Role of Translocation in the Activation and Function of Protein Kinase B*

We have investigated the role of subcellular localization in the regulation of protein kinase B (PKB) activation. The myristoylation/palmitylation motif from the Lck tyrosine kinase was attached to the N terminus of protein kinase B to alter its subcellular location. Myristoylated/palmitylated (m/p)-PKBα was associated with the plasma membrane of transfected cells, whereas the wild-type kinase was mostly cytosolic. The activity of m/p-PKBα was 60-fold higher compared with the unstimulated wild-type enzyme, and could not be stimulated further by growth factors or phosphatase inhibitors.In vivo 32P labeling and mutagenesis demonstrated that m/p-PKBα activity was due to phosphorylation on Thr308 and Ser473, that are normally induced on PKB following stimulation of the cells with insulin or insulin-like growth factor-1 (IGF-1). A dominant negative form of phosphoinositide 3-kinase (PI3-K) did not affect m/p-PKBα activity. The pleckstrin homology (PH) domain of m/p-PKBα was not required for its activation or phosphorylation on Thr308 and Ser473, suggesting that this domain may serve as a membrane-targeting module. Consistent with this view, PKBα was translocated to the plasma membrane within minutes after stimulation with IGF-1. This translocation required the PH domain and was sensitive to wortmannin. Our results indicate that PI3-K activity is required for translocation of PKB to the plasma membrane, where its activation occurs through phosphorylation of the same sites that are induced by insulin or IGF-1. Following activation the kinase detached from the membrane and translocated to the nucleus.

Many growth factors elicit cellular responses by activating phosphoinositide 3-kinase (PI3-K 1 ; reviewed in Ref. 1). Re-cently, protein kinase B (PKB), also known as RAC protein kinase or c-Akt (2)(3)(4) was recognized as a downstream target of PI3-K (5,6). Three mammalian isoforms of PKB have been identified so far, termed PKB␣, -␤, and -␥ (7-9). 2 All three isoforms contain a pleckstrin homology (PH) domain at the N terminus (10), followed by a catalytic domain related to protein kinases A and C, and a C-terminal regulatory region. PKB␣ was found to mediate insulin-and insulin-like growth factor (IGF-1)-induced cellular responses, such as the inhibition of glycogen synthase kinase-3 (11), the stimulation of glucose uptake (12), and the promotion of cell survival by inhibiting apoptosis (Ref. 13; reviewed in Refs. 14 and 15). PKB␣ is the cellular homologue of the oncogene product v-Akt encoded by the AKT8 retrovirus, which induces thymic lymphomas in mice (16). Cloning of v-akt revealed that it was created by fusion of viral Gag sequences to the N terminus of mouse PKB␣, which adds an N-terminal myristoylation signal to the oncoprotein and could account for its transforming ability (2,17). Overexpression of PKB␣ or -␤ is associated with some human ovarian, pancreatic, and breast carcinomas (8, 18 -20).
PKB␣ is activated by a variety of growth factors and phosphatase inhibitors (5, 6, 21) through a phosphorylation mechanism (21)(22)(23). The activation of PKB␣ by insulin or IGF-1 is mediated by phosphorylation of Thr 308 in the catalytic domain and Ser 473 at the C terminus (22). The phosphorylation of both sites is blocked by pretreatment of the cells with the PI3-K inhibitor wortmannin. Substitution of both regulatory sites by aspartic acid residues to mimic phosphorylation by the introduction of a negative charge, produces a constitutively active enzyme (22). This work predicted the existence of an upstream kinase(s) that phosphorylate(s) these sites, and recently a protein kinase activity was identified and purified capable of phosphorylating Thr 308 in the presence of phosphatidylinositol 3,4,5-trisphosphate (PtdIns(3,4,5)P 3 ) or phosphatidylinositol 3,4-bisphosphate (PtdIns(3,4)P 2 ) (Refs. 24 and 25; reviewed in Ref. 26). The enzyme has therefore been termed 3-phosphoinositide-dependent protein kinase-1 (PDK1).
