) (kindly provided by Dr. N. Hirokawa) was linealized by cutting with HindIII, and cRNA was transcribed using T3 RNA polymerase. For in vitro translation, these cRNAs were translated in the rabbit reticulocyte lysate (Promega) in the presence of [ 35 S]methionine.
Preparation of GST Fusion Proteins-pGST/NFL#21 for GST fused to NFL#21 (amino acids 1-349 of human NFL) was made by subcloning an EcoRI insert of plasmid 21 into the pGEX4T vector. pGST/NFLdelA for GST fused to NFLdelA (amino acids 1-175 and 335-349) was made by digesting a pGST/NFL#21 with PstI, removing the 480-base pair fragment, and self-ligating with T4 DNA ligase. pGST/NFLdelB for GST fused to NFLdelB (amino acid 245-349) was made by subcloning the BglII/EcoRI fragment of pGST/NFL#21 into the pGEX4T vector. pGST/NFLt for the GST fused to the tail domain of NFL was made by subcloning the ϳ700-base pair KpnI/EcoRI fragment of pBL into the pGEX4T vector. pGST/NFMhr for GST fused to the head-rod domain (411 amino acids of the amino-terminal region) of NFM was made by polymerase chain reaction amplification from a human hippocampus cDNA library and subcloning into the pGEX4T vector. pGST/NFLf for GST fused to the full length of NFL was constructed by subcloning the BamHI/EcoRI insert of pBL into the pGEX4T vector. pGST/NFMt for GST fused to the tail domain of NFM was made by subcloning the ϳ1.0-kilobase pair XhoI/NotI fragment of pBM (22) (kindly provided by Dr. N. Hirokawa) into the pGEX4T vector. pGST/NFHf for GST fused to the full length of NFH was constructed by digesting pBH with BglII, and the cDNA insert was subcloned into the BamHI restriction site of the pGEX4T vector. pGST/NFHhr for GST fused to the head-rod domain of NFH was made by digesting pGST/NFHf with Tth111I/EcoRI, filling in the ends with T4 DNA polymerase, and ligating them together with T4 DNA ligase. This removed the DNA sequences 3Ј-terminal to the Tth111I site, which encodes the tail domain of NFH. pGST/NFHt for GST fused to the tail domain of NFH was made by digesting pBH with Tth111I/BglII, filling in the ends, and subcloning the ϳ2-kilobase pair fragment into the 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 8.8 containing 1 mM EDTA, 1 mM DTT, and 0.1 g/ml leupeptin.
In Vitro Binding Assay--For the in vitro NF binding experiment, 2.5 l of in vitro translated PKNN2 or PKNC2 was mixed with 5 g of each GST-NF fusion protein or with 25 g of GST alone in 400 l of GST binding buffer (20 mM Tris/HCl at pH 7.5, 0.5 mM DTT, 150 mM NaCl, 0.05% Triton X-100, 1 mM EDTA, 1 g/ml leupeptin) and incubated for 1 h at 4°C. After the addition of 25 l of glutathione-Sepharose 4B pretreated with 10 mg/ml Escherichia coli extract to block nonspecific binding, the binding reaction was continued for an additional 30 min at 4°C. The glutathione-Sepharose 4B was then washed three times in GST wash buffer (20 mM Tris/HCl at pH 7.5, 0.5 mM DTT, 1 mM EDTA, 1 g/ml leupeptin) containing 0.5 M NaCl and 0.5% Triton X-100 and washed further with GST wash buffer. Bound proteins were eluted with GST elution buffer (100 mM Tris at pH 8.0, 20 mM glutathione, 120 mM NaCl, 1 mM EDTA, 1 mM DTT, 1 g/ml leupeptin) and subjected to 10% SDS-PAGE. The binding was visualized and quantitated by an imaging analyzer (Fuji BAS1000).
Preparation of NF Proteins from Native Bovine Tissues-NF proteins were prepared from bovine spinal cords as reported (23).
