More is not always better: Finding the right trade-off between affinity and selectivity of a G-quadruplex ligand

Article abstract of DOI:10.1093/nar/gky607

Guanine-rich nucleic acid sequences can fold into four-stranded G-quadruplex (G4) structures. Despite growing evidence for their biological significance, considerable work still needs to be done to detail their cellular occurrence and functions. Herein, we describe an optimized core-extended naphthalene diimide (cex-NDI) to be exploited as a G4 light-up sensor. The sensing mechanism relies on the shift of the aggregate-monomer equilibrium towards the bright monomeric state upon G4 binding. In contrast with the majority of other ligands, this novel cex- NDI is able to discriminate among G4s with different topologies, with a remarkable fluorescent response for the parallel ones. We investigate this sensing by means of biophysical methods, comparing the lead compound to a non-selective analogue. We demonstrate that mitigating the affinity of the binding core for G4s results in an increased selectivity and sensitivity of the fluorescent response. This is achieved by replacing positively charged substituents with diethylene glycol (DEG) side chains. Remarkably, the limit of detection values obtained for parallel G4s are more than one order of magnitude lower than those of the parallel-selective ligand N-methyl mesoporphyrin IX (NMM). Interestingly, the classical fluorescent intercalator displacement (FID) assay failed to reveal binding of cex-NDI to G4 because of the presence a ternary complex (G4-TO-cex-NDI) revealed by electrospray-MS. Our study thus provides a rational basis to design or modify existent scaffolds to redirect the binding preference of G4 ligands.

Full text of DOI:10.1093/nar/gky607

Published online 9 July 2018
Nucleic Acids Research, 2018, Vol. 46, No. 19 e115
doi: 10.1093/nar/gky607
More is not always better: finding the right trade-off
between affinity and selectivity of a G-quadruplex
ligand
Michela Zuffo1, Aurore Gue´din2, Emma-Dune Leriche2, Filippo Doria1, Valentina Pirota1,
1,*
Vale´rie Gabelica2, Jean-Louis Mergny2,3,* and Mauro Freccero
1Dipartimento di Chimica, Universita` di Pavia, Pavia 27100, Italy, 2ARNA Laboratory, Universite´ de Bordeaux, Inserm
U1212, CNRS UMR5320, Institut Europe´en de Chimie Biologie (IECB), Pessac 33607, France and 3Institute of
Biophysics, AS CR, Brno 61265, Czech Republic
Received April 14, 2018; Revised May 26, 2018; Editorial Decision June 22, 2018; Accepted June 25, 2018
ABSTRACT
INTRODUCTION
The last decades were characterized by extensive research
on nucleic acids (NA) folded into G-quadruplex (G4) sec-
ondary structures (1). These are formed by guanine-rich
DNA or RNA single strands, which can assemble into quar-
tets via Hoogsteen hydrogen bonds (2). Such quartets then
stack on top of each other constituting the core of the G4
structure. This is further stabilized by monovalent cations
sitting in the inner channel (3,4). Despite this common
backbone, G4s are highly polymorphic as for molecular-
ity (mono, bi or tetramolecular), topology (parallel, anti-
parallel or hybrid), glycosidic bond angle (syn or anti),
sugar conformation (ring puckering), loops and grooves
size (5,6). Reasons for this widespread interest in G4s re-
side in mounting evidence of their regulatory roles in key
biological processes, such as telomere maintenance, repli-
cation, transcription and translation (7–10). Interestingly,
many putative G4 sequences (PQS) are located in close
proximity to genes involved in a wide range of pathologies.
These include cancer (11–13), neurodegenerative diseases
(14,15), viral (16,17) and parasitic (18,19) infections. There-
fore, G4s hold great potential as novel therapeutic and ther-
anostic targets (20,21).
In order to validate them as such, however, their forma-
tion and functions must be proven in cells. In recent years,
monoclonal antibodies were engineered to signal G4 for-
mation in cellular models (22–24). However, imaging biases
arising from cell fixation and chaperone activity of such pro-
teins cannot be excluded. Among the investigation tools,
molecular probes undergoing a fluorescence light-up pro-
vide a complementary method to address the issue (25,26).
A number of fluorescent small molecules were already re-
ported to specifically emit upon G4 binding (27–30). Struc-
tures of such sensors usually mimic the scaffolds of pow-
erful G4 ligands, to ensure high affinity (31–34). Unfortu-
Guanine-rich nucleic acid sequences can fold into
four-stranded G-quadruplex (G4) structures. Despite
growing evidence for their biological significance,
considerable work still needs to be done to detail
their cellular occurrence and functions. Herein, we
describe an optimized core-extended naphthalene
diimide (cex-NDI) to be exploited as a G4 light-up
sensor. The sensing mechanism relies on the shift
of the aggregate-monomer equilibrium towards the
bright monomeric state upon G4 binding. In contrast
with the majority of other ligands, this novel cex-
NDI is able to discriminate among G4s with different
topologies, with a remarkable fluorescent response
for the parallel ones. We investigate this sensing by
means of biophysical methods, comparing the lead
compound to a non-selective analogue. We demon-
strate that mitigating the affinity of the binding core
for G4s results in an increased selectivity and sen-
sitivity of the fluorescent response. This is achieved
by replacing positively charged substituents with di-
ethylene glycol (DEG) side chains. Remarkably, the
limit of detection values obtained for parallel G4s
are more than one order of magnitude lower than
those of the parallel-selective ligand N-methyl meso-
porphyrin IX (NMM). Interestingly, the classical fluo-
rescent intercalator displacement (FID) assay failed
to reveal binding of cex-NDI to G4 because of the pres-
ence a ternary complex (G4-TO-cex-NDI) revealed by
electrospray-MS. Our study thus provides a rational
basis to design or modify existent scaffolds to redi-
rect the binding preference of G4 ligands.
*To whom correspondence should be addressed. Tel: +390382987668; Email: mauro.freccero@unipv.it
Correspondence may also be addressed to Jean-Louis Mergny. Tel: +33540003022; Email: jean-louis.mergny@inserm.fr
Present address: Michela Zuffo, Institut Curie, CNRS UMR9187, INSERM U1196, PSL Research University, Orsay 91400, France.
ꢀ
C
The Author(s) 2018. Published by Oxford University Press on behalf of Nucleic Acids Research.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License
(http://creativecommons.org/licenses/by-nc/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work
is properly cited. For commercial re-use, please contact journals.permissions@oup.com

e115 Nucleic Acids Research, 2018, Vol. 46, No. 19
PAGE 2 OF 13
some of the affinity, but at the same time we would let the
core selectivity emerge. We demonstrate this hypothesis by
testing two cex-NDIs with different substituents (Scheme 1),
one displaying positively charged dimethyl amine groups (1)
and the other neutral DEG pendals (2).
The newly employed functionalization pattern is also
aimed at improving the sensing mechanism. In particu-
lar, we introduce an additional hydrophilic and electron-
donating side chain at position 4 in contrast with previously
investigated cex-NDIs (3–5; Scheme 1). On the one hand,
this change should produce a significant red-shift in both
absorption and emission, as previously observed for NDIs
(42). On the other hand, the positioning of the hydrophilic
substituents away from the lipophilic core-extension should
enhance the aggregation propensity of the core. Moreover,
DEG functionalization should likely favour the aggrega-
tion. It would in fact maintain charge neutrality at physi-
ological pH, while still ensuring sufficient water solubility.
MATERIALS AND METHODS
General materials and methods
Reagents, solvents and chemicals were purchased from Alfa
Aesar or Sigma-Aldrich and were used as supplied without
further purification. For in vitro studies 5 × 10−3 M stock
solutions of cex-NDI 1 and 2 were prepared in MilliQ wa-
ter and DMSO respectively and were stored at −20◦C. The
DNA and RNA oligonucleotides (Supplementary Tables
S1–S3) were purchased from Sigma Aldrich or Eurogentec
and dissolved in MilliQ water to prepare 5 × 10−4 M stock
solutions. The exact concentrations were then determined
by absorption measurements at 260 nm. The solutions were
stored at −20◦C. The NAs were annealed by incubation at
95◦C for 5 min, in the presence of the relevant amounts of
Scheme 1. 4-Substituted cex-NDIs (1 and 2) investigated in the present
study and unsubstituted cex-NDI (3) or 11-substituted cex-NDIs (4 and
5), previously investigated as G4 sensors (39,40).
nately, most ligands are unable to discriminate among dif-
ferent topological classes of G4s, not to mention recognize a
specific sequence. Specific recognition is nonetheless essen-
tial to thoroughly understand G4-related processes. Iden-
tifying probes able to recognize specific structural motifs
is thus of primary importance and relevant efforts have al-
ready been made in this direction (29,35–38).
salt (see specific protocols for details) and of 1 × 10−2
M
buffer (pH 7.2, either lithium cacodylate or Tris-HCl). Solu-
tions were then let to equilibrate for at least 2 hours at room
temperature. For the fluorophore-labelled oligonucleotides,
the heating time was reduced to 2 min and the equilibration
was performed in ice for 30 min. The folded samples were
then stored at 0◦C.
Herein, we exploit a fluorescence light-up mechanism
based on core-extended naphthalene diimides (cex-NDIs).
cex-NDIs have high G4 affinity and form non-emitting ag-
gregates under physiological conditions (39,40). Their fluo-
rescence can be restored upon monomerization induced by
interaction with the G4. Despite proving the sensing mech-
anism both in vitro and in fixed cells, the prototype cex-
NDIs displayed sub-optimal properties: (i) the aggregation
of those compounds is only partial at ␮M concentration,
leading to non-negligible background fluorescence and thus
modest sensitivity; (ii) the compounds excitation and emis-
sion bands are both placed at the lower edge of Red-NIR
(near infrared) spectroscopic window; (iii) finally, little or
no selectivity for a peculiar G4 topology was observed.
Herein, we present novel cex-NDI probes designed to
overcome these issues. Concerning selectivity among differ-
ent G4s, we propose a new strategy, counterintuitively based
on the reduction of the overall ligand affinity for G4s. In
fact, the most potent ligands produce a high and possibly
saturated binding response with all structures, regardless of
their identity and conformation. We thus propose to remove
the less specific interactions resulting from the electrostatic
attraction between the phosphate backbone and the proto-
nated amine groups (41). In this way, we would indeed lose
TLC analysis was carried out on silica gel (Merck 60F-
254) with visualization at 254 and 366 nm. Reverse phase
HPLC analysis was performed using an Agilent system SE-
RIES 1260. The column was XSelectHSS C18 (2.5 Mm) (50
× 4.6 mm) (Waters). Flow was 1.4 ml/min. The following
method (Method 1) was used: Aqueous solvent: 0.1% triflu-
oroacetic acid in water; Organic solvent: Acetonitrile; Gra-
dient: 95% aqueous, gradually to 40% aqueous over 8 min,
then isocratic flow for 4 min. Cex-NDI 2 was purified by
reverse phase HPLC, specifically an Agilent Technologies
1260 Infinity preparative HPLC provided with a diode ar-
ray UV–vis detector. The preparative column was XSelect
CSH Prep Phenyl-Hexyl 5 ␮m (150 × 30 mm) (Waters). The
flow was 30 ml/min. Purifications were performed through
three different methods, (Method 2) Aqueous solvent: 0.1%
trifluoroacetic acid in water; Organic solvent: Acetonitrile;
Gradient: Isocratic flow over 2 min at 80% of aqueous sol-
vent, gradually to 60% aqueous over 14 min, then isocratic
flow for 2 min (␭ detection: 254, 360 and 520 nm); (Method

