Through the open door: Preferential binding of dasatinib to the active form of BCR‐ABL unveiled by in silico experiments

Abstract

Dasatinib is a second‐generation BCR‐ABL inhibitor approved for the treatment of patients with chronic myeloid leukemia, both in the frontline and in the imatinib‐resistant/intolerant settings. The high affinity of dasatinib for the protein is currently assumed to result from its ability to bind both the active and inactive conformations of the BCR‐ABL kinase. In the present work, using state of the art molecular simulation techniques we prove that dasatinib exhibits a highly selective preference for the active (open) BCR‐ABL conformation. By using three different BCR‐ABL conformations (active, inactive, and intermediate inactive) we show that, from a thermodynamic standpoint, the affinity of dasatinib for BCR‐ABL drastically decreases in the order: active > alternative inactive > inactive, as a result of differential contributions from the single residues lining the kinase binding pocket and the concomitant stabilization/destabilization of the kinase hydrophobic spine. Molecule‐pulling experiments also corroborate this trend as significantly lower forces and smaller times are required to extract dasatinib from its inactive BCR‐ABL complexes with respect to the active complex counterparts. Importantly, our results support recent NMR solution results demonstrating no evidence of dasatinib bound to the inactive form of BCR‐ABL.

Highlights

• By in silico methods we confirm dasatinib effectively binds only BCR‐ABL active form.

• Dasatinib low IC50 stems also from poor ability to leave active kinase binding site.

• T315I is resistant to dasatinib due to drug poor affinity and facilitated escape.

• The hydrophobic‐spine contribution to TKI binding is calculated for the first time.

1. Introduction

Chronic myelogenous leukemia (CML) accounts for 15% of leukemias in adults (Jabbour and Kantarjian, 2012), and typically evolves in 3 different phases: chronic phase (CP), accelerated phase (AP), and blast crisis (BC). The BCR‐ABL oncogene, a product of the reciprocal chromosomal translocation t(9;22) (q34;q11), is the main pathogenetic driver. The protein product is the cytoplasmic BCR‐ABL tyrosine kinase (TK) whose activity regulates survival, proliferation, and adhesion of CML cells (Kurzrock et al., 1988).

The discovery of the role of BCR‐ABL in the pathogenesis of CML provided the rationale for the design of the tyrosine kinase inhibitor (TKI) imatinib (Gleevec, Novartis Pharmaceuticals, Basel, Switzerland), an inhibitor that binds with high affinity the inactive conformation of BCR‐ABL (Nagar et al., 2002; Deininger et al., 2005). Imatinib showed remarkable activity as a single agent in CML particularly among patients in CP (Cortes and Kantarjian, 2012). Unfortunately, resistance to imatinib in patients with CML occurs in a significant number of patients failing imatinib therapy (Druker et al., 2006). Of all possible mechanisms mediating imatinib resistance (Ohanian et al., 2012; Yeung and Hughes, 2012), the selection of subclones bearing BCR‐ABL point mutations is the most frequent. At the protein level, these point mutations result in a distorted conformation of the TK/Imatinib interface, which leads to BCR‐ABL being unable to adopt the tailor‐fit inactive (close) conformation to which imatinib binds (Gibbons et al., 2012).

Dasatinib (Sprycel, Bristol Meyer, New York, NY, USA) is a second generation BCR‐ABL inhibitor able to inhibit the in vitro and in vivo kinase activity of most imatinib‐resistant BCR‐ABL mutants, with the exception of the “gatekeeper” mutation T315I (Shah et al., 2004). Noteworthy, in vitro experiments have shown that dasatinib is 325‐fold more potent than imatinib in cells expressing wild‐type BCR‐ABL (Martinelli et al., 2005).

