Table 1. Electrostatic binding free energies of protonated and
unprotonated imatinib with Abl
kinase. Protonated imatinib shows a more favorable electrostatic
interaction with WT and mutant Abl compared to unprotonated
imatinib.
30
Table 2. Electrostatic binding free energies of protonated and
unprotonated ponatinib with Abl
kinase. Protonated ponatinib shows a favorable electrostatic
interaction with WT and mutant Abl compared to unprotonated
ponatinib. Ponatinib shows a more favorable electrostatic energy
when bound to mutant than to WT Abl. Protonation improves the
electrostatic binding free energy in all cases. The
electrostatic
binding free energy of protonated drugs was consistently less than
that of unprotonated drugs by
about ~3 kcal/mol for imatinib and ~4 kcal/mol for ponatinib.
Interestingly, ponatinib bound to
mutant Abl showed the greatest relative increase in electrostatic
binding affinity upon
protonation. The results agree well with MD free energy simulations
that showed a strong
preference for a drug to bind to Abl in its protonated state with a
net positive charge, as it
favorably interacts with negatively charged residues in the binding
site13-15. Also, the pKa of the
N atom in a freely solvated piperazinyl group is 9.85, and thus, at
a physiological pH of 7.4, the
equilibrium already favors protonation15.
Therefore, in all subsequent analyses, we will consider only the
protonated forms of the
drugs and will not explicitly refer to them as “protonated”.
Component analysis quantifies the contribution of drug moieties to
binding
The contribution of drug moieties to the overall electrostatic
binding free energy was determined
by the change in electrostatic binding free energy when charges on
moieties were set to zero . The results are shown in Figures 12 and
13.
31
Figure 12: Component analysis of imatinib for favorable
contribution. The Structure of imatinib colored by atom type with
Abl residues that form hydrogen bonds shown in yellow. The
energetic contributions of moieties that form hydrogen bonds with
the WT and mutant Abl residues are shown. The contribution of a
moiety is given by a value. Blue boxes represent favorable moieties
with a value greater than 1. None of the moieties shown contributed
unfavorably to binding ( < − .
Component analysis shows that many moieties that form hydrogen
bonds with Abl
residues contributed favorably to binding with a value greater than
1 kcal/mol.
Interestingly, moiety III, which forms hydrogen bonds with Asp 381
and Glu 286, had the most
favorable contribution to binding when bound to WT ( = 1.91
kcal/mol). However, its
contribution decreased by 0.86 kcal/mol when bound to T315I mutant.
Moiety VII, which forms
hydrogen bond with Met 318, also showed a favorable contribution
for imatinib bonded to WT
( = 1.67 kcal/mol) but became less favorable with the T315I mutant
( = 0.67 kcal/mol).
The decrease in binding affinity by about 1 kcal/mol in both cases
suggests that T315I mutation
may be responsible for the loss of binding at these moieties.
Imatinib with WT Imatinib with mutant
32
On the other hand, moiety I contributed more favorably to binding
for imatinib bound to
the mutant = 1.1 kcal/mol) than it did to imatinib with WT = 0.74
kcal/mol).
Notably, moiety V, which forms a hydrogen bond with residue Thr 315
on WT, did not
contribute significantly favorably or unfavorably to binding in
either WT or mutant Abl.
Figure 13: Component analysis of ponatinib for favorable
contribution. The Structure of ponatinib colored by atom type. Abl
residues that form hydrogen bonds are shown in yellow. The
energetic contributions of moieties that form hydrogen bonds with
the WT and mutant Abl residues are shown. The contribution of a
moiety is given by a value. Blue boxes represent favorable moieties
with a value greater than 1. None of the moieties shown contributed
unfavorably to binding ( < − .
Component analysis shows that many moieties of ponatinib that form
hydrogen bonds
with Abl residues contributed favorably to binding. Similarly to
imatinib, the moiety that
interacts with Glu 286 and Asp 381 (moiety IV), contributed most
favorably to binding in WT
( = 3.51 kcal/mol). Unlike imatinib, the binding affinity of the
moiety improved by 0.21
kcal/mol for ponatinib bound to mutant Abl (= 3.72 kcal/mol).
Ponatinib with WT Ponatinib with mutant
33
Ponatinib showed a slight loss of binding at moiety VII, which
forms hydrogen bond with
Met 318 when bound to mutant Abl, with a decrease in binding
contribution of 0.32 kcal/mol
while the contribution of moiety I increased by 0.68 kcal/mol in
mutant compared to that of
ponatinib with WT. As was the case with imatinib, the T315I
mutation may also be responsible
in affecting binding at these moieties.
All other moieties including moiety VI near residue 315 did not
contribute either
favorably or unfavorably to binding.
The hypothetical, optimal charge distribution for maximum binding
affinity
We carried out charge optimization to determine the charge
distribution that minimizes the
electrostatic binding free energy and therefore maximizes the
binding affinity of the drug for
Abl. Additionally, sensitivity analysis was carried out to
determine the impact of atoms' charge
values on the electrostatic binding free energy. The effect of
charge optimization on the overall
electrostatic binding free energy is shown in Table 3, and optimal
charge distributions are shown
on Figures 14 and 15 for imatinib and ponatinib respectively.
Imatinib Ponatinib
Table 3. Charge optimization and electrostatic binding free energy.
Charge optimization minimizes the electrostatic binding energy
producing the best possible electrostatic contribution to the
binding energy. The electrostatic binding energy improved by ~8-9
kcal/mol for imatinib and ~6-7 kcal/mol for ponatinib.
34
Imatinib
Figure 14. Charge optimization and sensitivity analysis of imatinib
with WT and mutant Abl. A) Charge distribution before optimization.
Charges were constrained to range from 1.0 e (blue) to -1.0 e
(red). B) Charge differences between optimal and original charge
distribution of imatinib bound to WT. C) Charge differences of
imatinib bound to mutant. Red indicates atoms that are too positive
in the original drug and need to be more negative to be optimal.
