Hyperosmotic-induced protein kinase N 1 activation in a vesicular compartment is dependent upon Rac1 and 3-phosphoinositide-dependent kinase 1.

Protein kinase N 1 (PKN1), which in part resembles yeast protein kinase C, has been shown to be under the control of Rho GTPases and 3-phosphoinositide-dependent kinase 1 (PDK1). We show here that green fluorescent protein-tagged PKN1 has the ability to translocate in a reversible manner to a vesicular compartment following hyperosmotic stress. PKN1 kinase activity is not necessary for this translocation, and in fact the PKN inhibitor HA1077 is also shown to induce PKN1 vesicle accumulation. PKN1 translocation is dependent on Rac1 activation, although the GTPase binding HR1abc domain is not sufficient for this recruitment. The PKN1 kinase domain, however, localizes constitutively to this compartment, and we demonstrate that this behavior is selective for PKNs. Associated with vesicle recruitment, PKN1 is shown to undergo activation loop phosphorylation and activation. It is established that this activation pathway involves PDK1, which is shown to be recruited to this PKN1-positive compartment upon hyperosmotic stress. Taken together, our findings present a pathway for the selective hyperosmotic-induced Rac1-dependent PKN1 translocation and PDK1-dependent activation.

Hyperosmotic stress is established as a potent activator of several signaling cascades, including stress-activated protein kinases (1), p38 (2), and extracellular signal-regulated kinases (3). In contrast, protein kinase B (PKB) 1 is shown to be downregulated by hyperosmotic stress via dephosphorylation of its regulatory Thr-308 and Ser-473 phosphorylation sites (4). Interestingly, the yeast PKC homologues, which in part resemble the PKNs, have been established as essential for the maintenance of cell wall integrity (5). The regulation of cell wall integrity has been likened to the hyperosmotic stress response in Saccharomyces cerevisiae and has suggested roles for PKCs in high osmolarity responses (6). Notably, Mkh1, a yeast MEK kinase, is also shown to be activated and indeed required for cell wall integrity and a normal response to osmotic stress in Schizosaccharomyces pombe, further establishing the link between cell wall integrity and the osmotic stress response (7). Classical and novel PKC activation is suggested to be a requirement for hyperosmotic induced extracellular signal-regulated kinase activation (8). This investigation has sought to address the involvement of PKN1 in the hyperosmotic stress response in mammalian cells.
PKNs (protein kinase novel, also known as PRKs 2 (9,10)) are a subfamily of serine/threonine kinases identified independently by molecular cloning, protein purification, and PCRbased screens for PKC related kinases. The carboxyl-terminal kinase domains of these proteins are closely related to those of PKC, and at their amino termini they have a conserved repeated domain (HR1a,b,c) followed by a C2-related domain; overall these proteins have a domain organization related to that of the yeast PKC-related proteins (11). PKNs are activated by fatty acids and phospholipids in vitro, although the in vivo significance of this remains unclear (12,13). The amino-terminal HR1 domain was identified as a Rho interacting region (14 -16), and RhoB has been shown to target PKN1 to an endosomal compartment where it is implicated in controlling the kinetics of epidermal growth factor receptor traffic (17,18).
The interaction of Rho with PKN1 has been demonstrated to facilitate PKN1 activation loop phosphorylation by 3-phosphoinositide-dependent kinase 1 (PDK1). PDK1 was originally purified as an activity responsible for PKB␣ activation loop phosphorylation (19,20). PDK1 has been demonstrated more recently to phosphorylate equivalent residues on many other AGC kinases, including p70 S6k , cAMP-dependent protein kinase, and PKCs (reviewed in Ref. 21). Based upon co-transfection experiments, the in vivo ternary complex of Rho-PKN1-PDK1 has been shown to be dependent on PI 3-kinase activity and to be critical for the catalytic activation of PKN1 (22). PKN1 has been linked to stress-induced pathways because it has been implicated upstream of c-Jun transcription via p38␥ (23), both are activated upon hyperosmotic stress. Another relevant PKN response involves Fyn tyrosine kinase, which has been shown to mediate PKN2 function in keratinocytes (24) and has recently been shown (25) to be essential in transcription from the osmotic response element.