The PH domain of PKB has been reported to play a role in the activation process (6), but PKB activation can also occur in its absence, depending on the agonist and the type of deletion mutants used (21,23,27). The PH domain of PKB binds PtdIns(3,4,5)P 3 and PtdIns (3,4)P 2 at low micromolar concentrations (28,29), but the precise role of inositol phospholipid binding to PKB is not fully understood (28 -31). Since PtdIns(3,4,5)P 3 and PtdIns(3,4)P 2 are located in the plasma membrane, the interaction of PKB with one or both of these phosphoinositides may play a role in recruiting PKB to the membrane. To address this question, we have added a membrane targeting signal to the N terminus of PKB␣ and find that this is sufficient for maximal phosphorylation of Thr 308 and Ser 473 , and activation of PKB␣ in the presence or absence of the PH domain. Furthermore, we provide evidence that the IGF-1-induced activation of PKB is accompanied by its translocation to the membrane, followed by the translocation to the nucleus.

Construction of Expression Vectors-
The pECE and pCMV constructs encoding hemagglutinin (HA) epitope-tagged PKB␣ have been described (21,22). Myristoylated/palmitylated (m/p)-HA-PKB␣ was created by polymerase chain reaction with HA-PKB␣ as template, using a 5Ј oligonucleotide encoding the 12-amino acid N-terminal sequence of Lck, which carries the myristoylation/palmitylation signal, followed by two Ala residues and the HA epitope, and a 3Ј oligonucleotide encoding amino acids 468 -480 of PKB␣. The resulting product was subcloned as a SalI/NotI fragment into pECE.HA-PKB␣. The pECE expression constructs encoding PKB␣ phosphorylation site mutants at Ser 473 and Thr 308 , or both have been described (22). To create the membrane targeted versions of these mutants, the CelII/EcoRII fragment of pECE.HA-PKB␣-S473A and NotI/EcoRI fragment of pECE.HA-PKB␣-T308A were subcloned into the respective restriction sites of pECE.m/ p-HA-PKB␣. The pECE construct encoding m/p-HA-PKB␣-⌬PH was made by polymerase chain reaction using a 5Ј oligonucleotide encoding the 12-amino acid N terminus of Lck, followed by two Ala residues, the HA epitope and amino acids 119 -125 of PKB␣, and a 3Ј oligonucleotide encoding amino acids 468 -480. The product was subcloned as a SalI/ EcoRI fragment into the same restriction sites of the pECE vector (32). The BglII/XbaI fragments from the above described pECE constructs were transferred into the same restriction sites of the pCMV5 vector (33). m/p-HA-PKB␣-T308D/S473D was created by subcloning NotI/XbaI fragments from pECE-HA-PKB␣-T308D/S473D and pECE.HA-PKB␣-⌬C14 into pCMV5.m/p-HA-PKB␣. The constructs were confirmed by restriction analysis and sequencing. The construct SR␣-p85␣⌬478 -513 has been described (34).
Cell Fractionation-Transfected 293 cells were collected in ice-cold hypotonic buffer containing 5 mM Tris, pH 7.4, 25 mM NaF, 5 mM MgCl 2 , 1 mM EGTA, and 0.1 mM sodium orthovanadate, and lysed by 10 strokes in a Dounce homogenizer. Nuclei were removed by centrifugation for 10 min at 1,000 ϫ g at 4°C. The P100 and S100 fraction were obtained by additional centrifugation at 100,000 ϫ g for 30 min at 4°C. P100 was resuspended in lysis buffer.
