Artemisinin-Polypyrrole Conjugates: Synthesis, DNA Binding Studies and Preliminary Antiproliferative Evaluation

Full text of DOI:10.1002/cmdc.201200536

CHEMMEDCHEM
COMMUNICATIONS
DOI: 10.1002/cmdc.201200536
Artemisinin–Polypyrrole Conjugates: Synthesis, DNA
Binding Studies and Preliminary Antiproliferative
Evaluation
Louise La Pensꢀe,[a] Sunil Sabbani,[a] Raman Sharma,[a] Inder Bhamra,[a] Emma Shore,[a]
Amy E. Chadwick,[b] Neil G. Berry,[a] James Firman,[b] Nuna C. Araujo,[e] Lꢁlia Cabral,[e]
Maria L. S. Cristiano,[e] Cerys Bateman,[b] Omar Janneh,[c] Adelina Gavrila,[c] Yi Hang Wu,[c]
Afthab Hussain,[c] Stephen A. Ward,[d] Paul A. Stocks,[d] Rick Cosstick,[a] and Paul M. O’Neill*[a]
Artemisinin-based combination therapies (ACTs) are currently
the recommended treatment for uncomplicated and severe
cases of malaria.[1] Additionally, artemisinins, as well as
a number of other sesquiterpene lactones (SLs), are currently
in phase I–II clinical trials against breast, colorectal and non-
small-cell lung cancers.[2] As outlined by the iron-dependent
activation hypothesis,[3] the activity of artemisinin (ART) is de-
pendent on the endoperoxide bridge.[4] The peroxide is
cleaved by endogenous sources of FeII to generate highly reac-
tive carbon-centred radicals (CCRs), which are believed to react
with critical cellular targets.[3] ART demonstrates selectivity to-
wards rapidly proliferating cancer cell lines that possess a high
intracellular iron content required to sustain their characteristic
high rates of multiplication.[5] Iron activation links this particu-
lar potency of ART towards rapidly proliferating cancer cell
lines; differentiation between healthy and cancerous cells by
variation of iron concentration provides a strategy for selective
cytotoxicity by ART and its derivatives.[4] The mechanism by
which ART exerts its cytotoxic activity still remains elusive. ART
acts by disruption of proliferation,[6, 7] oxidative stress,[8] anti-an-
giogenesis,[9] NF-kB signalling,[10] apoptosis[4] and interfering
with iron uptake and metabolism.[6] ART also induces DNA
breakage,[11] and it has been reported that artesunate-mediat-
ed DNA damage contributes to its therapeutic efficacy.[12]
Natural antibiotics like netropsin (NET) and distamycin (DST)
containing amide-linked N-methyl pyrrole units, show high
affinity for the minor groove of DNA at adenine and thymine
(AT) tracts.[13] Within the context of the hybrid approach, poly-
pyrrole derivatives of NET and DST have been exploited as
DNA-targeted delivery agents, such that conjugation to cyto-
toxic moieties (e.g., alkylators, intercalators and free radical
generators) has provided hybrids with enhanced cytotoxic
activity.[14]
Previously, we explored the antitumour and antimalarial
activity of ART–acridine hybrids,[15] resulting in a two- to four-
fold increase in activity against leukaemia and breast cancer
cell lines compared with dihydroartemisinin (DHA) alone.
These results encouraged us to prepare ART hybrids with di-
and tripyrrole polyamides as DNA recognition elements. The
effect that polypyrrole chains have on enhancing DNA cleav-
age potency, promoting cellular uptake,[16] and their use in the
development of novel DNA-directed anticancer treatments,[17]
make polypyrrole minor groove binders good candidates for
enhancing the antiproliferative effects of artemisinin deriva-
tives. We hypothesised that tethering the minor groove bind-
ing motif to ART might lead to enhanced cytotoxicity by di-
recting the endoperoxide moiety to DNA. Studying the DNA
binding affinity, cytotoxicity and the mechanism of cell death
caused by the ART portion would also give insight into the re-
lationship between the binding affinity and potential for DNA
damage by this novel class of DNA-directed drugs.
[a] Dr. L. La Pensꢀe,+ Dr. S. Sabbani,+ Dr. R. Sharma, Dr. I. Bhamra, E. Shore,
Dr. N. G. Berry, Prof. R. Cosstick, Prof. P. M. O’Neill
Department of Chemistry, University of Liverpool
Crown Street, Liverpool, L69 7ZD (UK)
In this paper, we describe the synthesis, DNA binding/mod-
elling study, and preliminary biological evaluation of a series of
hybrid ART–polypyrrole minor groove binders (1–7). The conju-
gates examined vary in the length of the pyrrole chain (2 or 3
pyrrole units), the nature of the linker (flexible or rigid), and
length of the linker (oligomethylene (n >1); R=p-aryl,
p-benzyl).[18]
E-mail: p.m.oneill1@liv.ac.uk
[b] Dr. A. E. Chadwick, J. Firman, C. Bateman
MRC Centre for Drug Safety Science
Department of Pharmacology & Therapeutics, University of Liverpool
Sherrington Buildings, Ashton Street, Liverpool, L69 3GE (UK)
[c] Dr. O. Janneh, A. Gavrila, Y. H. Wu, A. Hussain
Department of Biomolecular & Sports Sciences, Coventry University
Priory Street, Coventry, CV1 5FB (UK)
Work within our group has developed efficient methods for
modifying ART, allowing the introduction of synthetically
useful functional groups at the C-10 position of the artemisinin
framework. Using this experience, we selected and synthesised
alkoxy-,[19] phenoxy-[20] and carba-[21] linked ART carboxylic acids
to undergo amide coupling with the amine terminus of poly-
pyrrole chains.
