Balancing reactivity and antitumor activity: heteroarylthioacetamide derivatives as potent and time-dependent inhibitors of EGFR

Article abstract of DOI:10.1016/j.ejmech.2018.11.029

Second- and third-generation inhibitors of EGFR possess an acrylamide group which alkylates Cys797, allowing to overcome resistance due to insurgence of T790M mutation. Less reactive warheads, yet capable to bind the target cysteine, may be useful to design newer and safer inhibitors. In the present work, we synthesized a 2-chloro-N-(4-(phenylamino)quinazolin-6-yl)acetamide (8) derivative as a prototype of EGFR inhibitor potentially able to react with Cys797 by nucleophilic substitution. We then tuned the reactivity of the acetamide fragment by replacing the chlorine leaving group with (hetero)-aromatic thiols or carboxylate esters. Among the synthesized derivatives, the 2-((1H-imidazol-2-yl)thio)acetamide 16, while showing negligible reactivity with cysteine in solution, caused long-lasting inhibition of wild-type EGFR autophosphorylation in A549 cells, resulted able to bind recombinant EGFR L858R/T790M in a time-dependent manner, and inhibited both EGFR autophosphorylation and proliferation in gefitinib-resistant H1975 lung cancer cells (expressing EGFR L858R/T790M mutant) at low micromolar concentration.

Full text of DOI:10.1016/j.ejmech.2018.11.029

Accepted Manuscript
Balancing reactivity and antitumor activity: Heteroarylthio acetamide derivatives as
potent and time-dependent inhibitors of EGFR
Riccardo Castelli, Nicole Bozza, Andrea Cavazzoni, Mara Bonelli, Federica Vacondio,
Francesca Ferlenghi, Donatella Callegari, Claudia Silva, Silvia Rivara, Alessio Lodola,
Graziana Digiacomo, Claudia Fumarola, Roberta Alfieri, Pier Giorgio Petronini, Marco
Mor
PII:
S0223-5234(18)30988-7
DOI:
https://doi.org/10.1016/j.ejmech.2018.11.029
EJMECH 10886
Reference:
To appear in: European Journal of Medicinal Chemistry
Received Date: 9 July 2018
Revised Date: 8 November 2018
Accepted Date: 9 November 2018
Please cite this article as: R. Castelli, N. Bozza, A. Cavazzoni, M. Bonelli, F. Vacondio, F. Ferlenghi,
D. Callegari, C. Silva, S. Rivara, A. Lodola, G. Digiacomo, C. Fumarola, R. Alfieri, P.G. Petronini, M.
Mor, Balancing reactivity and antitumor activity: Heteroarylthio acetamide derivatives as potent and
time-dependent inhibitors of EGFR, European Journal of Medicinal Chemistry (2018), doi: https://
doi.org/10.1016/j.ejmech.2018.11.029.
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.

ACCEPTED MANUSCRIPT
Balancing reactivity and antitumor activity:
heteroarylthio acetamide derivatives as potent and
time-dependent inhibitors of EGFR
‡1
‡1
2
2
1
Riccardo Castelli, Nicole Bozza, Andrea Cavazzoni, Mara Bonelli, Federica Vacondio,
1
1
1
1
1
Francesca Ferlenghi, Donatella Callegari, Claudia Silva, Silvia Rivara, Alessio Lodola,*
2
2
2
2
Graziana Digiacomo, Claudia Fumarola, Roberta Alfieri, Pier Giorgio Petronini, and Marco
1
Mor*
1
Department of Food and Drug, University of Parma, Parma, Italy
2
Department of Medicine and Surgery, University of Parma, Parma, Italy
‡
These authors contributed equally to this work
*
Corresponding Authors
Prof. Alessio Lodola,
Department of Food and Drug, University of Parma, Parma, Italy
E-mail: alessio.lodola@unipr.it
&
Prof. Marco Mor,
Department of Food and Drug, University of Parma, Parma, Italy
E-mail: marco.mor@unipr.it
1

ACCEPTED MANUSCRIPT
ABSTRACT
Second- and third-generation inhibitors of EGFR possess an acrylamide group which alkylates
Cys797, allowing to overcome resistance due to insurgence of T790M mutation. Less reactive
warheads, yet capable to bind the target cysteine, may be useful to design newer and safer
inhibitors. In the present work, we synthesized a 2-chloro-N-(4-(phenylamino)quinazolin-6-
yl)acetamide (8) derivative as a prototype of EGFR inhibitor potentially able to react with
Cys797 by nucleophilic substitution. We then tuned the reactivity of the acetamide fragment by
replacing the chlorine leaving group with (hetero)-aromatic thiols or carboxylate esters. Among
the synthesized derivatives, the 2-((1H-imidazol-2-yl)thio)acetamide 16, while showing
negligible reactivity with cysteine in solution, caused long-lasting inhibition of wild-type EGFR
autophosphorylation in A549 cells, resulted able to bind recombinant EGFR L858R/T790M in a
time-dependent manner, and inhibited both EGFR autophosphorylation and proliferation in
gefitinib-resistant H1975 (EGFR L858R/T790M) lung cancer cells at low micromolar
concentration.
KEYWORDS
EGFR; 4-anilinoquinazoline; warhead; cysteine; time-dependent inhibition; QM/MM
simulations.
Abbreviations
DMF, dimethylformamide; PDDG, Pairwise Distance Directed Gaussian; QM/MM, quantum
mechanics/molecular mechanics; SMD, steered molecular dynamics; TBTU, O-(benzotriazol-1-
yl)-N,N,N’,N’-tetramethyluronium tetrafluoroborate; THF, tetrahydrofuran; TS, transition state.
2

