Rho GTPase Control of Protein Kinase C-related Protein Kinase Activation by 3-Phosphoinositide-dependent Protein Kinase*

The protein kinase C-related protein kinases (PRKs) have been shown to be under the control of the Rho GTPases and influenced by autophosphorylation. In an-alyzing the relationship between these inputs, it is shown that activation in vitro and in vivo involves the activation loop phosphorylation of PRK1/2 by 3-phos-phoinositide-dependent protein kinase-1 (PDK1). Rho overexpression in cultured cells is shown to increase the activation loop phosphorylation of endogenous PRKs and is demonstrated to influence this process by controlling the ability of PRKs to bind to PDK1. The interaction of PRK1/2 with PDK1 is shown to be dependent upon Rho. Direct demonstration of ternary (Rho z PRK z PDK1) complex formation in situ is provided by the observation that PDK1 is recruited to RhoB-containing endosomes only if PRK is coexpressed. Furthermore, this in vivo complex is maintained after phosphoinositide 3-kinase inhibition. The control of PRKs by PDK1 thus evidences a novel strategy of substrate-directed control involving GTPases. The protein kinase C-related protein kinases (PRKs) 1 are a subfamily of serine/threonine-specific kinases independently identified by molecular cloning, protein purification, and polymerase chain reaction-based screens for novel PKC isoforms Two members of subfamily have been fully cloned and characterized: PRK1

The protein kinase C-related protein kinases (PRKs) 1 are a subfamily of serine/threonine-specific kinases independently identified by molecular cloning, protein purification, and polymerase chain reaction-based screens for novel PKC isoforms (1)(2)(3). Two members of this subfamily have been fully cloned and characterized: PRK1 (also termed protein kinase N) and PRK2. They are activated by fatty acids and phospholipids in vitro, although the in vivo significance of this potential is as yet uncharacterized (4,5). Two-hybrid screens and affinity chromatography identified the PRKs as potential effectors of Rho family GTPases (6 -8). The novel amino-terminal HR1 domain (9) was subsequently identified as the Rho-interacting region (10,11), an interaction that is presumed to disrupt the autoinhibitory effect produced by a pseudosubstrate-catalytic domain contact (12). Indeed, the GTPase interaction has been shown to increase modestly the activity of the intact kinase (6 -8). The observed proteolytic activation of these kinases would support the role of an allosteric amino-terminal GTPase interaction in the regulation of activity (13,14). In addition to any allosteric component, Rho GTPases can also be responsible for the location of these kinases, with RhoB causing localization to an endosomal compartment in fibroblasts (15), a translocation event that is associated with the accumulation of a hyperphosphorylated form of the kinase.
In addition to GTPases, other PRK interactions have been identified. The adapter protein NCK has been shown to interact with a proline-rich region just N-terminal of the kinase domain of PRK2 (16). A similar region is absent in PRK1, suggesting a specificity in the upstream recruitment of these kinases. A potential role for PRKs in the regulation of the cytoskeleton has been proposed following the demonstrated disruption of fibroblast actin stress fibers by the expression of a catalytically inactive PRK2 (8) and the observed PRK1 interaction with the head domain of intermediate filament subunits, the subsequent phosphorylation of which results in an inhibition of polymerization (17,18). PRK1 has also been implicated in GTP␥S and insulin-stimulated glucose uptake via Glut4 mobilization in adipocytes (19). However, the true cellular target(s) of PRK action in both this process and established Rho responses have remained elusive.
3-Phosphoinositide-dependent protein kinase-1 (PDK1) was originally purified as an activity responsible for the activation loop phosphorylation of PKB␣ (20,21). It was found to phosphorylate PKB␣ at threonine 308 in a PtdIns(3,4,5)P 3 -dependent manner, resulting in activation of the substrate kinase. This PH domain-containing serine/threonine kinase has subsequently been demonstrated to phosphorylate equivalent residues in many members of the AGC kinase family, including p70 S6K , PKC, and protein kinase A (22)(23)(24)(25)(26)(27). PKB phosphorylation and activation by PDK1 show a much greater lipid dependence when the PKB PH domain is intact, suggesting that the lipid interaction with the substrate causes the removal of an intramolecular interaction masking the activation loop from PDK1. The deletion of this domain results in a more lipidindependent phosphorylation event (28,29). The equivalent PDK1 phosphorylation of a truncated p70 S6K is also independent of lipid (22). The in vivo mechanism for PDK1 dependence on the action of PI3K activity is less clear. The observations that PDK1 activity immunopurified from cultured cells is independent of mitogenic stimulation (20,23) and that PI3K-dependent translocation of PDK1 to the membrane is associated with its ability to activate PKB (30) suggest that PDK1 is constitutively active and that substrate phosphorylation occurs only when both kinases are co-localized at the same cellular site. However, PDK1 is seen to have some in vivo dependence on PtdIns(3,4,5)P 3 even when the substrate lacks a lipid-binding site, suggesting that lipid regulation operates at both the level of kinase and substrate (25).
