Searching new structural scaffolds for BRAF inhibitors. An integrative study using theoretical and experimental techniques

Article abstract of DOI:10.1016/j.bioorg.2019.103125

The identification of the V600E activating mutation in the protein kinase BRAF in around 50% of melanoma patients has driven the development of highly potent small inhibitors (BRAFi) of the mutated protein. To date, Dabrafenib and Vemurafenib, two specifi

Full text of DOI:10.1016/j.bioorg.2019.103125

Accepted Manuscript
Searching New Structural Scaffolds for BRAF Inhibitors. An Integrative Study
using theoretical and experimental techniques
Ludmila E. Campos, Francisco M. Garibotto, Emilio Angelina, Jiri Kos,
Tihomir Tomašič, Nace Zidar, Danijel Kikelj, Tomas Gonec, Pavlina
Marvanova, Petr Mokry, Josef Jampilek, Sergio E. Alvarez, Ricardo D. Enriz
PII:
S0045-2068(19)30532-2
https://doi.org/10.1016/j.bioorg.2019.103125
103125
DOI:
Article Number:
Reference:
YBIOO 103125
Bioorganic Chemistry
To appear in:
Received Date:
Revised Date:
Accepted Date:
10 April 2019
4 July 2019
11 July 2019
Please cite this article as: L.E. Campos, F.M. Garibotto, E. Angelina, J. Kos, T. Tomašič, N. Zidar, D. Kikelj, T.
Gonec, P. Marvanova, P. Mokry, J. Jampilek, S.E. Alvarez, R.D. Enriz, Searching New Structural Scaffolds for
BRAF Inhibitors. An Integrative Study using theoretical and experimental techniques, Bioorganic Chemistry (2019),
doi: https://doi.org/10.1016/j.bioorg.2019.103125
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Searching New Structural Scaffolds for BRAF Inhibitors. An
Integrative Study using theoretical and experimental techniques.
Ludmila E. Camposa, Francisco M. Garibottoa, Emilio Angelinab, Jiri Kosc, Tihomir Tomašičd,
Nace Zidard, Danijel Kikeljd, Tomas Gonece, Pavlina Marvanovae, Petr Mokrye, Josef
Jampilekc,f, Sergio E. Alvareza* and Ricardo D. Enriza*
aFacultad de Química, Bioquímica y Farmacia, Universidad Nacional de San Luis; Instituto
Multidisciplinario de Investigaciones Biológicas (IMIBIO-SL). Ejercito de los Andes 950, (5700)
San Luis, Argentina.
bLaboratorio de Estructura Molecular y Propiedades, Área de Química Física, Departamento
de Química, Facultad de Ciencias Exactas y Naturales y Agrimensura, Universidad Nacional
del Nordeste, Avda. Libertad 5460, (3400) Corrientes, Argentina
cDivision of Biologically Active Complexes and Molecular Magnets, Regional Centre of
Advanced Technologies and Materials, Faculty of Science, Palacky University Olomouc,
Slechtitelu 27, 78371 Olomouc, Czech Republic
dUniversity of Ljubljana, Faculty of Pharmacy, Aškerčeva 7, 1000 Ljubljana, Slovenia
eDepartment of Chemical Drugs, Faculty of Pharmacy, University of Veterinary and
Pharmaceutical Sciences Brno, Palackeho 1, 61242 Brno, Czech Republic
fDepartment of Analytical Chemistry, Faculty of Natural Sciences, Comenius University,
Ilkovicova 6, 84215 Bratislava, Slovakia
Corresponding authors: Ricardo D. Enriz* (denriz@unsl.edu.ar) and Sergio E. Alvarez*
(sealvarez98@gmail.com).
1

Abstract
The identification of the V600E activating mutation in the protein kinase BRAF in around 50%
of melanoma patients has driven the development of highly potent small inhibitors (BRAFi) of
the mutated protein. To date, Dabrafenib and Vemurafenib, two specific BRAFi, have been
clinically approved for the treatment of metastatic melanoma. Unfortunately, after the initial
response, tumors become resistant and patients develop a progressive and lethal disease, making
imperative the development of new therapeutic options. The main objective of this work was to
find new BRAF inhibitors with different structural scaffolds than those of the known inhibitors.
Our study was carried out in different stages; in the first step we performed a virtual screening
that allowed us to identify potential new inhibitors. In the second step, we synthesized and tested
the inhibitory activity of the novel compounds founded. Finally, we conducted a molecular
modelling study that allowed us to understand interactions at the molecular level that stabilize
the formation of the different molecular complexes.
Our theoretical and experimental study allowed the identification of four new structural
scaffolds, which could be used as starting structures for the design and development of new
inhibitors of BRAF. Our experimental data indicate that the most active compounds reduced
significantly ERK½ phosphorylation, a measure of BRAF inhibition, and cell viability. Thus,
from our theoretical and experimental results, we propose new substituted
hydroxynaphthalenecarboxamides, N-(hetero)aryl-piperazinylhydroxyalkylphenylcarbamates,
substituted piperazinylethanols and substituted piperazinylpropandiols as initial structures for
the development of new inhibitors for BRAF. Moreover, by performing QTAIM analysis, we
are able to describe in detail the molecular interactions that stabilize the different Ligand-
Receptor complexes. Such analysis indicates which portion of the different molecules must be
changed in order to obtain an increase in the binding affinity of these new ligands.
Keywords: BRAF inhibitors, virtual screening, synthesis, bioassays, molecular modeling.
1. Introduction
The BRAF/MEK½/ERK½ cascade is one of the major pathways deregulated in many types of
cancer. Oncogenic mutations in BRAF occur in around 50% of melanoma patients but also in
colon adenocarcinoma (5-12%), papillary thyroid carcinoma (39-69%), and others 1-2. Upon
activation of this pathway, BRAF phosphorylates and activates the dual kinase MEK½, which
in turn phosphorylates and activates its only target ERK ½ at Tyr 204/187 and Thr202/185
2