Blue indicates atoms that are too negative as they are and need to
be more positive to be optimal while white is for optimal atoms.
Radii of atoms in B and C indicate the sensitivity of the binding
free energy to the atoms’ charges with larger atoms yielding
greater sensitivity. The root mean square deviation of optimal
charge from original is also shown, in units of elementary
charge.
Charge optimization improved the electrostatic binding free energy
by approximately 8-9
kcal/mol in imatinib bound to WT and mutant Abl and by about 6-7
kcal/mol in ponatinib with
WT and mutant.
Charge optimization of imatinib resulted in a highly charged
methylpiperazine (moiety I)
with some H atoms of the methyl group shown to be too positive to
be optimal and C atoms
shown to be too negative to be optimal for binding. In particular,
the binding energy is shown to
RMSD = 0.37 RMSD = 0.38
Charges
35
be highly sensitive to changes in the partial charge of the H atom
on the protonated N29. This H
is shown to be slightly too positive for optimal binding while N29
is optimal for binding. The
binding energy is also somewhat sensitive to the H atom's charges
of moiety V near residue 315.
This H is shown is also slightly too positive for optimal binding.
The N atom of the same moiety
is too negative for optimal binding and its sensitivity value
suggests that its charge does not
greatly affect the binding energy.
Atoms of moiety III of imatinib that form hydrogen bonds with Glu
286 and Asp 381 are
optimal for imatinib bound to WT while O30 and N20 are too negative
for imatinib bound to the
mutant Abl. Other atoms that are shown to be far from their optimal
values for WT and mutant
including C and N atoms of moiety VI. The C atom is too positive
while the two N atoms are too
negative for optimal binding.
The results also show that the atoms in moieties II, IV and VII are
relatively close to their
optimal charge in both WT and mutant Abl.
36
Ponatinib
Figure 15. Charge optimization and sensitivity analysis of
ponatinib with WT and mutant Abl. A) Original charge distribution
before optimization. Charges were constrained to lie between -1.0 e
(red) to 1.0 e (blue). B) Charge differences between optimal and
original charge distribution of ponatinib bound to WT. C) Charge
differences of ponatinib bound to mutant. Red indicates atoms that
are too positive in the original drug and need to be negative to be
optimal; blue indicates atoms that are too negative as they are and
need to be positive to be optimal, while white is for optimal
atoms. Radii of atoms in B and C indicate the sensitivity of the
binding free energy to the atoms’ charges with larger atoms
yielding greater sensitivity. The root mean square deviation of
optimal charge from original is also shown, in units of elementary
charge.
Charge optimization of the methylpiperazine moiety (moiety I)
yielded a somewhat
similar optimal charge distribution for WT and mutant complexes,
except for one H atom of the
methyl group which is too positive for ponatinib bound to WT and
shown to be optimal for
mutant. σevertheless, the atom’s small radius suggests that the
change of its charges would not
necessarily affect the overall binding energy.
In both WT and mutant, the differences in optimal and original
charge distributions
reveal that the carbon atom of moiety III is far from its optimal
charge, and it needs to be more
B. WT C. Mutant A. Original
Charges
37
negative to be optimal while the F atoms of the same moiety are
slightly too negative for optimal
binding. The C atom of moiety IV that interacts with Glu 286 and
Asp 381 is slightly too
negative to be optimal in ponatinib bound to WT, but optimal in
mutant with similar sensitivity.
Interestingly, the electrostatic binding free energy is highly
sensitive to the charges of
atoms in moiety V, which is also found in imatinib, as well as to
the charges of atoms in moiety
VII which, like moiety VII of imatinib, interacts with residue Met
318.
The triple bond of moiety VI is also shown to be optimal and
binding energy is only
slightly sensitive to charges of its atoms.
Imatinib Ponatinib
I, IV & VII Gain in I,
IV & VII
VII
VII
Optimal II, III, IV, V, VII II, III, IV, V &
VII
IV, V & VI IV, V & VI
Not Optimal I , VI I, VI I, II & VII I, II & VII
Table 4. Summary of component analysis, charge optimization and
sensitivity analysis results for
imatinib and ponatinib bound to WT and mutant Abl. Favorable
moieties have a > 1 contribution to the overall binding energy.
Electrostatic binding free energy is sensitive to the changes of
atoms within moieties listed in “Sensitive atoms” in the
Table.
The GROMACS structure is a reasonable starting structure for MD
simulations
To assess the robustness of a subset of our results above to the
conformational dynamics, we
carried out a 150 ns MD simulation using ponatinib bound to WT Abl.
The MD simulation was
38
carried out on a structure prepared in GROMACS using united atoms
radii and GROMACS
charges (herein referred to as “GROMACS structure”) whereas the
results shown above were
using the PARSE radii and charges, which have been parameterized
especially for continuum
electrostatic calculations40 (and will be referred to as “PARSED
structure”). In order to generate
a proper “static” control to which we can compare our dynamical
analyses, we repeated charge
optimization within the continuum electrostatic framework using the
GROMACS structure.
Table 5 and Figure 16 show a comparison of electrostatic binding
free energy and charge
optimization respectively between the PARSE and GROMACS starting
structures.
Electrostatic Binding Free Energy GROMACS PARSED
(kcal/mol) 8.90 7.89
(kcal/mol) 2.11 0.87 − - 6.79 - 7.02
Table 5. Charge optimization results comparing GROMACS and PARSED
structures in salt
concentration of 0.145M. The results show that both structures have
similar electrostatic binding free
energies and charge optimization improves binding energy in both
cases.