The solved crystal structure of dasatinib in complex with BCR‐ABL (Tokarski et al., 2006) clearly showed that, at variance with imatinib, this TKI is able to target the active (open) form of the TK. Notwithstanding, in the same work Tokarski et al. (2006) hypothesized that the reason for the greater BCR‐ABL affinity of dasatinib towardtoward the BCR‐ABL kinase with respect to imatinib was due, at least in part, to its ability to bind the inactive conformation of the BCR‐ABL kinase. Indeed, these authors claimed ‐ in a narrative fashion based on “the modeling of dasatinib in the imatinib‐bound and PD173855‐bound forms of the kinase” (not reported in the paper) – that no “major sterical clashes would preclude dasatinib from also binding the inactive conformations found for in these other ABL structures”. In support of this assertion, Gambacorti‐Passerini et al. (2005) predicted that dasatinib could not only bind to the inactive form of the kinase but also to another BCR‐ABL conformation, termed the intermediate conformation – in which the kinase activation loop (A‐loop) assumes a conformation utterly similar to the one found in the active enzyme. According to the binding mode of Gambacorti‐Passerini et al. (2005), dasatinib was likely to be more active against the inactive and intermediate forms of BCR‐ABL than versus the active open A‐loop form, although no numerical data were offered to sustain these claims. To support these basic modeling results, these authors just invoked the lack of involvement of T315I in binding ATP; however, the importance of the T315I mutation in dasatinib resistance is against the inactive and intermediate forms rather than the active kinase, because dasatinib does not act as a true inhibitor of ATP binding.

Herein we performed a systematic in silico investigation based on extensive and state‐of‐the art molecular simulations and ultimately verified that dasatinib is unfit to bind different conformational states of the BCR‐ABL kinase. Furthermore, dasatinib shows a remarkable, almost exclusive preference towards the kinase open form, in full agreement with recent relevant NMR solution studies (Vajpai et al., 2008).

2. Methods

2.1. PDB files

The following X‐ray structures were employed in all simulations: i) 1IEP: BCR‐ABL/imatinib complex (Nagar et al., 2002), in which the TK is in its inactive form (IF), with the activation loop in its closed (i.e., DFG‐out) conformation and the Tyr‐393 residue is unphosphorylated; ii) 1M52: BCR‐ABL/PD173955 complex (Nagar et al., 2002), in which the TK is an alternative inactive form (IFa), with the activation loop in a closed (i.e., DFG‐out) conformation, although different from that found in 1IEP, and the Tyr‐393 residue unphosphorylated; iii) 2GQG (chain A): BCR‐ABL/dasatinib complex (Tokarski et al., 2006), in which the kinase is in its active form (AFA), with the activation loop in its open (i.e., DFG‐in) conformation, and the Tyr‐393 residue is phosphorylated; iv) 2GQG (chain B): BCR‐ABL/dasatinib complex (Tokarski et al., 2006), in which the kinase is its active form (AFB), with the activation loop in its open (i.e., DFG‐in) conformation, and the Tyr‐393 residue is unphosphorylated.

2.2. Computational details

All classical, parallel molecular dynamics (MD) simulations and the relevant analyses were performed the Amber 11 suites of programs (Case et al., 2010) running on the PLX‐GPU and FERMI supercomputers at the CINECA supercomputer center (Bologna, Italy). Each BCR‐ABL/TKI complex was then optimized in a box of TIP3P water molecules (Jorgensen et al., 1983) in the presence of 0.15 M NaCl. After each system equilibration and heating, 50 ns of data collection MD runs, necessary for the estimation of the free energy of binding (vide infra), were performed (see SI for full details). For the calculation of the binding free energy between BCR‐ABL and each inhibitor in water, a total of 50,000 snapshots were saved during the MD data collection period described above.

The binding free energy ΔG _bind of each BCR‐ABL1/TKI complex in water was calculated according to the procedure termed Molecular Mechanic/Poisson‐Boltzmann Surface Area (MM/PBSA), and originally proposed by Srinivasan et al. (1998). The theoretical background of this methodology is described in details in the original papers by Kollman et al. (2000), and has been successfully employed by our group in related studies (see SI) (Pierotti et al., 2011). Briefly, an MD simulation (typically in explicit solvent) is carried out which yields a representative ensemble of structures, from which the average total free energy of binding between each drug and the protein receptor ΔG _bind can then be calculated. The corresponding IC₅₀ values were calculated from the binding free energies using the following relationship (Wang et al., 2001): ΔG _bind = RT ln K _diss ≅ RT ln IC₅₀.

