Accepted Manuscript Title: Structure-based Grafting and Identification of Kinase Inhibitors to Target mTOR Signaling Pathway as Potential Therapeutics for Glioblastoma Author: Yu-Hui Cui Jiong Chen Tao Xu Heng-Li Tian PII: DOI: Reference:
S1476-9271(15)00002-X http://dx.doi.org/doi:10.1016/j.compbiolchem.2015.01.001 CBAC 6386
To appear in:
Computational Biology and Chemistry
Received date: Revised date: Accepted date:
11-9-2014 30-12-2014 1-1-2015
Please cite this article as: Cui, Yu-Hui, Chen, Jiong, Xu, Tao, Tian, Heng-Li, Structure-based Grafting and Identification of Kinase Inhibitors to Target mTOR Signaling Pathway as Potential Therapeutics for Glioblastoma.Computational Biology and Chemistry http://dx.doi.org/10.1016/j.compbiolchem.2015.01.001 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Structure-based Grafting and Identification of Kinase Inhibitors to Target mTOR Signaling Pathway as Potential Therapeutics for Glioblastoma
Yu-Hui Cui,* Jiong Chen, Tao Xu, Heng-Li Tian
Department of Neurosurgery, Shanghai Jiao Tong University Affiliated Sixth People’s Hospital, Shanghai, 200233, China
RI
PT
*Corresponding author at: Department of Neurosurgery, Shanghai Jiao Tong University Affiliated Sixth People’s Hospital, Shanghai 200233, China. E-mail address:
[email protected] (Y.-H. Cui).
A
N
U
SC
Graphical abstract An integrated computational protocol is described to graft hundreds of inhibitor ligands from their complex crystal structures with cognate kinases into the active pocket of mTOR kinase domain, and to virtually evaluate the binding strength of these inhibitors to their non-cognate target mTOR. Kinase assay is performed to solidify the findings suggested by computational investigations.
EP
TE
D
M
Highlights (i) An protocol is described to graft inhibitors from their cognate kinases to non-cognate mTOR. (ii) The grafted inhibitor–mTOR affinity is virtually evaluated using a consensus scoring strategy. (iii) A number of identified inhibitors are assayed to determine their inhibition against mTOR. (iv) Diverse nonbonded interactions are found at mTOR–inhibitor complex interface.
CC
Abstract
A
mammalian target of rapamycin (mTOR), a key mediator of PI3K/Akt/mTOR signaling pathway, has recently emerged as a compelling molecular target in glioblastoma. The mTOR is a member of serine/threonine protein kinase family that functions as a central controller of growth, proliferation, metabolism and angiogenesis,
but its signaling is dysregulated in various human diseases especially in certain solid tumors including the glioblastoma. Here, considering that there are various kinase inhibitors being approved or under clinical or preclinical development, it is expected that some of them can be re-exploited as new potent agents to target mTOR for glioblastoma therapy. To achieve this, a synthetic pipeline that integrated molecular grafting, consensus scoring, virtual screening, kinase assay and structure analysis was described to systematically profile the binding potency of various small-molecule inhibitors deposited in the protein kinase–inhibitor database against the kinase domain
PT
of mTOR. Consequently, a number of structurally diverse compounds were
RI
successfully identified to exhibit satisfactory inhibition profile against mTOR with IC50 values at nanomolar level. In particular, few sophisticated kinase inhibitors as well as a
SC
flavonoid myricetin showed high inhibitory activities, which could thus be considered
U
as potential lead compounds to develop new potent, selective mTOR inhibitors.
N
Structural examination revealed diverse nonbonded interactions such as hydrogen
A
bonds, hydrophobic forces and van der Waals contacts across the complex interface of
M
mTOR with myricetin, conferring both stability and specificity for the mTOR–inhibitor binding.
