Aberrations within the PI3K/AKT signaling axis are frequently observed in numerous cancer types, highlighting the relevance of these pathways in cancer physiology and pathology. However, therapeutic interventions employing AKT inhibitors often suffer from limitations associated with target selectivity, efficacy, or dose-limiting effects. Here we present the first crystal structure of autoinhibited AKT1 in complex with the covalent-allosteric inhibitor borussertib, providing critical insights into the structural basis of AKT1 inhibition by this unique class of compounds. Comprehensive biological and preclinical evaluation of borussertib in cancer-related model systems demonstrated a strong antiproliferative activity in cancer cell lines harboring genetic alterations within the PTEN, PI3K, and RAS signaling pathways. Furthermore, borussertib displayed antitumor activity in combination with the MEK inhibitor trametinib in patient-derived xenograft models of mutant KRAS pancreatic and colon cancer.
Borussertib, a first-in-class covalent-allosteric AKT inhibitor, displays antitumor activity in combination with the MEK inhibitor trametinib in patient-derived xenograft models and provides a starting point for further pharmacokinetic/dynamic optimization.
With its key roles in various cellular processes including cell proliferation, metabolism, and cell survival, the PI3K/AKT pathway is overactivated in several human cancers, contributing to tumor development, progression, and metastasis (1–3). Aberrations among members of this pathway have been described as oncogenic drivers in diverse cancer types, such as activating point mutations in PI3K and gene deletion or loss-of-function mutations in the tumor suppressor PTEN (4). These genetic lesions result in the augmented generation of the second messenger phosphatidylinositol (3,4,5)-trisphosphate, leading to hyperactivation of 3-phosphoinositide dependent protein kinase 1 (PDK1) and its substrate AKT, a serine/threonine-specific kinase, also known as protein kinase B. This step acts as a major signaling node in the PI3K/AKT pathway with hundreds of downstream substrates (5, 6).
In addition to aberrant AKT activity caused by genetic lesions in upstream-acting proteins, overexpression and activating mutations have been observed for all three AKT isoforms, for example in lung, prostate, breast, endometrium, and skin carcinomas (7). Mutations in AKT1 occur most frequently with a mutation rate of 2%–3% in urinary and bladder cancer, in which the somatic activating AKT1E17K mutation within the regulatory PH domain is the most prominent genetic lesion and also described as a driver mutation for the rare Proteus syndrome (8–10). Furthermore, overexpression of AKT is associated with resistance to several chemotherapeutics (11). Together, this evidence underlines the crucial role of the PI3K/AKT pathway in cancer progression and highlights the great potential for precisely targeted therapeutic intervention involving this signaling cascade.
Traditional ATP-competitive AKT inhibitors such as capivasertib (AZD5363; refs. 12, 13) and ipatasertib (GDC-0068; refs. 14, 15) are under clinical investigation in phase I and II studies. In contrast to these, allosteric inhibitors, including MK-2206 (16), miransertib (ARQ092; ref. 17), and BAY1125976 (18), which bind to the inactive kinase conformation of AKT, exhibit exquisite target selectivity solely for AKT1–3 while sparing structurally closely related AGC kinases, for example, p70S6K and protein kinase A.
Recently, we reported the development of a covalent-allosteric AKT inhibitor, borussertib. This inhibitor specifically binds to two noncatalytic cysteines in AKT at positions 296 and 310 by decorating allosteric ligands with electrophilic warheads at suitable positions, thus enabling the irreversible stabilization of the inactive conformation (19). Biochemical analyses have revealed superior inhibitory properties compared with reversible ATP-competitive and allosteric AKT inhibitors. Despite initial concerns associated with covalent kinase inhibitors, recently developed drugs demonstrated tremendous success in the clinics with major beneficial impact on patients’ survival rate [e.g., osimertinib (non–small cell lung cancer) and ibrutinib (chronic lymphocytic leukemia); refs. 20, 21].
In this article, we report the first crystal structure of AKT1 in complex with a covalent-allosteric inhibitor, borussertib (19), showing the unique covalent bond between inhibitor and Cys296. Furthermore, we demonstrate its antiproliferative activity for a panel of cancer cell lines harboring genetic alterations in PI3K, PTEN, and RAS. To investigate cellular pharmacodynamic changes induced by on-target inhibition of AKT, western blot studies were performed, and target selectivity was further corroborated by PathScan analyses. Subsequently, pharmacokinetic characterization and patient-derived xenograft (PDX) experiments were conducted, with results demonstrating the potential for optimization and development of orally bioavailable, targeted covalent-allosteric AKT inhibitors and their applicability in the mono- or combination therapy of different cancers.
Materials and Methods
Protein expression, purification, and crystallization
A gene encoding for AKT1(2-446, E114/115/116A) including an N-terminal His6-tag followed by a TEV protease recognition site was synthesized by GeneArt AG and cloned into the pIEx/Bac3 expression vector (Merck Millipore) using NcoI and BamHI restriction sites. Transfection, virus generation, and amplification as well as protein expression were carried out in Spodoptera frugiperda (Sf9) cells (Thermo Fisher Scientific) following the BacMagic protocol (Merck Millipore). Infected insect cells were grown in Erlenmeyer flasks for 72 hours at 27°C with shaking at 120 rpm, subsequently harvested by centrifugation at 3,000 × g for 15 minutes and washed once with PBS before being flash frozen in liquid nitrogen. Afterwards, cells were thawed and resuspended in lysis buffer [50 mmol/L Tris, 500 mmol/L NaCl, 1 mmol/L DTT, 10% glycerol, 0.1% Triton X-100, pH 8.0, EDTA-free protease inhibitor cocktail (Sigma-Aldrich)]. Cells were lysed using a microfluidizer, the lysate was cleared by centrifugation (40,000 × g, 1 hour). The supernatant was loaded onto a Ni-NTA Superflow Cartridge (Qiagen). Bound protein was eluted in buffer containing 50 mmol/L Tris, 500 mmol/L NaCl, 500 mmol/L imidazole, 1 mmol/L DTT, 10% glycerol, pH 8.0. For cleavage of the His6-tag, TEV protease was added to the pooled elution fractions and dialyzed overnight into buffer containing 25 mmol/L Tris, 50 mmol/L NaCl, 1 mmol/L DTT, 5% glycerol, pH 8.0 at 4°C. The cleaved protein was further purified by anion-exchange chromatography using a HiTrap Q HP Column (GE Healthcare) followed by size-exclusion chromatography on a HiLoad 16/60 Superdex 75 pg Column (GE Healthcare) using buffer containing 50 mmol/L HEPES, 200 mmol/L NaCl, 1 mmol/L DTT, 10% glycerol, pH 7.3. Afterwards, the protein was transferred into the storage buffer (25 mmol/L Tris, 100 mmol/L NaCl, 1 mmol/L DTT, 10% glycerol, pH 7.5) using a Superdex 75 10/300 GL Column (GE Healthcare), concentrated and stored at −80°C.