ACCEPTED MANUSCRIPT
1
. INTRODUCTION
Non-small-cell lung cancer (NSCLC) accounts for the 80% of all lung cancers, which
represent the leading cause of cancer-related deaths worldwide. NSCLC is featured by a 5-year
survival rate lower than 15% [1]. Gefitinib (1) and erlotinib (2) are the first generation of EGFR
inhibitors approved for the treatment of NSCLC [2]. Tumor responsiveness to these drugs has
been related to the insurgence of “activating mutations” such as short in-frame deletions in exon
1
9 (del19) or point mutations in exon 21, the latter resulting in arginine replacing leucine at
codon 858 (L858R) [3,4]. Gefitinib and erlotinib display a remarkable therapeutic outcome in
patients expressing activated forms of EGFR. However, their efficacy in clinical use is limited by
the occurrence of resistance. Nearly 60% of patients with EGFR alterations present a single
aminoacid mutation at the gatekeeper position of the ATP-binding site, from threonine to
methionine (T790M) [5]. The presence of this bulkier methionine reduced the potency of first
generation inhibitors for EGFR, favoring their competitive displacement by ATP, which is
present at millimolar concentrations inside the cells [6]. This resistance mechanism has been
overcome by equipping the scaffold of 4-anilinoquinazoline inhibitors with an acrylamide
warhead: by virtue of a hetero-Michael addition, the acrylamide moiety is capable of alkylating a
cysteine residue (i.e. Cys797) proximal to the hinge region of the EGFR active site [7,8]. The
formation of a covalent bond with Cys797 allowed second-generation inhibitors to exert
protracted occupancy of EGFR active site, overcoming competition with ATP [9]. Among the
second-generation inhibitors that entered into clinical studies, afatinib (compound 3) was
approved for the treatment of patients with metastatic NSCLC with EGFR mutations [10]. In
patients with acquired resistance to gefitinib (1) through the development of T790M mutation,
3

ACCEPTED MANUSCRIPT
afatinib (3) showed marginal therapeutic benefit. Its clinical use has been restricted by adverse
effects occurring in skin and gastrointestinal tract, resulting from the concomitant inhibition of
the wild type (WT) form of EGFR [11]. The disclosure of pyrimidine-based inhibitors [12],
selective for the T790M mutant over the WT, prompted the search for new candidates to be
advanced into the clinic [13]. Osimertinib (4) [14,15], the leader of this third generation of EGFR
inhibitors [16], is a 2-aminopyrimidine derivative featuring an acrylamide warhead portion. It
has been approved for patients affected by metastatic EGFR T790M mutation-positive NSCLC
who have progressed on after the therapy with first- and second-generation inhibitors [17], and
very recently for the first-line treatment of patients with metastatic NSCLC carrying EGFR-
activating mutation [18]. However, mutations conferring resistance to osimertinib have been
reported [19]. Among the identified mutations, C797S, in which the critical cysteine 797 is
replaced by a less nucleophilic serine, directly hampers EGFR alkylation, while loss of T790M
or appearance of L718Q mutation may affect the accommodation of the inhibitor in the active
site [20], preventing the formation of a protein-ligand reactive conformation that leads to target
alkylation [21]. These findings indicate that both the acrylamide group and the driving portion
need to be optimized to ensure selective inhibition of activated forms of EGFR (i.e. L858R, or
del19) in the presence of novel mutations. Moreover, the presence of a highly electrophilic
warhead (i.e. as for acrylamide) remains a toxicity burden [22]. A recent work on osimertinib has
shown that it inhibits several lysosomal cathepsins when administered to mice [23]. Considering
that cathepsins targeted by osimertinib are also expressed by human cells, some concerns have
been raised on the protracted use of osimertinib.
4

ACCEPTED MANUSCRIPT
Fig. 1. Chemical structures of selected EGFR inhibitors.
In this scenario, the availability of warheads able to form a covalent bond with Cys797 [24],
but less reactive than acrylamide present in EGFR inhibitors, would facilitate the design of novel
and selective inhibitors able to overcome resistance in NSCLC [25, 26]. Our previous work,
performed installing new warheads on a 4-anilinoquinazoline nucleus, has led to the
identification of long-lasting inhibitors of EGFR having electrophilic groups other than
acrylamide [27]. However, epoxide-based inhibitor 5 (Fig. 1) raised safety concerns similar to
those of acrylamides, while the 2-(pentafluorophenoxy)-acetamide 6 resulted incapable to
overcome T790M resistance. Herein, using a 4-(3-chloro-4-fluoroaniline)-7-ethoxyquinazoline
scaffold (compound 7, Scheme 1) featured by a substitution pattern similar to that of the
clinically approved drug afatinib (3, Fig. 1), we prepared a series of compounds possessing at the
position 6 an “activated” acetamide group, which may potentially alkylate Cys797 by a
N
nucleophilic substitution (S ). Our investigation started with the synthesis of a 4-(3-chloro-4-
fluoroaniline)-7-ethoxyquinazoline bearing a chloroacetamide at position 6 (compound 8,
5

ACCEPTED MANUSCRIPT
Scheme 1). The chloroacetamide was selected as the reference warhead for our work as this
group has been reported to alkylate cysteine both in solution [28] and within the ATP binding
site of kinases [29] and it has been already installed on a different 4-anilinoquinazoline to
achieve irreversible inhibition of EGFR [30]. As consequent step of our research, we modulated
the reactivity of the acetamide portion by replacing the chlorine atom in 8 with (hetero)-aromatic
thiols or carboxylate esters that may act as leaving groups in the framework of a S reaction. Our
N
choice of using the 4-anilinoquinazoline nucleus for the installation of the new warheads arises
from the necessity of having compounds with good inhibitory potency in the
autophosphorylation assay performed on A549 cells, which expressed WT EGFR. This test was
used as a screening method to identify potential irreversible inhibitors of EGFR. Regardless of
the chemical nature of the driver portion employed to target the ATP binding site of EGFR, our
approach can lead to the identification of soft warheads able to bind the reactive cysteine
(Cys797) of EGFR, while sparing all other thiols present in the cellular environment.
2
. CHEMISTRY
The newly synthesized inhibitors reported in this work (Scheme 1) feature the 6-amino-4-(3-
chloro-4-fluoroaniline)-7-ethoxyquinazoline core (7) equipped with specific amide-group based
warheads obtained by means of acylation of the 6-amino group with a panel of “electrophilic
traps”. The synthesis of the anilinoquinazoline scaffold 7 was performed as previously reported
with minor adaptations (see Supplementary Data) [31]. The general strategy features the
chloroacetamide-functionalized quinazoline core (compound 8) as starting material in the
synthesis of most of the compounds of the class. Compound 8 was prepared by reacting the
amino group of 7 with the mixed-anhydride obtained from the reaction of the trialkyl ammonium
6