The PRKs possess a putative activation loop phosphorylation site that fits the consensus of all characterized PDK1 substrates. In this study, we demonstrate that PDK1 does indeed phosphorylate the activation loop threonine of both human PRK isoforms and that this phosphorylation event is crucial for the activation of the PRKs. Furthermore, we show that a PRK⅐PDK1 complex can be formed in vivo and that this is dependent on Rho GTPases. These data provide a mechanism for the Rho-dependent activation of the PRKs and illustrate a novel specificity constraint on this PDK1 substrate.

EXPERIMENTAL PROCEDURES
Materials-Anti-PRK1 and anti-PRK2 monoclonal antibodies for general Western blot analysis and endogenous PRK2 immunoprecipitation were from Transduction Laboratories. The activation loop phospho-specific rabbit polyclonal antiserum was generated by immunization with a 7-residue phosphorylated peptide antigen (RTST(P)FCG). This was then affinity purified in the presence of excess dephosphorylated peptide by chromatography on an Acti-gel ALD column (Sterogene Bioseparations Inc.) coupled to the phosphopeptide. Rabbit anti-RhoB polyclonal antibody 119 was from Santa Cruz Biotechnology Inc. Anti-EE epitope monoclonal antibody was described previously (31). Cy2and Cy3-conjugated secondary antibodies were from Jackson Immu-noResearch Laboratories, Inc. The PI3K inhibitors LY294002 and wortmannin were from Calbiochem, as were the protein phosphatase 1A and 2A inhibitors okadaic acid and calyculin A. The lipids phosphatidylserine and phosphatidylcholine were from Sigma, and the D-enantiomer sn-1-stearoyl-2-arachidonyl-PtdIns (3,4,5)P 3 was a gift from P. Gaffney. The GST-PDK1 protein has been described previously (28). The 3-hydroxy-3-methylglutaryl-CoA reductase inhibitor mevastatin was obtained from Sigma.
Plasmid Constructs-The full-length carboxyl-terminal Myc-tagged PRK1 and PRK2 constructs have been described previously (11). The truncated GST-PRK1kin protein was constructed by polymerase chain reaction (using Vent TM polymerase, New England Biolabs Inc.) of the Myc-tagged construct incorporating an amino-terminal BamHI site and a carboxyl-terminal NotI site. Following restriction digestion, the fragment was subcloned into pGEX-4T1 (Amersham Pharmacia Biotech) for bacterial expression. EE-tagged PDK1 as well as Myc-PDK1 (fulllength) and Myc-PDK⌬PH (residues 51-404) were as described (28). The GTPase-deficient Rho mutants have been described previously (15). The PRK2 activation loop mutants T816A and T816E were generated in pcDNA3 (Invitrogen) using the QuickChange TM site-directed mutagenesis kit (Stratagene) in accordance with the manufacturer's instructions. All mutants were sequenced to confirm integrity.
Cell Culture and Transfection-HEK 293 and NIH 3T3 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum. 293 cells were transfected with plasmid DNA by a Ca 2 PO 4 procedure as described (32). Briefly, cells were plated at 1 ϫ 10 5 and incubated for 24 h prior to transfection, and the precipitate was washed off after a further 14 -16 h. Cells were harvested 24 h later after treatments described below. For immunofluorescence microscopy, cells were plated on acid-washed coverslips and allowed to adhere overnight. The following day, cells were transfected with various mammalian expression vector constructs using Transfast lipid (Promega) according to the manufacturer's protocol. After exposure to the lipid/DNA mixture for 7 h, the cells were washed into fresh medium and left for a further 12 h prior to experimentation.