Here we describe the acute translocation of GFP-PKN1 to vesicles in response to hyperosmotic stress. It is established that PKN1 translocation is dependent on Rac1 activation, and the kinase domain of PKN1 is shown to be an essential component of this response. The activity of PKN1 is found not to be required for vesicle recruitment, although activation loop phos-phorylation and catalytic activation of PKN1 is shown to increase upon hyperosmotic stress. This activation is catalyzed by PDK1, which is recruited to the PKN1 compartment in a PI 3-kinase-and PKN1-dependent manner. Thus a pathway is established that leads from the plasma membrane activation of Rac1 to the vesicular accumulation of an activated PKN1 complex.

EXPERIMENTAL PROCEDURES
Materials-Anti-PKN1 monoclonal antibodies were from Transduction Laboratories, and the anti-phospho-PKN polyclonal antibody was as described previously (22). For immunofluorescence, the anti-Myc 9E10 monoclonal antibody was from Cancer Research UK monoclonal services; Cy3 anti-mouse antibody was from Jackson ImmunoResearch, and anti-HA polyclonal antibody was from Santa Cruz Biotechnology. The Rac1 activation assay was purchased as a kit from Totam Biologicals. Anti-GFP polyclonal antibody used for immunoprecipitation was purchased from Clontech, and phosphatidylserine lipids were from Lipid Products. The PI 3-kinase inhibitor LY294002 and the ROCK/ PKN inhibitor HA1077 were purchased from Calbiochem.
Plasmid Constructs-GFP-PKN1 and GFP-PKN2 constructs were generated by PCR using as templates PKN1 and PKN2 constructs, respectively. PKN1 was amplified using primer 1, 5Ј-GCGCAAGCTT-GCATGGCCAGCGACGCCGTGCAGAGTGAGCC-3Ј; primer 2, 5Ј-GC-GCGGTACCTTAGCAGCCCCCGGCCACGAAGTCGAAGTCCAGGAA-GGC-3Ј; for PKN2 primer 3, 5Ј-GCGCAAGCTTCGATGGCGTCCAAC-CCCGAACGG-3Ј; and primer 4, 5Ј-GCGCGGTACCTTAACACCAATC-AGCAATGTAGTCAAAATCTCTGAACATTTCCTGC-3Ј. Primers 1 and 3 incorporated a HindIII restriction site, and primers 2 and 4 contained a KpnI site. The resulting products contained the full-length PKN sequences and were cloned into PCR blunt (Invitrogen). This was subsequently digested with HindIII and KpnI and cloned into pEGFP-C1 (Clontech) enabling the fusion of an amino-terminal GFP tag in-frame with the full-length PKN1 and PKN2. The DS-Red PKN1 was constructed in the same manner as GFP-PKN1, using the pDS-Red-C1 vector from Clontech. HA-HR1abc has been described previously (17); GFP-PKC⑀ was as described (26), and the GFP-PKC␣ kinase domain was derived from the full-length cDNA. 3 The Myc-tagged PKN1 kinase domain was derived from the full-length cDNA. 4 GFP-PKN1-KR, a kinase-dead mutant, was constructed by cloning the kinase-dead K644R mutation from the kinase-dead kinase domain construct 3 into the EcoRI/HincII sites of the GFP-PKN construct. GFP-PDK1 was a gift from Dr. Dario Alessi, Dundee, UK, and the Myc-tagged Rac1 wild type and N17 plasmids were gifts from Dr. Alan Hall, London, UK.
Cell Culture and Transfection-NIH-3T3 cells were maintained in Dulbecco's modified Eagle's medium and supplemented with 10% fetal calf serum (10% E4). Transient transfections were performed using LipofectAMINE 2000 (Invitrogen). The lipid DNA mix was incubated for 6 h according to the manufacturer's protocol, and all subsequent manipulations (see text and figure legends) were performed 48 h post-transfection.
