Discovery of lactoquinomycin and related pyranonaphthoquinones as potent and allosteric inhibitors of AKT/PKB: mechanistic involvement of AKT catalytic activation loop cysteines

In the past decade, molecular elucidation of the phosphoinositide-3-kinase (PI3K)/AKT (PKB) signaling pathway and epidemiologic studies have firmly established a central role for AKT in human malignancy (1). The AKT family (AKT1, AKT2, and AKT3) is characterized by an NH₂-terminal pleckstrin-homology domain and a COOH-terminal catalytic domain bearing the highest sequence homology to the AGC family founding members protein kinase A (PKA) and protein kinase C (PKC; ref. 2). AKT is the cellular homologue of the viral v-Akt encoded by the oncovirus Akt8 (3). In quiescent cells, AKT is expressed as an inactive form and becomes activated by phosphorylation upon translocation to the plasma membrane in response to growth factor–stimulated PI3K activation (4). This process is negatively regulated by the inositol 1,4,5-trisphosphate phosphatase PTEN, a major tumor suppressor in human. Elevated AKT activity occurs in ∼50% of all human malignancies via numerous mechanisms (reviewed in refs. 5–8), including constitutive activation of cell surface growth factor receptors, loss of PTEN, activation mutation of the PI3K catalytic subunit p110α (PIK3CA), as well as overexpression of various AKT family members. The role of AKT in cancer is mediated through a growing list of downstream targets that are directly phosphorylated by AKT (5–8). AKT promotes survival through several well-known apoptosis modulators such as the forkhead transcription factors, GSK3, BAD, Bcl-X_L, caspase 9, and nuclear signaling of nuclear factor-κB. AKT promotes cell growth and proliferation through activation of the mammalian target of rapamycin (mTOR) kinase, a central regulator of protein translation, and by influencing the levels of D-type cyclins and cell cycle inhibitor p27/Kip1. Moreover, AKT also plays a role in tumor-induced angiogenesis by regulating the expression of hypoxia-inducible factor 1α and vascular endothelial growth factor (9, 10).

AKT represents an appealing target for anticancer therapy. Numerous experimental approaches have further corroborated clinical evidence that deregulated AKT kinase activity causes oncogenic transformation in a variety of cell types and causes tumor growth in vivo. Inhibition of AKT by introducing either the PTEN tumor suppressor or a dominant-negative AKT into PTEN-deficient cancer cells leads to the inhibition of tumor growth (11, 12). Likewise, expression of antisense RNA against AKT resulted in growth inhibition and increased sensitivity to chemotherapeutic agents in a variety of cancer cell lines (13). The discovery of small molecule AKT inhibitors will further enhance our ability to validate the therapeutic potential of targeting AKT in cancer. These efforts have accelerated in recent years (reviewed in refs. 14, 15). Several groups have reported AKT inhibitors that target the ATP-binding pocket (16), the pleckstrin-homology domain (17), and upstream inhibitors that interfere with enzyme activation (18–20). Although significant progress has been made, the development of potent and selective AKT inhibitors is particularly challenging due to the extensive homology in ATP-binding sites of the AGC kinase family members. Consequently, a considerable interest now exists in searching for novel allosteric inhibitors of AKT. In this report, we describe our screening efforts that led to the identification of lactoquinomycin as a potent and selective inhibitor of AKT kinases. We show that this inhibitor class targets AKT through a novel allosteric mechanism that involves two critical activation loop (T-loop) cysteines neighboring the activating residue Thr³⁰⁸. Inhibition of cellular AKT by lactoquinomycin confirms the critical role of AKT in growth factor signaling and mRNA translation, which are essential for tumor cell growth and survival.

All general chemical reagents used for buffers and assays were purchased from Sigma Corporation unless otherwise specified. Lactoquinomycin and frenolicin B were obtained from Wyeth Natural Product sample collections. Cell cycle inhibitor-779 (rapamycin ester) and wortmannin were obtained from Wyeth Chemical & Pharmaceutical Development. Staurosporine was purchased from CalBiochem.