PAGE 3 OF 13
Nucleic Acids Research, 2018, Vol. 46, No. 19 e115
3) Aqueous solvent: 0.1% trifluoroacetic acid in water; Or-
ganic solvent: Acetonitrile; Gradient: Isocratic flow over 2
min at 80% of aqueous solvent, gradually to 60% aqueous
over 8 min, then isocratic flow for 4 min (␭ detection: 590,
254 and 520 nm); (Method 4) Aqueous solvent: 0.1% triflu-
oroacetic acid in water; Organic solvent: Acetonitrile; Gra-
dient: Isocratic flow over 2 min at 95% of aqueous solvent,
gradually to 60% aqueous over 13 min, then isocratic flow
6 (0.25 mmol) and 2-(2-aminoethoxy)ethanol (1 mmol) in 3
mL of acetic acid at the microwave in a closed vessel (12 min,
110◦C, 250 psi). The desired compound 8 was obtained in a
mixture with compounds 9 and 10, produced by partial or
total dehalogenation of the starting material and product 8.
The mixture was used as such for the following steps, veri-
fying its composition by TLC (CHCl3 95: MeOH 5). Acetic
acid was stripped away and the residues were removed by fil-
tration on silica (pure chloroform, then 2% MeOH in chlo-
roform). 100 mg of the mixture were then suspended in 3
mL of acetonitrile and 1,2-phenylendiamine (1.25 mmol)
was added. The suspension was heated at 110◦C for 20
min (MW, closed vessel, 250 psi). The reaction progress
was checked at the analytical HPLC (Method 1). The de-
sired product 12 was obtained together with by-product
13, which in turn results from the reaction of the previ-
ous by-product 8 with 1,2-phenylendiamine. Compounds
12 and 13 exclusively precipitated from the reaction mix-
ture (purple solid). The products remaining in fine suspen-
sion were recovered by filtration. Again, compound 12 was
not purified, since by-product 13 did not affect the follow-
ing step. Finally, the mixture was submitted to nucleophilic
aromatic substitution. It was dissolved in 3 mL of 1:1 so-
lution of DMA and 2-(2-aminoethoxy)ethanol and heated
at the MW in a closed vessel (10 min, 150◦C, 250 psi).
The reaction progress was evaluated by analytical HPLC
(Method 1). The mixture was diluted in acidic water (0.1%
TFA) and purified at the preparative HPLC (Method 4).
Possible residues of compound 13 were removed by subse-
quent column chromatography purification (pure CHCl3,
then gradually to 5% MeOH), reaching an overall yield of
7%. All the intermediates (8, 9, 10, 12 and 13) were puri-
fied by preparative HPLC to perform a full characteriza-
tion. The reported yields correspond to these purification
steps. 8: purified by preparative HPLC (Method 2). Yellow
solid. Yield = 42%. 1H NMR (300 MHz, DMSO-d6), ␦ =
8.69 (s, 1H); 8.64 (s, 1H); 4.26 (m, 4H); 3.71-3.65 (m, 12 H).
13C NMR (75 MHz, DMSO-d6), ␦ = 170.2; 162.5; 160.7;
160.5; 137.1; 130.5; 127.5; 126.5; 126.2; 126.1; 126.0; 125.6,
124.0; 72.1; 68.0; 66.7; 66.7; 66.5; 63.0; 60.2. 9: purified by
preparative HPLC (Method 2). Yellow solid. Yield = 23%.
1H NMR (300 MHz, DMSO-d6), ␦ = 8.63 (s, 4H); 4.24 (m,
4H); 3.74–3.62 (m, 12 H). 13C NMR (75 MHz, DMSO-d6),
␦= 160.8; 160.7; 131.1; 130.5; 126.5; 125.6, 120.0; 70.0; 65.3;
61.3. 10: purified by preparative HPLC (Method 2). Yellow
solid. Yield = 4%. Identity of the compound was verified
by NMR analysis. Data were in good agreement with those
available in the literature (45). 12: purified by preparative
HPLC (Method 3). Blue solid. Yield = 24%. 1H NMR (300
MHz, DMSO-d6), ␦ = 12.77 (s, 1H); 12.38 (s, 1H); 7.94 (s,
1H); 7.15 (m, 4H); 4.67 (s, 2H); 4.18 (m, 4H); 3.69–3.57
1
for 1 min (␭ detection: 630, 254 and 310 nm). H-, 13C-
NMR spectra were recorded on a Bruker ADVANCE 300
MHz. HRMS were recorded on Agilent G6550A ESI Q-
TOF instrument coupled with an UPLC Agilent 1290 Infin-
ity II. UV/Vis spectra were run on an Agilent Cary 60 UV-
Vis or a SAFAS spectrophotometer. Emission experiments
were run on Cary Eclipse Fluorescence Spectrophotomer
and on a JobinYvon Fluorolog-3 (HORIBA) fluorimeter.
All spectrometers were equipped with Peltier temperature
controllers. Quartz cuvettes of 0.4–10 cm path length were
used.
Synthesis of compound 1
Compound 1 was synthesized, as reported in Scheme 2, ac-
cording to a published procedure (43). Analytical HPLC
(method 1) and comparison of NMR data with those avail-
able in the literature confirmed the identity and purity of
the compound.
Synthesis of compound 2 (Scheme 2)
Compound 6 was synthesized according to a published pro-
cedure (44). Compound 8 was obtained heating compound
(m, 12 H). The product exhibited too low solubility for 13
C
NMR acquisition. HRMS m/z calcd. for C28H25BrN4O8
[M+1]+ 625.09, 627.09 Found: 625.0885, 627.0897. 13: pu-
rified by preparative HPLC (Method 3). Blue solid. Yield
= 48%.1H NMR (300 MHz, DMSO-d6), ␦ = 12.2 (s, 2H);
7.75 (s, 2H); 7.06 (m, 2H); 6.99 (m, 2H); 4.59 (s, 2H); 4.05
(m, 4H); 3.58 (t, 4H); 3.47 (m, 8H). 13C NMR (75 MHz,
DMSO-d6), ␦ = 164.5; 161.9; 141.7; 127.2; 126.1; 125.2;
124.3; 121.2; 117.0; 95.1; 75.3; 66.9; 60.3. 2: purified by
preparative HPLC (Method 4). Green solid. Yield: 82%.
Scheme 2. General synthetic scheme of 4-substituted cex-NDIs: (a) amine,
acetic acid as solvent, 120◦C, 30 min; (b) 1,2-phenylendiamine, ace-
tonitrile, MW, closed vessel, 110◦C, 20 min, 250 psi; (c) amine, N,N-
dimethylacetamide, MW, closed vessel, 150◦C, 10 min, 250 psi.

