Docking of Protein Kinase B Inhibitors: Implications in the Structure-Based Optimization of a Novel Scaffold

Docking of Protein Kinase B Inhibitors: Implications in the Structure-Based Optimization of a Novel Scaffold
of 8
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
  Docking of Protein Kinase B Inhibitors:Implications in the Structure-Based Optimizationof a Novel Scaffold Alicia Herna´ ndez-Campos 1 , IsraelVela´ zquez-Martı´nez 1 , Rafael Castillo 1 ,Fabian Lo´ pez-Vallejo 2 , Ping Jia 3 ,Yongping Yu 3 , Marc A. Giulianotti 2 andJose L. Medina-Franco 2, * 1 Facultad de Qumica, Departamento de Farmacia, Universidad Nacional Autnoma de Mxico, Mxico DF 04510, Mxico  2  Torrey Pines Institute for Molecular Studies, 11350 SW Village Parkway, Port St. Lucie, FL 34987, USA 3  College of Pharmaceutical Science, Zhejiang University, Hangzhou 310058, China  * Corresponding author: Jose L. Medina-Franco,  Protein kinase B (PKB   ⁄   AKT) is an attractive thera-peutic target in anticancer drug development. Wehave recently identified by docking-based virtualscreening a low micromolar AKT-2 inhibitor. Addi-tionally, the virtual screening hit represents anovel AKT-2 inhibitor scaffold. In this work, wediscuss a structure-based design strategy towardthe optimization of this hit. Following this strat-egy and using a herein validated docking protocol,we conducted the design of novel compoundswith expected improved activity over the parentcompound. The newly designed molecules havehigh predicted affinity for AKT-2; are syntheticallyaccessible and are contained within the kinase-rel-evant property space.Key words:  AKT, cancer, drug design, structure–activity relation-ships Abbreviations:  HBA, hydrogen bond acceptors; HBD, hydrogenbond donors; IFD, induce-fit docking; MOE, molecular operating environ-ment; MW, molecular weight; RB, number of rotatable bonds; SAR,structure-activity relationships; TPSA, topological surface area; XP, extraprecision.Received 6 January 2010, revised 4 March 2010 and accepted for publi-cation 25 May 2010 The serine   ⁄   threonine kinase B, also known as AKT, has severaldownstream targets that regulate a number of processes associatedwith cell growth, differentiation and division. AKT is frequentlyamplified and over-expressed in human cancer cells and its inhibi-tion is a promising therapeutic approach for the treatment ofcancers (1,2). There are three known subtypes, AKT-1   ⁄   PKB a , AKT-2   ⁄   PKB b  and AKT-3   ⁄   PKB c . Each one is associated with differenttypes of cancers. In particular, AKT-2 is amplified in pancreatic,breast and ovarian tumors. AKT-3 is over expressed in hormone-insensitive breast and prostate cancers (1). Aberrations in AKT-1 areless common. AKT has an N-terminal pleckstrin homology domain, ahinge region, a central kinase domain, and a C-terminal region (3).The kinase domains have a large similarity of more than 85% andthe binding pocket residues are the same (3,4). To date, small mole-cules targeting the ATP-binding site in the kinase domain, and allo-steric inhibitors interfering with the pleckstrin homology domainfunction have been reported, among others (3–6). AKT inhibitors,either ATP competitors or compounds that interact with regulatorydomains, have shown promising activity in cancer treatment. It isthought that subtype-selective inhibitors are needed for optimalefficacy with acceptable toxicity (1). However, it remains to bedetermined if subtype selective inhibitors have a larger therapeuticwindow over compounds that inhibit all three subtypes (4). Forexample, a pan-AKT inhibitor with nanomolar activity against thethree subtypes has been evaluated as an intravenous agent in clini-cal trials in patients with cancer (7). Small molecules targeting theATP-binding site have been reported. These include isoquinoline-5-sulfonamides (8,9), pyrazole (see below), (10) indazole (11) andaminofurazan analogs (7). Several of these inhibitors haven beendeveloped using structure-based design techniques and arereviewed in Ref. (4). Despite the fact that some current AKT-2 inhib-itors have shown promising activity for further development, thereare still issues that need to be addressed including toxicity and lackof suitable pharmacokinetic and pharmacodynamic properties (7,12).Thus, there is a critical need to develop novel scaffolds with AKT-2inhibitory activity.Using fragment-based