Human AKT1 cDNA was obtained by PCR from human placental Quick cDNA (BD-Clontech), sequenced and confirmed to be identical to the previous report (21). The AKT1 cDNA was inserted into the BamHI site of the mammalian expression vector pFlag-CMV5 (Sigma) in which AKT1 was COOH-terminally tagged with Flag-epitope. The myristoylation sequence derived from the human c-Src (22) was then added to the NH₂ terminus of AKT1 by PCR to generate myristoylated AKT1-Flag (myr-AKT1). This construct was then subjected to site-directed mutagenesis to generate T-loop cysteine mutants C310A, C296A, and C296A/C310A using the mutagenesis kit (Stratagene).

All cell culture media, supplements, and transfection reagents were obtained from Invitrogen. HEK293 cells were maintained in DMEM containing 10% fetal bovine serum, 100 μg/mL of penicillin, 50 μg/mL of streptomycin, and 1 mmol/L of glutamine. Plasmid DNA (50 μg per 150 mm culture plate) of various myr-AKT1 constructs were transiently transfected into HEK293 using LipofectAMINE 2000 reagent. Cells were harvested 48 h later with all steps done at 4°C. Cells (1–2 × 10⁷ per 150 mm plate) were washed with PBS and scraped off the plate in 1.5 mL AKT lysis buffer [25 mmol/L Hepes (pH 7.5), 100 mmol/L NaCl, 0.5% NP40, 1 mmol/L Na₃VO₄, 1 mmol/L EDTA, 1 mmol/L EGTA, 20 mmol/L β-glycerophosphate, 10 μg/mL aprotinin, 10 μg/mL leupeptin, 1 mmol/L phenylmethylsulfonyl fluoride, 1 μmol/L microcystin LR, 0.1% β-mercaptoethanol]. The cell suspension was then sonicated, incubated for 30 min with gentle shaking, and cleared by centrifugation for 30 min at 14,000 × g using a Beckman J2-HS centrifuge. The cleared lysate was collected and stored at −80°C. For purification of various AKT1 enzymes, frozen cell lysate was thawed on ice and added onto an anti-Flag M2 affinity column (Sigma) at 4°C following a ratio of 1 mL affinity beads per 100 mg crude lysate. The column was washed with 15 times the bed volume of TBS. The myr-AKT1 proteins were eluted using 100 μg/mL of Flag peptide (Sigma) in elution buffer [50 mmol/L Tris-HCl (pH 7.5), 150 mmol/L NaCl, 270 mmol/L sucrose, 1 mmol/L Na₃VO₄, 1 mmol/L benzamidine, 0.1 mmol/L EGTA, 0.2 mmol/L phenylmethylsulfonyl fluoride, 0.1% β-mercaptoethanol, 0.03% Brij-35]. Eluted proteins were quickly frozen in a dry-ice ethanol bath and stored at −80°C. Concentrations of all purified proteins were determined by the Bradford method (Bio-Rad) using bovine serum albumin as a standard.

The routine assays of AKT1 were done in low-binding 96-well plates (Corning). Myr-AKT1 (11.5 μL) or various mutant enzymes diluted in kinase assay buffer [50 mmol/L Hepes (pH 7.4), 100 mmol/L KCl, 25 mmol/L β-glycerophosphate, 10 mmol/L MgCl₂, 1 mmol/L orthovanadate and 0.5 mmol/L EGTA] with 50 μg/mL of bovine serum albumin were added to each well. One microliter of DMSO vehicle or test inhibitors was then added. The kinase reactions were initiated by the addition of 12.5 μL of assay mix containing kinase assay buffer, ATP, and a biotinylated GSK3α substrate peptide (SGRARTSSFA). The final reaction volume of 25 μL contained 6 ng of AKT1 (4 nmol/L), 10 μmol/L of biotin-GSK3α peptide, and 50 μmol/L of ATP. The assays were incubated for 1 h at room temperature and terminated with 25 μL of stop buffer [25 mmol/L Tris-HCl (pH 7.5), 20 mmol/L EDTA]. Phosphorylated biotin-GSK3α was detected by the time-resolved fluorescence resonance energy transfer Lance format with all reagents obtained from Perkin-Elmer. Product detection was done in a black low-binding plate (Dynex) in 50 μL of detection buffer [50 mmol/L Tris-HCl (pH 7.5), 150 mmol/L NaCl, and 0.5% bovine serum albumin] containing 100 ng/mL of phospho-GSK3 polyconal antibody (Cell Signaling Technology) labeled with Europium (Perkin-Elmer), 4 μg/mL of streptavidin-allophycocyanin, and 2.5 μL of the terminated kinase reaction mix. After incubating for 30 min at room temperature, the plates were read in a Victor plate reader (Wallac/Perkin-Elmer). In some experiments, AKT activity was also measured in a rate-based coupled enzyme assay in a black clear-bottomed 96-well plate (half well or regular well; Corning). In addition to appropriate concentrations of GSK3α peptide, ATP, and thiol-free AKT, the assay mixture contained 20 units/mL of pyruvate kinase, 28.5 units/mL of lactic dehydrogenase, 2 mmol/L of pyruvate enol phosphate, 0.25 mmol/L of NADH in an assay buffer of 100 mmol/L of Hepes (pH 7.5), 25 mmol/L of β-glycerophosphate, 10 mmol/L of MgCl₂, and 0.005% Brij-35. The assay (with a final volume of 120 μL in the half well plate or 300 μL in the regular plate) was monitored continuously by absorbance at 340 nm on a Gemini plate reader (Molecular Devices). Initial rates were calculated from the linear portion of progress curves, typically from 5 to 15 min.