residues. RAF family includes 3 members: ARAF, BRAF and CRAF, although oncogenic
mutations occur mostly in BRAF gene. The majority of BRAF mutations consist in substitutions
of valine for glutamic acid (BRAFV600E), or less frequently for lysine (BRAFV600K), at position
600 leading to a constitutive active protein that does not require external stimuli to be activated
1-3. To date, the U.S. Food and Drugs Administration have approved the use of two BRAF
4
5
inhibitors (BRAFi), Vemurafenib and Dabrafenib , and two MEK inhibitors (MEKi),
trametinib and cobimetinib 6-7, for the treatment of metastatic melanoma. BRAFi and MEKi
constitute the first-line treatment for BRAFV600E mutated melanomas, however the development
of acquired resistance invariably occurs, and in some cases no clinical benefit at all is observed
8-9. Although many mechanisms of resistance have been reported, most of them involve the
reactivation of ERK½ signaling 10-12. Moreover, BRAFi therapy also induces paradoxical
13
activation of ERK signaling in BRAFwt tissue that may promote the formation of benign
14
tumors in the skin . Resistance development and secondary effects to BRAFi therapy reduce
the effectiveness and durability of the clinical response, and justify the search and design of new
potential BRAFi.
The high sequence conservation within the large family of protein kinases makes it difficult to
develop selective inhibitors for a particular member of the family. Furthermore, nearly all of the
approved protein kinase antagonists are steady-state competitive enzyme inhibitors of ATP and
they interact with the ATP-binding pocket which is well conserved among the family of protein
kinases 15. However, protein kinases have high plasticity and it is well known that kinase
conformational state has a great impact in inhibitor potency and selectivity 16. Therefore,
conformational plasticity of kinases can be exploited to design inhibitors that target a specific
kinase type in a particular conformational state.
BRAF has an active site (including the so called catalytic loop) to which both ATP and inhibitors
bind. The conserved DFG triad shifts to an ‘out’ conformation where the phenylalanine side
chain vacates a lipophilic “allosteric” pocket that is available to inhibitor binding (i.e. DFG-out
conformation). Compounds that bind to kinase in DFG-out state are classified as type II
17
inhibitors , in contrast to the active DFG-in conformation. In the latter case, PHE is flipped
toward the allosteric pocket and inhibitors that target this conformation (Type I inhibitors) bind
at the ATP site but do not penetrate the allosteric pocket and therefore do not depend on specific
kinase conformations for binding. There is a small set of kinases including BRAF that achieves
a peculiar variant of the DFG-in conformation in which a conserved salt bridge between a
glutamate in C-helix and a lysine in the 3-sheet is broken. The disruption of this salt bridge
3

causes an outward movement of the αC-helix (αC-helix-out) and renders the kinase in an inactive
state despite the DFG triad residing in the ‘in’ conformation. Inhibitors that target the DFG-in
αC-helix-out conformation are classified as type IIB while inhibitors that bind to the more typical
inactive kinase conformation DFG-out are re-classified as type IIA inhibitors. Since only few
kinases are able to perform this DFG-in, αC-helix-out conformation, inhibitors designed to target
this particular kinase state have exhibited excellent selectivity towards the RAF kinase family.
These inhibitors accommodate a propyl group in a small lipophilic pocket formed by the outward
18
movement of αC-helix-out . Vemurafenib is an example of a selective BRAF inhibitor that
targets this kinase conformation.
Considering the important role of BRAF in processes related to cancer 1 and vemurafenib clinical
limitations, the main objective of this work was to develop new inhibitors possessing different
structural scaffolds than those of the already known inhibitors. Our study has been carried out in
different stages; in the first step we performed a virtual screening that allowed us to identify
potential new inhibitors. In the second step, we synthesized and tested the inhibitory activity of
the novel compounds. Finally, we conducted a molecular modeling study that allowed us to
understand interactions at the molecular level that stabilize the formation of the different
molecular complexes. Figure 1 shows an scheme of the various steps that have been carried out
in this study.
4

Figure 1. Flow chart showing the various steps and techniques carried out in our study. The
numbers inside the boxes indicate the number of compounds evaluated.
2. Results and discussion
2.1 Structure-based approach to prioritize novel scaffolds as BRAF inhibitors
At the time we started this work, there were around 60 BRAF crystal structures available in the
Protein Data Bank (PDB). Taking advantage of this structural information we pursued a
structure-based drug design project searching for novel scaffolds that may function as BRAF
inhibitors.
Structure-Based Virtual Screening (SBVS) strategies rely on the three dimensional structure of
a target and on the ability of docking algorithms to predict the binding mode and the binding
affinities of different compounds obtained from libraries.
As was previously remarked, protein kinases have high plasticity and several conformations, and
it is well known that kinase conformational state has a great impact in inhibitor potency and
16
selectivity . BRAF crystal structures deposited in the PDB up to date cover the three more
relevant kinase conformational states.
5

Figure 2 shows superposition of BRAF in two of its conformational states, DFG-out (gray) and
DFG-in αC-helix-out (cyan). In Figure 2a, a type IIA inhibitor (gray sticks) accommodates a
bulky aromatic group into the lipophilic pocket of DFG-out conformation (depicted as a gray
surface) which is otherwise occluded in DFG-in like conformations. Thus, type IIA inhibitors
that were designed to target the lipophilic pocket in DFG-out conformation would not bind
properly at DFG-in conformations.
Figure 2. Superposition of two conformations of BRAF (DFG-out is shown in grey color and
DFG-in αC-helix-out in cyan)). Fig 2a: Transition from DFG-in to DFG-out conformation. Fig
2b: Type IIB inhibitor induce an outward shift of αC-helix
In contrast, in Figure 2b, a type IIB inhibitor (cyan sticks) accommodates a small aliphatic group
into a small lipophilic pocket created by the outward shift of αC-helix which disrupt the salt
bridge between K483 and E501; such place can be occupied by the inhibitor in the DFG-in αC-
helix-out conformation . This small but deep pocket is generally absent in DFG-in and DFG-out
conformations in which the salt bridge holds αC-helix and β3-sheet together. Therefore, in order
to allow more structural variety among selected compounds, virtual screening libraries should
be docked against representative structures from each of the BRAF conformational states
(ensemble docking). In other words, if only one of the BRAF conformations are considered in
docking calculations, we could fish inhibitors targeting that conformation but we would miss
inhibitors that stabilize other relevant BRAF conformations.
2.1.1. Virtual screening and Docking power assessment
The docking process is usually divided into two major steps: first, the correct placement of the
6