For crystallization, purified protein at a concentration of 3 mg/mL was incubated with 3 equivalents of borussertib on ice for 60 minutes. The samples were centrifuged at 20,000 × g for 10 minutes before hanging drops were prepared in 15-well crystallization plates (EasyXtal Tool, Qiagen) by mixing protein–ligand complex with reservoir solution (1:1) containing 1.25 mmol/L sodium acetate pH 5.2, 3.75 mmol/L sodium citrate pH 5.2, and 15% PEG MME 2000 at 20°C. Diffraction-grade crystals grew within 3 days and were cryoprotected using 20% ethylene glycol before they were flash cooled in liquid nitrogen. X-ray diffraction data were collected at the PXII-X10SA beam line of the Swiss Light Source (Paul Scherrer Institute, Villingen, Switzerland) with wavelengths close to 1 Å. The diffraction data were integrated with XDS (X-ray Detector Software) program package (22) and scaled using the program XSCALE (22). The crystal structure was solved by molecular replacement with PHASER (23) using a cocrystal structure of Akt1 in complex with another covalent-allosteric inhibitor as template (24). The manual modification of the molecule of the asymmetric unit was performed using the program COOT (25) and with the help of the Dundee PRODRG server (26) the inhibitor topology files were generated. For multiple cycles of refinement, phenix.refine (27) was employed and the final structure was evaluated by Ramachandran plot analysis using the server MolProbity (28). Final validation of the model was performed with the help of the PDB_REDO server and the crystal structure was visualized using PyMOL (29, 30).
Cell culture and inhibitors
T-47D and ZR-75-1 cell lines were purchased from Sigma-Aldrich/ECACC. KU-19-19 cells were purchased from DSMZ. AN3-CA and HPAF-II cells were obtained from ATCC. BT-474, Dan-G, and MCF-7 were purchased from CLS Cell Lines Service. Cell lines were cultured in DMEM, MEM, or RPMI1640 medium (Gibco) supplemented with 10% FBS (PAN-Biotech) and 1% penicillin/streptomycin (Gibco). For AN3-CA and MCF-7, 1 mmol/L sodium pyruvate (Gibco) was added to the growth medium and media for BT-474 and T-47D were supplemented with 10 μg/mL insulin (Sigma-Aldrich). Bo103 cells were cultured in a 1:1 mixture of DMEM/F12 (Gibco) and DMEM (Gibco), supplemented with 5% FBS (Gibco), 2% penicillin/streptomycin (Gibco), 1.6 μg/mL Amphotericin B (Gibco), 10 μmol/L ROCK inhibitor Y-27632 (LC Laboratories), 10 μg/mL ciprofloxacin (Sigma-Aldrich), 8.4 ng/mL cholera toxin (Sigma-Aldrich), 10 μg/mL insulin (Sigma-Aldrich), 20 nmol/L 1-thioglycerol (Sigma-Aldrich), and 0.5 mmol/L sodium pyruvate (Gibco). Cells were cultured in a humidified incubator at 37°C, 5% CO2 and cell line authenticity was confirmed by STR analysis at Microsynth AG or by SNP profiling at Multiplexion. Mycoplasma testing has not been performed. Cells were used for viability and western blot analyses within 8 weeks after thawing.
Borussertib was synthesized as described previously (24). Reference inhibitors capivasertib, ipatasertib, MK-2206, and miransertib were purchased from SelleckChem; staurosporine and Y-27632 were obtained from LC Laboratories; trametinib was obtained from LC Laboratories and Hycultec.
Cell viability analysis
On day 0, cells were plated into white 384-well cell culture plates (Greiner Bio-One) using a Multidrop reagent dispenser (Thermo Fisher Scientific) at cell numbers that ensure linear and optimal luminescent signal intensity (AN3-CA: 800 cells/well; BT-474: 400 cells/well; Dan-G: 400 cells/well; HPAF-II: 400 cells/well; KU-19-19: 400 cells/well; MCF-7: 200 cells/well; T-47D: 800 cells/well; ZR-75-1: 400 cells/well; and Bo103: 800 cells/well). Following incubation for 24 hours in a humidified atmosphere at 37°C/5% CO2, cells were treated with inhibitors in serial dilutions ranging from 30 μmol/L down to 0.1 nmol/L using an Echo 520 acoustic liquid handler (Labcyte Inc.). Cell viability was analyzed on day 5 using the CellTiter-Glo Assay (Promega) as per the manufacturer's instructions. Luminescence was recorded using an EnVision Multilabel 2104 Plate Reader (PerkinElmer) using 500 ms integration time. The obtained data were normalized to the plate positive control (30 μmol/L staurosporine) and negative control (DMSO) and subsequently analyzed and fitted with the Quattro Software Suite (Quattro Research) using a four parameter logistic model. As quality control, the Z′-factor was calculated from 16 positive and negative control values. Only assay results showing a Z′-factor ≥ 0.5 were used for further analysis. All experimental points were measured in duplicates for each plate and were replicated in at least three plates.
Combination studies were conducted using the CompuSyn Software (Biosoft) for calculating the combination index (CI) equation to determine synergism of drug combinations using fixed drug ratios as described by Chou-Talalay (31).
