Targeting the KRAS oncogene: Synthesis, physicochemical and biological evaluation of novel G-Quadruplex DNA binders

Article abstract of DOI:10.1016/j.ejps.2020.105337

The oncogene KRAS is involved in the pathogenesis of many tumors such as pancreatic, lung and colorectal cancers, thereby representing a relevant target for the treatment of these diseases. The KRAS P1 promoter contains a nuclease hypersensitive, guanine-rich sequence able to fold into a G-quadruplex motif (G4). The stabilization of this G4 structure by small molecules is emerging as a feasible approach to downregulate KRAS expression. Here, a set of novel stabilizing molecules was identified through a virtual screening campaign on the NMR structure of the 22-mer KRAS G4. The most promising hits were then submitted to structure-activity relationships studies which allowed improving their binding affinity and selectivity over double helix DNA and different G4 topologies. The best derivative (19) underwent fluorescence titration experiments and further computational studies to disclose its binding mechanism to KRAS G4. Finally, biological assays showed that this compound is capable to reduce the viability of colorectal cancer cells in which mutated KRAS acts as a driver oncogene. Thus, 19 might represent the prototype of a new class of drugs for the treatment of tumors that, expressing mutated forms of KRAS, are refractory to current therapeutic regimens.

Full text of DOI:10.1016/j.ejps.2020.105337

Journal Pre-proof
Targeting the KRAS Oncogene: Synthesis, Physicochemical and
Biological Evaluation of Novel G-Quadruplex DNA Binders
Federica D’Aria Physicochemical Studies; Data curation; Validation; Writing - Original draft preparation ,
Vincenzo Maria D’Amore Computational Studies; Data curation; Validation; Writing- Original draft preparation ,
Francesco Saverio Di Leva Computational studies; Conceptualization; Supervision; Funding acquisition; Writing - r
Jussara Amato Physicochemical Studies; Superivision ,
Marco Caterino Physicochemical Studies ,
Pasquale Russomanno Computational Studies ,
Silvia Salerno Chemical Synthesis ,
Elisabetta Barresi Chemical Synthesis ,
Marinella De Leo Chemical Synthesis ,
Anna Maria Marini Chemical Synthesis ,
Sabrina Taliani Chemical Synthesis; Supervision; Funding acquisition ,
Federico Da Settimo Chemical Synthesis; Supervision ,
Gilmar F. Salgado Writing - reviewing & editing ,
Luca Pompili Biological evaluation ,
Pasquale Zizza Biological evaluation; Data Validation; Writing - Original draft preparation ,
Senji Shirasawa Biological evaluation ,
Ettore Novellino Supervision; Funding acquisition ,
Annamaria Biroccio Biological evaluation; Supervision; Funding acquisition ,
Luciana Marinelli Computational Studies; Conceptualization; Supervision; Funding acquisition ,
Concetta Giancola Phisicochemical Studies; Conceptualization; Supervision; Funding acquisition; Writing - reviewi
PII:
S0928-0987(20)30126-3
DOI:
Reference:
https://doi.org/10.1016/j.ejps.2020.105337
PHASCI 105337
To appear in:
European Journal of Pharmaceutical Sciences
Received date:
Revised date:
Accepted date:
10 February 2020
30 March 2020
1 April 2020
Please cite this article as: Federica D’Aria Physicochemical Studies; Data curation; Validation; Writing - Original dra
Vincenzo Maria D’Amore Computational Studies; Data curation; Validation; Writing- Original draft preparation ,
Francesco Saverio Di Leva Computational studies; Conceptualization; Supervision; Funding acquisition; Writing - r
Jussara Amato Physicochemical Studies; Superivision , Marco Caterino Physicochemical Studies ,
Pasquale Russomanno Computational Studies ,
Elisabetta Barresi Chemical Synthesis ,
Silvia Salerno Chemical Synthesis ,
Marinella De Leo Chemical Synthesis ,
Anna Maria Marini Chemical Synthesis , Sabrina Taliani Chemical Synthesis; Supervision; Funding acquisition ,
Federico Da Settimo Chemical Synthesis; Supervision , Gilmar F. Salgado Writing - reviewing & editing ,
Luca Pompili Biological evaluation , Pasquale Zizza Biological evaluation; Data Validation; Writing - Original draft p
Senji Shirasawa Biological evaluation ,
Ettore Novellino Supervision; Funding acquisition ,
Annamaria Biroccio Biological evaluation; Supervision; Funding acquisition , Luciana Marinelli Computational Stud
Concetta Giancola Phisicochemical Studies; Conceptualization; Supervision; Funding acquisition; Writing - reviewi
Targeting the KRAS Oncogene: Synthesis, Physicochemical and Biological Evaluation of

Novel G-Quadruplex DNA Binders, European Journal of Pharmaceutical Sciences (2020), doi:
https://doi.org/10.1016/j.ejps.2020.105337
This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition
of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of
record. This version will undergo additional copyediting, typesetting and review before it is published
in its final form, but we are providing this version to give early visibility of the article. 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.
© 2020 Elsevier B.V. All rights reserved.

Targeting the KRAS Oncogene: Synthesis,
Physicochemical and Biological Evaluation of Novel
G-Quadruplex DNA Binders
Federica D’Ariaa,‡, Vincenzo Maria D’Amorea,‡, Francesco Saverio Di Levaa,*, Jussara Amatoa,
Marco Caterinoa, Pasquale Russomanno,a Silvia Salernob, Elisabetta Barresib, Marinella De
Leob, Anna Maria Marinib, Sabrina Talianib, Federico Da Settimob, Gilmar F. Salgadoc, Luca
Pompilid, Pasquale Zizzad, Senji Shirasawae, Ettore Novellinoa, Annamaria Birocciod, Luciana
Marinellia, Concetta Giancolaa,*
aDepartment of Pharmacy, University of Naples Federico II, via D. Montesano 49, 80131 Napoli, Italy;
bDepartment of Pharmacy, University of Pisa, via Bonanno 6, 56126, Pisa, Italy;
cARNA Laboratory, IECB, University of Bordeaux, Inserm U1212, CNRS UMR 5320, F-33600 Pessac,
France;
dOncogenomic and Epigenetic Unit, IRCCS – Regina Elena National Cancer Institute, Via Elio Chianesi
53, 00144, Rome, Italy;
eCentral Research Institute for Advanced Molecular Medicine, Fukuoka University, Fukuoka, Japan.
*Corresponding authors:
E-mail: concetta.giancola@unina.it (C. Giancola), francesco.dileva@unina.it (F.S. Di Leva)
‡These authors contributed equally.
KEYWORDS: DNA G-quadruplex; KRAS; Drug Discovery; Cancer; Medicinal Chemistry;
Molecular Modeling; Physicochemical Studies
1