The contribution of each kinase residue belonging to the so‐called “hydrophobic spine” (Azam et al., 2008) to the stabilization/destabilization of the active/inactive protein conformation and its eventual relation to drug resistance was investigated by means of component analysis (Gohlke et al., 2003). Accordingly, a per residue binding free energy decomposition was performed exploiting the MD trajectory of each given inhibitor/kinase complex. This analysis was carried out using the MM/GBSA approach (Tsui and Case, 2000), and was based on the same snapshots used in the binding free energy calculation.

The entire MM/PBSA computational procedure was optimized by integrating AMBER 11 in modeFRONTIER, a multidisciplinary and multi‐objective optimization and design environment (http://www.esteco.com/home/mode_frontier/mode_frontier.html).

The steered molecular dynamics (SMD) simulations (Isralewitz et al., 2001) were performed using snapshots randomly taken from the MD equilibrated runs as initial structures. With the adopted settings (see SI), 2.5 nm were covered in 500 ps of SMD simulation.

3. Results

3.1. Differences between 3D models available for BCR‐ABL kinase

Panel A in Figure 1 compares the crystallographic structures of the active (AFA) and inactive (IF) BCR‐ABL conformations. The most notable differences between the two kinase conformations are the opposite orientation of the activation loop (A‐loop) and of the relevant DFG catalytic triad. When the open, active form of BCR‐ABL (AFA) is compared with the alternative, inactive form of the kinase (IFa), the A‐loop in the latter assumes an open conformation but the catalytic triad is still positioned in a DFG‐out (i.e., inactive) orientation (Figure 1, panel B). Lastly, panel C in Figure 1 compares the two active chains found in the unit crystal cell of BCR‐ABL in complex with dasatinib (AFA and AFB). Interestingly, chain A is phosphorylated on Try393 whilst chain B, although still complexed with the TKI in the original crystal, is unphosphorylated at the same residue. However, the two structures are almost entirely superimposable.

Figure 1.

Comparison of crystallographic structures available for inactive, active and intermediate state of the TK. (A) Superposition of the active (AFA, orange) and inactive (IF, cyan) structure of BCR‐ABL. (B) Superposition of the active (AFA, orange) and alternative, inactive (IFa, forest green) structure of BCR‐ABL. (C) Superposition of the two active structures of BCR‐ABL found in the same crystal structure (AFA, orange, and AFB, blue). All images are zoomed view of the TK A‐loop region. In all structures, the DFG motif and the Tyr393 residue are highlighted as sticks‐and‐balls.

3.2. Docking and binding affinity

In order to evaluate the possible interactions of dasatinib with the inactive (IF) and the alternative inactive (IFa) form of BCR‐ABL, the three‐dimensional (3D) model of the TKI was positioned within the two protein binding sites according to an established procedure (see also SI) (Pierotti et al., 2010). The validity of the adopted docking protocol was further assessed by reproducing the available orientations of imatinib and dasatinib as previously determined on crystal structure (see SI for details).

In contrast to previous reports (Gambacorti‐Passerini et al., 2005), the mere docking results relative to dasatinib binding to the different forms of the BCR‐ABL kinase are not sufficient to determine whether this TKI shows any real preference for the active, inactive or intermediate state of the enzyme. Indeed, a visual inspection of the binding modes of dasatinib in the binding site of the inactive (IF) and the alternative inactive (IFa) form of the protein does not reveal evident steric clashes or substantial difference/distortion of the binding site conformations when compared to those of the two active kinase chain conformations (AFA and AFB). Accordingly, all 4 complexes (IF/dasatinib, IF_a/dasatinib, AFA/dasatinib, and AFB/dasatinib) were subjected to extensive molecular dynamics (MD) simulations in the MM/PBSA framework of theory (Kollman et al., 2000; Pierotti et al., 2011) to estimate the free energy of binding of dasatinib (ΔG _bind) – and the relevant IC₅₀ – for each complex TKI/TK assembly. These results are shown in Table 1.

Table 1.

Molecular dynamics free energy of binding (ΔGbind) and IC50 values for dasatinib in complex with different active and inactive forms of BCR‐ABL.

BCR‐ABL	ΔGbind (kcal/mol)	IC50 (nM)a
AFA	−12.41 ± 0.01	0.8
AFB	−12.29 ± 0.02	1.0
IFa	−9.40 ± 0.02	130
IF	−8.89 ± 0.01	306

Data are reported with standard errors of the mean.