Recombinant PKA and PKCα were obtained from Upstate Biotechnology and assayed for phosphorylation of the myelin basic protein (23). The PKCα assay included lipids (Upstate Biotechnology). The kinase reaction in a final volume of 25 μL contained kinase buffer (as AKT assay), PKA (2 μg/mL) or PKCα (0.8 μg/mL), 100 μg/mL of bovine serum albumin, 100 μmol/L of ATP, and myelin basic protein (80 μg/mL; Upstate Biotechnology). The assays were incubated for 1 h at room temperature and terminated by adding 25 μL of stop buffer. The phosphorylation of myelin basic protein was measured by the dissociation-enhanced lanthanide fluorescence immunoassay using an anti–phosphorylated (Ser/Thr)-PKA substrate antibody (Cell Signaling Technology) and a Europium-labeled anti-rabbit secondary antibody (Perkin-Elmer) in 100 μL assay buffer as previously described (24). Assays for other kinases were done with recombinant enzymes produced from bacterial, insect, or human cells. Substrates used were peptides (IKKβ, PDK1, S6K1, Src), proteins (CDK4, CDK2, CDC2, mTOR, Mek1), and poly(glutamic acid₄-tyrosine; KDR), and autophosphorylation (epidermal growth factor receptor, HER-2, c-Met). Phosphorylation was measured using TMB peroxidase substrate (Pierce) for CDKs and dissociation-enhanced lanthanide fluorescence immunoassay/Lance (Wallac-Perkin-Elmer) for others. The assays of a panel of 45 kinases were done as described by Invitrogen SelectScreen profiling.⁴

HEK293, Rat1, and various tumor cell lines used in this study were obtained from American Type Culture Collection. DU145-AKT was created by stable transfection of myr-AKT1 into DU145. MDA-MB-361 (MDA-MB-361-DYT2) was obtained from Dr. C. Discafani (Oncology, Wyeth Research, Pearl River, NY). For evaluation of cellular effects by inhibitors, tumor cells were plated in six-well culture plates in growth medium for 1 day. Cells were then treated with inhibitors for 6 or 16 h in growth media, or as indicated. For insulin-like growth factor-I stimulation experiments, serum-starved cells were treated with inhibitors for 2 h and then stimulated with insulin-like growth factor-I for 0.5 h. For transient expression of Flag-AKT-WT and Flag-AKT-C296A/C310A, DU145 cells were plated in six-well culture plates for 24 h. Cells were then transfected without (mock) or with 2 μg (per well) each DNA construct using LipofectAMINE 2000 reagent (Invitrogen). Forty-eight hours posttransfection, cells were treated with inhibitor for 12 h. Total cellular lysates were prepared with NuPAGE-LDS sample buffer (Invitrogen), sonicated, clarified by centrifugation, and resolved in appropriate NuPAGE gels following the protocols provided by the vendor (Invitrogen). For 7-methyl-GTP pull-down, treated MDA-MB-468 cells in 100 mm culture plates were lysed on ice for 30 min in 500 μL of AKT lysis buffer in which NP40 was replaced with 1% Tween 20. Lysates were collected and centrifuged for 10,000 × g for 5 min, and the protein concentration of the cleared lysate was then determined. Forty microliters of a 50% slurry of 7-methyl-GTP Sepharose (Amersham) was added to 0.5 mg of cleared lysate, and incubated while rocking for 2 h at 4°C. The cap-complexes were then collected and washed four times with lysis buffer, dissociated from the Sepharose by adding 50 μL of NuPAGE LDS sample buffer, heated to 70°C for 10 min, and resolved in NuPAGE gels. Protein blots were probed with antibodies: phospho-(P)-AKT (T308), P-AKT (S473), AKT, P-GSK3, GSK3, P-FKHRL1 (T32), P-ERK (T202/Y204), ERK, P-S6K1 (T389), P-S6 (S240/244), P-4EBP1 (T70), 4EBP1, eIF4E, P-IKKγ (S376), and IκBα (Cell Signaling Technology); P-FKHRL1 (T32), P-HER2 (Y1248), and HER2 (Upstate Biotechnology); cyclin D3, eIF4A, eIF4G, and eIF3b (Santa Cruz Biotechnology); and p27/Kip1 (Transduction Laboratories).