Abstract
The oncogene KRAS is involved in the pathogenesis of many tumors such as pancreatic, lung and
colorectal cancers, thereby representing a relevant target for the treatment of these diseases. The
KRAS P1 promoter contains a nuclease hypersensitive, guanine-rich sequence able to fold into a
G-quadruplex motif (G4). The stabilization of this G4 structure by small molecules is emerging
as a feasible approach to downregulate KRAS expression. Here, a set of novel stabilizing
molecules was identified through a virtual screening campaign on the NMR structure of the 22-
mer KRAS G4. The most promising hits were then submitted to structure-activity relationships
studies which allowed improving their binding affinity and selectivity over double helix DNA
and different G4 topologies. The best derivative (19) underwent fluorescence titration
experiments and further computational studies to disclose its binding mechanism to KRAS G4.
Finally, biological assays showed that this compound is capable to reduce the viability of
colorectal cancer cells in which mutated KRAS acts as a driver oncogene. Thus, 19 might
represent the prototype of a new class of drugs for the treatment of tumors that, expressing
mutated forms of KRAS, are refractory to current therapeutic regimens.
2

1. Introduction
1.1 G-quadruplexes. Nowadays it is well known that DNA can form alternative higher-order
structures that are believed to play important roles in a number of biological processes and in the
progression of several diseases (Bacolla and Wells, 2009; Zhao et al., 2010, Amato et al., 2014b;
Balasubramanian et al., 2011). Among non-canonical DNA structures are G-quadruplexes (G4s)
which are formed by guanine-rich sequences that assemble into guanine tetrads stabilized by
Hoogsteen hydrogen bonds and monovalent cations (Bochman et al., 2012). In the last decades,
G4s have drawn growing attention as possible druggable anticancer targets because they are
located in key genome regions, including promoters of oncogenes such as MYC, VEGF, BCL2,
KIT, and KRAS (Caterino et al., 2020; Cogoi et al., 2004; Cogoi and Xodo, 2006; Dai et al.,
2006a, 2006b; Fernando et al., 2006; Rankin et al., 2005; Siddiqui-Jain et al., 2002; Simonsson et
al., 1998; Sun et al., 2005), whose expression can be downregulated by G4-stabilizing small
molecules (Balasubramanian et al., 2011).
1.2 KRAS. Remarkably, the KRAS gene is overexpressed in about 30% of all human cancers,
including pancreatic ductal adenocarcinoma, lung and colorectal cancers (Cogoi et al., 2013;
Kang et al., 2017; Miglietta et al., 2017; Stephen et al., 2014). KRAS (Gene ID: 3845) maps at
12p12.1 and encodes for a 21 kDa membrane GDP/GTPase (UniProt: P01116) whose on/off
state is governed by the GTP/GDP homeostasis. Missense point mutations that shift such balance
(most frequently at codon 12 and 13) are responsible for a constitutive activation of KRAS and
downstream cellular signaling pathways (e.g. RAF, MEK, or PI3K) (Cox et al., 2014). A direct
targeting of KRAS and its downstream effectors at protein level has proved unfruitful since long
(Haigis, 2017; Hobbs et al., 2016; McCormick, 2016, 2018). Nevertheless, in recent years
specific inhibitors of the constitutively active G12C KRAS mutant have been developed,
3

showing promising therapeutic potential (Janes et al., 2018; Lito et al., 2016; Ostrem et al., 2013;
Patricelli et al., 2016).
1.3 KRAS G4. The nuclease-hypersensitive element (NHE) within the KRAS P1 promoter
features a G-rich sequence that folds into a G4 whose stabilization by small molecules offers a
convenient alternative to downregulate KRAS at gene-level (Cogoi et al., 2008; Cogoi and Xodo,
2016; Hoffman et al., 1990; Jordano and Perucho, 1986). Diverse classes of compounds have
been tested to this end, such as porphyrins, acridines, anthraquinones, phenanthrolines, perylenes
and quinolines (Alzeer et al., 2009; Boschi et al., 2016; Calabrese et al., 2018; Carvalho et al.,
2018; Lavrado et al., 2013, 2015; Monchaud and Teulade-Fichou, 2008; Pattanayak et al., 2018).
The majority of these compounds share polycyclic and heteroaromatic moieties that grant decent
selectivity over the duplex DNA, mainly due to stronger π-π stacking interactions with the
guanine tetrads. However, several of these molecules target a large set of G4s rather than a single
one, which may cause off-target effects, and lack basic drug-like properties.
1.4 Targeting KRAS G4. In this study, a structure-based virtual screening (VS) was performed
on the NMR structure of the KRAS P1 promoter G-quadruplex (KRAS G4) (Kerkour et al., 2017),
with the aim to find novel stabilizers of this DNA motif. The VS selected hits were evaluated by
thermal melting experiments, and the most promising compounds were then chosen for likewise
lead optimization program. The newly synthesized molecules were again evaluated for their
KRAS G4 stabilizing properties and then assayed for selectivity toward both diverse G4
topologies and duplex DNA. The best performing derivative underwent further fluorescence
assays and molecular dynamics calculations to gain insight into the stoichiometry and binding
affinity of its interaction with KRAS G4, and to disclose its binding mode to this DNA motif.
Also, biological analyses were performed to evaluate the effect of this compound on both the
4

KRAS gene expression and viability in tumor cells expressing constitutively active forms of
mutated KRAS (Downward, 2003).
2. Materials and Methods
2.1 Virtual Screening and Molecular Docking. For our study, an in-house virtual database of
5,858 compounds held by the Computational Chemistry Lab of the Department of Pharmacy of
the University of Naples Federico II and the commercially available Asinex Platinum Collection
library (http://www.asinex.com) of 9,216 molecules were selected. Both the selected libraries
had been preliminary filtered to include molecules endowed with lead-like pharmacokinetic
properties, according to the Lipinski’s rule of five. All the possible tautomeric and protonation
states in the pH range 7.4 ± 1.5 were generated for each compound using Epik (Greenwood et al.,
2010; Shelley et al., 2007) for a total amount of 7,924 and 17,500 structures, respectively, for the
in-house and the Asinex libraries. A filtering procedure (see paragraph 3.1 for details) was then
applied to retain all the molecules having at least two aromatic rings and a positive total charge,
resulting in two final subsets containing, respectively, 875 and 14,280 structures. The NMR
structure of the KRAS G4 DNA motif (PDB code: 5I2V) (Kerkour et al., 2017) was selected as
target macromolecule and prepared through the Protein Preparation Wizard implemented in
Maestro Suite 2019 (Madhavi Sastry et al., 2013). During the preparation, all water molecules
were deleted, hydrogen atoms were added, and the complex was minimized. The docking search
area was set on the center of mass of the macromolecule so as to enclose the entire G4. The
interaction grids were thus computed through the grid generation tool of Glide 6.7 (Friesner et
al., 2004; Halgren et al., 2004). The OPLS 2005 force field (Banks et al., 2005) was employed
for docking. The best 500 initial poses per ligand were retained for post-docking energy
5

