Molecular Mechanisms of the Blockage of Glioblastoma Motility
Jing Xu, Federica Simonelli, Xiaoyun Li, Angelo Spinello, Sara Laporte, Vincent Torre,* and Alessandra Magistrato*
Metrics & More
*sı Supporting Information
⦁ Downloaded via BUTLER UNIV on May 15, 2021 at 21:09:58 (UTC).
⦁ See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.
Brain tumors are incurable and, in the great majority of the cases, fatal neoplasms characterized by a large and potent inﬁltrative growth.1−3 Glioblastoma (GBM; World Health Organization grade IV glioma) is the most common and lethal intrinsic tumor. Unlike other solid tumor cell types, GBM invades the surrounding brain and, in contrast to other kinds of cancers, rarely metastasizes to other organs.1−3 Although several attempts, for instance, using drugs such as bevacizumab or immunotherapies,1−3 have been made to stop and counteract GBM inﬁltration, GBM treatment is still mainly focused and primarily limited to surgical resection followed by concurrent radiation therapy with some chemotherapeutic reagents such temozolomide.4,5 GBM represents one of the most comprehensively genomically characterized cancer types,6,7 leading to recognition of groups of tumors deﬁned by four distinct transcription proﬁles (proneural, neural, classical, and mesenchymal). Mutations leading to the transformation of healthy astrocytes into malignant glioma and/or GBM5 are very diverse, and, indeed, at least, the four diﬀerent transcription proﬁles mentioned above are on the basis of brain tumors.
The molecular mechanisms on the basis of cellular motility are similar in all healthy cells and neurons and in malignant GBM. The process of polymerization of actin ﬁlaments is the main source of cellular motion and protrusion, which is regulated and controlled by several proteins such as the actin- related protein 2/3 complex (Arp2/3), coﬁlin, formin, and molecular motors, such as myosin, dynein, controlling diﬀerent
features of cellular motility.8 A key role in cellular motility and migration is played by the small GTPases, which are present in all migrating cells. Rho family GTPases have distinct and speciﬁc roles in the regulation of growth, maintenance, and retraction of growth cones.9 The mammalian Rho GTPase family currently consists of three subfamilies, Rho (RhoA, RhoB, and RhoC), Rac (Rac1, Rac2, and Rac3), and Cdc42 (Cell Division Cycle-42) (Cdc42Hs and G25K).10 RhoA, Rac1, and Cdc42 are well-studied members of the Rho GTPase family, controlling distinct cytoskeletal elements. Activation of Rac1 stimulates actin polymerization to form lamellipodia,11 Cdc42 induces the polymerization of actin to form ﬁlopodia, and Rho regulates the bundling of actin ﬁlaments into stress ﬁbers and the formation of focal adhesion complexes.12
The Rho family of GTP-binding proteins is activated by a variety of growth factors, cytokines, adhesion molecules, hormones, integrins, G-proteins, and other biologically active substances.13,14 Furthermore, the molecular chaperone tran- sient receptor potential family participates in cytoskeletal- dependent processes through Ca2+-mediated modulation of Rho GTPases,15,16 while Hsp90 can compromise the folding of RhoGTPase family members.17
© XXXX American Chemical Society
J. Chem. Inf. Model. XXXX, XXX, XXX−XXX
Biochemical approaches have shown that Rho GTPases are also involved in crosstalk. Depending on the concentration and localization of these proteins, mammalian cells show diﬀerent morphologies, movements, and behaviors.18 Rho GTPases exert their function via a cyclic mechanism in which they pass from an active guanosine triphosphate (GTP)-bound form to an inactive guanosine diphosphate (GDP)-bound form, after GTP hydrolysis occurs. The cycle is fostered by the GTPase- activating protein (GAP) and by the guanine exchange factor (GEF) proteins, which enhance the exchange of the GTP/ GDP nucleotide during the cycle. Structural studies on small GTPases pinpointed two key functional regions in these proteins, called as switch I (residues 27−37) and switch II (residue 59−73 for both Cdc42 and Rac1), which play critical roles in shaping the GTP-binding pocket and in engaging interactions with GTPases’ regulators (GEFs and GAPs) and protein eﬀectors (such as kinases).18
On the basis of the present manuscript, there is the
biological observation that the cellular motility allowing cells to move, migrate, and inﬁltrate is in essence very similar in all kinds of cells and is primarily based on the orchestration of the cytoskeleton and of a variety of adhesion molecules. The proteins involved in these biological processes and their inhibitors are known. In the present manuscript, we focus on three inhibitors of cellular motility, that is, ML141, EHT 1864, and R-ketorolac. These small molecules have been employed to monitor their ability to reduce cellular motility in GBM. Complementarily, molecular dynamics (MD) simulations have unveiled the binding mode and mechanisms of action of these inhibitors from an atomic-level perspective.
⦁ MATERIALS AND METHODS
⦁ Experimental Methods. 2.1.1. Cell Culture. U87 GBM cells (#89081402, Sigma-Aldrich) were cultured in Dulbecco’s modiﬁed Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum (FBS; Invitrogen, Life Technologies, Gaithersburg, MD) and 1% PenStrep (100 U/ mL penicillin and 100 μg/mL streptomycin; Invitrogen). The GFP-labeled U87 GBM cells were infected with a mix containing a lentiviral vector, LV-GFP. All the cells were cultured in an incubator at 37 °C, 5% CO2, and 95% relative humidity, and the medium was replaced every 3 days. Once 70−80% of conﬂuence had been reached, the cells were replated at a density of 2.5 × 103/cm2.
⦁ Transwell Assay. A total of 5 × 105 U87 GBM cells in
DMEM medium (without FBS) were seeded to the top chambers of 12-well transwell plates (Millipore; 8 μm pore size), and 10% FBS DMEM medium was added to the well. Inhibitors were employed in both the chamber and the well. After incubation for the indicated time, cells in the top of the chamber (non-migrating cells) were removed from the chambers, and cells in the bottom of the chamber (migrating cells) were ﬁxed with 4% PFA for 20 min and stained with 5% crystal violet for 30 min in room temperature. The migrated cells were counted with a microscope.
⦁ Live Cell Imaging. GFP-labeled U87 GBM cells were plated at a density of 8.0 × 104 cells into 35 mm dishes with a glass bottom and cultured for 1 day. Live cell imaging experiments were performed on an epi-ﬂuorescence micro- scope (Nikon Ti2-E) equipped with a chamber incubator and light-emitting diode illumination (λ = 490 nm for). During all imaging experiments, cells were kept at 37 °C, 5% CO2, and 95% humidity. Time lapse images were taken with 50 ms
exposure time, and one image was taken every 2 min. The videos were analyzed using the Fiji plugin TrackMate,19 which allows the selection of regions of interest for every cell and obtaining the average velocities for each cell.
⦁ Computational Methods. 2.2.1. Model Building. The starting conﬁgurations for building the models of the Cdc42 and Rac1 proteins were taken from the crystal structures deposited in the protein data bank (PDB) (id 5CJP and 2YIN for Cdc42 and Rac1, respectively). Conversely, in order to investigate the binding of the drug at the interface with the speciﬁc GEF proteins, we used the crystal structures of Dock9/Cdc42 and Dock2/Rac1 com- plexes deposited in the PDB (PDB id 2WMO and 2YIN, respectively). For each system, the protein structures were prepared, and the protonation state of the ionizable residues was determined using the Schrödinger software suite.
Since the investigated inhibitors are known to act via a non- competitive mechanism (i.e., they do not exert inhibition by competing with the GTP substrate), we looked for possible binding pockets able to host them following an established computational protocol.20 To this end, we used SiteMap21 and FTMap software.22 The search of druggable pockets was carried out on the crystal structure of the proteins and on selected frames of the equilibrated MD trajectories of each protein in the GTP- and GDP-bound form. As a result, we identiﬁed two possible druggable pockets on Cdc42 and one pocket on Rac1. Docking simulations of the experimentally tested inhibitors were then performed to ﬁnd a binding pose of the drugs on these pockets, following an ensemble docking approach. Namely, we used the following as target protein structures: (i) the crystal structures of Rac1 and Cdc42 in the free form and the protein structures in complex with their speciﬁc GEFs, (ii) a representative frame selected by a cluster analysis of the equilibrated part of the MD trajectory; (iii) a visually selected MD trajectory frame of the GDP-bound form of the proteins in which the pockets were able to accommodate the studied inhibitors. Prior to performing docking simulations, each drug was prepared, considering all possible protonation states. As a result, R-ketorolac was considered in its negatively charge form, ML141 was considered in its neutral form, and EHT 1864 was considered in its positively charged form, which are predicted to be the most abundant protonation states at physiological pH. Docking simulations were done using the Glide program of the Schrödinger suite.