TE
D
Keywords: rational drug design; mTOR; kinase inhibitor; glioblastoma
EP
1. Introduction
Glioblastoma is one of the most aggressive solid cancers and the most common
CC
primary brain tumor, which invades the surrounding brain, making complete surgical excision highly improbable. They are also among the most radiotherapy- and
A
chemotherapy-resistant cancer types with a median survival duration of 12–15 months after initial diagnosis (Stupp et al., 2005). Thus, new therapeutic approaches for glioblastoma are needed. Recently, the efficacy of small-molecule kinase inhibitors has been changed as standard clinical practice for several solid tumors (De Witt Hamer, 2010). The glioblastoma represents a compelling disease for kinase inhibitor therapy
because the majority of these tumors harbor genetic alterations that result in aberrant activation of growth factor signaling pathways (Mellinghoff et al., 2012). The PI3K/Akt/mTOR pathway regulates several normal cellular functions that are critical for tumorigenesis of glioblastoma, including cellular proliferation, growth, survival and mobility (Morgensztern and McLeod, 2005). The mammalian target of rapamycin (mTOR), a serine/threonine protein kinase, is a key mediator of the pathway; signaling functions of mTOR are distributed between at least two distinct mTOR protein complexes: mTORC1 and mTORC2. The mTORC1 is an association of
PT
mTOR with a number of proteins such as PRAS40 and the rapamycin-sensitive adapter
RI
protein of mTOR (Raptor), whereas the mTORC2 consists of mTOR and a separate protein complex including the rapamycin-insensitive companion of mTOR (Rictor)
SC
(Fan and Weiss, 2012). The rapamycin and its analogs (rapalogs) are the well
U
established first generation mTOR inhibitors, which serve as allosteric mediators to
N
irreversibly regulate the conformation and activity of mTOR. However, the response
A
rate in solid tumors where these first generation inhibitors have been used as a
M
single-agent therapy have been modest, due to partial mTOR inhibition that rapamycin and rapalogs are not sufficient for achieving a broad and robust anticancer effect
D
(Brachmann et al., 2009). Currently, a new generation of ATP-competitive inhibitor
TE
that targets the catalytic site of mTOR kinase domain exhibits potent efficacy and are in
EP
early clinical trials (Benjamin et al., 2011). In this study, we considered to perform a systematic virtual screening against a large
CC
pool of kinase inhibitors to identify those that may potentially target mTOR. The kinase inhibitors have been shown to have a broad specificity across protein kinase family
A
members due to their high conservation in sequence, structure and function. Recently, Shao and co-workers (2014) demonstrated that certain existing inhibitors that were originally designed for their cognate kinase targets can also bind efficiently to protein kinase C. Thus, some non-cognate ligands are thought to be the promising lead candidates and could be further optimized to develop new selective, potent mTOR
inhibitors for glioblastoma therapy. We also performed kinase inhibition assay to solidify the findings obtained from computational modeling and virtual screening.
2. Materials and Methods 2.1. Data compilation and setup Patel and Doerksen (2010) have developed a protein kinase–inhibitor database that contains 755 unique, curated and annotated PDB protein kinase–inhibitor complexes. We have systematically examined these complex crystal structures, from which only
PT
those with high resolution (< 3 Å) and low thermal motion (R-factor < 0.25) were
RI
considered. In addition, the kinase active sites should be complete; it was required that no missing atoms or other chemical components such as the functional moieties of
SC
backbones and side chains nearby the sites were found in the crystal structures. Next,
U
the absent hydrogen atoms of kinase proteins were added using REDUCE program
N
(Word et al., 1999), and the chemical attributes of cocrystallized ligand molecules such
A
as atomic hybridization state and bond order type were assigned with I-INTERPRET
M
program (Zhao et al., 2007).
In order to ensure the reliability of consensus scoring used in this study, we herein
D
employed a set of kinase–inhibitor binding data to test virtual scoring functions. The set
TE
contains 77 crystal structure-solved, affinity-known kinase–inhibitor complexes that
EP
were compiled by Wang et al. (2014) to develop a quantitative structure-activity relationship predictor for structure-based evaluation of inhibitor interaction with their
CC
kinase targets. The crystal structures and binding affinity of these kinase–inhibitor complexes were retrieved from the PDB (Berman et al., 2000) and PDBbind (Wang et
A
al., 2005) databases, respectively. This nonredundant set is structurally diverse in ligand chemicals and represents a wide range of kinase–inhibitor binding affinities.