minimization. Otherwise, default parameters were applied. The results from each set of
calculations were evaluated and ranked based on the Glide SP scoring function (Friesner et al.,
2004; Halgren et al., 2004). For each compound, only the best scored docking pose was retained.
The top-ranked molecules (the best 15 %) of each subset were then visually inspected for their
binding modes and for good chemical geometry. Compounds 1-4 were retrieved from our in-
house library, while compounds 5-12 were purchased from the Asinex vendor. Compounds
purity (> 95 %) was determined by HPLC, according to the procedure described in the chemistry
section.
2.2
d(AGGGCGGTGTGGGAATAGGGAA) (KRAS G4), d(AGGGAGGGCGCTGGGAGGAGGG)
(KIT G4), d(TAGGGTTAGGGTTAGGGTTAGGG) (Tel23), and
DNA
synthesis
and
sample
preparation.
The
sequences
d(CGAATTCGTTTTCGAATTCG) (hairpin duplex), were synthesized on a ABI394 DNA/RNA
synthesizer (Applied Biosystem, Foster City, CA, USA) by using the standard DNA synthesis
protocol on solid phase at the 5-μmol scale (Oliviero et al., 2010). A concentrated ammonia
aqueous solution at 55 °C for 12 h was used to DNA detachment from support and removal of
the semi-permanent protection groups. The combined filtrates and washings were concentrated
under reduced pressure, dissolved in water, and purified by HPLC on a Nucleogel SAX column
(Macherey-Nagel, 1000-8/46), using buffer A consisting of 20 mM KH2PO4/ K2HPO4 aqueous
solution (pH 7.0), containing 20% (v/v) CH3CN, buffer B consisting of 1 M KCl, 20 mM
KH2PO4/ K2HPO4 aqueous solution (pH 7.0), containing 20% (v/v) CH3CN, and a linear gradient
from 0% to 100% B for 30 min with a flow rate 1 mL min-1. The fractions of the oligomers were
collected and successively desalted by Sep-Pak cartridges (C-18).
6

Oligonucleotide samples were prepared by dissolving the lyophilized DNAs in potassium
phosphate buffer (60 mM KCl, 20 mM KH2PO4, 0.1 mM EDTA, pH 7.0). The solutions were
◦
heated at 90 C for 5 min and slowly cooled to room temperature. The annealing procedure was
followed by overnight storage at 4 ◦C prior to use. The concentration of the oligonucleotides was
◦
evaluated by UV measurements at 260 nm, at a temperature of 90 C, using molar extinction
coefficient values calculated by the nearest-neighbor model (Cantor et al., 1970). Stock solutions
of putative ligands were prepared at 10 mM concentration in DMSO. Appropriate amounts of
ligand were added to the oligonucleotide solutions to obtain the desired ligand/DNA ratio,
followed by mix at 20 ◦C for 10 min before data acquisition.
2.3 Circular Dichroism. CD spectra were recorded on a Jasco J-815 spectropolarimeter (JASCO
Inc., Tokyo, Japan) equipped with a PTC-423S/15 Peltier temperature controller. All the spectra
were recorded at 20 °C in the 220-360 nm wavelength range and averaged over three scans. The
scan rate was 100 nm min-1, with a 4 s response and 1 nm bandwidth. Buffer baseline was
subtracted from each spectrum. Sample concentration was 2 µM for all DNA samples. CD
spectra of DNA/ligand mixtures were obtained by adding 10 mol equiv of ligand. CD melting
experiments were carried out in the 20-100 °C range at 1 °C min-1 heating rate and followed at
an appropriate wavelength for each DNA sample (KRAS G4 and KIT G4 at 264 nm, Tel23 at 289
nm, and the duplex hairpin at 280 nm). The melting temperatures (Tm) were determined from
curve fit using Origin 7.0 software (OriginLab Corp., Northampton, MA, USA). CD melting
experiments were recorded on DNA samples both in the absence and presence of each ligand to
obtain ΔTm values, determined as:
∆푇푚 = 푇푚,퐷푁퐴/푙푖푔푎푛푑 − 푇푚,퐷푁퐴
7

where Tm,DNA/ligand is measured on ligand/DNA mixtures (10 mol equiv), Tm,DNA is from the
standalone DNA sample. All experiments were performed in duplicate and the reported values
are the average of two measurements.
2.4 Chemistry. The uncorrected melting points were determined using a Reichert Köfler hot-
stage apparatus. NMR spectra were obtained on a Bruker AVANCE 400 (1H, 400 MHz, 13C, 100
MHz) in DMSO-d6 or in CDCl3 (Fig. S1 and S2). The coupling constants are given in Hertz.
Magnesium sulfate was used as the drying agent. Evaporations were made in vacuo (rotating
evaporator). Analytical TLC have been carried out on Merck 0.2 mm precoated silica gel
aluminum sheets (60 F-254). Silica gel 60 (230–400 mesh) was used for column
chromatography. ESI-MS/MS were obtained in positive ion mode from an LCQ Advantage
ThermoFinnigan spectrometer (ThermoFinnigan, USA). Microwave assisted reactions were
carried out in BIOTAGE Initiator 2.5 microwave apparatus. Reagents, starting materials, and
solvents were purchased from commercial suppliers and used as received. The following
compounds were obtained according to methods previously described: 6-(2-((2-
hydroxyethyl)amino)ethyl)benzo[4,5]imidazo[2,1-b]quinazolin-12(6H)-one (1) (Dalla Via et al.,
2001), N,10-bis(2-(diethylamino)ethyl)-4-oxo-4,10-dihydrobenzo[4,5]imidazo[1,2-a]pyrimidine-
3-carboxamide
(2)
(Da
Settimo
et
al.,
2003),
7-(3-
(dimethylamino)propyl)benzo[4,5]imidazo[1,2-a]pyrido[3,2-e]pyrimidin-5(7H)-one (3) (Da
Settimo et al., 1998), 1,11-dimethyl-5-oxo-5,11-dihydrobenzo[4,5]imidazo[1,2-a]pyrido[2,3-
d]pyrimidin-1-ium iodide (4) (Caroti et al., 1987), 1-dialkylaminoalkyl-2-aminobenzimidazoles
(20a-d) (Dalla Via et al., 2001) and 11-methylbenzo[4,5]imidazo[1,2-a]pyrido[2,3-d]pyrimidin-
5(11H)-one (21) (Caroti et al., 1987).
8

