Full text of DOI:10.1002/chem.202005026

Full Paper
doi.org/10.1002/chem.202005026
Chemistry—A European Journal
&
Quadruplex–Duplex Junctions |Very Important Paper|
De Novo Design of Selective Quadruplex–Duplex Junction
Ligands and Structural Characterisation of Their Binding Mode:
Targeting the G4 Hot-Spot
Laura Díaz-Casado+,[a] Israel Serrano-Chacón+,[b] Laura Montalvillo-JimØnez,[a]
Francisco Corzana,[c] Agatha Bastida,[a] AndrØs G. Santana,*[a] Carlos Gonzµlez,*[b] and
Juan Luis Asensio*[a]
In memory of Enrique Pedroso, our dear colleague and friend
Abstract: Targeting the interface between DNA quadruplex
and duplex regions by small molecules holds significant
promise in both therapeutics and nanotechnology. Herein, a
new pharmacophore is reported, which selectively binds
with high affinity to quadruplex–duplex junctions, while pre-
senting a poorer affinity for G-quadruplex or duplex DNA
alone. Ligands complying with the reported pharmacophore
been determined by NMR methods. According to these
data, the remarkable selectivity of this structural motif for
quadruplex–duplex junctions is achieved through an unpre-
cedented interaction mode so far unexploited in medicinal
and biological chemistry: the insertion of a benzylic ammo-
nium moiety into the centre of the partially exposed G-
tetrad at the interface with the duplex. Further decoration of
exhibit a significant affinity and selectivity for quadruplex– the described scaffolds with additional fragments opens up
duplex junctions, including the one observed in the HIV-1
LTR-III sequence. The structure of the complex between a
quadruplex–duplex junction with a ligand of this family has
the road to the development of selective ligands for G-quad-
ruplex-forming regions of the genome.
tive targets for the development of new drugs[7,8] for a number
of pathologies such as cancer,[9] infective[10] and neurodegener-
ative diseases.[11] Unfortunately, compounds that bind G-quad-
ruplexes with high affinity do not usually have a high selectivi-
ty for a particular topology and several approaches have been
proposed to tackle this problem.[12] Among them, a promising
strategy is targeting the interface formed by quadruplexes
with adjacent double-stranded regions.[13] As DNA is mainly a
B-form double helix in the cell, local formation of non-canoni-
cal structures entails the formation of interphases or junctions
between DNA regions with different secondary structures. Ac-
cording to this reasoning, quadruplex–duplex junctions (QDJs)
must be common in the genome and a relevant example has
been identified in potential pharmacological targets, such as
the viral HIV LTR-III sequence.[14] Moreover, in recent years, sev-
eral structures of quadruplex–duplex junctions have been de-
termined by NMR spectroscopy[14–17] and X-ray crystallogra-
phy.[13a] Disappointingly, little progress has been made in the
design of selective binders of these or other quadruplex–
duplex junctions. Thus, molecular modelling studies on the po-
tential binding modes of several G-quadruplex ligands have
been described by Parkinson et al., although no experimental
evidence of selective interaction was documented.[13a] While
preparing the submission of this manuscript, the tight associa-
tion of indoloquinoline-derived ligands to particular junction
structures was reported by Weisz et al.[17] However, the struc-
tural basis for this interaction is presently unknown. This pio-
Introduction
G-quadruplexes are non-canonical DNA structures resulting
from the stacking of Hoogsteen paired G-tetrads.[1] These
motifs are known to be present in telomeric[2,3] and promoter
regions[4] of the eukaryotic genome, controlling its stability and
playing key roles in transcriptional regulation.[5,6] Indeed,
during the past decades, G-quadruplexes have become attrac-
[a] L. Díaz-Casado,+ Dr. L. Montalvillo-JimØnez, A. Bastida, Dr. A. G. Santana,
Dr. J. L. Asensio
Glycochemistry and Molecular Recognition group— Dpt. Bio-Organic
Chemistry
Instituto de Química Orgµnica General (IQOG-CSIC)
Juan de la Cierva 3. 28006 Madrid (Spain)
E-mail: andres.g.santana@csic.es
juanluis.asensio@csic.es
[b] I. Serrano-Chacón,+ Prof. Dr. C. Gonzµlez
Instituto de Química-Física Rocasolano (IQFR-CSIC)
Serrano 119. 28006 Madrid (Spain)
E-mail: cgonzalez@iqfr.csic.es
[c] F. Corzana
Department of Chemistry
Centro de Investigación en Síntesis Química
Universidad de La Rioja
Madre de Dios, 53. 26006 LogroÇo (Spain)
[+] These authors contributed equally to this work.
Supporting information and the ORCID identification number(s) for the
author(s) of this article can be found under:
https://doi.org/10.1002/chem.202005026.
Chem. Eur. J. 2021, 27, 6204 –6212
6204
ꢀ 2020 Wiley-VCH GmbH

Full Paper
doi.org/10.1002/chem.202005026
Chemistry—A European Journal
neering contribution constitutes the first and only experimen-
tal study on this topic reported so far.
To address this relevant issue, we have employed a multidis-
ciplinary approach that combines NMR spectroscopy with mo-
lecular modelling and binding studies. Our design strategy was
based on the observation that the centre of the G-tetrad at
the interface with the duplex, a region characterised by a
strongly negative electrostatic potential, is often partially ac-
cessible in junction architectures. We hypothesised that this
motif, whose binding properties should be strongly modulated
by the adjacent duplex extensions, could represent a hitherto
unexplored hot-spot for the selective recognition of quadru-
plex–duplex junctions. This idea has been examined by em-
ploying all the quadruplex–duplex junction structures reported
in the bibliography. Among them, some of the architectures
described by Phan et al. were selected as the main model
system for our study (QDJ1–QDJ5).[15a]
They comprise a two-G-tetrad antiparallel quadruplex con-
nected with a long stem-loop hairpin, as a duplex extension,
leading to a partially exposed interfacial G-tetrad suitable for
ligand targeting. The G-quadruplex moiety is structurally very
similar to the well-known thrombin binding aptamer (TBA)[18]
in which the three-residue axial loop is replaced by a longer
stem-loop hairpin.[19] We further validated our results with
other junctions available in the literature, including a potential
pharmacological target: the HIV-1 LTR-III sequence.[14,20–23]
Figure 1. Representative ensembles obtained for QDJ1 (a) and QDJ5 (b) by
employing MD simulations. Average structures are shown in grey. Regions
characterised by a large density of potassium cations, as revealed by the
analysis of the MD trajectories with Chimera, are highlighted in green.
(c) Representation of the quadruplex–duplex junction structure in QDJ1. The
centre of the interfacial G-tetrad is highlighted with a yellow shading, where
guanine O6 atoms are represented as red spheres. (d) Schematic representa-
tion of the proposed binding mode for a benzylamine-like pharmacophore
targeting the G4 spot together with examples of putative ligands.