e115 Nucleic Acids Research, 2018, Vol. 46, No. 19
PAGE 4 OF 13
1H NMR (300 MHz, DMSO-d6), ␦ = 11.8 (s, 1H); 11.0 (s,
1H); 8.83 (s, 1H); 6.71 (m, 2H); 6.66 (s, 1H); 6.43 (m, 2H);
4.62 (s, 3H); 3.86 (m, 4H); 3.68–3.46 (m, 18H). The prod-
uct exhibited too low solubility for 13C NMR acquisition.
HRMS m/z calcd. for C32H35N5O10 [M+1]+ 650.24, Found:
650.2419.
NaOH. The absorption spectrum was recorded after each
addition of base. Normalized absorption data in the two
maxima (617 and 670 nm) were plotted as a function of pH.
The presented results are the average of three replicates.
FRET melting
FRET melting experiments were run on a Stratagene
Mx3005P real-time PCR equipment in 96 wells plates, on
the DNA sequences reported in Supplementary Table S2.
Experiments were performed in 1 × 10−2 M lithium cacody-
late buffer (pH 7.2) and either 1 × 10−2 M KCl and 9 × 10−2
M LiCl (F21T, F32KRAST, F21CTAT, FBom17T, FTBAT,
FdxT) or 1 × 10−3 M KCl and 9.9 × 10−2 M LiCl (FcmycT,
F25CebT) concentrations, depending on the Tm of the G4s
alone. The DNA concentration was 2 × 10−7 M. The stabili-
sation (ꢀTm) induced by compounds 1 and 2 was calculated
as the difference between the mid-transition temperature of
the nucleic acid (NA) alone and measured with the relevant
ligand concentration (2.5 × 10−7 M for compound 1, 5 ×
10−6 M for compound 2). Data are presented as an average
of three independent measurements, each conducted in du-
plicate conditions (␭exc = 492 nm, ␭em = 516 nm, T interval
= 25–95◦C, ramp: 25◦C for 5 min, then 1◦C/min, measure-
ments every 1◦C, 8x magnification of the fluorescence sig-
nal). The G4 selectivity of the ligands with respect to duplex
and single stranded DNA was evaluated adding to the mix-
ture increasing amounts of the competitors (Supplementary
Table S3) and measuring the new ꢀTm.
Aggregation studies
Solvent dependence. Compounds solutions were prepared
diluting the appropriate amount of the stock solutions
in the various solvents. Temperature dependent measure-
ments were performed by gradually heating the solution
and recording the absorption spectrum after each step.
Concentration dependent measurements were performed
adding aliquots of 5 × 10−4 or 5 × 10−3 M stock solutions to
the tested ones. When recording spectra in the micelles, so-
lution at three cex-NDI concentrations (5, 10, 20 × 10−6 M)
were prepared containing 2 × 10−2 M SDS. When recording
the spectra at variable organic solvent – water compositions,
solutions in the pure solvents already at the final concentra-
tions were mixed in the appropriate percentages.
Aggregation constant calculation. Compound 1 was dis-
solved in buffered water (Tris–HCl, pH 7.2) at a concen-
tration of 5 × 10−8 M and its absorption spectrum was
measured using a 10 cm path length cuvette. The concen-
tration was then gradually increased, recording the spec-
trum after each addition of aliquots of a 1 × 10−3 M so-
lution of the same compound. The same was done for com-
pound 2 in buffered water and then in 1:1 mixture of water
and methanol. Molar absorptivity data in the maxima were
plotted as a function of the concentration and fitted accord-
ing to a isodesmic model with GraphPad 7.0 (Prism). The
data are presented as an average of three independent repli-
cates.
Thiazole orange displacement assay (46)
Fluorescence displacement assay was run on a TECAN
plate reader, Infinite M1000 PRO, in 384 wells plates. Thi-
azole orange (TO) displacement was performed adding in-
creasing amounts of ligands 1 or 2 to the pre-folded NA (2.5
× 10−7 M) – TO mixture (5 × 10−7 M). Ligands concentra-
tion was incremented from 0 to 2.5 × 10−6 M, correspond-
ing to 10 molar equivalents with respect to the NA. Exper-
iments were performed in 1 × 10−2 M lithium cacodylate
buffer (pH 7.2) and 0.1 M KCl. The fluorescence spectra
were measured for each ligand addition between 500 and
Ionic strength dependence. 1 × 10−5 M solutions of com-
pounds 1 and 2 were prepared in water (1.0 × 10−2
M
lithium cacodylate buffer, pH 7.2) and aliquots of 1 M MCl
(M = Li, K, Na) were added, varying the salt concentration
from 0 to 2.0 × 10−1 M. Absorption spectra were recorded
after each addition (T = 25◦C). Spectra were corrected for
the dilution.
800 nm on 4 × 10−5 L of solution (␭ = 485 nm, band-
exc
width = 5 nm, z position and gain were adapted for each
specific NA-TO complex, T = 25◦C). The % of displace-
ment was calculated as 100 × (1 – F/F0), where F is the TO
fluorescence in the maximum (535 nm) and F0 is the initial
fluorescence. Data are presented as an average of three in-
dependent replicates, all conducted in duplicate conditions.
Temperature dependence. 2 × 10−5 M solutions of com-
pounds 1 and 2 were prepared in water (1.0 × 10−2
M
lithium cacodylate buffer, pH 7.2, 0.1 M KCl) and their
absorption spectra were recorded raising the temperature
from 25 to 95◦C, with 5◦C steps (1 min equilibration time).
Mass spectrometry experiments
pH dependence. 1 × 10−5 M solutions of compound 2
were prepared in buffered water (1.0 × 10−2 M buffer,
either phosphoric acid-sodium dihydrogenphosphate, or
sodium dihydrogenphosphate-sodium monohydrogenphos-
phate, or sodium monohydrogenphosphate-sodium phos-
phate) and their absorption was subsequently recorded (T
= 25◦C). Compound 1 was instead dissolved in phosphoric
acid-sodium dihydrogenphosphate buffered solution (pH
3.5, 1 × 10−3 M buffer, 1 × 10−2 M NaCl) and the pH was
then gradually increased up to 12 upon addition of 0.1 M
Electrospray ionization mass spectrometry (ESI-MS) ex-
periments were performed on a Thermo- Exactive Orbitrap
mass spectrometer in the negative ion mode, equipped with
a standard ESI source. Samples were injected at 3.5 ␮l/min
by a syringe pump. The full scan mass range was [200–
4000]. The Exactive was tuned to soft conditions using the
bimolecular quadruplex d[G4T4G4]2 in 0.1 M ammonium
acetate (47). G4 solutions were prepared in 0.1 M trimethyl
ammonium acetate (TMAA), pH 7.0, and 1 × 10−3 M KCl.

PAGE 5 OF 13
Nucleic Acids Research, 2018, Vol. 46, No. 19 e115
were measured at a 3 × 10−6 M concentration of cex-NDIs.
cex-NDI-G4 complex ␸ was measured after the addition of
three molar equivalents of the relevant G4 (25Ceb, c-myc or
22AG).
The nucleic acids were analysed at a concentration of 5 ×
10−6 M, adding the relevant amounts of compounds 1 or
2 and/or thiazole orange. The mixture was supplemented
with 10% of methanol just before measurement. Data were
analysed using Xcalibur 2.2.0 software (Thermo Scientific).
LOD measurements
In vitro sensing studies
Limit of detection (LOD) / sensitivity assay was run on
a TECAN plate reader, Infinite M1000 PRO, in 384 wells
plates. Solutions containing equimolar amounts of the rele-
vant G4 (25Ceb or c-myc) and compound (2 or NMM por-
phyrin) at decreasing concentration were obtained by sub-
sequent dilution of a 4 × 10−6 M one. Blank was measured
on a 4 × 10−6 M solution of either compound 2 or NMM
alone. All solutions contained 1 × 10−2 M lithium cacody-
late buffer (pH 7.2) and 0.1 M KCl. Experiments on 2 were
carried out exciting at 650 nm for 2 and at 400 nm for NMM
and collecting the emission respectively in the 670–900 nm
and 500–800 nm intervals. Sensitivity was assessed by plot-
ting the fluorescence in the maxima (695 and 600 nm respec-
tively) as a function of the concentration and calculating the
LOD according to the following formula (50):
1 × 10−3 L of 4 × 10−6 M solutions of compounds 1
and 2 were titrated with 1 × 10−4 M solutions of the rele-
vant nucleic acid, after annealing (0.1 M KCl, 0.01 M Tris–
HCl buffer, pH 7.2). All measurements were performed at
25◦C. Absorption or emission (␭exc = 650 nm) spectra were
recorded after each NA aliquot addition (NA concentra-
tion: 0, 0.2, 0.4, 0.6, 0.8, 1.0, 1.2, 1.4, 1.6, 1.8, 2.0, 2.4, 2.8,
3.2, 3.6, 4.0, 4.8, 5.6, 6.4, 7.2, 8.0 × 10−6 M). They were then
corrected for the dilution. Data are presented as an average
of three independent replicates.
Fitting
Fitting was performed according to the following equation
(48), using GraphPad 7.0:
sb ∗ k
LOD =
m
Finf
F
=
2Ltot
⎡
⎤
⎦
ꢀ
where sb is the standard deviation calculated out of 20 inde-
pendent measurements on blank solutions; k is 3, according
to IUPAC recommendations; m is the curve slope, obtained
from data linear fitting. Data are presented as an average of
three independent replicates, each conducted in duplicate
conditions.
ꢁ
ꢂ
2 − 4nLtot
x
1
1
⎣
∗
nx + Ltot
+
−
nx + Ltot
+
Ka
Ka
where F is the normalized fluorescence response, Finf is the
fluorescence at infinite G4 concentration, Ltot is the analyt-
ical ligand concentration, n is the number of binding sites
and Ka is the affinity constant per site (Table 1). The equa-
tion thus optimizes both the binding stoichiometry and the
binding constant based on the supplied dataset. Such fit-
ting relies on the assumption, usually met with G4s, that
all binding sites display similar affinity and thus they are
involved as first binding sites according to a statistical dis-
tribution.
Circular dichroism studies
CD experiments were run on a Jasco J-1500 spectropo-
larimeter, equipped with a Peltier temperature controller.
Titrations were performed on 3 × 10−6 M solutions of the
relevant NAs (0.1 M KCl, 0.01 M lithium cacodylate buffer,
pH 7.2) upon addition of aliquots of a 5 × 10−4 M solu-
tion of compound 1 or compound 2. Parameters: scan speed
50 nm/min, three acquisitions, bandwidth 2 nm, integration
time 1 s, T = 25◦C.
Fluorescence quantum yield measurements
Fluorescence quantum yield values for the compounds
alone or upon G4 complexation were obtained using rho-
damine 800 as primary reference [␸ = 0.25 0.03 in ethanol
(49)]. Emission was measured exciting the compounds at
three different excitation wavelengths (600, 605 and 610 nm)
and integrating the resulting spectrum between 630 and 900
nm. ꢁ values were calculated applying the following for-
mula:
Native gel electrophoresis
Twelve percent polyacrylamide gels were prepared dilut-
ing 40% acrylamide-bisacrylamide (19:1) with water and
adding the relevant amounts of KCl 100×, TBE 5× (Tris-
Borate, EDTA buffer), APS (ammonium persulfate) solu-
tion and TEMED (tetramethylenediamine). The solution
was poured between the glass slides and let polymerize (0.4
cm thickness, 20 wells comb). Samples were prepared dilut-
ing the appropriate amount of 5 × 10−5 M pre-folded NA
solutions and compounds stock solutions in buffer (0.01 M
lithium cacodylate, pH 7.2, and 0.1 M KCl). 20 ␮l of each
sample were supplemented with 6 ␮l of sucrose solution
(50% m/v) and loaded. A ladder was prepared containing
coloured markers (0.1% xylene cyanol and 0.1% bromphe-
nol blue) and 5 × 10−6 M single stranded NAs (dT9, dT15,
dT21, dT30, dT57). The gels were run at 4 W (110 V, 37 mA)
for 2 h or until the front marker dye had reached the lower
Areasample
ϕsample = ϕreferences
∗
Areareference
ηs2ample
Absreference
Abssample
∗
∗
ηr2eference
where Area is the area subtended by the curve, Abs is the
absorption value at the excitation wavelength and η is the
refractive index of the solution. All data are presented as
a mean of the values obtained at the three wavelengths and
are the average of three replicates. Residual ␸ values in water