ligand at the protein binding-site and second, the estimation of the ligand affinity by a scoring
function 19. Scoring functions launched with docking software are usually calibrated with
structurally very diverse complexes because they are intended for general use in any bimolecular
system of interest. In part because of their general nature, standard docking scoring functions
often might fail either to recover ligand known binding affinities (scoring power) or poses
(docking power) or might fail to rank ligands by their affinities (ranking power) or also recover
known ligands from decoys libraries (screening power). Therefore, before undertaking the search
for novel BRAF inhibitors, we considered it was prudent to perform a retrospective study with
known BRAF inhibitors to evaluate performance of docking algorithm in our system of study.
Metrics employed to evaluate and tune docking performance depends on the intended use of
docking programs. Docking, scoring, ranking or screening power can be tuned depending on the
particular task and experimental information available. The scoring function optimal for the
different tasks may have some overlap, although improvements in affinity prediction are not
always related to improvements in the docking performance 20. To test screening power we
measured the ability of AutoDock4 21 standard scoring function to retrieve BRAF inhibitors from
a decoys library. The decoys library was compiled by seeding about 40 known BRAF inhibitors
into a larger set of presumably non-binding compounds (decoys) that were automatically
22
generated with the DUD-E server . Decoy molecules have similar physical properties than
actual ligands so that enrichment is not a mere separation by trivial physical characteristics, but
are chemically distinct from ligands, so that they are likely to be non-binders.
BRAF inhibitors seeded into the decoys library were extracted from crystal structures of BRAF
bound to inhibitors to later evaluate both docking power (i.e. by RMSD measurement) and
screening power. Inhibitors were chosen considering that the three types of inhibitors are fairly
represented within the inhibitors pool (i.e. types I, IIA and IIB are roughly in the same number).
The decoy library was docked against a representative structure from each of the three most
relevant BRAF conformational states, i.e. DFG-in, DFG-out and DFG-in αC-helix-out
conformations. After docking the decoys library with AutoDock4, ROC (Receiver Operating
Characteristic) curves were constructed and the Area Under the Curve (AUC) was computed.
Figure 3 depicts ROC curves and AUCs for virtual screening on each of the BRAF
conformations as well as for the ensemble of BRAF conformations.
Ensemble docking helps to mitigate the lack of flexibility of docking receptor models by
allowing docking a single ligand against multiple conformations of the protein and selecting the
highest scoring binding mode from the ensemble of receptors. Any model with an AUC >0.5
7

performs better than random in discriminating the most active compounds from less active ones.
Unsurprisingly, the virtual screening (VS) prediction on the ensemble of BRAF conformations
performs better than on single BRAF conformations (see AUCs in Figure 3). This is in line with
the fact that different types of BRAF inhibitors stabilize different conformations of the enzyme.
As observed in the figure, the ensemble docking model performs reasonably well in the
classification of the dataset (AUC ~ 0.75).
Figure 3. Screening power evaluation of docking algorithms for retrieving known BRAF
inhibitors from a decoys library.
Moreover, since we know the experimental binding modes of the BRAF inhibitors in the decoys
library we subsequently exploited that information to evaluate docking power, namely the ability
of docking algorithm to predict the correct conformation of the ligand and placement in the
receptor molecule (pose).
Docking power usually is evaluated by computing the root mean square deviation (RMSD)
between the heavy atoms of the docked ligand and native crystal pose. Usually, a threshold of
2-3 Å RMSD is used to calculate the fraction of correct predictions.
Heat maps on Figures 4a, 4b, and 4c depict the RMSD values between crystallographic and
docking poses of type IIA, type IIB and type I BRAF inhibitors, respectively. Each inhibitor was
docked against the three relevant conformations of the enzyme and the conformation that
achieves the best docking scoring is labeled with a green star.
8

Figure 4. Heat maps depicting RMSD differences between crystallographic and docking poses
for type IIA (a), IIB (b) and type I (c) BRAF inhibitors as docked in the three relevant
conformations of the enzyme. Best scoring conformation for each ligand is labeled with a green
star. Ligands are named by the PDB ID of the BRAF native complex they belong to. Underscored
PDB IDs represent different protonation states of the same ligand.
As depicted in Figure 4a, Type IIA inhibitors can reproduce their crystallographic poses (i.e.
considering a 2-3 Å RMSD tolerance) only when they are docked to the DFG-out conformation
of BRAF. This makes sense because type IIA inhibitors were designed to target the allosteric
pocket that is otherwise occluded in DFG-in conformations. There is also a good correspondence
9

