Characterization of a Protein Kinase B Inhibitor In Vitro and in Insulin-Treated Liver Cells

Actrapid (human insulin) was from Novo Nordisk (Bagsværd, Denmark); 2× Universal PCR Master Mix and No AmpErase UNG were from Applied Biosystems. Complete protease inhibitor cocktail tablets were from Roche and 8-(4-chloro-phenylthio)-cAMP from Calbiochem. All primers and probes were synthesized and purified by Sigma-Aldrich. All other chemicals were of the highest grade obtainable.

PKB isoform-specific antibodies have been characterized elsewhere (37). Antibodies to phospho-PKB (Thr³⁰⁸), phospho-PKB (Ser⁴⁷³), phospho-FOXO1 (Ser²⁵⁶), phospho-p42/p44 mitogen-activated protein kinase (MAPK) (Thr²⁰²/Tyr²⁰⁴), and phospho-S6 ribosomal protein (Ser^240/244) were from Cell Signaling Technology (Beverly, MA); anti-PKB was from Upstate Biotechnologies (Lake Placid, NY), FOXO antibody was from Dr. Graham Rena, University of Dundee (Dundee, U.K.), and total and phospho-TSC2 and -PRAS40 antibodies were generated in the Division of Signal Transduction and Therapy, University of Dundee. Anti–β-actin was purchased from Sigma-Aldrich (St. Louis, MO).

The dual PKBα/β inhibitor (compound 16h) was synthesized in house through a modification of the procedure of Lindsley et al. (39). Purity was established as >98% by H-NMR and LCMS. This compound (termed Akti-1/2 or Akt inhibitor VIII in the Calbiochem catalog) is a derivative of Akti-1/2a (a compound originally termed Akti-1/2 [38]) and exhibits increased potency.

Details of the kinase selectivity screens have previously been provided (40), and complete assay conditions are provided in supplementary data (available in an online appendix at http://dx.doi.org/10.2337/db07-0343). Briefly, all assays (25.5 μl at 21°C for 30 min) were performed using a Biomek 2000 Laboratory Automation Workstation in a 96-well format (Beckman Instruments, Palo Alto, CA). Reactions contained 5–20 mU purified kinase along with substrate peptide or protein and were initiated by the addition of 10 mmol/l MgAcetate and 5, 20, or 50 μmol/l ATP ([γ-33P]-ATP, 800 cpm/pmol).

HL1c rat hepatoma cells were maintained in 1g/l glucose containing Dulbecco's modified Eagle's medium supplemented with 5% FCS and 100 units/ml penicillin and 100 ug/ml streptomycin.

Following hormone treatments, total cellular RNA was extracted from HL1c cells using TRIreagent (Sigma-Aldrich) according to the manufacturer's instructions. cDNA was synthesized from total cellular RNA using the Superscript II reverse transcriptase kit (Invitrogen). Real-time PCR analysis was carried out using an ABI Prism 7700 sequence detector (Applied Biosystems). PCRs were carried out using the following cycling conditions: 50°C for 2 min ×1 and 95°C for 10 min ×1, followed by 40 cycles of 95°C for 15 s and 60°C for 1 min.

A reporter construct encompassing a luciferase cDNA under the control of a thymidine kinase gene promoter containing the IGFBP1 thymine-rich insulin response element (BP1-TIRE) was transfected with or without an expression construct for glutathione-S-transferase-FOXO1 (pEBG-FOXO1) into HL1c cells as previously described (26,33). Cells were incubated for 16 h with hormones as indicated in the figure legends before harvesting and assaying for luciferase activity.

Luciferase activity was determined using the Luciferase Assay system (Promega) according to the manufacturer's instructions. The data are expressed as relative light units (RLU) of luciferase activity per microgram of protein.

Following hormone treatments, cells were harvested by scraping into ice-cold lysis buffer (50 mmol/l Tris/HCl, pH 7.5; 50 mmol/l NaF; 500 mmol/l NaCl; 1 mmol/l Na vanadate; 1 mmol/l EDTA; 1% [vol/vol] triton X-100; 5 mmol/l Na pyrophosphate; 0.27 mmol/l sucrose; and 0.1% [vol/vol] 2-mercaptoethanol) and cell debris removed by centrifugation at 4°C for 10 min at 13,000g. Cell lysate (10–20 μg) was separated on Novex 4–12% NuPAGE gels. Following transfer to nitrocellulose, blots were blocked with 5% (wt/vol) nonfat milk in Tris-buffered saline (containing 0.1% [vol/vol] Tween 20) for 1 h and incubated with primary antibodies at 4°C overnight before incubation for 1 h at room temperature with the appropriate secondary antibody. Protein bands were visualized using an enhanced chemi-luminescence kit (Amersham Biosciences).