ACCEPTED MANUSCRIPT
salt of chloroacetic acid with pivaloyl chloride. This method proved superior to the direct use of
chloroacetyl chloride both in terms of purity of the final compound and overall yield, and
allowed the use of tetrahydrofuran as the solvent, being 7 sparingly soluble in either chlorinated
solvents and acetonitrile. The obtained chloroacetyl-functionalized anilinoquinazoline 8 was
further elaborated by reacting with the appropriate nucleophile to furnish compounds 10, 12, 14-
2
2, as depicted in Scheme 1. The general procedure for this subset of compounds features the
alkylation of the heterocyclic thiol or carboxylate salt by 8 in the presence of a base.
Scheme 1. Reagents and conditions: a) i. chloroacetic acid, diisopropylethylamine, THF; ii.
pivaloyl chloride, 30 min; iii. 7, room temperature (yield 83%); b) See Experimental Section for
7

ACCEPTED MANUSCRIPT
specific conditions (yield range: 19-89%); c) 3-pyridylmercaptoacetic acid, TBTU, DMF,
triethylamine, room temperature, (for 11, yield: 25%); 3-pyridin-3-yl-propionyl chloride, THF,
pyridine, room temperature, (for 13, yield: 39%); d) NaOCH , THF, CH OH, room temperature
3
3
(23, yield: 62%).
The exact conditions for each compound varied extensively depending on the reactivity of
each thiol or carboxylate salt, details are given in the Experimental Section. The synthesis of
compounds 11 and 13 had to be devised differently, and was performed condensing 7 and 3-
pyridylmercaptoacetic acid in the presence of O-(benzotriazol-1-yl)-N,N,N’,N’-
tetramethyluronium tetrafluoroborate (TBTU) for 11 and 3-pyridin-3-yl-propionyl chloride for
1
3 respectively. Compound 23 (the derivative of glycolic acid) was derived from 20 (the
acetoxy-derivative) by means of a transesterification reaction performed with a substoichiometric
amount of sodium methoxide in THF and methanol.
3
3
. RESULTS AND DISCUSSION
.1. Reactivity with cysteine in solution
We set to investigate whether our newly prepared warheads could indeed engage in a
nucleophilic substitution, and evaluate their potential reactivity in a purely chemical setting, in
order to obtain a scale of relative reactivity. Quinazoline derivatives 8-23 were tested for their
ability to undergo a nucleophilic substitution in the presence of a large molar excess (1:100) of
cysteine at a 10 µM concentration in 50 mM N-cyclohexyl-3-aminopropanesulfonic acid (CAPS;
pH 10) buffer at 37 °C (Table 1). We set the pH of the buffer to 10 in order to have the thiol
8

ACCEPTED MANUSCRIPT
group of cysteine (pKa = 8.5 [32]) predominantly in its thiolate form and thus more prone to
react with electrophiles.
Consumption of the starting compounds and formation of the corresponding cysteine-
conjugate were monitored by LC-MS, adopting our previously reported protocol [33]. An
acrylamide
derivative
(N-(4-((3-chloro-4-fluorophenyl)amino)-7-ethoxyquinazolin-6-
yl)acrylamide, compound 9 [31a]) was included in this study as a benchmark, for its established
reactivity towards thiols. Acrylamide derivatives have high propensity to react with several sites.
This results in a high non-specific binding in cells, also at rather low compound concentration
1
8
(
i.e. 0.06 µM) as reported for [ F]-ML04, a labelled 4-anilinoquinazoline inhibitor bearing a 6-
acrylamide group [34]. This further stress the importance of finding chemical selective
warheads. Compound 8 was included considering the ability of the chloroacetamide group to
alkylate cysteines [28, 29]. Chloroacetamide derivatives have been recently shown to exhibit low
cross-reactivity versus a panel of eleven enzymes, suggesting that chloroacetamide itself could
be intrinsically less promiscuous than what initially expected in the presence of biological
nucleophiles [35].
Both the acrylamide 9 and the chloroacetamide 8 reacted with a large molar excess of cysteine
at pH 10 following pseudo first-order kinetics. LC-UV analysis showed that the disappearance of
compound 8 paralleled the formation of the cysteine-conjugate (see Supplementary Data, Fig.
S1) indicating the absence of competing reactions (e.g. hydrolysis). Compound 8 displayed a
half-life time (t1/2) of nearly 60 min (Table 1), resulting 4-fold less reactive than the acrylamide
analogue 9 (t1/2 = 15 min). Additionally, stability assays performed at pH 10 confirmed that both
compounds were stable for the tested time period. As opposed to 8 and 9, all the other
compounds reported in Table 1 reacted with cysteine either very slowly (t1/2 > 1440 min for
9

ACCEPTED MANUSCRIPT
compounds 10, 12, 14-22) or not at all (compounds 11, 13, 23), thereby precluding a quantitative
ranking of their relative electrophilicity. To qualitatively rank them, the area of the peak of the
cysteine-conjugate was measured by LC-MS after 24 h of reaction time (see Supplementary
Data, Fig. S2), and used as a relative descriptor of reactivity. The area of the covalent conjugate
given between 8 and cysteine was arbitrarily set at the reference value of 10,000, while the
cysteine-conjugate areas for the other compounds were referred to this value. The 2-((1H-
tetrazol-5-yl)thio)acetamide 14 and the 2-((1H-imidazol-2-yl)thio)acetamides 15 and 16 resulted
significantly more reactive than the 2-((pyridin-2-yl)thio)acetamide 10, the 2-((pyrimidin-2-
yl)thio)acetamide 12 and the 2-((1-methyl-benzimidazol-2-yl)thio)acetamide derivative 17, as
well as than its analogues 18 and 19. For 2-((pyridin-3-yl)thio)acetamide 11 and 3-(pyridin-3-
yl)-propanamide 13 no cysteine-conjugate could be observed. Among the carboxylate esters, 20,
2
1 and 22, the benzoyloxyacetamide derivative 21 showed a reactivity comparable to that of the
-((1H-imidazol-2-yl)thio)acetamide 16, while the acetoxy derivative 20 resulted one of the less
2
reactive compounds of the whole set. Additionally, for esters 20 and 21 significant amounts of
the hydrolyzed product (glycolate derivative, compound 23) could be observed after 24 h.
Compound 13, in which the sulfur atom on the acetamide portion is replaced by a methylene
unit, expectedly failed to give the conjugate, as well as 11, where the sulfur atom is not in the 2-
position of a nitrogen-containing heterocycle. Moreover, the hydroxyacetamide 23 did not show
any conjugate when incubated with cysteine.
As an additional test, we measured the chemical stability of the cysteine-conjugate, obtained in
situ by incubating a prototypical acetamide (i.e. 2-chloroacetamide 8) with a large excess of
cysteine, under the conditions of the reactivity assay (see above) by LC-MS. The cysteine-
conjugate was stable for at least 24 hours as evidenced by Fig. S3 of the supplementary data
10