⦁ MD Simulations. The topology of the system was built with the Amber 2018 tool tleap using the amber ﬀ14SB force ﬁeld (FF).23 The parameters of the GTP and GDP cofactors and of the tested drugs were built according to the following procedure: each molecule was subjected to structure minimization using the Jaguar program at the density functional theory B3LYP level of theory and the 6-31G** basis set with the Gaussian program.24 Next, electrostatic potential-derived charges were computed according to the Merz-Kolmann partitioning scheme using Gaussian software (Gaussian 09, R. A. G. I., Wallingford CT, 2016) with the same basis set and converted into RESP charges with the resp module of amber tools 2018. For the other FF parameters, the General Amber FF (GAFF) was employed.25
The Mg2+ ion present in the active site was described using the Aqvist parameters.26 The systems were solvated by adding a layer of 10 Å of TIP3P water molecules27 and neutralized with Na+ ions, using the Joung and Cheatman parameters.28 This led to a total number of 44,662 atoms for GDP-bound
Figure 1. Analysis of RAC1 and CDC42 genes in GBM: diﬀerential expression and association with patient survival in TCGA data. (A,C) RAC1
⦁ and CDC42 (C) are signiﬁcantly overexpressed in glioma tumor samples, compared to the matched normal brain tissue. P-value calculated using DESeq2. (B) GBM patients with higher RAC1 expression are associated with higher risk than patients with lower RAC1 expression. (D) Low-grade glioma patients with higher CDC42 expression are associated with higher risk than patients with lower CDC42 expression. P-value calculated using log-rank test. (E) Sketch of the R-ketorolac, ML141, and EHT 1864 molecular structure, highlighting the preferential protonation state at physiological pH as obtained by in silico predictions.
Cdc42; 35,813 atoms for GDP-bound Cdc42 in complex with R-ketorolac; 36,611 atoms for GDP-bound Cdc42 in complex with ML141; 92,293 atoms for GDP bound in complex with Cdc42 and Dock9; 92,310 atoms for GDP-bound Cdc42 in complex with ML141 and Dock9; 34,986 atoms for GDP- bound Rac1; 34,739 atoms for GDP-bound Rac1 in complex with R-ketorolac; 34,415 atoms for GDP-bound Rac1 in complex with EHT 1864; 95,331 atoms for GDP-bound Rac1 in complex with Dock2; and 95,156 atoms for GDP-bound Rac1 in complex with Dock2 + EHT 1864. Overall, 10 diﬀerent systems were simulated and extensively analyzed in this study. The system topology was then converted to the GROMACS format with acpype software.29
A short minimization was run before annealing the system to
300 K. The pressure was equilibrated to 1 atm. In the
simulations with the drugs, after equilibration, the position of the drug in the binding site was restrained for 30 ns, and subsequently, a production run was started removing the constraint. For all simulations, the pressure was kept at the equilibrium value with the Parrinello−Rahman barostat,30 while the temperature was controlled with the velocity rescale thermostat.31 MD simulations were performed using GRO- MACS 2018.232 using an integration time step of 2 fs, and all covalent bonds involving hydrogen atoms were constrained with the LINCS algorithm. The length of the simulations of each system is reported in Table S1. Due to the highly ﬂexible nature of the inhibitor’s binding pockets, to better assess the drug/target interactions, a second MD replica was simulated for each Rho GTPase/drug adducts with the same protocol, thus increasing the cumulative simulation time.
Figure 2. . Rac1 inhibitor EHT 1864 reduces transwell migration of U87 GBM cells. Transwell assay was conducted for three days (ﬁrst, second, and third row) in the presence of 0, 1, 10, and 20 μM EHT 1864 (ﬁrst, second, third, and fourth column).
Figure 3. Inhibitors EHT 1864 (A), ML141 (B), and R-ketorolac (C) inhibit U87 GBM cell migration in transwell assay depending on concentration and duration of treatment.
⦁ Analysis. The root mean square deviation (RMSD), translational and rotational motion; then, the mass-weighted
the root mean square ﬂuctuation (RMSF), the principal component analysis (PCA), and the per residue correlation matrix were derived using both GROMACS 2018.232 and
covariance matrix was computed for the Cα atoms and diagonalized.35 The eigenvectors exhibiting the largest eigenvalues pinpoint the most relevant motions sampled
AMBER 18 programs.33 In particular, GROMACS’s tools were during the MD simulation, also referred to as PCs.36
used to compute the RMSD
rms) and RMSF
The motion along the ﬁrst eigenvector (essential dynamics),
rmsf), while the hydrogen (H) bonds and cross-correlation matrix (CCM) were computed with AMBER’s tool cpptraj. A cluster analysis was performed with GROMACS’s cluster tool using the algorithm described in ref 34. Only Cα atoms of each residue were considered to compute RMSD, RMSF, and the
that is the vector corresponding to the largest eigenvalue, represents the most relevant motion of the system. This was visualized with the VMD program, and arrows highlighting the direction and the amplitude of motion were drawn using the porcupineplot.tcl plugin of the VMD program.
correlation matrix. To compute the RMSD, the whole
⦁ Energetic Analysis. Binding free energies between the
trajectory was used, while all other properties were evaluated on a stable (almost ﬂat rmsd) ﬁnal 100 ns part of the whole trajectory. PCA was performed with the GROMACS 2018 tools (gmx covar, and gmx anaeig). To obtain the principal components (PCs), we applied the following protocol: the trajectory was ﬁrst ﬁtted on the reference structure to remove
proteins and selected ligands (ΔGb) were calculated using the molecular mechanics-generalized born surface area (MM- GBSA) method37 with the Amber18 program. The value of the igb ﬂag was set to 2, and a salt concentration of 0.1 M was used. MM-GBSA calculations were performed on 100 equally distant frames taken from the last 100 ns of the equilibrated
Figure 4. ML141 inhibits U87 GBM cell motility analyzed with live cell imaging. (A,B) Live cell imaging frames at diﬀerent recording time points were merged before and after treatment with 50 μM ML141. Red indicates cells at T = 0 min, and cyan indicates the same cells at T = 60 min. When two channels are co-localized, they appear in white, indicating immobilizing cells. (C,D) Migration trajectories before and after treatment with 50 μM ML141 treatment were reconstructed.
Figure 5. Eﬀect of ML141 on the velocities of U87 GBM cells. (A−C) Mean velocity was signiﬁcantly decreased with 20 mM (B) and 50 mM (C) ML141 and 20 mM ETH 1864. Data are shown as mean ± sem, **p < 0.01. (D,E) Velocity distribution of U87 GBM cells before and after 10 and 20 mM ML141 treatment. (F) Velocity distribution of U87 GBM cells before and after 50 mM ML141 and 20 mM EHT 1864 treatment.
MD trajectory, following a protocol used in previous studies.38 The conformational entropic contribution of the free energy was not considered, as this term usually does not improve the
quality of the results.37 In this analysis, we also used the per- residue decomposition tool in order to dissect how each residue lining the binding site contributes to the binding of the
drug, selecting only those residues whose contribution to the
ΔGb is larger than 1 kcal/mol.
According to The Cancer Genome Atlas (TCGA) (cancer- genome.nih.gov), the analysis of the expression level of proteins involved in cellular motility in the normal tissue and GBM, reported in Figure 1A,C for Rac1 and Cdc42, shows that Rac1 and Cdc42 are both upregulated in primary and recurrent tumors. The overexpression of these proteins is strongly entwined with a negative outcome of the patient (Figure
able to invade into the empty space at later days and that this migration is reduced with an increasing amount of EHT 1864. Considering the number of GBM cells present in the empty space at diﬀerent days obtained with crystal violet and calculated with image J, we found that 1 μM EHT 1864 has a small eﬀect on GBM migration, while 20 μM EHT 1864 almost halved the number of migrating GBM for the GBM cell line U87 both on day 2 and day 3 (Figure 2). We also computed the number of migrating cells per ﬁeld at diﬀerent days and in the presence of a distinct amount of the tested inhibitors. Collected data from at least three diﬀerent experiments show that 50 μM ML141 almost completely
39 blocked migration on day 1 (Figure 3B), while the action of
Cdc42 and Rac1 are primarily involved in cellular motility, and their upregulation leads to a higher inﬁltration ability of malignant GBM to invade the healthy tissue. To monitor the impact of Cdc42 and Rac1 inhibition on the migration and inﬁltration of GBM, we used three inhibitors R-ketorolac, ML141, and EHT 1864 (R-ketorolac and ML141 for Cdc42 and R-ketorolac and EHT 1864 for Rac1), both in live GBM cells and in in silico studies. These inhibitors have rather diﬀerent structures (Figure 1E) both in size and in chemical properties. R-ketorolac contains a carboxylic moiety, which is negatively charged; ML141 is preferentially neutral; and EHT 1864 is positively charged at physiological pH. This may inﬂuence their penetration rate into cells and inside GBM (vide infra). R-ketorolac is known to aﬀect the activity of both Rac1 and Cdc42, with the following half-inhibitory concentrations (IC50): R-ketorolac 0.57 and 1.07 μM for Rac1 and Cdc42, respectively, evaluated using HeLa cells.40,41 Conversely, ML141 and EHT 1864 are exclusive inhibitors of Cdc42
EHT 1864 (Figure 3A) and of R-ketorolac (Figure 3C) was more prominent on day 2. This observation is consistent with the more pronounced ability of ML141 to cross the lipid membrane and act on the interior of cells.