e115 Nucleic Acids Research, 2018, Vol. 46, No. 19
PAGE 6 OF 13
Table 1. Affinity constant values (Ka /M−1) obtained from the fitting of the data presented in Supplementary Figure S16. n is the number of binding sites
and Ka is the affinity constant per site.
cex-NDI 2
cex-NDI 1
NA
Ka (M−1
)
n
Ka (M−1
)
n
25Ceb
c-myc
32KRAS
22AG
ds26
5.8 0.5 × 106
4.1 0.5 × 106
1.0 0.1 × 105
2.20 0.50 × 104
2.84 0.06 × 103
1
1
1
1
1
a
a
a
a
a
a
2
2
2.5 0.3 × 106
3.2 0.8 × 106
a) Data not yielding a reliable fitting with our method.
end of the gel. The gels were then extracted and imaged
with a Typhoon Trio variable-mode imager (GE Health-
tration increase or cooling (Supplementary Figure S1C and
D). We thus ran the same experiments in DMSO, a highly
disaggregating solvent. Effectively, cex-NDI 2 was present
mostly in the monomeric form under these conditions (Sup-
plementary Figure S2).
care) (␭ = 633 nm, emission: 670 nm filter, PMT = 600,
exc
200 ␮m resolution, + 3 mm focal plane). Subsequently they
were exposed to SYBR Gold staining solution (1:10 000
stock solution dilution in water, containing 1 × 10−3
M
Moreover, the absorption bands for both 1 and 2 have
a significantly different shape and lower molar absorptiv-
ity in buffered aqueous solvent than in organic media.
Such bands were assigned to the H-aggregates. In addi-
tion, the absorption behaviour reflected in high fluores-
cence in organic solvent and negligible emission in water
(Figure 1A and B, dashed line). More specifically, in aque-
ous solvent, emission of compound 1 was low but still de-
tectable, whereas that of compound 2 was almost com-
pletely quenched. On the contrary, emission was restored in
methanol, resulting in 0.25 0.01 and 0.33 0.01 fluores-
cent quantum yields (␸) for cex-NDIs 1 and 2, respectively.
The lower value obtained for compound 1 might be due to
partial quenching via back-electron transfer to the amine
moieties.
The residual ␸ in buffered water (pH 7.2) were 0.017
0.002 and 0.0051 0.0006 for cex-NDIs 1 and 2, respec-
tively. This, together with the broader band shape observed
in absorption for cex-NDI 2 likely indicates tighter aggre-
gation. In order to quantify this difference, we measured
aggregation constants for the two compounds in buffered
water (10−2 M Tris–HCl buffer, pH 7.2). For cex-NDI 1 we
were able to measure an aggregation constant of 9.8(6) ×
105 M−1 (number in parenthesis indicates uncertainty in
the measurement; Supplementary Figure S3A and B). How-
ever, this was not possible for cex-NDI 2, as the aggregate
was the only species observed even at the lower detection
limit (Supplementary Figure S3C). We thus measured this
latter aggregation constant in a 1:1 mixture of water and
methanol. The measured constant [7.1(5) × 105 M−1] was
comparable to that of cex-NDI 1 in neat water (Supplemen-
tary Figure S3D and E).
lithium cacodylate buffer, pH 7.2, 40 min exposure), washed
twice and reimaged (␭exc = 473 nm, emission: 530 nm filter,
PMT = 600 V, 200 microns resolution, + 3 mm focal plane).
RESULTS AND DISCUSSION
Synthesis and characterization of the aggregation process
Compound 1 was synthesized as already reported (43), via
imidation of 6 with N,N-dimethylpropandiamine, yielding
7, subsequent nucleophilic aromatic substitution (SNAr)
with 1,2-phenylendiamine, ring closure and final SNAr
with N,N-dimethylpropanediamine at position 4 (Scheme
2). The same synthetic scheme was applied for com-
pound 2, replacing N,N-dimethylpropanediamine with 2-
(2-aminoethoxy)ethanol. The conditions, mostly for purifi-
cation, were however adapted to the increased hydrophobic-
ity of compound 2 and intermediates. In fact, compound
8 could not simply be recovered, unlike 7, by acetic acid
neutralization and extraction, since this latter step was in-
efficient. We thus decided not to isolate compound 8, us-
ing the crude for the following step. Similarly, the poor sol-
ubility of intermediate 12 in both water and organic sol-
vents hampered its purification. This was instead achieved
in good yields for 11 by reverse phase HPLC. Nevertheless,
such low solubility made possible the co-precipitation of
a 50:50/12:13 mixture, leaving the by-products in the ace-
tonitrile solution. Finally, the last SNAr step, proved to be
quantitative for both compounds 11 and 12. However, com-
pound 2 had to be purified by HPLC to eliminate the afore-
mentioned by-product 13.
To assess self-aggregation, we next evaluated the spectro-
scopic characteristics of 1 and 2 as a function of solvent,
concentration and temperature. Both cex-NDIs displayed
an intense charge transfer (CT) band with maxima at 595
and 647 nm in methanol (Figure 1A and B), consistently
red-shifted in comparison to the spectra of compounds 3–5
(Figure 1C) (39,40). We assigned the CT band observed in
methanol to the monomer of cex-NDI 1, as its shape was
not affected by compound concentration and temperature
(Supplementary Figure S1A and B). In contrast, compound
2 displayed evidence of aggregation in methanol as its ab-
sorptivity decreased and the band broadened upon concen-
Finally, we ran experiments detailing the aggregation
process dependence on the various environmental condi-
tions, including solvent composition, ionic strength, tem-
perature and pH. Concerning the solvent composition, the
disaggregation process was observed in both absorption
and emission when increasing the organic solvent content in
the mixture with water (Supplementary Figure S4). In fact,
the spectrum of the monomer reappeared for both com-
pounds upon addition of MeOH to the mixture and so did
the emission band. Interestingly, a much higher amount of
organic solvent was required to obtain effective monomer-