Immunoblot Analysis-Cell extracts and immunoprecipitates were resolved by 7.5% SDS-polyacrylamide gel electrophoresis, and transferred to Immobilon P membranes (Millipore). The filters were blocked for 30 min with 5% skimmed milk in 1 ϫ Tris-buffered saline, 1% Triton X-100, 0.5% Tween 20, followed by a 2-h incubation with 50-fold diluted rabbit polyclonal anti-PKB␣ antisera specific for the C terminus (Ab ␣469/480 ; Ref. 4), or recombinant PH domain containing N-terminal 131 amino acids (Ab ␣1/131 ) or with the anti-HA epitope 12CA5 monoclonal antibody that was 1,000-fold diluted in the same blocking solution. The secondary antibodies were 1,000-fold diluted alkaline-phosphatase conjugated anti-rabbit IgG (Sigma) and anti-mouse Ig (Southern Biotechnology Associated), or 5,000-fold diluted horseradish peroxidase-linked Ig (Amersham). Detection was performed using the AP color development reagents from Bio-Rad or by enhanced chemiluminescence (Amersham). To normalize expression levels of PKB, FITClabeled secondary antibodies were employed at a 200-fold dilution, the signal detected by chemifluorescence using a Storm 840/860 Phosphor-Imager and quantified with ImageQuant Software (Molecular Dynamics).
Immunofluorescence-293 cells were plated and transfected on sterile coverslips. Fixation of cells with formaldehyde and permeabilization with 0.2% Triton X-100 were performed according to Ref. 37. The 12CA5 monoclonal antibody diluted 50-fold in phosphate-buffered saline (PBS) was applied for one hour at 37°C. The cells were subsequently washed twice with PBS and incubated with FITC-conjugated anti-mouse IgG (Sigma) at a 50-fold dilution, or with 100-fold diluted biotinylated anti-mouse IgG (Sigma), followed by 200-fold diluted streptavidin coupled to Texas Red (Amersham). DNA was stained with 4,6-diamidino-2-phenylindole. The coverslips were washed twice with PBS, once with H 2 O, mounted on glass slides using Gelvatol, and photographed with a LEITZ DMRD Leica camera. Confocal images were collected on a Leica TCS 4D microscope. REF-52 cells were subcultured on either 25-mm glass coverslips (Schutt Labortechnik, Göttingen, Germany) or acidwashed coverslips. Microinjection was performed with a normal Leitz micromanipulator, as described previously (38). Cells were injected with 0.5 mg/ml PKB construct in the presence of 1 mg/ml biotinylated rabbit IgG (Sigma). Cells were serum-starved for 36 -48 h before stimulation with 1 M okadaic acid/10% FCS. Cells were fixed with 3.7% formaldehyde and further treated as described previously (36). PKB was detected using Ab ␣469/480 followed by a FITC-conjugated anti-rabbit antibody, and HA-PKB␣ was detected with the 12CA5 monoclonal antibody followed by a FITC-conjugated anti-mouse antibody. Microinjected cells were identified by costaining the cells with streptavidin coupled to Texas Red (Amersham). DNA was stained using Hoechst stain 33358 (1 g/ml bisbenzemidine).
Immunoelectron Microscopy and Quantification-The cells were fixed in 4% formaldehyde in 0.2 M Pipes, pH 7.2, for at least 20 min, washed in PBS, scraped from the dish using a rubber policeman, and embedded in 10% pig skin gelatin before cryoprotection in 2.3 M sucrose in PBS. Ultrathin sections were cut at Ϫ110°C in a Reichert Ultracut E cryomicrotome, mounted on carbon/Formvar-coated grids, and labeled using the 12CA5 monoclonal antibody followed by a rabbit antimouse antibody and finally protein A gold (7 or 5 nm particle size prepared as described by Lucocq;Ref. 39). Sections were embedded in methylcellulose uranyl acetate as described in Ref. 40. To quantitate immunolabeling, sections were scanned systematically and gold label identified. All visible parts of the plasma membrane and adjacent cytoplasm of labeled cells were photographed at magnification ϫ15,000. Cytoplasm areas and membrane profile length were estimated using a square lattice grid with 1-cm line spacing as described in Ref. 41. Gold particles over the nucleus were not included in the analysis and gold particles were only assigned to the plasma membrane if they lay within 2 particle widths of the plasma membrane profile. 32 (Fig. 10). The cells were then stimulated for 10 min at 37°C in the presence or absence of 100 ng/ml IGF-1 and placed on ice. The medium was aspirated, the cells washed twice with ice-cold DMEM buffer and then lysed. HA-PKB␣ was immunoprecipitated from the lysates, alkylated with 4-vinylpyridine, digested with trypsin, and analyzed by C18 chromatography exactly as described previously (22).