Results and Discussion
Cation-binding properties of quadruplex–duplex interfaces
and design of a selective pharmacophore
considerations helped to conceive a general pharmacophore
represented by chemically accessible highly versatile frame-
works integrating an electron-rich aromatic unit attached to a
methylene-amine moiety (Figure 1d). Junction recognition
would benefit from two complementary modes of interaction:
stacking of the aromatic platform onto the p-deficient surface
of the G-tetrad at the interface with the duplex plus insertion
of the ammonium moiety into the G4 centre, referred to as the
G4 hot-spot throughout this manuscript (Figure 1d). To test
this idea, we performed binding experiments with bare ver-
sions of the proposed pharmacophore as well as more com-
plex synthetic derivatives thereof.
In contrast with occluded or fully exposed G-tetrads character-
istic of common quadruplex regions, those located at the inter-
face between quadruplex and duplex fragments might display
distinctive electrostatic properties, and thus are potential hot-
spots for the molecular recognition of properly designed cat-
ionic ligands. To test this idea, we performed molecular dy-
namics simulations employing two different QDJ structures re-
ported by Phan et al. and herein referred to as QDJ1 and QDJ5
(Figure 1a and b).[15a] The relative orientation of the duplex
and quadruplex moieties is co-axial in QDJ1, which determines
the presence of an interfacial tetrad partially occluded by the
duplex extension. On the contrary, this motif is absent in QDJ5.
Long MD trajectories collected in the presence of potassium
ions revealed distinct electrostatic properties in both junctions.
Whereas QDJ1 displays a high density of cations at the inter-
face between the quadruplex and the duplex stem, a low den-
sity of ions is observed in QDJ5, either at the terminal tetrads
or duplex base pairs. This observation supports the notion that
the electrostatic properties of terminal tetrads can be affected
by co-axial duplex extensions, generating a singular recogni-
tion site. Targeting these epitopes would require the displace-
ment of the bound potassium, which poses an energy cost on
the recognition process. However, this could be compensated
by additional stacking and hydrogen bonding interactions,
leading to high-affinity ligands. Thus, geometrical and chemical
Binding experiments with model QDJs
As a proof of concept, we assayed first the association of the
simplest fragment that incorporates the proposed pharmaco-
phore, that is, benzylamine 1, to the oligonucleotide QDJ1. We
employed NMR spectroscopy, as this technique is especially
suitable for the detection and characterisation of low-affinity
binding processes (Figure 2a). Thus, according to titration ex-
periments performed at 278 K, 1 induced clear chemical shift
perturbations in the DNA imino region. Satisfactorily, these are
consistent with the selective recognition of the quadruplex–
duplex interface, with major Dd values located at the junction
duplex base pair G21/C7 and C20/G8. The change in Dd values
Chem. Eur. J. 2021, 27, 6204 –6212
www.chemeurj.org
6205
ꢀ 2020 Wiley-VCH GmbH

Full Paper
doi.org/10.1002/chem.202005026
Chemistry—A European Journal
abolishes the association, thus proving the need for at least
+
one primary NH3 moiety to effectively interact with the G4
hot-spot. According to this trend, guanidinylated derivatives
8–9 were synthesised, showing a similar affinity pattern (see
Figure 2b and Figures S3–S4 in the Supporting Information).
Finally, the parallel stacking between complementary p-rich
and p-deficient aromatic surfaces proved to be an important
contributor to the free energy of binding. Indeed, the obtained
ITC data showed that increasing the p-rich aromatic surface of
the ligand enhances the complex stability. Thus, com-
pounds 10 and 11, with just one or two positive charges, re-
spectively, display binding affinities comparable to that mea-
sured for 4 (Figure 2b and Figure S5 in the Supporting Infor-
mation). Further extension of the aromatic platform with addi-
tional fused phenyl rings, as in 12–14, renders these deriva-
tives too hydrophobic for microcalorimetry experiments.
However, CD melting experiments confirmed these derivatives
bind to QDJ1 with much higher affinities than 4 (CD experi-
ments with high-affinity ligands are shown in Figure 2c. Data
sets measured for 4 are also shown for comparison purposes).
Figure 2. (a) Low-affinity scaffolds. Kb values for the association of 1–3 to
QDJ1 at 278 K derived from NMR spectroscopy. Junction residues (CPK) and
its electrostatic surface (only bases considered) are shown. Binding curve for
1 together with 1D NMR data from the corresponding titration are also rep-
resented. (b) Medium-affinity ligands. Binding data derived for the associa-
tion of selected aromatic ligands incorporating benzylamine-like fragments
to QDJ1. Kb (mÀ1) values for compounds 4–11 were obtained from ITC
(298 K) experiments (ITC data measured with 4 is shown on the left).
(c) High-affinity ligands and preliminary selectivity assays. Typical CD thermal
denaturation curves measured for quadruplex TBA (orange; left panels),
QDJ1(red; middle panels) and the duplex fragment present in QDJ1 (violet;
right panels) in the absence and presence of ligands 4 and 12–14 (from top
to bottom). The observed shifts in melting temperatures are highlighted
with a coloured shading and represented on the right.
Selectivity for QDJs versus isolated quadruplex or duplex
fragments
Chemical shift perturbations observed in NMR titration experi-
ments performed with compound 4 and QDJ1 are consistent
with the association of this ligand to the junction region (Fig-
ure S2 in the Supporting Information). Similar effects are ob-
served for other compounds of the family (Figures S3–S4 in
the Supporting Information). To assess the selectivity of the
molecular recognition process, the interaction of the proposed
pharmacophore with oligonucleotides containing the constitu-
ent duplex or quadruplex moieties present in QDJ1 was ex-
plored by different techniques. The stem-loop hairpin se-
along the titration experiments confirmed that benzylamine (1)
binds to oligo QDJ1 with a Kb =16002mÀ1, corresponding to a
DG value of À5.3 kcalmolÀ1 at 278 K, a significant free energy
of interaction for such a limited set of contacts (values at 298 K
are also shown in Figure 2a). Significantly, homologation or
shortening of the aromatic/ammonium linker (as in phenethyl-
amine 2 or aniline 3) translates into a dramatic decrease in
binding affinity, ranging from 19-fold for the case of 2 to a vir-
tually undetectable association for 3 (Figure 2a and Figure S1
in the Supporting Information). These preliminary observations
are fully consistent with the proposed binding mode and en-
couraged us to pursue this line of research.