Immunoprecipitation of PRK and PDK1 Proteins-Cells were transfected as described above. After the treatments below or in the figure legends, the cells were either lysed directly into 4ϫ Laemmli SDS sample buffer for Western blot analysis or washed twice in cold Trisbuffered saline prior to lysis in 50 mM Hepes (pH 7.5), 100 mM NaCl, 20 mM NaF, 1 mM DTT, 1% Triton X-100, 5 mM EDTA, 10 g/ml leupeptin, 0.1 mM phenylmethylsulfonyl fluoride, 5 mM sodium pyrophosphate, and 1 M microcystin LR. Preclearance of cell lysates with insoluble protein A (Sigma) was followed by incubation with the relevant antibody (anti-PRK2 monoclonal antibody for both endogenous and exogenous PRK2 or anti-EE monoclonal antibody for PDK1). Proteins were immunoprecipitated on protein G-Sepharose (Sigma). For the co-immunoprecipitation experiments, the Sepharose was washed three times in ice-cold lysis buffer, followed by protein elution in 4ϫ SDS sample buffer. Samples were then fractionated on a 10% SDS-polyacrylamide gel. For samples to be used in subsequent kinase assays, the Sepharose was washed (1ϫ lysis buffer, 1ϫ lysis buffer ϩ 0.5 M NaCl, and 2ϫ storage buffer), followed by storage on ice in the final wash buffer.
Phosphorylation and Activation of GST-PRK1kin by PDK1-In a 40-l reaction volume, 100 ng of purified GST-PRK1kin was incubated with or without 12 nM GST-PDK1 in the presence of 10 mM Mg 2ϩ and 100 M ATP. Reactions took place at 30°C and were terminated at the time points indicated by the addition of 20 l of 4ϫ Laemmli SDS sample buffer. GST-PRK1kin Thr-774 phosphorylation was assessed by Western analysis using the phospho-specific polyclonal antibody. Lipids were presented as mixed vesicles containing 100 M phosphatidylserine, 100 M phosphatidylcholine, and 10 M PtdIns(3,4,5)P 3 . For the assessment of GST-PRK1kin activation, a two-stage assay was carried out, similar to that previously described (28). Briefly, stage 1 involved the GST-PRK1kin phosphorylation by GST-PDK1 as described above (40 min at 30°C). In stage 2, 20 l of storage buffer containing 4% Triton X-100, 0.5 mg/ml MBP, 100 M ATP, and 5 Ci of [␥-32 P]ATP/ reaction was added to the reactions from stage 1, followed by a further 15-min incubation at 30°C. Reactions were terminated by the addition of 20 l of sample buffer and heating to 95°C prior to fractionation on a 15% SDS-polyacrylamide gel. The acrylamide gel was Coomassie Blue-stained and dried, and the incorporation of [ 32 P]orthophosphate into MBP was assessed by Cerenkov counting.
Assessment of Immunoprecipitated Kinase Activity-Immunoprecipitates of either endogenous PRK2 or the overexpressed mutants were prepared as described above. The washed protein G-Sepharose containing the purified kinases was resuspended in storage buffer containing 0.25% Triton X-100 as a 30% slurry and stored on ice. Using a positive displacement pipette, 10 l of the slurry was added to a kinase reaction (storage buffer containing 2.5 mM Mg 2ϩ , 100 M ATP, 5 Ci of [␥-32 P]ATP, 0.5 mg/ml MBP, and 0.25% Triton X-100) to a final volume of 40 l. Reactions were incubated at 30°C for 15 min in a shaking incubator (Microtherm TM , Camlab) and terminated by the addition of 20 l of sample buffer. Orthophosphate incorporation into MBP was assessed as described above.
Immunofluorescence Microscopy-Cells were processed for indirect immunofluorescence microscopy as described previously (15). Briefly, cells were fixed in 4% (w/v) paraformaldehyde for 15 min and permeabilized with 0.2% (v/v) Triton X-100 for 5 min. Autofluorescence was quenched by incubation with 0.1% (w/v) sodium borohydride for 10 min. Cells were incubated with primary antibody in phosphate-buffered saline containing 1% (w/v) bovine serum albumin for 1 h and with fluorescent dye-conjugated secondary antibody for 45 min. All incubations were performed in phosphate-buffered saline. Cells were mounted under MOWIOL 4-88 (Calbiochem) containing 0.6% (w/v) 1,4diazabicyclo[2.2.2]octane as an anti-photobleaching agent and were viewed using a Leica DM RBE confocal microscope equipped with a Plan APO ϫ63/1.4 oil immersion lens. Cy2 and Cy3 were excited using the 488 and 568 nm lines of a krypton-argon laser. Series of images were taken at 0.5-m intervals through the Z-plane of the sample and were processed to form a projected image.