Membrane Targeting Promotes the Activation of PKB␣-We
previously proposed that PKB activation occurs by phosphorylation, following recruitment of the kinase to the membrane via its PH domain (21,22). To investigate the role of membrane targeting in the activation of PKB, the N-terminal membrane localization sequence from Lck was attached to the N terminus of HA-PKB␣ (see Fig. 1). This signal was chosen because it contains the consensus sequence for both myristoylation and palmitylation (42) and has been shown to be sufficient to localize a number of cytosolic proteins to the plasma membrane (43). Several mutants of m/p-HA-PKB␣ were also prepared in which the ATP-binding site (Lys 179 ) or the phosphorylation sites (Thr 308 , Ser 473 ) were mutated to Ala, or in which the N-terminal 118 amino acids containing the PH domain, or the Cterminal 14 amino acids encompassing the Ser 473 phosphorylation site were deleted (Fig. 1).
To confirm that the Lck myristoylation/palmitylation signal provides membrane attachment of m/p-HA-PKB␣, we deter-mined the subcellular localization of the proteins expressed in 293 cells by immunofluorescence using the anti-HA epitope antibody. HA-PKB␣ was found in the cytosol of serum-starved, unstimulated 293 cells (Fig. 2, A and B). However, all forms of m/p-HA-PKB␣ (Fig. 2, C-G) were highly concentrated at the plasma membrane. No immunostaining occurred if the 293 cells were transfected with vector alone (Fig. 2H). Overexpression of m/p-HA-PKB␣, m/p-HA-PKB␣-S473A, or m/p-HA-PKB␣-⌬PH in 293 cells resulted in rounding of the cells, which was not observed when either wild-type HA-PKB or other m/ p-HA-PKB␣ mutants were overexpressed.
The activity of m/p-HA-PKB␣ in unstimulated cells was over 60-fold higher than that of HA-PKB␣. This is higher than the activity of PKB␣ obtained after stimulation of the wild-type kinase with insulin, IGF-1, or vanadate (Fig. 3A). Consistent with this finding, the activity of m/p-HA-PKB␣ could not be increased further by stimulation of the cells with insulin, IGF-1, or vanadate (Fig. 3A). m/p-HA-PKB␣ from unstimulated cells, like HA-PKB␣ from IGF-1-stimulated cells, could be in- The intracellular localization of wild-type HA-PKB␣ and m/ p-HA-PKB␣ were confirmed by biochemical studies. About 80% of HA-PKB␣ activity and 75% of HA-PKB␣ protein were detected in the 100,000 ϫ g supernatant (S100) of unstimulated 293 cells, whereas virtually all of the m/p-HA-PKB␣ activity and protein was recovered in the 100,000 ϫ g pellet (P100) (Fig.  4, A and B).
Membrane-targeted PKB␣ Is Constitutively Active due to Phosphorylation on Thr 308 and Ser 473 -The m/p-HA-PKB␣-T308A and m/p-PKB␣-S473A mutants possessed only ϳ2% and ϳ10 -15% of the activity, respectively, of the membrane-targeted wild-type enzyme in unstimulated or stimulated 293 cells (Fig. 3A), suggesting the involvement of these sites in the activation of m/p-HA-PKB␣. Furthermore, all membrane-targeted forms of the kinase expressed in 293 cells displayed reduced electrophoretic mobility indicative of phosphorylation (Fig. 3B). To establish which residues in PKB were phospho-rylated in vivo, 293 cells were 32 P-labeled, and either m/p-HA-PKB␣ or HA-PKB␣ was immunoprecipitated, digested with trypsin, and the resulting phosphopeptides analyzed by C18 chromatography. As observed previously (22), two peptides phosphorylated on Ser 124 and Thr 450 were obtained from HA-PKB␣ immunoprecipitated from unstimulated cells (Fig. 5A), while stimulation with IGF-1 induced the appearance of two further phosphopeptides labeled on Thr 308 and Ser 473 , without significantly affecting the labeling of Ser 124 and Thr 450 (Fig.  5B). In contrast, m/p-HA-PKB␣ from unstimulated cells was heavily phosphorylated on residues Thr 308 and Ser 473 , as well as at Ser 124 and Thr 450 (Fig. 5C), and IGF-1 treatment did not lead to any further increase in the 32 P labeling of m/p-HA-PKB at Thr 308 or Ser 473 (Fig. 5D). Membrane targeting led to more efficient incorporation of phosphate into Thr 308 and Ser 473 than did IGF-1 stimulation of HA-PKB␣ (Fig. 5, B-D). m/p-HA-PKB␣ was also phosphorylated on two peptides each containing phosphothreonine (peptides TX1 and TX2 in Fig. 5C), that were absent in HA-PKB␣ (Fig. 5A).