5’
quence CGCGAAGCATTCGCG3’ was used as a model for the
duplex. Similarly, the TBA sequence (5’GGTTGGTGTGGTTGG3’)
was used as a model of the quadruplex fragment considering
its structural similarity with the one present in QDJ1. Most im-
portantly, melting curves acquired with these quadruplex and
duplex fragments confirmed an overwhelming preference for
the junction structure. Additional evaluation of the selectivity
provided by this recognition motif was derived from micro-di-
alysis experiments (Figures S6–S7 in the Supporting Informa-
tion, and see the Experimental Section).[24] The results obtained
demonstrate that 4 recognises the quadruplex–duplex inter-
face with exquisite selectivity with respect to the constituent
duplex and quadruplex fragments alone. Indeed, numerical sim-
ulations of this experiment indicate that Kb values for the
duplex or quadruplex fragments must be <20000mÀ1, at least
a 40-fold decrease in affinity with respect to QDJ1 (see Figur-
es S6–S7 in the Supporting Information). This result is further
supported by NMR titrations of the isolated quadruplex and
duplex moieties, which revealed no significant association
even at much higher concentrations of 4 (see Figure S8 in the
Supporting Information).
Next, and as a straightforward strategy to increase the stabil-
ity of the complex, we prepared a multivalent derivative
equipped with three alternating methylene-amine fragments
attached to a single phenyl ring (compound 4). NMR titration
assays (Figure S2 in the Supporting Information) confirmed the
enhanced affinity of this scaffold, which prompted us to
employ microcalorimetry to fully dissect the association pro-
cess (Figure 2b). According to these data sets QDJ1 recogni-
tion by 4 is exothermic and characterised by a Kb value in the
106 mÀ1 range. Concomitant with our rationale, N-methylation
and N,N’-dimethylation of 4 (compounds 5 and 6) preserves
the ligand’s capacity to selectively recognise the junction struc-
ture. However, the introduction of a third N’’-Me group (7)
The weak to null affinity for the TBA fragment exhibited by
ligands 13 and 14 is intriguing as derivatives with similarly
Chem. Eur. J. 2021, 27, 6204 –6212
www.chemeurj.org
6206
ꢀ 2020 Wiley-VCH GmbH

Full Paper
doi.org/10.1002/chem.202005026
Chemistry—A European Journal
large aromatic surfaces have been reported as reasonable
quadruplex binders.[9] Taking this into account, we decided to
extend our selectivity studies with these derivatives to other
quadruplex topologies, including both parallel and antiparallel
arrangements such as the proto-oncogene promoter KRAS,[25]
the human Bcl2 promoter region,[26] a natural human telomere
quadruplex[27] and the T30695 structure described by Hogan
et al.[28] In Figure 3, it can be observed that ligand 13 promotes
moderate stabilisations of several of the DNA fragments
tested, with DT values up to 7.7 K in the most favourable situa-
tion. However, a clear preference for the QDJ1 structure is still
apparent, providing further support for the junction-selective
character of this derivative.
(50 mm) fragment was titrated with 0.5 equivalents of 13 to
generate a mixture of free and complexed junctions (50%
each) with well-resolved NMR signals in the 12–14 ppm spec-
tral region. Next, different quadruplex fragments (100 mm each)
were subsequently added to the NMR sample and their influ-
ence on the QDJ1 free/bound equilibrium was determined
through integration of the appropriate peaks. The obtained re-
sults demonstrate that these quadruplexes are unable to com-
pete with QDJ1 for 13, having a negligible impact on the frac-
tion of complexed junction (an observation that is maintained
after a 24 h equilibration). Notably, addition of the QDJ1 frag-
ment onto a KRAS/13 complex mixture immediately showed
migration of the ligand from the quadruplex toward the junc-
tion structure (Figure S9 in the Supporting Information). Alto-
gether the obtained results confirmed the significant prefer-
ence of 13 for quadruplex–duplex junctions, regardless of the
topology of the competing quadruplex fragments considered.
Regarding the more extended compound 14, this time sig-
nificant stabilisations of most of the analysed quadruplexes, in
the 15–17 K range, were detected. Then again, the observed
stabilisation of the QDJ1 fragment is higher (>19 K), a trend
that is further accentuated with the HIV LTR-III junction (see
below). Overall, the obtained results suggest that three fused
aromatic units (such as those present in 13) represent the ap-
propriate size to provide optimal junction selectivity with re-
spect to plain quadruplexes.
Structure determination of a QDJ/ligand complex
To get further insight into the interaction mode of the ana-
lysed scaffolds with DNA QDJs, we resorted to NMR methods.
Complete assignment of the NMR spectra of QDJ1/4 and
QDJ1/13 complexes was carried out (Figures S10–S12; assign-
ment Tables S1 and S2 are shown in Supporting Information).
Chemical shift perturbations promoted by both ligands on the
DNA receptor were similar (Figure S13 in the Supporting Infor-
mation) and are limited to the residues in the quadruplex–
duplex junction, consistently with a unique binding mode in
this region. Especially illustrative is the strong change of chem-
ical shifts of C20 base protons upon ligand binding (Figure 4a).
Although melting studies provide an indication of the rela-
tive complex stabilities, they do not constitute quantitative af-
finity measurements by themselves (which are otherwise diffi-
cult to perform with high-affinity, low-solubility compounds
like 13 and 14). Therefore, we carried out competition experi-
ments by NMR spectroscopy (Figure 3). Accordingly, the QDJ1
Figure 4. (a) NMR spectra acquired in potassium phosphate buffer, 20 mm
KCl at pH 6.9. DNA and ligand concentrations were 500 and 600 mm, respec-
tively. Left: Cytosine H5/H6 region in TOCSY experiments (298 K) measured
with QDJ1 free (red) and complexed to 13 (black). Chemical shift perturba-
tions detected at residue C20 have been emphasised. Right: Two different
sections of a NOESY (278 K) experiment measured for the QDJ1/13 complex
are shown. NOE cross-peaks involving the ligand key ammonium group are
represented. (b) NMR structure of the QDJ1/13 complex (PDB 6FC9). Ob-
served ligand/DNA interactions are highlighted on the top-right corner. A
CPK representation of the junction region (top view) is shown on the
bottom-right corner. (c) 1H-15N HSQC experiments for the QDJ1 complexes
with 15N-labelled derivatives 11 and 14 (see the main text).
Figure 3. Selectivity assays. Left: Melting temperature perturbations mea-
sured for different quadruplex fragments (5 mm, orange) and QDJ1 (5 mm,
red) in the presence of two equivalents of ligands 13 (left) or 14 (right) as
determined from CD thermal denaturation measurements. Structures for se-
lected quadruplexes are shown above. The respective PDB codes are indicat-
ed. Right: NMR competition experiments. QDJ1 (50 mm) was titrated with
0.5 equivalents of 13 to generate a mixture of free and complexed junctions
(50% each) and different competing quadruplexes (100 mm each) were sub-
sequently added. Free/bound equilibrium for the junction fragment was
evaluated by integration of the appropriate NMR signals (yellow-shaded
area).