2.2. Computational modeling of mTOR–inhibitor complex structures
A method proposed by Lu et al. (2014) was modified here to carry out computational modeling of non-cognate mTOR–inhibitor complex structures. Briefly, ligand molecule is assumed to adopt a similar binding mode to interact with different kinase receptors since the kinase family members are highly conserved. Hence, the inhibitor binding mode can be directly grafted from crystal complex template into the active pocket of mTOR. As shown in Figure 1, the crystal structure of mTOR kinase domain in apo state was obtained from the PDB database (Berman et al., 2000) under the accessible code 4JSV; the structure has only a moderate resolution so that it was
PT
first refined with a FoldX minimization procedure (Schymkowitz et al., 2005) to
RI
eliminate those unreasonable phenomena involved in the coarse-grained structure, such as atomic collisions and bond distortions. The refined mTOR kinase domain was then
SC
superposed onto the crystal template of inhibitor complex with its cognate kinase
U
target, resulting in mTOR/inhibitor/kinase system. The superposition was simply
N
carried out by least-squares fitting of the backbone Cα atoms between the two kinase
A
proteins using SPDBV program (Johansson et al., 2012). Subsequently, the kinase
M
template was removed manually from the system to obtain an artificial mTOR–inhibitor complex structure model, which was further optimized using the
TE
D
FoldX force field (Schymkowitz et al., 2005).
EP
2.3. Consensus scoring
A number of widely used scoring functions were considered to derive consensus
CC
scoring for the structure-based evolution of non-cognate mTOR–inhibitor binding potency, including ChemScore (Eldridge et al., 1997), X-Score (Wang et al., 2002),
A
DrugScore (Velec et al., 2005), DFIRE (Zhang et al., 2005), DOCK score (Meng et al., 1992) and Autodock score (Goodsell et al., 1996). The ChemScore and X-Score are knowledge-based potentials derived from statistical observations of close contacts between certain types of atoms or functional groups that occur more frequently than one would expect by a random distribution. The DrugScore and DFIRE are empirical
potentials that count the number of various types of interactions in a protein–ligand complex. The DOCK score and Autodock score are force field-based potentials that describe protein–ligand affinity by summing the strength of intermolecular van der Waals and electrostatic interactions between all atoms of the protein and ligand. Here, the Autodock score, DOCK score, ChemScore and X-Score were calculated using stand-alone programs AutoDock, DOCK, GOLD and X-ScoreTM, respectively. The DrugScore and DFIRE were computed with online servers DSX-ONLINE and dDFIRE/DFIRE2, respectively.
PT
A supervised strategy was employed to derive consensus evaluation for
RI
kinase–inhibitor systems from above six scoring functions. The strategy performed least squares fitting of the six scores to experimental affinity based on the 77
SC
structure-solved, affinity-known kinase–inhibitor complex samples, resulting in a
N
U
linear regression formula:
M
A
pK d c0 c1 Score1 c2 Score2 c3 Score3 c4 Score4 c5 Score5 c6 Score6 (1)
D
where the pKd is the negative logarithm of kinase–inhibitor dissociation constant Kd
TE
(affinity), Score1, Score2 ······ Score6 are the calculated score values separately using the six scoring methods, the c1, c2 ······ c6 are corresponding weight terms that
EP
determine the relative contributions of the six scores to kinase–inhibitor affinity, and c0 is the constant term that reflects those additional factors that are not included in the six
A
CC
scores, such as flexibility and entropy effect.
2.4. mTOR kinase inhibition The kinase assay for mTOR inhibition was as described previously (Knight et al., 2006; Feldman et al., 2009). Briefly, the assay was performed in a volume of 30 μL buffer containing 1 μg PHAS-I, 120 mM NaCl, 40 mM HEPES pH 7, 0.3% CHAPS, 4 mM MnCl2, 10 mM DTT, 1X Roche inhibitor cocktail-EDTA, 2 μg/mL heparin, 100
μM ATP, 2 μCi γ-32P ATP. Inhibitors were tested in a four-fold dilution series from 10 μM to 600 pM, and four measurements were made at each concentration. The kinase reaction was terminated by spotting onto nitrocellulose, which was washed several times with NaCl phosphoric acid. The radioactivity remaining on the nitrocellulose sheet was quantified by phosphorimaging, and IC50 values were determined by fitting the data to a sigmoidal dose-response curve.