between docking power and scoring, namely enzyme conformation achieving the best pose for
a particular ligand is also the one that show the best energy score, as indicated by the green stars
in Figure 4a. In practice, that means that prospective virtual screening campaigns will likely
prioritize candidates that reproduce known inhibitors binding modes and therefore that have
more chance of being true positives upon experimental corroboration.
Similarly, most of Type IIB inhibitors are docked correctly in their native-like BRAF
conformation, DFG-in αC-helix-out. These inhibitors have a hydrophobic tail attached to a
sulfonamide group and target a small hydrophobic pocket created between β3-sheet and αC-
helix. In DFG-in conformation a conserved salt bridge between GLU501 and LYS483 holds
together the αC-helix and the β3-sheet and prevent the formation of the small hydrophobic pocket
between both segments. This explains why Type IIB inhibitors cannot dock properly to BRAF
in the DFG-in conformation (see Figure 4b). Regarding DFG-out state, several type IIB
inhibitors bind to this conformation in a native-like pose because they can accommodate the
small hydrophobic tail in the adjacent allosteric pocket characteristic to that conformation
without a significant degradation in the RMSD value.
On the other hand, and unlike previous BRAF inhibitors types, several type I inhibitors bind
better to DFG-out or DFG-in αC-helix-out BRAF conformations than to its own DFG-in
conformation. Moreover, many of them cannot reproduce the crystallographic poses within a 2-
3 Å RMSD tolerance in any of the BRAF conformations. Even though type I inhibitors do not
perform as well as type IIA and IIB inhibitors, we consider that our proposed virtual screening
protocol involving the three most relevant BRAF conformations performs well in terms of
screening and docking power.
2.1.2. Pre-selection of candidate compounds and results of the virtual screening
Most kinase inhibitors discovered to date bind to kinase ATP binding cleft and roughly mimic
the hydrogen bonds normally formed by the adenine, ribose and/or triphosphate moieties of ATP.
Structure of ATP analog bound to kinase AKT1 (PDB code 1O6L) provides evidence about ATP
binding mode at kinases ATP binding cleft. ATP binds in the cleft with the adenine ring forming
hydrogen bonds with the kinase 'hinge', i.e. the segment that connects the amino- and carboxy-
terminal kinase domains. The ribose and triphosphate groups enter in a tight channel and
triphosphate moiety interacts with the activation segment that contains the common DFG motif,
via two magnesium ions (PDB code 1O6L).
It is clear that the compounds selected for further testing should bear some structural resemblance
10

to vemurafenib and other known ligands; i.e. part of the molecule mimicking adenine and other
portion that mimics the triphosphate moiety. In addition some conformational flexibility could
be convenient. Based on these considerations, we selected a group of 153 compounds including
chalcones, acetogenins, protoberberines, nitrosopyrimidines, carbamates and carboxamides
among others that already form part of our collection of compounds. We also include some new
compounds (not reported yet) structurally related. The fingerprints of these 153 compounds were
calculated using Open Babel 23, a chemical expert system mainly used for converting chemical
file formats, and the results were compared with fingerprints obtained for the 40 compounds
used previously to evaluate the docking “powers”. Compounds that have a Tanimoto index (Tc)
19 greater than 0.7 were discarded. Note that the fingerprints are a way to encode the structure of
a molecule in order to compare it with other compounds; whereas the Tc is a measure of the
similarity between two structures ranging from 0 to 1. Based on Tc index, we selected 108
structurally diverse compounds which possess structural differences with respect to already-
known BRAF inhibitors, justifying their further study. Each of the 108 compounds was docked
against the three relevant conformations of the enzyme (ensemble docking). Then, the
compounds were ranked according to the score achieved in the best enzyme conformation and
the 30 top ranked compounds were selected. From these compounds, only the ones that bind in
a pose that resemble the binding mode of known BRAF inhibitors like Vemurafenib (see below)
were selected. Thus, 23 compounds were selected as potential inhibitors, while the remaining 7
were discarded because they unite in a different pose to that of the crystallized structures. All
these 23 compounds have shown a better score when docked to the DFG-in αC-helix-out
conformation of the enzyme, hence they were regarded as potential BRAF type IIB inhibitors.
2.2. Chemistry
Our model predicted 23 compounds as potential inhibitors of BRAF (Table 1). Those
compounds were tested in an in vitro ERK1/2 phosphorylation assay as an indirect measure of
BRAF inhibition in melanoma cells (results shown in the next section).
Table 1. Structural features of compounds obtained from the virtual screening and evaluated as
BRAF inhibitors.
11

HO
HO
H
H
N
N
CF3
1
3
2
4
O
O
Br
H
H
N
N
Cl
Cl
Cl
OH
O
OH
OH
O
Cl
OH
H
N
Cl
Cl
H
N
Cl
5
6
O
O
Cl
OH
OH
H
H
N
Cl
N
Cl
Cl
7
8
O
O
Cl
Cl
N
S
N
O
O
H
N
NH
H
NH
N
S
COOH
COOH
N
N
Br
Br
9
10
12
H
H
O
O
COOH
Br
Br
SO2NH2
H
S
O
N
HN
S
N
Br
Br
N
Br
Br
N
HN
O
11
H
O
O
NH
N
H
O
O
H
N
HN
O
N
OH
NH
O
13
15
14
16
HN
F
N
COOH
CF3
F
OH
N
H
N
OH
N
O
N
H
O
N
N
O
O
N
OH
N
N
H
OH
N
O
N
N
H
N
17
18
O
N
O
O
12

N
N
OH
N
OH
N
N
H
N
19
20
22
O
O
N
N
O
O
O
O
N
H
O
N
O
N
OH
N
N
OH
NH
H
N
21
23
O
N
H
N
Cl
O
O
O
O
N
OH
NH
O
N
H
Cl
The compounds 1–8 were prepared by means of microwave-assisted synthesis in one step.
Briefly, the carboxyl group was activated with phosphorus trichloride, and then aminolysis of
acyl chloride by ring-substituted aniline in dry chlorobenzene yielded a final amide. All the
compounds were purified by recrystallization from ethanol 24-26
.
27-30
Synthesis and characterization of compounds 9-13 have been previously reported
(See
Suplemmentary Information). These compounds have been reported DNA gyrase inhibitors
interacting with the ATP-binding site of the enzyme.
Reaction of 2- or 3- or 4-aminoacetophenone and a suitable alkyl chloroformiate gave alkyl (2-
/3-/4-acetylphenyl)carbamates 14a–21a, which reacted with bromine or with copper (II) bromide
in chloroform/ethyl acetateto yield alkyl [2-/3-/4-(bromoacetyl)phenyl]carbamates 14b–21b;
these compounds by treatment with different N-monosubstituted aryl(heteroaryl)piperazines
provided
yl)acetyl]phenyl}carbamates 14c–21c. Subsequent reduction of keto group using NaBH4
provided target alkyl {2-/3-/4-[1-hydroxy-2-(4-aryl(heteroaryl)piperazin-1-
the
corresponding
alkyl
{2-/3-/4-[(4-aryl(heteroaryl)piperazin-1-
yl)ethyl]phenyl}carbamates 14–21, see Scheme 1.
13