To establish an enzyme assay for novel inhibitors of active AKT1, we constructed an expression vector encoding myristoylated human AKT1 with a COOH-terminal Flag-epitope tag (myr-AKT1-Flag/myr-AKT1). This approach permitted the expression of a constitutively active form of recombinant AKT1 in HEK293 cells, obviating the requirement for in vitro phosphorylation of the purified enzyme by PDK1 (4). Anti-Flag affinity chromatography of the transfected HEK293 cell lysate yielded a relatively pure enzyme with a stoichiometry of Thr³⁰⁸ phosphorylation comparable to that observed with a commercial preparation of in vitro–phosphorylated AKT1 (Fig. 1A). In a homogeneous time-resolved fluorescence resonance energy transfer Lance assay, the myr-AKT1 efficiently phosphorylated a biotinylated-GSK3α peptide on Ser²¹ in a dose- and time-dependent manner (Fig. 1B). Apparent Michaelis constant (K_m) values of the enzyme for ATP and substrate peptide were determined as 46 ± 2.5 and 1.35 ± 0.37 μmol/L, respectively (Fig. 1C). These kinetic parameters determined in the Lance assay are in good agreement with a recent report on kinetic characterization of the AKT family enzymes (25). Based on these data, we developed and optimized the Lance assay containing 6 ng of myr-AKT1 (4 nmol/L), 50 μmol/L of ATP, and 10 μmol/L of substrate peptide with a kinase reaction time of 1 h.

Screening of the Wyeth Natural Products collection led to the identification of lactoquinomycin from the fermentation broth of a microbial strain Streptomyces sp. LL-AF101. A subsequent structural similarity search also uncovered a related inhibitor, frenolicin B. Both lactoquinomycin and frenolicin B are antibiotics possessing a common pyranonaphthoquinone core (Fig. 2A). Dose titration experiments with lactoquinomycin yielded mean IC₅₀ values of 0.133 ± 0.017 and 0.149 ± 0.045 μmol/L, with myr-AKT1 and commercially prepared AKT1, respectively (Fig. 2B, left), indicating an identical potency against both forms of active AKT1. Similar results were obtained for frenolicin B, in which IC₅₀ values against myr-AKT1 and AKT1 were 0.157 ± 0.038 and 0.313 ± 0.115 μmol/L, respectively (Fig. 2B, right). Because both compounds have the same pyranonaphthoquinone core, these data imply that the pyranonaphthoquinone moiety is critical for AKT inhibition. To evaluate the specificity of lactoquinomycin in AKT inhibition, we first determined the effects of this compound on PKA and PKCα, two AGC family members the catalytic domains of which exhibited the highest homology to AKT. Interestingly, we found that lactoquinomycin did not inhibit PKA or PKCα at concentrations up to 200 μmol/L (Fig. 2C, left), whereas staurosporine potently inhibited all three enzymes (Fig. 2C, right). IC₅₀ determination in a panel of 13 additional protein kinases revealed that lactoquinomycin weakly inhibited HER2/ErbB2 (IC₅₀, 2.42 μmol/L) and IKKβ (IC₅₀, 3.9 μmol/L), and had no effect on other protein kinases in this panel at concentrations up to 10 μmol/L (Fig. 2D). These initial interesting observations prompted us to examine its activity against a broader range of kinases. Lactoquinomycin was tested at 1 μmol/L in a panel of 45 kinases in duplicate (Invitrogen SelectScreen profiling). Mean percent inhibition values confirmed the potent inhibition of AKT1 (92%) and AKT2 (99%; Supplementary Table S1).⁵