[d] Prof. S. A. Ward, Dr. P. A. Stocks
Liverpool School of Tropical Medicine
Pembroke Place, Liverpool, L3 5QA (UK)
[e] Dr. N. C. Araujo, L. Cabral, Prof. M. L. S. Cristiano
Centre of Marine Sciences (CCMAR), University of the Algarve
Campus de Gambelas, 8005-139, Faro (Portugal)
[+] These authors contributed equally to this work.
Treatment of DHA (8) with hydroxy esters 9a–d provided
the required phenoxy or alkoxy esters 10a–d. Mild hydrolysis
and neutralising work-up provided carboxylic acid intermedi-
Supporting information for this article is available on the WWW under
http://dx.doi.org/10.1002/cmdc.201200536.
ꢂ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2013, 8, 709 – 718 709

CHEMMEDCHEM
COMMUNICATIONS
www.chemmedchem.org
ates 11 a–d (Scheme 1A). Carba-linked intermediate 15 was
synthesised in four steps as previously described
(Scheme 1B).[15,21]
Scheme 1. Synthesis of A) carboxylic acid intermediates 11 a–d and B) carba-
linked intermediate 15. Reagents and conditions: a) BF3·Et2O, Et2O, RT, over-
night, 40–50%; b) TMSOTf, AgClO4, NEt3, CH2Cl2, ꢁ788C, 3 h, 50–80%; c) 5%
KOH in MeOH, RT, 4 days; or 2.5% KOH in MeOH/H2O (1:1), 3–4 h, 75–98%;
d) BzCl, Py, CH2Cl2, 08C, 16 h, 91%; e) Allyltrimethyl silane, ZnCl2, sieves (4 ꢃ),
C2H2Cl2, 08C, 3 h, 83%; f) 1. O3, MeOH, ꢁ788C, 1 h; 2. PPh3, MeOH, ꢁ788C!
RT, 18 h, 69%; g) NaClO2, 2-methyl-2-butene, NaH2PO4, tBuOH/H2O (5:1), RT,
2 h, 62%.
N-Methylpyrrole 16 was treated with trichloroacetyl chloride
to give 17, which was then nitrated to give minor groove
binder building block 18.[22] To synthesise the terminal unit of
the carboxamide chain (19), compound 18 was reacted with
3-dimethylaminopropyl amine (Scheme 2A). This terminal
group is often selected as a DNA recognition moiety as it
allows a positive charge to be distributed across the dimethyl-
amine terminus allowing electrostatic interactions with DNA.
The di- and tripyrrole units were then built from the terminal
unit (19) by reduction of the nitro group to the reactive amine
and subsequent reaction with 18 (Scheme 2B).
Table 1. DTm values of compounds 1–8 with netropsin (NET) and dipyr-
role 20 as controls.
A number of amide coupling conditions were trialled and
found that N,N,N’,N’-tetramethyl-O-(1H-benzotriazol-1-yl)uroni-
um hexafluorophosphate (HBTU) and ethyldiisopropylamine
(EDIPA) were the most suitable carboxylic acid activating
agents. The nitro reduction of polypyrrole units 20 and 21
gave reactive amines 20a and 21a, which were then coupled
overnight with activated forms of carboxylic acids 11 a–d and
15 to give target ART–polyamide conjugates 1–7 (Scheme 3).
Compd
DTm[a] [8C]
Compd
DTm[a] [8C]
NET
20
1
2
3
9.2ꢀ0.0
5.1ꢀ0.5
3.7ꢀ0.1
4.7ꢀ0.1
4.7ꢀ0.4
4
5
6
7
8
4.8ꢀ0.5
3.0ꢀ1.0
3.5ꢀ0.2
3.4ꢀ0.3
3.9ꢀ0.3
[a] DTm =Tm(duplex+drug)ꢁTm(duplex); data represent the mean ꢀstan-
dard deviations (SD) of three experiments. Experiments were carried out
at a duplex concentration of 1 mm with a duplex/drug ratio of 1:1 in H2O
buffered with 1m NaCl, 0.1m MgCl2, 10 mm Tris·HCl (pH 7). The absorb-
ance was measured at 260 nm every 0.58C whilst heating from 18–908C
at increments of 0.58Cminꢁ1. Further experimental details are given in
the Supporting Information.
Thermal
denaturation
of
the
self-complementary
d(CGCGAATTCGCG)[23] duplex in the presence of 1–7 and NET
was carried out to ascertain the effect of the ART moiety on
the DNA binding affinity of the polypyrrole portions (Table 1).
The difference in the thermal denaturation temperatures (DTm)
ꢂ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2013, 8, 709 – 718 710

CHEMMEDCHEM
COMMUNICATIONS
www.chemmedchem.org
groove; this binding is supported by modelling studies con-
ducted herein (see Supporting Information).