Thr 308 and Ser 473 were also heavily phosphorylated in the membrane-targeted mutant, which lacks the PH domain in both unstimulated and IGF-1-stimulated cells (Fig. 5, E and F). Consistent with this finding, m/p-HA-PKB␣-⌬PH was as active as m/p-HA-PKB. Significantly, only the TX1 phosphopeptide was detected in m/p-HA-PKB␣-⌬PH, implying that TX2 resides within the N-terminal 118 amino acids and that it is not important for the activation by membrane targeting. The TX1 peptide was found to be phosphorylated on Thr at position 4, and predicted tryptic cleavage sites suggested that it could be Thr 371 . However, this phosphopeptide was found not to be crucial for m/p-HA-PKB␣ activity (see below), and further characterization of this phosphorylation site was not pursued.
Deletion of the C-terminal 14 amino acids of HA-PKB␣ prevented kinase activation by insulin, IGF-1, or vanadate, and m/p-HA-PKB␣-⌬C14 possessed a similarly low activity in unstimulated or stimulated cells (Fig. 3A). This suggests that the extreme C terminus not only contains an activating phosphorylation site, but may also provide an important conforma-tional determinant for activation.
The kinase-inactive m/p-HA-PKB␣-K179A mutant (which was expressed at much lower levels than m/p-HA-PKB␣) was phosphorylated on Thr 450 and Thr 308 in unstimulated cells, but the level of phosphorylation of Ser 473 was much lower (Fig. 5G). Following IGF-1 stimulation, 32 P-incorporation into Thr 308 doubled and incorporation into Ser 473 increased 5-fold (Fig.  5H). As reported previously for HA-PKB␣-K179A (22), the 32 P labeling of Ser 124 was extremely low in the membrane-targeted, kinase-inactive mutant (Fig. 5G).  D, F, and H). Each 32 P-labeled peptide that eluted from the C18 column was subjected to phosphoamino acid analysis, and the location of each phosphorylation site was established by solid phase sequencing in which the 32 P radioactivity released after each cycle of Edman degradation was measured (51). The peptides corresponding to Ser 124 (S124), Thr 308 (T308), Thr 450 (T450), and Ser 473 (S473) contained the expected phosphoamino acid and released 32 P radioactivity at the 3rd, 1st, 14th, and 8th cycles of Edman degradation as expected (Ref. 22 and data not shown). Similar results were obtained in three separate experiments.
PI3-K Activity-A mutant form of the p85 regulatory subunit of PI3-K (⌬p85) that fails to bind and activate the p110 catalytic subunit has been reported to act as a dominant negative mutant of this enzyme (34) and to abolish platelet-derived growth factor-induced activation of PKB in cotransfection experiments (5). Coexpression of ⌬p85 with HA-PKB␣ reduced insulin-induced PKB activation by 80% (Fig. 6A), but had no effect on the activity of m/p-HA-PKB␣ in unstimulated cells (Fig. 6B).