Western blot and PathScan analysis
For protein isolation, cells were seeded into 6-well tissue culture plates (Sarstedt) yielding 80%-90% confluency after overnight incubation. Afterwards, cells were treated with various concentrations of inhibitors or DMSO and incubated for additional 24 hours before the medium was removed and cells were washed once with ice-cold PBS. Cell lysis was initiated by addition of 100 μL RIPA buffer (Cell Signaling Technology) per well supplemented with phosphatase and protease inhibitor cocktails (Sigma-Aldrich) followed by incubation on ice for 30 minutes. Cells were then harvested by scraping and transferred into precooled microcentrifuge tubes. Whole-cell lysates were cleared by centrifugation at 14,000 × g/4°C for 10 minutes and transferred into fresh, precooled microcentrifuge tubes. Protein concentrations were determined using the Pierce BCA Protein Assay (Thermo Fisher Scientific) as per the manufacturer's instructions. Equal amounts of protein were separated by SDS-PAGE and transferred to Immobilon-FL PVDF Membranes (Merck Millipore) using Pierce 1-step transfer buffer (Thermo Fisher Scientific) and the Pierce Power Blotter (Thermo Fisher Scientific). Membranes were washed for 5 minutes with ddH2O, blocked with Odyssey Blocking Buffer TBS (Li-COR Biosciences) for 1 hour at room temperature and then incubated with primary antibodies diluted in Odyssey Blocking Buffer TBS overnight at 4°C with gentle agitation. The next day, membranes were washed three times with TBS-T (50 mmol/L Tris, 150 mmol/L NaCl, 0.05% Tween 20, pH 7.4) for 5 minutes before being incubated with secondary antibodies diluted in Odyssey Blocking Buffer TBS for 1 hour at room temperature with gentle agitation. Finally, the membranes were washed three times for 5 minutes with TBS-T and then scanned using an Odyssey CLx Imaging System (LI-COR Biosciences).
For capillary western blot analysis, lysates of the cell pellets were prepared as described above and protein concentration was estimated. Simple Wes assay was performed and analyzed according to the manufacturer's instructions (Protein Simple). For pAKTS473 detection, 0.55 μg of protein was loaded per capillary with 1:20 dilution of anti-pAKTS473 antibody while tAKT was detected with anti-tAKT antibody, dilution 1:100 and 0.05 μg of total protein per capillary.
For PathScan Akt Signaling Array (Cell Signaling Technology), 1× array wash buffer, 1× detection antibody cocktail, 1× horseradish peroxidase (HRP)-linked streptavidin, and the multi-well gasket were prepared as per the manufacturer's instructions and glass slides and blocking buffer were calibrated to room temperature. Subsequently, 100 μL array blocking buffer was added to each well for 15 minutes at room temperature covered with sealing tape and placed on an orbital shaker (as well as all following incubation steps). After removal of the blocking buffer, 50 μL diluted lysate (0.5 mg/mL protein concentration) were added to each well and incubated for 2 hours at room temperature. Wells were washed three times with 100 μL 1× array wash buffer for 5 min at room temperature, followed by incubation for 1 hour at room temperature in 75 μL 1× detection antibody cocktail. Wash steps were repeated four times before adding 75 μL 1× HRP-linked streptavidin to each well for a 30 minute incubation at room temperature. The multi-well gasket was removed from the slides another four wash steps later and the slides were washed with 10 mL 1× array wash buffer. Slides were then covered with LumiGlo/Peroxide reagent (as per the manufacturer's instructions) and images with different exposure times were captured using a digital imaging system.
Half-time determination (Western blot analysis)
For protein isolation, AN3-CA cells were seeded into 10-cm dishes (Sarstedt). Treatment with the inhibitors (borussertib, MK-2206, and miransertib) was initiated at a confluency of 60%–70% with the indicated concentrations for 24 hours. Then the medium was removed and cells were washed twice with PBS. Medium without borussertib was added to start the wash out experiments, which were terminated at the indicated time points. Prior to the cell harvest, cells were either not stimulated at all or stimulated with EGF (100 ng/mL, PeproTech), TGF-α (100 ng/mL, R&D Systems), or insulin (100 nmol/L, Sigma-Aldrich) for 15 minutes. Then the medium was removed and cells were washed twice with ice-cold PBS. Cell lysis was initiated by addition of 500 μL RIPA buffer per dish supplemented with phosphatase (Sigma-Aldrich) and protease inhibitor cocktails (Roche) followed by harvesting the cells by scraping and transferred into precooled microcentrifuge tubes. Cell lysates were incubated on ice for 30 minutes. After that, cells were lysed by sonication, cleared by centrifugation at 14,000 × g at 4°C for 10 minutes and transferred into fresh, precooled microcentrifuge tubes. Protein concentrations of the lysates were determined by the Bradford protein assay system (Bio-Rad). Equal amounts of protein (36 μg protein each lane) were separated by SDS-PAGE and transferred to polyvinylidene difluoride (PVDF) membranes (Roth). Immunoblots were blocked with 5% BSA in 1× TBS and Tween-20 (0.1% v/v) for 1 hour at room temperature. The membrane was incubated overnight at 4°C with primary antibodies. Afterwards, the membrane was incubated with the corresponding secondary antibody conjugated with HRP (Dianova). Bands were visualized with enhanced chemiluminescence Western blot detection system (Thermo Fisher Scientific).
Anti-pAkt(Ser473; Cell Signaling Technology, catalog nos. 3787 and 4060), anti-pAKT (Thr308;Cell Signaling Technology, catalog no. 2965), anti-tAKT (Cell Signaling Technology, catalog nos. 2920 and 4691), anti-pS6 ribosomal protein (Ser235/236;Cell Signaling Technology, catalog nos. 2211 and 4858), anti-pPRAS40(Thr246;Cell Signaling Technology, catalog no. 3997), anti-pERK1/2 (Thr202/Tyr204;Cell Signaling Technology, catalog no. 4370), anti-p4E-BP1(Ser65;Cell Signaling Technology, catalog no. 13443), anti-pGSK-3β (Ser9;Cell Signaling Technology, catalog no. 9323), anti-GSK-3β (Cell Signaling Technology, catalog no. 9315), anti-PARP (Cell Signaling Technology, catalog no. 9542), anti-HSP90, (Cell Signaling Technology, catalog no. 4874), anti-β-Actin (Cell Signaling Technology, catalog no. 4970/Sigma-Aldrich, catalog no. A5441), anti-GAPDH (Cell Signaling Technology, catalog no. 2118), anti-mouse IgG (H+L; DyLight 680 Conjugate; Cell Signaling Technology, catalog no. 5470), anti-rabbit IgG (H+L; DyLight 800 4X PEG Conjugate; Cell Signaling Technology, catalog no. 5151).
In vitro pharmacokinetic studies
For determining kinetic solubility, the compound was diluted from a 10 mmol/L stock in DMSO to a final concentration of 500 μmol/L in 50 mmol/L HEPES, pH 7.4. Following an incubation of 90 minutes at room temperature on a shaker, the aqueous dilution was filtered through a 0.2-μm PVDF filter, and the optical density between 250 and 500 nm was measured at intervals of 10 nm. The kinetic solubility was calculated from the AUC between 250 and 500 nm and normalized to absorption of a dilution of the compound in acetonitrile.
Metabolic stability under oxidative conditions was measured in human and murine liver microsomes by LC/MS-based analysis of depletion of compound at a concentration of 3 μmol/L over time up to 50 minutes at 37°C. On the basis of compound half-life, t1/2, the in vitro intrinsic clearance, Clint, was calculated.
Plasma stability was measured by LC/MS-based determination of percent remaining of selected compound at a concentration of 5 μmol/L after incubation in 100% plasma obtained from different species for 1 hour at 37°C.
Assessment of plasma protein binding was measured by equilibrium dialysis by incubating the compound of interest at a concentration of 5 μmol/L for 6 hours at 37°C in 50% plasma in buffer (v/v) followed by LC/MS-based determination of final compound concentrations. The resulting fraction unbound at 50% plasma (fu50%) was extrapolated to the fraction unbound at 100% plasma (fu100%) using the following equation: fu100% = fu50%/(2 − fu50%).
In vivo pharmacokinetic studies
For in vivo pharmacokinetic analysis, RjOrl:SWISS mice (Janvier Labs), age 8–10 weeks, were treated with borussertib by oral gavage (20 mg/kg), intraperitoneal (20 mg/kg), or intravenous (2 mg/kg) administration. The compound was formulated in PBS/PEG200 (60:40) for oral and intraperitoneal administration, whereas it was dissolved in DMSO for intravenous injection. Blood was collected 5, 15, 45, and 135 minutes after compound administration, immediately centrifuged at 15,000 × g for 10 minutes at 4°C and plasma samples were stored at −80°C for subsequent LC-MS/MS analysis. Three mice were analyzed per experimental condition (for each time point and route of administration). Mice were fed ad libitum with Allein-Futter für Ratten-/Mäusehaltung (Sniff Special Diets GmbH). They had free access to water and were kept in a 12 hour day/night rhythm. All experiments were approved by the local authorities.
Samples and blanks were prepared by adding 2.5 μL blank DMSO and 80 μL of ice-cold acetonitrile containing the internal standard (griseofulvin, 1 μmol/L) to 20 μL plasma followed by centrifugation at 13,000 rpm (4°C) for 10 minutes. A total of 65 μL of the supernatant were diluted with 65 μL of LC/MS grade water. Samples were filtered (MSRLN0450, Millipore) and subjected to LC/MS measurement. Analyte stock solution (10 mmol/L in DMSO) was diluted in DMSO to yield DMSO stock solutions with the following concentrations: 10, 5, 2.5, 1, 0.5, 0.25, 0.1, 0.05, 0.025, 0.01, 0.005, and 0.0025 μmol/L. A total of 2.5 μL of the corresponding DMSO stock solution were added to 20 μL of blank plasma followed by the addition of 80 μL of ice-cold acetonitrile containing the internal standard. The samples were centrifuged for 10 minutes at 4°C and 13,000 rpm. A total of 65 μL of the supernatant were diluted with 65 μL of LC/MS grade water and subjected to LC/MS analysis. A set of three different quality controls (n = 3) was prepared by adding 2.5 μL DMSO stock solution (5, 0.5, and 0.05 μmol/L) to 20 μL of plasma. Samples were subsequently handled as described above. All samples were analyzed using a Shimadzu LC20ADXR Solvent Delivery Unit, a Shimadzu SIL30ACMP autosampler, and an ABSciex Qtrap5500 LC-MS/MS system. Therefore, 2 μL of sample were injected and separated using an Agilent Poroshell C18, 2.7 μm column (2.1 mm× 50 mm) at 60°C starting at 5% of solvent B for 0.3 minutes followed by a gradient up to 100% of solvent B over 0.6 minutes (flow rate 1 mL/min) with 0.1% formic acid in water as solvent A and 0.1% formic acid in acetonitrile as solvent B. Data evaluation was performed using Analyst 1.6.2 Software (Sciex).
Animal models and treatments
Tissue samples to establish the PDX models were collected from patients following surgical intervention for colon cancer or pancreatic adenocarcinoma at the Ruhr-University Comprehensive Cancer Center (Bochum, Germany). From all patients informed and written consent were obtained. The studies were approved by the Ethics Committees of the Ruhr-University Bochum (registry nos. 3534-09, 3841-10, and 16-5792, Bochum, Germany). Animal experiments and care were in accordance with the guidelines of institutional authorities and approved by local authorities (nos.: 8.87-50.10.32.09.018, 84-02.04.2012.A328, 84-02.04.2012.A360, 84-02.04.2015.A135, and 81-02.04.2017.A423). Nondiagnostic tissue samples were selected by the pathologist within 2–6 hours postsurgery. Selected tumor pieces (1–2 mm) were soaked in undiluted Matrigel (Becton Dickinson) for 15–30 minutes and subsequently implanted subcutaneously onto 5- to 10-week-old female mice (NMRI-Foxn1nu/Foxn1nu, Janvier Labs) at two sites (scapular region, one mouse per tumor) using as many as 4 pieces per site. To establish treatment cohorts, early passage (≤F5 generation) PDX tumor pieces were implanted as described above into nude mice and were allowed to grow to a size off approximately 100–200 mm3, at which time mice were randomized in the treatment and control groups with 3–4 mice in each group. Tumor volumes were estimated from two-dimensional tumor measurements by bi-weekly caliper measurements using the following formula: tumor volume (mm3) = [length (mm) × width (mm)2]/2. Response was defined in analogy to RECIST 1.1 criteria with at least 30% reduction in mean tumor volume compared with the mean tumor volume at start of treatment being a partial response (PR) and an undetectable tumor being a complete response. Disease progression was defined as more than 20% increase in mean tumor volume to the tumor volume at the beginning of treatment. All other measurements were defined as stable disease. Mice were treated with borussertib by daily intraperitoneal injection dosed at 20 mg/kg and trametinib (Hycultec) by oral gavage at 0.5 mg/kg per day with a weekly treatment cycle comprising of 5 consecutive days of treatment followed by 2 days treatment pause.
Borussertib binds covalently between the kinase and PH domain of AKT
Recently, we described a novel class of AKT inhibitors that bind into the allosteric pocket of AKT and harbor a Michael acceptor to covalently bind to noncatalytic cysteines. These features combine the advantage of outstanding selectivity of PH domain–dependent allosteric inhibition with the therapeutic benefit of irreversible modification, leading to increased drug target residence time and gain of potency. In this class of inhibitors, we identified a potent lead compound, borussertib, exhibiting an exquisite kinase selectivity profile (19). To investigate the binding mode of this novel AKT inhibitor at the atomic level, we solved the cocrystal structure of AKT1 in complex with borussertib to a resolution of 2.9 Å.
The crystal structure discloses the inactive, autoinhibited conformation with the PH domain folded onto the kinase domain (PH-in conformation) between the N- and the C-lobe, thereby displacing the regulatory helix αC and simultaneously shaping an allosteric binding pocket at the interface between these two domains (Fig. 1A; Supplementary Fig. S1; Supplementary Table S1). Borussertib binds to this allosteric pocket and forms a key aromatic π–π stacking interaction between the 1,6-naphthyridinone scaffold and the indole side chain of Trp80 in the PH domain. Additional hydrophobic contacts can be observed between the phenyl ring in the 3-position and Leu210, Leu264, and Ile290. Water-mediated hydrogen bonds between Glu17, Arg273, Tyr326, and the benzo[d]imidazolone moiety of borussertib foster the high-affinity reversible binding of the ligand to the kinase (Fig. 1B). Furthermore, the acrylamide moiety is preoriented by a hydrogen bond formed between the amide oxygen of the warhead and the backbone NH of Glu85, facilitating covalent bond formation between the electrophilic β-carbon and the thiol side chain of Cys296.
Borussertib potently inhibits proliferation of PI3K/PTEN-mutated cell lines
To investigate the in vitro antiproliferative activity of borussertib, we employed breast, bladder, pancreas, and endometrium cancer cell lines harboring genetic alterations in the PI3K/AKT and RAS/MAPK pathways, that is, AN3-CA (endometrium), BT-474 (breast), Dan-G (pancreas), HPAF-II (pancreas), KU-19-19 (bladder), MCF-7 (breast), T-47D (breast), and ZR-75-1 (breast). Genetic alterations include in-frame deletions in PIK3R1, frame shifts and point mutations in PTEN, and (activating) point mutations in PIK3CA, NRAS, and KRAS (Supplementary Table S2). Each cell line was treated for 96 hours with borussertib or reference inhibitors in a concentration range from 30 μmol/L to 0.1 nmol/L. In these cell lines, we observed outstanding sensitivity to borussertib with EC50 values in the submicromolar range, indicating a 1.5- to 43-fold greater potency than miransertib, MK-2206, ipatasertib, and capivasertib (Fig. 2). Only in the bladder cancer cell line KU-19-19, which harbors additional activating mutations in AKT1 (E17K/E49K) and NRAS (Q61R), micromolar antiproliferative activities could be observed (EC50 = 3.1–5.0 μmol/L) for the tested compounds being in good correlation with previously published data for the allosteric AKT inhibitor BAY 1125976 (18). Furthermore, for pancreatic cancer cell lines Dan-G and HPAF-II, generally lower sensitivities to AKT inhibition were observed with only minor differences between the individual inhibitors (Supplementary Table S2). The lack of mutations within PI3K and/or PTEN in combination with codon 12 mutations in KRAS substantiates the relatively high EC50 values observed for these two cell lines.
Of note, the breast cancer cell line ZR-75-1 exhibited a pronounced sensitivity to borussertib, with an EC50 of 5 ± 1 nmol/L and thus an approximately 7- to 12-fold higher potency compared with the reversible allosteric inhibitors miransertib (EC50 = 35 ± 18 nmol/L) and MK-2206 (EC50 = 63 ± 21 nmol/L); ATP-competitive inhibitors capivasertib (EC50 = 191 ± 68 nmol/L) and ipatasertib (EC50 = 219 ± 83 nmol/L) showed a 38- to 43-fold lower activity, respectively. The significant differences in antiproliferative activity observed for some of the tested cell lines can only to some extent be explained by the biochemical inhibitory potencies toward AKT1; capivasertib (IC50 = 0.9 ± 0.1 nmol/L) exhibits a higher potency with respect to inhibition of AKT1 in vitro compared with ipatasertib (IC50 = 3.5 ± 0.6 nmol/L) and MK-2206 (IC50 = 10.0 ± 2.1 nmol/L; ref. 24). However, MK-2206 showed a 5- to 9-fold higher activity in MCF-7 cells, whereas capivasertib and ipatasertib inhibited growth of AN3-CA, BT-474, and T-47D cells with similar activities. Differences in cellular pharmacokinetic properties as well as AKT1 expression and activity levels might have contributed to these observations. In addition, kinase-independent functions related to specific conformations stabilized by either ATP-competitive or allosteric AKT inhibitors could affect in vitro as well as in vivo potency (32, 33). Moreover, the molecular impact of AKT isoforms 1–3 on cell survival and proliferation is not fully understood, and a potential influence of the compounds’ selectivity profiles toward AKT1, AKT2, and AKT3 on their antiproliferative efficacy cannot be excluded (6). In summary, the experimental inhibitor borussertib exhibited superior antiproliferative properties in our experimental setup compared with the clinical candidates of ATP-competitive and allosteric reference compounds, indicating a potentially beneficial impact of our approach to irreversibly target AKT.