RESULTS
The PRK1 Kinase Domain Is Phosphorylated and Activated by PDK1 in Vitro-Previous studies on the phosphorylation of the PRK family of kinases has provided evidence for an autophosphorylation event in the regulation of the mammalian proteins (33) and demonstrated that hyperphosphorylated species can accumulate in endosomal compartments in a RhoB-dependent manner (15). To investigate both the mechanisms and possible modulators of this control in vitro and in vivo, we initially expressed the PRK1 kinase domain (amino acids 602-942) in E. coli as a GST fusion protein, GST-PRK1kin. This truncation mutant bypasses any requirement for Rho⅐GTP, which interacts with the inhibitory amino-terminal HR1 domain (11,12). Purified GST-PRK1kin was found to have very low intrinsic protein kinase activity in vitro (Ͻ0.2 units/mg) against the substrate MBP. This activity was not increased on preincubation of the kinase domain with MgATP (see below), suggesting that the bacterially expressed protein may require heterologous modification for full activity.
The recent finding that the closely related PKC family members are subjected to phosphorylation by PDK1 in their activation loop sites (24 -26) led us to examine whether PRK1 was also phosphorylated by PDK1. As shown in Fig. 1A, PDK1 phosphorylated GST-PRK1kin within its activation loop in a time-dependent manner. This was detected using a polyclonal antibody specific for the phosphorylated form of threonine 774. No phosphorylation of this site was detectable under the same conditions in the absence of PDK1. The in vitro phosphorylation of PKB␣ by PDK1 has been shown to be dependent upon the presence of PtdIns(3,4,5)P 3 (28). However, the phosphorylation of Thr-774 on PRK1 by PDK1 was not affected by the addition of mixed lipid vesicles containing 10 M PtdIns(3,4,5)-P 3 (Fig. 1B). Under the same conditions, the activity of GST-PDK1 against MBP was PtdIns(3,4,5)P 3 -dependent (data not shown). Optimum phosphorylation of GST-PRK1kin by PDK1 was associated with a Ͼ60-fold activation of the kinase domain, resulting in a specific activity of 12.5 units/mg (Fig. 1C). Consistent with the phosphorylation data, this activation/activity is not dependent upon the presence of PtdIns(3,4,5)P 3 .
PRK Activation Loop Phosphorylation Is Affected by PDK1 and Rho in Vivo-To assess the physiological potential of PDK1 as a PRK activation loop kinase, we monitored the phosphorylation state of endogenous PRK1 and PRK2 in HEK 293 cells transfected with PDK1 or with the GTPases RhoA and RhoB (as GTPase-deficient Q63L mutants). As shown in Fig. 2, although both PRK isoforms showed very low basal activation loop phosphorylation, expression of either Rho mutant or PDK1 led to a substantial increase in the specific phosphorylation of both kinases. Thus, both controls act to influence the phosphorylation of PRK1 and PRK2.
The combination of transfected Rho protein plus PDK1 did not further increase phosphorylation of either PRK. This is not due to a high stoichiometry of phosphorylation since treatment of cells with the protein phosphatase 1/2A inhibitor okadaic acid (or calyculin A) led to a more extensive activation loop phosphorylation of both kinases, coincident with other undefined modifications that caused a shift in SDS-polyacrylamide gel electrophoresis mobility (Fig. 2, upper and center panels). This implies that requirements additional to Rho and PDK1 are involved in the observed phosphorylation and mobility shift. The products of PI3K have been shown to play a key role in the phosphorylation of PKB and other AGC kinases by PDK1  (29). We therefore assessed the effect of the PI3K inhibitor LY294002 on PRK activation loop phosphorylation. Although the basal Thr-774 occupation of endogenous PRK1 was only slightly altered by the inhibitor treatment (Fig. 3), the increased signals associated with the transient transfection of either GTPase-deficient Rho or PDK1 constructs were reduced to almost basal levels by a block of PI3K activity. This is consistent with a PtdIns(3,4,5)P 3 -dependent PDK1 localization/allosteric event being necessary for PRK phosphorylation. Interestingly, the PRK activation loop phosphorylation induced by PDK1⌬PH (residues 51-404) was also seen to be dependent upon PI3K products, being inhibited by LY294002. This suggests that a separate pathway exists between PI3K and PRK other than through PDK1, perhaps through endogenous GTPases.