To define the kinetic mechanism of lactoquinomycin, we first did inhibitor versus ATP matrix competition assays. AKT activity was measured in a rate-based pyruvate kinase/lactic dehydrogenase–coupled enzyme assay in the presence of constant substrate peptide and varying amounts of ATP and lactoquinomycin. Double-reciprocal plots of the initial enzyme rate versus ATP concentration were best-fit to noncompetitive inhibition (Fig. 3A). To assess whether AKT might be irreversibly inhibited, lactoquinomycin dose-response was assayed at various time points after initiation of the kinase reaction. IC₅₀ values determined at 2, 15, 30, 45, and 60 min were 1.7, 0.66, 0.31, 0.26, and 0.18 μmol/L, respectively (Fig. 3B), indicating a progressive, time-dependent decrease in IC₅₀ values, a characteristic of irreversible inhibitors that interact covalently with their target enzyme. Similar assays with a known reversible ATP-pocket inhibitor staurosporine yielded nearly identical IC₅₀ values in these assays (data not shown). Thus, lactoquinomycin seemed to act via a distinct kinetic mechanism that may involve its irreversible binding to AKT at a site outside of the ATP-pocket.

To further define the lactoquinomycin-AKT interaction, we examined the effect of this compound on an AKT1 mutant lacking the pleckstrin-homology domain. This deletion mutant was equally sensitive to lactoquinomycin and frenolicin B (data not shown), implying that these inhibitors bind within the catalytic domain. Because the pyranonaphthoquinone core of lactoquinomycin and frenolicin B could potentially react with thiols, we speculated that these inhibitors might interfere with catalytic activity through the T-loop cysteine(s) surrounding the critical activating phosphorylation residue Thr³⁰⁸. This model was particularly attractive in light of the inactive AKT2 crystal structure which shows that the T-loop is disordered and contains a redox-sensitive cysteine-disulfide bond (26, 27). To test this hypothesis, we did cysteine-competition experiments. The IC₅₀ values for lactoquinomycin and frenolicin B were increased by 66- and 55-fold, respectively, when AKT was assayed after preincubation with 1 mmol/L of cysteine (Fig. 4A), consistent with a potential involvement of cysteine(s) in the AKT-inhibitor interaction. To definitively test this model, we created T-loop cysteine mutants in myr-AKT1 for assessing inhibitor effects. AKT enzymes harboring either C310A or C296A and the double mutant C296A/C310A were expressed in HEK293 cells and affinity-purified (Fig. 4B, left). The purified AKT Cys-mutants were all active in the in vitro kinase assays; however, we noted that the activities of the C310A and C296A/C310A mutants were moderately reduced relative to the wild-type myr-AKT1 (Fig. 4B, right). In subsequent experiments, the wild-type and Cys-mutants of AKT were assayed for sensitivity to lactoquinomycin and frenolicin B (Fig. 4C). We found that although lactoquinomycin (Fig. 4C, top) efficiently inhibited the wild-type AKT (IC₅₀, 0.36 μmol/L), it only weakly inhibited C310A (IC₅₀, 38.5 μmol/L) and C296A (IC₅₀, 30.4 μmol/L), and was inactive against the double mutant C296A/C310A at concentrations up to 40 μmol/L. Similar results were obtained for frenolicin B (Fig. 4C, middle) with IC₅₀ values of 0.46, 39.6, and 27.4 μmol/L against the wild-type, C310A, and C296A, respectively. In contrast to the results with the pyranonaphthoquinone inhibitors, the ATP-competitive inhibitor staurosporine (Fig. 4C, bottom) as well as a chemically unrelated compound (data not shown) indiscriminately inhibited all four enzymes with nearly identical potencies. These observations suggest that the active site conformation in the Cys-mutants likely remained comparable to that of the wild-type AKT because no significant difference in the inhibitory activity was detected for these ATP inhibitors. We conclude from these data and additional experiments (see below) that the pyranonaphthoquinones inhibit AKT kinase activity through a binding mechanism that requires the T-loop Cys²⁹⁶ and Cys³¹⁰ residues in the catalytic domain.