Remarkably, at later days, such as on day 3, we often observed a larger number of cells per ﬁeld (comparing day 3 and day 2 for EHT 1864 in Figure 2) with respect to previous days since the inhibitors used block exclusively the migration and not the replication of GBM. Collected data indicate that the inhibitor concentrations required to block half of the GBM motility is approximately 20, 30, and 50 μM for EHT 1864, ML141, and R-ketorolac, respectively.
We inspected the action of these inhibitors with live cell imaging (Figures 4 and 5) and observed a rapid action of ML141 on GBM motility: the addition of 50 μM halved the motility within some minutes and similar to EHT 1864 exerted a rather fast action. In contrast, we could not detect a fast
action of R-ketorolac. This may be due to the presence of the
(IC50 of 2.1 μM)45 and Rac1 (IC50 1−5 μM),40 respectively.
carboxylic group, which limits its ability to cross the lipid
The latter values refer to the biochemical assay performed on the puriﬁed proteins. In the next section, we will examine the eﬀect of these inhibitors on the migration of GBM, where more complex conditions hold.
⦁ Action of the Cdc42 and Rac1 Inhibitors in GBM Migration. We analyzed the eﬀect of the small-molecule inhibitors on the GBM cell line U87, which is very often used as a model for understanding properties of high-grade GBM. Migration and inﬁltration were studied by the transwell assay (Figures 2 and 3) and by live cell imaging (Figures 4 and 5). The transwell migration assay is a standard method of measuring cell movement through an empty space, based on the use of a hollow plastic chamber, which is sealed at one end with a porous membrane and suspended over a larger well containing speciﬁc medium. Migrating cells, that is, inﬁltrating GBM, are plated inside the chamber and allowed to migrate through the pores to the other side of the membrane. Migrated cells are ﬁxed after a given time, stained, and counted. Live cell imaging methods visualize individual cells in a dish and take an image every minute, and in this way, can follow the motility of individual cells. These two methods are complementary: live cell imaging requires substantial illumination of the cells under investigation and cannot be used for experiments lasting several days, while the transwell assay can be used over several days with only minor side eﬀects on the investigated cells (Figure 2). The elongated and fusiform shapes in Figure 2 are the proﬁle of the migrating U87 GBM under control conditions (ﬁrst column) and in the presence of 1, 10, and 20 μM EHT 1864 (second, third, and fourth column). The assay was conducted for three days (ﬁrst, second, and third row in Figure 2). Visual inspection shows that more GBM cells are
membrane. The inhibitory action of R-ketorolac is best seen with the transwell assay and requires at least 24 h.
Live cell imaging allows the tracking over time of an individual migrating GBM and therefore provides an estimate of the mean velocity of migrating cells. As GBM cells replicate very eﬃciently, both the transwell assay and live cell imaging are not exempt from limitations: in the presence of a high replication activity, the transwell assay can underestimate the eﬀect of an inhibitor, and the live cell imaging tracking can be confused during a mitosis event. Nevertheless, the combination of the two assays provides a reliable characterization of the eﬀect of the used inhibitors on GBM motility.
⦁ Molecular Mechanism of Cdc42 and Rac1 Inhibition from All-Atom Simulations. According to the experiments detailed above, the tested inhibitors, known to target the Rac1 and Cdc42 proteins, were able to reduce the migration and/or the inﬁltration propensity of the U87 cell lines. As such, we employed docking and all-atom explicitly solvated MD simulations to unravel the molecular mechanism of inhibition and to dissect similarities and diﬀerences in hampering the function of the targeted Rho GTPases.
⦁ Identiﬁcation of the Inhibitors Binding Pose via
Docking and MD Simulations. Since all drugs investigated above are known to exert a non-competitive inhibition mechanism (i.e., they inhibit the Rho GTPase without competing with the protein’s cofactor, the GTP), we have initially identiﬁed druggable cavities, distinct from the GTP binding site, possibly able to bind the studied inhibitors. This search was done using site-detecting algorithms, considering both the GTP- and GDP-bound forms of each protein. As a result, we identiﬁed druggable cavities only in the GDP-bound
form of both proteins. In particular, we detected two cavities (sites 1 and 2) on Cdc42 and only one cavity (site 1) on Rac1 (Figure S1). In order to assess the druggability of these pockets, we docked the investigated inhibitors on site 1, which ﬂanks the GTP-binding cavity. Remarkably, only the R- enantiomers of ketorolac and of ML141 could be docked in this site, consistent with experimental ﬁndings, showing that only R-ketorolac can inhibit the activity of Rho GTPases.40 To the best of our knowledge, the enantioselectivity of ML141 has never been previously investigated; thus, we tested both the R- and S-enantiomers of this drug, but only the R-enantiomer was found to stably bind to site 1.
The binding stability of the drugs inside the allosteric cavity
was monitored by performing 100−400 ns-long MD simulations40 for each drug/GTPase adduct (Tables S1, S2, Figure S2). Site 2, identiﬁed exclusively for Cdc42, is rather small and ﬂexible. Thus, all molecules docked in this site rapidly dissociated (within 30 ns of MD simulation). This ﬁrst set of simulations unprecedentedly allowed predicting the binding pose of R-ketorolac and R-ML141 (no pose for the S- enantiomers was found) on Cdc42 and that of R-ketorolac and EHT 1864 on Rac1 (Figure 6). In the following, we perform
Our set of simulations predicts that the carboxylic moiety of R-ketorolac coordinates the Mg2+ ion. This coordination may weaken the interaction of the metal ion with the GDP phosphates, thus facilitating the release of GDP from Cdc42 and/or preventing the binding of a new GTP molecule in an optimal position to undergo the next GTP hydrolysis cycle.40 Thus, consistent with experimental evidences, our simulations pinpoint R-ketorolac as a GTP-binding inhibitor.42 Besides coordinating the Mg2+ ion with its carboxylic group, in our simulations, the dihydro-1H-pyrrolizine moiety of R-ketorolac establishes hydrophobic interactions with Val36. Moreover, the Phe37, Phe56, and Tyr64 residues appear to be creating a ﬂexible hydrophobic pocket where the drug binds. Addition- ally, the aromatic ring of R-ketorolac lies in the vicinity of one of the functionally important regions of Cdc42, named as switch II (Figure 6), where the GEF proteins bind, as revealed in crystallographic studies.43 Therefore, the binding of R- ketorolac to Cdc42 may also interfere with the binding of the GEF proteins (Figure 6).
The relevance of the carboxylic moiety observed in our simulations is conﬁrmed by the activity of other R-enantiomers of non-steroidal anti-inﬂammatory drugs (NSAIDs).41 Indeed,
R-naproxen was observed to be active on both Rac1 and
Figure 6. Binding pose of R-ketorolac (R-keto, A) and ML141 (B) on Cdc42, R-keto (C), and EHT 1864 (D) on Rac1. Cdc42 and Rac1 are shown in gray and blue new cartoons, respectively; switches I (from residue 27 to residue 37) and II (from residue 59 to residue 73) are shown in lime and mauve, respectively. Mg2+ ions are pictured in orange van der Waals spheres, while the GDP and the inhibitors are shown as licorice and colored by the atom name. In panel C, the hydrogen bond between R-keto and Tyr32 (depicted as green licorice) of Rac1 is displayed.
systematic analysis to unravel the molecular mechanism of inhibition exerted by the used inhibitors on the small Rho GTPase subject of this study. In addition, a comparison with inactive structural analogues of these drugs better allows us to pinpoint the key structural traits stabilizing the binding of the active drugs to the target protein, being possibly responsible for their eﬃcacy.
⦁ Inhibition Mechanism of Cdc42. We initially assessed the mechanism of the two Cdc42 inhibitors (Figure 6A,B). Diﬀerent non-competitive inhibitory mechanisms may be operative for small Rho GTPases depending on whether the small molecules exert their action by interfering with GTP or with GEF binding.40
Cdc42 proteins in previous experimental studies.41 Conversely, R-ketoprofen, lacking the extended dihydro-1H-pyrrolizine cyclic moiety of R-ketorolac, thus being characterized by lower conformational restraints, is inactive on Cdc4242 in spite of its ability to coordinate the Mg2+ ion with its carboxylic moiety as predicted by docking simulations (Figures S3 and S4). A decrease in the size of the cyclic moiety ﬂanking the carboxylic group, such as in acetylsalicylic acid, also results in eﬃcacy loss (Figures S3 and S4).41 As such, the larger rigidity of the carboxylic group induced by the presence of an extended cyclic moiety, such as the dihydro-1H-pyrrolizine moiety, of R- ketorolac appears to be a critical structural trait to confer the binding stability and activity of NSAID to small Rho GTPases.41
At variance with R-ketorolac, we identiﬁed two possible binding poses of ML141 from combined docking and MD simulations. In the ﬁrst, ML141 protrudes toward the Mg2+ ion with its sulfonamide moiety (Figure 6B). These interactions with the Mg2+ ion and with the pocket ﬂanking the GDP- binding site conﬁrm its interference with the binding of the GTP cofactor, as suggested experimentally.40 Instead, the methoxyphenyl ring inserts below switch I. As a result of ML141 binding, the guanine moiety of GDP loses its H- bonding interaction with Asp118, which usually anchors the GTP/GDP cofactor in its binding pocket, being therefore destabilized.44 In this binding pose, the drug is stabilized with Val36 and Tyr64 (Table S3).