and iterative structure-based screening meth-odologies, the activity of the pyrazole derivatives  1 – 9  was opti-mized leading to the nanomolar inhibitors  8a  and  9  (Figure 1) (10).A crystal structure of the AKT-2 in complex with compound  9  hasbeen published (10). As part of our on-going effort involving anti-cancer drug discovery (13–15), we recently reported the discoveryof the low micromolar AKT-2 inhibitor  11  with a distinct scaffold(Figure 1) (13). In that work, we conducted a multistep docking-based virtual screening of a large compound database with thecrystal structure of AKT-2 in complex with  9 . Compound  11  wasthe top ranked molecule in docking and our next goal is to optimizeits activity. Based on previous results (13) and molecular modeling 269 Chem Biol Drug Des 2010;   76:  269–276 Research Letter ª  2010 John Wiley & Sons A/S  doi: 10.1111/j.1747-0285.2010.01002.x  studies discussed in this work, we hypothesize that the crystalstructure of AKT-2 in complex with  9  can be used to conduct thestructure-guided optimization of  11 .In this work, we report the docking of the eleven AKT-2 inhibitorsin Figure 1 with the crystal structure of AKT-2 in complex with com-pound  9  using the program Glide. Compounds  1 – 10  were obtainedfrom the literature (8,10). We also docked the virtual screening hit 11  (Figure 1). As part of the validation of the docking protocol, wefound that docking with Glide was able to reproduce the bindingmode of the co-crystal pyrazole. In addition, the docking scores ofthe eleven compounds had an excellent agreement with the experi-mental activity. Comparison of the predicted binding mode of  11 with the co-crystal inhibitor  9  suggested modifications to the chem-ical structure of the virtual screening hit to increase its potency.Newly designed compounds showed the expected binding mode indocking and have a more favorable docking energy than the parentcompound. In addition, the new molecules are contained within thekinase-relevant property space and are synthetically accessible.Details of the docking of AKT-2 inhibitors and molecular modelingstudies toward the optimization of the activity of compound  11  arepresented in the following paragraphs. Methods and Materials In order to conduct the docking studies, the structures of all theligands were prepared using Molecular Operating Environment(MOE), version 2008.10. a The crystal structure of AKT-2 in complexwith the pyrazole-based inhibitor  9  was retrieved from the ProteinData Bank (PDB entry 2UW9) (16). Docking was performed with theGlide (Grid-based Ligand Docking with Energetics) program, version5.0. b The Protein Preparation Wizard module of Maestro was usedto prepare the protein. c During protein preparation, water moleculesand peptide substrate (GSK-3 b ) were deleted. For docking, the scor-ing grids were centered on the crystal structure of  9  using thedefault bounding box sizes; with an inner box of 10  on each sideand an outer box of 24  on each side. Flexible docking withdefault parameters was used. Glide XP (extra precision) wasemployed for all docking calculations (17). The best docked poseswere selected as the ones with the lowest GlideScore; the morenegative the Glide Score, the more favorable the binding. In orderto explore the putative binding interactions of the ligands withAKT-2, the top ranked binding mode found by Glide in complex withthe binding pocket of the enzyme were subjected to full energyminimization using the MMFF94x force field implemented in MOE Figure 1:  Chemical structuresand IC 50  values of AKT-2 inhibitorsconsidered in this study. Activityof  6a  and  6b  corresponds tothe racemic mixture. Compounds 1 – 10  were assayed under thesame experimental conditions (10).Compound  11  was discoveredrecently by docking-based virtualscreening (13). Herna´ ndez-Campos et al. 270  Chem Biol Drug Des   2010; 76: 269–276  until the gradient 0.001 was reached. The default parametersimplemented into the MOE's LigX application were used. a Results and Discussion Before docking compounds in Figure 1 with AKT-2, the docking pro-tocol was validated. The co-crystal inhibitor  9  was removed fromthe active site and docked back into the binding site. The rootmean square deviation between the predicted conformation and theobserved X-ray crystallographic conformation was 0.40, indicatingthe ability of the