Although the GROMACS structure uses “united atoms”, the original
electrostatic free
energy was similar to the PARSED “all atoms” structure while PARSED
had a smaller optimal
electrostatic binding free energy
39
Figure 16. Charge optimization results comparing the difference
between optimal charge distribution and original charge
distribution of GROMACS and PARSED structures. Optimal charges were
not constrained in this case, as they were previously, in order to
make sure that the observed robustness is not an artifact of these
constraints. Red indicates atoms that are too positive in the
original drug and need to be negative to be optimal; blue indicates
atoms that are too negative as they are and need to be positive to
be optimal, while white is for optimal atoms. Unconstrained charge
optimization of PARSED structure yielded some charges that were
greater than 1 and –1 for moiety I. For ease visualization, we
colored these atoms 1 (blue) and -1 (red). These atoms were also
excluded in RMSD calculation.
Both structures show a similar optimal charge distribution of the
drug especially for
moieties II, IV, V VI and VII. For example, partial atomic charges
on the triple bond and moiety
III are shown to be optimal in both structures while some atoms of
moiety IV and VII are
similarly shown to be far from their optimal value in GROMACS and
PARSED.
As is the case with the PARSED structure, some atoms of moiety I in
GROMACS are
shown to be far from their optimal charges, specifically the C38
(refer to Figure 18A), which is
too positive for optimal binding in GROMACS. This charge may
correlate with a very red H
atom on PARSED, which suggests that it is too positive for optimal
binding. It is interesting to
RMSD = 0.32 RMSD = 0.35
40
note that the deviations from optimality in moiety I for atoms in
the six-membered ring are
actually “inverted” when comparing GROMACS and PARSED structures –
atoms that are too
positive in GROMACS are too negative in PARSE within this moiety.
This could partially be
consequence of GROMACS using a united atom model for the methyl
group and PARSE not
doing so – the loss of flexibility in creating additional dipoles
and polar groups in the
GROMACS optimal charge distribution could have a “ripple” effect,
leading to this inverted
pattern.
Though there are some differences, there are still several overall
similarities between the
PARSED and GROMACS results, and the GROMACS structure is a
reasonable starting
structure for MD simulation.
Stability of the system during MD simulation
After carrying out the MD simulation, we determined the stability
of the dynamic system by
using the GROMACS analysis tools to calculate the Root Mean Square
Deviation (RMSD) of
the drug and the protein from the reference crystal structure. We
also determined the Root Mean
Square Fluctuation (RMSF) of atoms on the drug relative to the
reference minimized ponatinib
structure. The results are shown in Figures below.
41
Figure 17. System Stability. The RMSD plot shows that ponatinib
(red) equilibrates quickly after 5 ns while Abl (blue) does not
become stable until approximately 100 ns. The RMSD stability of the
drug through the simulation indicates that there is little mobility
of the molecule within the Abl binding pocket.
The RMSD analysis indicated that the drug was equilibrated after 5
ns while the protein
became equilibrated only after 100 ns. The average RMSD for the
drug was approximately 0.15
nm and that of the protein was 0.25 nm from the aligned minimized
reference crystal structure.
The results suggested that the system needs at least 100 ns of
simulation to stabilize.
42
Figure 18. RMSF of ponatinib’s atoms averaged over 50 ns. Radii of
atoms in B represent the RMSF value of each atom (scaled by a
factor of 50 for ease of visualization). B shows that F34- 36 of
moiety III fluctuated the most during the simulation. Slightly
higher fluctuations were also seen in all four hydrogen atoms of
moiety VII and the methyl group of moeity I i.e., C38.
Figure 14 shows that moiety III was the most mobile area of the
drug during the
simulation with an average RMSF value of 0.12 nm. Notably, all
hydrogen atoms of moiety VII
and the methyl group of moiety I were also shown to fluctuate
during the simulation. We will
later investigate the effect of these fluctuations on the
robustness of the optimal charge
distribution.
The optimal charge distribution is somewhat affected by the
conformational dynamics of
the complex.
Charge optimization was carried out on 20 trajectory snapshots
taken between 100 ns and 150 ns,
sampled every 2.5 ns. The mean optimal charge distribution of the
samples (herein referred to as
“dynamic structure”) was determined and compared to the charge
distribution of the static
model. Figure 16 below shows the comparison between optimal charge
distributions of the static
model and dynamic structure.
8.87
8.42
1.0
2.34 − - 6.80 - 4.01 1.0
Table 19. Charge optimization and electrostatic free energy. The
mean of the dynamic structure was very similar to that of the
static structure. Charge optimization showed a greater improvement
of binding energy in the static structure than it did in the mean
dynamic structure; mean of dynamic structure was greater than
optimal of the static structure.
44
Figure 20. Charge optimization and sensitivity analysis in static
and dynamic structure B) Charge differences between optimal and
original charges of the static structure. C) Charge differences
between mean optimal and original charge distribution for the
dynamic structure. Red indicates atoms that are too positive in the
original drug and need to be negative to be optimal. Blue indicates
atoms that are too negative as they are and need to be positive to
be optimal, while white is for optimal atoms. Radii of atoms in B
and C indicate the sensitivity of the binding free energy to the
atoms’ charges, with large atoms yielding greater
sensitivity.
Charge optimization on the dynamic structure yielded a more
hydrophobic drug
compared to the static optimization. However, charge optimization
of moiety VII yielded similar
optimal charge distributions in the mean dynamic and static
structures; N19 is shown to be too
negative for optimal binding while σ20 is too positive. The change
of these atoms’ charges has a
slight effect on the overall electrostatic binding free energy in
both cases.
Additionally, the optimal charge distribution of the triple bond
(moiety VI) is robust to
conformational changes and the electrostatic binding free energy is
only slightly sensitive to the
RMSD = 0.32 RMSD = 0.20
45
change of its atom charges. Similarly, O28 (moiety IV) and H39
(moiety I) are shown to be
optimal and robust to dynamics, however, change of their charges
affect the overall binding
energy. Interestingly, the binding energy is also very sensitive
toward the change of charges of
H6, H7 and H28 (moiety V) and moiety VII in both cases and these
atoms are close to their
optimal charges and remain so during the simulation.