^aΔGbind and IC50 (i.e., the concentration of dasatinib that inhibits the kinase activity by 50%) are related by the following fundamental equation: ΔGbind = −RT ln 1/IC50.

Regarding dasatinib in complex with the two open forms of BCR‐ABL (AFA and AFAB, differing only by the presence/absence of a phosphate group on the Tyr393 respectively), no significant differences in the corresponding binding free energy values were observed. Specifically, for both chain A and B bound to dasatinib the largely negative values of ΔG _bind in Table 1 show the highly favorable interaction energy of the TKI with the open/active state of the TK. Importantly, the calculated IC₅₀ is in concert with experimental data, thus indicating an affinity of the drug for the protein in the nanomolar range (Martinelli et al., 2005).

In the case of the inactive (IF, A‐loop in close conformation and DFG‐out conformation) and alternative inactive (IFa, A‐loop in open conformation but DFG‐out conformation) forms of BCR‐ABL in complex with dasatinib, the estimated ΔG _bind values sit above (i.e., are less negative than) −10 kcal/mol, indicating a net decrease of affinity of the drug for the kinase. As mentioned above, the major difference between these two BCR‐ABL/TKI complexes resides in the small alteration of the conformation of the catalytic triad DFG, which in both cases is in the out conformation. Although these residues are not proximal to the drug binding site, this difference reflects in a structural rearrangement of the protein which, in turn, influences the dasatinib binding site, as detailed in Figure 2.

Figure 2.

Molecular dynamics snapshots of dasatinib in complex with active and inactive conformations of BCR‐ABL. (A) Equilibrated snapshot from MD simulation of the AFA kinase form (Y393 phosphorylated, DFG‐in conformation) in complex with dasatinib. The TKI is in an atom‐colored stick‐ and‐ball representation (C, gray; O, red; N, blue; yellow, S; green, Cl). The main residues involved in the binding are highlighted as labeled sticks. Hydrogen atoms, ions and counterions, and water molecules are omitted for clarity. (B) Equilibrated snapshot from MD simulations of (top left) dasatinib/AFA (kinase open form, Y393 phosphorylated, DFG‐in conformation), vs. dasatinib/AFB (kinase open form, Y393 unphosphorylated, DFG‐in conformation); (top right) dasatinib/AFA vs. dasatinib/IF (kinase closed form, DFG‐out conformation); (bottom left) dasatinib/AFA vs. dasatinib/IFa (kinase alternative inactive form, DFG‐out conformation); (bottom right) dasatinib/IF (kinase inactive form, DFG‐out conformation) vs. dasatinib/IFa (kinase alternative inactive form, DFG‐out conformation). In the first case, the two binding sites are very similar, as is the binding mode and conformation of the inhibitor in the two protein binding pockets. In the other cases, the differences in the activation loops and binding site are evident. Dasatinib/AFA (orange); dasatinib/IF (cyan); dasatinib/IFa (forest green); dasatinib/AFB (blue).

3.3. Per‐residue deconvolution of the free energy of binding

In order to gain further insight into the binding modes of dasatinib to the different forms of the BCR‐ABL kinase, we calculated the energetic contribution for those residues which afford a substantial contribution to the binding. As shown in Table 2, in the case of the dasatinib/AFA and AFB complexes, no meaningful differences were observed at the individual residue binding level. Conversely, in the case of both dasatinib/IF and IFa assemblies, the overall reduction of drug affinity cannot be attributed either to a single residue or a particular cluster of binding site residues. Instead, a general decrease of the binding energy for all residues involved is predicted. In other words, the different protein conformations (inactive or alternative inactive conformation) do not result in a marked decrease of the binding energy localized in a specific region of the protein binding pocket but, on the contrary, they exert a sort of domino effect that affects the entire protein binding site.

Table 2.

Per‐residue free‐energy decomposition for AFA, AFB, IF and IFa in complex with dasatinib.