O
O
O
O
O
R1O
R1O
Br
a
b
CH3
CH3
H2N
HN
HN
14a-21a
14b-21b
c
R2
R2
OH
N
O
N
N
O
O
R1O
R1O
N
d
HN
HN
14-21
14c-21c
Scheme 1: Synthesis of alkyl {2-/3-/4-[1-hydroxy-2-(4-arylpiperazin-1-yl)ethyl]
phenyl}carbamates 3–6. R1: 2-Et, 3-Pr, 3-Bu, 4-Bu; R2: 2-fluorophenyl, 4-fluorophenyl, 4-
(trifluoromethyl)phenyl, pyridine-2-yl, pyridine-4-yl, pyrimidine-2-yl. Reagents and conditions:
a) ClCOOR1, pyridine, acetone, reflux 3h; b) Br2, CHCl3, ambient temperature 3 hor CuBr2,
CHCl3/EtAc, reflux 8 h; c) N-(hetero)arylpiperazine, TEA, THF, ambient temperature 3 h; d)
NaBH4, MeOH, reflux 3 h.
Studied 1-(3-{4-[(alkoxycarbonyl)amino]benzoyloxy}-2-hydroxypropyl)-4-phenylpiperazin-1-
ium chlorides 22 and 23 were prepared by multiple-step reaction described in Scheme 2.
Epoxides 22a and 23a were prepared from 4-aminobenzoic acid through reaction with propyl
and methylchloroformiates giving appropriate 4-[(alkoxycarbonyl)amino]benzoic acids.
Chlorides of these acids formed by thionyl chloride treatment gave desired epoxides 22a and 23a
31
after reaction with 2,3-epoxypropan-1-ol . In the last step, final compounds 22 and 23 were
prepared by a reaction of the epoxides with 1-(4-phenyl)piperazine and then converted to the
hydrochloride salts using ethereal HCl to enhance their solubility in water 32.
COOH
COCl
COOH
O
N
O
a
b
R
R
O
N
O
N
H2N
H
H
c
O
O
Cl
NH
O
O
O
O
O
d,e
R
O
OH
R
N
O
N
H
H
22, 23
22a 23a
,
Scheme 2: Synthesis of studied 1-(3-{4-[(alkoxycarbonyl)amino]benzoyloxy}-2-
hydroxypropyl)-4-phenylpiperazin-1-ium chlorides 22 and 23. R: Bu and Pr. Reagents and
conditions: a) ClCOOR, pyridine, acetone, reflux 3 h; b) SOCl2, toluene, reflux; c) 2,3-
epoxypropan-1-ol, THF, TEA, 0°C then ambient temperature; d) 1-(4-phenyl)piperazine, i-
PrOH, reflux; e) HCl, Et2O, ambient temperature.
14

2.3.
Biological assays
Next, it was important to study whether or not these 23 compounds inhibit BRAF. Since the
BRAF/MEK/ERK pathway is lineal, it is possible to relate the BRAF activity to the levels of
phosphorylated ERK (pERK) in a western blot assay. Indeed, ERK½ kinase is the only substrate
of MEK ½ that has been described to date 33. To this end, Lu1205 melanoma cell line that express
the mutant kinase BRAFV600E, was employed. To establish the reliability of our model, we first
tested the response of Lu1205 cells to the known BRAFV600E inhibitor Vemurafenib. In line with
previous reports, we show here that Vemurafenib strongly reduced ERK phosphorylation at both
1 and 10 μM 34. The effect of Vemurafenib is less pronounced after 48-72 h, probably indicating
that ERK reactivation occurs, as was shown by other authors (Figure 5) 35-36
.
Figure 5: Lu1205 melanoma cells were treated with 1 and 10 μM of Vemurafenib or Control
(untreated cells incubated with the carrier DMSO) at the indicated times. Proteins from cell
lysates were separated by SDS-PAGE and transferred onto a nitrocellulose membrane.
Membranes were incubated with the primary antibodies as indicated. The image is representative
from at least 5 independent experiments.
After determining that Lu1205 cells are responsive to Vemurafenib treatment as we expected
and it is reported in literature, we considered that this cell model is appropriate for the following
assays. Next, we evaluated the biological activity of the compounds listed in the table 1. Since
the goal of this study was to find new inhibitors of BRAF, we consider that 2 hours treatment
should be enough to show whether or not the new compounds have any noticeable activity. Thus,
to perform a quick initial screening of BRAF kinase inhibition, we determined the
phosphorylation of ERK in Lu1205 cells treated with the compounds at 10 μM for 2 hours by
In-Cell Western (ICW) or western blot (Figure 6).
15

Figure 6: Results obtained from the initial screening for compounds of table 1. Substituted
hydroxynaphthalenecarboxamides (dark green), N-(hetero)aryl-
piperazinylhydroxyalkylphenylcarbamates (blue), substituted piperazinylethanols (light green),
and substituted piperazinylpropandiols (orange). The red line indicates the cut between
compounds with significant activity of the rest. The graph corresponds to the quantification of
pERK of 2 experiments.
As shown in Figure 6, compounds 1, 2, 9, 10, 19-22 reduced ERK phosphorylation by more
than 30% and were selected to perform a more detailed study. Compounds 17, 18 and 23, which
did not show, or showed little inhibitory activity in the previous screening, were also included
to verify the specificity of the compounds that did show activity. Next, Lu1205 cells were treated
with the 11 selected compounds for 2 h at 10 and 1 μM. Remarkably, all the compounds selected
reduce ERK phosphorylation at different magnitude indicating that they have BRAF inhibitory
activity, although only compounds 1, 2, 9, 19, 20-22 and 23 reduced significantly pERK
suggesting that they have a stronger inhibitory effect (Figure 7). It is important to notice that,
although some compounds (ie. compound 17) decreased ERK phosphorylation, that reduction is
not statistically significant.
16