Notably, N29 is optimized to be negatively charged in static model,
in disagreement with
the dynamic model, which shows that the atom is on average optimal
during the simulation.
Additionally, H29 (moiety IV) is shown to be slightly too negative
for optimal binding in static
model while slightly too positive in the dynamic structure, and the
overall electrostatic binding
energy in both cases is affected by the change of its charge.
Additionally, charge optimization of
the static model suggested that C38 (moiety I) is too positive in
disagreement with the results
from the dynamic structure, which shows that the atom is on average
optimal during
conformational changes.
Interestingly, F34, F35 and F36 (moiety III) are not only optimal
and robust to
conformational change, but also slightly affect the overall
electrostatic binding free energy. C33
on the other hand, is optimized to be more negative in the mean
dynamic structure.
There is no correlation between the standard deviation of an atom’s
optimal charge in the
dynamic model and its flexibility in the binding pocket
As a first step toward understanding the relationship between
conformation and design
predictions, we plotted the standard deviation of optimal charges
for each atom vs. its RMSF in
the MD simulation to test the hypothesis that atoms that fluctuated
more would have a greater
variation in optimal charge.
46
Figure 21. There is no clear relationship between standard
deviation of the atom’s optimal charge and its RMSF value during
the simulation. Atom radius in A indicates the standard deviation
in the optimal charge of the atom while atom radius in B represents
the RMSF for each atom.
47
Figure 21 shows that most atoms did not fluctuate much from their
reference structure
and did not show much variation in their optimal charges. Highly
flexible atoms (F35- F36) are
shown to have small standard deviation while N39 that has the
largest variation in its optimal
charge has a median RSMF value. Thus our results showed that there
is no clear correlation
between the standard deviation of the atom and its flexibility in
the binding pocket.
48
5. Discussion
In the first part of this study, we analyzed the electrostatic
component of the binding free energy
between two leukemia drugs, imatinib and ponatinib, and their
biological target, the Abl kinase,
using component analysis, charge optimization and sensitivity
analysis within the continuum
electrostatic framework. Component analysis enabled us to determine
the contribution of drug
moieties to the binding affinity, while by carrying out charge
optimization, we determined the
hypothetical, optimal charge distribution of the drug that will
have the maximum possible
binding affinity.
Our electrostatic energy results showed that imatinib bound the
T315I mutant with a
higher electrostatic binding energy, by nearly 2 kcal/mol when
compared to WT. These results
are in good agreement with previous computational studies
suggesting that the worsening of
electrostatic interactions is partly responsible for the loss of
imatinib affinity towards T315I
mutant27.
More specifically, the resistance of T315I mutants to imatinib was
once hypothesized to
be caused mainly by the loss of a hydrogen bond between the
“gatekeeper”, Thr 315 and
imatinib (at moiety V) due to substitution of Thr by a nonpolar
Ile. Interestingly, our component
analysis and charge optimization show similar results for imatinib
bound to WT and imatinib
bound to T315I mutant at this moiety. Component analysis of both
complexes shows that this
moiety contributes neither favorably nor unfavorably to binding,
and charge optimization yielded
an optimal H atom on the moiety whose change in charge would have a
great effect on the
overall binding energy. Also, in both cases the N atom of the
moiety is not optimal, but rather, it
is too negative for optimal binding. Therefore, the direct
interaction of imatinib with residue 315
49
did not seem to fully explain the energetic differences between WT
and mutant Abl and thus
does not explain why resistance occurs. This is explored further
below.
Some of the imatinib moieties were shown to lose their binding
affinities when bound to
the T315I mutant. A good example is the loss of binding affinity
for the pyridine moiety (moiety
VII, Figure 10) that forms a hydrogen bond with Met 318. Other
moieties with noticeable loss of
binding affinities in mutant Abl included those that interact with
residues Glu 286 and Asp 381.
This loss of binding suggests that the T315I mutation may in fact
alter interactions of the drug
with other moieties. Our results agree well with several recent
computational studies that have
shown that induced conformational change of the binding site to
accommodate the bulky Ile 315
side chain causes a loss of binding affinity of other remote
residues which in turn leads to drug
resistance27, 37. For example, in a MD simulation study, Zhou et
al., predicted loss of binding due
to a slight outward displacement of the imatinib moiety from the
binding pocket to accommodate
Ile 31524.
Ponatinib, on the other hand, was designed to have a linear triple
bond moiety at this
position (moiety VI) in order to surpass the interaction with the
residue altogether. Thus, the
T315I mutation should not significantly affect its binding
affinity. Our analysis of the
electrostatic binding energy shows that indeed, ponatinib binds to
both WT and mutant Abl with
similar binding affinities. Additionally, component analysis shows
that this moiety contributes
neither favorably nor unfavorably to the overall binding. Charge
optimization shows that the
moiety is optimal and the binding energy is insensitive to the
changes of its charges. Notably, our
MD analysis of ponatinib bound to WT showed that the optimal charge
distribution of the moiety
is robust to conformational dynamics of the complex.
50
Consequently, our analysis shows that the design for better CML
inhibitors should
consider conserving this triple bond as one way to maintain the
drug’s activity towards the T315I
mutant. Interestingly, this group is preserved in PF-114 (Figure
6), a recent drug intended to
improve upon the selectivity profile of ponatinib.
Interestingly, the methylpiperazine ring that forms hydrogen bonds
with Ile 360 and His
361 is shown to contribute more favorably to binding in imatinib
bound to the T315I mutant than
it does to the WT. Ponatinib also shows a gain in binding affinity
at this moiety and the moiety
that interacts with Glu 286 and Asp 381 (moiety IV in Figure 9).
This gain in binding suggests
that T315I mutation may also be cause favorable conformational
changes in the binding pocket.
However, different moieties are affected for different drug-protein
complexes.