Purity of the target compounds was determined using a Shimadzu LC-20AD SP liquid
chromatograph equipped with a DDA Detector at 254 nm (column C18 (250 mm x 4.6 mm,
5m, Shim-pack)). The mobile phase, delivered at isocratic flow, consisted of acetonitrile (60%)
and phosphate buffered saline (PBS) (40%) and a flow rate of 1.0 mL/min. All the compounds
showed percent purity values of ≥ 95%.
The structural characterization of previously described target compounds 1-4 and 13-16 has been
herein implemented by 13C NMR data.
6-(2-((2-Hydroxyethyl)amino)ethyl)benzo[4,5]imidazo[2,1-b]quinazolin-12(6H)-one (1). 13C
NMR (100 MHz, DMSO-d6):  42.27, 47.17, 51.77, 60.78, 109.94, 115.53, 116.60, 122.03,
122.97, 125.63, 125.93, 126.42, 126.84, 132.53, 134.82, 146.87, 149.48, 159.61.
N,10-Bis(2-(diethylamino)ethyl)-4-oxo-4,10-dihydrobenzo[4,5]imidazo[1,2-a]pyrimidine-3-
carboxamide (2).13C NMR (100 MHz, CDCl3):  12.07, 12.30, 37.62, 41.93, 47.59, 47.64, 50.90,
52.08, 107.16, 110.11, 117.73, 123.80, 125.95, 127.10, 131.54, 149.75, 159.34, 160.27, 164.85.
7-(3-(Dimethylamino)propyl)benzo[4,5]imidazo[1,2-a]pyrido[3,2-e]pyrimidin-5(7H)-one (3).13C
NMR (100 MHz, CDCl3):  26.29, 40.57, 45.51, 56.43, 110.08, 114.96, 116.30, 122.20, 123.86,
125.64, 126.49, 131.49, 138.65, 149.20, 151.00, 152.71, 168.85.
1,11-Dimethyl-5-oxo-5,11-dihydrobenzo[4,5]imidazo[1,2-a]pyrido[2,3-d]pyrimidin-1-ium
13
iodide (4). C NMR (100 MHz, DMSO-d6):  29.14, 42.02, 111.34, 115.39, 116.96, 124.60,
124.81, 127.30, 131.50, 144.78, 148.79, 150.44, 152. 52, 156.86.
General procedure for the synthesis of 6-(aminoalkyl)benzo[4,5]imidazo[2,1-b]quinazolin-
12(6H)-one derivatives 13-16. A mixture of the appropriate 1-dialkylaminoalkyl-2-
aminobenzimidazole 20a-d (0.50 mmol), 2-chlorobenzoic acid (0.078 g, 0.50 mmol), cuprous
bromide (0.004 g, 0.030 mmol), anhydrous potassium carbonate (0.083 g, 0.60 mmol) and a
9

catalytic amount of potassium iodide in 1.0 mL of dimethylformamide was irradiated at a T =
150 °C, P = 10 bar, power = 40-150 Watt, ramp time = 2 min for 12 min. After cooling the
reaction mixture was poured into ice. The precipitated crude products 13-16 were collected and
purified by recrystallization from ethanol.
6-(2-(Dimethylamino)ethyl)benzo[4,5]imidazo[2,1-b]quinazolin-12(6H)-one (13). Yield: 70%;
m.p. 161-162 °C (lit. ref. 43, m.p. 160-165 °C); 13C NMR (100 MHz, CDCl3):  40.44, 45.88,
56.75, 108.51, 116.62, 117.12, 122.37, 123.17, 126.02, 126.20, 126.35, 127.26, 131.99, 134.59,
146.69, 149.55, 160.31.
6-(2-(Diethylamino)ethyl)benzo[4,5]imidazo[2,1-b]quinazolin-12(6H)-one (14). Yield: 65%;
m.p. 143-144 °C (lit. ref. 43, m.p. 140-145 °C); 13C NMR (100 MHz, CDCl3):  11.49, 40.13,
47.70, 50.07, 108.78, 116.60, 117.16, 122.47, 123.26, 126.00, 126.16, 126.47, 127.33, 131.99,
134.65, 146.68, 149.53, 160.32.
6-(3-(Dimethylamino)propyl)benzo[4,5]imidazo[2,1-b]quinazolin-12(6H)-one (15). Yield: 55%;
13
m. p. 163-164 °C (lit. ref. 43, m.p. 163-165 °C); C NMR (100 MHz, CDCl3):  25.85, 40.31,
45.25, 56.51, 108.55, 116.61, 117.10, 122.39, 123.21, 126.01, 126.12, 126.42, 127.30, 132.15,
134.66, 146.74, 149.60, 160.34.
6-(3-(Diethylamino)propyl)benzo[4,5]imidazo[2,1-b]quinazolin-12(6H)-one (16). Yield: 65%;
m.p. 131-133 °C (lit. ref. 43, m.p. 130-135 °C); 13C NMR (100 MHz, CDCl3):  10.64, 24.99,
40.57, 46.96, 50.27, 108.59, 116.66, 117.14, 122.53, 123.29, 125.93, 126.14, 126.52, 127.35,
131.93, 134.71, 146.77, 149.52, 160.30.
General synthetic procedures for 3-(2-substituted-ethyl)-12-methyl-6-oxo-2,3,6,12-
tetrahydro-1H-benzo[4,5]imidazo[1,2-a]imidazo[1',2':1,6]pyrido[2,3-d]pyrimidin-14-ium
bromide 17-19. The appropriate amine (10.0 mmol) was added dropwise to a suspension of 22
10

(0.110 g, 0.25 mmol) in 10 ml of ethanol and the mixture was refluxed for 4 h. After cooling, the
resulting suspension was filtered, and the crude products obtained were purified by
recrystallization from ethanol.
3-(2-Hydroxyethyl)-12-methyl-6-oxo-2,3,6,12-tetrahydro-1H-benzo[4,5]imidazo[1,2-a]imidazo
1
[1',2':1,6]pyrido[2,3-d]pyrimidin-14-ium bromide (17). Yield: 45%; H NMR (400 MHz,
DMSO-d6):  3.71 (s, 4H), 3.87 (s, 3H), 4.20 (t, J = 10.0 Hz, 2H), 4.65 (t, J = 10.0 Hz, 2H), 5.07
(s, 1H), 6.95 (d, J = 9.2 Hz, 1H), 7.53 (t, J = 7.4 Hz, 1H), 7.65 (t, J = 7.6 Hz, 1H), 7.83 (d, J =
13
8.0 Hz, 1H), 8.47-8.51 (m, 2H); C NMR (100 MHz, DMSO-d6):  28.90, 45.40, 48.39, 48.60,
57.74, 100.68, 100.81, 110.96, 115.36, 123.97, 125.22, 126.64, 131.48, 141.79, 148.96, 151.43,
156.47, 156.61. ESI-MS/MS m/z 336 [M]+. Anal. Calcd. for C18H18BrN5O2 (%): C, 51.94; H,
4.36; N, 16.82. Found: C, 52.01; H, 4.29; N, 16.75.
3-(2-(Dimethylamino)ethyl)-12-methyl-6-oxo-2,3,6,12-tetrahydro-1H-benzo[4,5]imidazo[1,2-
1
a]imidazo[1',2':1,6]pyrido[2,3-d]pyrimidin-14-ium bromide (18). Yield: 49%; H NMR (400
MHz, DMSO-d6):  2.23 (s, 6H), 2.56 (t, J = 6.0 Hz, 2H), 3.74 (t, J = 6.0 Hz, 2H), 3.86 (s, 3H),
4.19 (t, J = 9.8 Hz, 2H), 4.64 (t, J = 10 Hz, 2H), 7.00 (d, J = 9.2 Hz, 1H), 7.52 (t, J = 7.2 Hz,
1H), 7.64 (t, J = 7.2 Hz, 1H), 7.82 (d, J = 8.0 Hz, 1H), 8.45-8.50 (m, 2H); 13C NMR (100 MHz,
DMSO-d6):  28.78, 43.44, 45.22, 48.26, 55.62, 100.14, 100.80, 110.84, 115.21, 123.86, 125.03,
126.53, 131.31, 141.85, 148.76, 151.34, 156.06, 156.28. ESI-MS/MS m/z 363 [M]+. Anal. Calcd.
for C20H23BrN6O (%): C, 54.18; H, 5.23; N, 18.96. Found: C, 54.32; H, 5.19; N, 18.81.
3-(2-(Diethylamino)ethyl)-12-methyl-6-oxo-2,3,6,12-tetrahydro-1H-benzo[4,5]imidazo[1,2-
1
a]imidazo[1',2':1,6]pyrido[2,3-d]pyrimidin-14-ium bromide (19). Yield: 50%; H NMR (400
MHz, DMSO-d6):  0.93 (t, J = 7.0 Hz, 6H), 2.53 (q, J = 7.2 Hz, 4H), 2.68 (t, J = 5.6 Hz, 2H),
3.70 (t, J = 5.8 Hz, 2H), 3.86 (s, 3H), 4.21 (t, J = 10 Hz, 2H), 4.66 (t, J = 9.8 Hz, 2H), 6.96 (d, J
11