3. Results and Discussion
PT
3.1. Scoring evaluation analysis
RI
The six scoring methods considered in this study were tested against the set of 77 structure-solved, affinity-known kinase–inhibitor complex samples (Wang et al.,
SC
2014). The atomic-resolution crystal structures of the 77 complexes were retrieved
U
from the PDB database (Berman et al., 2000), and corresponding binding affinities
N
(dissociation constant, Kd) were obtained from the PDBbind database (Wang et al.,
A
2005). Before performing analysis the crystal structures were processed with repairing
M
missing side chains, adding hydrogen atoms, and structure minimization. Based on the treated structures the six scoring methods were separately applied to deriving affinity
D
scores for the 77 complexes, and the obtained score values were plotted against the
TE
experimentally measured affinity (pKd = –log10Kd) (Figure 2). As can be seen, all the
EP
six scoring functions exhibit negative correlations with experimental affinity; this is expected because these scores are all indirect indicators of free energy change upon
CC
protein–ligand binding, that is to say, high affinity is always associated with negative score value. However, all the six scoring methods can only obtain a modest or moderate
A
correlation with experimental affinity, with Pearson’s correlation coefficients R ranging between –0.25 and –0.50. The results suggested that the affinity scores can only be used to qualitatively rank the relative binding capability of different small-molecule inhibitors to kinase receptors, although these scores seem to be able to properly reflect affinity tendency for different kinase–inhibitor interactions. For this it
is explained that the binding of inhibitor ligands to kinase receptors is dominated by various physicochemical factors, and any one of the six scoring methods can only independently characterize a limited aspect (but not all) of these factors and thus can only exhibit a moderate correlation with inhibitor affinity (Zhou et al., 2012). In this respect, we herein proposed combination of different scores to reproduce inhibitor affinity by using a supervised approach. To achieve this, multiple linear regression was employed to reconcile the six scoring methods by least-squares fitting of them to the experimentally measured affinity of the 77 complex samples, resulting in
PT
a typical linear regression equation that is expected to perform extrapolation on those
RI
unknown kinase–inhibitor complexes, including the non-cognate mTOR–inhibitor system. Therefore, we carried out statistical validation to test the internal stability and
SC
external predictability of the regression equation. It is clearly seen that a good linear
U
correlation between the calculated and experimental affinity values can be observed in
N
Figure 3 with coefficient of determination r = 0.786, suggesting a high internal stability
A
of the regression. Further, 7-fold cross-validation was performed to test the predictive
M
power of regression equation, that is, the 77 samples were randomly partitioned into seven equal-size subsets; a single subset was retained as the validation data for testing
D
the equation, and the remaining six subsets were used as training data. The
TE
cross-validation process was then repeated for 7 times, with each of the seven subsets
EP
used exactly once as the validation data. Consequently, a satisfactory prediction profile was obtained with cross-validation coefficient of determination rcv = 0.674, which
CC
satisfied the rcv > 0.5 threshold value for a predictable regression model as
A
recommended by Golbraikh and Tropsha (2002).
3.2. High-throughput virtual screening and kinase inhibition assay Here, we performed high-throughput virtual screening of potential mTOR binders against various kinase inhibitors deposited in the protein kinase–inhibitor database (Patel and Doerksen, 2010) by using the established regression predictor. In the
procedure, the cocrystallized inhibitor ligands were grafted from crystal complex templates with their cognate kinase targets into the active pocket of mTOR kinase domain, which were then minimized with FoldX force field (Schymkowitz et al., 2005) to eliminate atomic collisions and bond distortions in the artificial mTOR–inhibitor systems. Subsequently, the consensus scores were one-by-one calculated for these mTOR–inhibitor systems using regression predictor, and the histogram distribution of resulting score values is visualized in Figure 4. As can be seen, predicted pKd values exhibit a normal distribution with peak position at the 7–8 bin and only very few
PT
inhibitors can bind tightly to mTOR, suggesting that most inhibitor ligands are not the
RI
good binders of mTOR kinase since these inhibitors were originally not designed for the mTOR. There are two important regions in the distribution histogram: the region 1
SC
ranges from 6 to 9 that contains a large number of mTOR–inhibitor interaction pairs,
U
indicating that many inhibitors can only bind moderately to their non-cognate mTOR
N
target; the region 2 ranges from 11 to 13 that represents few high-affinity mTOR
A
binders.