Structure-guided design of new CML drugs aims to optimize several
moieties of earlier
drugs to improve their potency toward the T315I mutant. For
example, the pyridine ring (moiety
VII) of imatinib that forms hydrogen bond with Met 318 was changed
to the imidazole
pyridazine in ponatinib to improve the binding affinity towards the
T315I mutant27. As
mentioned earlier, the binding affinity at this moiety is lost when
imatinib binds to the T315I
mutant.
On the other hand, our results show that the moiety in ponatinib
contributes favorably to
binding and changing its charges would have great effect on the
overall electrostatic binding
energy. Charge optimization resulted in more positively charged N19
and C21 atoms and a more
negatively charged N20 atom for optimal binding. MD analysis
revealed that the optimal charges
of N19 and N20 atoms are robust to conformational dynamics, while
C21 remains optimal in the
dynamic model.
51
Studies have also associated this moiety to the drug’s selectivity
in binding27. For
example, the design of PF-114 involved replacement of N19 with a C
atom to disrupt hydrogen
bond formation in active sites of some off-target kinases thus
improving its selectivity profile.
Therefore, our study suggests that further optimization of the
group to make better interactions
with residue Met 318 might further increase the potency of the
future CML drugs within the
constraints of maintaining selectivity.
In addition to making hydrogen bonds with Ile 360 and His 361, MD
simulations showed
that methylpiperazine (moiety I) increases the drug's potency and
molecular recognition 27, 80.
Expectedly, the hydrogen atom of the protonated N (N29 on imatinib
and N39 on ponatinib) is
shown to have great effect in electrostatic binding energy because
as we have seen in our study,
and other previous studies13-15, protonation at this position
improves electrostatic binding
interactions. The hydrogen atom is not optimal for imatinib bound
to WT or mutant and its
optimal charge varies during conformational dynamics in the case of
ponatinib bound to WT.
Future studies should carry out MD simulations on ponatinib bound
to the mutant to see if
comparisons between WT and mutant interactions observed with the
static structures are robust
to conformational dynamics.
Furthermore, our study showed that the moieties that interact with
Glu 286 and Asp 381
in either imatinib or ponatinib have the greatest contribution to
the electrostatic binding energy.
Our results are in good agreement with the previous Molecular
Mechanics/Poisson Boltzmann
surface area study of binding energy that suggested that Glu 286
interactions with the NH group
of the moiety is one of the strongest contact points81. In
addition, charge optimization results
showed that charges of the atoms of this moiety affect the binding
energy.
52
N20, O30, C and the hydrogen atom in imatinib bound to WT are
optimal while only the
oxygen atom within the moiety is optimal in imatinib bound to
mutant. O28 and the hydrogen
atom of this moiety are also optimal in ponatinib except C27 of WT
and N29 of the mutant.
However, MD analysis of ponatinib bound to the WT shows that on
average, all atoms of this
moiety are in fact optimal during conformational changes.
Our results show that qualitatively, the binding energy is also
sensitive towards the
change of charges of the trifluoromethyl group (moiety III on
ponatinib). The moiety is also
shown to have a negligible contribution towards binding. Charge
optimization of the group
reveals that the three F atoms are not optimal and need to be more
positive for optimal binding
while the C atom is optimized to have a negative charge. The
function of the trifluoromethyl
group is to increases the solubility and lipophilicity of the drug
for easier membrane
permeability82. Thus, the design for better drugs may consider
altering it for better electrostatics
only if it is possible to maintain these other qualities.
Our study compared the electrostatic binding energetics of the
crystal structure
conformation and those at different conformations obtained from
ponatinb-WT MD simulations,
assuming rigid binding in both cases. Although simulations results
strongly rely on the quality of
the starting model, our results show that the optimal charge
distributions of many atoms did not
change much during the simulation and are thus robust to
conformational dynamics of the
complex. Such atoms include the hydrogen atom of the protonated
methypiperazine moiety, the
highly flexible F atoms of the trifluoromethyl moiety, the O atom
of moiety IV, and he triple
bond and hydrogen atoms of moieties V and VII. In addition, the
variations in optimal charge
values did not relate to the degree of spatial fluctuations of drug
atoms. For instance, the fluorine
atoms, which were shown to have the most flexibility, had small
standard deviations in their
53
optimal charges, while N39, which showed the most variation in its
optimal charge, did not show
large fluctuations.
Interestingly, the average optimal charge distribution of the
conformational ensemble
yielded a more hydrophobic drug. A study on binding specificity
suggested that hydrophobic
ligands tend to bind more generally to multiple partners with equal
affinity than charged ligands
i.e. they are more promiscuous83. In deed ponatinib has been shown
to bind to multiple targets
including all of the clinically active mutants39. Unfortunately,
the lack of selectivity is also
associated to the toxicity level of the drug33. PF-114 on the other
hand, was designed to have a
better selectivity profile35. It would be interesting to see if the
optimal charge distribution of PF-
114 is more charged as compared to ponatinib. Future work should
carry out similar MD
analyses on PF-114 to determine its average optimal charge
distribution and perform a
comparison study with ponatinib.
It is important to note that although we looked only at the
electrostatic component
of binding to predict and analyze the binding of CML drugs, other
components of binding
energy, such as van der Waals interactions, contribute to the
relative binding energy of these
drugs37, 39. Furthermore, our study assumed rigid binding even for
the dynamic model. As
discussed earlier, conformational changes of the protein and drug
heavily influence their binding
affinities.
Additionally, there is no available crystal structure of imatinib
bound to the T315I mutant
Abl. Thus, the relatively crude model of the complex modeled using
CHARMM from the WT-
imatinib crystal structure limited our analysis of the complex. By
carrying out MD simulations
using our crude model as a starting point, we plan to overcome this
current limitation. Also, as
discussed earlier, MD simulation analysis was also limited by the
starting structure and the
54
parameter set used – understanding the robustness of these model
inputs can also be potential
future work.