= 9.2 Hz, 1H), 7.52 (t, J = 7.6 Hz, 1H), 7.64 (t, J = 7.8 Hz, 1H), 7.82 (d, J = 8.0 Hz, 1H), 8.45-
8.50 (m, 2H); 13C NMR (100 MHz, DMSO-d6):  11.79, 28.90, 44.12, 45.33, 46.30, 48.54,
49.65, 100.54, 100.85, 110.97, 115.34, 123.97, 125.17, 126.63, 131.44, 141.72, 148.91, 151.45,
156.21, 156.45. ESI-MS/MS m/z 391 [M]+. Anal. Calcd. for C22H27BrN6O (%):C, 56.05; H,
5.77; N, 17.83. Found: C, 56.21; H, 5.71; N, 17.68.
1-(2-Bromoethyl)-11-methyl-5-oxo-5,11-dihydrobenzo[4,5]imidazo[1,2-a]pyrido[2,3-
d]pyrimidin-1-ium bromide 22. Compound 21 (0.215 g, 0.86 mmol) was suspended in 1.5 mL
of 1,2-dibromoethane (17.0 mmol). The resulting mixture was heated at 90 °C for 40 h. During
this time an additional amount of 1,2-dibromoethane (0.2 ml) was slowly added. After cooling,
the suspension was filtered to give a yellow solid, which was purified by recrystallization from
1
ethanol. Yield: 65%; H NMR (400 MHz, DMSO-d6):  3.97 (s, 3H), 4.17 (t, J = 6.0 Hz, 2H),
5.25 (t, J = 6.0 Hz, 2H), 7.63 (t, J = 6.8 Hz, 1H), 7.72 (t, J = 7.0 Hz, 1H), 7.76-7.79 (m, 1H),
7.91 (d, J = 8.0 Hz, 1H), 8.49 (d, J = 8.0 Hz, 1H), 9.20 (d, J = 6.0 Hz, 1H), 9.26 (dd, J = 1.8 Hz,
J = 7.8 Hz, 1H).
2.5 Fluorescence. Fluorescence experiments were performed at 20 °C on an FP-8300
spectrofluorimeter (Jasco Inc., Tokyo, Japan) equipped with a Peltier temperature controller
accessory (Jasco PCT-818). A 1 cm path length, sealed quartz cuvette was used. Both excitation
and emission slits were set at 5 nm. The ligand solutions had a concentration in the 1.5-2.0 µM
range in phosphate buffer and were prepared by weighting using an analytical scale. The
titrations were carried out by stepwise additions of DNA (5 µL at 100 µM) to a cell containing a
fixed concentration (1.5-2.0 µM) of ligand. The solution was stirred and allowed to equilibrate
for 5 min after each addition. Emission spectra were recorded by exciting each compound at the
12

appropriate wavelength. The fraction of bound ligand (α) at each point of the titration was
calculated following fluorescence changes at the maximum of intensity:
퐼휆 − 퐼휆푚푎푥
훼 =
퐼휆푚푖푛 − 퐼휆푚푎푥
in which λ is the wavelength of the ligand maximum emission; 퐼휆 is the fluorescence intensity at
each point of titration; 퐼휆푚푎푥 is the fluorescence intensity of the ligand alone; and 퐼휆푚푖푛 is the
fluorescence intensity of the saturated sample.
Titration curves were obtained by plotting α versus the DNA G4 concentration. The
equilibrium binding constant (Kb) and the stoichiometry of the binding (n) were estimated by
fitting the resulting curve to an independent and equivalent binding site model (Cummaro et al.,
2011):
퐿0 + 푛푄0 + 1 퐾푏 2 − 4퐿0푛푄0]
⁄ )
( ⁄
) (
⁄
)
√(
훼 = 1 2 퐿0 [ 퐿0 + 푛푄0 + 1 퐾푏
−
in which α is the bound ligand fraction, 퐿0 is the total ligand concentration, and 푄0 is the added
G4 concentration. Appropriate excitation wavelength was use for each ligand. Experiments were
repeated twice and the results are reported as the mean ± S.D..
2.6 Native polyacrylamide gel electrophoresis (PAGE) analysis. Native gel electrophoreses
were carried out in 15% (w/v) polyacrylamide gel. An 89 mM Tris HCl, 89 mM boric acid 10×
TB solution at pH 7.0 was used as a run buffer, APS as a radical initiator, TEMED as a radical
stabilizer and bromophenol blue as a run marker. Stock ligand solutions concentration was 10
mM in DMSO. Two native gel electrophoreses were carried out in excess of either the ligand or
the oligonucleotide. In the first case, an oligonucleotide concentration of 50 μM for each sample
and an increasing amount (1, 2, 10 mol equiv) of ligand were used. In the second case, an
oligonucleotide concentration of 50 μM for each sample and an increasing amount (0.01, 0.02,
13

0.1 mol equiv) of ligand were used. The total volume loaded in each well was 10 μL. Gels were
imaged by UV-shadowing, using 254 nm exposure wavelength (Amato et al., 2014a).
2.7 Molecular Dynamics. Docking simulations on compound 19 were performed applying the
same protocol followed in VS calculations. Prior to docking, the ligand tridimensional structure
was generated with the Maestro Build Panel, and its tautomeric and protonation states at
physiological pH (7.4 ± 1.5) were then predicted by Epik (Greenwood et al., 2010; Shelley et al.,
2007). The docking predicted 19/KRAS G4 complexes were solvated in a 12.0 Å layer cubic
water box using the TIP3P water model parameters (Jorgensen et al., 1983). 19 K+ cations were
used to neutralize each system, with two of these ions placed at the center of the G-tetrads.
Further 2 K+ and 2 Cl− ions were added to reach the standard 150 mM KCl concentration. The
parmbsc1 (Ivani et al., 2015) and gaff (Wang et al., 2004) Amber force fields were used to
parameterize the nucleic acid and ligand, respectively. Amber charges were applied to the DNA
and water molecules, whereas ligand charges were computed using the restrained electrostatic
potential (RESP) fitting procedure (Bayly et al., 1993). The ESP was first calculated by means of
the Gaussian package (Frisch et al., 2009) using a 6–31G* basis set at B3LYP level of theory,
and then the RESP charges (Wang et al., 2000) were obtained by a two-stages fitting procedure
using Antechamber (Wang et al., 2006). The NAMD 2.13 (Phillips et al., 2005) code was used to
perform the simulations. A cutoff of 10 Å was used for short-range interactions. The long-range
electrostatic interactions were computed by means of the particle mesh Ewald method (Darden et
al., 1993) using a 1.0 Å grid spacing in periodic boundary conditions. The SHAKE algorithm
(Ryckaert et al., 1977) was applied to constraint bonds involving hydrogen atoms and the non-
iterative SETTLE method (Miyamoto and Kollman, 1992) to keep waters rigid, which allowed a
14