Immunopurified endogenous PRK2 displayed an increase in catalytic activity associated with increased activation loop phosphorylation. The comparatively modest effect of PDK1 expression on phosphorylation was insufficient to elicit a significant change in catalytic activity (Fig. 4). However, the much greater activation loop phosphorylation associated with phosphatase inhibition, coupled with the modification resulting in the shift on SDS-polyacrylamide gel electrophoresis, produced a Ͼ3-fold increase in activity against MBP.
Activation Loop Phosphorylation Is Required for Catalytic Activity in Vivo-Although the in vitro data generated with GST-PRK1kin suggested an absolute requirement for activation loop phosphorylation, the endogenous full-length PRK proteins seemed to require additional modifications. To assess the direct influence of activation loop phosphorylation on in vivo activity, activation loop mutants were employed. HEK 293 cells were transiently transfected with either wild-type PRK2 or the T816A or T816E mutant. Immunoprecipitation was carried out using an anti-PRK2 monoclonal antibody, which resulted in endogenous as well as exogenous protein being purified. The in vitro activity of immunoprecipitates was corrected to untransfected cells. As shown in Fig. 5, the T816A mutant had an activity below the basal level, and the T816E mutant produced only 7% of the wild-type activity. This indicates that an intact activation loop threonine is required for activity and further that the acidic mutation is a poor surrogate for phosphorylation. To determine whether Thr-816 phosphorylation was required for the increased activity in response to phosphatase inhibition, we compared the two mutants with wild-type PRK2 following okadaic acid treatment of cells. As seen previously for  4. Effect of activation loop phosphorylation on immunopurified activity. 293 cells transiently transfected with PDK1 and/or treated with 100 nM calyculin A for 40 min were lysed, and endogenous PRK2 was immunopurified. Kinase activities associated with the immunoprecipitates were assessed in vitro against MBP, resulting in the levels displayed graphically (in arbitrary units). Kinase assays were performed in the presence of 0.25% Triton X-100 for 15 min at 30°C. Western blot analysis of the immunoprecipitates displayed equal PRK2 loads in each assay, whereas the phospho-specific polyclonal antibody displayed the relative phosphate occupation of the PRK activation loops in the whole cell lysates of the relative samples. endogenous PRK2, exogenous wild-type PRK2 showed a marked activation due to the phosphatase inhibitor. Neither mutant displayed any significant activity increase over endogenous cellular activity. This demonstrates that activation of PRK under these conditions has an absolute requirement for phosphorylation of the activation loop threonine.
PRK and PDK Interaction Is Rho-dependent-In vivo interactions between PDK1 and PKC isotypes have been assessed previously and are dependent upon intact kinase domains in both proteins (25). Furthermore, it has been reported recently that a fragment of PRK2 (residues 908 -984) interacts with PDK1 (34). We investigated the presence of an in vivo complex formed between PDK1 and PRK. Coexpression of PDK1 and PRK2 in 293 cells followed by immunoprecipitation of PDK1 led to the co-immunoprecipitation of PRK2 (Fig. 6A). No PRK was present in immunoprecipitates from cells not expressing the tagged PDK1. Interestingly, the complex formation was enhanced with the coexpression of RhoA (Q63L mutant). Consistent with the requirement for endogenous Rho in the formation of PRK2⅐PDK1 complexes, the expression of Clostridium toxin C3 transferase was found to block the constitutive kinasekinase interaction. It can be concluded therefore that PRK2 binds PDK1 in a Rho-dependent manner. A similar situation was seen for a PRK1-PDK1 co-immunoprecipitation, where the presence of an activated Rho greatly increased the extent of complex formation. However, unlike PRK2, the constitutive low level interaction seen in the absence of exogenous Rho was not eliminated by C3 transferase (Fig. 6B). This implies that although a similar potential for Rho dependence exists for the PRK1-PDK1 interaction, an alternative input to PRK1 must also exist (see "Discussion").