Despite these limitations, our study offers a tool to qualitatively
and quantitatively
understand the determinants of binding in this system, and it
provides insights and predictions
that can be tested and corroborated by experiments and other
computational studies. We hope
that our study will provide more insights into understanding and
optimizing the electrostatic
component of the binding energy and will aid in the design of
improved future CML drugs.
55
References
1. Zamecnikova, A., Targeting the BCR-ABL tyrosine kinase in
chronic myeloid leukemia as a model of rational drug design in
cancer. Expert Rev of Hematol 2010, 3 (1), 45-56.
2. Sawyers, C. L.; Druker, B., Tyrosine kinase inhibitors in
chronic myeloid leukemia. Cancer J Sci Am
1999, 5 (2), 63-9. 3. Liu, Y.; Gray, N. S., Rational design of
inhibitors that bind to inactive kinase conformations. Nature
Chem Biol 2006, 2 (7), 358-364. 4. Nagar, B.; Bornmann, W. G.;
Pellicena, P.; Schindler, T.; Veach, D. R.; Miller, W. T.;
Clarkson, B.;
Kuriyan, J., Crystal structures of the kinase domain of c-Abl in
complex with the small molecule inhibitors PD173955 and imatinib
(STI-571). Cancer Res 2002, 62 (15), 4236-4243.
5. Bikker, J. A. B., Natasja; Wissner, A.; Mansour, T. S., Kinase
Domain Mutations in Cancer: Implications for Small Molecule Drug
Design Strategies. J Med Chem 2009, 62 (15). 4236-4243.
6. Goldman, J. M.; Melo, J. V., Targeting the BCR-ABL tyrosine
kinase in chronic myeloid leukemia. N Engl J Med 2001, 344 (14),
1084-1086.
7. Reddy, E. P.; Aggarwal, A. K., The ins and outs of bcr-abl
inhibition. Genes Cancer 2012, 3 (5-6), 447-54.
8. Schindler, T.; Bornmann, W.; Pellicena, P.; Miller, W. T.;
Clarkson, B.; Kuriyan, J., Structural mechanism for STI-571
inhibition of abelson tyrosine kinase. Science 2000, 289 (5486),
1938-42.
9. Cowan-Jacob, S. W.; Fendrich, G.; Floersheimer, A.; Furet, P.;
Liebetanz, J.; Rummel, G.; Rheinberger, P.; Centeleghe, M.; Fabbro,
D.; Manley, P. W., Structural biology contributions to the
discovery of drugs to treat chronic myelogenous leukaemia. Acta
Crystallogr D Biol Crystallogr
2007, 63 (Pt 1), 80-93. 10. Eck, M. J.; Manley, P. W., The
interplay of structural information and functional studies in
kinase
drug design: insights from BCR-Abl. Curr Opin Cell Biol 2009, 21
(2), 288-95. 11. Liu, Y.; Shah, K.; Yang, F.; Witucki, L.; Shokat,
K. M., A molecular gate which controls unnatural
ATP analogue recognition by the tyrosine kinase v-Src. Bioorg Med
Chem 1998, 6 (8), 1219-26. 12. Liu, X.; Kung, A.; Malinoski, B.;
Prakash, G.; Zhang, C., Development of Alkyne-Containing
Pyrazolopyrimidines To Overcome Drug Resistance of Bcr-Abl Kinase.
J Med Chem 2015, 58 (23), 9228-9237.
13. Lin, Y.; Meng, Y.; Jiang, W.; Roux, B., Explaining why Gleevec
is a specific and potent inhibitor of Abl kinase. Proc Natl Acad
Sci USA 2013, 110 (5), 1664-1669.
14. Aleksandrov, A.; Simonson, T., Molecular Dynamics Simulations
Show That Conformational Selection Governs the Binding Preferences
of Imatinib for Several Tyrosine Kinases. J Biol Chem
2010, 285 (18), 13807-13815. 15. Aleksandrov, A.; Simonson, T., A
Molecular Mechanics Model for Imatinib and Imatinib:Kinase
Binding. J Comput Chem 2010, 31 (7), 1550-1560. 16. Grante, I.;
Actins, A.; Orola, L., Protonation effects on the UV/Vis absorption
spectra of imatinib: A
theoretical and experimental study. Spectrochimica Acta Part
A-Molecular and Biomolecular
Spectroscopy 2014, 129, 326-332. 17. Szakacs, Z.; Beni, S.; Varga,
Z.; Orfi, L.; Keri, G.; Noszal, B., Acid-base profiling of
imatinib
(Gleevec) and its fragments. J Med Chem 2005, 48 (1), 249-255. 18.
Druker, B. J.; Guilhot, F.; O'Brien, S. G.; Gathmann, I.;
Kantarjian, H.; Gattermann, N.; Deininger,
M. W. N.; Silver, R. T.; Goldman, J. M.; Stone, R. M.; Cervantes,
F.; Hochhaus, A.; Powell, B. L.; Gabrilove, J. L.; Rousselot, P.;
Reiffers, J.; Cornelissen, J. J.; Hughes, T.; Agis, H.; Fischer,
T.; Verhoef, G.; Shepherd, J.; Saglio, G.; Gratwohl, A.; Nielsen,
J. L.; Radich, J. P.; Simonsson, B.; Taylor, K.; Baccarani, M.; So,
C.; Letvak, L.; Larson, R. A.; Investigators, I., Five-year
follow-up of patients receiving imatinib for chronic myeloid
leukemia. N Engl J Med 2006, 355 (23), 2408-2417.
19. Santos, F. P. S.; Kantarjian, H.; Quintas-Cardama, A.; Cortes,
J., Evolution of Therapies for Chronic Myelogenous Leukemia. Cancer
J 2011, 17 (6), 465-476.