Hyperosmotic Stress-Cells were subjected to hyperosmotic conditions for 30 min by the addition of Dulbecco's modified Eagle's medium containing 0.4 M sucrose and 25 mM Hepes, pH 7.2.
For recovery experiments, cells were shocked for 30 min, and then medium was replaced with normosmotic medium. The average number of vesicles after 30 min of hyperosmotic stress and subsequent 15 and 30 min of recovery was determined. From each condition, 10 GFP-PKN1 vesicle-containing cells were sampled. Images were taken as confocal sections through the center of the cell, and the number of vesicles from each image was counted.
Microscopy-Cells were seeded on acid-washed glass coverslips, and transfection and stimulation were as indicated in the figure legends. Cells were then washed and fixed in 4% paraformaldehyde for 15 min. When processing for immunofluorescence, cells were permeabilized with 0.2% (v/v) Triton X-100 for 5 min, washed, and incubated in 1% (w/v) bovine serum albumin for 20 min. Coverslips were incubated with primary antibody containing 1% (w/v) bovine serum albumin for 1 h. Cells were then washed and incubated with fluorescent dye-conjugated secondary antibody for 1 h and washed three times with the final wash in water. All washes and incubations were performed in phosphatebuffered saline unless otherwise stated. Cells were mounted on glass slides under Mowiol (100 mM Tris-HCl, pH 8.8, 10% (w/v) Mowiol 4-88 (Calbiochem), and 25% (v/v) glycerol) containing anti-photobleaching agent (2.5% (w/v) 1,4-diazabicyclo[2.2.2]octane (Sigma)). Slides were examined using a confocal laser scanning microscope (Axioplan2 with LSM 510; Carl Zeiss Inc.) equipped with 63ϫ/1.4 Plan-APOCHROMAT oil immersion objectives. GFP and Cy3 were excited with 488-and 543-nm lines of Kr-Ar lasers, respectively, and individual channels scanned in series to prevent bleed through. Each image represents a single 1.0-m "Z" optical section.
Rac1 Activation-NIH-3T3 cells (500,000 cells per 6-cm dish) were seeded and grown in 10% fetal calf serum E4 and subsequently starved for 16 h with 0.05% E4. Cells were then subjected to treatments of hyperosmotic stress as indicated in the figure legends. The Rac activation assay was performed on cell extracts using reagents contained in the kit from Totam Biologicals. This assay employs the p21 binding domain (PBD or CRIB domain) of p21-activated kinase 1 (PAK) as a GST fusion protein that selectively binds GTP-Rac1 in a "pull-down" assay enabling the detection of activated Rac1. All samples were prepared and processed according to the manufacturer's guidelines. Quantitation was obtained by image capture and densitometry using NIH TM image software.
Phosphorylation of GFP-PKN1-Cells were transiently transfected on 6-cm 2 plates with GFP-PKN1 and subject to hyperosmotic stress as indicated. Cells were harvested in 400 l of sample buffer and fractionated on a 10% SDS-polyacrylamide gel. Activation loop Thr-774 phosphorylation of PKN1 was assessed by Western blotting using a phospho-specific polyclonal antibody with excess of dephospho-peptide as described (22). Relative PKN1 phosphorylation was determined as a function of PKN1 protein levels defined using the PKN1 monoclonal antibody. GFP-PKN1 phosphorylation is taken from three duplicate experiments, where represented graphically the error bars denote the S.E.