Lactoquinomycin was examined for inhibition of AKT substrate phosphorylation in tumor cells. Treatment of PTEN-negative U87MG cells with 10, 5, 2.5, and 1.25 μmol/L of lactoquinomycin for 6 h resulted in a dose-dependent inhibition of GSK3 phosphorylation (Fig. 5A, left). To verify that the inhibition was not a result of targeting upstream kinases, we studied a stable DU145-myr-AKT1 (DU-AKT) cell line in which AKT is activated constitutively, in a PI3K-independent fashion. Phosphorylation of FKHRL1 was completely suppressed in cells treated for 6 h with lactoquinomycin, but not with the PI3K inhibitor wortmannin (Fig. 5A, middle). Lactoquinomycin at 2 μmol/L also blocked insulin-like growth factor-I–stimulated phosphorylation of FKHRL1 in serum-starved Rat1 cells (Fig. 5A, right). Notably, Thr³⁰⁸ and Ser⁴⁷³ phosphorylations of AKT were minimally inhibited under these acute treatment conditions (Fig. 5A, middle and right), further supporting a direct targeting of cellular AKT rather than upstream kinases. The lack of suppression of AKT phosphorylation in cells is consistent with the in vitro enzyme inhibition data in which lactoquinomycin did not inhibit PDK1 and mTOR, two known kinases of Thr³⁰⁸ and Ser⁴⁷³, respectively (Fig. 2D; data not shown). It was noted that lactoquinomycin dose-dependently increased phospho-ERK in these cells (Fig. 5A), which might be a consequence of AKT suppression and/or cellular stress (28, 29). To further address the role of T-loop cysteines in mediating AKT inhibition in cells, DU145 cells were transiently transfected with the Flag-tagged wild-type and C296A/C310A myr-AKT. Both constructs expressed comparable myr-AKT and induced phospho-FKHRL1 and phospho-GSK3 (Fig. 5B). Treatment with lactoquinomycin (1–6 μmol/L) resulted in a dose-dependent suppression of phospho-FKHRL1 and phospho-GSK3 in the wild-type AKT-transfected cells, and the inhibition of both markers were substantially reduced in the C296A/C310A AKT cells (Fig. 5B). These data are consistent with the biochemical data in Fig. 4 in confirming the critical involvement of T-loop cysteines in the inhibition mechanism of lactoquinomycin.

AKT is an upstream regulator of the mTOR in the canonical growth factor–stimulated cellular mRNA translation (reviewed in ref. 30). We found that both mTOR substrates S6K1 and 4EBP1 were constitutively phosphorylated in serum-starved DU-AKT cells as compared with the DU145 parent (Fig. 6A, left), indicating mTOR activation by AKT. We therefore determined whether lactoquinomycin would interfere with mTOR signaling and mRNA translation. Treatment of DU-AKT cells for 16 h with 2, 1, and 0.5 μmol/L of lactoquinomycin resulted in a dose-dependent decrease in phospho-S6K and phospho-4EBP1, whereas the inhibition was not observed in wortmannin-treated cells (Fig. 6A, right). Notably, Ser⁴⁷³ phosphorylation of AKT was also partially reduced with 2 μmol/L of lactoquinomycin (Fig. 6A, right), which may be a result of blocking the signaling of mTOR complex 2 after a relatively prolonged treatment. In the PTEN-negative MDA-MB-468 cells, lactoquinomycin at concentrations as low as 0.25 μmol/L substantially inhibited phosphorylation of S6K, S6, 4EBP1, and FKHRL1 (Fig. 6B). Notably, these alterations in mTOR substrate phosphorylation were accompanied by a marked reduction in cyclin D3- and an increase in p27/Kip1-expression. Both of the cell cycle regulatory proteins are indirectly regulated by mTOR signaling and were shown to be modulated by the mTOR inhibitor cell cycle inhibitor-779 in MDA-MB-468 cells (31). In additional studies, we examined the effects of lactoquinomycin on the structure of the cap-binding translation initiation complex, which is known to be modulated by AKT/mTOR signaling. 7-Methyl-GTP pull-down assays showed a profound dose-dependent increase in the binding of 4EBP1 to the translation initiation factor eIF4E (Fig. 6C), consistent with the known inhibitory effects of 4EBP1 on the assembly of an active, cap-binding translation initiation complex (32, 33). The increase in eIF4E-bound 4EBP1 was accompanied by a decrease in the coprecipitation of eIF4G and eIF4A with eIF4E (Fig. 6C). Treatment with the mTOR inhibitor cell cycle inhibitor-779 (50 nmol/L) caused a similar but relatively modest increase in eIF4E-bound 4EBP1 and loss of eIF4G and 4A (Fig. 6C). Similar results were obtained when we examined the levels of eIF3b-bound eIF4E. Although the control sample contained abundant eIF4E, samples from treated cells exhibited a concentration-dependent loss of eIF4E (Fig. 6D). As expected, a similar but partial loss of eIF4E was also observed in the cell cycle inhibitor-779–treated cells (Fig. 6D). Thus, it is clear from the data in Fig. 6 that inhibition of AKT by lactoquinomycin in these cells resulted in a substantial suppression of mTOR signaling and disruption of cap-initiation complexes eIF4F and eIF3, thereby inhibiting cap-dependent translation initiation, a rate-limiting step of cellular protein synthesis.