During the MD simulations, a second ML141 binding pose is also observed. In this pose, the sulfate moiety of the drug interacts with the metal ion, triggering a diﬀerent relative orientation of the phenyl and methoxy phenyl rings (Figure S5). In this second binding pose, the benzyl ring heads toward the small cavity lined by switch II, approaching the GEF- binding pocket near the Tyr64, Leu67, and Phe56 residues (Figure 6) and thus being stabilized by hydrophobic interactions (Table S3).
In order to dissect the key structural elements underlying ML141 binding stability and eﬃcacy, we also considered celecoxib and valdecoxib, two inactive structural analogues.41,44 A superposition of the benzenesulfonamide moiety of the
Figure 7. Per-residue CCM of GDP-bound Cdc42 (A). Diﬀerence between the per-residue CCM of GDP-bound Cdc42 in complex with R- ketorolac (B) and ML141 (C) and the CCM of GDP-bound Cdc42 with no drug. Diﬀerence between the per-residue CCM of GDP-bound Cdc42 upon binding of Dock9 and ML141 and the CCM of GDP-bound Cdc42 alone (D). Pearson’s cross-correlation coeﬃcients vary from −1 (anticorrelated motion, blue) to +1 (correlated motion, red). Horizontal and vertical lines deﬁne diﬀerent regions of the protein: β1 (residues 1− 10), α1 (residues−26), β2 (residues 38−47), β3 (residues 49−58), β4 (residues 77−84), α3 (residues 87−106), β5 (residues 110−116), αi (residues 123−132), α4 (residues 139−150), β6 (residues 153−158), and α5 (residues 165−177). Domain partitioning of Cdc42 is shown in panels E and F. Residues corresponding to switch I and II are highlighted by red lines.
inactive analogues to that of the ﬁrst binding pose of ML141 reveals that their phenyl rings establish hydrophobic interactions with the small cavity lined by switch II, therefore missing the stabilizing interaction due to the insertion of the methoxybenzyl ring of ML141 below switch I (Figures 6, S3 and S4). Conversely, when a superposition of celecoxib and valdecoxib is performed on the second binding pose of ML141, celecoxib inserts the triﬂuoromethyl substituent in the small cavity lined by switch II, while the methyl benzene ring is solvent-exposed. Instead, in valdecoxib, due to the diﬀerent orientation of the oxazol substituents, neither the methyl nor phenyl substituents establish relevant interactions with Cdc42 (Figure S5). This provides a rationale to the experimentally
We next inspected the impact of R-ketorolac and ML141, in their ﬁrst and most relevant binding pose, on the structural and functional properties of the drug/Cdc42 adduct as compared to that of the Cdc42 protein per se. Ostensibly, the RMSFs (Figure S6) reveal that switches I and II are the regions of Cdc42 most aﬀected by R-ketorolac binding. Indeed, the binding of R-ketorolac increases their ﬂexibility. In contrast, ML141 only slightly aﬀects the switch I region. In order to understand how drug binding could impact the internal motion of the protein, we also computed the per-residue CCM. Positive values of the CCM indicate dynamically coupled regions, associated to a lockstep motion of the protein, while negative regions indicate the negatively correlated motion with
observed lack of activity of Cdc42.41,44
the corresponding parts of the protein moving in opposite directions.34 The CCM of the GDP-bound Cdc42 shows
Figure 8. Porcupine plot representing the essential dynamics of Cdc42 in the GDP-bound form (A) in complex with R-ketorolac (B) and with ML141 (C). Cdc42 is depicted in gray new cartoons, with switch I and II highlighted in lime and mauve, respectively. The arrows indicate the direction of the motion; their length and color (from blue to red) are representative of the motion amplitude.
dynamical coupling of two functionally relevant regions, switch I and switch II, with switch I moving in a lockstep motion with α1, α4, β6, and β1 while being negatively correlated with αi (Figure 7A). Instead, switch II results to be negatively correlated with switch I, β1, and αi. Consistently, the essential dynamics (Figure 8) of Cdc42, capturing the most relevant slow vibrational motion of the protein from an MD trajectory, reveals an opening/closing motion of switch I, which is most likely instrumental to the load/release of the GTP/GDP cofactor and to engage the interactions with the variety of protein eﬀectors that mediate Rho GTPase signaling. To monitor the impact of R-ketorolac and ML141 binding on the internal motion of Cdc42, we also computed the CCM of the drug-bound protein. This CCM, plotted as a diﬀerence with respect to that of the undrugged Cdc42 (Figures 7B,C), reveals that both R-ketorolac and ML141 dampen the internal motion of the Cdc42 (Figure 8), thus preventing the GTP/GDP exchange and, in turn, Rho GTPase activation.
The computed binding free energies (ΔGb) disclose that the binding of R-ketorolac to Cdc42 (ΔGb of −154 ± 5 kcal/mol) is energetically favored as compared to that of ML141 (ΔGb of
−45 ± 6 kcal/mol). The electrostatic interactions between the negatively charged carboxyl moiety of R-ketorolac and the Mg2+ ion markedly contribute to the ΔGb. In addition, residues Lys16 and Val36 stabilize the drug binding via electrostatic and hydrophobic interactions, respectively. We remark that the computed ΔGb is calculated considering a FF-based description of the system and neglects the electronic rearrangements of the charge density induced by the Mg2+ ions (charge transfer and polarization eﬀects) to the coordinating ligands.45 As a result, the calculated ΔGb is most likely overestimated. In addition to Lys16, also, Val36, Ty40, Tyr64, and Leu67 contribute to the ΔGb of ML141 by establishing hydrophobic interactions (Table S3).
As a further check, we also inspected whether ML141 could bind at the Cdc42/GEF interface. Indeed, small molecules able to stabilize the GEF/GTPase adduct may permanently inactivate the GTPase cycle. Among the GEF proteins, known to bind to Cdc42, we selected Dock9 since an X-ray structure of the Cdc42/Dock9 adduct is available in the PDB database. ML141 was docked in a cavity identiﬁed at the interface of the two proteins, and the binding pose of ML141 resulted to be stable in MD simulations (Figures S2 and S7). The binding of ML141 had a limited impact on the Cdc42/ Dock9 ﬂexibility (i.e., does not markedly aﬀect the RMSF of the two switch regions, which, in the adduct, are locked by Dock9 binding, Figure S8). Upon ML141 binding, an increase in positive correlation of the two Cdc42 switches is observed (Figures 7D and S9), underlining a partial recovery of Cdc42 internal motion, which may reduce the stability of the Dock9/ Cdc42 adduct. Consistently, the ΔGb of ML141 at the Dock9/
Cdc42 interface is smaller than that in the Cdc42 protein per se (ΔGb = −28 ± 3 kcal/mol), being stabilized by hydrophobic and electrostatic interactions with residues Gln342 and Glu403, respectively, from Dock9 and by hydrophobic interactions with Leu67 from Cdc42 (Table S3). This suggests that ML141 most likely binds to site 1 of the Cdc42 protein.
⦁ Inhibition Mechanism of Rac1. The same simulation protocol was also employed to elucidate the binding mode and the inhibition mechanism exerted by R-ketorolac and EHT 1864 to Rac1. Also, in this case, docking and MD simulations predict that both drugs bind nearby the GDP-binding site. R- ketorolac coordinates the Mg2+ ion with its carboxylic moiety, similar to what is observed in Cdc42, while inserting between the GDP molecule and switch I (Figures 6 and ⦁ S10), as is also the case for EHT 1864. In spite of the structural similarity of Cdc42 and Rac1, the pocket lined by switch II is smaller in the latter due to the presence of Trp56 as compared to Phe56 in Cdc42.
In Rac1, the pyrrolidine moiety of R-ketorolac interacts hydrophobically with Val36, while the phenyl moiety interacts with Pro29, Leu160, and Ile21. Its carbonyl moiety, instead, establishes a H-bonding interaction with Tyr32.
In order to assess the importance of the interactions observed, we compared the binding mode of R-ketorolac to that of two inactive analogues, R-ketoprofen and acetylsalicylic acid.41 Docking simulations of R-ketoprofene reveal that in spite of the coordination of Mg2+ ions with the drug’s carboxylic moiety, the larger ﬂexibility of its propionic moiety induces a diﬀerent binding mode as compared to R-ketorolac (Figures S3 and S10). Conversely, a reduction in the size of the conjugated moiety, such as in acetylsalicylic acid, allows the molecule to ﬁt in a orientation diﬀerent from R-ketorolac, coordinating the Mg2+ ion with both carboxylic groups (Figure S10).