docking protocol to reproduce the binding mode of 9  (Figure 2A). Docking of AKT-2 inhibitors  Pyrazole derivatives  1 – 9  (10), the isoquinoline-5-sulfonamide  10 (8), and the lead compound  11  (13) were docked into the ATP bind-ing site of AKT-2. The docking scores and the biological activity aresummarized in Table 1. The biological activity of compounds  1 – 11 was obtained from the literature (10,13). Activities were convertedinto the corresponding  ) logIC 50  (pIC 50 ) values, were IC 50  is theeffective concentration of the compound required to achieve 50%of inhibition of AKT-2. To note, the activity reported for compounds 1 – 10  was determined using the same conditions (10). The activityreported for compound  11  was measured by a different group (13).The activity of compounds  6a  and  6b  in Table 1 was  approximated  as the activity reported for the racemic  6  (10). In contrast, theactivity for compounds  8a  and  8b  corresponds to the activityreported for the ( R  )- and ( S  )-enantiomers, respectively.The docking scores calculated with Glide had an excellent correla-tion with the biological activity, particularly with the rank orderingof the compounds (Table 1); compounds with the highest activitiessuch as  9  (pIC 50  = 7.74; IC 50  = 18 n M ) and  8a  (pIC 50  = 7.47;IC 50  = 34 n M ) also displayed the top-ranked docking scores ( ) 9.07and  ) 9.12 kcal   ⁄   mol, respectively). In contrast, compounds with thelowest activities such as  1  (pIC 50  = 4.1; IC 50  = 80  l M ) and  5 (pIC 50  = 3.87; IC 50  = 135  l M ) showed the lowest docking scores( ) 6.41 and  ) 6.05 kcal   ⁄   mol, respectively). The isoquinoline-5-sulfon-amide  10  was included in this work as a control; its activity wasmeasured under the same experimental conditions as the pyrazolederivatives; however, compound  10  has a different scaffold. Inter-estingly,  10  was ranked in position twelve by docking (dockingscore of  ) 6.37 kcal   ⁄   mol); this is in good agreement with its rankingposition of ten according to its experimental activity (pIC 50  = 5.12;IC 50  = 7.5  l M ). In turn, compound  11 , also with a distinct scaffold,showed a better docking score than the isoquinoline-5-sulfonamide 10  and pyrazoles  1  and  5 . This is also consistent with the superiorAKT-2 inhibitory activity of  11  over  1 ,  5  and  10  (Table 1). More-over, compound  11  showed a lower docking score than the submi-cromolar or nanomolar pyrazole inhibitors  6 – 9 , in furtheragreement with the experimental activities. Taken together, theseobservations suggest that the binding modes predicted with Glidefor compounds  1 – 11  are reasonable.A comparison of the docking models for compounds  8a  and 10  is depicted in Figure 2B. Crystallographic  9  is shown as a AB Figure 2:  (A) Comparison between the binding position of  9 found within the crystal structure (yellow) and the binding modepredicted by Glide (red). (B) Comparison of the binding conforma-tions of the pyrazole  8a  (green) and isoquinolin-5-sulfonamide  10 (cyan) predicted by Glide. Crystallographic  9  (yellow) is shown as areference with hydrogen bonds displayed as magenta dashes. Non-polar hydrogen atoms are omitted. Table 1:  Results of docking of the AKT-2 inhibitors  1 – 11 a CompoundDockingscore (kcal   ⁄    mol)Dockingscore rank pIC 50  pIC 50  rank 1  ) 6.41 11 4.10 12 2  ) 7.08 9 4.92 11 3  ) 7.82 8 5.28 9 4  ) 7.85 7 5.52 8 5  ) 6.09 13 3.87 13 6a  ) 8.72 3 6.29 b 5.5 6b  ) 8.47 5 6.29 b 5.5 7  ) 8.56 4 6.70 3 8a  ) 9.12 1 7.47 2 8b  ) 8.28 6 6.55 4 9  ) 9.07 2 7.74 1 10  ) 6.37 12 5.12 10 11  ) 6.72 10 5.82 7 a All pIC 50  values were obtained from Ref. (10) except for  11  that wasobtained from Ref. (13). b Approximated as the activity reported for the racemic  6  (10). Docking of Protein Kinase B Inhibitors Chem Biol Drug Des   2010; 76: 269–276  271  reference. Two-dimensional representations of the optimized dock-ing models of  1 – 10  are shown in Figure S1. Compounds  8a  and 9  (Figure 2B) and all other pyrazole derivatives, have a similarbinding mode. The pyrazole ring forms hydrogen bonds with thebackbone NH of Ala232 and backbone carbonyl of Glu230 (Fig-ure S1). This is in agreement with the observations derived fromcrystallographic structure of AKT-2 in complex with  9  and thecrystallographic structures of the PKA-PKB chimera in complexwith  3  and  8b  (10). A hydrogen