To determine the need for Rho membrane localization in supporting PRK⅐PDK1 complex formation, we blocked Rho isoprenylation with the inhibitor mevastatin. The presence of the inhibitor suppressed Rho-dependent complex formation (Fig.  7A), indicating that modification and consequent membrane localization are necessary. Consistent with this, a RhoA SAAX mutant, which cannot be isoprenylated, was unable to support complex formation (data not shown). The loss of complex formation with PDK1 paralleled a loss of endogenous PRK1 Thr-774 and PRK2 Thr-816 phosphorylation (Fig. 7B).
Previous studies have shown the RhoB-dependent recruitment of PRK1 to endosomes (15). This observation allows visualization of a putative RhoB⅐PRK⅐PDK1 complex in intact cells. NIH 3T3 cells transfected with wild-type RhoB displayed a characteristic punctate staining of the protein (Fig. 8A) that has been shown to reflect early endosomal compartments (35). The staining of PDK1 coexpressed with RhoB was diffuse cytoplasmic and non-nuclear (Fig. 8B); this was also the case when PDK1 was coexpressed with PRK1 in the absence of Rho. However, when PDK was expressed with RhoB in the presence of PRK1, it was found to display a punctate distribution coincident with RhoB (Fig. 8D). A similar translocation event occurred in the presence of overexpressed PRK2 (data not shown). Hence, the Rho dependence of PRK⅐PDK1 complex formation evidenced in the co-immunoprecipitation assay above is shown directly here with ternary (Rho⅐PRK⅐PDK1) complex formation in intact cells.
PDK1 has previously been shown to translocate from the cytoplasm of serum-starved porcine aortic endothelial cells to the plasma membrane on stimulation with platelet-derived growth factor BB (30). This PDK1 translocation event was seen to be dependent upon both an intact PH domain and the products of PI3K. We therefore assessed whether the observed ternary complex formation was also dependent upon PtdIns(3,4,5)P 3 or PtdIns(3,4)P 2 . Treatment of the cells with 100 nM wortmannin for 20 min was seen to disrupt the morphology of the endosomal compartment, but did not disrupt PDK1 co-localization with PRK1 and RhoB (Fig. 8F). Such a wortmannin treatment has been shown to cause rapid displacement of PtdIns(3,4,5)P 3 -specific binding proteins from the plasma membrane (36). It can be concluded therefore that both the translocation to and maintenance of PDK1 in the endosomal compartment is due to a protein-protein interaction with PRK rather than a lipid-specific event. DISCUSSION It has been proposed that the interaction of Rho family GTPases with the amino-terminal HR1 domain of PRK acts to disrupt an autoinhibitory intramolecular interaction (12). Indeed, incubation of PRK1 or PRK2 (in whole cell lysates, immuno-or affinity-purified) with Rho GTPases results in a significant increase in the phosphorylation of the kinase and its activity against in vitro substrates (6 -8). The increased phosphorylation was assumed to be autokinase activity promoted by the intramolecular disruption. However, it is shown here that the PRK1 kinase domain expressed in bacteria has a very low kinase activity, suggesting that a heterologous input is required prior to any autocatalytic activation. This input is established by the demonstration that the recently identified AGC kinase family activator PDK1 phosphorylates the PRKs at their conserved activation loop threonines (Thr-774 and Thr-816 for PRK1 and PRK2, respectively) both in vitro and in vivo. Furthermore, mutation of the activation loop threonine in PRK1 results in an inhibition of both basal and agonist-induced activities in vivo. Investigation of the behavior of intact PRKs in vitro and in vivo has identified a permissive role for Rho GTPases in the association of PRKs with PDK1. The evidence indicates that there is a sequential assembly of regulators on the membrane that leads to PRK activation through activation loop phosphorylation. This provides an interesting contrast with other PDK1 effectors, where the specificity of input appears to rely upon an interaction with lipid effectors, i.e. PtdIns(3,4,5)P 3 -PKB (21, 28) and diacylglycerol-novel PKC (25,37). It is concluded that for PRK activation, the specificity of the PDK1 input is affected by a requirement for Rho⅐GTP.