2 fs integration time step. A conjugate gradient energy minimization of 10,000 steps was
performed on each system in two stages. In the first step, the system heavy atoms were kept
fixed in order to relax only the hydrogens and the water molecules, while during the second step
all the atomic positions were minimized. Then, each complex was heated up to 300 K using the
Langevin thermostat (Grest and Kremer, 1986) in three stages of 250 ps each, and then
pressurized to 1 atm for 250 ps using the Langevin piston Nose-Hoover method implemented in
NAMD (Feller et al., 1995; Martyna et al., 1994). During the equilibration process harmonic
constraints were put on both the ligand and DNA heavy atoms, which were gradually released
along the process. The production runs were performed in the NPT ensemble at 1 atm and 300 K.
The stability of each docking-predicted pose of 19 along MD simulations was evaluated through
RMSD calculations on the heavy atoms of the ligand polycyclic scaffold. The effects of the
ligand binding on the overall architecture of KRAS G4 were measured through the RMSD plots
of the DNA backbone heavy atoms. Prior to each RMSD calculation, trajectory frames were
aligned on the DNA guanine residues heavy atoms.
2.8 Cell lines, culture conditions and treatments. HCT116 colorectal cancer cells were
obtained by Dr Vogelstein, Johns Hopkins University. HK2–6 and HKE-3 were kindly provided
by Dr. Shirasawa and performed by gene targeting technique (Shirasawa et al., 1993). BJ-EHLT
derived from the transformation of BJ human fibroblasts with hTERT and SV40 early region. All
the cells were grown in high glucose Dulbecco modified eagle medium (DMEM; Euroclone)
supplemented with L-glutamine, penicillin/streptomycin and 10% fetal bovine serum (FBS,
Hyclone). The G-quadruplex ligand EMICORON (Franceschin et al., 2012; Zizza et al., 2019)
was used at 1 M for 24 h.
15

2.9 Real-time PCR. To quantify gene expression by real-time quantitative polymerase chain
reaction (qRT-PCR), total RNA was isolated from cell pellets by using TRIzol reagent
(Ambion). The quality of the extracted RNA was assessed by 1% agarose gel electrophoresis and
from the A260/A280 absorbance ratio (Nanodrop 1000, ThermoFisher Scientific). Reverse
transcription (RT) from 500 ng of RNA was performed using the QuantiTect Reverse
Transcription Kit (Qiagen) according to the manufacturer’s instructions. Real-time qRT-PCRs
were performed on the obtained cDNAs by using Fast power SYBR green master mix (Apply
Biosystem) on 7900HT Fast Real-Time PCR System (Applied Biosystem) thermocycler. For
each sample 5 µl of the 1:10 diluted cDNA was mixed with 0.5 µl of each primer (10 µM), 10 µl
of the SYBR green master mix and water at final volume of 20 µl. Standard qPCR thermal
parameters were used: one cycle of 95 °C for 10 min then 40 cycles at 95 °C for 15 sec and 60
°C for 1 min followed by dissociation curve (95 °C for 15 sec, 60 °C for 1 min, 95 °C for 15
sec).
Primer pairs (KRAS: FW 5'-ACACAAAACAGGCTCAGGACT-3'; REV: 5'-
TGTCGGATCTCCCTCACCAA-3' and -actin: FW 5'-AGCACTGTGTTGGCGTACAG-3';
REV: 5'-TCCCTGGAGAAGAGCTACGA-3') were synthesized by Integrated DNA
Technologies (BVBA Leuven, Belgium).
All experiments were run in triplicate and the gene expression levels were normalized to the
-actin.
2.10 Western Blotting. For western blot analysis total cell lysates were prepared as previously
reported (Biroccio et al., 2003). The proteins were fractionated by SDS-polyacrylamide gel
electrophoresis and transferred to nitrocellulose membrane (Amersham, Arlington Heights,
16

USA). Membranes were probed with primary antibodies and the signal was detected using
peroxidase-conjugated anti-mouse or anti-rabbit secondary antibodies (Jackson Immunoresearch
Labs, Inc., Baltimore, USA). The enhanced chemi-luminescence (ECL) system (Thermo
Scientific) was used for detection. The following primary antibodies were used: rabbit mAb anti-
RAS (D2C1, Cell Signaling Technology Beverly, MA, USA) (Naidu et al., 2017), mouse mAb
anti-β-actin (clone AC-15, Sigma, St. Louis, USA) and mouse mAb anti-γH2AX (Millipore,
Billerica, MA, USA).
2.11 Clonogenic assay. HCT116 cells, were seeded in 35 mm-Petri dishes at the clonogenic
density of 300 cells/plate in DMEM medium with 10% FBS and, after 24 h, cells were treated
with DMSO (negative control) or the indicated dose of 19. After 10 days, the cells were stained
with 2% methylene blue in 50% ethanol and the number of colonies was counted. Surviving
fractions were calculated as the ratio of absolute survival of the treated sample/absolute survival
of the untreated sample.
2.12 Viability Assay (Crystal Violet). HK2-6 and HKE-3 cells were seeded in a 24-well plate at
a density of 7x104 for well. After 24 h, the cells were treated with compound 19 at different
doses for 72 h. Then cells were washed twice in PBS and fixed with 4% formaldehyde for 15
min at RT. After washing, 500 μl of crystal violet staining solution (Sigma) was added to each
well, and incubated for 30 min at RT. Finally, the plates were rinsed twice with water, air-dried
at RT, and the cell pellets were dissolved in 400 μl of acetic acid. One hundred l of each sample
were transferred to a 96-well plate and the optical density of each well was measured at 570 nm
(OD570) with an ELISA reader (Thermo Scientific). The average absorbance in each condition
was used to calculate the survival expressed as percent of treated vs untreated condition. IC50
(the dose necessary to reduce the survival of 50%) was calculated by Calcusyn software.
17