HL1c cells were harvested as above and protein concentration measured. Cell extract (0.1 mg) was incubated for 1 h on a shaking platform with protein G sepharose conjugated to the appropriate antibody. The immunocomplexes were pelleted and washed twice with 1 ml lysis buffer and twice with 1 ml assay buffer (50 mmol/l Tris/HCl, pH 7.5; 0.1 mmol/l EGTA; and 0.1% [vol/vol] 2-mercaptoethanol). The immunoprecipitated kinase activity was measured in a total volume of 50 μl, containing 50 mmol/l Tris/HCl, pH 7.5; 0.1 mmol/l EGTA; 2.5 μmol/l PKI; 10 mmol/l MgCl₂; 0.1 mmol/l [γ³²P]-ATP (2 × 10⁶ cpm/nmol); and 30 μmol/l crosstide (GRPRTSSFAEG). One unit of kinase activity is the amount that catalyzes the phosphorylation of 1 nmol/l substrate in 1 min.

Akti-1/2 was synthesized as previously described (39) and dissolved in DMSO for in vitro and cell culture experiments. The compound was assayed against a panel of 70 protein kinases that includes representatives of all the major protein kinase families (Table 1). Akti-1/2 is known to inhibit PKBα (58 nmol/l half-maximal inhibitory concentration) and PKBβ (210 nmol/l) much more potently than PKBγ (2,119 nmol/l) (39). PKB lacking a PH domain is not inhibited by this class of compound, as previously reported (39); however, full-length PKBα and -β were inhibited significantly in vitro by submicromolar concentrations of Akti-1/2 (Table 1). CamKI was the only other protein kinase significantly affected by 1 μmol/l Akti-1/2 (78% inhibition). Meanwhile, smooth muscle myosin light-chain kinase (smMLCK) was potently inhibited at 10 μmol/l, and RSK1, SGK (serum- and glucocorticoid-inducible kinase)1, MAPK-activated kinase 2, Aurora B, MELK, PIM (proviral integration site)1, PIM3, EFK2, p21-activated kinase 4, and CSK all show some sensitivity to Akti-1/2 at this concentration. Previously, analysis of related compounds had suggested a strong selectivity over other members of the AGC family (e.g., cAMP-dependent protein kinase, SGK, and protein kinase C) (38,39). Our data confirm a selective advantage toward PKBα and PKBβ of 10- to 100-fold when compared with 70 protein kinases, with the exception of CamKI and PKBγ (at least in vitro).

Previous work had suggested that PKBγ was not a major contributor to total PKB activity in isolated rat hepatocytes (37). Using isoform-specific antibodies, each PKB isoform was sequentially immunoprecipitated and assayed from cells incubated with or without insulin. The main insulin-induced PKB activity was PKBα, with some PKBβ stimulation (Fig. 1A and Table 2); however, no PKBγ activity could be detected (although this antibody will immunoprecipitate PKBγ activity from L6 myotubes [37]). Agents that induce gluconeogenic gene expression, such as the synthetic glucocorticoid dexamethasone and/or cell-permeable cAMP, did not alter the PKB isoform activities or response to insulin in these cells (Fig. 1A). The activation of PKBα by insulin is rapid and sustained for at least 3 h, even in the presence of dexamethasone and cAMP (Fig. 1B), while significant activation can be achieved with 0.1–1 nmol/l insulin—although 10 nmol/l is required for full activation (Fig. 1C).