Previously, the Rho-inactivating Clostridium toxin C3 transferase has been seen to inhibit lysophosphatidic acid-induced phosphorylation of PRK in vivo (6). The basis of this is established here by the demonstration that the PRK⅐PDK1 complex is Rho-dependent. Although apparently accounting for the lysophosphatidic acid response, a component of the PRK1⅐PDK1 complex was not inhibited by C3; other small GTPases have been seen to interact with the PRKs, suggesting the potential for alternative allosteric inputs (8). The Rho-dependent interaction between the two kinases suggests that Rho binding is required to allow access to PDK1. This is supported by the recent observation that both intact PDK1 and a minimal kinase domain (residues 51-404) can interact directly with the carboxyl-terminal 77 amino acids of PRK2 (34). Thus, it would seem that the regulatory inputs required by the PRK isoforms are interdependent. Initially, a GTPase interaction exposes the carboxyl-terminal region of the catalytic domain, permitting the binding of PDK1. This can, in turn, phosphorylate the PRK activation loop threonine in the presence of PtdIns(3,4,5)P 3 (see below), producing a PRK molecule capable of autophosphorylation and further activation. In vitro, the HR1 domains of PRK1 and PRK2 are sufficient for binding to GTP-loaded, bacterially expressed Rho, i.e. non-prenylated Rho⅐GTP (11). However, the assembly of the Rho⅐PRK⅐PDK1 complex and the consequent PRK activation loop phosphorylation in vivo require prenylation of Rho. Thus, there is a requirement for assembly on a membrane prior to the PtdIns(3,4,5)P 3 -dependent phosphorylation of PRK. This implies that events downstream of this specific pathway are themselves dependent upon modified Rho (Rho family) protein. Thus, for example, this pathway could not account for the induction of the serum response element by RhoB (38). PDK1 has been shown to be translocated to the membrane of porcine aortic endothelial cells upon acute platelet-derived growth factor stimulation in a manner dependent upon the products of PI3K (30). This translocation event was associated with the ability of the kinase to enhance PKB activity. Here it is shown that PDK1 was recruited to the early endosomal compartment in a PRK-dependent manner and that the resulting ternary complex of RhoB⅐PRK⅐PDK1 was unaffected by PI3K inhibition. This suggests that the recruitment and maintenance of such a complex are due to protein-protein interactions and are independent of 3-phosphorylated lipids. Interestingly, the PRK activation loop phosphorylation by PDK1 was found to be inhibited by LY294002, indicating an in vivo PtdIns(3,4,5)P 3 or PtdIns(3,4)P 2 requirement. It appears that the recruitment of the kinases to this membrane compartment is independent of PI3K products, but that the subsequent activation step requires a 3-phosphorylated lipid. This conclusion is supported by the finding that the activity of PDK1 becomes PtdIns(3,4,5)P 3 -dependent when complexed to the C-terminal fragment of PRK2 (34).
Many members of the AGC kinase family require the dual phosphorylation of an activation loop threonine by PDK1 as well as a conserved carboxyl-terminal serine or threonine residue (for example, Ser-473 in PKB␣) for full activation (39). The kinase responsible for carboxyl-terminal site modification has until recently remained elusive. The PRK and atypical PKC ( and ) kinases possess an acidic residue at this second site; and in the study by Balendran et al. (34), this motif has been shown to be critical for the PDK1 interaction. When bound to PDK1, a 77-amino acid peptide covering the carboxyl terminus of PRK2 modifies PDK1 activity, enabling it to phosphorylate both Thr-308 and Ser-473 of PKB␣. Evidence for the involvement of the atypical class of PKCs in the phosphorylation of equivalent sites is offered by the observed role of a PKC complex in the phosphorylation of PKC␦/␣ at this residue (40). A prediction arising out of the studies here is that there are downstream targets of PDK1 that are dependent upon Rho family protein functions and that are mediated by the RhoA/B⅐PRK1/2 (and related) complexes. The implication of Rac in the phosphorylation of PKB and p70 S6K (41)(42)(43) is intriguing considering the observed PRK2-Rac interaction (8).
The results presented in this study demonstrate PDK1 to be the kinase responsible for the phosphorylation of Thr-774 and Thr-816 of PRK1 and PRK2, respectively. This phosphorylation event results in the activation of these Rho effectors. The Rho dependence of the PRK⅐PDK1 complex formation and co-localization suggests a mechanism for the previously observed PRK activation by the GTPases. Such a Rho-dependent phosphorylation event would explain the previously observed block of insulin-and GTP␥S-stimulated glucose transport in rat adipocytes by PI3K, Rho GTPase, and broad-specificity PKC inhibitors (19). Precisely how this Rho-PRK-PDK1 pathway is assimilated into insulin-induced and other agonist responses is the subject of further investigation.