Despite widespread screening efforts, few small molecule inhibitors of AKT have been reported (14, 15). Moreover, the development of selective AKT inhibitors is hindered by the extensive conservation of the ATP-binding pockets of the AGC kinase family. Members of this family participate in a wide array of critical cellular functions, and broad targeting of this family might lead to adverse side effects. Consequently, AKT inhibitors that avoid the conserved ATP binding site have attracted considerable interest. In this regard, a class of AKT pleckstrin-homology domain inhibitors was reported to possess remarkable specificity for AKT over other AGC family members and to inhibit the phosphorylation of AKT in cells (17). It remains to be seen whether these inhibitors potently block the phosphorylation of AKT downstream substrates in tumor cells with constitutive PI3K/AKT signaling pathways (17).

In the present study, we have identified two pyranonaphthoquinones that potently inhibited AKT through targeting of the catalytic T-loop of the enzyme. Binding of these inhibitors to AKT did not require the pleckstrin-homology domain and did not discriminate appreciably between AKT1 and AKT2 isoenzymes. A novel mechanism involving an irreversible inhibitor–T-loop interaction was strongly supported by mutational analysis that highlighted the T-loop Cys²⁹⁶ and Cys³¹⁰ residues to be critical for sensitivity to the pyranonaphthoquinone inhibitors. This conclusion was further supported by the observation that transiently expressed C296A/C310A myr-AKT substantially attenuated the cellular inhibition of AKT substrate phosphorylation in response to lactoquinomycin. Mass spectrometry studies have been used to elucidate the specific covalent interaction of the same T-loop cysteines in AKT2 with lactoquinomycin and related pyranonaphthoquinones. These results and the proposed mode of chemical inactivation of AKT will be described elsewhere. In various tumor cell models, lactoquinomycin potently and acutely inhibited the phosphorylation of AKT downstream substrates GSK3α/β, FKHRL1, as well as mTOR (data not shown). In serum-starved Rat1 cells, this compound blocked the insulin-like growth factor-I–stimulated phosphorylation of known AKT substrates. These data thus indicate that lactoquinomycin could inhibit both the growth factor–stimulated as well as the constitutively active AKT signaling pathways in exponentially proliferating tumor cells. The observation that lactoquinomycin did not inhibit Thr³⁰⁸ phosphorylation of AKT itself in these cells indicates that binding of the inhibitor to the AKT T-loop does not interfere with the PDK1-dependent Thr³⁰⁸ phosphorylation. Alternatively, lactoquinomycin might preferentially target the phosphorylated form of AKT. Lactoquinomycin caused an increase of phospho-ERK in cells. Although the mechanism for ERK phosphorylation was not investigated in this study, previous reports have indicated a negative regulation of the Raf/Mek/ERK pathway by AKT as well as ERK activation in stress response (28, 29). Thus, an enhanced phospho-ERK induced by lactoquinomycin may reflect a consequence of AKT suppression in these cells and/or stress response. Nevertheless, despite its induction of phospho-ERK, lactoquinomycin inhibited the proliferation of a broad panel of cancer cells in the culture (data not shown).