PAGE 7 OF 13
Nucleic Acids Research, 2018, Vol. 46, No. 19 e115
Figure 1. Absorption and emission (␭ = 620 nm in A and B, 550 nm in C) of A) compound 1, (B) compound 2 and (C) compound 3 in water (pH 7.2,
exc
0.01 M lithium cacodylate buffer) and methanol. Absorption spectra were recorded at 2 × 10-5 M, while emission spectra at 5 × 10-6 M concentration.
ization of compound 2 with respect to 1, confirming its
tighter aggregation. Finally, the monomer spectrum could
be observed in water for both compounds in the presence of
SDS micelles (Supplementary Figure S5).
cex-NDI 2 gives interesting information on its selectivity
for G4s over duplex DNA. In fact, no stabilization was
observed for FdxT even at such a high ligand concentra-
tion. To get a clearer picture on G4/ds selectivity, we per-
formed FRET competition experiments in the presence of
the duplex competitor ds26 (0, 3 and 10 ␮M). Interestingly,
whereas compound 1 underwent a dramatic drop in stabi-
lization for all analysed G4s already at low competitor con-
centration, compound 2 did not (Figure 2B and C). This
held true also at higher competitor concentration (5 × 10−5
M) and with different duplex and single strand competi-
tors (Supplementary Table S3, Supplementary Figure S12).
Only when confronted with single stranded DNA, com-
pound 1 did not show any decrease in stabilization activity
(Supplementary Figure S12). These results are particularly
important, because they confirm that positive charges on
compound 1 enhance the affinity for NAs, but in a highly
unspecific manner.
An additional assay addressing ligands affinity for G4s
is fluorescent intercalator displacement assay (G4-FID), in
which the ligand complexation drives the displacement of
thiazole orange (TO), a fluorescent probe (46). Affinity of
the competing ligand can be assessed by monitoring the
changes in TO emission. When tested by G4-FID assay,
compound 1 proved capable of displacing TO from all NAs,
including ds26, although to different extents (Figure 3A).
In fact, more than 95% of displacement was obtained only
with parallel G4s 25Ceb, c-myc and 32KRAS and with the
hybrid 22AG. The displacement was complete upon addi-
tion of only two molar equivalents (eq.) of 1 for the paral-
lel G4s, whereas around five eq. were required for 22AG,
suggesting a lower affinity. Concerning anti-parallel G4s
(21CTA and Bom17) and ds26, displacement could not be
higher than 80%. The curve shapes suggest a less efficient
competition with TO for their binding sites, particularly
when considering 21CTA. The ability to displace TO from
ds26 highlights once again the lack of selectivity for G4s
over duplex DNA, as already observed in FRET melting
assays.
When testing compound 1, the environmental condi-
tions significantly affected the equilibrium position in wa-
ter. In particular, the aggregate was favoured upon increase
the ionic strength and pH (Figures S6 and S7), while the
monomer emerged upon heating in neat water (Supplemen-
tary Figure S8). Among these, pH proved to be the most
influential factor, as aggregation was massively triggered
by amine deprotonation. Plotting the normalized molar
absorptivity coefficient data as a function of pH (Supple-
mentary Figure S7B) revealed a mid-transition pH of 8.3.
Smaller changes were observed for compound 2, when vary-
ing the ionic strength or the temperature (Figures S9 and
S10), and pH variation did not prove meaningful (Supple-
mentary Figure S11). Once again, this is due to the higher
aggregation propensity of 2, reducing the effect of external
factors.
G4 affinity
We first evaluated cex-NDIs 1 and 2 affinities for G4s
by FRET melting assays on a small panel of fluores-
cently labelled NAs (Supplementary Table S2). This high-
throughput test quantifies the stabilization imparted by the
ligand to the NA secondary structure, by comparing its
melting alone and upon complexation (51).
When testing compound 1, we observed an overall high
stabilization for G4s, even at compound concentration as
low as 2.5 × 10−7 M (Figure 2A). In more detail, telomeric
F21T was the most stabilized G4, with a ꢀTm of 19.4◦C.
Lower ꢀTm values (∼10◦C) were observed with FmycT,
F25CebT, F21CTAT and F32KRAST. Finally, FTBAT and
F17BomT displayed the smallest ꢀTm values, between 4
and 5◦C. Hairpin duplex DNA FdxT did not show any in-
crease in stabilization at this compound concentration, al-
though measurable stabilizations were found with higher
ligand quantities (data not shown). In order to obtain com-
parable ꢀTm values with compound 2 we had to increase
its concentration to 5 × 10−6 M, corresponding to 20-fold
the amount of cex-NDI 1 (Figure 2A). This confirms our
preliminary hypothesis that the replacement of protonable
amine groups by uncharged DEG moieties reduces the over-
all ligand affinity for the G4. Besides, the radar plot for
Surprisingly, compound 2 was completely unable to dis-
place TO even at high concentrations (Figure 3B), although
high ꢀTm values were found in FRET-melting experiments.
We reasoned that the absence of displacement might be due
to the simultaneous binding of TO and cex-NDI. To verify
this hypothesis, we analysed mixtures of Pu24 G4, TO and
2 by native mass spectrometry, at fixed stoichiometric ratios

e115 Nucleic Acids Research, 2018, Vol. 46, No. 19
PAGE 8 OF 13
Figure 2. (A) Results of FRET melting assays on a panel of labelled NAs (2 × 10−7 M, Supplementary Table S2), in the presence of 1 (2.5 × 10−7 M) or 2
(5 × 10−6 M); lithium cacodylate buffer 1 × 10−2 M, pH 7.2, KCl 1 × 10−2 M and LiCl 9 × 10−2 M (FTBAT, FBom17T, F21T, F21CTAT, F32KRAST)
or KCl 1 × 10−3 M and LiCl 9.9 × 10−2 M (F25CebT, FmycT); (B, C) results of FRET melting assays carried out in the same conditions but adding
increasing amounts of ds26 competitor (0, 3, 10 ␮M) with ligands 1 (B) or 2 (C).
Figure 3. FID assays of (A) cex-NDI 1 and (B) cex-NDI 2 on a panel of G4s (c-myc, 25Ceb, 32KRAS, 22AG, Bom 17, 21CTA) and on control ds26 (pH
7.2, 1 × 10−2 M lithium cacodylate buffer, 1 × 10−1 M KCl, T = 25◦C). Dotted lines indicate 50% displacement. Curves in A are only a guide for the eye.
(C) Native MS analysis of a Pu24, TO, cex-NDI 2 mixture (concentrations 5, 10, 50 ␮M respectively, 0.1 M TMAA, pH 7, 1 mM KCl). All charge states
shown are 5-. The annotated peaks correspond to complexes with 2 specific K+ ions bound. Asterisks indicate extra non-specific K+ adducts.
resembling FID conditions (Supplementary Figure S13 for
FID). When analysing 5 ␮M Pu24 in the presence of 2 eq.
of TO (10 ␮M) and increasing concentrations of 2 (0, 10, 25
and 50 ␮M; Supplementary Figure S14), we detected G4
complexes with the cex-NDI or TO individually, and also
relatively abundant ternary complexes: [Pu24·2·TO]5− and
[Pu24·(2)2·TO]5− (highlighted in green in Figure 3C), with
the latter becoming more abundant at higher concentra-
tions of 2. These complexes were observed at charge states
5- (Figure 3C), 4- and 6- (Supplementary Figure S14). The
native mass spectrometry experiment illustrates that nega-
tive FID results do not always mean that a compound does
not interact with the nucleic acid target: lack of TO displace-
ment by compound 2 may simply result from binding to two
different sites on the G-quadruplex. In this case, FID gave
a false negative result regarding the binding capabilities of
2.
myc and 32KRAS), three anti-parallel ones (Bom17, TBA
15 and hRAS1), a hybrid G4 (22AG in K+), plus a sin-
gle strand (ss SCR) and a duplex (ds26) as controls. Both
compounds had their fluorescence restored upon interac-
tion with the G4 NAs, such as 25Ceb (Figure 4A and B). We
confirmed by spectrophotometric titrations with the same
G4 that this was due to monomerization induced by G4
binding (Figure 4C and D). Such process induced changes
in the absorption spectra, producing monomer-like bands.
These were significantly redshifted with respect to those of
the monomer alone (675 and 617 nm, compared to 661 and
607 nm in SDS). However, remarkable differences were ob-
served for the two compounds, when analysing the whole
NA panel. In fact, cex-NDI 1 displayed a moderate emis-
sion light-up (up to 14-fold) at 687 nm after a modest initial
quenching, upon gradual NA addition (Figure 5A). More-
over, a robust turn-on effect was also detected in the pres-
ence of duplex and single-strand controls and no real selec-
tivity was observed for any G4 topology. In contrast, cex-
NDI 2, displayed much higher fluorescence light-up (up to
250-fold). This was selectively observed for parallel G4s,
compared to the moderate responses to other G4s (hybrid
and anti-parallel) or the negligible ones to the controls (Fig-
ure 5B). Fluorescent quantum yields values (␸) were calcu-
lated for the complexes of the two cex-NDIs with G4s. In
G4 fluorescent sensing
After evaluation of G4 affinity and fluorescent light up
upon monomerization for both 1 and 2, we investigated
their G4 sensing activity. We thus tested both cex-NDIs on
a panel of NA sequences in fluorescence titration experi-
ments. This NA list includes three parallel G4s (25Ceb, c-