GFP-PKN1 Kinase Assay-For each condition a 15-cm 2 plate of cells was transfected with GFP-PKN1 and stimulated with hyperosmotic medium as indicated in the figure legends. Cells were harvested in a lysis buffer comprising 20 mM Tris-HCl, pH 7.5, 0.5% (w/v) CHAPS, 1 Complete tablet (protease inhibitor from Roche Applied Science) per 50 ml of lysis buffer, 2 mM EDTA, 150 mM NaCl. Preclearance of cell lysates with pre-equilibrated protein G was followed by incubation with 9 g of GFP polyclonal antibody for 20 min at 4°C. Immunocomplexes were captured by incubation for 1 h with protein G. Immunoprecipitates were washed three times with lysis buffer, once with 0.5 M NaCl, and a final wash in reaction buffer, 20 mM Hepes, pH 7.5, 10 mM MgCl 2 . For each reaction, 10 l of immunopurified GFP-PKN1 was incubated in a 40-l reaction mix of 20 mM Tris-HCl, pH 7.5, 10 mM MgCl 2 , 25 g of bath-sonicated phosphatidyl-L-serine, 10 g of myelin basic protein (MBP), 50 M ATP, and 1 Ci of [␥-32 P]ATP for 20 min at 30°C with agitation. Reactions were terminated by the addition of 40 l of sample buffer prior to fractionation on a 15% SDS-polyacrylamide gel, and subsequently Western blotting analysis was performed. To assess MBP phosphorylation, membranes (Immobilon-P from Millipore) were exposed to PhosphorImager (STORM, Amersham Biosciences) prior to immunoblotting with anti-PKN1. Relative specific activity was determined as a function of immunoreactive GFP-PKN1. The data presented are representative of three duplicate experiments, and where represented graphically, error bars denote the S.E. Additional duplicate reactions incorporated 20 M HA1077 (a PKN inhibitor) as a control for specific PKN activity (27,28).

RESULTS
GFP-PKN1 expressed in NIH-3T3 cells displayed a punctate cytoplasmic localization with an absence of nuclear localization. The effect of a range of potential PKN1 agonists on the subcellular distribution of GFP-PKN1 was investigated including the following: epidermal growth factor, fibroblast growth factor, insulin, platelet-derived growth factor, and stress conditions such as oxidative, temperature, hypo-and hyperosmotic stress. Notably, cells treated with hyperosmotic media for 30 min displayed a dramatic change in localization, with GFP-PKN1 accumulating in large cytoplasmic vesicular structures (Fig. 1A).
The dynamic nature of the hyperosmotic-induced structures was investigated by placing cells back into osmotically balanced media following 30 min of hyperosmotic stress. Cells containing GFP-PKN1 vesicles were randomly selected, and sectional images taken by confocal microscopy. These vesicles were clearly dissipated on removal from hyperosmotic conditions; quantitation showed that the average number of vesicles per cell decreased in a time-dependent fashion after re-addition of osmotically balanced medium. It is evident therefore that the accumulation of vesicular PKN1 is a reversible process (Fig. 1B).
GFP-PKC⑀ has been shown to accumulate in vesicular structures following chronic PKC inhibition in MEF cells (26). To investigate the selectivity of the PKN1 translocation in response to hyperosmotic stress, DS-Red PKN1 and GFP-PKC⑀ were co-expressed. DS-Red-PKN1 has been co-expressed with GFP-PKN1 to confirm that they show the same localization behavior before and after hyperosmotic stress (data not shown). The ability of DS-Red PKN1 to translocate into vesicles upon hyperosmotic stress was unaffected by the presence of GFP-PKC⑀; furthermore, PKC⑀ did not locate to the DS-Red PKN1containing structures (Fig. 1C), indicating that the observed behavior of PKN1 in response to hyperosmotic stress shows some selectivity. The ability of other PKN subfamily members to respond to hyperosmotic stress was also followed. GFP-PKN2 when unstimulated was largely cytoplasmic with some limited nuclear accumulation (Fig. 1D). After hyperosmotic stress, GFP-PKN2 showed the same translocation pattern as GFP-PKN1 (Fig. 1D), indicating that the described behavior is a common PKN response. The influence of catalytic activity on this process was investigated by employing the drug HA1077, which has been shown to inhibit both PKN1 and PKN2 (27,28). Pretreatment of GFP-PKN1-transfected cells with HA1077 followed by hyperosmotic stress did not prevent PKN1 vesicle recruitment. Interestingly cells treated with HA1077 alone displayed an accumulation of vesicular GFP-PKN1 ( Fig. 2A). To investigate further the contribution of PKN1 activity, we performed localization studies using the kinase-dead GFP-PKN-K644R mutant. When unstimulated this inactive PKN1 mutant shows a punctate cytoplasmic distribution, and after hyperosmotic stress accumulates in vesicles ( Fig. 2A) as observed for wild type GFP-PKN1. DS-Red-PKN1 and GFP-PKN1-KR were coexpressed to investigate whether either form of PKN1 could behave in a dominant fashion with respect to vesicle recruitment. Both wild type and kinase-dead PKN1 were cytoplasmic in unstimulated cells and co-localized in vesicles upon hyperosmotic stress (Fig. 2B). These findings imply that the kinase activity of PKN1 is not necessary for vesicle recruitment. However, given the effect of HA1077 in unstimulated cells, PKN1 may play a role in the exit from this vesicular compartment, which itself may be part of a constitutive endocytic pathway (see below and "Discussion").