bond between the nitrogen atomof the isoquinoline ring of compound  10  and the backbone NH ofAla232 was also predicted by Glide. Compounds  2 – 4  showed twoadditional hydrogen bonds with the side chain of Glu236 and thebackbone carbonyl of Glu279 (Figure S1). These observations arealso consistent with the crystallographic structure of the PKA-PKBchimera in complex with  3 . As suggested previously, the inter-actions of the basic amine with the electronegative pocket ofAKT-2 formed by residues Glu236, Glu279, Asn280 and Asp293explain the increased activity of  2 – 4  over  1  and  5  (10). Similarhydrogen bonds with Glu236 and Glu279 were predicted for thelow micromolar inhibitor  7  and the nanomolar inhibitors  8a  and 8b , respectively (Figure S1).A comparison of the binding model of  11  obtained with Glide andthe crystallographic position of  9  is shown in Figure 3A. The 1,3-benzoxazol-2(3 H  )-one ring of  11  occupies the same binding pocketas the pyrazole ring of  9  making hydrogen bonds with the sidechains of Thr213 and Thr292 (13). The 2 H  -1,4-benzoxazin-3(4 H  )-onering of  11  partially overlaps the chlorophenyl ring of  9  making andhydrogen bond with Asp293. A very similar binding model for  11 was obtained previously with two different docking programs (13).Notably, the phenyl ring of  11  is close to the secondary amine of 9  and to Glu236 in the electronegative pocket discussed above. Itis also worth mentioning the three-dimensional similarity of  9  and 11  (Figure 3A); both compounds have a Y-shape and similar flexibil-ity (e.g., three rotatable bonds).In view of the fact that  11  was a top-ranked hit in virtual screen-ing with the crystallographic structure of AKT-2 bound to  9  (13) andthat the predicted binding mode for  11  is similar to the co-crystalinhibitor  9 , we hypothesize that the predicted binding interactionsof  11  with AKT-2 can be the basis to initiate a structure-basedoptimization program. To further test this hypothesis considering theprotein flexibility, we conducted induce-fit docking (IFD) for com-pound  11  with the same crystallographic structure used during therigid-receptor docking with Glide. We used the validated IFD proto-col developed by Schrçdinger Inc. with default parameters. d The IFDapproach implemented in Schrçdinger iteratively combines rigidreceptor docking using Glide with sampling side chain degrees offreedom in the receptor while allowing backbone movementsthrough minimization with the program Prime (18). For the Glidecomponent, Glide XP was employed. We observed a very similarbinding mode for compound  11  predicted by induce-fit and rigid-receptor docking (data not shown). These observations further sup-ported the  approach   of using this crystal structure of AKT-2 (e.g.,PDB code 2UW9) to guide the optimization of  11 . The protocol dis-cussed earlier to explain the structure–activity relationships (SAR)of the pyrazole derivatives, and the good agreement between theexperimental activity and docking scores, including the parent ABC Figure 3:  (A) Comparison of the predicted binding mode of com-pound  11  (cyan) with the crystallographic position of  9  (yellow). (B)Comparison of the binding modes of  11  (cyan) with the representa-tive designed molecules  16  (purple),  20  (gray) and  21  (orange). (C)Comparison of the crystallographic position  9  (yellow) with the pre-dicted binding modes of  16  (purple) and  21  (orange). Hydrogenbonds for  21  are displayed as magenta dashes. Non-polar hydrogenatoms are omitted. Herna´ ndez-Campos et al. 272  Chem Biol Drug Des   2010; 76: 269–276  compound  11 , supports the use of docking to guide the structure-based optimization of  11 . Molecular modeling studies toward the structure-based optimization of the virtual screening hit 11 The overall goal of the design was to identify synthetically accessibleanalogs of  11  with docking scores better than  11  and, preferably,scores close to or better than the nanomolar inhibitor  9 . In addition,we confirmed that the predicted binding modes for the new mole-cules showed the expected binding interactions with the proteinas derived from the SAR of experimentally known AKT-2 inhibitors.The structure-based design was divided into two main parts. In thefirst part, a series of compounds containing the same or similarheterocyclic rings present in  11  such as 