PAGE 9 OF 13
Nucleic Acids Research, 2018, Vol. 46, No. 19 e115
Table 2. LOD values for 25Ceb and c-myc sensing by 2 and NMM, mea-
sured by fluorescence emission analysis with a plate reader
Probe
25Ceb
c-myc
2
10 nM
120 nM
8 nM
104 nM
NMM
any other non-parallel G4s. Similar effects were found with
RNA and DNA parallel G4s, suggesting that the paral-
lel topology itself is the key for this selective fluorescent
turn on. Statistical analysis of the enhancement factors in
the presence of 2 eq. of NAs revealed that the average re-
sponse recorded for parallel G4s is significantly different
from those observed with non-parallel G4s and controls (P
< 0.001, Supplementary Table S4). This preference was also
highlighted by apparent binding constants (Supplementary
Figure S16, Table 1). In fact, Ka values for the most respon-
sive G4s c-myc [4.1(5) × 106 M−1] and 25Ceb [5.8(5) × 106
M
−1] were two and three orders of magnitude higher than
those for 22AG and ds26, respectively [2.20(1) × 104 M−1
and 2.38(6) × 103 M−1]. Interestingly, they were also more
than one order of magnitude higher than the Ka calculated
for 32KRAS [1.0(1) × 105 M−1]. Moreover, comparison of
the fits for the two cex-NDIs revealed comparable affinity of
cex-NDI 1 for two binding sites (1 G4: 2 ligands), in contrast
with cex-NDI 2 (1 G4: 1 ligand). Finally, cex-NDI 1 bound
both 22AG and ds26 with Ka values [2.5(3) × 106 M−1 and
3.2(5) × 106 M−1, respectively] considerably higher than
those of cex-NDI 2. Such values were comparable in mag-
nitude to those calculated for cex-NDI 2 with the most re-
sponsive G4s 25Ceb and c-myc.
Finally, we addressed the sensitivity of compound 2 in
parallel G4s signalling by limit of detection (LOD) mea-
surement. We measured the fluorescence of solutions con-
taining equimolar amounts of ligand 2 and parallel G4s
25Ceb and c-myc over a range of concentrations. We did
the same with N-methyl mesoporphyrin IX (NMM), which
is a renowned selective ligand for parallel G4s (52,53). Inter-
polation of the resulting curves revealed that cex-NDI 2 has
LOD values more than one order of magnitude lower than
NMM (Table 2, Supplementary Figure S17). LOD values
in the low nanomolar range were also reported for green-
emitting triaryl-imidazole based ligands, selective for paral-
lel G4s (54). Because of this high sensitivity, together with
the considerably more red-shifted excitation and emission
wavelengths necessary for the cex-NDI, our new ligand pro-
vides a significant improvement of the available parallel G4
probes.
Figure 4. Fluorometric (A and B) and spectrophotometric (C and D) titra-
tions of 4 × 10−6 M solutions of (A, C) cex-NDI 1 and (B, D) cex-NDI 2 (1
× 10−2 M Tris–HCl buffer, pH 7.2, 1 × 10−1 M KCl) with 25Ceb G4 (from
0 M in black to 8 × 10−6 M in red, start and end curves of the titrations
are highlighted in bold, see in vitro sensing studies subsection for 25Ceb
concentration details).
particular, in the presence of three eq. of 25Ceb and c-myc,
2 displayed ␸ = 0.090 0.005 and 0.084 0.005, respec-
tively, whereas only 0.032
0.004 with 22AG. The latter
confirms the incomplete complexation by the telomeric G4
even when present in large excess. 1 exhibited comparable ␸
regardless to the G4 topology [0.095 0.002 with 25Ceb,
0.086 0.003 with c-myc and 0.099 0.005 in the presence
of 22AG].
In order to probe the selectivity of 2 for parallel G4s,
we decided to expand the NA panel, adding five parallel
DNA G4s (vav1, VEGF, c-kit2, c-kit1, bcl2) and two par-
allel RNA G4s (TERRA and NRAS). Compound 2 dis-
played a remarkable light-up with all of these sequences, al-
though to variable levels (from 85-fold with c-kit1 to 260-
fold with 25Ceb, Supplementary Figure S15). More specif-
ically, 25Ceb, c-myc and bcl2 proved to be the highest re-
sponding G4s. ckit-2, VEGF, 32KRAS, NRAS, TERRA
and Vav1 gave slightly lower light-up factors and ckit-1
was the least responding one, although still higher than
Binding selectivity or sensing selectivity?
Next, we tried to prove whether this selectivity of compound
2 observed for parallel G4s really concerned the bind-
ing event. Alternatively, it could arise from the formation
of non-fluorescent complexes with non-parallel quadru-
plexes and thus be limited to the sensing. We started by
qualitatively assessing the complexation by an emission-
independent method. We thus monitored the CD signature
of some selected G4s upon gradual addition of the two lig-

e115 Nucleic Acids Research, 2018, Vol. 46, No. 19
PAGE 10 OF 13
Figure 5. Normalized fluorescence in the maximum (685-95 nm), calculated from fluorescence titrations data of 4 × 10−6 M water solutions of (A) cex-NDI
1, B) cex-NDI 2 (pH 7.2, 1 × 10−2 M Tris–HCl, 1 × 10−1 M KCl, T = 25◦C) with a panel of NAs, c = 0–8 × 10−6 M (c-myc, 25Ceb, 32KRAS, 22AG,
hRAS1, Bom17, TBA 15, ss SCR, ds26).
ands. Even in this case, the two compounds displayed quite
different behaviours. In fact, compound 1 strongly affected
the signatures of all the tested NAs, while compound 2 af-
fected only those of parallel G4s. For example, both cex-
NDIs induced a small decrease in the absolute intensities
of both the negative and positive maxima for c-myc, respec-
tively at 240 nm and 265 nm (Figure 6A and B), maintaining
however the overall shape of the signal. On the other side,
only cex-NDI 1 induced significant changes on 22AG signa-
ture (Figure 6C and D). In fact, the 264 and 295 nm positive
maxima intensities increased considerably upon complexa-
tion. Contextually, the whole band was slightly blue-shifted
and the positive 250 nm maximum disappeared. Similar
considerations can be made about the other NAs (Supple-
mentary Figure S18), highlighting once again cex-NDI 2 se-
lectivity for parallel G4s.
Additional proofs of such binding preference were ob-
tained by native gel electrophoresis. Complexes formed in
solutions containing equimolar amounts of selected NAs
and ligand 1 or 2 were let to migrate in non-denaturing
conditions, so as to assess their stabilities. We tested a wide
panel of NAs containing G4s of different topologies (par-
allel 25Ceb and c-myc, anti-parallel TBA 15 and hybrid
22AG), an RNA parallel G4 (TERRA), single- and double-
stranded controls (DNA: ss37, ds26, RNA: SL2, mir122).
Interestingly, only complexes of 2 with parallel G4s, either
DNA or RNA, could effectively be observed (Figure 7A,
see Supplementary Figure S19A for SYBR Gold staining).
Instead, none could be visualized with G4s having other
topologies, although some complex was shown to be formed
at least with 22AG by fluorescence titrations. The same
outcome was observed with control NAs. This suggested
that only the complexes with the parallel G4s are stable
enough to survive during migration in a native gel. On the
other hand, compound 1 was found to stain all NAs, ex-
cept TBA (Figure 7B; see Supplementary Figure S19B for
SYBR Gold staining), providing a positive control for the
selectivity of the DEGylated analogue 2.
The modest solubility of ligand 2 and to lower extent
of ligand 1, at concentration much lower than mM, pre-
vented us from performing 1D-NMR titrations in the pres-
ence of parallel G4s. Therefore, in order to provide an addi-
tional and conclusive proof of 2 binding selectivity to par-
allel G4s over other topologies, we monitored competition
experiments by mass spectrometry. We chose Pu24, a paral-
lel G4 resembling c-myc, and 26TTA, a hybrid one related
to 22AG, which are well characterized with respect to their
behaviour in native MS conditions (55). Moreover, they dis-
play sufficient differences in mass to clearly visualize and
identify all free and complexed species. Fluorescence (Sup-
plementary Figure S20) and CD titrations (Supplementary
Figure S21) confirmed that they behave as expected in the
usual conditions, as well as in trimethylammonium acetate
(TMAA) buffer + 1 mM KCl, used for MS experiments. A
1:1 mixture of Pu24 and 26TTA (5 ␮M each) was analysed
in the presence of increasing amounts of compound 2 (0, 1,
25, 50 ␮M; Figure 8). When zooming on the 5- peaks, we
detect the selective formation of the 1:1 and 1:2 complexes
of NDI 2 with the parallel Pu24 G4, while the 1:1 complexes
with hybrid 26TTA appeared only at high cex-NDI concen-
tration. Concurrently, the signal of free Pu24 considerably
decreases compared to that of free 26TTA, as the complexes
with Pu24 increase. This is in line with the different extent
of involvement in the binding. As a control, we ran the same
experiment on compound 1 (Supplementary Figure S22). In
that case, complexes with both G4s were observed from the
very first ligand aliquot addition, suggesting a comparable
affinity for both G4s and confirming the reliability of the
methodology. We were thus able to confirm that cex-NDI 2
selectivity extends beyond G4 sensing, being an actual bind-
ing preference.