Molecular modelling was performed to rationalise the rela-
tive binding affinities of NET and compounds 2, 3, 5 and 20 to
the minor groove of the d(CGCGAATTCGCG) duplex. Molecular
docking was initially employed to predict binding modes, and
then more rigorous molecular dynamics–molecular mechanics/
Poisson–Boltzmann surface area (MM/PBSA) simulations were
performed. These ligands were chosen on the basis that they
span the range of experimentally determined DTm values. In-
terestingly, compounds 3 and 5 varied only in the nature of
the linker from the ART moiety to the dipyrrole head group,
yet their binding affinities and DTm values were markedly dif-
ferent (DTm =4.78C for 3 cf. DTm =3.08C for 5). The protona-
tion state of the ligands was adjusted so that it was consistent
with that found at the physiologically relevant pH 7.4. Full de-
tails of how the ligands were prepared for docking and molec-
ular dynamics simulation are discussed in the Computational
Methods part of the Experimental Section.
The X-ray crystal structure of d(CGCGAATTCGCG) duplex in
complex with NET (PDB code: 1D86)[24] was used as the basis
for all structure-based modelling work. This 2.2 ꢃ-resolution
structure represents the only example of the structural charac-
terisation of an unmodified d(CGCGAATTCGCG)–NET complex
currently described in the literature. For further information on
the binding modes of NET characterised in co-crystals with var-
ious B-DNA duplexes, see the Supporting Information.
The molecular docking program GOLD[25] was employed to
initially predict the binding pose and strength of interaction of
the ligands in complex with the dodecameric DNA duplex.
GOLD has been shown to accurately reproduce the crystallo-
graphically determined binding poses of minor groove binders
to DNA, and a diverse range of ligands to RNA targets, ranking
amongst the top methods in each study.[26,27] Previous studies
have shown that the binding pose of NET in the crystal struc-
ture PDB code 1D86 can be accurately reproduced by using
the GOLDscore scoring function[25,26] within GOLD 3.2.[26] Dock-
ing validation was repeated for the current version
(GOLD 5.0.1) used in this study. Full computational details of all
docking calculations are discussed in the Computational Meth-
ods part of the Experimental Section. Figure 1A shows that
GOLD 5.0.1 accurately reproduces the native binding pose of
NET in the minor groove of d(CGCGAATTCGCG) in the co-crys-
tal structure (PDB code: 1D86). The root-mean-square (RMS)
deviation from the experimental binding pose for the top
three scoring poses as predicted by the GOLDscore scoring
function was found to be within 1.1 ꢃ, which is concordant
with the performance of GOLD 3.2 for this system.[26] This vali-
dation gave further assurance that the binding poses of com-
pounds 2, 3, 5 and 20, for which crystal structures are not de-
termined, can be accurately predicted by GOLDscore as imple-
mented in GOLD 5.0.1.
Scheme 2. Synthesis of polypyrrole minor groove binding components. Re-
agents and conditions: a) Cl3CCOCl, CH2Cl2, RT, 3 h, 90%; b) 70% HNO3, Ac2O,
ꢁ408C, 0.5 h, 45%; c) 3-Dimethylaminopropylamine, THF, 08C!RT, 1 h, 86%;
d) H2, 1 atm, PtO2 (or 10% Pd/C), MeOH, 16 h (or 3–4 h); e) 18, DMF, 08C!
RT, 1 h, 20: 60%, 21: 45% (two steps).
Scheme 3. Synthesis of conjugates 1–7. Reagents and conditions: a) 1. HBTU,
EDIPA, DMF, 08C, 0.5 h; 2. amine 20a or 21a, DMF, RT, overnight, 28–71%.
of the DNA duplex with and without the drug were measured
as described in Table 1.
All the conjugates exhibited a stabilising effect on the
duplex, and this observation demonstrates the suitability of
the polyamide unit as a DNA delivery moiety. Conjugation with
ART had some detrimental effects on binding when compared
with both NET and 20; however, affinity for the duplex by the
conjugates was evident, with 2, 3 and 4 demonstrating the
greatest duplex stabilisation. Structurally, conjugate 2 has the
longest and most flexible linker giving the minor groove
binder portion “freedom” to bind closely into the DNA minor
After removal of the native ligand, NET, from the crystal
structure PDB code 1D86, compounds 2, 3, 5 and 20 were
docked into the minor groove, and the top three scoring
poses were examined for each compound. In all cases, at least
two out of the three top scoring poses were in agreement
ꢂ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2013, 8, 709 – 718 711

CHEMMEDCHEM
COMMUNICATIONS
www.chemmedchem.org
d(CGCGAATTCGCG) DNA duplex. The amide and pyrrole
groups that form the core of all the molecules are generally
shifted relative to their position in the native (class II) NET
binding pose seen in crystal structure PDB code 1D86.
From Figure 2, it can be seen that each ligand is predicted
to interact with the minor groove of the d(CGCGAATTCGCG)
duplex in a unique manner. Compound 2 is predicted to bind
to the length of the minor groove with the pyrrole–amide
motif sitting more towards the 5’-(C1) end of strand 1 (C1!