2.13 Immunofluorescence Microscopy. BJ-EHLT cells were fixed in 2% formaldehyde in
phosphate buffered saline (PBS) for 10 min at room temperature and permeabilized in 0.25%
Triton X-100 in PBS for 5 min at RT. For immune-labeling, cells were incubated with the mouse
anti-γH2AX (Millipore, Billerica, MA, USA) primary antibody 2h at RT, washed twice in PBS
and finally incubated with the anti-mouse IgG (H+L), F(ab’)2 Fragment (Alexa Fluor 555
Conjugate) secondary antibodies for 1 h. Nuclei were stained with 4’,6-diamidino-2-
phenylindole (DAPI, Sigma). Fluorescence signals were recorded by using a Leica DMIRE2
microscope equipped with a Leica DFC 350FX camera and elaborated by Leica FW4000
deconvolution software (Leica, Solms, Germany). For quantitative analysis of γH2AX positivity,
at least 250 cells on triplicate slices were scored.
3. Results and Discussion
3.1 Virtual Screening. As briefly reported in the introduction, the solved KRAS G4 NMR
structure (PDB code: 5I2V) (Kerkour et al., 2017) was here used as target for a structure-based
virtual screening (VS) campaign aimed at identifying brand new KRAS G4 stabilizers. Indeed,
this G4 motif possesses two wide stacking surfaces and four grooves of medium size that can
potentially be recognized by small organic molecules (Kerkour et al., 2017). For VS calculations,
we docked molecules from both the commercial Asinex Platinum Collection library and an in-
house database. Notably, the latter library is composed of all the compounds designed,
synthesized and tested so far by our research group against common anticancer targets including
kinases, integrins, chemokine receptors and nucleic acids. Advantageously, these compounds are
either already available in stock or can be easily resynthesized. Moreover, for many molecules, a
number of analogues are already present allowing a rapid structure-activity relationship analysis.
18

Both the selected libraries were prepared as described in Materials and Methods. Then, since the
use of appropriate molecular filtering procedures have largely proved to improve the VS success
rate (Da Costa et al., 2019; La Pietra et al., 2018; Pinto et al., 2019), all the neutral and
negatively charged compounds as well as all the molecules featuring less than two aromatic rings
were discarded from both the datasets. In fact, positively charged moieties and extended
aromatic systems are generally required to establish favorable contacts with the DNA phosphate
backbone and with the G-tetrads stacking surfaces or possibly the G4 grooves, respectively
(Castillo-González et al., 2015; Di Leva et al., 2014, 2013). The final subsets from both the
libraries were then submitted to VS calculations.
NH
N
N
N
O
N
N
O
N
N
N
O
N
N
O
N
OH
N
HN
N
1
2
3
N
HN
O
N
N
N
O
N
N
N
N
HN
N
I-
N
N
O
4
5
6
H
N
N
S
O
N
O
N
N
N
H
N
N
N
N
H
N
N
N
N
7
8
9
N
N
O
N
S
O
N
OH
NH
O
N
N
N
N
F
N
N
N
N
N
N
10
11
12
Chart 1. Chemical structures of VS hits 1-4 from the in-house library and 5-12 from the Asinex
Platinum Collection.
19

The top 15% of the ranked solutions from each database was selected to obtain a subset of
molecules, all showing a docking score < -7.0, feasible for the following visual examination. In
fact, these compounds were subjected to a careful visual inspection in order to evaluate their
predicted interaction mode with KRAS G4. In particular, the formation of π-stacking or polar
interactions (i.e. salt bridges) with the target DNA was investigated. Finally, the selected
compounds were further inspected for good chemical binding geometry. Based on these criteria
we chose 12 compounds (Chart 1 and Table S1): 4 were picked from the in-house database (1-4),
while 8 were taken from the Asinex Platinum Collection library and then purchased from the
vendor (5-12). Interestingly, the predicted binding modes of the best ranked compounds from
each library, namely 1 and 9, suggested that the chosen molecules could in principle occupy
either the G4 stacking surface (Fig. S3A) or the DNA grooves (Fig. S3B). All the compounds
were then tested in biophysical assays to verify their G4 binding properties.
3.2 CD Experiments of the selected hits (1-12) with KRAS G4. Initially, the conformation of
KRAS G4 was proved by circular dichroism (CD) spectroscopy. The CD spectrum displayed a
positive band at 264 nm and a negative band at 240 nm, which is the characteristic CD profile of
parallel-stranded quadruplexes (Zhang et al., 2010). Notably, no CD spectra change has to be
reported for KRAS G4 upon ligand binding (Fig. S4). The stabilizing effect of each of the
selected hits (1-12) on KRAS G4 was then measured in terms of ΔTm by CD thermal melting
experiments at 264 nm (Fig. 1 and Fig. S5). Compounds 1 and 4 proved the best, albeit not fully
satisfactory, stabilizing effect on KRAS G4 (ΔTm = +4.5 °C and ΔTm = +7.5 °C, respectively; Fig.
2). Thus, the selectivity of these compounds against the 20-mer hairpin DNA was then assessed
through CD thermal melting experiments at 280 nm (Fig. S6). These assays showed no
20

stabilizing effects for 1 and 4 on the duplex DNA, thus prompting their selection as hit
compounds for chemical optimization.
Fig. 1. ΔTm of KRAS G4 upon interaction with VS hits.
Fig. 2. Normalized CD thermal denaturation of KRAS G4 in the absence and in the presence of 1
(left panel) and 4 (right panel).
3.3 Hit Optimization. In our lead optimization program, we intended to investigate the influence
of side chains of different length and nature on the binding of 1 and 4 to the target DNA as well
as on their G4/duplex selectivity. In this perspective, we first selected from our in-house database
a set of four analogues of 1 (compounds 13-16, Chart 2), featuring a terminal amino group in
21

place of the hydroxyl function. In fact, as briefly reported in the introduction, our in-house
library offers the advantage that many compounds already possess a number of analogues, so a
rapid SAR analysis is possible.
Chart 2. Chemical structures derivatives 13-19.
Scheme 1. Synthesis of compounds 13-16. Reagents and conditions: (i) K2CO3, CuBr, KI, DMF,
150 °C, P = 10 bar, power = 40-150 W, ramp t = 2 min, t = 12 min.
However, the synthesis of 13-16 (Dalla Via et al., 2001) has been here improved using
microwaves (MW), with higher yields and shorter reaction time (12 min versus 5-7 h, Scheme
22