The absence of PKBγ from HL1c cells suggested that Akti-1/2 could be used to fully inhibit PKB activity at concentrations that will not affect most other protein kinases. Although the reported half-maximal inhibitory concentration of Akti-1/2 is 58, 210, and 2,119 nmol/l against PKBα, PKBβ, and PKBγ, respectively (39), 5 μmol/l Akti-1/2 was previously used to completely inhibit PKBα and PKBβ activity in intact cells (with no effect on endogenous PKBγ) (41). Therefore, we examined insulin activation of PKB in the presence of up to 10 μmol/l Akti-1/2 (Fig. 2 and Table 2). Significant reductions in both PKBα and PKBβ activities are observed in cells exposed to 0.1–1.0 μmol/l Akti-1/2 acutely or chronically, irrespective of the presence of dexamethasone/cAMP (Table 2). In addition, PKB phosphorylation in response to insulin is almost completely lost in the presence of 0.5 μmol/l Akti-1/2 (Fig. 2A). This is consistent with no PKBγ activity being present in this cell line, as the phospho-specific antibody recognizes all three isoforms of PKB. Cell lysates were then probed for insulin-induced phosphorylation of three known PKB substrates, namely, TSC2, PRAS40, and FOXO1 (Fig. 2A). A similar profile of Akti-1/2 sensitivity is observed for these substrates, with complete loss of insulin-induced phosphorylation in the presence of 0.5–1.0 μmol/l Akti-1/2. This confirms that these proteins are PKB substrates and demonstrates that 1 μmol/l Akti-1/2 almost completely prevents acute insulin activation of PKB (Table 2). Meanwhile, insulin fully activates p42/p44 MAPK even in the presence of 30 μmol/l Akti-1/2 (Fig. 2B).

It was noted that insulin-induced PKBα activity (Fig. 1B), as well as Thr³⁰⁸ and Ser⁴⁷³ phosphorylation (Fig. 3A), is sustained for several hours, although it returns to basal after 16 h of exposure to insulin. The presence of 1 μmol/l Akti-1/2 prevents the appearance of insulin-induced phosphorylation of PKB throughout this time period (Fig. 3A). However, some phosphorylation of the PKB substrates FOXO1 (Ser²⁵⁶) and TSC2 (Thr¹⁴⁶²) is detectable after 1–3 h of exposure to insulin even in the presence of 1 μmol/l Akti-1/2 (Fig. 3A and C). This suggests that prolonged insulin exposure induces a small amount of PKB signaling even in the presence of this concentration of inhibitor, although we cannot measure this at the level of PKB phosphorylation. Meanwhile, higher concentrations of Akti-1/2 completely block downstream PKB signaling in response to insulin (Fig. 3B). Indeed, the presence of 10 μmol/l Akti-1/2 not only blocks insulin-induced PKB signaling but also reduces basal phosphorylation of some substrates (e.g., TSC2 and FOXO1) (Fig. 3B and D). Therefore, we conclude that 1 μmol/l Akti-1/2 prevents PKB activation by insulin over short incubation periods but that higher concentrations of this compound are required to maintain a complete block of PKB signaling for 3 h and longer.

Insulin inhibits the transcription of the PEPCK, G6Pase, and IGFBP1 genes (Fig. 4). These gene promoters are normally active in the fasted state (glucagon and glucocorticoids) but are then turned off after feeding (insulin) (17). In the HL1c cell line, treatment with dexamethasone (synthetic glucocorticoid) and 8CPT-cAMP (to mimic glucagon signaling) strongly induces the transcription of each of the genes. Simultaneous exposure to insulin completely blocks the effects of the inducing hormones (Fig. 4). Interestingly, 1 μmol/l Akti-1/2 has almost no effect on the regulation of the genes by any hormone (Fig. 4A). However, 10 μmol/l Akti-1/2 significantly enhances the induction of all three genes by glucocorticoids and cAMP (Fig. 4B) and also prevents insulin repression of IGFBP1 and PEPCK. This concentration of Akti-1/2 only partially reduces the ability of insulin to repress G6Pase (Fig. 4B). The data argue that PKB activity is required for full insulin repression of these genes; however, it also suggests that only a relatively small amount of PKB activation (<5%) is actually required to permit insulin action, since 1 μmol/l Akti-1/2 has little apparent effect on the action of insulin, despite strong antagonism of insulin signaling (Fig. 3A and C). The enhanced dexamethasone/cAMP induction of all three genes by complete PKB inhibition is similar to the effect of PI3 kinase inhibitors (20). Basal PI3 kinase activity is relatively high in this cell line, and there is measurable PKB activity (Fig. 1) in the absence of insulin. Therefore, it is likely that this inductive effect is due to removal of this basal PKB activity.