56
20. Lu, X. Y.; Zhang, Z.; Ren, X. M.; Pan, X. F.; Wang, D. P.;
Zhuang, X. X.; Luo, J. F.; Yu, R. M.; Ding, K., Hybrid pyrimidine
alkynyls inhibit the clinically resistance related Bcr-Abl(T315I)
mutant. Bioorg Med Chem Lett 2015, 25 (17), 3458-3463.
21. Hughes, T.; Deininger, M.; Hochhaus, A.; Branford, S.; Radich,
J.; Kaecla, J.; Baccarani, M.; Cortes, J.; Cross, N.; Druker, B.;
Gabert, J.; Grimwade, D.; Hehlmann, R.; Kamel-Reid, S.; Lipton, J.;
Longtine, J.; Martinelli, G.; Saglio, G.; Soverini, S.; Stock, W.;
Goldman, J., Monitoring CML patients responding to treatment with
tyrosine kinase inhibitors: review and recommendations for
harmonizing current methodology for detecting BCR-ABL transcripts
and kinase domain mutations and for expressing results. Blood 2006,
108 (1), 28-37.
22. Branford, S.; Rudzki, Z.; Walsh, S.; Grigg, A.; Arthur, C.;
Taylor, K.; Herrmann, R.; Lynch, K.; Hughes, T., High frequency of
point mutations clustered within the adenosine triphosphate-binding
region of BCR/ABL in patients with chronic myeloid leukemia or
Ph-positive acute lymphoblastic leukemia who develop imatinib
(STI571) resistance. Blood 2002, 99 (9), 3472-3475.
23. von Bubnoff, N.; Schneller, F.; Peschel, C.; Duyster, J.,
BCR-ABL gene mutations in relation to clinical resistance of
Philadelphia-chromosome-positive leukaemia to STI571: a prospective
study. Lancet 2002, 359 (9305), 487-491.
24. Zhou, T.; Commodore, L.; Huang, W.; Wang, Y.; Thomas, M.;
Keats, J.; Xu, Q.; Rivera, V.; Shakespeare, W.; Clackson, T.;
Dalgarno, D.; Zhu, X., Structural Mechanism of the Pan-BCR-ABL
Inhibitor Ponatinib (AP24534): Lessons for Overcoming Kinase
Inhibitor Resistance. Chem Biol
Drug Des 2011, 77 (1), 1-11. 25. Frankfurt, O.; Licht, J. D.,
Ponatinib-A Step Forward in Overcoming Resistance in Chronic
Myeloid
Leukemia. Clin Cancer Res 2013, 19 (21), 5828-5834. 26. Deininger,
M.; Buchdunger, E.; Druker, B. J., The development of imatinib as a
therapeutic agent for
chronic myeloid leukemia. Blood 2005, 105 (7), 2640-53. 27. Miller,
G. D.; Bruno, B. J.; Lim, C. S., Resistant mutations in CML and
Ph(+)ALL - role of
ponatinib. Biologics 2014, 8, 243-54. 28. Shah, N. P.; Tran, C.;
Lee, F. Y.; Chen, P.; Norris, D.; Sawyers, C. L., Overriding
imatinib resistance
with a novel ABL kinase inhibitor. Science 2004, 305 (5682),
399-401. 29. le Coutre, P.; Ottmann, O.; Giles, F.; Kim, D.;
Cortes, J.; Gattermann, N.; Apperley, J.; Larson, R.;
Abruzzese, E.; O'Brien, S.; Kuliczkowski, K.; Hochhaus, A.; Mahon,
F.; Saglio, G.; Gobbi, M.; Kwong, Y.; Baccarani, M.; Hughes, T.;
Martinelli, G.; Radich, J.; Zheng, M.; Shou, Y.; Kantarjian, H.,
Nilotinib (formerly AMN107), a highly selective BCR-ABL tyrosine
kinase inhibitor, is active in patients with imatinib-resistant or
-intolerant accelerated-phase chronic myelogenous leukemia. Blood
2008, 111 (4), 1834-1839.
30. O'Hare, T.; Shakespeare, W. C.; Zhu, X.; Eide, C. A.; Rivera,
V. M.; Wang, F.; Adrian, L. T.; Zhou, T.; Huang, W. S.; Xu, Q.;
Metcalf, C. A.; Tyner, J. W.; Loriaux, M. M.; Corbin, A. S.;
Wardwell, S.; Ning, Y.; Keats, J. A.; Wang, Y.; Sundaramoorthi, R.;
Thomas, M.; Zhou, D.; Snodgrass, J.; Commodore, L.; Sawyer, T. K.;
Dalgarno, D. C.; Deininger, M. W.; Druker, B. J.; Clackson, T.,
AP24534, a pan-BCR-ABL inhibitor for chronic myeloid leukemia,
potently inhibits the T315I mutant and overcomes mutation-based
resistance. Cancer Cell 2009, 16 (5), 401-12.
31. Cortes, J. E.; Kim, D. W.; Pinilla-Ibarz, J.; le Coutre, P.;
Paquette, R.; Chuah, C.; Nicolini, F. E.; Apperley, J. F.; Khoury,
H. J.; Talpaz, M.; DiPersio, J.; DeAngelo, D. J.; Abruzzese, E.;
Rea, D.; Baccarani, M.; Mueller, M. C.; Gambacorti-Passerini, C.;
Wong, S.; Lustgarten, S.; Rivera, V. M.; Clackson, T.; Turner, C.
D.; Haluska, F. G.; Guilhot, F.; Deininger, M. W.; Hochhaus, A.;
Hughes, T.; Goldman, J. M.; Shah, N. P.; Kantarjian, H.;
Investigators, P., A Phase 2 Trial of Ponatinib in Philadelphia
Chromosome-Positive Leukemias. N Engl J Med 2013, 369 (19),
1783-1796.