PKN has been shown to bind to and become activated by members of the Rho family of small GTPases via the regulatory HR1 domain and indeed has been shown previously to be recruited to an endosomal compartment by RhoB (17). We tested the involvement of Rho proteins in the PKN response to hyperosmotic stress by employing the C3 toxin from Clostridium botulinum. C3 toxin has been shown to inhibit specifically the Rho subfamily over other Rho family GTPases such as Rac1 (29). NIH-3T3 cells were transfected with GFP-PKN1 and then pre-loaded with C3-GST for 6 h at 5 g/ml followed by subjection to hyperosmotic stress. The effectiveness of C3-GST cell loading was measured by an in vitro ribosylation assay and also by monitoring the extent of stress fiber disruption assessed by phalloidin staining (data not shown). The accumulation of PKN1-containing vesicles was independent of C3-GST treatment, implying that Rho function is not required for this response (data not shown).
The small GTPase Rac1 has been implicated previously in the control of PKN2 (16) and recently in response to hyperosmotic shock (30). Hence in the absence of a Rho input, the potential involvement of Rac1 in the translocation of PKN1 was investigated. Myc-tagged Rac1 was co-expressed with GFP-PKN1. Under control conditions GFP-PKN1 was cytoplasmic, and Rac1 located to lamellipodia, the plasma membrane, and to some vesicular structures (Fig. 3A). After 30 min of hyperosmotic stress, Myc-Rac1 and GFP-PKN1 co-localized in vesicles (Fig. 3A). Co-expression of the dominant negative Myc-Rac1-N17 with GFP-PKN1 resulted in no vesicular translocation of GFP-PKN1 after hyperosmotic stress (Fig. 3B). These observations imply that GTP loading of Rac1 is necessary for the hyperosmotic stress-induced movement of GFP-PKN1. Rac1 has been shown recently to become GTP-loaded upon hyperosmotic stress in neutrophils (30). This was confirmed here by employing a Rac1 pull-down assay, using the CRIB domain of PAK that selectively binds GTP-bound Rac1. After 30 min of hyperosmotic stress, a 1.5-fold increase in the GTP loading of endogenous Rac1 was detected (Fig. 3D).
Rac1, like Rho, binds to the HR1 domain of PKN proteins (31). Thus the impact of the regulatory HR1abc region on hyperosmotic stress-induced vesicle recruitment was investigated through localization studies of the ectopically expressed HA-tagged HR1abc domain of PKN1. This domain localized around the cell periphery under control conditions, and after hyperosmotic stress did not change its distribution. HA-HR1abc was co-expressed with full-length GFP-PKN1 to determine whether their combined presence would influence each other's localization. The distribution of the HR1abc domain remained unchanged when co-expressed with GFP-PKN1 either before or after hyperosmotic stress. The presence of HR1abc did not abolish GFP-PKN1 vesicle recruitment (Fig.  3C). The evidence thus indicates that Rac1 recruits PKN1 at the plasma membrane where HR1abc can compete for interaction and that the PKN1-Rac1 complex but not the HR1abc-Rac1 complex is then endocytosed. We speculate that because HR1abc was unable to prevent GFP-PKN1 vesicle association, either endogenous Rac1 remains in excess under these conditions or GFP-PKN1 can compete effectively for Rac1 in the presence of co-expressed HR1abc. Irrespective of the lack of the competition between these two, the behavior of the HR1abc domain indicates that although the regulatory input of Rac1 into PKN1 might be essential in priming PKN1 for endocytosis/ vesicle recruitment, it cannot be the sole device by which PKN1 vesicle association is mediated.