PAGE 11 OF 13
Nucleic Acids Research, 2018, Vol. 46, No. 19 e115
Figure 6. Circular dichroism spectra of 3 × 10−6 M solutions (pH 7.2, 0.01 M lithium cacodylate buffer, 0.1 M KCl, T = 25◦C) of (A, B) c-myc; (C, D)
22AG in the presence of 0–9 × 10−6 M cex-NDI 1 (A and C) or cex-NDI 2 (B and D), corresponding to 0–3 molar equivalents. Red: 0 eq; green: 0.5 eq;
blue: 1 eq; yellow: 1.5 eq; light blue: 2 eq; purple: 3 eq.
Figure 7. Native gel electrophoretic analysis of solutions containing either the NA alone or the NA with 1 equivalent of cex-NDI 2 (A) or 1 (B) (NA:
30 pmol, compound: 0–30 pmol, 12% polyacrylamide, 18◦C, 2 h, 150 V). (A) cex-NDI 2 fluorescence (␭ = 633 nm; ␭ = 670 nm); (B) cex-NDI 1
exc
em
fluorescence (␭ = 633 nm; ␭ = 670 nm). Green: parallel G4s; these are the only fluorescent bands revealed with 2 on the left gel. blue: non-parallel
exc
em
G4s, red: controls. L = ladder and fluorescent markers.
CONCLUSION
Several complementary techniques were used to obtain a
consistent view on the different selectivity of compounds
1 and 2. All methods converged to the same conclusion,
with the notable exception of FID, for which no compe-
tition could be found for 2. This false negative result was
explained by native mass spectrometry experiments, which
unambiguously demonstrate that the same quadruplex may
simultaneously accommodate TO and one or two 2. Finally,
compound 2 affinity and turn-on properties allow it to re-
veal the presence of parallel G-quadruplexes with limits of
detection in the low (8-10) nM range.
Achieving this level of G4 selectivity is still rare among
G4 sensors and ligands. Our study provides a rational basis
to design or modify existent scaffolds to redirect the binding
preference. We demonstrated that reducing the unspecific
electrostatic interactions was key to let the core (and possi-
bly DEG-substituents) selectivity emerge. Interestingly, the
compound NMM, exhibiting a clear preference for parallel
structures, does not bear any positive charges, and is actu-
In this work, we presented two core-extended naphthalene
diimides with a novel functionalization pattern, which we
exploited as G4 fluorescent sensors. Such rational design
improved the sensing effectiveness by affecting the ligand
binding selectivity. Sensing was greatly enhanced through
the increased self-aggregation and the shift of the com-
pounds emission in the red-NIR. The most remarkable re-
sult is the relevant selectivity for parallel G4s versus other
conformations obtained for compound 2, achieved by DEG
functionalization. Polyethylene glycol (PEG) preferential
interaction with parallel G4 structures was previously re-
ported for the telomeric sequence in K+ rich buffer, for
which a structural shift from hybrid to parallel topology
was observed (56). Initially presented as a result of molec-
ular crowding or dehydration (56,57), the effect was later
associated to specific CH-␲ and lone pair-␲ interactions of
the PEG chains with the G-quartets and loops (58,59). The
observed binding preference could rely on similar effects.

e115 Nucleic Acids Research, 2018, Vol. 46, No. 19
PAGE 12 OF 13
6. Burge,S., Parkinson,G.N., Hazel,P., Todd,A.K. and Neidle,S. (2006)
Quadruplex DNA: sequence, topology and structure. Nucleic Acids
Res., 34, 5402–5415.
7. Rhodes,D. and Lipps,H.J. (2015) G-quadruplexes and their
regulatory roles in biology. Nucleic Acids Res., 43, 8627–8637.
8. Xu,Y. (2011) Chemistry in human telomere biology: structure,
function and targeting of telomere DNA/RNA. Chem. Soc. Rev., 40,
2719–2740.
9. Bugaut,A. and Balasubramanian,S. (2012) 5ꢁ-UTR RNA
G-quadruplexes: translation regulation and targeting. Nucleic Acids
Res., 40, 4727–4741.
10. Dexheimer,T.S., Sun,D. and Hurley,L.H. (2006) Deconvoluting the
structural and drug-recognition complexity of the
G-quadruplex-forming region upstream of the bcl-2 P1 promoter. J.
Am. Chem. Soc., 128, 5404–5415.
11. Brooks,T.A. and Hurley,L.H. (2010) Targeting MYC expression
through G-quadruplexes. Genes Cancer, 1, 641–649.
12. Moye,A.L., Porter,K.C., Cohen,S.B., Phan,T., Zyner,K.G.,
Sasaki,N., Lovrecz,G.O., Beck,J.L. and Bryan,T.M. (2015) Telomeric
G-quadruplexes are a substrate and site of localization for human
telomerase. Nat. Commun., 6, 7643.
13. Balasubramanian,S., Hurley,L.H. and Neidle,S. (2011) Targeting
G-quadruplexes in gene promoters: a novel anticancer strategy? Nat.
Rev. Drug Discov., 10, 261–275.
14. Dai,J., Liu,Z.-Q., Wang,X.-Q., Lin,J., Yao,P.-F., Huang,S.-L.,
Ou,T.-M., Tan,J.-H., Li,D., Gu,L.-Q. et al. (2015) Discovery of small
molecules for Up-Regulating the translation of antiamyloidogenic
secretase, a disintegrin and metalloproteinase 10 (ADAM10), by
binding to the G-Quadruplex-Forming sequence in the 5ꢁ
untranslated region (UTR) of its mRNA. J. Med. Chem., 58,
3875–3891.
15. Zhou,B., Liu,C., Geng,Y. and Zhu,G. (2015) Topology of a
G-quadruplex DNA formed by C9orf72 hexanucleotide repeats
associated with ALS and FTD. Sci. Rep., 5, 16673.
16. Amrane,S., Kerkour,A., Bedrat,A., Vialet,B., Andreola,M.-L. and
Mergny,J.-L. (2014) Topology of a DNA G-Quadruplex structure
formed in the HIV-1 Promoter: a potential target for anti-HIV drug
development. J. Am. Chem. Soc., 136, 5249–5252.
17. Artusi,S., Nadai,M., Perrone,R., Biasolo,M.A., Palu,G., Flamand,L.,
Calistri,A. and Richter,S.N. (2015) The herpes simplex Virus-1
genome contains multiple clusters of repeated G-quadruplex:
Implications for the antiviral activity of a G-quadruplex ligand.
Antiviral Res., 118, 123–131.
18. Belmonte-Reche,E., Mart´ınez-Garc´ıa,M., Gue´din,A., Zuffo,M.,
Are´valo-Ruiz,M., Doria,F., Campos-Salinas,J., Maynadier,M.,
Lo´pez-Rubio,J.J., Freccero,M. et al. (2018) G-Quadruplex
identification in the genome of protozoan parasites points to
naphthalene diimide ligands as new antiparasitic agents. J. Med.
Chem., 61, 1231–1240.
19. Smargiasso,N., Gabelica,V., Damblon,C., Rosu,F., De Pauw,E.,
Teulade-Fichou,M.-P., Rowe,J.A. and Claessens,A. (2009) Putative
DNA G-quadruplex formation within the promoters of Plasmodium
falciparum var genes. BMC Genomics, 10, 362–362.
20. Balasubramanian,S., Hurley,L.H. and Neidle,S. (2011) Targeting
G-quadruplexes in gene promoters: a novel anticancer strategy? Nat.
Rev. Drug Discov., 10, 261–275.
21. Collie,G.W. and Parkinson,G.N. (2011) The application of DNA and
RNA G-quadruplexes to therapeutic medicines. Chem. Soc. Rev., 40,
5867–5892.
22. Biffi,G., Antonio,, M.,Tannahill, D. and Balasubramanian,S. (2014)
Visualization and selective chemical targeting of RNA G-quadruplex
structures in the cytoplasm of human cells. Nat. Chem., 6, 75–80.
23. Biffi,G., Tannahill,D., McCafferty,J. and Balasubramanian,S. (2013)
Quantitative visualization of DNA G-quadruplex structures in
human cells. Nat. Chem., 5, 182–186.
24. Liu,H.-Y., Zhao,Q., Zhang,T.-P., Wu,Y., Xiong,Y.-X., Wang,S.-K.,
Ge,Y.-L., He,J.-H., Lv,P., Ou,T.-M. et al. (2016) Conformation
selective antibody enables genome profiling and leads to discovery of
parallel G-Quadruplex in human telomeres. Cell Chem. Biol., 23,
1261–1270.
25. Largy,E., Granzhan,A., Hamon,F., Verga,D. and
Teulade-Fichou,M.-P. (2013) In: Chaires,JB and Graves,D (eds).
Quadruplex Nucleic Acids. Springer, Berlin, Heidelberg, 111–177.
Figure 8. Mass spectra of a 1:1 mixture of Pu24 and 26TTA G4s (5 × 10−6
M each, 1 × 10−1 M TMAA, pH 7, 1 × 10−3 M KCl) in the presence of
increasing amounts of cex-NDI 2 (0–5 × 10−5 M, corresponding to 0, 2, 5
and 10 molar equivalents).
ally negatively charged at neutral pH. Our results thus argue
against the use of cationic chains in order to increase affinity
towards parallel G4 structures. We anticipate that the out-
lined strategy could be more widely applied to the design
of novel G4 ligands and optimization of existent ones, be-
yond naphthalene diimides. We envisage this would provide
remarkable benefits to the G4 field. Differentiated and se-
lective targeting by small fluorescent probes would certainly
improve the understanding of G4s biological roles.
SUPPLEMENTARY DATA
Supplementary Data are available at NAR Online.
FUNDING
European Research Council [ERC Consolidator
grant 615879 to M.F.]; Italian Association for
Cancer Research [14708 to M.F.]; European Re-
gional Development Fund [SYMBIT project reg. no.
CZ.02.1.01/0.0/0.0/15 003/0000477 to J.L.M.]. Funding
for open access charge: ERC Consolidator [615879].
Conflict of interest statement. None declared
REFERENCES
1. Ha¨nsel-Hertsch,R., Di Antonio,M. and Balasubramanian,S. (2017)
DNA G-quadruplexes in the human genome: detection, functions
and therapeutic potential. Nat. Rev. Mol. Cell. Biol., 18, 279–284.
2. Huppert,J.L. (2008) Four-stranded nucleic acids: structure, function
and targeting of G-quadruplexes. Chem. Soc. Rev., 37, 1375–1384.
3. Largy,E., Mergny,J.-L. and Gabelica,V. (2016) In: Sigel,A, Sigel,H
and Sigel,RKO (eds). The Alkali Metal Ions: Their Role for Life.
Springer International Publishing, Cham, 203–258.
4. Bhattacharyya,D., Mirihana Arachchilage,G. and Basu,S. (2016)
Metal cations in G-Quadruplex folding and stability. Front. Chem., 4,
38.
5. Phan,A.T., Luu,K.N. and Patel,D.J. (2006) Different loop
arrangements of intramolecular human telomeric (3+1)
G-quadruplexes in K+ solution. Nucleic Acids Res., 34, 5715–5719.