Chem. Eur. J. 2021, 27, 6204 –6212
www.chemeurj.org
6207
ꢀ 2020 Wiley-VCH GmbH

Full Paper
doi.org/10.1002/chem.202005026
Chemistry—A European Journal
Unfortunately, ligand signals in QDJ1/4 complex could not be
identified. This is probably a consequence of an intermediate
exchange rate between the three equivalent poses in which a
three-fold symmetric compound can be accommodated in the
binding pocket. However, signals of compound 13 were clearly
identified in the NMR spectra of the QDJ1/13 complex, for
which more than 20 intermolecular cross-peaks were observed
(Table S3 in the Supporting Information). The solution structure
of this complex was determined by restrained molecular dy-
namics methods on the basis of 232 experimental distance
constraints. The structure is well-defined with an RMSD for the
heavy atoms of 0.6 Š in less mobile regions. Complex coordi-
nates were deposited in the PDB (code 6FC9). Statistical data
are summarised in Table S4 and the conformational ensemble
shown in Figure S14 (in the Supporting Information).
these assays is that fast exchange with solvent usually pre-
cludes detection of ammonium groups in water solution
unless this process is strongly slowed down by tight binding
interactions with the DNA receptor. Indeed, this behaviour
would be expected for the proposed complexes, in which the
ammonium moiety is inserted into the centre of the QDJ inter-
facial G-tetrad, establishing numerous hydrogen bonding and
electrostatic interactions. On the contrary, alternative, non-spe-
cific binding modes might prove less efficient at slowing down
the ammonium exchange, rendering the corresponding HSQC
cross-peaks undetectable. Control assays performed with 15N-
labelled 11 and 14 either free or in the presence of duplex or
quadruplex fragments were consistent with this view, thus fail-
ing to provide detectable cross-peaks. In contrast, clear HSQC
signals were apparent for the QDJ1 complexes. Moreover,
1
The obtained DNA structure is very similar to that of the
free oligonucleotide, being the complex stabilised by two key
interactions involving the interfacial tetrad. Although the
ligand aromatic moiety stacks on the exposed guanines of the
G-tetrad, one of the ligand ammonium groups points towards
the tetrad central position (the proposed G4 hot-spot), inter-
acting with the electron-rich guanine O6 groups (Figure 4b
and Figure S15 in the Supporting Information). This interaction
is firmly supported by a number of NOEs between the ammo-
nium signal at 7.6 ppm with imino protons of G1, G6, G22 and
G27 (Figure 4a). The fact that this ammonium signal is unusu-
ally narrow and that it is observed even at relatively high tem-
perature (298 K) indicates that this functional group is not ex-
posed to the solvent. Ligand contacts with the duplex region
also contribute to the stability of the complex, in particular a
partial stacking with C20 and a salt bridge with the phosphate
of C20 (see Figure 4b). These structural data are fully consis-
tent with the binding assays previously described with alterna-
tive ligands (a model of the QDJ1/4 complex generated by
docking plus molecular dynamics simulations, also consistent
with all the experimental information herein reported as
shown in Figure S16 in the Supporting Information).
these presented downfield shifted H and 15N ammonium reso-
nances in agreement with values reported by Plavec and co-
workers for chelated intertetrad ammonium ions.[29] Interest-
ingly, in the particular case of 14, the obtained spectrum re-
vealed the co-existence of a mixture of two complexes, which
probably reflects the non-symmetrical nature of this ligand.
Similar experiments performed with the weaker binder 4 (and
derivatives) failed to detect the desired signals presumably ex-
tremely broadened by the free/bound exchange process. How-
ever, for tight binders such as 11 and 14 this simple assay pro-
vides a signature of the ammonium/G4 spot association
shared for all the analysed ligands.
Binding to different QDJs
Next, we analysed the structural/topological requirements for
the recognition process to happen. To this end, binding studies
were performed with 4 and different junction architectures.
First, we explored the alternative structures described by Phan
et al. (herein referred to as QDJ2–QDJ5; see Figure 5a and Fig-
ures S17–S20 in the Supporting Information),[15a] employing
both NMR and ITC. As can be observed, oligos QDJ1–QDJ4 dis-
play a similar coaxial orientation of quadruplex and duplex do-
mains (Figure 5). However, although QDJ1 and QDJ2 exhibit a
partially exposed G4 spot, this site is occluded in QDJ3 and
QDJ4. Consistently, the latter provided no detectable associa-
To assess the validity of this general binding mode for other
junction ligands, selected compounds were labelled with 15N
and 1H/15N HSQC experiments were recorded on the corre-
sponding DNA complexes (Figure 4c). The rationale behind
Figure 5. (a) Binding studies (298 K) performed by ITC with ligand 4 and oligos QDJ2–QDJ5 (sequences shown above). Overall topology (up, ribbon) and junc-
tion residues (down, CPK) are represented. G4 spot oxygen atoms are coloured in yellow (structures are ordered according to the increasing degree of G4
spot exposure). (b) Kb (mÀ1, 298 K) values obtained by ITC with ligand 4 and mutated QDJ1 variants. Substituted residues at the interfacial and adjacent
duplex base pairs (I/II and III/IV, respectively) are indicated. ITC data measured with a selected variant, bearing a TTT overhang at the 5’ end is represented on
the left.
Chem. Eur. J. 2021, 27, 6204 –6212
www.chemeurj.org
6208
ꢀ 2020 Wiley-VCH GmbH

Full Paper
doi.org/10.1002/chem.202005026
Chemistry—A European Journal
tion. In contrast, QDJ1 and QDJ2 displayed comparable stabili-
ties. Regarding QDJ5, as mentioned earlier, this oligonucleotide
exhibits a radically different topology, characterised by an or-
thogonal orientation between the duplex and the quadruplex
domains. As a result, the terminal G-tetrad is exposed with its
central cavity virtually unaffected by the duplex regions (see
Figures 1b, 5a and S20 in the Supporting Information). Inter-
estingly, in this particular case no association to the fully ex-
posed G4 hot-spot was detected either.
namic duplex base pair at the boundary between the quadru-
plex and duplex fragments (Figure 6a).[14] Close inspection at
the reported structure (pdb code 6H1K) revealed that this fea-
ture leads to a partially accessible interfacial G4 spot potential-
ly targetable by the ligands, complying with the pharmaco-
phore herein reported. Satisfactorily, NMR titration experiments
clearly show that ligand 4 indeed binds to the HIV LTR-III se-
quence with significant affinity to form a 1:1 complex (Fig-
ure 6b). No evidence of additional binding events was ob-
served even at high ligand/DNA ratios, which indicates that
the ongoing association process is highly selective. Chemical
shift changes upon complex formation were analysed on the
basis of the assignments previously reported for the free
DNA[14] and confirmed that protons located in the interfacial
quadruplex/duplex region are the most affected by the ligand.