To assess the potential contribution of other PKN domains in vesicle recruitment, the myc-PKN1 kinase domain was trans- fected into NIH-3T3 cells. Under control conditions, myc-PKN1 kinase domain already displayed a partially vesicular distribution, and upon hyperosmotic stress, this vesicular localization became more pronounced (Fig. 4A). To test whether the kinase domain locates to the same structures as the full-length protein, the Myc kinase domain was co-expressed with GFP-PKN1; after hyperosmotic stress they were found to co-localize in large vesicular structures (Fig. 4B). The observation that the PKN1 kinase domain can by itself localize to vesicles raises the possibility that the kinase domains of highly related PKC isoforms could behave in a similar manner. Myc-PKN1 kinase domain was co-expressed with the GFP-PKC␣ kinase domain. After hyperosmotic stress the kinase domain of PKN1 could be detected in vesicles; in these structures the kinase domain of PKC␣ was absent (Fig. 4C). These data indicate that vesicle targeting of PKN1 through the kinase domain is specific for this compartment and must follow distinct mechanisms when compared with a closely related kinase domain.
The HA1077 effects on the basal state and osmotic induced distribution of PKN1 ( Fig. 2A) imply that its activity could be required for vesicle turnover. However, it is not clear whether activation of PKN takes place in response to osmotic stress. Activation loop phosphorylation is required for optimum catalytic activity of PKN1 (22). Phospho-specific polyclonal antibodies were used to assess the effect of hyperosmotic shock on GFP-PKN1 activation loop phosphorylation. A 2-fold increase in phos-  phorylation after 30 min of hyperosmotic stress was detected on GFP-PKN1 (Fig. 5A). We were also able to detect an increase in endogenous PKN1 phosphorylation after hyperosmotic shock.
Immunoprecipitated GFP-PKN1 from transiently transfected NIH-3T3 cells, either prior to or post-hyperosmotic shock, was used to determine directly the effect of osmotic stress on activity. After hyperosmotic shock, immunoprecipitated GFP-PKN1 displayed a specific activity ϳ2-fold above un-shocked GFP-PKN1 (Fig. 5B). The possibility of associated kinases being responsible for this activity was precluded with parallel control assays incorporating the PKN inhibitor HA1077. This inhibitor has been reported to be specific for PKN1 over kinases that can associate with it (e.g. PDK1, see below) (27,28). Immunoprecipitated GFP-PKN1 activity after hyperosmotic shock was reduced to background levels in the presence of 20 M HA1077. PDK1 has been shown previously to bind to and facilitate the activation loop phosphorylation of PKN1 (22). Given the finding that PKN1 is both phosphorylated and activated after osmotic stress, we assessed the potential involvement of PDK1 in the control of vesicular PKN1. GFP-PDK1 was co-expressed with DS-Red-PKN1. The localization of both proteins was cytoplasmic in unstimulated cells, but after hyperosmotic stress, GFP-PDK1 and DS-Red-PKN1 were found to be co-localized in vesicles (Fig. 6). PI 3-kinase influences PDK1 via its pleckstrin homology domain (32,33) and the subsequent phosphorylation and activation of PKN1 (22). We examined the effect on osmotic responses following pretreatment with the PI 3-kinase inhibitor LY294002. After hyperosmotic stress, GFP-PDK1 was no longer recruited to DS-Red-PKN1-positive vesicles. Notably, DS-Red-PKN1 was still recruited to vesicles; however, these were smaller after the pretreatment with LY294002 (Fig. 6). FIG. 5. GFP-PKN1 is phosphorylated and activated by hyperosmotic stress. A, NIH-3T3 cells were transiently transfected with GFP-PKN1 and harvested (1) unstimulated or (2) after 