Figure 7: Lu1205 melanoma cells were treated with the indicated compounds for 2 hours at 1
and 10 μM. Proteins from cell lysates were separated by SDS-PAGE and transferred onto a
nitrocellulose membrane. Membranes were incubated with the primary antibodies as indicated.
The software Image Studio 5.2 was employed for the quantifications. Optical density of pERK
band for each treatment was normalized to the respective ERK total band. pERK/ERK total
relation for each treatment was relativized to pERK/ERK total relation of the control
corresponding to untreated cells incubated with the carrier DMSO. Quantifications correspond
to 3 independent experiments and the blots are representative from them. Statistic: One-way
ANOVA and Tukey’s post-test; (* p = ˂ 0,05; ** p = ˂0,01; *** p = ˂ 0,001).
Considering that BRAFV600E is important in cell growth, we decide to determine whether or not
the 11 selected compounds reduce cell viability as result of BRAF inhibition by using MTT
assay. In order to determine the adequate experimental conditions, first we tested cell viability
of Lu1205 cells treated with Vemurafenib 1 and 10 μM at 48 and 72 h. As shown in Figure 8A,
Lu1205 cells treated with Vemurafenib proliferate less in comparison to the control both at 48
17

and 72 h. Furthermore, ICW was performed to corroborate that BRAF remains inhibited at 48-
72 h of Vemurafenib treatment (Figure 8B). Since the results were similar at both 48 and 72 h,
we decided to use only 72 h for the following experiments.
Figure 8: Lu1205 cells cultured in 96-well plate were treated with Vemurafenib 1 and 10 μM or
control. MTT reagent was added at the indicated times and incubated for 3 additional h. Then
formazan product was dissolved and absorbance read at 540 nm (A). ICW was performed in
parallel to MTT assay (B).
Subsequently, Lu1205 cells were treated with the 11 selected compounds at 1 and 10 μM for 72
h. Our results revealed that only compounds 1 and 22 reduced significantly Lu1205 cell viability
(Figure 9).
Figure 9: Lu1205 cells cultured in 96-well plate were treated with the 11 compounds selected
previously, Vemurafenib or control (DMSO), at 10 and 1 μM. After 72 h of treatment, MTT
reagent was added and incubated for 3 additional h. Then formazan product was dissolved and
absorbance read at 540 nm. Each condition was assayed by 6 replicates and the results
correspond to 3 independent experiments. One-way ANOVA and Tukey’s post-test were
performed.
18

In summary, we evaluated 23 compounds with different structural scaffolds. From the
preliminary study we selected 11 compounds, of which 7 showed significant inhibitory activities.
The results of the most exhaustive study show that these 7 compounds reduce ERK
phosphorylation. These results indicate the good performance of our screening model.
Compounds 1, 9, 20 and 22 have different scaffold and showed the most interesting activities.
Among active compounds it is possible to clearly visualize four different types of structural
scaffolds: a) substituted hydroxynaphthalenecarboxamides (compounds 1 and 2), N-
(hetero)aryl-piperazinylhydroxyalkylphenylcarbamates (9), substituted piperazinylethanols (19
and 20) and substituted piperazinylpropandiols (22 and 23). Considering that these compounds
have been obtained through a primary screening, their inhibitory activities can be considered
quite strong and promising. In this sense, the inhibitory effect obtained for compounds 1 and 22
are very interesting since they showed a strong inhibitory activity even at relatively low
concentrations (1 M). Furthermore, both compounds also reduced significantly cell viability,
indicating that the inhibition of ERK signaling is accompanied by an anti-proliferative effect.
Although these compounds have a strong inhibitory activity, their effects are weaker than that
obtained for Vemurafenib. In the next section we report a molecular modeling study in which
we evaluate the molecular interactions of the different L-R (Ligand-Receptor) complexes, which
allows us to explain the different activities observed.
2.4. Molecular modeling
In order to understand at molecular level the behavior of these compounds, we conducted a
molecular modeling study. The main objective of this study was to evaluate the molecular
interactions that stabilize the different molecular complexes. This information could be very
useful for the design and modification of these starting structures to obtain compounds with
higher affinity for this protein. To this end, we selected a representative compound for each
structural scaffold to make a comparative study for compounds 1, 9, 20 and 22. We also included
in our study Vemurafenib, which allowed us to perform a comparative analysis of the different
activities displayed by these molecules in relation to their structural differences.
The molecular modeling study was conducted in three different stages. First, we performed a
21
docking analysis using the Autodock program . In the second stage of this study, we carried
37
out molecular dynamics (MD) simulations using the AMBER software package . From the
trajectories obtained with MD simulations, we performed an analysis per residue for the different
compounds. Finally, to better understand the molecular interactions involved in the different L-
19