ACCEPTED MANUSCRIPT
section. This result indicates that cysteine alkylation by activated acetamide is an irreversible
process.
3
.2. Simulation of Cys797 alkylation by QM/MM simulations
To support the hypothesis of covalent bond formation within EGFR, we evaluated whether the
2
putative warheads could undergo attack by Cys797 via a SN reaction pathway by means of a
quantum mechanics/molecular mechanics (QM/MM) approach [36]. The reaction mechanism
(depicted in Fig. 2) was simulated for the chloroacetamide 8 and for the two 2-(imidazol-2-
ylthio)acetamide derivatives 15 and 16, which emerged as the most promising compounds in the
reactivity assay. QM/MM mechanistic modelling is currently providing useful indications for
inhibitor design specifically when the formation and breakage of covalent bonds are involved in
the inhibitory process [37]. Models of enzyme-ligand non-covalent complexes were built by
docking 8, 15 and 16 within the ATP-binding pocket of the X-ray structure of the covalent
adduct of WT EGFR with a N-(4-anilinoquinazolin-6-yl) acrylamide inhibitor [8]. Previous
calculations performed by our group [38] suggest that Cys797 can act as nucleophile in the form
of thiolate anion, with the nearby Asp800 as the natural acceptor of the thiol hydrogen. The
-
Cys797/Asp800 pair was therefore modelled in Cys-S /Asp-COOH form (as reported in Fig. 2A).
Among the obtained docking poses, we selected those in which the quinazoline portion of the
inhibitors well fitted the hinge region of the ATP binding site, and the angle formed by the sulfur
3
atom of Cys797, the Csp of the acetamide fragment and the leaving group atom (i.e., chlorine
for 8, sulfur for 15 and 16) was higher than 130°. Such an angle approaches the value of 180° in
2
an ideal transition state (TS) for a SN reaction. The docking complexes were solvated and
equilibrated by classic molecular dynamics simulations (see Experimental Section for details).
2
The relaxed structures were then employed to model the SN reaction by means of a PDDG/PM3
11

ACCEPTED MANUSCRIPT
Hamiltonian [39] coupled to the AMBER99SB potential [40]. The PDDG/PM3 Hamiltonian has
2
N
been shown to yield energy barriers for S reactions in gas phase similar to those calculated by
ab initio methods [41]. In our approach, the substituted acetamide fragment of the inhibitor and
the side chains of Cys797 and of Asp800 were treated at PDDG/PM3 level, while the remaining
part of the system was described by AMBER99SB force field [42]. To account for solvation and
structural flexibility of both EGFR active site and inhibitor, the hybrid PDDG/PM3-
AMBER99SB Hamiltonian was used in steered molecular dynamics (SMD) simulations. To
simulate the nucleophilic substitution involving Cys797, we applied the SMD approach
implemented in AMBER, applying an adaptive force along a given reaction coordinate (RC) to
form or break covalent bonds, similarly to what described in reference [43]. In the case of
compound 8, thiolate addition and chlorine expulsion were modeled using the difference of
distances [d(Cαwarhead − Clwarhead) - d(Scys797 − Cαwarhead)]. For compounds 15 and 16, cysteine
thiolate attack and imidazol-2-ylthiolate expulsion were described by the difference [d(Cαwarhead
−
Swarhead) - d(Scys797 − Cαwarhead)]. While for compound 8 our QM/MM steered-MD approach led
to the identification of a stable alkylation product with low free-energy, this was not the case for
compounds 15 and 16 (data not shown). We thus hypothesized that protonation at the imidazole
ring could stabilize the imidazolylthio leaving group. In fact, the carboxylic acid of Asp800 was
close enough to the distal nitrogen at the imidazole ring to protonate it. For compounds 15 and
1
6, therefore, Cys797 alkylation was modelled starting from the cationic species resulting from
protonation of the nitrogen atom at the imidazole ring, with deprotonated carboxylate group of
Asp800 (Fig. 2B).
12

ACCEPTED MANUSCRIPT
Fig. 2. Reactions modelled by QM/MM simulation for compound 8, with expulsion of a chloride
ion (panel A) and for compounds 15 and 16 with expulsion of a protonated imidazolylthiolate
group (panel B). In the former case, Asp800 was modelled as carboxylic acid, while in the latter
as a carboxylate anion. In panel B, R stands for -CH (compound 15) or -H (compound 16).
3
The free-energy profiles calculated for Cys797 alkylation by compounds 8, 15 and 16 are
reported in Fig. 3. A single energy barrier separates reactants (R) from products (P). No
intermediates are identified as stable minima in QM/MM simulations, which is consistent with a
2
SN mechanism. The calculated activation energy for Cys797 alkylation, obtained applying the
Jarzynski equality [44], (see Experimental section) was 27.0 ± 0.7 kcal/mol for compound 8,
4
0.4 ± 1.2 kcal/mol for compound 15 and 34.2 ± 0.7 kcal/mol for compound 16. The calculated
reaction energy was -12.6 ± 0.7 and -10.2 ± 1.2 kcal/mol for 8 and 16 respectively, indicating
that Cys797 alkylation by these two compounds has an exergonic character. Conversely, the
calculated reaction energy for 15 was 0.9 ± 1.0 kcal/mol, suggesting that for this specific
compound the reaction was not energetically favored.
13