In particular, changes in chemical shifts of base protons of C13
and T14 and imino protons of G15 and G26 are clearly ob-
served, whereas other regions are mainly unaffected (Fig-
ure 6b,c). To demonstrate the relevance of the amino groups
in the pharmacophore for QDJ6 binding, we carried out addi-
tional NMR titration experiments employing ligand 7. As previ-
ously observed for model junction architectures QDJ1–QDJ5,
N-methylation of the three amino moieties totally abolishes
the capacity of the compound to recognise the interfacial G4
spot motif. Indeed, no association of 7 to QDJ6 was detected
even at high ligand/DNA ratios. Finally, the thermodynamic
In conclusion, neither an occluded (QDJ3-4) nor a totally ex-
posed (QDJ5) G4 spot allows complex formation at the junc-
tion binding-site with a benzylamine-like probe. Enhancement
of the G4 hot-spot binding properties by adjacent co-axial
duplex regions seems to be an essential requirement for tight
association, which provides significant opportunities for the
design of selective binders. These results have been confirmed
with other junctions available in the literature. Thus, binding of
4 to the junction studied by Karg et al.,[16] which exhibits a par-
tially occluded G4 hot-spot, can be observed by NMR titration
experiments (Figure S22 in the Supporting Information) where-
as no binding is apparent with the junction studied by Russo
et al.,[13a] the crystallographic structure of which shows the
centre of the QDJ interfacial G-tetrad buried by a TAT triad
(Figure S22 in the Supporting Information).
Following this line of thought, interfacial duplex base pairs
should strongly modulate the association properties of this
family of ligands to the G4 spot. Regarding the QDJ1 architec-
ture, we have observed that swapping the G21/C7 base pair to
C21/G7 or its replacement by A21/T7 led to a >40-fold de-
crease in binding affinity (Figure 5b and Figure S21 in the Sup-
porting Information). Remarkably, the modulating influence of
the contiguous duplex base pair (C20/G8) seems significant
too, with Kb values in the 107 mÀ1 range for the G20/C8 mutant
(see Figure 5b and Figure S21 in the Supporting Information).
Finally, the extension of the 5’-end (located in this case beside
the QDJ interfacial G-tetrad) with a TTT overhang promotes a
3–4-fold increase in the stability of the complex (Kb =
2980000mÀ1), probably by providing additional harnessing
points for the ligand (Figure 5b and Figure S21 in the Support-
ing Information). Overall, these binding experiments point to a
tighter interaction at the junction hot-spot especially when the
adjacent duplex base pairs contain G-C.
Binding to biologically relevant sequences: The case of
HIV-1 LTR-III
Figure 6. (a) 3D NMR structure (PDB code: 6H1K) of HIV oligonucleotide
QDJ6 (ribbon representation). Non-interfacial duplex and quadruplex regions
are represented in cyan and orange, respectively. Similarly, interfacial duplex
base pair and G-tetrad are shown in blue and red. (b) QDJ6/4 titration ex-
periment. [QDJ6]=0.1 mm, T=298 K. (c) Top: Cytosine H5-H6 region in
TOCSY experiments acquired for isolated QDJ6 (red) and 1:1 QDJ6/4 com-
plex (black). Bottom: NOESY spectra (150 ms mixing time) of isolated QDJ6
(red) and 1:1 QDJ6/4 complex (black). (d) ITC profile for the titration of 4
into a solution containing QDJ6 at 298 K. (e) 1D NMR spectra for the QDJ6/
Encouraged by these results, we decided to further test the
potential of the proposed pharmacophore for selective recog-
nition of quadruplex–duplex junction architectures found in
natural sequences. We focused our attention on a recently de-
scribed target of great biomedical relevance: the LTR-III region
of the HIV-1 virus (herein referred to as QDJ6).[14] Stabilisation
of this site performed by small ligands has been proposed as a
promising strategy for the inhibition of the viral transcription.
Interestingly, NMR structural studies have shown that it com-
prises a peculiar quadruplex–duplex junction, which, in con-
trast to previously employed models, incorporates a highly dy-
1
13 and QDJ6/14 complexes (imino region) together with a H/15N-HSQC
spectrum for the latter acquired with a 15N-labelled ligand. (f) Melting tem-
perature perturbations deduced for QDJ6 (5 mm) in the presence of 10 mm of
ligands 4, 13 or 14 as deduced from CD thermal denaturation measure-
ments.
Chem. Eur. J. 2021, 27, 6204 –6212
www.chemeurj.org
6209
ꢀ 2020 Wiley-VCH GmbH

Full Paper
doi.org/10.1002/chem.202005026
Chemistry—A European Journal
this purpose, they were placed in a water bath at 293 K, heated to
358 K for 5 min and then slowly cooled back to 293 K over a 2 h
period.
features of the QDJ6/4 complex were dissected by microca-
lorimetry experiments (Figure 6d). According to these data, 4
binds to QDJ6 with a significant affinity (Kb =6.68”105 mÀ1),
corresponding to a free energy of association DG=À7.91 kcal
molÀ1, almost identical to that previously observed for the
QDJ1/4 interaction. However, in contrast with the latter, the
binding event between 4 and QDJ6 is mainly entropy-driven,
whereas it is only slightly exothermic (DH=À2.08 kcalmolÀ1),
which might reflect the distinct dynamic behaviour exhibited
by interfacial duplex base pairs in both QDJ6 and QDJ1.
Microcalorimetry (ITC) binding experiments
Binding studies were performed at 298 K, in 20 mm KCl, 20 mm
K2HPO4 at pH 6.9, by using a VP-ITC titration calorimeter (MicroCal,
LLC) with a reaction cell volume of 1.467 mL. Typically, 10 mm DNA
oligonucleotide in the reaction cell was titrated with a 500–700 mm
solution of the different ligands contained in a 300 mL syringe. At
least 30 consecutive injections of 5 mL were applied at 5 min inter-
vals while the DNA solution was stirred at a constant speed of
300 rpm. Dilution heats of ligand into DNA solutions (which
agreed with those obtained by injections of ligands into the same
volume of buffer) were subtracted from measured heats of bind-
ing. Titration curves were analysed with Origin, provided with the
instrument by MicroCal LLC. For every single DNA fragment, ther-
modynamic parameters were derived from two independent ex-
periments and averaged.
Larger aromatic ligands, such as 13 and 14, also bind to the
HIV-LTIII to form 1:1 complexes. The addition of sub-stoichio-
metric amounts of the ligands reveal that now the exchange
between free and bound DNA is slow on the chemical shift
timescale. In addition, the final complexes render high quality
spectra characterised by sharp and well-resolved signals in the
imino region (Figure 6e). Moreover, HSQC spectra with 15N-la-
belled ligand 14 exhibits a single cross-peak, consistent with
the insertion of the ammonium group in the centre of the G-
tetrad at the interface with the duplex (Figure 6e and Fig-
ure S23 in the Supporting Information). All these experimental
pieces of evidence strongly suggest that both 13 and 14 pres-
ent again increased affinities for the target DNA with respect
to 4. This impression was further confirmed by CD thermal de-
naturation experiments, which showed more significant incre-
ments in the receptor melting temperature upon formation of
the complexes (Figure 6 f). Indeed, the observed DT values are
larger than those previously measured for the model junction
QDJ1 (Figure 2b). In addition, they clearly surpass the thermal
stabilisations previously detected with quadruplex fragments
of alternative topologies (Figure 3), further supporting the
junction-selective character of this pharmacophore.