R complexes, a Quantum Theory of Atoms In Molecules (QTAIM) study was carried out for the
most representative structures. We have previously demonstrated the importance of these
QTAIM studies for the understanding of the details of the different molecular interactions that
stabilize or destabilize the various complexes 38-45
.
Analysis per residue obtained from molecular dynamics simulations allowed us to define the
main interactions that stabilize the different complexes (Figure 10). Our simulations indicate the
importance of the following residues: ILE463, LYS483, LEU505, PHE516, ILE527, TRP531,
GLN530, CYS532, PHE583, PHE595 and mainly ASP594 for binding of these ligands to BRAF.
These results are in agreement with previously reported experimental data 46.
Figure 10. Over imposed histograms showing the interaction energies of compounds 1 (dark
green, a), 9 (light-blue, b), 20 (light-green, c) and 22 (orange, d) with the main amino acids
involved in the complex formation. The histogram obtained for Vemurafenib is shown in gray
in all the histograms for comparison.
Superpositions of interactions obtained for compounds 1, 9, 20 and 22 with those displayed by
Vemurafenib are shown in Figure 10a-d. According to this data, it is evident that these
compounds bind in a similar manner to Vemurafenib because they interact with essentially the
same amino acids. However, these interactions are weaker than those shown for Vemurafenib
along the simulations. These results are in agreement with the experimental data and might
explain, at least in part, the lower inhibitory effects of the compounds reported here than
Vemurafenib. Active compounds displayed their pharmacophoric portions in a closely related
spatial form to that displayed by Vemurafenib 46. This might be well appreciated in Figure 11.
20

Figure 11. View of the spatial ordering adopted by the different ligands in the active site of
BRAF. Compounds 1 (dark green), 9 (light-blue), 20 (light-green) and 22 (orange) are over
imposed with Vemurafenib in gray color. All these compounds are docked to a similar
conformation of BRAF.
The main objective of our molecular modeling study is to evaluate the molecular interactions
that stabilize the different ligand-enzyme complexes. It is well known that simulations of
molecular dynamics are too crude, and that they might miss important information and many
details involved in the stability of molecular complexes. Thus, we decided to conduct a QTAIM
study that can better quantify the molecular interactions obtained for the different complexes. In
fact, our research’s group has been the pioneer in this type of studies and in many of them we
have demonstrated the ability of QTAIM calculations to evaluate the molecular interactions for
this type of complexes 38-45
.
21

Figure 12. Charge density sum at the intermolecular BCPs for complexes of BRAF with selected
inhibitors. The interactions corresponding to the adenine-like moiety of inhibitors has been
colored in orange, whereas the interactions of the inhibitors that resemble the ATP triphosphate
interactions are depicted in magenta.
Figure 12 shows the charge density sum at intermolecular bond critical points (BCPs). Inhibitors
contributions to anchoring at the DFG-loop and hinge regions of ATP binding site are stacked
on top of each other. As depicted in this figure, Vemurafenib is more strongly anchored to the
enzyme ATP binding site than inhibitors reported here, as evidenced by the total height of
stacked bars. Moreover, separate contributions to anchoring at the Hinge and DFG-loop regions
of ATP binding cleft tell us which interactions need to be optimized in order to increase binding
affinity of the new inhibitors.
On the other hand, when comparing Σρ values and inhibitory activities of moderate inhibitors as
the ones reported here, no evident relationship is found. It should note that the four compounds
reported here displayed similar inhibitory effect; however compounds 1 and 20 are more weakly
anchored to the enzyme binding cleft according to the Σρ values.
Separate contributions to anchoring at the Hinge and DFG-loop regions of ATP binding cleft
(see Figure 12) show that Vemurafenib as well as the two most active inhibitors reported here
(1 and 22) show a stronger anchoring at the hinge region as compared to compounds 9 and 20.
Moreover, these inhibitors also have in common that anchoring at both DFG-loop and hinge has
similar strength, namely they show a balanced anchoring at both ATP binding cleft regions.
Comparing ATP binding mode with those found for Vemurafenib and the new inhibitors from
Molecular Dynamic simulations (Figure 11) it is easy to identify which part of the inhibitors
mimic the interactions of adenine and which other parts mimic the triphosphate (and less clearly
22

ribose) interactions. Adenine moiety from ATP binds to the hinge region of the binding cleft
while triphosphate group binds to the DFG-loop. Therefore, the hinge and DFG-loop interacting
parts of inhibitors are depicted in red and blue respectively, in the subsequent charge density
molecular graphs.
The charge density molecular graphs in Figure 13 displayed the main interactions of
Vemurafenib at ATP binding site of BRAF kinase. For easy viewing, interactions of
arylsulfonamide moiety that substitutes position 3 of the azaindole ring are shown in Figure 13A
whereas interactions of azaindole as well as its substituent at position 5 are depicted in Figure
13B.
23

Figure 13. Charge density molecular graphs depicting the molecular interactions of
Vemurafenib at the ATP binding site. Yellow lines connecting the nuclei are the Bond Paths
(BPs) and small red circles over them are the Bond Critical Points (BCPs). For easy viewing,
interactions with DFG-loop and Hinge region are depicted separately in panels A and B,
respectively. The interacting portions of DFG-loop and Hinge of Vemurafenib are colored in
purple and orange, respectively.
24