ACCEPTED MANUSCRIPT
Fig. 3. Free-energy curves with uncertainties for Cys797 alkylation by compounds 8 (blue), 15
(orange) and 16 (magenta), obtained by the Jarzynski equality as exponential average of five
independent replicas for each EGFR-inhibitor complex, using different snapshots taken from the
equilibrating QM/MM MD trajectory. Reaction coordinate is expressed in Å, while the free
energy is given in kcal/mol.
The geometries identified along the SMD trajectory indicate that the TS structure for the
alkylation of Cys797 by 8 and 16 features a trigonal bipyramidal geometry (Fig. 4A and Fig.
4
B). For compound 15 such a geometry is distorted, due to steric hindrance between the methyl
group of the imidazole ring and the p-loop of the protein (Fig. 4C). Calculations thus indicate
that compounds 8 and 16 (but not 15) can easily react with WT EGFR, through alkylation of the
Cys797 residue and leading to stable adducts. It is worth mentioning that these calculations were
performed contextually to the synthesis of the compounds and their biological testing. These
results were in fact critical for the decision to synthesize and test compound 16, which resulted
one of the most interesting ones on the gefitinib resistant cell line (vide infra).
14

ACCEPTED MANUSCRIPT
Fig. 4. Representative TS structures for Cys797 alkylation by compound 8 (panel A, cyan carbon
atoms), 16 (panel B, purple carbon atoms) and 15 (panel C, orange carbon atoms). Protein
carbon atoms are depicted in white, with EGFR secondary elements depicted as white cartoons.
3
.3. EGFR Inhibitory activity on A549 lung cancer cells
Compounds 8-23 were tested for their ability to inhibit EGFR autophosphorylation in A549
lung cancer cells, harboring WT EGFR. We decided to investigate the effect of the novel
warheads on this cell line because the 4-anilinoquinazoline scaffold possesses high affinity for
WT EGFR [13,16]. Gefitinib (1) and the acrylamide derivative (9) were tested as reference
inhibitors. All the synthesized compounds inhibited EGFR autophosphorylation, with IC50 values
ranging from the single digit nM for compounds 8, 11 and 14, to 83 and 197 nM for the 2-
(benzothiazolyl-2-thio)acetamides 18 and 19, respectively (Table 1). These data confirm that the
structural modifications introduced by the warheads on the 4-anilinoquinazoline nucleus allowed
to retain the high inhibitory potency of the parent structure.
To assess the persistency of EGFR inhibition within cells, test compounds were added to cell
cultures at 1 µM concentration and autophosphorylation in A549 cells was measured both 1 h
15

ACCEPTED MANUSCRIPT
after the addition and 8 h after washing of the inhibitor from the extra-cellular medium.
Inhibition enduring after compound removal could be ascribed to irreversible inhibition of the
kinase domain [45]. 8 h after inhibitor washing, gefinitib (1) gave 46% of inhibition, while
acrylamide 9 and chloroacetamide 8 inhibited EGFR autophosphorylation by more than 80%. In
the same conditions, compounds 10-12 inhibited EGFR by more than 70%, which is evocative of
a covalent interaction with the target Cys797. However, 60% of inhibition observed for
compound 13, which lacks a group able to undergo nucleophilic substitution, suggests that
persistent accumulation of the inhibitor within the cells may account for the observed inhibition
data, as it had been observed for other EGFR inhibitors [46]. The tetrazolyl (14), imidazolyl (15,
1
6), benzimidazolyl (17) and benzothiazolyl (18) derivatives also blocked EGFR
autophosphorylation 8 h after washing. Within this subset, the two imidazole and the
benzimidazole derivatives were able to reach at least 90% of inhibition.
Finally, we tested the set of ester derivatives 20-22 on A549 cells. Compounds with similar
groups were reported in the literature to covalently bind activated cysteines [47]. All these
derivatives were able to ensure EGFR inactivation 8 h after dilution. Despite its failure to give
detectable amounts of conjugate in the presence of cysteine at basic pH, also the
hydroxyacetamide 23 displayed significant long-term inhibition (77% of inhibition after 8 h).
Overall, the evaluation of the autophosphorylation levels of EGFR in A549 cells allowed to
identify several 4-anilinoquinazoline derivatives giving a percentage of EGFR inhibition close to
or even higher than 80% 8 h after inhibitor removal, which may indicate an irreversible binding
to the target [48]. A representative set of these compounds was further evaluated in vitro (vide
infra), to assess whether this prolonged inhibition can be ascribed to a covalent interaction with
the target.
16

ACCEPTED MANUSCRIPT
Table 1. LC-MS reactivity assay in the presence of cysteine and EGFR autophosphorylation
inhibition for compounds 1, 8–23 in A549 cell line.
Reactivity
Assay
c
Cpds
R
-
Autophosphorylation A549 cells
a
%
inhibition @
1µM
t
1/2
Peakb
IC50
(nM)
d
(
min)
Area
N. T.
1
h
8 h
1
e
e
N. T.
60±25
7.0±5.7
15±6.4
70±21
96 ± 4.0
100 ± 0.5
96 ± 4.0
73 ± 4.3
96 ± 4.0
99 ± 1.0
99 ± 0.50
46 ± 5.0
81 ± 2.9
99 ± 1.0
71± 8.1
89 ± 4.6
88 ± 16
60± 2.1
(
Gefitinib)
8
9
60 ± 5.0 10000
15 ± 3.0
> 1440
-
1
1
1
1
0
1
2
3
0.51
f
> 1440 N. D.
0.60±0.14
27±18
> 1440
0.66
f
> 1440 N. D.
14±1.4
17