Residue	AFA	AFB	IF	IFa
	$\Delta {\text{G}}_{\text{bind}}^{\text{bs}}$	$\Delta {\text{G}}_{\text{bind}}^{\text{bs}}$	$\Delta {\text{G}}_{\text{bind}}^{\text{bs}}$	$\Delta {\text{G}}_{\text{bind}}^{\text{bs}}$
L248	−0.92	−0.95	−0.78	−0.76
M290	−1.15	−1.02	−0.93	−0.91
V299	−1.29	−1.08	−0.95	−1.10
I313	−1.32	−1.41	−1.15	−1.12
T315	−1.21	−0.92	−0.75	−0.83
E316	−0.54	−0.49	−0.35	−0.40
F317	−1.68	−1.87	−1.37	−1.45
M318	−0.56	−0.61	−0.42	−0.36
T319	−1.51	−1.56	−1.34	−1.32
$\sum \Delta {\text{G}}_{\text{bind}}^{\text{bs}}$ (kcal/mol)	−10.18	−9.91	−8.04	−8.25

All values are in kcal/mol. Data standard deviations are in the range ±0.01 ÷ ±0.05.

This analysis confirms how the conformation adopted by the kinase in the open forms AFA and AFB is a key factor for the drug high stability within its protein binding site with respect to the other two kinase conformations (IF and IFa) considered. Indeed, the sums of all individual residue contribution to binding in the first two complexes (dasatinib/AFA and dasatinib/AFB, equal to −10.18 and −9.91 kcal/mol, respectively) are more favorable (by approximately 2 kcal/mol) than those calculated for the dasatinib/IF (−8.04 kcal/mol) and dasatinib/IFa (−8.25 kcal/mol).

3.4. Hydrophobic spine analysis

The hydrophobic spine is a highly conserved network of hydrophobic interactions characteristic of the active kinase conformation which is stabilized by the gatekeeper mutation T315I in BCR‐ABL (Azam et al., 2008). Mutations at spine residues may disrupt the hydrophobic connectivity and inactivate the kinase. In the BCR‐ABL protein, the residues making up the hydrophobic spine are M290, L301, H361, and F382, this last amino acid further belonging to the catalytic triad DFG involved in kinase activation.

To describe and quantify the role of the hydrophobic spine upon dasatinib binding to the 4 different BCR‐ABL conformations, we evaluated the free energy of binding deconvolution analysis considering the BCR‐ABL residues involved in the spine. A pictorial view of the results is shown in Figure 3.

Figure 3.

Conformation of BCR‐ABL hydrophobic spine residues for active and inactive forms of BCR‐ABL in complex with dasatinib. Equilibrated molecular dynamics snapshots of dasatinib in complex with the two BCR‐ABL active forms (top left, AFA; top right, AFB), with the inactive form (bottom left), and with the alternative inactive form IFa (bottom right). All residues belonging to the hydrophobic spine are highlighted as red sticks, and labeled. Dasatinib is portrayed as atom‐colored balls‐and‐sticks (C, gray; O, red; N, blue; S, yellow; Cl, green). Hydrogen atoms, ions and counterions, and water molecules are omitted for clarity.

The different conformational state of the residue F382 in the DFG‐in and DFG‐out configuration influences the spine stabilization process. More importantly, both in the case of dasatinib/IF and dasatinib/IFa complexes (in which the kinase features the DFG‐out conformation), the spine stabilization is much less effective than for the two alternative complexes (AFA and AFB) in which the kinase has an open, active conformation (DFG‐in), as quantified in Table 3. In fact, the contribution afforded by the hydrophobic spine stabilization is quite larger in the case of the active conformation for which, in turn, the binding of dasatinib is more favorable. In those BCR‐ABL/dasatinib complexes where the hydrophobic spine is less stabilized, the interaction between M290 and L301 is preserved, but the interaction energy between M290 and F382 becomes less favorable (from −0.60 kcal/mol for the AFA and −0.20 and −0.26 kcal/mol for the IF and IFa, respectively), and especially the stabilizing energies between H361 and F382 become remarkably weaker (from −1.7 kcal/mol for the AFA to −0.15 and −0.11 kcal/mol for IF and IFa, respectively).

Table 3.

Hydrophobic spine per residue free‐energy decomposition for AFA, AFB, IF and IFa in complex with dasatinib.