M
Here, we only considered 27 inhibitor compounds in the region 2 (consensus scores > 11) as promising candidates. The 27 selected inhibitors are tabulated in Supporting
D
Information, Table S1. As can be seen, these compounds are structurally diverse,
TE
including derivatives of azepine, imidazole, indole, pyridazine, pyrimidine, quinoline
EP
and others. In particular, two flavonoids were identified as promising mTOR binders; the flavonoid compounds possess a broad spectrum of kinase inhibitory activity and
CC
recently Syed et al. (2013) reported that the dietary flavonoid fisetin exhibited potent mTOR inhibition. In addition, two FDA-approved drugs erlotinib (for lung cancer
A
therapy) and dabrafenib (for metastatic melanoma therapy) were also listed here. In order to test the results suggested by computational study, we further performed
kinase assay to determine the inhibitory potency of seven compounds in the 27 promising candidates against mTOR kinase activity, including ABL1 inhibitor PD180970, PAK1 inhibitor IPA-3, INSR inhibitor GS3, PDK1 inhibitor BX-517,
IKKβ inhibitor Bay 65-1942, B-Raf inhibitor dabrafenib and PDGFR inhibitor crenolanib (Table 1). Consequently, four out of the seven tested compounds showed (relatively) strong inhibition against mTOR kinase, with determined IC50 values at nanomolar level. In particular, the INSR inhibitor GS3 and IKKβ inhibitor Bay 65-1942 exhibited very high activities (IC50 = 4 ± 0.6 and 28 ± 4 nM, respectively), which could be considered as potential lead compounds to develop new potent, selective mTOR inhibitors. However, other three candidates, i.e. IPA-3, BX-517 and the approved dabrafenib, were measured to only show a weak mTOR inhibition profile
PT
(IC50 > 1000 nM), suggesting that these identified kinase inhibitors may not be the good
RI
choice for further mTOR inhibitor developments, although they were predicted to have strong binding capability toward mTOR kinase domain. This is not unexpected because
SC
a strong binding of small-molecule ligands to mTOR receptor does not mean a potent
U
inhibition associated with the binding; instead, there are many additional factors such
A
N
as chemical condition and allosteric effect that may also relate closely to the inhibition.
M
3.3. Identification and modification of mTOR flavonoid inhibitors Flavonoid compounds have long been widely used as small-molecule inhibitors of
D
various protein kinases for cancer chemoprevention (Hou and Kumamoto, 2010). Here,
TE
we selected a number of natural flavonoids as lead molecular entities to identify mTOR
EP
inhibitors, including fisetin, apigenin, luteolin, naringenin, myricetin, catechin, cyanidin, delphinidin, malvidin, pelargonidin, taxifolin, petunidin, hesperetin,
CC
quercetin, isorhamnetin, kaempferol, eriodictyol and flavopiridol (Table 2). These natural products have previously been found to target a broad spectrum of kinase family
A
members, such as PKC, Akt, CDK, Lck and Fyn (Faivre et al., 2006). First, the complex structures of these flavonoid ligands with mTOR kinase domain were modeled using the strategy shown in Figure 1, and their binding affinities were then calculated by consensus scoring predictor. Consequently, most of these flavonoid candidates were suggested as high or moderate binders of mTOR, with predicted
affinity values larger than 7.0. However, the five positively charged compounds, i.e. cyanidin, delphinidin, malvidin and pelargonidin, have been shown to interact weakly with mTOR (pKd < 6.0); this could be explained as that the positive charge commonly possessed by these compounds may disrupt the interaction considerably. Subsequently, five compounds (fisetin, myricetin, taxifolin, quercetin and flavopiridol) with predicted pKd > 7.0 were tested for their mTOR inhibition, and resultant IC50 values were listed in Table 2. Most of the five tested compounds exhibited high inhibitory potency against mTOR (IC50 < 100 nM), while only one has moderate activity (IC50 > 100 nM). Among
PT
the five compounds the myricetin possessed the highest potency (IC50 = 16 ± 2 nM),
RI
which has been shown to exhibit inhibitory capability for a number of protein kinases, such as PI3K, Fyn and p38 (Walker et al., 2000).