Nuclear magnetic resonance (NMR) binding experiments
The binding of ligands 1–4 to oligonucleotides QDJ1 and QDJ5
was monitored by recording 1D 600 MHz 1H NMR spectra of a
series of samples with variable ligand concentrations at 278 K. In
all cases, the ongoing association processes were apparent from
the chemical shift perturbations detected in the DNA imino region
(12–15 ppm) of the receptor. The DNA samples were prepared at
100 mm concentration (calculated from the UV absorbance at
260 nm) from stock solutions in 90% H2O:10% D2O, 20 mm KCl,
20 mm K2HPO4 at pH 6.9. The 1H NMR spectrum for the sample
with the highest ligand/DNA ratio was recorded by dissolving the
ligand (ꢀ20 mm) in 0.5 mL of the QDJ1-QDJ5 100 mm solutions. Ti-
tration curves were obtained by adding small aliquots of this high
ligand/DNA ratio sample to a ligand-free NMR sample. In all cases,
the data were found to fit well assuming a 1:1 complex.
Conclusion
Micro-dialysis binding experiments
We have described a novel binding-motif, which relies on the
recognition of the strongly negative electrostatic potential of
the G-tetrad centre. Although, this centre is usually occluded
in most G-quadruplex structures, we have observed that it is
partially exposed in some quadruplex–duplex junctions. By
taking advantage of this interaction hot-spot, we have de-
signed a family of compounds able to bind selectively to quad-
ruplex–duplex junctions, including biomedically relevant tar-
gets as the HIV LTR-III fragment. The benzylamine-like motifs
herein described are very convenient scaffolds, which can be
readily conjugated with other complementary duplex or quad-
ruplex binders, thus paving the way to the development of a
whole new class of quadruplex-discriminating ligands.
A solution (400 mL) containing oligonucleotide QDJ1 (50 mm) and
ligand 4 (20 mm) was placed in the central compartment (herein re-
ferred to as II) of a three-chamber micro-equilibrium dialyser (Har-
vard apparatus) equipped with 5 kDa cut-off membranes and dia-
lysed against the two separate DNA(50 mm)/ligand 4 (20 mm) mix-
tures: one containing TBA quadruplex (in chamber I) and the other
with the duplex fragment present in QDJ1 (in chamber III). To ach-
ieve a complete equilibration of the three micro-dialysis compart-
ments, this assay was left to proceed for 2 days at 298 K. After-
wards, solutions from each chamber were collected and the DNA
fragments were digested with DNAase (1–2 mm) for 6 h. Next,
ligand 4 present in the three samples was derivatised by treatment
with 13C-labelled formaldehyde (20 mm) and sodium cyanoborohy-
dride (20 mm). After 12 h at room temperature, the solutions were
loaded into NMR tubes and HSQC experiments were acquired. Inte-
gration of the single N,N-dimethyl cross-peak present in these
spectra provided a simple means to determine the relative concen-
tration of 4 in each chamber after equilibration.[24]
Experimental Section
DNA oligonucleotides were purchased in purified desalted form
from IDT (Integrated DNA technologies) and dialysed against
20 mm KCl, 20 mm K2HPO4 at pH 6.9 buffer before binding experi-
ments. To ensure a proper folding of the different quadruplex,
duplex or quadruplex–duplex junction structures, the correspond-
ing oligonucleotides were subjected to an annealing protocol. For
Micro-dialysis theoretical simulations
We performed extensive modelling studies on the micro-dialysis
competition experiment described in the manuscript, employing
the biochemical kinetic simulator GEPASI 21.[30] Our model com-
Chem. Eur. J. 2021, 27, 6204 –6212
www.chemeurj.org
6210
ꢀ 2020 Wiley-VCH GmbH

Full Paper
doi.org/10.1002/chem.202005026
Chemistry—A European Journal
prised a solution, containing the quadruplex–duplex junction oli-
gonucleotide, placed in the central chamber (herein referred to as
II) of a three-compartment micro-dialysis device. Two alternative
DNA solutions (one containing TBA quadruplex and the other with
the duplex fragment present in QDJ1) are confined in lateral cham-
bers I and III. Chamber volumes and DNA concentrations were
fixed at 400 mL and 100 mm, respectively, in agreement with the
actual conditions employed in our experimental assay. The ligand
was assumed to be initially present in the three chambers at iden-
tical concentrations (20 mm). Numeric integration of the corre-
sponding kinetic equations allowed a theoretical evaluation of the
equilibrium ligand concentrations in chambers I–III by assuming a
variety of binding constants for the duplex and quadruplex frag-
ments (Kbduplex and Kbquadruplex, respectively. For QDJ, KbQDJ1 was fixed
as 106 mÀ1).
mental distance constraints was used to determine the three-di-
mensional structure by using restrained molecular dynamics meth-
ods.
Structures were calculated with the SANDER module of the molec-
ular dynamics package AMBER 12.[33] The coordinates of the free
DNA structures (PDB 2M8Z)[15a] were taken as starting points. First,
a 3D model of compound 13 was built with the program Sybyl
and docked manually into the duplex–quadruplex interface. Then,
the docked structures were taken as starting coordinates for the
AMBER refinement, consisting of an annealing protocol in vacuo,
followed by trajectories of 1 ns each in which explicit solvent mole-
cules were included. The Particle Mesh Ewald method was used to
evaluate long-range electrostatic interactions.[34] The specific proto-
cols for these calculations have been previously described.[35] The
BSC1 force field was used to describe the DNA,[36] and the TIP3P
model[37] was used to simulate the water molecules. Analysis of the
representative structures was carried out with the program
MOLMOL.[38]
CD melting experiment
Thermal unfolding transitions were measured by employing 1 cm
path length quartz cells with samples containing 5.0 mm DNA in
20 mm KCl, 20 mm K2HPO4 at pH 6.9. Junction-selective ligands 4
and 12–14 were employed at 10 mm concentration. The transitions
were monitored by the decrease of the CD signal at 250 nm
(duplex), 295 nm (QDJ1, HIV LTR-III quadruplex–duplex junctions
and TBA, BCL2 and telomeric quadruplexes) or 265 nm (KRAS
quadruplex) by using 2 nm band width. Heating rates were
20 KhÀ1. Transitions were evaluated by using a nonlinear least
squares fit assuming a two state model with pre- and post-transi-
tional baselines sloping.