The selective targeting of AKT over those tested AGC kinases by these pyranonaphthoquinones indicates that T-loop inhibition may be a viable strategy for developing novel inhibitors of AKT. This selectivity likely reflects the fact that, in contrast to the high degree of ATP binding site conservation among the AGC kinases, the T-loops of these protein kinases are relatively polymorphic (27). Although one or both of the equivalent T-loop cysteines are conserved in all AGC family members, lactoquinomycin was inactive against these highly related enzymes. It therefore seems that the T-loop region of the AKT family may confer a distinct conformational feature(s) that renders the two target cysteines susceptible to modification by the pyranonaphthoquinones. The AKT2 crystal structures reveal that the inactive AKT2 has an unstructured T-loop, which becomes structured upon phosphorylation of Thr³⁰⁹ by PDK1 (26, 27). This structuring of the T-loop is accompanied by the realignment of the rest of the catalytic domain, resulting in compatibility with both ATP and substrate binding. It is conceivable that binding of lactoquinomycin in the T-loop may either directly interfere with interaction of AKT with substrate or cause an inhibitor-induced conformational misalignment that abrogates catalysis. Interestingly, although the pyranonaphthoquinones represent the first class of T-loop inhibitors targeting AKT, the antiinflammatory natural product parthenolide was previously shown to bind the T-loop cysteine of IKKβ and block its kinase activity (34). At present, it is not known whether the T-loop cysteines of active AKT generally play a role in physiologic regulation of AKT activity in cells, an earlier report implicated redox regulation of T-loop cysteines of AKT2 in response to hydrogen peroxide–induced apoptosis, which involves disulfide bond formation between Cys²⁹⁷ and Cys³¹¹ and dephosphorylation by phosphatase 2A (35). We observed that inhibition of AKT by lactoquinomycin did not involve disulfide formation (data not shown) and did not require dephosphorylation of AKT in cells. It is noteworthy that lactoquinomycin also inhibited HER2 and IKKβ in vitro and in select cell models at higher doses (data not shown). Interestingly, the active site Cys⁸⁰⁵ of HER2 was previously implicated in the covalent inhibition by HKI-272 (36). However, the mechanism of action by lactoquinomycin in targeting HER2 remains to be elucidated.

The PI3K/AKT/mTOR pathway positively regulates the cap-dependent mRNA translation through dynamic phosphorylation of the mTOR substrates 4EBP1 and S6K1, which are crucial regulatory events required for the functional assembly of key translation initiation protein complexes (30, 32, 33, 37, 38). In particular, mTOR-dependent phosphorylation of 4EBP1 promotes its release from eIF4E bound to the mRNA 5′ cap structure (7-methyl-GTP; refs. 32, 33), allowing for the recruitment of the RNA helicase eIF4A and the scaffolding protein eIF4G. This process is essential for the recruitment of the 40S ribosomal subunit to the mRNA and is abrogated by the mTOR inhibitor rapamycin (32). Constitutive phosphorylation of 4EBP1 and/or S6K1 is believed to contribute to oncogenic transformation in tumors harboring deregulated PI3K/AKT/mTOR signaling (30, 39, 40). In DU-AKT cells, constitutive phosphorylation of S6K1 and 4EBP1 was normalized by lactoquinomycin. In exponentially proliferating MDA-MB-468 cells, lactoquinomycin-induced dephosphorylation of 4EBP1 correlated with a drastic increase in binding of 4EBP1 to eIF4E and disruption of cap-initiation complexes eIF4F and eIF3. Thus, our data, obtained with a selective chemical inhibitor of AKT, corroborated previous genetic and biochemical data in further establishing AKT as a critical positive regulator of cap-dependent mRNA translation.

Given its critical role in tumor growth and survival, AKT is an attractive target for the development of new therapies. Our results highlight T-loop targeting as a new strategy for the development of potent and selective AKT inhibitors for the treatment of cancer and other proliferative diseases. Nevertheless, lactoquinomycin itself is not suitable for therapy because the general redox properties of the naphthoquinones confer nonspecific cytotoxicity and might limit the therapeutic window (41). Interestingly, the AKT inhibition by the pyranonaphthoquinones does not seem to use the general redox mechanism, and AKT inhibition is not a common property of benzoquinones (data not shown). The specific structural features of the pyranonaphthoquinone scaffold are critical for AKT inactivation. Hence, new chemical analogues with further improved AKT inhibition potency, selectivity, and reduced redox properties might offer higher therapeutic potential. Our studies on the elucidation of the chemical structure and activity relationship of the pyranonaphthoquinones will be described elsewhere.

We thank Frank Erardi, Frank Ritaco, Jason Lotvin, Mark Tischler, Mairead Young, and Celine Shi for technical assistance; and Drs. Philip Frost, Guy Carter, and Jerauld Skotnicki for helpful discussions and generous support.