SC
The mTOR active site is composed of charged, hydrophobic and aromatic residues,
U
which can readily form a variety of nonbonded interactions with myricetin and other
N
flavonoid compounds that possess a number of hydrophobic aromatic rings and polar
A
hydroxyl moieties. The structure architecture of mTOR complexed with myricetin as
M
well as nonbonded interactions across the complex interface are shown in Figure 5. It is evident that there is a complicated network of diverse nonbonded interactions at the
D
complex interface (Zhou et al., 2012); the myricetin can form four specific hydrogen
TE
bonds separately with protein residues Pro2169, Glu2190, Asp2195 and His2340 as
EP
well as a number of nonspecific hydrophobic and van der Waals contacts with residues Ile2163, Val2227, Val2240, Ile2356, and Phe2358. All these nonbonded effects come
CC
together to confer a high affinity and specificity for the mTOR–myricetin binding,
A
resulting in a strong inhibition with IC50 = 16 ± 2 nM. 4. Conclusions Glioblastoma is the most common and most aggressive malignant primary brain tumor in humans and the mTOR signaling pathway has long been considered as a promising target for glioblastoma therapy. In the current study, we developed an integrated protocol to systematically profile the inhibitory potency of kinase inhibitors
against mTOR, from which a number of active compounds were identified to show a moderate or high inhibition profile for mTOR. In particular, the INSR inhibitor GS3 and IKKβ inhibitor Bay 65-1942 as well as a flavonoid compound exhibited very high activities, which could be considered as potential lead compounds to develop new potent, selective mTOR inhibitors. Further, complex structure analysis was performed to dissect the structural basis and energetic property of intermolecular interaction between the mTOR kinase domain and identified inhibitors, revealing a complicated network of nonbonded interactions across the tightly packed interface of non-cognate
RI
PT
mTOR–inhibitor complexes.
References
SC
Berman, H.M., Westbrook, J., Feng, Z., Gilliland, G., Bhat, T.N., Weissig, H.,
U
Shindyalov, I.N., Bourne, P.E., 2000. The protein data bank. Nucleic Acids Res.
N
28, 235–242.
A
Brachmann, S., Fritsch, C., Maira, S.M., García-Echeverría, C., 2009. PI3K and mTOR
M
inhibitors: a new generation of targeted anticancer agents. Curr. Opin. Cell Biol. 21, 194–198.
D
Benjamin, D., Colombi, M., Moroni, C., Hall, M.N., 2011. Rapamycin passes the torch:
TE
a new generation of mTOR inhibitors. Nat. Rev. Drug Discov. 10, 868–880.
EP
De Witt Hamer, P.C., 2010. Small molecule kinase inhibitors in glioblastoma: a systematic review of clinical studies. Neuro. Oncol. 12, 304–316.
CC
Eldridge, M.D., Murray, C.W., Auton, T.R., Paolini, G.V., Mee, R.P., 1997. Empirical
A
scoring functions: I. The development of a fast empirical scoring function to estimate the binding affinity of ligands in receptor complexes. J. Comput. Aided Mol. Des. 11, 425–445.
Fan, Q.W., Weiss, W.A., 2012. Inhibition of PI3K-Akt-mTOR signaling in glioblastoma by mTORC1/2 inhibitors. Methods Mol. Biol. 821, 349–359.
Faivre, S., Djelloul, S., Raymond, E., 2006. New paradigms in anticancer therapy: targeting multiple signaling pathways with kinase inhibitors. Semin. Oncol. 33, 407–420. Feldman, M.E., Apsel, B., Uotila, A., Loewith, R., Knight, Z.A., Ruggero, D., Shokat, K.M., 2009. Active-site inhibitors of mTOR target rapamycin-resistant outputs of mTORC1 and mTORC2. PLoS Biol. 7, e38. Goodsell, D.S., Morris, G.M., Olson, A.J., 1996. Automated docking of flexible ligands: applications of AutoDock. J. Mol. Recog. 9, 1–5.
PT
Hou, D.X., Kumamoto, T., 2010. Flavonoids as protein kinase inhibitors for cancer
RI
chemoprevention: direct binding and molecular modeling. Antioxid. Redox Signal. 13, 691–719.
SC
Johansson, M.U., Zoete V., Michielin O., Guex N., 2012. Defining and searching for
U
structural motifs using DeepView/Swiss-PdbViewer. BMC Bioinformatics 13,
N
173.
A
Knight, Z.A., Gonzalez, B., Feldman, M.E., Zunder, E.R., Goldenberg, D.D., Williams,
M
O., Loewith, R., Stokoe, D., Balla, A., Toth, B., Balla, T., Weiss, W.A., Williams, R.L., Shokat, K.M., 2006. A pharmacological map of the PI3-K family defines a
D
role for p110alpha in insulin signaling. Cell 125, 733–747.