G12), much “higher” than those in the native NET pose (Fig-
ure 2A). Compound 3 is predicted to bind lower in the minor
groove, with the pyrrole–amide motif located much more to-
wards 3’ (strand 1) end the of AATT region than found in the
native binding pose of NET (Figure 2B). Furthermore, conju-
gate 3 is predicted to bind lower than NET in a class I complex,
which is centrally located, as exemplified by the crystal struc-
ture of NET in complex with the d(CGCGAATTC5BrGCG) duplex
(PDB code: 6BNA; see Figure S1 in the Supporting Informa-
tion).[29] Compound 5 is predicted to be located in a similar po-
sition to 2, with the pyrrole–amide motif again being located
much “higher” than in the native NET pose see in the crystal
structure PDB code 1D86 (Figure 2C). Compound 20 is predict-
ed to be located centrally in AATT region of the minor groove,
in a very similar position to the class I binding mode of NET
found in 6BNA (Figure 2D).[29] This similarity is exemplified by
the significant overlap of the pyrrole–amide motifs in class I
NET and 20 (see Figure S2 in the Supporting Information). Full
details of each binding modes are reported in the Computa-
tional Methods section of the Supporting Information.
The GOLDscore for compounds 2, 3, 5 and 20 was found to
correlate poorly with the DTm values (R2 =0.1; Table 2). The
binding strength or potency of a ligand can be normalised by
Table 2. Correlation of DTm values and GOLDscore results for ligands
docked into the minor groove of d(CGCGAATTCGCG) DNA duplex (PDB
code: 1D86[24]).[a]
Figure 1. A) Best scoring docked pose of netropsin (NET) as predicted by
GOLD 5.0.1 in the context of d(CGCGAATTCGCG) DNA (PDB code: 1D86[24]).
The DNA section shown in blue line representation indicates the AATT rec-
ognition region. Also shown is a close up of the best docked pose (cyan) in
comparison to native crystallographic binding mode (green). B) The best
docked pose (cyan) for compounds 2, 3, 5 and 20 (left to right) in compari-
son with the native crystallographic binding mode of NET (green). Images
created using PyMol.[28]
Compd
DTm
[8C]
DTm (HAC)
[8C]
HAC
Goldscore
LigEff (HAC)
NET
20
2
3
5
9.20
5.10
4.70
4.70
3.00
0.30
0.19
0.08
0.09
0.05
31
27
56
53
55
88.64
77.16
96.95
93.58
104.72
2.86
2.86
1.73
1.77
1.90
[a] DTm/HAC=heavy atom count (HAC)-normalised DTm.
with the top scoring pose. Figure 1 shows the position of top
poses for compounds 2, 3, 5 and 20 relative to the native
binding pose of NET with d(CGCGCAATTCGCG). From Fig-
ure 1A, it can be seen that all compounds are predicted to
adopt a crescent shape and display a coarse overlap with the
native binding pose of NET. Compounds 2, 5 and 20 are pre-
dicted to bind with their propyl ammonium groups pointing
towards the guanidinium terminus of NET and the ART groups
towards the propyl amidinium of NET.
size of molecule or other molecular properties to give ligand
efficiency indices; this can help to mitigate against size effects.
Efficiency indices have been shown to be particularly useful in
lead identification of the drug discovery process.[30–33] The con-
cept of calculated ligand efficiency has been successfully em-
ployed in structure-based virtual screening, showing improved
correlation against experimental values.[34–36] In the present
docking study, the GOLDscore was found to increase with size
of molecule and, in an effort to quantify this possible size
effect, the heavy atom count (HAC) was calculated for each of
Compound 3, however, is predicted to bind in the opposite
configuration with the ART moitey being positioned towards
the guanidinium group of NET. This might be of limited conse-
quence due to the palindrome nature of the
ꢂ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2013, 8, 709 – 718 712

CHEMMEDCHEM
COMMUNICATIONS
www.chemmedchem.org
order to assess their predicted
binding modes and energies of
interaction. Simulations were
performed as specified in the Ex-
perimental Section. In particular,
starting configurations for each
of the ligand–d(CGCGAATTC-
GCG) complexes were taken as
the best scoring docking poses
for each ligand. Simulations
were performed using explicit
solvation and periodic boundary
conditions and were followed by
single trajectory MM/PBSA calcu-
lations, as recommended in the
literature for accurate calculation
of interaction energies of DNA
minor groove binders.[37] Analysis
of the representative binding
poses for NET shows that NET
has shifted up by half a base
unit to accommodate bifurcated
H-bonds as supported by NMR
studies.[38] In general, the stron-
gest binders (NET and 20)
change least both in absolute
terms from the starting structure
and fluctuation throughout the
simulation. Ligands 2, 3 and 5
are found to deviate more from
their minimised starting configu-
rations and fluctuate more sub-
stantially throughout the simula-
tions (even after a substantial
16 ns equilibration phase). Fluc-
tuations in the DNA duplexes for
each simulation are less pro-
nounced, although NET and 20
are still found to distort their
DNA duplexes to a lesser degree
and fluctuate less throughout
the simulations. A detailed analy-
sis of the simulation trajectories
and discussion of the prevalent
binding modes for each ligand–
DNA duplex complex is dis-
cussed in the Supporting Infor-
mation.
Figure 2. The best docked pose for ligands 2, 3, 5 and 20 (A–D) in the context of d(CGCGAATTCGCG) duplex DNA
(PDB code: 1D86[24]). The DNA section shown in blue line representation indicates the AATT recognition region.