Kinase Assay-Soluble cytosolic extract of spinal cord or purified NF proteins were boiled for 5 min to destroy endogenous protein kinase activity before use as phosphate acceptors. The phosphorylation of NF preparations was carried out at 30°C in an assay mixture containing 20 mM Tris/HCl at pH 7.5, 8 mM MgCl 2 , 100 M ATP, 185 kBq of [␥-32 P]ATP, phosphate acceptors, 20 ng/ml PKN purified from rat testis (6), with/without 40 M arachidonic acid as indicated in each experiment. After incubation for various times, the reaction was terminated by the addition of an equal volume of Laemmli's sample buffer and separated on 7% SDS-polyacrylamide gels. The gels were dried under vacuum, and the phosphorylation was visualized by an imaging analyzer (Fuji BAS1000). Dephosphorylation of NF proteins was conducted with calf intestine alkaline phosphatase as described (24), and NF proteins were boiled for 5 min to destroy phosphatase activity.
Effects of Phosphorylation on the Polymerization of NFL-For the in vitro NF polymerization experiment, phosphorylated forms of bacterially synthesized GST fused to the full length of NFL and GST fused to NFL#21 were prepared by incubation of 5 g of these proteins with 1 mM MgCl 2 , 60 ng of PKN, and 100 M ATP for 2 h at 30°C. Bacterially synthesized proteins incubated with PKN in the absence of ATP were employed as a nonphosphorylated control. 2.5 l of in vitro translated NFL was added to the above mixtures in depolymerization buffer (20 mM Tris/HCl at pH 8.5, 1 mM DTT, 1 g/ml leupeptin, 1 mM MgCl 2 ), then the pH of the reaction mixture was shifted to 7.2 by the addition of appropriate volumes of 1 M PIPES at pH 6.8 and incubated for 1 h at 35°C. After adding 25 l of glutathione-Sepharose 4B pretreated with 10 mg/ml E. coli extract to block nonspecific binding, the binding reactions were continued for an additional 30 min at 4°C. The glutathione-Sepharose 4B was then washed twice in depolymerization buffer containing 0.5% Triton X-100 and washed further with depolymerization buffer. Bound proteins were eluted with GST elution buffer and subjected to 10% SDS-PAGE. Quantitation of the binding reactions was carried out by an imaging analyzer (Fuji BAS1000). protein. Specificity of this interaction was tested further by measuring the ability of other combinations of the two-hybrid constructs, LexAbd (instead of Gal4bd)-PKN and VP16ad (instead of Gal4ad)-NF, or LexAbd-NF and VP16ad-PKN, to support lacZ expression in L40 cells. Fig. 2 summarizes these data and demonstrates a specific interaction between the aminoterminal region of PKN and the head-rod domain of NFL. To characterize further the interaction between PKN and NFL, we tested the ability of in vitro translated PKN to bind to GST-NFL in an in vitro binding assay. As shown in Fig. 3A, the amino-terminal region of PKN bound to the head-rod domain of NFL, and little binding of the carboxyl-terminal catalytic domain of PKN to the head-rod domain of NFL was observed in this assay.

Isolation of PKN-binding Proteins
We reported previously that RhoA binds to the amino-terminal region of PKN. Given the interaction of NFL with the amino-terminal regulatory region of PKN, it was possible that this protein interacts with the Rho binding domain of PKN and might compete with Rho for binding to PKN. In an in vitro binding assay, bacterially synthesized NFL did not compete with the bacterially synthesized RhoA for the binding to the amino-terminal region of PKN (data not shown).
The Amino-terminal Region of PKN Binds to the Head-Rod Domain of Each Subunit of NF-NF was composed of three subunits of low (66 kDa; NFL), medium (130 kDa; NFM), and high (180 kDa; NFH) molecular masses (25,26). The ␣-helical rod domain of NF protein is conserved throughout these subunits of NF and the family of intermediate filaments, such as vimentin, and confers common structural characteristics to the polymers (27). To determine whether the amino-terminal region of PKN binds to NFM or NFH as well as NFL, corresponding regions of the head-rod domain of NFM and NFH were also ligated to pVP16, and LacZ expression was measured in the two-hybrid system. The amino-terminal region of PKN bound to the head-rod domain of each subunit of NF (Fig. 2B) and also interacted with the corresponding head-rod domain of vimentin, a ubiquitously expressed member of intermediate filament, in the two-hybrid system (cDNA of vimentin was kindly provided by Dr. M. Inagaki, Tokyo Metropolitan Institute of Gerontology; data not shown). We also tested the ability of in vitro translated PKN to bind to each subunit of NF and found that PKN bound directly to each subunit of NF at the head-rod domain in vitro (Fig. 3B).