The data above demonstrate that ∼90% inhibition of PKB (1 μmol/l Akti-1/2) has little affect on insulin repression of PEPCK, G6Pase, and IGFBP1 (Fig. 4). However, these experiments were conducted with a supraphysiological concentration of insulin (10 nmol/l). Insulin repression of each gene is almost maximal in cells exposed to 1 nmol/l insulin. Activation of PKB at these insulin concentrations is relatively poor (Fig. 1C), consistent with only a small amount of PKB activation being required for gene repression. Incubation of cells with 1 μmol/l Akti-1/2 greatly reduces repression of the genes by 0.01–1.00 nmol/l insulin, but a strong repression remains at 10 nmol/l (Fig. 5). Therefore, insulin resistance due to reduced PKB activity can be overcome by hyperinsulinemia in this system.

PKB regulation of these genes potentially requires the phosphorylation and inhibition of the FOXO family of transcription factors. Insulin-induced PKB phosphorylation of FOXO1 at Ser²⁵⁶ is blocked by Akti-1/2, with phosphorylation reduced to below basal in cells exposed to 10 μmol/l Akti-1/2 (Fig. 3D). However, phosphorylation is not completely blocked over a 16 h incubation period in the presence of 1 μmol/l Akti-1/2 (Fig. 3C). Therefore, we compared the effects of 1 and 10 μmol/l Akti-1/2 on insulin inhibition of FOXO1 in the HL1c cells (Fig. 6). Cells tranfected with BP1-TIRE (a FOXO reporter construct) express luciferase activity that is reduced ∼50% by incubation with insulin (26). The expression of a glutathione S-transferase–FOXO1 fusion protein induces luciferase expression, but this remains sensitive to repression by insulin (Fig. 6B) (26). The ability of insulin to reduce luciferase expression is blocked by the presence of 1 or 10 μmol/l Akti-1/2, regardless of whether FOXO1 is overexpressed (Fig. 6). Meanwhile, the induction of luciferase expression by FOXO1 was enhanced in the presence of 10 μmol/l Akti-1/2 (Fig. 6B), similar to the effect on glucocorticoid induction of endogenous PEPCK, G6Pase, and IGFBP1 mRNA (Fig. 4B).

This work establishes Akti-1/2 as a selective tool for the characterization of PKB function in liver. The lack of hepatic PKBγ combined with a 10- to 100-fold selective preference for PKB over 70 protein kinases means that PKB can be completely inhibited in liver with little effect on other phosphorylation pathways. Of course, we cannot be certain that other enzyme classes are not affected by Akti-1/2, but our data provide details of the minimal effective concentrations required to block PKB signaling in cells, thereby reducing nonspecific effects.

Insulin action in the liver and muscle is vital for glucose homeostasis and, when disrupted, produces prolonged hyperglycemia (diabetes). In type 2 diabetes (the major form of the disease), insulin and the insulin receptor appear to be present and functional. However, much evidence is accumulating that the signaling processes that link the receptor to the regulation of glucose metabolism are defective (34,42–45). Reduced PKB signaling is a proposed mechanism in the development of insulin resistance and diabetes in humans (34). Complete ablation of PKB is embryonic lethal; however, isoform-specific genetic analysis suggests that PKBα is required for normal hepatic function and growth, while PKBβ is required for regulation of hepatic glucose output (34). The precise action of PKB in the regulation of this process is not established, although the data were consistent with a reduced regulation of gluconeogenesis following feeding. One likely explanation for this defect in the PKBβ-deficient animals is loss of insulin repression of the gluconeogenic genes PEPCK and G6Pase. However, there is some evidence that PKB activity is not required for insulin repression of these genes (22,29,32). In this work, we have used a biochemical and pharmaceutical approach to fully establish a key role for PKB in the insulin regulation of specific gene promoters in the liver. We show that complete pharmacological inhibition of PKB in a liver cell line prevents insulin repression of PEPCK and reduces regulation of G6Pase. This argues that PKB is required for proper insulin repression of these genes and suggests that the defects in the PKBβ-deficient animals are due to loss of their repression following feeding. It will be of interest to establish whether a PKBβ-specific inhibitor has a similar effect on this action of insulin; however, an isoform-specific inhibitor is not currently available. It could be argued that the inhibitory effect of 10 μmol/l Akti-1/2 on gene regulation is due to inhibition of CAMK1, smMLCK, or potentially one of the other kinases partly affected at this concentration of compound in vitro (Table 1). However, CAMK1 is poorly expressed in liver and is a calcium- but not insulin-induced enzyme. Similarly, smMLCK is a calcium-responsive rather than insulin-regulated kinase and is not really affected by the inhibitor at 1 μmol/l, a concentration that alters gene regulation (Fig. 5). In addition, neither of these kinases is regulated by PI3 kinase or PDK1 signaling, both of which are required for insulin regulation of the genes (19,20,22).