Furthermore, EHT 1864 also inserts itself between switch I and GDP. In this pocket, EHT 1864 is stabilized by hydrophobic and π-stacking interactions between its purine ring and Cys18, Ile21, and Tyr32 (Table S3) and the GDP cofactor, respectively. In addition, the amine group of the morpholine ring H-bonds with Asp38, and this ring is lined by the hydrophobic residues Ala59 and Leu67 and by Tyr64.
We also performed a comparison of ETH 1864 with its inactive structural analogue EHT 8560, which diﬀers from the ﬁrst only by the presence of a second amine moiety on the morpholine ring (Figure S3).46 EHT 8560 can exist in two possible protonation states with one of the two amine moieties predicted with equal probability to be positively charged.
If the tertiary amine is positively charged, as in EHT 1864, the presence of a second nitrogen, in the place of the oxygen of the morpholine ring, may reduce the acidity of the tertiary amine, thus weakening the H-bond with Asp38. Conversely, if
Figure 9. Porcupine plot representing the essential dynamics of Rac1 in the free GDP-bound form (A), in complex with R-ketorolac (B) and in complex with EHT 1864 (C). Cdc42 is depicted in blue new cartoons, with switch I and II highlighted in lime and mauve, respectively. The arrows indicate the direction of the motion; their length and color (from blue to red) are representative of the motion amplitude.
Figure 10. Per-residue correlation matrix (CCM) for the GDP-bound form of Rac1 (A). Diﬀerence between the per-residue CCM of GDP-bound Rac1 in complex with R-ketorolac (B) and EHT 1864 (C) with respect to that of GDP-bound Rac1 with no drug. Diﬀerence between the per- residue CCM of GDP-bound Rac1 in complex with Dock2 and EHT 1864 and GDP-bound Rac1 (D). Pearson’s cross correlation coeﬃcients vary from −1 (blue, anti-correlated motion) to +1 (red, correlated motion). Horizontal and vertical lines deﬁne diﬀerent regions of the protein: β1 (residues 1−10), α1 (residues 16−26), β2 (residues 38−47), β3 (residues 49−58), β4 (residues 77−84), α3 (residues 87−106), β5 (residues 110−116), αi (residues 123−132), α4 (residues 139−150), β6 (residues 153−158), and α5 (residues 165−177). Domain partitioning of Rac1 is shown in panels E and F. Switch I and II are highlighted by red lines under the matrices.
the secondary amine is positively charged, the tertiary amine lacks the proton and thus the possibility of establishing a H- bond with Asp38, reducing even further the aﬃnity of the ligand for Rac1 and thus providing a rationale to the experimentally observed lack of activity.46
Consistent with the observed binding poses, both drugs
(Figure 6C,D) principally perturb switch I of Rac1 (Figure
S6). Interestingly, the binding of R-ketorolac locks the internal dynamics of the protein, hampering the opening/closing motion of the switch loops (Figure 9), whereas EHT 1864 does not interfere with it. The CCM of both systems also highlights marked diﬀerences. While R-ketorolac introduces small changes in the per-residue CCM (Figure 10), EHT 1864 largely perturbs the protein dynamics, increasing the
negatively/positively coupled motions of switch I and II, respectively (Figure 10).
The calculated ΔGbs reveal that R-ketorolac binds more strongly than EHT 1864 with Rac1 (ΔGb = −179 ± 15 kcal/ mol and ΔGb = −43 ± 6 kcal/mol for R-ketorolac and EHT 1864, respectively) and that R-ketorolac has a larger binding aﬃnity for Rac1 as compared to Cdc42, consistent with its higher inhibitory potency (lower IC50).40 Similar to what is observed for Cdc42, the binding pose of R-ketorolac in Rac1 is mostly due to the electrostatic interactions of its carboxyl
eﬃcacy of the compounds, as compared to in vitro tests and migration characteristics in other cell line types, also suggests that other pathways may be relevant to stop the inﬁltration and migration propensities of the GBM.
4.2. Switches and Binding and Docking Sites in Cdc42 and Rac1. The set of docking and MD simulations performed supply a molecular basis for understanding the inhibition mechanism exerted by the GTPase inhibitors tested in this study. R-ketorolac, bearing a carboxylic moiety, coordinates the Mg2+ ions, in line with the high aﬃnity of
moiety with the Mg2+ ions, being here further stabilized by a H-bond between the hydroxyl group of Tyr32 and its carbonyl oxygen (Table S3). In contrast, EHT 1864 mostly establishes hydrophobic interactions with the binding pocket and electrostatic interactions between its amine moiety and Asp38 (Table S3).
As for Cdc42, we also monitored if EHT 1864 could bind at the interface of Rac1 and a speciﬁc GEF. In this case, we selected Dock2 for which a crystal structure in complex with Rac1 was available in the PDB database. Docking of EHT 1864 followed by MD simulations conﬁrmed the presence of a druggable pocket at the Rac1/Dock2 interface (Figures S7 and S11). The binding of EHT 1864 at this site slightly increases the ﬂexibility of switch I (Figure S8). Consistently, the CCM of the Rac1/EHT 1864 adduct registers an increase in the positive correlation (Figures 10 and S12) of this critical region. Hence, the binding of this drug at the Rac1/Dock2 interface is not eﬀective in blocking functional dynamics of protein but rather destabilizes the Rac1/Dock2 adduct. This is further conﬁrmed by a decrease in the ΔGb (−28 ± 6 kcal/mol) with respect to that of EHT 1864 binding to site 1 of Rac1. Consistent with its suggested activity as a nucleotide-displacing inhibitor,47 EHT 1864 most likely targets Rac1 rather than the Rac1/Dock2 adduct.
The present study illustrates the eﬀect of three inhibitors (ketorolac, ML141, and EHT 1864) on the migration and motility of GBM, and a detailed analysis based on all-atom simulations discloses their binding pose and the inhibition mechanism on the selected small GTPases, Cdc42 and Rac1, involved in cellular motility. The complementary use of experimental and theoretical approaches provides better insights and clearly dissects similarities and diﬀerences among the investigated inhibitors.
⦁ In Vitro and in Silico Comparison. In the experimental analysis, as shown in Figures 2−5, the inhibitors are added to the extracellular medium bathing the GBM. Therefore, these molecules must cross the cellular membrane before interacting with Rac1 and Cdc42. As a result, it is expected that the concentration of the inhibitors halving the migration K1/2migration to be diﬀerent from the IC50 concen- tration, inhibiting by half the GTP hydrolysis of Rac1 and Cdc42 under in vitro conditions. Indeed, from the transwell experiments performed in the U87 GBM cell line, we have estimated K1/2migration to be approximately 20, 30, and 50 μM for EHT 1864, ML141, and R-ketorolac, respectively. For EHT 1864, the corresponding values obtained in vitro are 1−547 and
20 μM on MDA-MB-231 and MCF-7 breast cancer cells,
respectively.48 For ML141, 2 μM under in vitro conditions,49 3 μM was used for 50% reduction on ovarian cancer cell migration.38 R-ketorolac against Rac1 and Cdc42 in HeLa cells has values of 0.57 and 1.07 μM,41 respectively. The reduced
carboxylic moieties toward Mg2+ ions observed in protein enzymes.50,51
The binding of the R-ketorolac carboxylic moiety to Mg2+ may reduce the binding strength of GDP phosphate toward the Mg2+ ion since the latter can tolerate the binding of a limited number of negatively charged ligands.44
This may facilitate the dissociation of GDP and/or possibly inactivate the small Rho GTPase function by impeding GDP/ GTP exchange. Nevertheless, the binding of R-ketorolac is slightly diﬀerent in Cdc42 and Rac1, where in the former, both switch I and II regions are lined by the drug, thus hampering the opening/closing motion of the cavity and possibly aﬀecting also the interactions with the GEF proteins. Conversely, when binding to Rac1, R-ketorolac exclusively aﬀects switch I, which is less involved in the GEF interaction/stabilization. The discussed diﬀerence in the binding poses is also conﬁrmed by the predicted binding mode of ML141 and EHT 1864, the ﬁrst heading toward the switch II cavity and the second being also lined by switch I. While ML141 engages interactions with site 1 and hampers the internal motion of switches I/II, necessary for GTP/GDP upload/release, EHT 1864 does not completely freeze this movement in spite of its larger size.
Rho GTPases being at the crossroads of an intricate signaling interactome and their action being co-adjuvated by several interacting partners, we have attempted at predicting if/ how the binding of the studied inhibitors could aﬀect the interactions with diﬀerent proteins from their interactome. Notably, by superimposing the drug-bound form of Rac1/ Cdc42 proteins with selected Rho GTPase partners, for which a crystal structure in complex with Rac1/Cdc42 is available in the PDB database, it emerges that R-ketorolac and EHT 1864 may more markedly aﬀect the binding of speciﬁc Rac1 protein partners such as plexin B1,52 P-Rex1, and epithelial cell- transforming protein 2 (ETC2),53 which are involved in GBM migration and invasive propensity of cancer cells while aﬀecting to a lower extent the interaction with typical GEF proteins such as Dock2 (Figures S13−S16). Conversely, the binding of both R-ketorolac and ML141 to Cdc42 appears to potentially impact even the binding of the GEFs (Dock9) (Figures S17 and S18).