Since NFL bound to the regulatory region of PKN, a possibility was raised that NFL affects the kinase activity of PKN. To test this possibility, we added various concentrations of the bacterially synthesized NFL to the PKN assay mixture and examined the phosphorylation activity using a synthetic peptide based on the pseudosubstrate site of PKC as a substrate (1). We found no obvious effect of NFL on the peptide phosphorylation activity of PKN (data not shown).
PKN Phosphorylates the Head-Rod Domain of Each Subunit of NF-To test the ability of PKN to phosphorylate each subunit of NF, NF purified from spinal cord was subjected to in vitro phosphorylation by PKN. PKN phosphorylated efficiently all three NF subunits (Fig. 4). Initial velocities of phosphorylation of each subunit by PKN in the presence of arachidonic acid were ϳ5-10 times higher than those in the absence of modifier (Fig. 4A). In the presence of arachidonic acid, phosphate incorporation into each subunit of NF reached a maximum at ϳ60 min and then continued to plateau up to 120 min (Fig. 4, B and C). Although purified PKN was labile in its diluted condition, it was unlikely that the decline in phosphorylation speed was mainly the result of the inactivation of PKN because the phosphorylation level did not decrease appreciably from 60 to 120 min in the absence of arachidonic acid (data not shown). The maximal phosphorylation by PKN per mol of protein subunit was estimated by image quantitation to be ϳ2 mol/mol of NFH, ϳ6 mol/mol of NFM, and ϳ1 mol/mol of NFL, respectively. The NFH, which was reported to be the most intensely radiolabeled subunit in vivo (28), was a relatively poor substrate for PKN in vitro compared with NFM. It seemed possible that potential phosphorylation sites of NFH and NFM for PKN had already been masked since these proteins were purified from bovine tissue. Therefore, we tested whether we could reveal additional phosphorylation sites using enzymatically dephosphorylated NF. As shown in Fig. 4D, the change of electrophoretic mobility was accompanied by the dephosphorylation of NFH and NFM (24). Dephosphorylation of bovine NF with alkaline phosphatase did not result in any significant difference in phosphorylation of the NFH and NFM subunits by PKN, suggesting that native NF contains sites that are accessible to phosphorylation (Fig. 4, B and C).
To examine the phosphorylation site of NF by PKN, we prepared bacterially synthesized GST fused to the head-rod domain and GST fused to the tail domain of each subunit of NF and subjected them to an in vitro phosphorylation assay. As shown in Fig , lanes 3 and 7). The position of labeled protein is indicated by a white arrowhead.
sites were located exclusively in the head-rod domain of these intermediate filaments.
The Effect of Phosphorylation on Filamentous Structure of NFL in Vitro-It is widely accepted that the NFL subunit forms the "core" of the NF, suggesting that NFL may assemble first and may then provide the signal as well as the scaffold for coassembly or polymerization of NFM and NFH. It has been shown in vitro that phosphorylation of NFL by PKA and PKC inhibited its polymerization and also depolymerized the filaments (29 -31).
We investigated whether NFL polymerizes in an in vitro binding analysis. The bacterially produced GST fused to the head-rod domain of NFL or GST fused to the full length of NFL was mixed with in vitro translated NFL in buffer at pH 8.5 with 1 mM MgCl 2 . Then the pH of the reaction mixture was shifted to 7.2 and incubated for 1 h at 35°C. After extensive washing,  proteins bound to the beads were analyzed by autoradiography. As shown in Fig. 6, polymerization of NFL was clearly detected. Next, we determined whether phosphorylation of NFL by PKN inhibits the polymerization of NFL in this assay system. The GST fused to the head-rod domain of NFL and GST fused to the full length of NFL were phosphorylated by PKN, transferred to the reaction mixture, and mixed with in vitro translated NFL. As shown in Fig. 6, the binding of the head-rod domain and the full length of NFL to in vitro translated NFL was very weakly detected, indicating that phosphorylation of NFL by PKN inhibited the polymerization of NFL. DISCUSSION Recently, increasing evidence has accumulated to suggest that continuous turnover of intermediate filaments is involved in maintaining the intermediate filament structure and function in living cells. Transfection of native and mutated intermediate filament genes or microinjection of biotin-labeled intermediate filaments has revealed the successive incorporation of newly synthesized or microinjected intermediate filament proteins into preexisting intermediate filament networks (32)(33)(34)(35)(36)(37) and the disruption of endogenous intermediate filament networks by the integration of assembly-incompetent mutant proteins (38 -40). NF, a neuronal intermediate filament, was also reported to have a dynamic structure, which turns over within a small region of the axoplasm by exchanging subunits at discrete sites on the filaments (41). One of the possible mechanisms of the growth-dependent regulation of NF turnover is the phosphorylation of NF triplet proteins. Protein kinases responsible for the phosphorylation of NF subunits in the intact tissue are yet unclear. Several laboratories have directed their attention to phosphorylating activities that remain bound to vertebrate NFs through purification procedures (42)(43)(44)(45). Wible et al. (46) reported the isolation of a NF-associated protein kinase of 67 kDa from bovine spinal cord which phosphorylates NF. PKN is clearly distinguished from this protein kinase by molecular weight, by their marked preference for the tail domain of NFH, and by their lack of autophosphorylation activity. Protein kinases that phosphorylate all three subunits of NF were partially purified by several other laboratories, although their molecular weights and structural characteristics are unknown. One of these, partially purified by Toru-Delbauffe and Pierre (47) and Toru-Delbauffe et al. (48), has a high specificity for NFH and is insensitive to 5 mM Ca 2ϩ . Another one, purified by Dosemeci et al. (49), has marked preference for casein rather than histone or NF as substrate.