Molecular dynamics simulations
The conformational and dynamical properties of selected junctions
were tested through unconstrained Molecular Dynamics (MD) sim-
ulations performed with the AMBER 12 package,[33] employing the
BSC1 force field.[36] MD trajectories, 1 ms long, were collected in the
presence of explicit TIP3P water,[37] periodic boundary conditions
and Ewald sums for the treatment of long-range electrostatic inter-
actions.[34] DNA charges were neutralised with potassium ions in-
cluding the presence of the well-known stabilising cations placed
between the G-tetrads. The time-step was 1 fs in all the simula-
tions. Finally, the distribution of potassium cations along the MD
trajectories was analysed by employing the program Chimera.[39]
NMR spectroscopy of the complexes
Samples for NMR experiments were dissolved (in Na+ form) in
either D2O or 9:1 H2O/D2O, 20 mm potassium phosphate buffer,
20 mm KCl at pH 6.9. Oligonucleotide concentration was 0.5 mm. Acknowledgements
Experiments were acquired with Bruker spectrometers operating at
600 and 800 MHz, equipped with cryoprobes and processed with
This investigation was supported by research grants from the
Ministerio de Ciencia, Innovación y Universidades (CTQ2016-
79255-P, PID2019-107476GB-I00, BFU2017-89707-P and
the TOPSPIN software. NOESY spectra were acquired with mixing
times ranging from 75 to 250 ms. TOCSY spectra were recorded
with the standard MLEV-17 spin-lock sequence and a mixing time
of 80 ms. In the experiments in H2O, water suppression was ach-
ieved by including a WATERGATE[31] module in the pulse sequence
prior to acquisition. The spectral analysis program SPARKY[32] was
used for semiautomatic assignment of the NOESY cross-peaks and
quantitative evaluation of the NOE intensities. The high similarity
between the NMR spectra of free QDJ1 and its complexes with 4
and 13 allowed for a straightforward assignment of the later on
the basis of the assignment published by Phan et al.[15a] Chemical
shift lists are given in Tables S1–S2 (in the Supporting Information).
RTI2018-099592-B-C21). A.G.S. acknowledges the Ministerio de
Ciencia, Innovación y Universidades for a Juan de la Cierva
contract. L.D.-C., I. S-R and L.M.-J. acknowledge the Ministerio
de Ciencia, Innovación y Universidades for FPI contracts. NMR
experiments were performed in the ‘‘Manuel Rico’’ NMR labora-
tory (LMR), a node of the Spanish Large-Scale National Facility
(ICTS R-LRB).
Conflict of interest
NMR constraints and structural calculations
The authors declare no conflict of interest.
Qualitative distance constraints were obtained from NOE intensi-
ties. NOEs were classified as strong, medium or weak, and distance
constraints were set accordingly to 3, 4 or 5 Š. In addition to these
experimentally derived constraints, hydrogen bond constrains for
Watson–Crick base pairs and G-tetrads were used. Target values for
distances and angles related to hydrogen bonds were set to values
obtained from crystallographic data in related structures. Owing to
the relatively broad line-widths of the sugar proton signals, J-cou-
pling constants were not accurately measured, but only roughly
estimated from DQF-COSY cross-peaks. Loose values were set for
the sugar dihedral angles d, n1 and n2 to constrain the deoxyri-
bose conformation to North or South domain. A set of 232 experi-
Keywords: NMR structures · nucleic acids ligands
pharmacophores · quadruplex–duplex junctions · selective
molecular recognition
·
[1] J. Y. Davis, Angew. Chem. Int. Ed. 2004, 43, 668–698; Angew. Chem.
2004, 116, 684–716.
[2] C. Schaffitzel, I. Berger, J. Postberg, J. Hanes, H. J. Lipps, A. Pluckthun,
Proc. Natl. Acad. Sci. USA 2001, 98, 8572–8577.
[3] G. Biffi, D. Tannahill, J. McCafferty, S. Balasubramanian, Nat. Chem. 2013,
5, 182–186.
Chem. Eur. J. 2021, 27, 6204 –6212
www.chemeurj.org
6211
ꢀ 2020 Wiley-VCH GmbH

Full Paper
doi.org/10.1002/chem.202005026
Chemistry—A European Journal
[4] S. Neidle, Nat. Rev. Chem. 2017, 1, 0041.
[21] D. Piekna-Przybylska, M. A. Sullivan, G. Sharma, R. A. Bambara, Biochem-
istry 2014, 53, 2581–2593.
[5] T. Tian, Y.-Q. Chen, S.-R. Wang, X. Zhou, Chem 2018, 4, 1314–1344.
[6] B. Onel, M. Carver, G. Wu, D. Timonina, S. Kalarn, M. Larriva, D. Yang, J.
Am. Chem. Soc. 2016, 138, 2563–2470.
[7] G. W. Collie, G. N. Parkinson, Chem. Soc. Rev. 2011, 40, 5867–5892.
[8] R. Hänsel-Hertsch, M. Di Antonio, S. Balasubramanian, Nat. Rev. Mol. Cell
Biol. 2017, 18, 279–284.
[9] It is important to note that when TBA is formed, two of the DNA bases
on the top loop (G8 and T9) actually stack strongly on the G4 tetrad.
This means that any G4 binder needs to displace those DNA bases to
stack on the tetrad with a quite big energetic cost. See: S. Balasubra-
manian, L. H. Hurley, S. Neidle, Nat. Rev. Drug. Discovery 2011, 10, 261–
275.
[22] K. W. Lim, P. Jenjaroenpun, Z. J. Low, Z. J. Khong, Y. S. Ng, V. A. Kuznet-
sov, A. T. Phan, Nucleic Acids Res. 2015, 43, 5630–5646.
[23] S. Amrane, A. Kerkour, A. Bedrat, B. Vialet, M. L. Andreola, J. L. Mergny, J.
Am. Chem. Soc. 2014, 136, 5249–5252.
[24] E. JimØnez-Moreno, L. Montalvillo-JimØnez, A. G. Santana, A. M. Gómez,
G. JimØnez-OsØs, F. Corzana, A. Bastida, J. JimØnez-Barbero, F. J. CaÇada,
I. Gómez-Pinto, C. Gonzµlez, J. L. Asensio, J. Am. Chem. Soc. 2016, 138,
6463–6474.
[25] A. Kerkour, J. Marquevielle, S. Ivashchenko, L. A. Yatsunyk, J. L. Mergny,
G. F. Salgado, J. Biol. Chem. 2017, 292, 8082–8091.
[26] J. Dai, D. Chen, R. A. Jones, L. H. Hurley, D. Yang, Nucleic Acids Res. 2006,
34, 5133–5144.
[10] L. M. Harris, C. J. Merrick, PloS Pathog. 2015, 11, e1004562.