In summary, migration and motility in cells are regulated by a complex and highly interconnected network of proteins. The pharmacological inhibition of the proteins entailed in this network may be of therapeutic beneﬁt.47 As an example, the knockdown of Rac1 in melanoma cells depresses the formation of invadopodia48 and also decreases Rac1 capacity to promote cancer metastasis. The already-approved drug ketorolac and in particular its R-enantiomer inhibit Rac1 and Cdc42 in ovarian cancer49 and potentially contributes to the observed survival beneﬁt. This observation suggests using the same drug for treating glioma/GBM at its early stages also in the case of brain cancer. Ketorolac is a drug that has already been approved by the US Food and Drug Administration (FDA) in 1989, while
ML141 and EHT 1864 are not yet approved. Therefore, dissecting the molecular mechanism and establishing the working concentration of these three inhibitors in brain cancer will beneﬁt their clinical promotion and use.
The outcomes of this study remark the therapeutic potential of abrogating the cellular mobility in inﬁltrative tumor types such as GBM and provide a fundamental advancement in understanding the mechanism of small-molecule inhibitors of Rho GTPases. This information is of pivotal importance to devise novel drug candidates that are able to more eﬀectively block cellular motility in GBM and in other inﬁltrative tumor types.
*sı Supporting Information
The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jcim.1c00279.
Length (ns) of the molecular dynamics simulations, binding free energies, putative binding sites, and docking poses of inactive structural analogues (PDF)
⦁ AUTHOR INFORMATION
Vincent Torre − International School for Advanced Studies (SISSA/ISAS), Trieste 34136, Italy; Email: [email protected]
Alessandra Magistrato − National Research Council of Italy - Institute of Materials (CNR-IOM) c/o SISSA, Trieste 34136, Italy; orcid.org/0000-0002-2003-1985;
Email: [email protected]
Jing Xu − International School for Advanced Studies (SISSA/ ISAS), Trieste 34136, Italy; Institute for Systems Medicine, Suzhou, Jiangsu 215123, P. R. China
Federica Simonelli − National Research Council of Italy - Institute of Materials (CNR-IOM) c/o SISSA, Trieste 34136, Italy
Xiaoyun Li − International School for Advanced Studies (SISSA/ISAS), Trieste 34136, Italy
Angelo Spinello − National Research Council of Italy - Institute of Materials (CNR-IOM) c/o SISSA, Trieste 34136, Italy; orcid.org/0000-0002-8387-8956
Sara Laporte − National Research Council of Italy - Institute of Materials (CNR-IOM) c/o SISSA, Trieste 34136, Italy;
Complete contact information is available at: https://pubs.acs.org/10.1021/acs.jcim.1c00279
The authors declare no competing ﬁnancial interest.
Data and Software Availability Data. Data-concerning top- ologies, input ﬁles, and PDB ﬁles of the most representative structure of each drug/protein adduct are available at https:// d r i v e . g o o g l e . c o m / d r i v e / f o l d e r s / 12qkMWgkHzLN4JQmMeTQ7DTkW7TuSg2c6?usp= sharing. The Fiji plugin TrackMate is available at the https:// imagej.net/TrackMate website. The GROMACS 2018.2 program to run molecular dynamics simulations and Amber tools 18 to build Amber topologies and perform analysis are freely available in their respective web sites (https://www. gromacs.org/Downloads and https://ambermd.org/ AmberTools.php). FTPMap software can be used at the web
server https://ftmap.bu.edu/login.php. The Gaussian program used for the parametrization of the inhibitor can be purchased at the http://gaussian.com/ website. Glide and SiteMap software can be purchased from the Schroedinger company at https://www.schrodinger.com/products/glide and https:// www.schrodinger.com/products/sitemap, respectively. A demo license is usually provided by the vendor upon request.
This research was funded by the project “AgainstbRain cancEr: ﬁnding personalized therapies with in silico and in vitro strategies” (ARES)CUP:D93D19000020007 POR FESR 2014
2020-1.3.b-Friuli Venezia Giulia. A.S. was supported by a FIRC-AIRC “Mario e Valeria Rindi” fellowship for Italy. A.M. thanks the ﬁnancial support of the Italian Association for Cancer Research (AIRC) (IG grant 24514).
⦁ Weller, M.; Roth, P.; Preusser, M.; Wick, W.; Reardon, D. A.; Platten, M.; Sampson, J. H. ⦁ Vaccine-based Immunotherapeutic ⦁ Approaches to Gliomas and Beyond. Nat. Rev. Neurol. 2017, 13, 363−374.
⦁ Bagley, S. J.; Desai, A. S.; Linette, G. P.; June, C. H.; O’Rourke,
D. M. CAR T-cell Therapy for Glioblastoma: Recent Clinical Advances and Future Challenges. Neuro-Oncology 2018, 20, 1429− 1438.
⦁ Diaz, R. J.; Ali, S.; Qadir, M. G.; De La Fuente, M. I.; Ivan, M. E.; Komotar, R. J. ⦁ The Role of Bevacizumab in the Treatment ⦁ of ⦁ Glioblastoma. J. Neuro Oncol. 2017, 133, 455−467.
⦁ Thomas, A. A.; Brennan, C. W.; DeAngelis, L. M.; Omuro, A. M. ⦁ Emerging⦁ ⦁ Therapies⦁ ⦁ for⦁ ⦁ Glioblastoma.⦁ JAMA Neurol. 2014, 71, 1437− 1444.
⦁ Stupp, R.; Mason, W. P.; van den Bent, M. J.; Weller, M.; Fisher, B.; Taphoorn, M. J. B.; Belanger, K.; Brandes, A. A.; Marosi, C.; Bogdahn, U.; Curschmann, J.; Janzer, R. C.; Ludwin, S. K.; Gorlia, T.; Allgeier, A.; Lacombe, D.; Cairncross, J. G.; Eisenhauer, E.; Mirimanoff, R. O. ⦁ Radiotherapy Plus Concomitant and Adjuvant ⦁ Temozolomide⦁ ⦁ for⦁ ⦁ Glioblastoma. N. Engl. J. Med. 2005, 352, 987− 996.
⦁ Bhat, K. P. L.; Balasubramaniyan, V.; Vaillant, B.; Ezhilarasan, R.; Hummelink, K.; Hollingsworth, F.; Wani, K.; Heathcock, L.; James, J. D.; Goodman, L. D.; Conroy, S.; Long, L.; Lelic, N.; Wang, S.; Gumin, J.; Raj, D.; Kodama, Y.; Raghunathan, A.; Olar, A.; Joshi, K.; Pelloski, C. E.; Heimberger, A.; Kim, S. H.; Cahill, D. P.; Rao, G.; Den Dunnen, W. F. A.; Boddeke, H. W. G. M.; Phillips, H. S.; Nakano, I.; Lang, F. F.; Colman, H.; Sulman, E. P.; Aldape, K. ⦁ Mesenchymal ⦁ Differentiation⦁ ⦁ Mediated⦁ ⦁ by⦁ ⦁ NF-⦁ κ⦁ B⦁ ⦁ Promotes⦁ ⦁ Radiation⦁ ⦁ Resistance⦁ ⦁ in ⦁ Glioblastoma. Cancer Cell 2013, 24, 331−346.
⦁ Behnan, J.; Finocchiaro, G.; Hanna, G. ⦁ The⦁ ⦁ Landscape⦁ ⦁ of⦁ ⦁ the ⦁ Mesenchymal⦁ ⦁ Signature⦁ ⦁ in⦁ ⦁ Brain⦁ ⦁ Tumours. Brain 2019, 142, 847− 866.
⦁ Pak, C. W.; Flynn, K. C.; Bamburg, J. R. ⦁ Actin-binding⦁ ⦁ Proteins ⦁ Take⦁ ⦁ the⦁ ⦁ Reins⦁ ⦁ in⦁ ⦁ Growth⦁ ⦁ Cones.⦁ Nat. Rev. Neurosci. 2008, 9, 136− 147.
⦁ Stankiewicz, T. R.; Linseman, D. A. ⦁ Rho Family GTPases: ⦁ Key ⦁ Players in Neuronal Development, Neuronal Survival, and ⦁ Neuro- ⦁ degeneration. Front. Cell. Neurosci. 2014, 8, 314.
⦁ Sayyad, W. A.; Fabris, P.; Torre, V. ⦁ The⦁ ⦁ Role⦁ ⦁ of⦁ ⦁ Rac1⦁ ⦁ in⦁ ⦁ the ⦁ Growth⦁ ⦁ Cone⦁ ⦁ Dynamics⦁ ⦁ and⦁ ⦁ Force⦁ ⦁ Generation⦁ ⦁ of⦁ ⦁ DRG⦁ ⦁ Neurons. PLoS One 2016, 11, No. e0146842.
⦁ Ridley, A. J.; Paterson, H. F.; Johnston, C. L.; Diekmann, D.; Hall, A. ⦁ The Small GTP-Binding Protein Rac Regulates Growth ⦁ Factor-induced Membrane Ruffling. Cell 1992, 70, 401−410.