Another, purified by Hollander and Bennett (50), phosphorylates efficiently the carboxyl-terminal tail domain of NFM. Judging from these characteristics, PKN seems to be different from these kinases. Recently, Xiao and Monteiro (51) detected four prominent kinases (NAKs) with molecular masses of 115, 95, 89, 84 kDa by in situ gel kinase assay in an affinity-purified fraction by using recombinant NFH from mouse brain extract (51). The 115-kDa molecular mass of NAK is close to that of human PKN, which is120 kDa (2). These investigators reported that NAKs phosphorylate all three NF subunits and that partial dephosphorylation of NF with alkaline phosphatase does not affect the phosphorylation of the NFH and NFM by NAKs. Thus PKN is similar in this respect. However, PKN bound to the head-rod domain of NFH, and NAKs were found on the basis of their retention by the tail domain of the NFH affinity column. Purification and information about molecular aspects of NAKs are expected. Dosemeci and co-workers (49,52) reported that bovine NF preparation contains known Ca 2ϩ /calmodulin-dependent kinase, PKC, and PKA as well as activatorindependent kinase activities. In vitro phosphorylation of NF at the amino-terminal domain by PKA or PKC inhibits the assembly process of NF and induces disassembly of preexisting filaments (53). In the present study, we revealed that PKN phosphorylated NF at the amino-terminal domain, was able to inhibit the assembly of NFL in vitro, and that PKN was another candidate for being a regulator of NF assembly.
We reported here that the regulatory domain of PKN binds to the head-rod domain of NF and vimentin. What is the meaning of this association? Although additional evidence is required for determining whether PKN phosphorylates NF in vivo, one possible explanation of the interaction between the regulatory domain of PKN and NF is that an association interface outside the catalytic domain is required for substrate recognition and could serve to target PKN to physiologically important substrates. Precedence for protein association outside the conserved catalytic domain of protein kinase as an important targeting mechanism during signal transduction is provided by studies of Src homology domains (54 -57) and the regulatory region of PKC (58,59). In our preliminary experiment, some clones that were isolated by this two-hybrid screening were indicated to be phosphorylated in vitro by PKN. It is possible that PKN uses its regulatory domain as an association domain for all physiologically important substrates. An alternative possibility is that the binding of PKN to one substrate serves to promote the phosphorylation of other proteins in the same complex or that NF acts only as an adaptor that promotes the activity of PKN toward adjacent specific cytoskeletal targets.
We recently found that PKN directly binds to the GTP-bound form of Ras-related small GTPase Rho and is activated by this binding both in a cell-free system and in cultured cells (7,8). Rho works as a molecular switch in diverse cellular processes such as stimulus-evoked cell adhesion (60 -62), cell motility (63), regulation of smooth muscle contraction (64), and cytokinesis (65,66). Rho was presented to exert these actions by inducing certain types of cytoskeletal structures (60,65,66). Paterson et al. (67) reported the collapse of the intermediate filament into irregular thick bundles after microinjection of constitutive activated Rho (Val 14 Rho) into subconfluent Swiss 3T3 cells; however, the biochemical basis of the Rho effect is not known. From the data presented in this paper, PKN, positioned in the downstream of Rho signal, might regulate the assemblydisassembly of intermediate filaments.