ACCEPTED MANUSCRIPT
1
1
1
1
1
1
2
2
4
5
6
7
8
9
0
1
> 1440
> 1440
> 1440
> 1440
> 1440
> 1440
6.5
4.5
5.0±1.5
11±5.0
27±3.5
12±0.71
83±28
100 ± 6.4 75 ± 4.1
97 ± 2.5
95 ± 4.8
15
99 ± 0.10 90 ± 4.6
100 ± 1.0 100 ± 1.0
88 ± 0.90 75 ± 0.50
63 ± 0.10 93 ± 0.50
0.38
0.39
1.4
197±18.4
34±11
1
400 ±
00
0.64
97 ± 0.40
88 ± 3.1
75 ± 19
89 ± 5.7
1
> 1440
> 1440
11
5
25±12
44±0.71
27±11
2
2
2
3
97 ± 1.9
90 ± 3.6
97 ± 2.6
77 ± 11
f
> 1440 N. D.
a
b
Half-life measured at pH 10 and 37 °C by LC-MS. Area of Cys-conjugate of compound 8 was
arbitrarily set at the value of 10,000 and all the Cys-conjugate areas of remaining compounds
c
were referred to this value. Inhibition of EGFR autophosphorylation in A549 cells was
d
evaluated by Western blot analysis. Percentage of inhibition at 1 µM concentration was
measured after 1 h of treatment and 8 h after washing of the compound from the medium (1 h
e
f
incubation). Not tested. Not detected.
18

ACCEPTED MANUSCRIPT
3
.4. Intracellular dosage in A549 cells of selected inhibitors
In order to discriminate whether the observed EGFR inhibition 8 h after washing had to be
ascribed to a covalent interaction or to accumulation into A549 cells, intracellular dosages of
selected compounds were carried out. Chloroacetamide 8, 2-((1H-tetrazol-5-yl)thio)acetamide
1
4, 2-((1H-imidazol-2-yl)thio)acetamides 15 and 16 and 2-acetoxyacetamide 20 were selected as
representatives for highly (8), moderately (14-16) and weakly reactive (20) compounds,
according to their susceptibility to nucleophilic substitution in the presence of free cysteine
(Table 1). Intracellular concentrations of these compounds were measured in A549 cells by LC-
MS under conditions reproducing those of the EGFR autophosphorylation assay, i.e.
immediately after 1 h of exposure to the inhibitor (1 µM) or 8 h after washing the inhibitor from
the extracellular medium (Table 2).
Gefinitib (1) was chosen as a representative of non-covalent inhibitor of WT EGFR. The
results previously obtained for the acrylamide derivative PD168393 [48] (compound 24) are also
reported in Table 2 for comparison. Considering the complexity associated with the
measurement of intracellular volume, the concentrations reported herein are approximate.
Gefitinib (1) displayed an intracellular concentration of 271 µM after 1 h, when EGFR
autophosphorylation is inhibited almost completely (96%), while 8 h after washing its
concentration was approximately 100-times lower (2.4 µM) and autophosphorylation was
inhibited by 46% (Table 1). The intracellular concentration of acrylamide 24 was below the limit
of detection (LOD) of the LC-MS method both at 1 h and 8 h, as expected for its high propensity
to react promiscuously within the cell. This result clearly indicates irreversible inhibition of
19

ACCEPTED MANUSCRIPT
EGFR, while the case of gefitinib pointed out that residual concentrations of a non-covalent
inhibitor may cause some inhibition even 8 h after compound removal.
With the sole exception of the 2-acetoxyacetamide 20, all other tested compounds (8, 14-16)
showed high intracellular concentrations (i.e. > 100 µM) after 1h of incubation. This observation
corroborates the assumption of higher stability in the intracellular environment of our new set of
inhibitors with respect to acrylamide-based derivatives. However, it does not provide
information regarding the mechanism of action of the tested compounds. EGFR inhibition by
compound 20 in A549 cells after 1 h of incubation, when it showed a much lower concentration
than other inhibitors, may be ascribed to its hydrolyzed counterpart (compound 23), which
possesses the ability to inhibit EGFR autophosphorylation.
A significant drop in the intracellular concentration was observed 8 h after the washing from
A549 cells for all the tested compounds. However, only for chloroacetamide 8 and 2-((1H-
imidazol-2-yl)thio)acetamide 16 sub-micromolar levels were observed, with a 144- and 222-fold
reduction compared to the values measured after 1 h, respectively. While not a conclusive
evidence of covalent inhibition, it is remarkable that both compounds, having the ability to
alkylate cysteine groups, show long-term inhibition of EGFR autophosphorylation in spite of a
large decrease of their intracellular concentrations.
Table 2. Intracellular concentrations in A549 cells and inhibitory potency on recombinant WT
EGFR for selected anilinoquinazoline derivatives.
20

ACCEPTED MANUSCRIPT
a
Kinase
Assay
WT EGFR)
IC50 (nM)
Intracellular concentration
b
Cpds
R
-
Ratio 1 h/8 h
(µM)
(
1
h
8 h
1
271 ± 34
2.4 ± 0.2
112
-
0.47 ± 0.05
0.27 ± 0.04
0.87 ± 0.10
0.53 ± 0.09
N.T.
(
Gefitinib)
2
4
c
c
<
0.05
<0.05
(
PD168393)
8
115 ± 10
181 ± 2
179 ± 6
118 ± 10
5.8 ± 0.6
0.80 ± 0.10
5.6 ± 0.1
10.2 ± 2.1
0.53 ± 0.10
<0.05
144
32
18
222
-
14
15
16
20
23
0.62 ± 0.08
0.52 ± 0.03
0.34 ± 0.05
d
d
d
132 ± 10
1.1 ± 0.1
120
a
A549 cells were incubated with 1 µM solution of test compound for 1 h. Intracellular content
was quantified immediately after incubation and 8 h after removal of the compound from the
medium as pmol/mg of protein in each sample. Reported µM concentrations were calculated
dividing the compound content (in pmol) by an average value of cell volume of 4 µL per mg of
protein, estimated by measurements of the equilibrium distribution in A549 cells of 3-O-methyl-
3
b
D-[1- H]glucose using the method of Kletzien et al, [49]. Measured by LanthaScreen Eu kinase
21