⦁ Nobes, C. D.; Hall, A. ⦁ Rho,⦁ ⦁ Rac,⦁ ⦁ and⦁ ⦁ Cdc42⦁ ⦁ GTPases⦁ ⦁ Regulate ⦁ the Assembly of Multimolecular Focal Complexes Associated ⦁ with ⦁ Actin⦁ ⦁ Stress⦁ ⦁ Fibers,⦁ ⦁ Lamellipodia,⦁ ⦁ and⦁ ⦁ Filopodia.⦁ Cell 1995, 81, 53− 62.
⦁ Ridley, A. J. ⦁ Rho GTPases and Actin Dynamics in⦁ ⦁ Membrane ⦁ Protrusions⦁ ⦁ and⦁ ⦁ Vesicle⦁ ⦁ Trafficking.⦁ Trends Cell Biol. 2006, 16, 522− 529.
⦁ Hall, A. ⦁ Rho⦁ ⦁ Family⦁ ⦁ GTPases.⦁ Biochem. Soc. Trans. 2012, 40, 1378−1382.
⦁ Lavanderos, B.; Silva, I.; Cruz, P.; Orellana-Serradell, O.; Saldías, M. P.; Cerda, O. ⦁ TRP Channels Regulation of Rho GTPases ⦁ in⦁ ⦁ Brain⦁ ⦁ Context⦁ ⦁ and⦁ ⦁ Diseases.⦁ Front Cell Dev Biol 2020, 8, 582975.
⦁ Masgras, I.; Sanchez-Martin, C.; Colombo, G.; Rasola, A. ⦁ The ⦁ Chaperone⦁ ⦁ TRAP1⦁ ⦁ As⦁ ⦁ a⦁ ⦁ Modulator⦁ ⦁ of⦁ ⦁ the⦁ ⦁ Mitochondrial⦁ ⦁ Adaptations ⦁ in⦁ ⦁ Cancer⦁ ⦁ Cells.⦁ Front. Oncol 2017, 7, 58.
⦁ Paladino, A.; Woodford, M. R.; Backe, S. J.; Sager, R. A.; Kancherla, P.; Daneshvar, M. A.; Chen, V. Z.; Bourboulia, D.; Ahanin,
E. F.; Prodromou, C.; Bergamaschi, G.; Strada, A.; Cretich, M.; Gori, A.; Veronesi, M.; Bandiera, T.; Vanna, R.; Bratslavsky, G.; Serapian, S. A.; Mollapour, M.; Colombo, G. Chemical Perturbation of Oncogenic Protein Folding: from the Prediction of Locally Unstable Structures to the Design of Disruptors of Hsp90-Client Interactions. Chem. Eur. J. 2020, 26, 9459−9465.
⦁ Etienne-Manneville, S.; Hall, A. ⦁ Rho⦁ ⦁ GTPases⦁ ⦁ in⦁ ⦁ Cell⦁ ⦁ Biology.
Nature 2002, 420, 629−635.
⦁ Tinevez, J.-Y.; Perry, N.; Schindelin, J.; Hoopes, G. M.; Reynolds, G. D.; Laplantine, E.; Bednarek, S. Y.; Shorte, S. L.; Eliceiri,
K. W. TrackMate: An open and extensible platform for single-particle tracking. Methods 2017, 115, 80−90.
⦁ Sgrignani, J.; Bon, M.; Colombo, G.; Magistrato, A. ⦁ Computational Approaches Elucidate the Allosteric Mechanism ⦁ of ⦁ Human Aromatase Inhibition: a Novel Possible Route to ⦁ Small- ⦁ Molecule⦁ ⦁ Regulation⦁ ⦁ of⦁ ⦁ CYP450s⦁ ⦁ Activities?⦁ J. Chem. Inf. Model. 2014, 54, 2856−2868.
⦁ Halgren, T. A. ⦁ Identifying and Characterizing Binding Sites ⦁ and ⦁ Assessing⦁ ⦁ Druggability.⦁ J. Chem. Inf. Model. 2009, 49, 377−389.
⦁ Brenke, R.; Kozakov, D.; Chuang, G.-Y.; Beglov, D.; Hall, D.; Landon, M. R.; Mattos, C.; Vajda, S. ⦁ Fragment-based⦁ ⦁ Identification⦁ ⦁ of ⦁ Druggable⦁ ⦁ ’⦁ Hot⦁ ⦁ Spots⦁ ’⦁ ⦁ of⦁ ⦁ Proteins⦁ ⦁ Using⦁ ⦁ Fourier⦁ ⦁ Domain⦁ ⦁ Correlation ⦁ Techniques. Bioinformatics 2009, 25, 621−627.
⦁ Maier, J. A.; Martinez, C.; Kasavajhala, K.; Wickstrom, L.; Hauser, K. E.; Simmerling, C. ⦁ ff14SB:⦁ ⦁ Improving⦁ ⦁ the⦁ ⦁ Accuracy⦁ of ⦁ Protein Side Chain and Backbone Parameters from ff99SB. J. Chem. Theory Comput. 2015, 11, 3696−3713.
⦁ Friesner, R. A.; Banks, J. L.; Murphy, R. B.; Halgren, T. A.; Klicic, J. J.; Mainz, D. T.; Repasky, M. P.; Knoll, E. H.; Shelley, M.; Perry, J. K.; Shaw, D. E.; Francis, P.; Shenkin, P. S. ⦁ Glide:⦁ ⦁ a⦁ ⦁ New ⦁ Approach for Rapid, Accurate Docking and Scoring. 1. Method⦁ ⦁ and ⦁ Assessment of Docking Accuracy. J. Med. Chem. 2004, 47, 1739− 1749.
⦁ Wang, J.; Wolf, R. M.; Caldwell, J. W.; Kollman, P. A.; Case, D.
A. Development and Testing of a General Amber Force Field. J. Comput. Chem. 2004, 25, 1157−1174.
⦁ Aqvist, J. ⦁ Ion-water Interaction Potentials Derived from ⦁ Free ⦁ Energy Perturbation Simulations. J. Phys. Chem. 1990, 94, 8021− 8024.
⦁ Jorgensen, W. L.; Chandrasekhar, J.; Madura, J. D.; Impey, R. W.; Klein, M. L. ⦁ Comparison of Simple Potential Functions ⦁ for ⦁ Simulating Liquid Water. J. Chem. Phys. 1983, 79, 926−935.
⦁ Joung, I. S.; Cheatham, T. E. ⦁ Determination of Alkali ⦁ and ⦁ Halide Monovalent Ion Parameters for Use in E⦁ x⦁ plicitly Solvated ⦁ Biomolecular⦁ ⦁ Simulations.⦁ J. Phys. Chem. B 2008, 112, 9020−9041.
⦁ Sousa da Silva, A. W.; Vranken, W. F. ⦁ ACPYPE⦁ ⦁ -⦁ ⦁ AnteChamber ⦁ PYthon Parser interfacE. BMC Res. Notes 2012, 5, 367.
⦁ Parrinello, M.; Rahman, A. ⦁ Crystal Structure and Pair ⦁ Potentials:⦁ ⦁ A⦁ ⦁ Molecular-Dynamics⦁ ⦁ Study. Phys. Rev. Lett. 1980, 45, 1196.
⦁ Bussi, G.; Donadio, D.; Parrinello, M. ⦁ Canonical⦁ ⦁ Sampling ⦁ through Velocity Rescaling. J. Chem. Phys. 2007, 126, 014101.
⦁ Van Der Spoel, D.; Lindahl, E.; Hess, B.; Groenhof, G.; Mark,
A. E.; Berendsen, H. J. C. GROMACS: Fast, Flexible, and Free. J. Comput. Chem. 2005, 26, 1701−1718.
⦁ Case, D. A.; Ben-Shalom, I. Y.; Brozell, S. R.; Cerutti, D. S.; Cheatham, T. E., III; Cruzeiro, V. W. D.; Darden, T. A.; Duke, R. E.; Ghoreishi, D.; Gilson, M. K.; Gohlke, H.; Goetz, A. W.; Greene, D.; Harris, R.; Homeyer, N.; Huang, Y.; Izadi, S.; Kovalenko, A.; Kurtzman, T.; Lee, T. S.; LeGrand, S.; Li, P.; Lin, C.; Liu, J.; Luchko, T.; Luo, R.; Mermelstein, D. J.; Merz, K. M.; Miao, Y.; Monard, G.; Nguyen, C.; Nguyen, H.; Omelyan, I.; Onufriev, A.; Pan, F.; Qi, R.; Roe, D. R.; Roitberg, A.; Sagui, C.; Schott-Verdugo, S.; Shen, J.; Simmerling, C. L.; Smith, J.; Salomon Ferrer, R.; Swails, J.; Walker, R. C.; Wang, J.; Wei, H.; Wolf, R. M.; Wu, X.; Xiao, L.; York, D. M.; Kollman, P. A. AMBER 2018; University of California: San Francisco, 2018.