1,3-benzoxazol-2(3 H  )-one,2 H  -1,4-benzoxazin-3(4 H  )-one, or 3,4-dyhidroquinoxalin-2(1 H  )-one weredesigned. To note,  11  is a nearly symmetric molecule (Figure 1). Atthis stage, we decided to keep two heterocyclic rings in the designedmolecules because, as discussed earlier, the 1,3-benzoxazol-2(3 H  )-one and 2 H  -1,4-benzoxazin-3(4 H  )-one ring in  11  seem to form keyhydrogen bonds with Thr213, Thr292 and Asp293. Following thisapproach, compounds such as  12  and  17  in Table 2 were designed.These two molecules are very similar to  11  with the only differencebeing that  12  and  17  have two identical heterocyclic rings, either2 H  -1,4-benzoxazin-3(4 H  )-one or 3,4-dihydroquinoxalin-2(1 H  )-one,respectively. Docking models of  12  and  17  with AKT-2 (see below)are very similar to the binding model of  11  indicating that the corescaffold for the new molecules (Table 2) is reasonable.In the second part of the design, the replacement of the phenyl ringof  11  with polar groups was proposed. This is based on the dockedposition of the phenyl ring of  11  in the electronegative pocket ofAKT-2, and the proximity of the phenyl group to the secondaryamine of  9  as discussed earlier (Figure 3B). Representative mole-cules are depicted in Table 2 along with their corresponding molec-ular weight (MW) and docking scores. Following this approach, thephenyl group is replaced, for instance, by a hydroxymethyl group( 13  and  18 ), pyrrolidin-2-yl ( 14  and  19 ), pyrrolidin-1-ylmethyl ( 15 and  20 ) or and aminomethyl group ( 16  and  21 ).The series of 2 H  -1,4-benzoxazin-3(4 H  )-ones and 3,4-dihydroquinoxa-lin-2(1 H  )-ones were docked with AKT-2 using the same validateddocking protocol described earlier. Docking results are summarizedin Table 2. Noteworthy, the docking scores of compounds  12 – 21 were more favored (i.e., more negative) than the docking score ofthe parent inhibitor  11 ,  ) 6.72 kcal   ⁄   mol (Table 1). Furthermore,docking scores of compounds  20  ( ) 8.82 kcal   ⁄   mol) and  21 ( ) 9.14 kcal   ⁄   mol) were similar to the docking score of the nanomo-lar inhibitor  9  ( ) 9.07 kcal   ⁄   mol). We also docked bis-1,3-benzoxazol-2(3 H  )-ones but these molecules showed lower docking scores thanthe corresponding 3,4-dihydroquinoxalin-2(1 H  )-ones and 2 H  -1,4-benz-oxazin-3(4 H  )-ones (data not shown). Based on these observationsand the good agreement between the docking scores with theexperimental  rank  -order of compounds  1 – 10  and the parent inhibi-tor  11  discussed earlier, it is expected that compounds  12 – 21  willbe more active than  11 . We want to emphasize that that the dock-ing scores of the newly designed analogs of  11  were obtained indocking experiments with the crystallographic structure of a differ-ent scaffold  9 . No quantitative predictions of potency for eachdesigned compound are made.Tridimensional binding models obtained with docking representative2 H  -1,4-benzoxazin-3(4 H  )-ones and 3,4-dihydroquinoxalin-2(1 H  )-onesare depicted in Figure 3B. Two-dimensional representations of theoptimized docking models are shown in Figure S2. According to theoptimized binding models, all analogs  12 – 21  adopt a very similarbinding mode to the parent  11 . The corresponding 2 H  -1,4-benzoxazin-3(4 H  )-one or 3,4-dihydroquinoxalin-2(1 H  )-one rings make hydrogenbonds with side-chain atoms of Thr213, Thr292 and Asp293, similarto the hydrogen bonds predicted for  11 . In addition, compoundswith polar groups at the fourth position of the pyridine ring( 13 – 16  and  18 – 21 ) make a hydrogen bond with the side chain ofGlu236 (Figure S2). A similar hydrogen bond is observed in the crys-tallographic structure of  9 , and it is also predicted for othersubmicromolar and nanomolar AKT-2 inhibitors. This hydrogen bond Table 2:  Chemical structure, molecular weight (MW) and dock-ing scores with AKT-2 of representative newly structure-baseddesigned analogs of  11 NHNX XHNO OR Compound X R MWDocking score(kcal   ⁄    mol) 12  O 449.5  ) 7.81 13  O  HO  403.4  ) 7.88 14  O NH 442.5  ) 8.30 15  O N 456.5  ) 8.67 16  O  H 2 N  402.4  ) 8.98 17  NH 447.5  ) 6.76 18  NH  HO  401.4  ) 8.04 19  NH NH 440.5  ) 8.73 20  NH N 454.5  ) 8.82 21  NH  H 2 N  400.5  ) 9.14 Docking of Protein Kinase B Inhibitors Chem Biol Drug Des   2010; 76: 269–276  273
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks