Structure of the KRas4B-PDEδ complex
At the earliest stages of this work, the 3D structure of the KRas4B-PDEδ complex was unknown. Homology modeling of this complex was carried out using the Molecular Operating Environment package  employing as a template the previously reported RHEB-PDE6δ crystallographic structure (PDB ID 3T5G). A large set of structures was modeled resulting from different side-chain rotamers of newly incorporated residues. The structure with the best packing index was subjected to a global energy-minimization analysis with the CHARMM27 force field to yield the final model. Later on, two different crystallographic structures of the KRas4B-PDE6δ complex were reported and deposited at PDB ID: 5TAR and 5 TB5. They differ in the KRas4B C terminus close to the farnesylation site: the former shows an ordered structure, while the latter has a partially disordered segment. We used both, reporded 3D models and our model, the 5TAR and 5 TB5 structures to represent KRas4B-PDEδ intermolecular contacts and to guide the search for small organic compounds capable to simultaneously form interactions to both proteins, thus acting on the complex as molecular staples. Since in 3T5G structure RHEB is in contact with PDE6δ, we directly used solvent-exposed cavities composed of atoms from both proteins as targets for potential-ligand search. In the case of 5TAR, we used as targets the pockets close to the KRas4B-PDE6δ interface as found in the crystallographic lattice. The KRas4BG12C mutant corresponding to the predominant mutation present in MIA PaCa cell line was modeled with PyMOL v0.99 (https://pymol.org/2/).
ENAMINE’s Discovery Diversity Set database (DDS) containing 50,240 low molecular weight compounds was selected for virtual screening. The 2D structures were translated into 3D structures using MOE-Import Search. Hydrogens and partial charges were assigned according to MMFF94 force field. Strong acids and bases are deprotonated and protonated, respectively. In order to simulate the molecular flexibility shown in real systems, structural conformers were constructed for each compound in DDS with MOE-Conformer Search and using a conformational energy cut-off of 3 kcal/mol with respect to the minimum energy conformer of each compound calculated according to the MMFF94 force field. The new database was then used for virtual screening. Potential binding sites, i.e. concave pockets at the protein-protein interface region in the KRas4B-PDE6δ model and crystalographic structures were identified with MOE-SiteFinder and CASTp server . Previously, all crystallographic water and other organic molecules were removed. Hydrogen atoms and partial charges were added to the KRas4B-PDE6δ complex using the CHARMM27 force field. Virtual screening was carried out using MOE_Dock function and setting the Alpha-Site-Triangle and the London dG as the methods to bias the orientation search on potential binding sites and docking scoring function, respectively. At least 10,000 different orientations or poses on potential binding sites were proved and evaluated for each conformer, and the ten best coupling scores for each confomer were saved for further analysis. Finally, the KRas4B-PDEδ-ligand complexes with the best binding energies and frequencies were selected and evaluated with respect to the specific contacts of the compounds and the binding strengths, with preference given to the more polar compounds.
Molecular dynamics (MD) simulations and binding free energy calculations
MD simulations of protein-protein and protein-ligand complexes were performed using AMBER 16 package  and the ff14SB forcefield . Ligand charges for ligands and for no parameterized residues in proteins were determined using the AM1-BCC level and the general Amber force field (GAFF) . For protein-protein and protein-ligand complexes a 15 Å and 12 Å, respectively, a rectangular-shaped box of TIP3P water model  was applied to solvate the complex and Cl− and Na+ ions for protein-protein and protein-ligand systems were placed to neutralize the positive or negative charges around the complex models at pH 7. Before MD simulations, each molecular system was minimized through 3000 steps of steepest descent minimization followed by 3000 steps of conjugate gradient minimization. Then, systems were heated from 0 to 310 K during 500 ps (ps) of MD with restrained positions under an NVT ensemble. Next, MD simulations for 500 ps, in an isothermal-isobaric ensemble (NPT), were carried out to adjust the solvent density, followed by 600 ps of constant pressure equilibration at 310 K, using the SHAKE algorithm  on hydrogen atoms, and Langevin dynamics for temperature control. Equilibration runs were tailed by 100 ns-long MD simulations without position restraints, under periodic boundary conditions using an NPT ensemble at 310 K. The particle mesh Ewald method was utilized to describe the electrostatic term , and a 10 Å cut-off was used for the van der Waals interactions. Temperature and pressure were preserved using the weak-coupling algorithm  with coupling constants τT and τP of 1.0 and 0.2 ps, respectively. The time step of the MD simulations was set to 2.0 femtoseconds, and the SHAKE algorithm  was used to constrain bond lengths at their equilibrium values. Coordinates were saved for analyses every 50 ps. AmberTools14 was used to examine the time-dependence of the root mean squared deviation (RMSD), and the radius of gyration (RG), as well as for clustering analysis to identify the most populated conformation during the equilibrated simulation time.
Calculation of binding free energies
Calculation of binding free energies was carried out using the MMGBSA approach [18,19,20] provided in the Amber16 suite . 500 snapshots were chosen at time intervals of 100 ps from the last 50 ns of MD simulations, using a salt concentration of 0.1 M and the Generalized Born (GB) implicit solvent model . The binding free energy of protein-protein and protein-ligand systems was determined as follows: ΔGbind = Gcomplex – Greceptor – Gligand. ΔGbind = ΔEMM + ΔGsolvation – TΔS. ΔEMM represents the total energy of the molecular mechanical force field that includes the electrostatic (ΔEele) and van der Waals (ΔEvdw) interaction energies. ΔGsolvation signifies the desolvation free energy price upon complex formation, estimated from GB implicit model and solvent-accessible surface area (SASA) calculation that yield ΔGele/sol and ΔGnpol/sol. Whilst, –TΔS is the solute entropy arising from structural changes that occur in the degrees of freedom of the free solutes and during formation of the protein-protein or protein-ligand complex.
Small organic compounds identified by virtual screening were purchased from ENAMINE (https://enamine.net/index.php?option=com_content&task=view&id=11) (Kyiv, Ukraine). The compounds were dissolved in 1.5% DMSO (SIGMA-ALDRIHC, catalog No. 276855-1 L). Deltarasin (hydrochloride) was purchased from Cayman Chemical (catalog No. 1440898–82-7).
Human pancreatic cancer cell lines MIA PaCa-2, PanC-1, BxPC-3 and hTERT-HPNE were obtained from the American Type Culture Collection (ATCC; Manassas, VA). Cell lines were grown as monolayers in the specific medium suggested by ATCC.
Cell viability assay
Cell lines were seeded at a density of 30,000 cells per well in a 96-well microtiter plate in growth medium and allowed to adhere for 24 h. Then, they were treated with 200 μM of each of the 38 compounds. Cell proliferation was assessed every 24 h during 3 days. Cell viability was determined by MTT (MTT Cell Proliferation Assay ATCC 30-1010K), by adding 10 μL of MTT per well, in dark conditions and incubated for 4 h. To solubilize the formazan crystals, 100 μL of acid isopropanol (50 mL of Triton X-100, 4 mL of HCl, 446 mL of isopropanol) was added, stirred continuously at room temperature and darkness for 3–4 h. The absorbance was measured in a spectrophotometer (Infinite F500 TECAN) at a wavelength of 570 nm. Each concentration was evaluated in triplicate, the solvent of the fractions and the untreated cells were taken as negative controls. The data are presented as the average percentage of proliferation and the standard deviation of the mean.
Cell lines were seeded at a density of 20,000 cells per well in a 96-well microtiter plate in growth medium and allowed to adhere for 24 h. Following the treatment with 200, 100, 50, 25, 12.5 and 6.25 μL of D14 and C22, respectively, cell viability was assessed for 5 days every 24 h. At the end of treatment, cell viability was determined by the CellTiter-Glo Luminescent Cell Viability Assay (Promega, catalog No. G7573). The dose-response curve was used to calculate the concentration of drug resulting in 50% inhibition of cell viability (IC50). The assays were repeated 5 times.
Approximately 5 X 105 cells were seeded in 6-well plates for 24 h. Then, cells were treated with an IC50 concentration of D14 and C22 compounds and vehicle for 24 h. Cells were harvested with 0.25% trypsin, washed with phosphate buffered saline (PBS), and collected together by centrifugation. Apoptosis was determined using the Apoptosis/Necrosis Detection kit (Abcam, catalog No. ab176749, Cambridge, England) according to the manufacturer’s instructions and analyzed by flow cytometry using a FACSCalibur instrument (BD Biosciences), followed by data analysis using FlowJo software (Tree Star Inc). All experiments were performed in triplicate. Proteome Profiler Apoptosis Array (R&D Systems: ARY009) was used to evaluate the activity of D14 and C22 compounds on MIA-PaCa-2 cancer cells to determine the signaling pathways associated with cell death via Kras4B inhibition, which were done following the manufacturer’s instructions.
Ras activation assay
The inactivation of Ras by D14 and C22 was determined using a G-LISA Ras activation assay kit (Cytoskeleton, catalog No. # BK131). The cells were serum-starved for 16 h and pre-treated with D14 and C22 at 99.3 μM and 137.5 μM, respectively, for 1 h; or Deltarasin at 5 μM for 3 h. Subsequently, the cells were stimulated with epidermal growth factor (EGF) (100 ng/mL) for 10 min. Lysates (1 mg/ml) were added to 96-well plates coated with Ras GTP-binding protein (Raf-RBD), following the manufacturer’s instructions. Experiments for each cell type were repeated three times.