PAGE 13 OF 13
Nucleic Acids Research, 2018, Vol. 46, No. 19 e115
26. Vummidi,B.R., Alzeer,J. and Luedtke,N.W. (2013) Fluorescent
probes for G-quadruplex structures. ChemBioChem, 14, 540–558.
27. Bhasikuttan,A.C. and Mohanty,J. (2015) Targeting G-quadruplex
structures with extrinsic fluorogenic dyes: promising fluorescence
sensors. Chem. Commun., 51, 7581–7597.
28. Faverie,A., Gue´din,A., Bedrat,A., Yatsunyk,L.A. and Mergny,J.-L.
(2014) Thioflavin T as a fluorescence light-up probe for G4
formation. Nucleic Acids Res., 42, e65.
29. Grande,V., Doria,F., Freccero,M. and Wu¨rthner,F. (2017) An
aggregating amphiphilic squaraine: a light-up probe that
discriminates parallel G-Quadruplexes. Angew. Chem. Int. Ed. Engl.,
56, 7520–7524.
30. Laguerre,A., Wong,J.M. and Monchaud,D. (2016) Direct
visualization of both DNA and RNA quadruplexes in human cells
via an uncommon spectroscopic method. Sci. Rep., 6, 32141.
31. Doria,F., Oppi,A., Manoli,F., Botti,S., Kandoth,N., Grande,V.,
Manet,I. and Freccero,M. (2015) A naphthalene diimide dyad for
fluorescence switch-on detection of G-quadruplexes. Chem.
Commun., 51, 9105–9108.
32. Zuffo,M., Doria,F., Spalluto,V., Ladame,S. and Freccero,M. (2015)
Red/NIR G-quadruplex sensing, harvesting blue light by a
coumarin–naphthalene diimide dyad. Chem.– Eur. J., 21,
17596–17600.
33. Arthanari,H., Basu,S., Kawano,T.L. and Bolton,P.H. (1998)
Fluorescent dyes specific for quadruplex DNA. Nucleic Acids Res.,
26, 3724–3728.
34. Yang,P., De Cian,A., Teulade-Fichou,M.-P., Mergny,J.-L. and
Monchaud,D. (2009) Engineering bisquinolinium/thiazole orange
conjugates for fluorescent sensing of G-Quadruplex DNA. Angew.
Chem. Int. Ed. Engl., 48, 2188–2191.
35. Zhang,L., Er,J.C., Li,X., Heng,J.J., Samanta,A., Chang,Y.-T. and
Lee,C.-L.K. (2015) Development of fluorescent probes specific for
parallel-stranded G-quadruplexes by a library approach. Chem.
Commun., 51, 7386–7389.
36. Hu,M.-H., Chen,S.-B., Wang,Y.-Q., Zeng,Y.-M., Ou,T.-M., Li,D.,
Gu,L.-Q., Huang,Z.-S. and Tan,J.-H. (2016) Accurate
high-throughput identification of parallel G-quadruplex topology by
a new tetraaryl-substituted imidazole. Biosens. Bioelectron., 83,
77–84.
37. Jin,B., Zhang,X., Zheng,W., Liu,X., Zhou,J., Zhang,N., Wang,F. and
Shangguan,D. (2014) Dicyanomethylene-functionalized squaraine as
a highly selective probe for parallel G-quadruplexes. Anal. Chem., 86,
7063–7070.
38. Zhang,L., Er,J.C., Ghosh,K.K., Chung,W.J., Yoo,J., Xu,W., Zhao,W.,
Phan,A.T. and Chang,Y.-T. (2014) Discovery of a structural-element
specific G-quadruplex “light-up” probe. Sci. Rep., 4, 3776.
39. Zuffo,M., Doria,F., Botti,S., Bergamaschi,G. and Freccero,M. (2017)
G-quadruplex fluorescence sensing by core-extended naphthalene
diimides. Biochim. Biophys. Acta (BBA) - Gen. Sub., 1861,
1303–1311.
40. Doria,F., Nadai,M., Zuffo,M., Perrone,R., Freccero,M. and
Richter,S.N. (2017) A red-NIR fluorescent dye detecting nuclear
DNA G-quadruplexes: in vitro analysis and cell imaging. Chem.
Commun., 53, 2268–2271.
43. Perrone,R., Doria,F., Butovskaya,E., Frasson,I., Botti,S.,
Scalabrin,M., Lago,S., Grande,V., Nadai,M., Freccero,M. et al.
(2015) Synthesis, binding and antiviral properties of potent
Core-extended naphthalene diimides targeting the HIV-1 long
terminal repeat promoter G-Quadruplexes. J. Med. Chem., 58,
9638–9652.
44. Doria,F., Di Antonio, M., Benotti,M., Verga,D. and Freccero,M.
(2009) Substituted heterocyclic naphthalene diimides with unexpected
acidity. Synthesis, properties, and reactivity. J. Org. Chem., 74,
8616–8625.
45. Bevers,S., O’Dea,P.T. and McLaughlin,L.W. (1998) Perylene- and
naphthalene-Based linkers for duplex and triplex stabilization. J. Am.
Chem. Soc., 120, 11004–11005.
46. Monchaud,D., Allain,C., Bertrand,H., Smargiasso,N., Rosu,F.,
Gabelica,V., De Cian,A., Mergny,J.L. and Teulade-Fichou,M.P.
(2008) Ligands playing musical chairs with G-quadruplex DNA: a
rapid and simple displacement assay for identifying selective
G-quadruplex binders. Biochimie, 90, 1207–1223.
47. Marchand,A. and Gabelica,V. (2014) Native electrospray mass
spectrometry of DNA G-quadruplexes in potassium solution. J. Am.
Soc. Mass Spectrom., 25, 1146–1154.
48. Stootman,F.H., Fisher,D.M., Rodger,A. and Aldrich-Wright,J.R.
(2006) Improved curve fitting procedures to determine equilibrium
binding constants. Analyst, 131, 1145–1151.
49. Alessi,A., Salvalaggio,M. and Ruzzon,G. (2013) Rhodamine 800 as
reference substance for fluorescence quantum yield measurements in
deep red emission range. J. Lumin., 134, 385–389.
50. Hu,M.-H., Chen,S.-B., Guo,R.-J., Ou,T.-M., Huang,Z.-S. and
Tan,J.-H. (2015) Development of a highly sensitive fluorescent
light-up probe for G-quadruplexes. Analyst, 140, 4616–4625.
51. De Rache,A. and Mergny,J.-L. (2015) Assessment of selectivity of
G-quadruplex ligands via an optimised FRET melting assay.
Biochimie, 115, 194–202.
52. Nicoludis,J.M., Barrett,S.P., Mergny,J.-L. and Yatsunyk,L.A. (2012)
Interaction of human telomeric DNA with N-methyl mesoporphyrin
IX. Nucleic Acids Res., 40, 5432–5447.
53. Sabharwal,N.C., Savikhin,V., Turek-Herman,J.R., Nicoludis,J.M.,
Szalai,V.A. and Yatsunyk,L.A. (2014) N-methylmesoporphyrin IX
fluorescence as a reporter of strand orientation in guanine
quadruplexes. FEBS J., 281, 1726–1737.
54. Hu,M.H., Chen,X., Chen,S.B., Ou,T.M., Yao,M., Gu,L.Q.,
Huang,Z.S. and Tan,J.H. (2015) A new application of click chemistry
in situ: development of fluorescent probe for specific G-quadruplex
topology. Sci. Rep., 5, 17202.
55. Marchand,A. and Gabelica,V. (2016) Folding and misfolding
pathways of G-quadruplex DNA. Nucleic Acids Res., 44,
10999–11012.
56. Xue,Y., Kan,Z.-Y., Wang,Q., Yao,Y., Liu,J., Hao,Y.-H. and Tan,Z.
(2007) Human telomeric DNA forms parallel-stranded
intramolecular G-quadruplex in K+ solution under molecular
crowding condition. J. Am. Chem. Soc., 129, 11185–11191.
57. Miyoshi,D. and Sugimoto,N. (2008) Molecular crowding effects on
structure and stability of DNA. Biochimie, 90, 1040–1051.
58. Buscaglia,R., Miller,M.C., Dean,W.L., Gray,R.D., Lane,A.N.,
Trent,J.O. and Chaires,J.B. (2013) Polyethylene glycol binding alters
human telomere G-quadruplex structure by conformational selection.
Nucleic Acids Res., 41, 7934–7946.
59. Tateishi-Karimata,H., Ohyama,T., Muraoka,T., Podbevsek,P.,
Wawro,A.M., Tanaka,S., Nakano,S.I., Kinbara,K., Plavec,J. and
Sugimoto,N. (2017) Newly characterized interaction stabilizes DNA
structure: oligoethylene glycols stabilize G-quadruplexes CH-pi
interactions. Nucleic Acids Res., 45, 7021–7030.
41. Cuenca,F., Greciano,O., Gunaratnam,M., Haider,S., Munnur,D.,
Nanjunda,R., Wilson,W.D. and Neidle,S. (2008) Tri- and
tetra-substituted naphthalene diimides as potent G-quadruplex
ligands. Bioorg. Med. Chem. Lett., 18, 1668–1673.
42. Sakai,N., Mareda,J., Vauthey,E. and Matile,S. (2010)
Core-substituted naphthalenediimides. Chem. Commun., 46,
4225–4237.