Importantly, we find that PKB inhibition (1 μmol/l, where smMLCK is not affected) reduces the sensitivity of the genes to insulin. Therefore, reduced PKB signaling capacity results in reduced insulin sensitivity of the gene promoters but does not prevent complete repression of the genes if the cells are exposed to high insulin concentrations. This is very similar to the presumed phenotype in an insulin-resistant (pre-diabetic) human. In these individuals, hyperglycemia does not ensue because hyperinsulinemia overcomes the effect of insulin resistance on glucose metabolism.

We find that only a small fraction of maximal PKB activation is required to repress PEPCK, G6Pase, and IGFBP1 transcription. First, 0.1 nmol/l insulin weakly activates PKB (<10% of maximum) yet strongly inhibits gene expression. Second, >90% inhibition of maximal PKB induction (in cells exposed to 1 μmol/l Akti-1/2) has no effect on the degree of repression of the genes by 10 nmol/l insulin. This is consistent with PKB expression studies where only ∼5% activation of an MER-PKB (a tamoxifen-inducible form of PKB) gives maximal effects on a downstream gene reporter (23). The concept that “excess” signaling capacity allows increased sensitivity to insulin was first proposed by Kono and Barham (46), and this concept of functional reserve may now be extended to PDK1 and PKB, as well as insulin-binding capacity. Alternatively, it is possible that PKB is part of an amplification cascade, and thus only small amounts of activity are sufficient to maximally regulate downstream components. However there was little evidence that the downstream components that we measured (FOXO, TSC2, and PRAS40) were becoming fully phosphorylated at reduced PKB activation; indeed, their phosphorylation profile mirrored that of PKB quite closely. Hence, either these are not the PKB targets that link the pathway to the PEPCK, G6Pase, or IGFBP1 gene promoters or they are not part of an amplification cascade.

Our data also suggest that the contribution of PKB signaling to each of these gene promoters is different. For example, complete loss of PKB activation does not completely prevent insulin action on G6Pase transcription. Similarly, genetic deletion of PDK1 from liver prevents regulation of PEPCK and IGFBP1 by feeding but only partially reduces the response of the G6Pase gene promoter (19). Therefore, there is a PKB-independent regulation of this gene promoter that does not regulate the PEPCK or IGFBP1 gene promoters. This is consistent with several other studies that suggest that insulin regulates these three gene promoters by different mechanisms despite the existence of a related insulin response sequence in each (18,21,26,28,47). Nakae et al. (32) have previously demonstrated a PKB-independent regulation of FOXO, a transcription factor closely associated with the insulin repression of these gene promoters. However, we find that the isolated IGFBP1 insulin-response element (BP1-TIRE) is not responsive to insulin in cells exposed to only 1 μmol/l Akti-1/2. Overexpression of FOXO1 does not alter this sensitivity to Akti-1/2, suggesting that insulin repression of FOXO1 activity is completely PKB dependent. It is perhaps dangerous to directly compare the regulation of endogenous gene promoter with this overexpression system; however, the data would suggest that FOXO1 is not the PKB-independent target of this insulin cascade. Similarly, the distinct Akti-1/2 sensitivity of the endogenous gene promoters and the FOXO reporter construct would argue that FOXO is not the only PKB target that regulates the endogenous promoters. IGFBP1 (but not PEPCK or G6Pase) repression by insulin is sensitive to the mTOR inhibitor rapamycin (26). Clearly, PKB and mTOR activation are therefore required for full repression of IGFBP1, but only PKB activation is required for repression of PEPCK, while PKB activation and an additional PKB-independent pathway contribute to full repression of G6Pase expression.

In summary, we have established the importance of PKB activation in the insulin regulation of PEPCK, G6Pase, and a third insulin-regulated gene, IGFBP1; found further evidence of a difference in the mechanism by which insulin regulates each gene; and identified Akti-1/2 as a useful tool to delineate PKB function in liver.

C.S. is the recipient of the Diabetes UK Senior Research Fellowship (BDA:RD02/0002473), and Y.L.W. is a recipient of a Foulkes Foundation Fellowship. This work was supported by the Medical Research Council and the Chief Scientist Office, Scottish Executive, Edinburgh, U.K.

We thank the members of the Division of Signal Transduction and Therapy for help with the kinase profiling and Dr. Sam Barnett (Merck) for helpful discussions.