Borussertib downregulates AKT-mediated signaling
To gain further insights into the molecular mode of action and how it affects AKT signaling, we performed PathScan AKT signaling assays in combination with Western blot studies. Furthermore, washout experiments were performed to reconstruct the average half-life of irreversibly inhibited AKT and thus anticipate the mean projected duration of compound treatment. For PathScan analysis, MCF-7 and Dan-G cells were treated with dimethyl sulfoxide (vehicle) or 1 μmol/L borussertib for 24 hours prior to lysis and array-based readout (Supplementary Fig. S2). For MCF-7 cells, the results indicate low basal levels of activated AKT (pAKTS473), whereas PRAS40 as well as GSK-3α/β exhibited pronounced phosphorylation in vehicle-treated cells. Upon treatment with 1 μmol/L borussertib, pAKT levels were reduced and phosphorylation of PRAS40 and GSK-3α/β significantly decreased, hinting at potent inhibition of AKT signaling. Moreover, phospho-S6 and phospho-p70 S6 kinase signals were diminished upon inhibitor treatment. Notably, borussertib exhibited no off-target inhibition toward activating kinase PDK1 and RAS/MAPK signaling. Comparable results were observed for Dan-G cells including the downregulation of pAKTS473, pS6S235/236, pPRAS40T246, and pGSK-3αS21.
Western blot analyses for ZR-75-1, AN3-CA, Dan-G, T-47D, and MCF-7 cell lines resolved the dose-dependent downregulation of pAKTT308 and pAKTS473 as well as downstream targets pPRAS40T246, pS6S235/236, and p4E-BP1S65 (Fig. 3A; Supplementary Fig. S3). For all cell lines, inhibition of AKT phosphorylation was observed upon drug treatment demonstrating the highest sensitivity for ZR-75-1 and AN3-CA cells, correlating well with the pronounced inactivation of AKT-mediated downstream signaling, as can be seen by the substantial dephosphorylation of PRAS40, S6, and 4E-BP1 (Fig. 3A). In addition, to investigate the underlying mechanism of borussertib's antiproliferative activity, cell lysates were probed for cleavage of PARP revealing a distinct induction of apoptosis in AN3-CA and ZR-75-1 cells after compound treatment at nanomolar concentrations (Fig. 3A). In contrast, no increase in cleaved PARP (cPARP) level could be observed for KRAS-mutant Dan-G cells at concentrations as high as 10 μmol/L, indicating a lower dependence on AKT-mediated signaling.
To determine the cellular half-life of covalently inhibited AKT, AN3-CA cells were treated with 0, 100, and 200 nmol/L borussertib, respectively, for 24 hours, followed by medium renewal and serum starvation for 0–48 hours. Prior to cell lysis, cells were stimulated with EGF for 15 minutes. pAKTS473 levels remained significantly downregulated up to 24 hours after medium renewal and drug withdrawal (Fig. 3B). Similar results were obtained for unstimulated, EGF-related TGF-α–treated, and insulin-treated cells (Supplementary Fig. S4A-S4D). In contrast to borussertib, higher concentrations were required for MK-2206 and miransertib to completely inhibit AKT-mediated signaling, as shown for pGSK-3βS9 and pS6S235/236 (Supplementary Fig. S4E), despite efficient downregulation of pAKTS473 at 200 nmol/L. With the slow recovery of pAKTS473 levels upon irreversible inhibition with borussertib, we propose that with respect to in vivo studies, single-compound administrations might be efficacious for a relatively long period of time, independent of the in vivo pharmacokinetics, provided that a sufficient amount of inhibitor reaches its target before being cleared. However, also reversible inhibition using MK-2206 and miransertib resulted in prolonged downregulation of pAKTS473 after compound washout (Supplementary Fig. S5).
AKTi borussertib and MEKi trametinib act synergistically in vitro
In addition to the potent antiproliferative efficacy of borussertib toward PI3K/PTEN-mutated cell lines, we were interested in the identification of potential additive or synergistic effects of AKT inhibition in combination with targeted MEK inhibition or chemotherapy. Therefore, Dan-G cells were treated with borussertib and MEKi trametinib or gemcitabine, respectively, at concentrations close to the respective EC50 (Fig. 4A and B). Employing the Chou-Talalay method (31), combination indices (CI) were calculated for the tested drug combinations to determine potential additive (CI = 1), synergistic (CI < 1), or antagonistic (CI > 1) effects. For the combination of borussertib and trametinib, strong synergy was observed over the entire range of concentrations tested, whereas no significant beneficial effect could be observed for combination treatment with borussertib and gemcitabine even at high doses (Fig. 4C). Inhibition of MAPK pathway activity via trametinib has failed so far in clinical trials in KRAS-driven tumors, which may be due to concomitant activity of PI3K/AKT pathway activity or induction of resistance via cross-talk among other reasons. Recent data showed PI3K activation upon complete ablation of KRAS in PDAC cells (30), further supporting strategies to block AKT activation in combination with RAS/MAPK-acting drugs. With regard to gemcitabine, efficacy of this chemotherapeutic agent is regulated on multiple levels including expression of transporter genes (e.g., hENT1), intracellular drug metabolism, and cell-cycle state among others. Because the combination of borussertib with gemcitabine showed no clear synergistic signal in PDAC cells, we did not follow this path in more detail but focused on rational drug combinations with favorable combinatory index.