To gain further insight into the regulation of m/p-HA-PKB␣, cells were treated with inhibitors of PI3-K, following a 12-h serum starvation. Treatment with either 250 nM wortmannin or 100 M LY 294002 for 3 h reduced m/p-HA-PKB␣ activity by ϳ50% (data not shown). However, longer treatment could only be performed with LY 294002, as prolonged wortmannin treatment was toxic. Following a 12-h incubation with LY 294002 in the absence of serum, m/p-HA-PKB␣ and m/p-HA-PKB␣-S473A activities were reduced by ϳ80% and ϳ60%, respectively, and the activity of m/p-HA-PKB␣ was rapidly restored to the initial level within 30 min following the removal of LY 294002 (Fig. 7, A and B). During this 30-min period, no activity changes for untreated or LY 294002-treated m/p-HA-PKB␣ were observed (data not shown). LY 294002 withdrawal did not promote the activation of wild-type HA-PKB␣ (data not shown). The activation of m/p-HA-PKB␣ in the absence of any growth factors after the inhibitor removal could be explained by an autocrine loop mechanism in which membrane-targeted kinase might induce the production and secretion of a signal, which then stimulates the activation of PI3-K. However, this possibility is unlikely, since we found that conditioned medium obtained from cells expressing m/p-HA-PKB␣ failed to activate the endogenous PKB in unstimulated 293 cells (data not shown).
PKB Associates with the Plasma Membrane and Subsequently Translocates to the Nucleus following IGF-1 Stimulation-To investigate whether dynamic changes in the subcellular distribution of PKB occur during cell stimulation, we employed quantitative immunogold localization on ultrathin cryosections. In unstimulated 293 cells expressing HA-PKB␣, there was substantial immunolabeling in the cytoplasm with little evidence for membrane localization (Figs. 9A and 10A). However, after stimulation for 2 min with IGF-1, there was clear evidence for accumulation of gold labeling at the plasma membrane (Fig. 9A). Labeling was more marked 5 min after stimulation with IGF-1, by which time the concentration of Changes in PKB localization during IGF-1 stimulation were also monitored by indirect immunofluorescence. HA-PKB␣ was detected mainly in the cytoplasm of unstimulated 293 cells ( Fig. 11A; see also Fig. 3, A and B). Membrane staining could be observed 2 and 5 min after IGF-1 stimulation (Fig. 11, B and C), whereas 30-min IGF-1 treatment led to considerable nuclear labeling, excluding nucleoli (Fig. 11D). To study immunolocalization of endogenous PKB, we used REF-52 cells treated with 1 M okadaic acid, 10% FCS. These agents exert a synergistic effect on PKB activation (21), and their concomitant application resulted in maximal stimulation of kinase activity in REF-52 fibroblasts (data not shown). Significant nuclear staining could be detected following 15-min stimulation of quiescent REF-52 cells with 1 M okadaic acid, 10% FCS, which increased further with 50-min treatment (Fig. 12A). Similar changes in subcellular distribution were observed upon 30-min stimulation of REF-52 fibroblasts expressing HA-PKB␣ (Fig.  12B), as well as kinase-deficient HA-PKB␣-K179A (data not shown). Taken together, the data indicate that PKB transiently associates with the plasma membrane of stimulated cells, which is followed by translocation to the nucleus.

DISCUSSION
Growth factor-induced activation of PKB␣ is mediated by PI3-K (5,6,11,27), and inhibitors of PI3-K prevent the insulinand IGF-1-induced phosphorylation of PKB on Thr 308 and Ser 473 (22). Based on the fact that the PH domain binds 3phosphoinositides at low micromolar concentrations (28,29), it was proposed that these lipids may act to recruit PKB␣ to the plasma membrane where the activating phosphorylations take place (15,26). The present study has provided three pieces of evidence supporting this concept. First, HA-PKB␣ translocates to the plasma membrane within minutes following IGF-1 stimulation (Figs. 9 and 11), whereas an HA-PKB␣ mutant that lacks the PH domain does not. Second, translocation to the membrane is blocked by wortmannin, indicating that it is dependent on the activation of PI3-K. Third, the addition of a membrane targeting domain to the N terminus of PKB␣ causes it to become active in unstimulated cells. The translocation of PKB␣ to the plasma membrane may not only be critical for its activation, but may also enable the kinase to phosphorylate protein substrates at this location.