[11] A. R. Haeusler, C. J. Donnelly, G. Periz, E. A. Simko, P. G. Shaw, M. S. Kim,
N. J. Maragakis, J. C. Troncoso, A. Pandey, R. Sattler, J. D. Rothstein, J.
Wang, Nature 2014, 507, 195–200.
[27] A. T. Phan, V. Kuryavyi, K. N. Luu, D. J. Patel, Nucleic Acids Res. 2007, 35,
6517–6525.
[28] N. Jing, R. F. Rando, Y. Pommier, M. E. Hogan, Biochemistry 1997, 36,
12498–12505.
[12] a) A. Gupta, L.-L. Lee, S. Roy, F. A. Tanious, W. D. Wilson, D. H. Ly, B. A. Ar-
mitage, ChemBioChem 2013, 14, 1476–1484; b) S. Roy, K. J. Zanotti, C. T.
Murphy, F. A. Tanious, W. D. Wilson, D. H. Ly, B. A. Armitage, Chem.
Commun. 2011, 47, 8524–8526; c) S. Lusvarghi, C. T. Murphy, S. Roy,
F. A. Tanious, I. Sacui, W. D. Wilson, D. H. Ly, B. A. Armitage, J. Am. Chem.
Soc. 2009, 131, 18415–18424; d) I. G. Panyutin, M. I. Onyshchenko, E. A.
Englund, D. H. Appella, R. D. Neumann, Curr. Pharm. Des. 2012, 18,
1984–1991; e) M. I. Onyshchenko, T. I. Gaynutdinov, E. A. Englund, D. H.
Appella, R. D. Neumann, I. G. Panyutin, Nucleic Acids Res. 2009, 37,
7570–7580; f) T. I. Gaynutdinov, R. D. Neumann, I. G. Panyutin, Nucleic
Acids Res. 2008, 36, 4079–4087; g) M. Tassinari, M. Zuffo, M. Nadai, V.
Pirota, A. C. Sevilla Montalvo, F. Doria, M. Freccero, S. N. Richter, Nucleic
Acids Res. 2020, 48, 4627–4642.
[29] P. Podbevsek, N. V. Hud, J. Plavec, Nucleic Acids Res. 2007, 35, 2554–
2563.
[30] P. Mendes, Comput. Applic. Biosci. 1993, 9, 563–571.
[31] M. Piotto, V. Saudek, V. Sklenµr, J. Biomol. NMR 1992, 2, 661–665.
[32] D. T. Goddard, D. G. Kneller, SPARKY, 3rd ed., University of California,
San Francisco.
[33] D. A. Case, T. A. Darden, T. E. Cheatham, C. L. Simmerling, J. Wang, R. E.
Duke, R. Luo, R. C. Walker, W. Zhang, K. M. Merz, B. Roberts, S. Hayik, A.
Roitberg, G. Seabra, J. Swails, A. W. Gçtz, I. Kolossvµry, K. F. Wong, F. Pae-
sani, J. Vanicek, R. M. Wolf, J. Liu, X. Wu, S. R. Brozell, T. Steinbrecher, H.
Gohlke, Q. Cai, X. Ye, J. Wang, M. J. Hsieh, G. Cui, D. R. Roe, D. H. Math-
ews, M. G. Seetin, R. Salomon-Ferrer, C. Sagui, V. Babin, T. Luchko, S. Gu-
sarov, A. Kovalenko, P. A. Kollman, 2012, AMBER12, University of Califor-
nia, San Francisco.
[34] T. Darden, D. York, L. Pedersen, Chem. Phys. 1993, 98, 10089–10092.
[35] R. Soliva, V. Monaco, I. Gómez-Pinto, N. J. Meeuwenoord, G. A. Marel,
J. H. Boom, C. Gonzµlez, M. Orozco, Nucleic Acids Res. 2001, 29, 2973–
2985.
[13] a) I. R. Russo Krauss, S. Ramaswamy, S. Neidle, S. Haider, G. N. Parkinson,
J. Am. Chem. Soc. 2016, 138, 1226–1233; b) T. Q. N. Nguyen, K. W. Lim,
A. T. Phan, Sci. Rep. 2017, 7, 11969.
[14] E. Butovskaya, B. Heddi, B. Bakalar, S. N. Richter, A. T. Phan, J. Am. Chem.
Soc. 2018, 140, 13654–13662.
[15] a) K. W. Lim, A. T. Phan, Angew. Chem. Int. Ed. 2013, 52, 8566–8569;
Angew. Chem. 2013, 125, 8728–8731; b) D. J. Y. Tan, F. R. Winnerdy, K. W.
Lim, A. T. Phan, Nucleic Acids Res. 2020, 48, 11162–11171; c) T. Q. N.
Nguyen, K. W. Lim, A. T. Phan, Nucleic Acids Res. 2020, 48, 10567–10575.
[16] B. Karg, S. Mohr, K. Weisz, Angew. Chem. Int. Ed. 2019, 58, 11068–11071;
Angew. Chem. 2019, 131, 11184–11188.
[36] I. Ivani, P. D. Dans, A. Noy, A. PØrez, I. Faustino, A. Hospital, J. Walther, P.
Andrio, R. GoÇi, A. Balaceanu, G. Portella, F. Battistini, J. L. Gelpí, C. Gon-
zµlez, M. Vendruscolo, C. A. Laughton, S. A. Harris, D. A. Case, M. Orozco,
Nat. Methods 2015, 12, 55–58.
[37] W. L. Jorgensen, J. Chandrasekhar, J. D. Madura, R. W. Impey, M. L. Klein,
J. Chem. Phys. 1983, 79, 926–936.
[17] Y. M. Vianney, P. Preckwinkel, S. Mohr, K. Weisz, Chem. Eur. J. 2020, 26,
16910–16922.
[38] R. Koradi, M. Billeter, K. J. Wüthrich, Mol. Graph. 1996, 14, 51–55.
[39] E. F. Pettersen, T. D. Goddard, C. C. Huang, G. S. Couch, D. M. Greenblatt,
E. C. Meng, T. E. Ferrin, J. Comput. Chem. 2004, 25, 1605–1612.
[18] R. F. Macaya, P. Schultze, F. W. Smith, J. A. Roe, J. Feigon, Proc. Natl.
Acad. Sci. USA 1993, 90, 3745–3749.
[19] L. M. Zhu, S. H. Chou, J. D. Xu, B. R. Reid, Nat. Struct. Biol. 1995, 2, 1012–
1017.
Manuscript received: November 19, 2020
Accepted manuscript online: December 25, 2020
Version of record online: January 22, 2021
[20] R. Perrone, M. Nadai, I. Frasson, J. A. Poe, E. Butovskaya, T. E. Smithgall,
M. Palumbo, G. Palu, S. N. Richter, J. Med. Chem. 2013, 56, 6521–6530.
Chem. Eur. J. 2021, 27, 6204 –6212
www.chemeurj.org
6212
ꢀ 2020 Wiley-VCH GmbH