ACCEPTED MANUSCRIPT
binding assay (see Experimental Section for details). Mean values of three independent
c
d
experiments ± SD are reported). Data from Vacondio et al. [45]. Generated in A549 cells from
the hydrolysis of compound 20.
3
.5. Binding to recombinant WT EGFR and EGFR L858R/790M double mutant
The ability of compounds 1, 8, 14-16, 20, 23 and 24 to displace a fluorescent tracer bound to
the kinase domain of WT EGFR was assessed by means of a time-resolved fluorescence
resonance energy transfer (TR-FRET) assay [50]. Compared to EGFR autophosphorylation in
intact cells, this assay yields inhibitory potency in the absence of cellular transport and/or
metabolism. Sub-nanomolar potencies were observed for all the tested compounds, with IC50
values ranging from 0.27 nM for the acrylamide derivative 24 to 0.87 nM for chloroacetamide 8.
Thus, the concentrations in the micromolar range, detected for compounds 8, 14, 15, 16 and 23 at
8
h, were significantly higher than IC values measured on the isolated enzyme. Therefore, these
50
concentrations could be sufficient to explain the inhibition of EGFR autophosphorylation
observed at 8 h (75-95%, see Table 1). On the other hand, being the IC50 values obtained in an
artificial environment, the TR-FRET assay provides only a relative order of potency among
different compounds, while the high intracellular concentration of ATP can compete with
reversible inhibitors at its binding site. In fact, gefitinib (1), a non-covalent inhibitor, only
inhibited by 46% EGFR autophosphorylation in A549 cells after 8 h, in spite of an intracellular
concentration of 2.4 µM. By comparison, 8 and 16 possess similar IC values on recombinant
5
0
WT EGFR, but they inhibited EGFR autophosphorylation in A549 cells, 8 h after washing,
significantly more than 1 (81% and 90% for 8 and 16 vs. 46% for 1). Considering that 8 and 16
reached intracellular concentration at 8 h lower than 1 (0.80 µM and 0.53 µM for 8 and 16 vs.
2
.4 µM for 1), it is conceivable that these two compounds might inhibit EGFR through a
covalent mechanism of action.
22

ACCEPTED MANUSCRIPT
We next tested the ability of 8 and 16 to inhibit EGFR L858R/T790M double mutant using the
TR-FRET binding assay (see Experimental section for details). Quinazoline-based inhibitors
usually display lower inhibitory potency for this mutant, when compared to WT or EGFR L858R
single mutant, due to steric clashes with M790 [9]. The insertion in the anilinoquinazoline core
of an electrophilic group, able to covalently bind the target, usually allows the inhibitor to
recover activity, but the impact on the measured potency depends on the rate of target alkylation.
Thus, when a slow covalent reaction occurs at the binding site, the observed IC50 value becomes
dependent on the time of incubation [51]. Fig. 5 reports the inhibitory potency for gefitinib (1,
panel A), 8 (panel B) and 16 (panel C) measured with no pre-incubation (left column) or pre-
incubating the inhibitor for 5 h with EGFR L858R/T790M double mutant (right column). The
IC value of 1 did not change with the pre-incubation, as expected for a non-covalent inhibitor,
5
0
while for compounds 8 and 16 a significant reduction in the IC values was observed. This
5
0
further supports a covalent mechanism of inhibition for these two compounds, also on EGFR
L858R/T790M enzyme.
A
23

ACCEPTED MANUSCRIPT
B
C
Fig. 5. Effect of the pre-incubation time on the potency (expressed as IC ) of gefitinib, 1 (panel
5
0
A), compound 8 (panel B) and compound 16 (panel C) on recombinant EGFR L858R/T790M.
IC were measured by means of a time-resolved fluorescence resonance energy transfer (TR-
5
0
FRET) assay using LanthaScreen technology. Values are reported as mean ± standard error of
the mean (n = 5). Statistical significance was set at P<0.05. **: P<0.01; ***: P<0.001.
TR-FRET experiments on EGFR L858R/T790M with 5 h inhibitor pre-incubation were
repeated either in the absence or in the presence of 500 nM ATP. Although the displacement
curves for compounds 1, 8 and 16 were barely affected by ATP (see supplementary data, Fig.
24

ACCEPTED MANUSCRIPT
S4), the limited signal-to-noise ratio of these TR-FRET measures precludes a conclusive
assessment of the effect of ATP on the inhibitory process for the tested inhibitors.
3
.6. Selectivity on kinases bearing a conserved cysteine on the ATP front pocket
We next tested the ability of compounds 8, 9 and 16 to inhibit a panel of kinases reported to
have a cysteine in a position corresponding to that of Cys797 in EGFR [52]. In this investigation,
the following kinases were included: i. ErbB2 and ErbB4 of the Erb family, the same of EGFR,
ii. BMX and BTK belonging to the Tek family, iii. JAK3 representative of the Janus kinase
family and iv. BLK, member of the Src family. Acrylamide 9 inhibited the activity of nearly all
the kinases of the panel at 1 µM, in agreement with previous findings reported in the literature
[52]. On the contrary, 2-chloroacetamide 8 and imidazol-2-ylthioacetamide 16, markedly
inhibited EGFR and ErbB4 only (i.e. 87% of higher), as they moderately (ErbB2, BTX and
BLK) or marginally (BMX, JAK3) affected the activity of the other kinases included in the
panel. This assay confirms further that the replacement of acrylamide with less reactive warheads
ensures higher selectivity.
Table 3. Selectivity of compounds 8, 9 and 16 on a panel of kinases bearing a front-pocket
a
cysteine corresponding to EGFR Cys797.
Cpds
EGFR
98
ErbB2
76
ErbB4
100
BMX
38
BTK
53
JAK3
32
BLK
62
8
9
90
92
100
88
87
92
97
1
6
90
65
87
30
54
0
58
a
Inhibitory activity determined with an activity-based assay (see methods for details). Percent
of inhibition measured at 1 M. Color code: red,
80% inhibition; yellow, 50−80% inhibition;
green, 50% inhibition. Reported data are the average of two independent experiments.
µ
≥
≤
25