TE
Lu, J., Zhang, Z., Ni, Z., Shen, H., Tu, Z., Liu, H., Lu, R., 2014. QM/MM–PB/SA
EP
scoring of the interaction strength between Akt kinase and apigenin analogues. Comput. Biol. Chem. 52, 25–33.
CC
Mellinghoff, I.K., Schultz, N., Mischel, P.S., Cloughesy, T.F., 2012. Will kinase
A
inhibitors make it as glioblastoma drugs? Curr. Top. Microbiol. Immunol. 355, 135–169.
Meng, E.C., Shoichet, B.K. and Kuntz, I.D., 1992. Automated docking with grid-based energy evaluation. J. Comput. Chem. 13, 505–524. Morgensztern, D., McLeod, H.L., 2005. PI3K/Akt/mTOR pathway as a target for cancer therapy. Anticancer Drugs 16, 797–803.
Patel, R.Y., Doerksen, R.J., 2010. Protein kinase-inhibitor database: structural variability of and inhibitor interactions with the protein kinase P-loop. J. Proteome Res. 9, 4433–4442. Schymkowitz, J., Rousseau, F., Martins, I.C., Ferkinghoff-Borg, J., Stricher, F., Serrano, L., 2005. Prediction of water and metal binding sites and their affinities by using the Fold-X force field. Proc. Natl. Acad. Sci. USA 102, 10147–10152. Shao, Q.C., Zhang, C.J., Li, J., 2015. Structure-based lead discovery for protein kinase C zeta inhibitor design by exploiting kinase-inhibitor complex crystal structure
PT
data and potential therapeutics for preterm labour. Arch. Pharm. Res. (Epub ahead
RI
of print, DOI: 10.1007/s12272-014-0495-1).
Stupp, R., Mason, W.P., van den Bent, M.J., Weller, M., Fisher, B., Taphoorn, M.J.,
SC
Belanger, K., Brandes, A.A., Marosi, C., Bogdahn, U., Curschmann, J., Janzer,
U
R.C., Ludwin, S.K., Gorlia, T., Allgeier, A., Lacombe, D., Cairncross, J.G.,
N
Eisenhauer, E., Mirimanoff, R.O., 2005. Radiotherapy plus concomitant and
A
adjuvant temozolomide for glioblastoma. N. Engl. J. Med. 352, 987–996.
M
Syed, D.N., Adhami, V.M., Khan, M.I., Mukhtar, H., 2013. Inhibition of Akt/mTOR
995–1001.
D
signaling by the dietary flavonoid fisetin. Anticancer Agents Med. Chem. 13,
TE
Velec, H.F., Gohlke, H., Klebe, G., 2005. DrugScore (CSD) –– knowledge-based
EP
scoring function derived from small molecule crystal data with superior recognition rate of near-native ligand poses and better affinity prediction. J. Med.
CC
Chem. 48, 6296–6303.
A
Wallace, A.C., Laskowski, R.A., Thornton, J.M., 1995. LIGPLOT: a program to generate schematic diagrams of protein-ligand interactions. Protein Eng. 8, 127–134.
Walker, E.H., Pacold, M.E., Perisic, O., Stephens, L., Hawkins, P.T., Wymann, M.P., Williams, R.L., 2000. Structural determinants of phosphoinositide 3-kinase
inhibition by wortmannin, LY294002, quercetin, myricetin, and staurosporine. Mol. Cell. 6, 909–919. Wang, B., Shen, W., Yang, H., Shen, J., Sun, T., 2014. Targeting EGFR mutants with non-cognate kinase inhibitors in non-small cell lung cancer. Med. Chem. Res. 23, 4510–4530. Wang, R., Lai, L., Wang, S., 2002. Further development and validation of empirical scoring functions for structure-based binding affinity prediction. J. Comput. Aided Mol. Des. 16, 11–26.
RI
methodologies and updates. J. Med. Chem. 48, 4111–4119.
PT
Wang, R., Fang, X., Lu, Y., Yang, C.Y., Wang, S., 2005. The PDBbind database:
Word, J.M., Lovell, S.C., Richardson, J.S., Richardson, D.C., 1999. Asparagine and
U
orientation. J. Mol. Biol. 285, 1735–1747.