The docked binding poses are shown in stick representation, with nitrogen atoms in dark blue and carbon atoms
in cyan. For the ligand and duplex H-bond donors and acceptors, interatomic distances within 3.5 ꢃ are shown as
dashed lines, with those that are concordant with the directionality of a H-bond being further annotated with the
interatomic distances (ꢃ). Images created using PyMol.[28]
Figure 4 shows a plot of the
the ligands. The HAC values were found to positively correlate
with the GOLDscore (R2 =0.43).
calculated MM/PBSA interaction free energies against the ex-
perimentally determined DTm values. It can be seen that there
is an excellent correlation between the calculated and experi-
mental values, with a clear differentiation achieved between
the predicted binding energies of NET and 20. Furthermore,
the DG(MM/PBSA) value for NET is in agreement with that
found in a previous MM/PBSA study.[26] DG(MM/PBSA) of inter-
action is itself constructed from individual energies of com-
GOLDscore normalised by HAC, LigEff(HAC), was calculated
and found to correlate moderately (R2 =0.46) with DTm values
(Figure 3A). Furthermore, LigEff(HAC) was found to correlate
better (R2 =0.81) with HAC-normalised DTm values (Figure 3B).
Molecular dynamics simulations and MM/PBSA calculations
were performed for NET and compounds 2, 3, 5 and 20 in
ꢂ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2013, 8, 709 – 718 713

CHEMMEDCHEM
COMMUNICATIONS
www.chemmedchem.org
plexation for van der Waals interactions (DEvdW), electrostatic
interactions (DEElec) and electrostatic and nonpolar solvation
terms (DESolv(PB) and DESolv–Cav, respectively).
Table 3 shows a decomposition of the contributions of the
individual energetic terms to DG(MM/PBSA) of binding for
each ligand, and correlation of these individual contributions
Table 3. Decomposition of the contributions of the individual energetic
terms to DG (MM/PBSA).[a]
Compd DTm DG(MM/PBSA)
[8C]
DEvdW
DEElec
[kcalmolꢁ1
DESolv (PB) DESolv–Cav
]
NET
20
2
3
5
9.20
5.10
4.70
4.70
3.00
ꢁ63.38
ꢁ29.10
ꢁ28.30
ꢁ27.88
ꢁ25.06
ꢁ69.28 ꢁ1168.12
1178.50
569.82
486.40
408.51
557.90
ꢁ4.47
ꢁ3.26
ꢁ3.81
ꢁ3.81
ꢁ5.12
ꢁ47.56
ꢁ50.55
ꢁ50.91
ꢁ63.19
ꢁ548.10
ꢁ460.35
ꢁ381.67
ꢁ514.65
R
R2
–
–
ꢁ0.97
ꢁ0.50
ꢁ0.90
0.89
0.80
0.06
0.00
0.93
0.25
0.82
[a] R: correlation coefficient; R2: coefficient of determination.
to the experimental DTm values. The electrostatic interaction
terms and electrostatic component of solvation terms correlate
well with the DTm value (R2 =0.82 and 0.80, respectively), al-
though DEElec is favourable and whereas DESolv(PB) is unfavoura-
ble (due to the ligand and DNA duplex losing their individual
solvent stabilisation on complexation). In contrast, the van der
Waals and nonpolar solvation terms do not correlate with DTm
(R2 =0.25 and 0.00, respectively). The van der Waals term is,
however, significant in predicting the correct strength of bind-
ing for 5, with the large van der Waals energy of interaction
for this molecule mitigating against the unfavourable DEElec
and DESolv(PB) values. From the decomposition of the DG(MM/
PBSA), it is apparent that binding is principally driven by differ-
ential solvation of both the ligand and DNA duplex on com-
plexation and extent of polar interactions with the DNA
duplex; these electrostatic terms only partially accounted for in
the docking calculations discussed earlier.
Figure 3. A) Heavy atom count (HAC)-normalised GOLDscore ligand efficien-
cy (LigEff (HAC)) against DTm (R2 =0.46). B) LigEff (HAC) against HAC-normal-
ised DTm (DTm (HAC)), where n is the number of HAC (R2 =0.81). GOLDscore
scoring function values were calculated for compounds 2, 3, 5, 20 and NET
when docked into the minor groove of d(CGCGAATTCGCG) duplex DNA
(PDB code: 1D86[24]).
As noted, an additional aim of this work was to determine
whether the cytotoxic or antimalarial activity of ART can be
augmented by conjugation to a DNA binding element. ART–
polyamide hybrids 1–7 were evaluated for their in vitro activity
against human leukaemia (HL-60) and colon adenocarcinoma
(HT-29) cancer cell lines in a methylthiazol tetrazolium (MTT)
assay, with DHA and NET as reference agents (Table 4). Further,
the mechanism of cell death caused by conjugates 3 and 5
was examined against HL-60 cells, selecting mitochondrial de-
polarisation and DNA fragmentation as early and late events in
the apoptotic pathway, evaluated using a tetramethylrhoda-
mine ethyl ester (TMRE) assay and flow cytometric analysis
after staining with propidium iodide (PI), respectively (for full
details, see the Supporting Information).
Against HT-29, compounds 2, 3 and 5 were more active
than DHA and NET (Table 4). Compound 2 expressed potent
nanomolar activity versus HL60 cells, in contrast to the other
Figure 4. DG (MM/PBSA) values against experimental DTm (R2 =0.93). Each
DG value was calculated for unrelated snapshots from 20 ns of production
simulation.