The membrane-targeted form of PKB␣ was activated in unstimulated cells to a level that was greater than that attained by wild-type HA-PKB␣ after stimulation of the cells with insulin, IGF-1, or vanadate (Fig. 3A). The activation of m/p-HA-PKB␣ resulted from its phosphorylation on Thr 308 and Ser 473 (Fig. 5), the same residues whose phosphorylation underlies the activation of wild-type PKB␣ by insulin and IGF-1 (22). Significantly, the kinase-inactive, membrane-targeted form of PKB␣ (m/p-HA-PKB␣-K179A) was highly phosphorylated on Thr 308 and (at a lower level) on Ser 473 in unstimulated cells, i.e. under conditions when the endogenous PKB is inactive (Fig. 5). These observations exclude the possibility that phosphorylation of Thr 308 or Ser 473 were catalyzed by PKB␣ itself. This is also indicated by the findings that PKB␣ that has been partially activated by phosphorylation on Thr 308 (24), or by its mutation to Asp (22), cannot autophosphorylate on Ser 473 , and that PKB␣ that has been partially activated by phosphorylation of Ser 473 (or by its mutation to Asp) cannot autophosphorylate on Thr 308 (22). Therefore, other kinases mediate the phosphorylation of PKB␣ at the plasma membrane. Moreover, the observation that m/p-HA-PKB␣-K179A is further phosphorylated on the key regulatory sites in response to IGF-1 implies that the upstream kinase(s) may also be regulated by growth factors. We have recently identified and highly purified a protein kinase that phosphorylates PKB␣ at Thr 308 and which is only active in the presence of PtdIns(3,4,5)P 3 or PtdIns(3,4)P 2 (24), and similar kinase activity was isolated by Stokoe et al. (25). This kinase, termed PDK1 (see Introduction), is likely to be the enzyme responsible for the phosphorylation of both wild-type and m/p-HA-PKB␣ at Thr 308 . The identity of the kinase which phosphorylates PKB␣ at Ser 473 is not yet known. MAPKAP kinase-2 is able to phosphorylate PKB␣ in vitro on Ser 473 , but it does not appear to be the in vivo PKB␣ kinase, as it is activated by stress stimuli and proinflammatory cytokines, which do not activate PKB␣ in the cells that we have studied (22). Moreover, the drug SB 203580, which prevents the activation of MAPKAP kinase-2 in cells (44), has no effect on the activation of PKB␣ by insulin or IGF-1 (22).
The PH domain was not required for the activation of membrane-targeted PKB␣ (Fig. 3A), indicating that it is not essential for phosphorylation of m/p-HA-PKB␣ on Thr 308 and Ser 473 by upstream protein kinase(s). Kohn et al. (23) also showed that a membrane-targeted form of PKB␣ lacking the PH do-main was highly active in unstimulated cells. However, it should be noted that PKB␣ constructs lacking the PH domain can be activated by insulin and IGF-1 (23,27), 3 under conditions where we detected no apparent translocation of the protein.
Basal PI3-K activity is required to maintain m/p-HA-PKB␣ in its phosphorylated, active state. This implies that the role of PI3-K in PKB activation is twofold: one is to provide membrane translocation in cooperation with the PH domain thus "priming" the kinase for the activation, and second is to enable phosphorylation and activation. Membrane targeting circumvents the former regulatory step, yielding the fully active, fully phosphorylated m/p-HA-PKB␣. This is different from v-Akt or Gag-PKB, whose activity can be stimulated further by mitogens in a wortmannin-sensitive manner (5,45). It appears therefore that the attachment of a short peptide carrying the Lck membrane-targeting motif provides an optimal conformation for maximal PKB activity, without the activation of PI3-K. This is in agreement with the data from Stokoe et al. (25,53), who demonstrated that PtdIns(3,4,5)P 3 through its binding to the PH domain facilitates phosphorylation and activation of PKB by Thr 308 kinase. In addition, the removal of the PH domain reduces or eliminates the requirement for PtdIns(3,4,5)P 3 for phosphorylation of Thr 308 (25,53), which also explains the ability of HA-PKB␣-⌬PH to be activated without the apparent membrane translocation. It therefore seems likely that the plasma membranes of unstimulated cells contain sufficient PtdIns(3,4,5)P 3 to maintain the activity of Thr 308 kinase/PDK1 at a high enough level to phosphorylate membrane-targeted PKB␣ at Thr 308 . PI3-K may also maintain m/p-HA-PKB␣ activity through the inhibition of PP2A (46), which is likely to be the in vivo PKB phosphatase (21).