⦁ Daura, X.; Gademann, K.; Jaun, B.; Seebach, D.; van Gunsteren,
W. F.; Mark, A. E. Peptide Folding: When Simulation Meets Experiment. Angew. Chem., Int. Ed. 1999, 38, 236−240.
⦁ Mark, L.; Palermo, G.; Spinello, A.; Rothlisberger, U.; Magistrato, A. ⦁ All-atom Simulations Disentangle the Functional ⦁ Dynamics Underlying Gene Maturation ⦁ in the ⦁ Intron Lariat ⦁ Spliceosome.⦁ Proc. Natl. Acad. Sci. U.S.A. 2018, 115, 6584−6589.
⦁ Borisek, J.; Saltalamacchia, A.; Galli, A.; Molteni, E.; Palermo, G.; Malcovati, L.; Magistrato, A. ⦁ Disclosing ⦁ the ⦁ Impact ⦁ of ⦁ Carcinogenic SF3b Mutations on Pre-mRNA Recognition Via ⦁ All- ⦁ Atom Simulations. Biomolecules 2019, 9, 633.
⦁ Borisěk, J.; Saltalamacchia, A.; Spinello, A.; Magistrato, A. ⦁ E⦁ x⦁ ploiting Cryo-EM Structural Information and All-Atom Simulations ⦁ To Decrypt the Molecular Mechanism of Splicing Modulators. J. Chem. Inf. Model. 2020, 60, 2510−2521.
⦁ Massova, I.; Kollman, P. A. Combined Molecular Mechanical and Continuum Solvent Approach (MM-PBSA/GBSA) to Predict Ligand Binding. Perspect. Drug Discovery Des. 2000, 18, 113−135.
⦁ Chi, X.; Wang, S.; Huang, Y.; Stamnes, M.; Chen, J.-L. ⦁ Roles⦁ ⦁ of ⦁ rho⦁ ⦁ GTPases⦁ ⦁ in⦁ ⦁ Intracellular⦁ ⦁ Transport⦁ ⦁ and⦁ ⦁ Cellular⦁ ⦁ Transformation. Int. J. Mol. Sci. 2013, 14, 7089−7108.
⦁ Maldonado, M. D. M.; Dharmawardhane, S. ⦁ Targeting⦁ ⦁ Rac⦁ ⦁ and ⦁ Cdc42 GTPases in Cancer. Cancer Res. 2018, 78, 3101−3111.
⦁ Oprea, T. I.; Sklar, L. A.; Agola, J. O.; Guo, Y.; Silberberg, M.; Roxby, J.; Vestling, A.; Romero, E.; Surviladze, Z.; Murray-Krezan, C.; Waller, A.; Ursu, O.; Hudson, L. G.; Wandinger-Ness, A. ⦁ Novel ⦁ Activities of Select NSAID R-Enantiomers against Rac1 and ⦁ Cdc42 ⦁ GTPases.⦁ PLoS One 2015, 10, No. e0142182.
⦁ Spinello, A.; Martini, S.; Berti, F.; Pennati, M.; Pavlin, M.; Sgrignani, J.; Grazioso, G.; Colombo, G.; Zaffaroni, N.; Magistrato, A. ⦁ Rational⦁ ⦁ Design⦁ ⦁ of⦁ ⦁ Allosteric⦁ ⦁ Modulators⦁ ⦁ of⦁ ⦁ the⦁ ⦁ Aromatase⦁ ⦁ Enzyme: ⦁ An⦁ ⦁ Unprecedented⦁ ⦁ Therapeutic⦁ ⦁ Strategy⦁ ⦁ to⦁ ⦁ Fight⦁ ⦁ Breast⦁ ⦁ Cancer.⦁ Eur. J. Med. Chem. 2019, 168, 253−262.
⦁ Yang, J.; Zhang, Z.; Roe, S. M.; Marshall, C. J.; Barford, D. ⦁ Activation⦁ ⦁ of⦁ ⦁ Rho⦁ ⦁ GTPases⦁ ⦁ by⦁ ⦁ DOCK⦁ ⦁ E⦁ x⦁ change⦁ ⦁ Factors⦁ ⦁ Is⦁ ⦁ Mediated ⦁ by⦁ ⦁ a⦁ ⦁ Nucleotide⦁ ⦁ Sensor.⦁ Science 2009, 325, 1398.
⦁ Hong, L.; Kenney, S. R.; Phillips, G. K.; Simpson, D.; Schroeder, C. E.; Nöth, J.; Romero, E.; Swanson, S.; Waller, A.; Strouse, J. J.; Carter, M.; Chigaev, A.; Ursu, O.; Oprea, T.; Hjelle, B.; Golden, J. E.; Aubé, J.; Hudson, L. G.; Buranda, T.; Sklar, L. A.; Wandinger-Ness, A. ⦁ Characterization of a Cdc42 Protein Inhibitor ⦁ and⦁ ⦁ Its⦁ ⦁ Use⦁ ⦁ as⦁ ⦁ a⦁ ⦁ Molecular⦁ ⦁ Probe⦁ *⦁ .⦁ J. Biol. Chem. 2013, 288, 8531− 8543.
⦁ Casalino, L.; Palermo, G.; Abdurakhmonova, N.; Rothlisberger, U.; Magistrato, A. ⦁ Development of Site-Specific Mg2+-RNA ⦁ Force ⦁ Field Parameters: A Dream or Reality? Guidelines from Combined ⦁ Molecular⦁ ⦁ Dynamics⦁ ⦁ and⦁ ⦁ Quantum⦁ ⦁ Mechanics⦁ ⦁ Simulations.⦁ J. Chem. Theory Comput. 2017, 13, 340−352.
⦁ Shutes, A.; Onesto, C.; Picard, V.; Leblond, B.; Schweighoffer, F.; Der, C. J. ⦁ Specificity⦁ ⦁ and⦁ ⦁ Mechanism⦁ ⦁ of⦁ ⦁ Action⦁ ⦁ of⦁ ⦁ EHT⦁ ⦁ 1864,⦁ ⦁ a ⦁ Novel⦁ ⦁ Small⦁ ⦁ Molecule⦁ ⦁ inhibitor⦁ ⦁ of⦁ ⦁ Rac⦁ ⦁ Family⦁ ⦁ Small⦁ ⦁ GTPases.⦁ J. Biol. Chem. 2007, 282, 35666−35678.
⦁ Onesto, C.; Shutes, A.; Picard, V.; Schweighoffer, F.; Der, C. J. ⦁ Characterization of EHT 1864, a Novel Small Molecule Inhibitor ⦁ of ⦁ Rac⦁ ⦁ Family⦁ ⦁ Small⦁ ⦁ GTPases.⦁ Methods Enzymol. 2008, 439, 111−129.
⦁ Molnár, J.; Fazakas, C.; Haskó, J.; Sipos, O.; Nagy, K.; Nyuĺ- Tóth, Á.; Farkas, A. E.; Végh, A. G.; Váró, G.; Galajda, P.; Krizbai, I.
A.; Wilhelm, I. Transmigration Characteristics of Breast Cancer and Melanoma Cells through the Brain Endothelium: Role of Rac and PI3K. Cell Adhes. Migr. 2016, 10, 269−281.
⦁ Surviladze, Z.; Waller, A.; Strouse, J. J. B.; Ologa, C.; Ursu, O.; Salas, V.; Parkinson, J. F.; Phillips, G. K.; Romero, E.; Wandinger- Ness, A.; Sklar, L. A.; Schroeder, C.; Simpson, D.; Nöth, J.; Wang, J.; Golden, J.; Aubé, J. A Potent and Selective Inhibitor of Cdc42 GTPase, Probe Reports from the NIH Molecular Libraries Program: Bethesda (MD), 2010.
⦁ Yang, W. ⦁ Nucleases: Diversity of Structure, Function ⦁ and ⦁ Mechanism.⦁ Q. Rev. Biophys. 2011, 44, 1−93.
⦁ Sgrignani, J.; Magistrato, A. ⦁ QM/MM⦁ ⦁ MD⦁ ⦁ Simulations⦁ ⦁ on⦁ ⦁ the ⦁ Enzymatic Pathway of the Human Flap Endonuclease (hFEN1) ⦁ Elucidating⦁ ⦁ Common⦁ ⦁ Cleavage⦁ ⦁ Pathways⦁ ⦁ to⦁ ⦁ RNase⦁ ⦁ H⦁ ⦁ Enzymes.⦁ ACS Catal. 2015, 5, 3864−3875.
⦁ Angelopoulou, E.; Piperi, C. ⦁ Emerging Role of Plexins Signaling ⦁ in⦁ ⦁ Glioma⦁ ⦁ Progression⦁ ⦁ and⦁ ⦁ Therapy.⦁ Cancer Lett. 2018, 414, 81−87.
⦁ Chen, M.; Pan, H.; Sun, L.; Shi, P.; Zhang, Y.; Li, L.; Huang, Y.; Chen, J.; Jiang, P.; Fang, X.; Wu, C.; Chen, Z. ⦁ Structure and ⦁ Regulation of Human Epithelial Cell Transforming 2 Protein. Proc. Natl. Acad. Sci. U.S.A. 2020, 117, 1027−1035.