30 min of hyperosmotic stress as indicated. Activation loop Thr-774 phosphorylation of PRK1 was assessed by Western blotting using a phospho-specific polyclonal antibody. Relative PKN1 phosphorylation was determined as a function of PKN1 protein levels by Western blotting and analyzed using NIH Image TM . The relative GFP-PKN1 phosphorylation is taken from three experiments run in duplicate, and error bars denote the S.D. B, a PKN1 kinase assay was performed using immunopurified GFP-PKN1 isolated from transfected NIH-3T3 cells. Immunopurified GFP-PKN1 was incubated with 10 mM MgCl 2 , 25 g of phosphatidyl-L-serine, 10 g of myelin basic protein (MBP), and 50 M ATP for 20 min at 30°C (1 and 2). Untransfected cell lysate was used as a bead control for the immunopurification (3)(4)(5)(6). GFP-PKN1 was immunopurified from transfected NIH-3T3 cells (1)(2)(3)(4)(5)(6). Cells were either unstimulated or subject to 30 min of hyperosmotic stress prior to immunopurification as indicated (4 and 6). Additional duplicate reactions incorporated the PKN inhibitor 20 M HA1077 in vitro. All reactions were performed in parallel as duplicates. Relative specific activity was determined as a function of immunoreactive GFP-PKN1 determined by Western blotting from the same filter used for detection of MBP phosphorylation. Band intensities were analyzed using NIH Image TM , and error bars represent the S.E. (n ϭ 3).

DISCUSSION
The results described here demonstrate that PKN1 is acutely regulated by hyperosmotic shock, through the assembly of a vesicular complex with the upstream kinase PDK1. This process is triggered by the activation of Rac1, although this alone is found to be insufficient for accumulation of vesicular PKN1. A distinct site of interaction between PKN1 and the described compartment appears to be required, and this is consistent with the finding that the kinase domain itself is selectively targeted to vesicles. Based upon use of an inactive PKN1 mutant, it is shown directly that PKN1 catalytic activity is not itself required for vesicle recruitment, a finding corroborated by the vesicular accumulation of PKN1 in the presence of the catalytic inhibitor HA1077. Once vesicular, the subsequent recruitment of PDK1 leads to the increased activation loop phosphorylation of PKN1 that parallels its catalytic activation.
Recent studies (30) have reported the GTP loading of Rac1 in response to hyperosmotic stress, a finding confirmed here in a distinct cell type. In fact a hyperosmotic stress response has been described previously for another Rac1 effector, p21-activated protein kinase ␥-PAK which binds to and is activated by Rac1. ␥-PAK has been shown to translocate from a soluble to a particulate fraction and become activated in response to hyperosmolarity (34). Interestingly, it was demonstrated that the activation but not translocation of ␥-PAK was sensitive to wortmannin, suggesting a two-step mechanism for the ␥-PAK response to hyperosmotic stress. This parallels the situation described here for PKN1 where inhibition of PI 3-kinase does not prevent vesicle accumulation of PKN1, while blocking recruitment of the upstream kinase PDK1. It has been shown previously (22) that inhibition of PI 3-kinase will block activation loop phosphorylation of PKN1.
The nature of the induced PKN1 vesicular compartment described here is not resolved. Immunostaining indicates that this is not an early endosomal compartment (EEA1-negative) nor an acidified compartment (lysotracker negative; data not shown). However, the effect of PI 3-kinase inhibition on the induced PKN1-positive compartment would be consistent with this being part of an endocytic pathway, where inhibition of PI 3-kinase arrests homotypic fusion (35).