The subcellular distribution of the wild-type kinase is tightly regulated. PKB␣ translocates to the plasma membrane, which can be detected after 2-min stimulation by IGF-1 (Figs. 9 -11) or vanadate. 4 The PH domain, which is required for the membrane translocation, may also be responsible for the detachment from the membrane, for example by binding inositol trisphosphate that is generated in the cell though the action of mitogen-stimulated phospholipase C␥ (47). The mechanism by which PKB translocates to the nucleus is still not clear. Kinase activity is not required for this process, judged by the fact that HA-PKB␣-K179A can also be detected in the nuclei of stimulated REF-52 cells. It is likely that the PH domain does not play a role in this process, as HA-PKB␣-⌬PH was also found in the nuclei of stimulated REF-52 cells. 5 According to the available data, we propose the following model for the regulation of PKB activity. After cell stimulation, the kinase translocates to the membrane, which allows a correct conformation for the activating phosphorylation. Following  Fig. 9. A, significant labeling of HA-PKB␣ in the cytoplasm of unstimulated cells with low labeling over the plasma membrane. B, the plasma membrane of cells stimulated with 100 ng/ml IGF-1 for 5 min is intensely labeled with gold particles and, in this case, there is little cytoplasmic label. C, lack of appreciable labeling at the plasma membranes in cells pretreated with 100 nM wortmannin for 10 min before IGF-1 stimulation. A and B, 7 nm protein A gold; C, 5 nm protein A gold. Bars, 100 nm.
FIG. 11. Changes in subcellular distribution of HA-PKB␣ during IGF-1 stimulation. 293 cells grown and transfected on coverslips were serum-starved for 12 h before stimulation with buffer or 100 ng/ml IGF-1. Cells were stained by the anti-HA epitope 12CA5 monoclonal antibody, followed by a biotinylated secondary antibody, and streptavidin-coupled to Texas Red and visualized by laser scanning microscopy. A, unstimulated cells showing mainly cytosolic staining. B and C, cells stimulated with IGF-1 for 2 and 5 min, respectively, with clear membrane staining. D, nuclear staining of cells stimulated with IGF-1 for 30 min. activation, PKB detaches from the membrane, which, in turn, enables it to phosphorylate its cytosolic, as well as nuclear substrates. Correct subcellular localization is therefore crucial for the activation of the kinase, and it may also allow its appropriate inactivation by phosphatases.
Unlike the wild-type kinase, the form of PKB␣ that is constitutively localized to the membrane escapes the physiological spatial regulation. The finding that PKB␣ is activated by membrane targeting may explain the oncogenic potential of v-akt, which encodes a fusion protein between the viral Gag and PKB␣ (2). Significantly, 40% of the v-Akt protein is membraneassociated since it is myristoylated at the N terminus (17). It was reported that the Gag-PKB␣ fusion protein, which mimics v-Akt, has increased activity and stimulates p70 s6k activity (5). Furthermore, expression of this form of the kinase in 3T3-L1 preadipocytes is sufficient to elicit cellular responses that normally require insulin/IGF-1, such as glucose transport and differentiation (12,48). The constitutively active, membranetargeted form of PKB may also induce other insulin/IGF-1evoked responses, such as cell survival caused by the inhibition of apoptosis (13,45,49,50).
In summary, we have demonstrated that forced membrane localization is sufficient for its full activation as a result of the essentially stoichiometric phosphorylation of Thr 308 and Ser 473 by upstream kinases. Transient membrane association is required for the physiological activation of PKB. The characterization of cytosolic and nuclear substrates of PKB is now critical to understand how the kinase mediates physiological responses to insulin and IGF-1, as well as its role in cell survival and transformation.