AFA	L301	H361	F382	AFB	L301	H361	F382
M290	−1.6	−0.02	−0.60	M290	−1.5	−0.07	−0.56
L301		−0.02	−0.01	L301		0.00	−0.01
H361			−1.7	H361			−1.8
$\sum \Delta G_{\text{bind}}^{\text{spine}}$			−4.0	$\sum \Delta G_{\text{bind}}^{\text{spine}}$			−3.9
IF	L301	H361	F382	IFa	L301	H361	F382
M290	−1.6	−0.04	−0.20	M290	−1.6	−0.06	−0.26
L301		−0.02	−0.03	L301		−0.01	−0.05
H361			−0.15	H361			−0.11
$\sum \Delta G_{\text{bind}}^{\text{spine}}$			−2.0	$\sum \Delta G_{\text{bind}}^{\text{spine}}$			−2.1

All values are in kcal/mol. Data standard deviations are in the range ±0.01 ÷ ±0.05.

3.5. Steered molecular dynamics and dasatinib unbinding process

MM/PBSA is a well‐known and validated methodology to estimate the free energy of e.g., drug/protein binding in fair to excellent agreement with experimental data. Usually, such technique can yield reliable information both on the main qualitative and quantitative aspects of drug binding; on the other hand, however, it deals only with the thermodynamic aspect of the association of a given ligand with its target protein. Thus, further confirmation of MM/PBSA‐based results concerning the remarkable preference of dasatinib toward the open form of the BCR‐ABL kinase was pursued by applying a steered molecular dynamics (SMD) protocol. Such computational experiments allow to estimate the force that is required to pull dasatinib out of its protein binding site, and in the present work it was applied for the first time to all 4 Dasatinib/BCR‐ABL complexes discussed above. As a result, also some kinetic aspects of the reversible drug binding (as is the present case) are accounted for in the analysis.

Figure 4 shows a prototypical smoothed force (F) vs. time (t) diagram obtained from the SMD runs for the 4 complexes of dasatinib with BCR‐ABL. As we can see, all curves present an initial phase in which the force increases up to a maximum, which represents the force value that needs to be applied to detach the drug from its binding site. In the case of the AFA/and AFB/dasatinib complexes, the TKI/TK complex rupture force is around 1200 pN, and this unbounded configuration is attained after approximately 300 ps. For these two complexes, the corresponding F vs. t profiles are very similar, suggesting a similar unbinding process event, as expected. On the contrary, and in line with all evidence gained from the MM/PBSA simulations discussed above, for the IF/and IFa/dasatinib complexes not only the force at rupture has a notably lower peak value (approx. 800 pN) but this peak is reached quite sooner, i.e., after approx. 180 ps. These SMD experiments ineluctably confirm that dasatinib is characterized not only by a remarkable lower affinity for both inactive forms of BCR‐ABL (as quantified by the relevant ΔG _bind values) but also by a kinetically different protein unbinding process among the different kinase conformations. In other words, should dasatinib be able to weakly bind to one of the inactive form of the BCR‐ABL, notably lower force and time values would be required to detach it from its binding site. These new findings are in full agreement with Vajpai et al. (2008), who used solution NMR to study dasatinib binding to phosphorylated and unphosphorylated forms of the BCR‐ABL kinase. According to their studies, no trace of dasatinib bound to the unphosphorylated form of the kinases was detected in solution.

Figure 4.

Average force profile of dasatinib unbinding from its complex with the different active and inactive forms of BCR‐ABL. Comparison of the unbinding force profiles of dasatinib from the IF (cyan), IFa (forest green), AFA (orange), and AFB (blue) forms of the BCR‐ABL TK. For each protein form, the plots show the resulting mean values from averaging the force profiles from five different SMD runs.

4. Discussion

In silico analyses performed in this work ‐ based on a combination of MD/SMD simulations – revealed that dasatinib shows a marked preference for binding to the active state of the BCR‐ABL kinase. The binding affinity of this TKI for the open form of its target enzyme is not only due to the highly favorable ΔG _bind (−12.41 kcal/mol), which translates into an IC₅₀ of 0.8 nM, thus ranking dasatinib as one of more potent small molecule inhibitors of wild‐type BCR‐ABL, both in biochemical and cellular assays (Martinelli et al., 2005) – but also from a kinetic point of view, in terms of reduced ability of escaping from the protein binding site when bound to its active conformation.