ACCEPTED MANUSCRIPT
3
.7. Activity on H1975 lung cancer cells
H1975 lung cancer cells are resistant to first generation inhibitors of EGFR due to the presence
of the T790M mutation which somehow hampers the interaction between the 4-
anilinoquinazoline nucleus and ATP-binding site (see above). The introduction in the
quinazoline scaffold of a group capable of alkylating Cys797 is expected to ensure a significant
improvement in the antiproliferative potency on H1975 cells compared to gefitinib [16]. With
this premise, we examined the ability of the synthesized compounds to block proliferation of
H1975 cells and to inhibit EGFR autophosphorylation at two concentrations (0.1 and 1 µM) in
the same cell line (Table 4) along with gefitinib (1) and acrylamide 9, which were used as
controls as they are expected to be the least and most potent compounds of the series,
respectively.
Compound 1 resulted poorly active on H1975 cells [53] as it inhibited cell growth with an
IC50 of 9.1 µM and it did not affect autophosphorylation at 0.1 or 1 µM concentration,
confirming its incapability to engage EGFR L858R/T790M mutant. On the contrary, the
acrylamide
9
showed sub-micromolar antiproliferative potency and inhibited
autophosphorylation almost completely at both concentrations. The same compound had shown
an antiproliferative IC50 value of 1.6 µM in a previous experiment [31a]. Among the newly
synthesized compounds, the hydroxyacetamide 23, which lacks a potential warhead, did not
show any antiproliferative activity up to 10 µM and did not inhibit EGFR L858R/T790M
autophosphorylation. Chloroacetamide 8 and the 2-((1H-imidazol-2-yl)thio)acetamide 16
significantly outperformed gefitinib and displayed a pharmacological activity superior (8) or
comparable (16) to that of the acrylamide 9. Indeed, compound 8 resulted extremely potent in
26

ACCEPTED MANUSCRIPT
inhibiting H1975 cell proliferation (IC50 of 0.06 µM) and capable of fully abolishing EGFR
L858R/T790M autophosphorylation at 0.1 µM. Compound 16 inhibited H1975 growth with an
IC50 of 1.4 µM and completely inhibited autophosphorylation at 1 µM concentration.
Remarkably, the ability to overcome the effect of T790M mutation in H1975 cells is achieved
introducing electrophilic groups featuring moderate (8) or very low reactivity (16) toward free
cysteine, when compared to acrylamide.
Table 4. Inhibition of cell proliferation and inhibition of EGFR autophosphorylation in H1975
cells for compounds 1, 8-12 and 14-23.
Cpds
R
H1975 cell line
Autophosphorylation
of inhibition @
1 µM
b
Cell growth
Inhibition
a
%
IC (µM)
0.1 µM
8.0 ± 4.9
100 ± 0.9
93 ± 0.70
1.5 ± 2.1
5
0
1
-
9.1 ± 1.1
0.06 ±0.01
0.70 ± 0.14
3.7 ±0.28
18 ± 4.2
100 ± 0.10
95 ± 0.10
42 ± 7.0
(
Gefitinib)
8
9
10
27

ACCEPTED MANUSCRIPT
11
12
14
15
16
17
18
19
20
3.0 ±0.42
3.6 ±1.8
2.1 ±0.78
6.9 ±0.64
1.4 ±0.10
9.1±1.0
29 ± 10
40 ± 12
16 ± 8.5
45 ± 14
80 ± 7.0
6.0 ± 4.2
29 ± 11
30 ± 13
27 ± 8.5
92 ± 9.9
87 ± 9.8
96 ± 5.7
67 ± 8.5
100 ±0.71
70 ± 6.8
24 ± 7.7
92 ± 0.90
68 ±13
9.8±0.14
3.2±0.71
3.4±0.57
21
22
23
8.3±1.8
6.1±1.4
> 10
0.5 ± 0.7
20 ± 3.9
0.0 ± 0.1
43 ± 4.4
16 ± 8.0
0.0 ± 0.1
a
Concentration able to inhibit 50% of cell proliferation as determined by the MTT assay, after 72
h of incubation with compounds (dose ranging from 0.1 to 10 µM). Osimertinib displays an IC50
28

ACCEPTED MANUSCRIPT
b
of 0.10 ± 0.1 µM in the employed conditions. Inhibition of EGFR autophosphorylation in
H1975 cells was evaluated by Western blot analysis.
4
. CONCLUSION
We report a structure-reactivity and structure-activity study aimed to identify novel cysteine-
targeting warheads, less reactive than acrylamide, that may be exploited for the design of
irreversible kinase inhibitors with improved chemical selectivity. In particular, we focused on the
design and synthesis of 4-anilinoquinazoline derivatives bearing warheads with reduced risk of
nonspecific covalent binding to thiols, yet able to exploit the high nucleophilicity of Cys797 of
EGFR and to overcome the resistance conferred by the T790M gatekeeper mutation to
reversible, anilinoquinazoline-based inhibitors. Starting from the chloroacetamide derivative 8,
2
reduction of reactivity for a series of potential SN substrates led to compound 16, a 2-((1H-
imidazol-2-yl)thio)acetamide derivative, which showed a inhibitory potency on H1975 cell
growth (IC50 = 1.4 ± 0.10 µM) comparable to that of the reference acrylamide-based inhibitor 9
(
IC = 0.70 ± 0.10 µM). Considering that 9 and 16 have the same 4-anilino-quinazoline scaffold,
50
the imidazol-2-ylthioacetamide group emerges as a bioisostere of the acrylamide warhead. It
should be pointed out that while compound 16 displays negligible reactivity towards free
cysteine in solution and possess higher selectivity than acrylamide 9 toward a panel of selected
kinases, its imidazol-2-ylthioacetamide warhead is a metabolic soft spot potentially able to give
toxic effects in vivo.
Several hints, including hybrid QM/MM simulations and pre-incubation studies on recombinant
EGFR L858R/T790M double mutant, support the hypothesis that compound 16 irreversibly
inhibits EGFR by reacting covalently with Cys797. By comparison with acrylamide-based
inhibitors, we propose that long-lasting inhibition can be achieved by modulating the reactivity
29