32. Berman, E., Where Exactly Does Ponatinib Fit in Chronic
Myelogenous Leukemia? J Natl Compr
Cancer Netw 2014, 12 (11), 1615-1620. 33. Talbert, D. R.; Doherty,
K. R.; Trusk, P. B.; Moran, D. M.; Shell, S. A.; Bacus, S., A
Multi-
parameter In Vitro Screen in Human Stem Cell-Derived Cardiomyocytes
Identifies Ponatinib- Induced Structural and Functional Cardiac
Toxicity. Toxicol Sci 2015, 143 (1), 147-155.
57
34. Mian, A.; Rafiei, A.; Metodieva, A.; Haberbosch, I.; Zeifman,
A.; Titov, I.; Stroylov, V.; Stroganov, O.; Novikov, F.; Ottmann,
O. G.; Chilov, G.; Ruthardt, M., PF-114, a Novel Selective Pan
BCR/ABL Inhibitor Targets The T315I and Suppress Models Of Advanced
Ph+ ALL. Blood 2013, 122 (21), 3907-3907.
35. Mian, A.; Rafiei, A.; Haberbosch, I.; Zeifman, A.; Titov, I.;
Stroylov, V.; Metodieva, A.; Stroganov, O.; Novikov, F.; Brill, B.;
Chilov, G.; Hoelzer, D.; Ottmann, O.; Ruthardt, M., PF-114, a
potent and selective inhibitor of native and mutated BCR/ABL is
active against Philadelphia chromosome- positive (Ph plus )
leukemias harboring the T315I mutation. Leukemia 2015, 29 (5),
1104-1114.
36. Laurini, E.; Posocco, P.; Fermeglia, M.; Gibbons, D.;
Quintas-Cardama, A.; Pricl, S., Through the open door: Preferential
binding of dasatinib to the active form of BCR-ABL unveiled by in
silico experiments. Mol Oncol 2013, 7 (5), 968-975.
37. Lee, T. S.; Potts, S. J.; Kantarjian, H.; Cortes, J.; Giles,
F.; Albitar, M., Molecular basis explanation for imatinib
resistance of BCR-ABL due to T315I and P-loop mutations from
molecular dynamics simulations. Cancer 2008, 112.
38. Banavath, S., Identification of novel tryosine kinase
inhibitors for drug resistant T315I mutant BCR- ABL: a virtual
screening and molecular dynamics simulations study. 2014.
39. Tanneeru, K.; Guruprasad, L., Ponatinib Is a Pan-BCR-ABL Kinase
Inhibitor: MD Simulations and SIE Study. PLoS ONE 2013, 8 (11),
e78556.
40. Liu, L. Y. Molecular Matchmaking: A Computational Study of the
Electrostatic Interaction Between Chronic Myeloid Leukemia Drugs
and Bcr-Abl Oncoprotein. Wellesley College, Honors Thesis
Collection, 2013.
41. Hiller, L.; Kasili, P.; Radhakrishnan, M., Molecular
Matchmaking: A computational Study of Two Chronic Myeloid Leukemia
Drugs and Their Target, the Bcr-Abl Kinase. Bunker Hill Community
College, Wellesley College
42. Kangas, E.; Tidor, B., Electrostatic specificity in molecular
ligand design. J Chem Phys 2000, 112 (20), 9120-9131.
43. Radhakrishnan, M., Designing electrostatic interactions in
biological systems via charge optimization or combinatorial
approaches: insights and challenges with a continuum electrostatic
framework. Theor Chem Acc 2012, 131 (8).
44. Busby, B.; Oashi, T.; Willis, C.; Ackermann, M.;
Kontrogianni-Konstantopoulos, A.; MacKerell, A.; Bloch, R.,
Electrostatic Interactions Mediate Binding of Obscurin to Small
Ankyrin 1: Biochemical and Molecular Modeling Studies. J Mol Biol
2011, 408 (2), 321-334.
45. Schutz, C.; Warshel, A., What are the dielectric "constants" of
proteins and how to validate electrostatic models? Proteins 2001,
44 (4), 400-417.
46. Baker, N., Poisson-Boltzmann methods for biomolecular
electrostatics. Methods Enzymol 2004, 383, 94-+.
47. Fogolari, F.; Brigo, A.; Molinari, H., The Poisson-Boltzmann
equation for biomolecular electrostatics: a tool for structural
biology. J Mol Recognit 2002, 15 (6), 377-92.
48. Wang, J.; Luo, R., Assessment of linear finite-difference
Poisson-Boltzmann solvers. J Comput
Chem 2010, 31 (8), 1689-98. 49. Li, L.; Li, C.; Sarkar, S.; Zhang,
J.; Witham, S.; Zhang, Z.; Wang, L.; Smith, N.; Petukh, M.;
Alexov,
E., DelPhi: a comprehensive suite for DelPhi software and
associated resources. Bmc Biophysics
2012, 5. 50. Carrascal, N.; Green, D. F., Energetic decomposition
with the generalized-born and Poisson-
Boltzmann solvent models: lessons from association of G-protein
components. J Phys Chem B 2010, 114 (15), 5096-116.
51. Lee, L.; Tidor, B., Barstar is electrostatically optimized for
tight binding to barnase. Nature
Structural Biology 2001, 8 (1), 73-76. 52. Lee, L.; Tidor, B.,
Optimization of electrostatic binding free energy. J Chem Phys
1997, 106 (21),
8681-8690.
58
53. Lee, L.; Tidor, B., Optimization of binding electrostatics:
Charge complementarity in the barnase- barstar protein complex.
Protein Sci 2001, 10 (2), 362-377.
54. Sims, P.; Wong, C.; McCammon, J., Charge optimization of the
interface between protein kinases and their ligands. J Comput Chem
2004, 25 (11), 1416-1429.
55. Karplus, M.; McCammon, J. A., Molecular dynamics simulations of
biomolecules. Nature Structural
Biology 2002, 9 (9), 646-5