To further investigate the potential of AKTi/MEKi combination therapy including our covalent-allosteric inhibitor borussertib, we utilized early passage pancreatic cancer cells (Bo103) harboring a KRAS mutation in codon 12 as the model of interest for viability studies in combination with Western blot analyses. Neither borussertib nor trametinib monotherapy resulted in complete inhibition of cell viability at concentrations up to 30 μmol/L; remaining cell viabilities at the highest concentrations of borussertib and trametinib were determined to be 51.5% ± 9.5% and 29.3% ± 5.9%, respectively (Fig. 5A). For the combination treatment, both compounds were added to the cells in a 1:1 stoichiometry yielding a highest total concentration of 30 μmol/L (15 μmol/L borussertib/15 μmol/L trametinib); a remaining cell viability of 6.9% ± 2.9% (mean ± SD) was determined from three independent experiments, indicating a substantial benefit of combination treatment as compared with single-agent therapy. The resulting EC50 of 82.7 ± 26.0 nmol/L additionally highlights the supreme efficacy of the combination therapy tested herein. EC50 values for either monotherapy were not calculated due to incomplete inhibition of cell viability (Fig. 5A).
For correlation of antiproliferative activity with downregulation of pS6S235/236 and p4E-BP1S65 induced by either AKT or MEK inhibition, Western blots were prepared for pancreatic cancer Bo103 cells treated with single agent or drug combination (Fig. 5B). Pronounced downregulation of pAKTS473 or pERK1/2T202/Y204 was detected upon treatment with borussertib and trametinib, respectively, correlating with decreasing amounts of detectable pS6S235/S236. However, p4E-BP1S65 was not affected by treatment with either of the two compounds. Similar effects were observed for cells treated with drug combination, yet showing a more distinct decrease of pS6S235/S236. These observations might explain the superior antiproliferative efficacy of combination compared with single-agent treatment and thus indicate the enormous potential of covalently targeting AKT.
Borussertib exerts antitumor activity in KRAS-mutant PDXs in combination with trametinib
To investigate the potential of borussertib as a drug candidate in preclinical studies, we performed in vitro and in vivo pharmacokinetic analyses for borussertib (Supplementary Fig. S6). Besides an unfavorable low solubility in aqueous media (13 μmol/L), in vitro analyses in both human and murine samples revealed promising features with generally low intrinsic clearance (Clint,human = 7 μL/min/μg, Clint,murine = 31 μL/min/μg), high plasma stability (human: 99% remaining, murine: 100% remaining) and high plasma protein binding (human: 100%, murine: 99%; Supplementary Fig. S6A). Subsequent pharmacokinetic studies were carried out in mice (2 mg/kg i.v.; 20 mg/kg oral gavage; and 20 mg/kg, i.p.; Supplementary Fig. S6B). Despite a rather low oral bioavailability (<5%), reaching only a maximum plasma concentration of 78 ng/mL (0.13 μmol/L), we found a significantly higher bioavailability upon intraperitoneal administration (39.6%), with maximum plasma levels of 683 ng/mL (1.14 μmol/L), indicating sufficient absorption of the compound to potentially exert antitumor activity in xenografts.
Given the promising pharmacokinetic and pharmacodynamic properties, we next examined the antitumor activity of borussertib in mouse xenograft studies, using implanted xenograft models derived from KRAS-mutant primary pancreas and colon cancers. These tumor entities are characterized by RAS/MAPK activity but considerable resistance to single-agent MAPK pathway inhibition clinically. Encouraged by our in vitro analyses indicating synergistic effects, we therefore focused on combined activity of MEK and AKT inhibition with trametinib and borussertib, respectively. For PDX studies, borussertib was administered at 20 mg/kg per intraperitoneal injection once daily, either as monotherapy or in combination with the targeted MEK inhibitor trametinib at 0.5 mg/kg administered perorally for 5 days per week.
First, PDAC PDX models were used to evaluate borussertib for its antitumor activity (Fig. 6A; Supplementary Fig. S7A and S7B). Borussertib monotherapy resulted in insignificant tumor growth delays in Bo103 PDX (Fig. 6A), Bo73 (Supplementary Fig. S7A), and Bo85 (Supplementary Fig. S7B) compared with untreated control mice. In contrast, trametinib monotherapy induced a decelerated tumor growth in all tested PDX models. However, only the combination of borussertib with trametinib resulted in a durable partial response in Bo103, whereas progressive diseases were observed for Bo73 and Bo85.
In addition, we employed KRAS-mutant colorectal carcinoma PDX models to further evaluate the antitumor activity of borussertib (Fig. 6B–D; Supplementary Fig. S7C and S7D). In four of five established model systems, borussertib monotherapy did not affect tumor growth as compared with untreated control mice (Fig. 6B and D; Supplementary Fig. S7C and S7D). Nevertheless, the PDX cohort engrafted with BoC105 exhibited significantly delayed tumor growth upon borussertib treatment (Fig. 6C). This effect was additionally augmented in mice treated with borussertib in combination with trametinib, resulting in a durable partial response. In total, the combination of the AKT inhibitor borussertib with the MEK inhibitor trametinib yielded three stable diseases (Fig. 6B and D; Supplementary Fig. S7C) and two PRs (Fig. 6C; Supplementary Fig. S7D), highlighting the potential benefit of this combination compared with MEK inhibitor monotherapy for treating colorectal cancer.
Taken together, these results underscore the general applicability of covalent-allosteric AKT inhibitors in vivo. Although the KRAS-mutant PDX models employed in this study did not show any response to AKTi monotherapy, alternative in vivo model systems harboring, for example, genetic lesions in PI3K or PTEN, could be more suitable for evaluation of borussertib monotherapy.
Numerous signaling cascades rely on AKT as a central integrator of diverse stimuli relevant for physiologic and pathobiologic processes. To date, few small-molecule AKT modulators have entered preclinical and clinical trials, largely because of selectivity issues caused by structurally similar kinases, lack of efficacy, and mechanism-based adverse effects. Moreover, aberrant AKT signaling resulting from activating mutations in PI3K or functional loss of PTEN might not give rise to oncogene addiction per se (14). However, (co-)targeting AKT in a clinical setting may result in beneficial therapy outcomes, as demonstrated for ipatasertib, capivasertib, MK-2206, and miransertib, respectively. Of note, low prevalent hyperactive AKT1E17K has recently been described to act as a classical driver oncogene in patients suffering from gynecologic and estrogen receptor–positive breast cancers, thus eliciting pronounced therapeutic responses upon administration of ATP-competitive AKT inhibitor capivasertib (34). The efficacy of borussertib for such indications remains to be determined.