A per residue deconvolution of the free energy of binding shows that, in a sort of domino effect, the efficiency of the amino acid lining the kinase binding pocket in binding dasatinib decreases of ∼2 kcal/mol (that is nearly 1.5 orders of magnitude) in passing from the open to the close form of the kinase. The same analysis applied to the protein residues making up the kinase hydrophobic spine also indicate that, correctly, this network of hydrophobic interaction is progressively disrupted when dasatinib is bound to the alternative inactive and the inactive form of the protein. Further, SMD simulations mimicking molecule‐pulling experiments also corroborate the MM/PBSA‐based dasatinib/BCR‐ABL affinity trend, as significantly lower forces and smaller times are required to extract dasatinib from its inactive BCR‐ABL complexes with respect to the active complex counterparts.

Our results explain why, by considering the stabilization/destabilization of the BCR‐ABL hydrophobic spine in the open/close TK forms, the gatekeeper mutation T315I is highly resistant to dasatinib. Indeed, it has been experimentally verified (Young et al., 2006) that the T315I BCR‐ABL variant induced equivalent levels of total cellular phosphotyrosine in cell lysates as the wild‐type counterpart. Also, since the gatekeeper residue T315 is situated near the tip of the hydrophobic spine, the substitution of a bulkier hydrophobic residue such as isoleucine at this position should result, as suggested (Young et al., 2006), in stabilization of the active state, to which dasatinib binds, by strengthening the hydrophobic spine. At the same time, however, the change in conformation of the TK binding pocket leads to an overall rearrangement of dasatinib within the protein cavity, with an overall loss of several stabilizing protein/drug interactions. Notwithstanding the stabilization of the BCR‐ABL hydrophobic spine and, hence, of its active conformation, the affinity of dasatinib for the TK is drastically decreased by virtue of a substantial penalty in the free energy paid by the TKI upon binding. The altered binding site conformation also results in a greater facility for dasatinib to escape from T315I BCR‐ABL; interestingly the SMD corresponding force–time profile is utterly similar to that predicted for the dasatinib/IF complex (see SI for details).

Current structural understanding of kinases is largely based on x‐ray crystallographic studies, whereas very little data exist on the conformations and dynamics that kinases adopt in the solution state. Recently, it has been reported the first characterization of ABL TK in complex with three “ib” TKIs (imatinib, nilotinib, and dasatinib) by solution NMR techniques (Vajpai et al., 2008). While imatinib and nilotinib in complex with the TK feature the A‐loop in the close (inactive) conformation, the BCR‐ABL/dasatinib complex preserves the active conformation, in stark contrast with previous predictions based upon molecular docking concepts (Gambacorti‐Passerini et al., 2005), but in full agreement with the present results. Interestingly, these NMR experiments were performed with non‐phosphorylated protein, a technical choice to ensure that the protein might adopt the inactive DFG‐out conformation (Vajpai et al., 2008). Since the presence of a phosphotyrosine residue at position 393 (PTR393, see Figure 1) in the A‐loop concurs to stabilize the active conformation of the protein by forming interactions with neighboring side chains (Young et al., 2006), the authors anticipated that this occurrence would further reduce the flexibility of the BCR‐ABL/dasatinib complex, thereby narrowing the conformational space available to the DFG‐in configuration. For the same reason, any propensity of the BCR‐ABL/dasatinib complex to adopt the inactive DFG‐out conformation would be even more reduced. Our results fully corroborate this hypothesis, and concur in ascertaining the highly selective preference of dasatinib for binding the active, open conformation of the BCR‐ABL kinase.

Authorship

Contribution: E.L. and P.P. designed and performed research and analyzed data; M.F., A.Q.C. and D.L.G. analyzed data and wrote the paper; S.P. performed research, analyzed data and wrote the paper.

Conflict of interest

The authors declare no competing financial interests.

Supporting information

Supplementary data

Supplementary data 1.

Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.molonc.2013.06.001.

Laurini Erik, Posocco Paola, Fermeglia Maurizio, Gibbons Don L., Quintás-Cardama Alfonso, Pricl Sabrina, (2013), Through the open door: Preferential binding of dasatinib to the active form of BCR-ABL unveiled by in silico experiments, Molecular Oncology, 7, doi: 10.1016/j.molonc.2013.06.001.