SC
glutamine: using hydrogen atom contacts in the choice of side-chain amide
N
Zhao, Y, Cheng, T., Wang, R., 2007. Automatic perception of organic molecules based
A
on essential structural information. J. Chem. Inf. Model. 47, 1379–1385.
M
Zhang, C., Liu, S., Zhu, Q., Zhou, Y., 2005. A knowledge-based energy function for
2325–2335.
D
protein-ligand, protein-protein, and protein-DNA complexes. J. Med. Chem. 48,
TE
Zhou, P., Huang, J., Tian, F., 2012. Specific noncovalent interactions at protein-ligand
EP
interface: implications for rational drug design. Curr. Med. Chem. 19, 226–238. Zhou, P. Wang, C., Ren, Y., Yang, C., Tian, F., 2013. Computational peptidology: a
CC
new and promising approach to therapeutic peptide design. Curr. Med. Chem. 20,
A
1985–1996.
Figure legends Figure 1. Scheme of modeling mTOR–inhibitor complex structure from the crystal template of the inhibitor bound with its cognate kinase target. Figure 2. Correlations between the experimental affinity and different scores for kinase–inhibitor complexes. (A) DOCK score (R = -0.38), (B) AutoDock score (R =
-0.46), (C) ChemScore (R = 0.25), (D) X-Score (R = -0.44), (E) DrugScore (R = -0.27) and (F) DFIRE (R = -0.50). Figure 3. Plot of calculated against experimental affinities for kinase–inhibitor complexes with correlation of determination r = 0.786. Figure 4. Histogram distribution of the predicted binding affinities of non-cognate inhibitors to mTOR.
A
CC
EP
TE
D
M
A
N
U
SC
RI
PT
Figure 5. (A) Stereoview of mTOR complex with myricetin, which was modeled from the crystal template of PI3K–myricetin complex structure (PDB: 1E90). (B) The nonbonded interactions across mTOR–myricetin complex. The plot was generated using LIGPLOT program (Wallace et al., 1995).
Table 1. The seven tested inhibitor compounds Inhibitor
Inhibition IC50 (nM) Cognate Affinity to Cognate Non-cognate target mTOR kinasea mTORb
Structure
PD180970
ABL1
12.3
70
560 ± 45
IPA-3
PAK1
11.9
2500
> 1000
INSR
12.6
H N
PT
O H N
N
N N N
NH
NH 2
4 ± 0.6
PDK1
Bay 65-1942
11.4
4
> 1000
12.5
2
28 ± 4
B-Raf
11.7
0.8
> 1000
PDGFR
11.2
1.8
146 ± 15
U
BX-517
SC
O
N
F
2
RI
GS3
M
A
IKKβ
EP
TE
D
Dabrafenib
CC
Crenolanib
obtained from the PDBbind database (Wang et al., 2005). bdetermined in this work.
A
a
Predicted affinity
Inhibition IC50 (nM)a
Fisetin
8.2
74 ± 8
Apigenin
6.5
n.d.
Luteolin
6.6
PT
T able 2. T he flavonoid compounds used to screen for mTOR inhibitors Structure
n.d.
SC
RI
Flavonoid
5.4
n.d.
N
U
Naringenin
7.3
16 ± 2
6.2
n.d.
5.4
n.d.
Delphinidin
4.6
n.d.
Malvidin
5.2
n.d.
A
Myricetin
HO
O
Catechin
M
OH
OH
D
OH
A
CC
EP
Cyanidin
TE
OH
Pelargonidin
4.8
n.d.
Taxifolin
7.1
32 ± 5
6.4
n.d.
7.2
PT
O
HO
O OH
Hesperetin OH
O
19 ± 3
SC
RI
Quercetin
6.5
n.d.
N
U
Kaempferol
5.7
n.d.
7.0
140 ± 24
M
A
Eriodictyol
TE
n.d., not determined.
A
CC
EP
a
D
Flavopiridol
TE
EP
CC
A D
.
PT
RI
SC
U
N
A
M
Figure1
TE
EP
CC
A D
.
PT
RI
SC
U
N
A
M
Figure2
TE
EP
CC
A D
.
PT
RI
SC
U
N
A
M
Figure3
TE
EP
CC
A D
.
PT
RI
SC
U
N
A
M
Figure4
TE
EP
CC
A D
.
PT
RI
SC
U
N
A
M
Figure5