1). In particular, a MW-assisted Ullmann condensation reaction between the appropriate 1-
dialkylaminoalkyl-2-aminobenzimidazole 20a-d (Dalla Via et al., 2001) and the commercially
available 2-chlorobenzoic acid (Scheme 1) was performed. After cooling, the reaction mixture
was poured into ice and the precipitated crude 6-(aminoalkyl)benzo[4,5]imidazo[2,1-
b]quinazolin-12(6H)-ones 13-16 were collected and purified by recrystallization from ethanol.
In the case of 4, we decided to introduce differently functionalized protonable pendant side
chains at the positively charged pyridine nitrogen since positively charged moieties are
recognized to establish favourable contacts with the phosphate backbone of DNA (Di Leva et al.,
2014, 2013). In this vein, the synthetic procedure previously developed by us for the preparation
of 4 (Caroti et al., 1987) was applied starting from the 11-methylbenzo[4,5]imidazo[1,2-
a]pyrido[2,3-d]pyrimidin-5(11H)-one 21 (Caroti et al., 1987), with few modifications (Scheme
2).
Scheme 2. Synthesis of compounds 17-19. Reagents and conditions: (i) 1,2-dibromoethane, 90
°C; (ii) H2N(CH2)2R, ethanol, reflux.
Reaction of compound 21 with an excess of 1,2-dibromoethane at 90 °C for 40 h furnished
22, which was suspended in ethanol and added with an excess of the appropriate amine. The
reaction mixture was refluxed for 4 h and then filtered to give crude products, finally purified by
recrystallization from ethanol. A deep analysis of the spectral data allowed us to unequivocally
23

assign to the obtained compounds 17-19 the structures outlined in Scheme 2, featuring a fused
pentacyclic system. Actually, this is in agreement with literature data (Parenty et al., 2007, 2005)
that report the reaction between a primary amine and a 2-bromoethylpyridinium derivative to
proceed via a nucleophilic attack, followed by a cyclisation and an oxidation step, leading to a
dihydroimidazopyridinium compound. Although the obtained derivatives 17-19 possessed a
rather unexpected structure, they were considered for biological evaluation, as their structural
features still met the requirements to act as G4 stabilizing molecules, that is a positively charged
pendant chain and an extended aromatic system.
3.4 Biophysical assays on 13-19. Compounds 13-19 were tested on KRAS G4 and duplex
hairpin DNA, evaluating ΔTm by CD thermal melting experiments (Fig. 3, 4 and S7-S10). In the
subset of derivatives 13-16, all the compounds except for 13 showed improved KRAS G4
stabilizing properties compared to the hit 1 and maintained satisfactory selectivity over duplex
hairpin (Fig. 3).
Fig. 3. Ligand-induced thermal stabilization of KRAS G4 and hairpin duplex measured by CD
melting experiments.
24

As shown in Fig. 3, the most interesting results were observed for the pentacyclic derivatives
17-19, with all the three compounds recording substantial performance improvement (ΔTm > 10
°C) with respect to the parent hit 4 (Fig. 1, ΔTm = 7.5 °C). Also, these compounds showed high
KRAS G4/duplex DNA selectivity ratio (Fig. 3). In view of these data, 17-19 were further tested
against the parallel KIT G4 and hybrid-1 Tel23 (Fig. S11 and S12).
Table 1. Comparison of 17, 18 and 19 ΔTm with KRAS G4 and control sequences.
KRAS G4
KIT G4
Tel23
ligand
ΔTm ± 0.5 °C ΔTm ± 0.5 °C ΔTm ± 0.5 °C
+ 13.5
+ 15.0
+ 18.0
+ 7.0
+ 10.0
+ 10.5
+ 2.5
+ 5.5
+ 7.5
17
18
19
These ligands showed preferential binding to the parallel G4s (Table 1). Indeed, all of them
turned out to increase the KRAS G4 thermal stability to a larger extent than the other G4s. In
particular, 19 was the most potent KRAS G4 stabilizing agent of the series and was thereby
selected for further analysis (Fig. 4).
Fig 4. CD spectra (left panel) and thermal melting profiles (right panel) of KRAS G4 without
(black line) and with (red line) 19.
25

Fluorescence titration experiments were thus performed for gaining information on the
binding affinity and stoichiometry of the interaction between KRAS G4 and 19 (Fig. 5) (Giancola
and Pagano, 2013; Pagano et al., 2012). Results of fluorescence analysis showed a stoichiometry
ratio of 1:1 KRAS G4/19 and a binding constant (Kb) of 7.2 ± 0.4 106 M-1. In order to strengthen
the CD results, fluorescence experiments were also performed with KIT G4 (Fig. S13), revealing
an interaction of 1:1 KIT G4/19 with Kb = 2.3 ± 0.5 106 M-1, which is threefold lower compared
to that of KRAS G4.
Fig. 5. Fluorescence titration of 19 with KRAS G4.
To investigate the behavior of the KRAS G4/19 complex upon addition of different amounts
of 19, a non-denaturing gel electrophoresis (PAGE) experiment was performed (Fig. S14). In
fact, KRAS G4 moves as a single band in the gel, thus suggesting the absence of high-order DNA
structures. The addition of 19 to the KRAS G4 did not affect the G4 mobility at all investigated
ratios, in agreement with the stoichiometry revealed by the fluorescence assay.
3.5 Computational Studies. In order to investigate at an atomic level the binding mode of the
newly synthesized derivatives to KRAS G4, molecular modeling studies were carried out on the
most promising compound of the series, 19. Initially, molecular docking of this molecule was
26

performed on the whole KRAS G4 NMR structure (PDB code: 5I2V) (Kerkour et al., 2017).
Docking calculations predicted two similar binding modes, A and B (Fig. S15), in which the
ligand stacks at the 3’ region of the target DNA to interact with residues G9, G13, and A22.
However, the two poses differ in the arrangement of the ligand planar scaffold which is mutually
flipped by 180°. The stability and the energetics of these poses were then investigated through
more accurate calculations. In particular, to fully take into account the receptor flexibility and the
water and ions effects, each of the two docking-predicted 19/DNA complexes was submitted to
1.5s molecular dynamics (MD) simulations in explicit solvent.
A comparison between the ligand RMSD plot computed for the two MD trajectories showed that
only one of the two docking solutions (pose A) evolves in a stable binding conformation over the
simulated time scale (Fig. S16). In fact, starting from this pose, in the first half of the MD run the
ligand experiences a slight rearrangement to reach a well-defined binding mode, which is
conserved throughout the rest of the simulation. In fact, this pose is stabilized by tight
interactions with the DNA. Specifically, 19 lies at the G4 3’ end where it engages favorable π-
stacking with G9, G13 and G20 (Fig. 10). Furthermore, the N,N-(diethyl)aminoethyl group can
alternatively establish a salt bridge with the DNA phosphate backbone of either A14 (Fig. 6) or
A22 (Fig. 6C and S17). At this regard, we report that in MD calculations on pose B this kind of
ligand-DNA interaction is conversely much less conserved along the whole simulation time (Fig.
S18). It is also interesting to note that 19 can form additional stacking interactions with the
terminal A22, which shifts from its initial position to be packed against the ligand aromatic
scaffold. Remarkably, these contacts stabilize not only the ligand binding conformation but also
the overall KRAS G4 architecture. This effect can be appreciated by looking at Fig. S19 where
the RMSD plot of the G4 backbone along the MD trajectory is shown.
27

Fig. 6. (A) Front (left panel) and bottom (right panel) view of the MD predicted binding pose of
19 (yellow sticks) at the NMR structure of the KRAS G4 (PDB code: 5I2V36). DNA is shown as
cyan cartoons and transparent surface. Nucleotides are highlighted as sticks; aromatic rings are
filled with thin slabs. The color code for the heteroatoms is: blue for nitrogen, red for oxygen. K+
ions are depicted as purple spheres. Nonpolar hydrogens are omitted for clarity. Hydrogen bonds
28