ꢂ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2013, 8, 709 – 718 714

CHEMMEDCHEM
COMMUNICATIONS
www.chemmedchem.org
In summary, conjugate 2 emerges as the most active mole-
cule being twofold more active than DHA against HT-29 cells
and exhibiting around five times the potency of DHA against
HL-60 cells. Conjugate 2 also exhibits one of the strongest
affinities for the DNA duplex d(CGCGAATTCGCG). Structurally,
these results are not surprising as conjugate 2 has the longest
flexible linker, allowing for the greatest steric freedom to both
portions of the hybrid. Modelling studies have accurately pre-
dicted the experimental DTm data, and decomposition of these
calculated free energies of binding has shown that the binding
affinity of these hybrid compounds depends on multiple fac-
tors, in particular electrostatic interactions and differential sol-
vation of the ligand and DNA duplexes on complexation.
Moreover, these factors are intrinsically linked to the varying
degrees of conformational change the ligands and DNA
duplex undergoes on binding. Further mechanistic studies are
needed in HT-29 cells to clarify whether ART maintains mito-
chondrial membrane depolarisation as a main mode of action
or whether the presence of the DNA binding structure has in
fact enhanced activity by causing a dual mode of action.
Table 4. Cytotoxic (HL-60 and HT-29 cells) and antimalarial (P. falciparum
3D7) activities of compounds 1–8 with netropsin (NET) and dihydroarte-
misinin (DHA) as controls.[a]
Compd
IC50[b] [mm]
IC50[c] [nm]
(HL-60)
(HT-29)
(3D7)
DHA
NET
1
2
3
4
5
6
7
0.50ꢀ0.10
12.92ꢀ2.31
11.00ꢀ1.87
22.87ꢀ1.56
6.26ꢀ0.45
7.38ꢀ0.52
39.94ꢀ3.93
8.78ꢀ1.47
58.76ꢀ2.77
48.01ꢀ3.16
2ꢀ0.5
ND
>100
4.41ꢀ0.49
0.11ꢀ0.03
1.19ꢀ0.09
6.35ꢀ2.93
1.42ꢀ0.66
4.60ꢀ1.40
11.67ꢀ4.38
65ꢀ9
40ꢀ6
50ꢀ12
126ꢀ11
24ꢀ8
119ꢀ14
120ꢀ18
[a] All values are the mean ꢀstandard deviations (SD) of three experi-
ments. [b] Cells were cultured in RPMI 1640 with 10% fetal bovine serum
and 1% w/v l-glutamine. Cultures were incubated in humidified air with
5% CO2 at 378C and kept below a density of 1ꢄ106. Compounds were
dissolved in DMSO before being added at various concentrations (0.01–
100 mm) to the cells. Plates were incubated for 24 and 48 h. Cell viability
was quantified using an MTT assay, and the absorbance of the samples
was measured at 570 nm by a plate reader. Results were expressed as
a percentage of the vehicle-only control, and IC50 values were calculated
using GraFit (version 4.0). [c] Parasites were maintained in continuous cul-
ture using the method of Jensen and Trager.[39] Antimalarial activity was
assessed with an adaption of the 48 h sensitivity assay of Desjardins
et al.[40] using [3H]hypoxanthine incorporation as an assessment of para-
site growth.
Experimental Section
Organic synthesis
Full details on the synthesis of intermediate compounds, including
characterisation data, general remarks and instrument particulars,
can be found in the Supporting Information. 10b-(2-Carboxyethyl)-
deoxoartemisinin (13) was synthesised according to Jones and co-
workers.[15]
conjugates that were between two- and 20-fold less potent
than DHA. It is interesting to note that NET is active against
HT-29 but inactive against HL-60 cells. Compounds 3, 5 and
DHA were further examined in HL-60 cells to determine the
mechanism of cell death. Although these conjugates induced
mitochondrial membrane depolarisation, no significant impact
on the cellular DNA content was seen. These results are strong-
ly indicative of 3 and 5 inducing cell death by mitochondrial
membrane depolarisation and apoptotic death in HL-60 cell
lines.
Artemisinin conjugates 1–7: Nitrodipyrrole 20 or 21 (1.04 equiv
with respect to ART carboxylic acid) was reduced according to the
procedure described for the synthesis of compounds 20 or 21
from 19 (3–4 h reaction time; see the Supporting Information for
details). The crude reaction mixture was filtered through celite and
concentrated in vacuo. The residue was redissolved in anhyd DMF
(30 mL), and the solution was concentrated to approximately half
the original volume in vacuo. Resulting amine 20a or 21a was
cooled to 08C and stored under an atmosphere of N2. In another
flask, the appropriate ART-derived carboxylic acid (1.0 equiv) was
dissolved in anhyd DMF, and the solution was cooled to 08C. This
solution was then treated with HBTU (1.1 equiv) and DIPEA
(2.0 equiv) and stirred at 08C under N2 for 1.0 h. The ice bath was
removed, and this solution containing the activated ART-derived
acid was transferred to the 20a or 21a solution in DMF. The result-
ing reaction mixture was stirred at RT for 23 h. The solvent was re-
moved in vacuo, and the crude material was redissolved in a mini-
mum amount of MeOH, and partitioned between EtOAc (50 mL)
and H2O (50 mL). The aqueous layer was extracted with EtOAc (3ꢄ
50 mL) and CH2Cl2 (2ꢄ50 mL). The organic layers were separately
washed with saturated NaHCO3 (1ꢄ50 mL) and brine (1ꢄ50 mL),
and the combined organic layers were dried (MgSO4), filtered and
concentrated in vacuo. Purification by flash chromatography
(column flush=MeOH/EtOAc (1:1) then MeOH/EtOAc (1:1) +3%
NH4OH) gave near-pure material, which was redissolved in CH2Cl2
and filtered to removed dissolved silica gel. Concentration of the
filtrate in vacuo gave the desired ART hybrid (1–7).