The sulfonamide moiety of Vemurafenib resembles ATP phosphate groups. In that way, it forms
strong interactions with the DFG motif of the activation segment as long as it is in the “in”
conformation. As shown in Figure 13A, the sulfonamide oxygen atoms are connected through
bond paths to the backbone of ASP594 while the amide proton is interacting with ASP594 side
chain oxygen. The propyl tail of Vemurafenib is anchored in a hydrophobic pocket where it
interacts with side chains of residues PHE595, ILE527, PHE516 and LEU505. The formation of
this hydrophobic pocket is a consequence of the disruption of the conserved salt bridge between
GLU501 and LYS483 that holds together the αC-helix and the β3-sheet. The propyl tail of
Vemurafenib pushes out the αC-helix which in turn causes the salt bridge breaking and the
formation of the hydrophobic pocket between αC-helix and β3-sheet. The hydrophobic pocket
is shown in surface representation in Figure 13A. Moreover, LYS483 is stacked over the phenyl
ring of the arylsulfonamide moiety and there are several bond paths connecting both groups.
Moreover, PHE498 is also helps hold in place the phenyl ring as well as the sulfonamide moiety
of Vemurafenib. This residue hangs over the binding cleft and usually do not interact directly
with inhibitor because the conserved salt bridge blocks its access to the binding cleft. Thus,
interactions of PHE498 are a feature specific to this type of DFG-in αC-helix-out BRAF
conformation where the salt bridge is broken.
In that regard, the heterocyclic or azaindole ring of Vemurafenib mimics the interactions of ATP
adenine ring and, as it was described in Figure 13B, heterocyclic ring lies in the plane of the
hinge loop and forms several interactions with backbone atoms of this loop. The pyridine
nitrogen atom acts as hydrogen bond acceptor against the backbone amide hydrogen of CYS532
whereas the proton attached to pyrrole nitrogen acts as hydrogen bond donor against the carbonyl
oxygen of GLN530. Other residues above and below the plane of the hinge loop also help to
keep the heterocyclic ring in place by forming interaction with the π electrons of this ring. Above
the plane of the hinge loop, side chain of residues ILE463 and ALA481 act as hydrogen bond
donor against the π system of the heterocyclic ring while side chain of TRP531 forms face to
face stacking interactions with that ring. Below the plane of the hinge loop, PHE583 forms
stacking interaction with azaindole ring and sulfhydryl proton from CYS532 acts as hydrogen
bond donor against the pyridine nitrogen. The chlorophenyl ring attached to the azaindole also
forms stacking interactions with PHE582 and TRP531.
Figures 14, 15, 16 and 17 show the charge density molecular graphs of inhibitors reported in
this work as they bind in the BRAF binding cleft.
25

Figure 14. Compound 9 interactions at the BRAF binding cleft. Topological elements of the
charge density for the intermolecular interactions are depicted: bond paths in yellow lines and
BCPs in red circles. The atoms of the inhibitor interacting with DFG-loop and hinge regions are
colored in purple and orange, respectively.
As can be seen in Figure 14, carboxylic acid oxygen atom of compound 9 forms strong
interactions with backbone of DFG-loop that resemble interactions of sulfonamide oxygen atom
of Vemurafenib with the same loop. In fact, DFG-loop interacting part of compound 9 is
anchored with almost the same strength to the ATP binding cleft as Vemurafenib (Figure 12).
On the other hand, hinge interacting part of compound 9 fails in adopting the ATP adenine-like
binding mode in which heterocycle lies in the same plane as the hinge backbone loop. Instead,
the compound 9 brominated pyrrole ring is anchored above that plane with an interaction pattern
similar to Vemurafenib but much weaker. It forms stacking interactions with side chains of
residues TRP531 and ILE463 above and PHE583 below the hinge plane, resembling interactions
of Vemurafenib with the same residues. However, some of the strong interactions formed by
Vemurafenib with the backbone of the hinge are either missed or very weak. It is evident that
anchoring of compound 9 is driven by the DFG-loop interacting part of the inhibitor which might
explain the deficient binding of the hinge interacting portion.
26

Figure 15. Interactions of compound 20 at the BRAF binding cleft. Topological elements of the
charge density for the intermolecular interactions are depicted: bond paths in yellow lines and
BCPs in red circles. Inhibitor atoms interacting with DFG-loop and hinge regions colored in
purple and orange, respectively.
In a similar fashion, compound 20 (Figure 15) also shows a deficient binding to enzyme hinge
loop that can be explained in the same way as in compound 9 complex, namely it is the DFG-
loop interacting part which governs the overall binding mode of the inhibitor (Figure 12). In
addition to interactions with backbone of DFG-loop, compound 20 is anchored in the
hydrophobic pocket between αC-helix and β3-sheet through its terminal n-butyl chain that
resemble Vemurafenib n-propyl chain. Hydrophobic interactions formed by n-butyl chain are
represented in the molecular graph of Figure 15 by bond paths connecting hydrogen atoms of
that chain with the same atoms of residues VAL501, LEU505, PHE516, ILE527 and PHE595.
These hydrophobic interactions together make a non-negligible contribution to anchoring.
Unlike to compounds 9 and 20, the remaining inhibitors reported in this work, compounds 1 and
22 are either more strongly anchored to the hinge loop as in the case of 22 or show a balanced
anchoring at both ATP binding cleft regions as in the case of 1.
27

Figure 16. Interactions of compound 1 at the BRAF binding cleft. Topological elements of the
charge density for the intermolecular interactions are depicted: bond paths in yellow lines and
BCPs in red circles. Inhibitor atoms interacting with DFG-loop and hinge regions colored in
purple and orange, respectively.
As evidenced in Figure 16, hinge interacting ring of compound 1 lies close to the plane of the
hinge region forming interactions with residues ILE463 and PHE583 above and below the plane,
respectively. Additionally, hydroxyl group atacched to β-naphtyl moiety of compound 1 is
connected through several bond paths to the side chain of residues SER535 and SER536 from
the hinge loop and also to HIS539. However, compound 1 lacks interactions with backbone of
the hinge characteristic of ATP adenine ring which explains its modest binding to that loop as
compared to Vemurafenib. On the other side, bromophenyl moiety from DFG interacting part of
compound 1 is anchored to the ATP binding cleft almost exclusively through hydrophobic
interactions involving the bromine atom. This rather weak anchoring of DFG interacting part
would allow a better positioning (and slighly stronger binding) of the hinge interacting part of
compound 1 as compared to 9 and 20. The modest improvement in hinge interacting part
anchoring seems to be related to increased inhibition of the first compound with respect to the
last ones.
28

Figure 17. Charge density molecular graphs depicting the molecular interactions of compound
22 at the ATP binding site. For easy viewing, interactions with DFG-loop and Hinge region are
depicted separately in panels A and B, respectively.
Similarly than compound 1, DFG-loop interacting part of compound 22 (Figure 17) binds to the
ATP binding cleft mostly through weak non-polar interactions involving its n-butyl terminal
chain. As shown in Figure 12, anchoring of that portion of compound 22 is rather weak as
29