Borussertib proved to be a highly selective and irreversible allosteric inhibitor of AKT with potent in vitro antiproliferative activity and the ability to synergize with other targeted therapies such as MEKi in KRAS-mutant colon and pancreatic cancer PDX models thereby overcoming potential limitations regarding therapeutic efficacy observed for MEKi monotherapy in the types of cancer mentioned above. The X-ray crystallographic complex structure presented here supports the anticipated binding mode and will foster the rational derivatization and optimization of our lead molecule borussertib concerning binding affinity and inhibitory potency. We provide evidence for the potent inhibition of cancer cell proliferation, especially for cell lines featuring genetic alterations in the PI3K/AKT signaling cascade, resulting from the targeted downregulation of pAKT and downstream effectors, including pS6, p4E-BP1, and pPRAS40, as deduced from immunoblot analyses. Future efforts will be directed toward the profiling of cancer cell lines from additional primary sites and the evaluation of potential drug combination strategies in combination with expanded comprehensive pharmacodynamic analyses. In addition, we provide proof-of principle data for the in vivo efficacy of what is to our knowledge the first-in-class covalent-allosteric AKT inhibitor, as shown for KRAS-mutant pancreatic ductal adenocarcinoma and colorectal carcinoma PDX models. Additional efforts will be directed toward the optimization of aqueous solubility to generate an oral bioavailable derivative of borussertib with improved biochemical potency, cellular antiproliferative activity, and in vivo efficacy. Furthermore, detailed in vivo pharmacodynamic analyses, optimal dosage identification and toxicity profiling are mandatory for the subsequent development of covalent-allosteric AKT inhibitors as drug-like candidates. Eventually, besides combination studies, it will be of interest to employ borussertib or its optimized derivatives in PDX models harboring genetic alterations in the PI3K/AKT signaling axis to enable characterization of its potential as a monotherapeutic agent in immediately relevant disease settings, for example, breast or endometrium cancer.
Disclosure of Potential Conflicts of Interest
J. Weisner has ownership interest (including stock, patents, etc.) in chemical patent (borussertib). R. Gontla has ownership interest (including stock, patents, etc.) in chemical patent (borussertib). M. Pohl has received speakers bureau honoraria from Roche Pharma AG, Amgen AG, Sanofi Aventis, Servier, Dres. Schlegel+Schmidt, and AbbVie and is a consultant/advisory board member for Roche Pharma AG, Amgen AG, Lilly, MSD Sharp & Dome, MCI Deutschland GmbH, Merck Serono, Sanofi Aventis, BMS, Baxalta Shire, Chugai, and Celgene. D. Rauh has received speakers bureau honoraria from Astra-Zeneca, Merck-Serono, Takeda, Pfizer, Novartis, Boehringer, Sanofi, and BMS. No potential conflicts of interest were disclosed by the other authors.
Conception and design: J. Weisner, I. Landel, C. Reintjes, N. Uhlenbrock, C. Schultz-Fademrecht, S.A. Hahn, J.T. Siveke, D. Rauh
Development of methodology: C. Reintjes, R. Scheinpflug, A. Maghnouj, J.T. Siveke, D. Rauh
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. Weisner, I. Landel, C. Reintjes, N. Uhlenbrock, M. Trajkovic-Arsic, N. Dienstbier, J. Hardick, S. Ladigan, M. Lindemann, S. Smith, L. Quambusch, R. Scheinpflug, L. Depta, R. Gontla, G. Günther, M. Pohl, C. Teschendorf, H. Wolters, R. Viebahn, A. Tannapfel, W. Uhl, J.G. Hengstler, S.A. Hahn, J.T. Siveke
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. Weisner, I. Landel, C. Reintjes, N. Dienstbier, J. Hardick, A. Unger, H. Müller, M. Baumann, C. Schultz-Fademrecht, M.P. Müller, A. Tannapfel, J.G. Hengstler, S.A. Hahn, J.T. Siveke, D. Rauh
Writing, review, and/or revision of the manuscript: J. Weisner, I. Landel, C. Reintjes, N. Uhlenbrock, M. Trajkovic-Arsic, N. Dienstbier, J. Hardick, M. Lindemann, S. Smith, L. Quambusch, R. Gontla, G. Günther, M.P. Müller, M. Pohl, H. Wolters, R. Viebahn, W. Uhl, J.G. Hengstler, S.A. Hahn, J.T. Siveke, D. Rauh
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Weisner, I. Landel, N. Uhlenbrock, A. Maghnouj, M.P. Müller, C. Teschendorf
Study supervision: J. Weisner, I. Landel, C. Reintjes, J.T. Siveke, D. Rauh
We thank Axel Choidas and Bert Klebl for helpful discussions and we are thankful to Prof. Dr. Philippe I. H. Bastiaens for granting access to the Odyssey CLx Imaging System (Li-Cor). This work was supported by the MERCATOR Foundation (grant no. Pr-2016-0014). D. Rauh is thankful for support from the German Federal Ministry for Education and Research (NGFNPlus and e:Med; grant No. BMBF 01GS08104, 01ZX1303C), the Deutsche Forschungsgemeinschaft and the German federal state North Rhine Westphalia, and the European Union (European Regional Development Fund: Investing In Your Future; grant no. EFRE-800400). J.T. Siveke is supported by the European Union's Seventh Framework Programme for Research, Technological Development and Demonstration (FP7/CAM-PaC) under grant agreement no. 602783, the German Cancer Aid (grant no. 70112505), the Deutsche Forschungsgemeinschaft (DFG; KFO337/SI 1549/3-1), and the Erich and Gertrud Roggenbuck Foundation and the German Cancer Consortium (DKTK).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.