In the case of HT-29 cells, the addition of a minor groove
binding element in compounds 2, 3 and 5 enhanced the anti-
proliferative activity. The primary mechanism by which these
compounds cause cell death in HT-29 would also appear to be
by mitochondrial depolarisation and, as such, localisation of
the compounds to cellular DNA is not advantageous.[41] Further
investigations into the mechanism of cell death employed by
these compounds in HT-29 cell lines is required to see if their
improved activity is a result of a shift in the mechanism of
action.
The human malaria parasite Plasmodium falciparum possess-
es AT-rich DNA (>80% cf. to 60% in man).[42] We proposed the
AT-specific ART conjugates have the potential for enhanced
and specific antimalarial activity. The conjugates were evaluat-
ed for their in vitro antimalarial activity against the chloro-
quine-sensitive 3D7 strain of P. falciparum, however, none of
the conjugates were more active than DHA and were, in some
cases, up to 60-fold less active. Nevertheless, the antimalarial
activities of all these hybrid compounds 1–7 were in the nano-
molar range.
Compound 1 (187 mg, 42% yield): 1H NMR (400 MHz, CDCl3): d=
7.77 (m, 2H), 7.62 (s, 1H), 7.15 (d, J=1.7 Hz, 1H), 7.08 (d, J=1.7 Hz,
1H), 6.59 (d, J=1.7 Hz, 1H), 6.45 (d, J=1.7 Hz, 1H), 5.40 (s, 1H),
ꢂ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2013, 8, 709 – 718 715

CHEMMEDCHEM
COMMUNICATIONS
www.chemmedchem.org
4.76 (d, J=3.2 Hz, 1H), 3.89 (s, 6H), 3.81 (dt, J=9.6, 6.7 Hz, 1H),
3.46 (dd, J=11.6, 5.9 Hz, 2H), 3.44 (dt, J=9.6, 6.4 Hz, 2H), 2.61 (m,
1H), 2.45 (t, J=6.2 Hz, 2H), 2.40 (m, 1H), 2.33 (m, 2H), 2.28 (s, 6H),
2.05–1.20 (m, 26H), 1.43 (s, 3H), 0.95 (d, J=6.2 Hz, 3H), 0.90 ppm
(d, J=7.3 Hz, 3H); 13C NMR (100 MHz, CDCl3): d=171.3, 162.2,
159.4, 124.3, 123.5, 122.1, 121.6, 119.5, 119.1, 104.5, 104.0, 103.4,
102.3, 88.3, 81.6, 68.8, 59.4, 53.0, 45.9, 45.9, 44.9, 39.8, 37.9, 37.3,
37.0, 37.0, 36.9, 35.1, 31.4, 30.0, 29.8, 29.7, 29.6, 29.6, 26.6, 26.5,
26.3, 26.2, 25.1, 24.9, 20.8, 13.4 ppm; IR (neat): n˜max =3310, 2937,
1651, 1099, 944, 873, 826 cmꢁ1; HRMS (ES+): m/z [M+H]+ calcd
for C38H58N6O8: 727.4394, found: 727.4415.
2.11 (m, 1H), 1.75–1.94 (m, 3H), 1.71 (quin, J=6.3 Hz, 2H), 1.62 (dq,
J=13.1, 3.0 Hz, 1H), 1.46–1.56 (m, 2H), 1.44 (s, 3H), 1.19–1.38 (m,
2H), 0.94 (d, J=6.0 Hz, 3H), 0.93 (d, J=7.3 Hz, 3H), 0.89–0.92 ppm
(m, 1H); 13C NMR (100 MHz, CDCl3): d=172.7, 168.6, 161.8, 158.9,
137.4, 133.9, 129.2, 127.7, 123.7, 123.1, 121.4, 121.3, 119.2, 118.7,
104.1, 103.7, 103.2, 101.2, 88.0, 81.1, 69.3, 58.5, 52.5, 45.2, 44.3,
43.4, 39.0, 37.4, 36.5, 36.5, 36.4, 34.5, 30.8, 26.1, 25.9, 24.6, 24.5,
22.5, 20.3, 13.0 ppm; IR (neat): n˜max =2943, 1648, 1532, 1437, 1403,
1255, 1140, 1099, 1011, 826, 732 cmꢁ1; HRMS (ES+): m/z [M+H]+
calcd for C41H57N6O8: 761.4238, found: 761.4273.