Based upon experiments with the ADP-ribosylation factor C3 toxin, it is concluded that Rho-GTP interaction is not required for the recruitment of PKN1 in response to hyperosmotic shock. However, because Rac1 has also been reported to bind the HR1abc domain of PKN1 and also because Rac1 becomes GTP-loaded in response to hyperosmotic shock, the possible role of this PKN effector was investigated. The ability of the dominant negative Rac1 to suppress PKN accumulation in the vesicular compartment indicates that Rac1 plays a key role in this hyperosmotic induced entry event. However, this role is permissive for subsequent events rather than sufficient, because the HR1abc domain of PKN1 is not recruited to vesicles despite retaining the Rac1 (and Rho) interacting domain (14 -16). The finding that the PKN1 kinase domain is in part constitutively vesicular, coupled with the observation that fulllength PKN1 requires Rac1 activation, suggests that Rac1 binding to PKN1 on hyperosmotic stress induces a conformational change exposing the catalytic domain and hence allowing vesicular recruitment. Consistent with this, it has been suggested that the interaction of Rho GTPases at the aminoterminal HR1 motif acts to disrupt an autoinhibitory intramolecular interaction thereby allowing activation, i.e. an open conformation (36).
The incubation of Rho-GTP with PKN1 has been shown to increase the phosphorylation and catalytic activity of PKN1 (14 -16), and the behavior of PKN1 described here supports the view that the allosteric input through the amino-terminal HR1 domain is required for complex formation with and subsequent phosphorylation by PDK1. The recruitment of PDK1 occurs through the kinase domain (results not shown) and is likely to involve the FXXFDY motif described as a PDK1-docking site (37). This vesicular interaction is not responsible however for the accumulation of PKN1, because this still occurs when PDK1 recruitment is blocked by the 3-kinase inhibitor LY294002. Thus the kinase domain docking in the vesicle membrane must be determined by a distinct protein (or perhaps lipid) contact. The inhibitory effect of LY294002 on PDK1 recruitment to PKN1 indicates that PI(3,4,5)P 3 is required to influence the conformation of PKN1 or that of PDK1 to facilitate complex formation. Both of these proteins have been shown to be influenced by PI(3,4,5)P 3 ; however, in the case of PKN1 no specificity was observed relative to the precursor lipid PI(4,5)P 2 (13). Hence it is likely that the role of PI(3,4,5)P 3 is to enhance membrane occupancy of PDK1 through its pleckstrin homology domain (32) and so facilitate recruitment to the membrane-bound PKN1. The reduction in size of PKN1-positive vesicles on treatment of shocked cells with LY294002 suggests that the larger vesicles observed are a consequence of a PI 3-kinase-dependent vesicle fusion event; this is a characteristic of homotypic endosome fusion (35).
The combined Rac1/PI(3,4,5)P 3 regulatory input that facilitates PDK1 phosphorylation of PKN1 provides further evidence for the view that the specificity of PDK1 actions is driven by the co-association of regulatory inputs to target kinases (21). In this context it is notable that despite the requirement of PI 3-kinase for PDK1 recruitment, PKB, which is also recruited by 3-phosphoinositides, is not recruited to this hyperosmotic induced endosomal compartment (data not shown). This suggests that PKB is either actively removed or that under appropriate activating conditions other stabilizing interactions are required for PDK1 phosphorylation of PKB.
Hyperosmolarity is a known apoptotic stimulus (38, 39), and PKNs have been shown to undergo caspase cleavage in re- sponse to apoptotic stimuli and under ischemic conditions (40,41). However, we have not observed cleavage of PKN1 under osmotic stress. The responses defined here indicate that in fact the behavior of PKN1 observed under hyperosmotic conditions reflects an underlying constitutive process. The partial vesicular localization of the kinase domain in the absence of hyperosmotic shock and the observation that HA1077 induces some PKN1 vesicle accumulation suggest that this is a constitutive trafficking pathway up-regulated by hyperosmotic shock. The findings indicate that the PKN1 response to hyperosmolarity is not simply targeting PKN1 for degradation but that PKN1 activity is involved in the turnover or exit from this vesicular compartment.
In summary we describe PKN1 to be a stress-responsive kinase, with the induced translocation being selective over related PKCs, although shared with PKN2. Many stress induced signaling cascades are generalized stress responses, although upon investigation of other stresses, such as temperature shock and oxidative stress, no such PKN1 translocation was observed (data not shown). We conclude that the described PKN1 response is specific for hyperosmotic stress. The selective translocation of PKN1 coupled with its subsequent activation details a mechanism by which hyperosmotic shock can elicit a particular repertoire of responses through this kinase.