The cells were serum-starved for 16 h and pre-treated with D14 at 99.3 μM or C22 at 137.5 μM for 1 h; or Deltarasin at 5 μM for 3 h. After pre-treatment, cells were stimulated with EGF at 100 ng/mL for 10 min. Whole-cell extracts were obtained by lysis of the Mia PaCa-2 cells in lysis buffer [20 mM Tris–HCl (pH 7.5), 1 mM EDTA, 150 mM NaCl, 1% Triton X-100, 1 mM NaVO3, 1 mM NaF, 10 mM β-glycerophosphate, 1 mM phenylmethylsulfonyl fluoride, and 1.2 mg/ml complete™ Lysis-M (Roche, Mannheim Germany) protease inhibitor cocktail]. The protein extracts were forced through a 22-gauge needle 10 times and centrifuged for 10 min at 14,000 rpm at 4 °C, and the protein concentration was determined using the Pierce™ BCA Protein Assay kit (Thermo Fisher Scientific, Waltham, MA, USA). Approximately 25 μg of protein was separated by 10% SDS-PAGE and transferred to nitrocellulose membranes. Then it was incubated with the following primary antibodies: Total ERK (Cell Signaling-9102; 1: 1000), pERK (Cell Signaling-9101; 1: 1000), Total AKT (Cell Signaling-9272 1: 1000), pAKT(Cell Signaling-4060 1: 1000), and anti-GAPDH (Gene Tex-GTX100118 1:100,000). Immunodetection was performed using a ChemiDoc™ Imaging Systems (BIO-RAD). Densitometry analysis was performed using the software ImageJ version 1.45 (National Institute of Health, USA).
MAPK activation profiling
Cells were rinsed with cold PBS and immediately lysed in buffer supplemented with 4xcOmplete™ EDTA-free Ultra Protease Inhibitor Cocktail (Sigma-Aldrich) and 1xPhosSTOP™ (Sigma-Aldrich) at 4 °C for 30 min. Following centrifugation at 14,000×g for 5 min, supernatants were transferred into a clean tube and protein concentrations were determined using the Precision Red Advanced Protein Assay (Cytoskeleton, Inc. ADV02-A). Lysates were diluted and analyzed using the Human Phospho-MAPK Arrays (Proteome Profiler; R&D Systems; Minneapolis, MN, USA) according to the manufacturer’s instructions. Nitrocellulose membranes were scanned using a ChemiDoc™ Imaging Systems (BIO-RAD Laboratories, Inc.).
Treatment of subcutaneous pancreatic carcinoma xenografts
Male immune-deficient Nu/Nu nude mice at 6 weeks of age (CINVESTAV, Mexico) were maintained in pathogen-free conditions with irradiated chow. The animals were subcutaneously injected in the back with 5 × 106 MIA PaCa-2 cells per tumor in 0.1 ml of sterile phosphate-buffered saline. When MIA PaCa-2 cells reached palpable tumors (>100mm3), mice were divided randomly into three groups receiving vehicle (10% DMSO, 0,05% Carboxy Methyl Cellulose and 0,02% Tween 80 in PBS) (n = 10), D14 at 20 mg kg− 1 (n = 7), or C22 at 10 mg kg− 1 (n = 5 subcutaneous injected in the both flanks) and 20 mg kg− 1 (n = 10) administered by intra-peritoneal injection three times per week. Body weight was measured once a week, whereas tumors were measured twice weekly. Tumor sizes were calculated by the following formula: [(length x width2)/2 in mm.
Histology and immunohistochemical staining of xenograft tumors
One day after the last treatment, mice were sacrificed in a CO2 chamber and the xenograft tumors were resected, fixed in 4% buffered formalin and embedded in paraffin. The tumors were cut using a microtome obtaining 2 μm slices. For hematoxylin and eosin (H & E) staining, the tissues were deparaffinized in xylene, hydrated in dehydrated alcohol starting from absolute ethanol to distilled water, stained for 2 min with Harris Hematoxylin, decolorized with 0.5% acid alcohol and fixing the color in lithium carbonate for 1 min, washed in distilled water, in 96% ethanol and stained with Sigma Eosin, washed and dehydrated in gradual alcohol changes until absolute alcohol was reached, allowed to dry at room temperature, mounted and observed, to identify the site of the injury. For immunohistochemical staining, the tissues were deparaffinized in xylene, hydrated in alcohols starting from absolute ethanol to distilled water, the epitopes were unmasked with 10 mM Citrate buffer at pH 6.03, washed with PBS pH 7.4. Endogenous peroxidase was blocked with 0.9% H2O2 for 15 min, then cross-sections were block with 3% BSA for 1 h. The antibodies Ki-67 (BIOCARE MEDICAL API 3156 AA) and CK 19 (GENETEX GTX110414) were diluted in PBS containing 1% BSA, the primary antibody was incubated at room temperature for 40 min, washed with PBS for 3 min, incubated with the biotinylated secondary antibody for 20 min at room temperature, washed with PBS for 3 min, incubated with streptavidin for 15 min, washed with PBS for 3 min. Reactions were incubated with 4% diaminobenzidine (DAB),counterstained with Harry’s Hematoxylin for 30 s, washed with distilled water, dehydrated in gradual changes of ethanol from distilled water to absolute Ethanol, allowed to dry at room temperature, mounted and observed.
The statistical significances of the differences among the data were determined by Tukey’s multiple comparisons test, using GraphPad Prism® 6 software (San Diego, CA, USA). P < 0.05 was considered statistically significant. Values are presented as the means ± s.e.m. (standard error of the mean).