1
Compound 6 (165 mg, 32%): H NMR (400 MHz, CDCl3): d=8.36 (br
1
s, 1H), 8.18 (br s, 1H), 8.00 (br s, 1H), 7.67 (br s, 1H), 7.23–7.34 (m,
4H), 7.19 (br s, 1H), 7.11–7.16 (m, 1H), 6.95–7.07 (m, 1H), 6.56–6.69
(m, 3H), 5.46 (s, 1H), 4.76–5.00 (m, 2H), 4.35–4.54 (m, 1H), 3.76–
3.90 (m, 9H), 3.58–3.68 (m, 2H), 3.39 (d, J=4.3 Hz, 2H), 2.58–3.20
(m, 2H), 2.40 (t, J=6.2 Hz, 2H), 2.23 (s, 6H), 2.09–2.17 (m, 1H),
1.75–2.04 (m, 3H), 1.56–1.73 (m, 3H), 1.42–1.53 (m, 5H), 1.16–1.36
(m, 2H), 0.80–1.00 ppm (m, 7H); 13C NMR (100 MHz, CDCl3): d=
169.7, 169.0, 161.9, 159.0, 137.3, 134.8, 133.9, 129.2, 128.2, 127.6,
123.6, 123.2, 123.0, 121.6, 121.4, 121.4, 119.5, 119.3, 118.7, 104.2,
103.8, 103.2, 101.2, 88.0, 80.5, 69.3, 58.4, 52.5, 46.8, 45.2, 38.9, 37.4,
36.6, 36.6, 36.5, 36.4, 35.8, 34.5, 33.3, 29.6, 26.1, 26.0, 24.8, 24.6,
20.4, 13.0 ppm; IR (neat): n˜max =2923, 1647, 1583, 1531, 1464, 1435,
1402, 1255, 1080, 1004, 775 cmꢁ1; HRMS (ES+): m/z [M+H]+ calcd
for C47H63N8O9: 883.4718, found: 883.4720.
Compound 2 (153 mg, 34%): H NMR (400 MHz, CDCl3): d=7.78 (s,
1H), 7.76 (s, 1H), 7.71 (s, 1H), 7.16 (d, J=1.7 Hz, 1H), 7.07 (d, J=
1.7 Hz, 1H), 6.63 (d, J=1.7 Hz, 1H), 6.45 (d, J=1.7 Hz, 1H), 5.40 (s,
1H), 4.74 (d, J=3.4 Hz, 1H), 3.89 (s, 6H), 3.74 (1H, dt, J=9.7,
6.6 Hz), 3.44 (m, 2H, ), 3.36 (dt, J=9.8, 6.6 Hz, 1H), 2.60 (m, 1H),
2.43 (t, J=6.2 Hz, 2H), 2.40 (m, 1H), 2.33 (m, 2H), 2.28 (s, 6H),
2.05–1.20 (m, 16H), 1.43 (s, 3H), 0.95 (d, J=6.2 Hz, 3H), 0.88 ppm
(d, J=7.3 Hz, 3H); 13C NMR (100 MHz, CDCl3): d=171.1, 162.2,
159.4, 124.4, 123.5, 122.0, 121.7, 119.6, 119.1, 104.6, 104.1, 103.4,
102.3, 88.4, 81.6, 68.7, 59.4, 52.9, 45.9, 45.9, 44.8, 39.9, 37.9, 37.1,
37.0, 37.0, 36.8, 35.0, 31.3, 29.8, 26.6, 26.3, 26.3, 25.9, 25.1, 24.9,
20.8, 13.4 ppm; IR (neat): n˜max =3299, 2940, 1651, 1099, 937, 876,
822 cmꢁ1
783.5020, found: 783.5056.
;
HRMS (ES+): m/z [M+H]+ calcd for C42H66N6O8:
1
Compound 7 (159 mg, 35%): H NMR (400 MHz, CDCl3): d=8.91 (br
1
Compound 3 (283 mg, 68%): H NMR (400 MHz, CDCl3): d=8.86 (br
s, 1H), 8.11 (br s, 1H), 7.63 (t, J=5.2 Hz, 1H), 7.57–7.73 (m, 1H),
7.49 (d, J=7.7 Hz, 1H), 7.23–7.31 (m, 1H), 7.20 (d, J=1.5 Hz, 1H),
7.14 (dd, J=8.1, 1.6 Hz, 1H), 6.97 (br s, 1H), 6.66 (s, 1H), 6.54 (d,
J=1.6 Hz, 1H), 5.68 (d, J=2.8 Hz, 1H), 5.42 (s, 1H), 3.86 (s, 3H),
3.81 (s, 3H), 3.33–3.45 (m, 2H), 2.71–2.83 (m, (1H), 2.41 (t, J=
6.6 Hz, 2H), 2.34 (td, J=14.0, 3.2 Hz, 1H), 2.24 (s, 6H), 1.27 (s, 3H),
1.20–2.00 (m, 11H), 1.01 (d, J=7.3 Hz, 3H), 0.94 (d, J=5.3 Hz, 3H),
0.90–0.97 ppm (m, 1H); 13C NMR (100 MHz, CDCl3): d=165.2, 161.9,
158.9, 157.2, 135.9, 129.7, 123.5, 123.2, 121.5, 121.2, 120.6, 119.9,
118.6, 114.9, 104.9, 104.4, 103.0, 99.7, 88.3, 80.9, 58.3, 53.4, 45.1,
44.2, 38.7, 37.4, 36.5, 36.4, 36.2, 34.5, 30.8, 26.2, 25.7, 24.6, 24.4,
20.2, 12.7 ppm; IR (neat): n˜max =2943, 1646, 1581, 1532, 1435, 1403,
1263, 1227, 1095, 1036, 979, 875 cmꢁ1; HRMS (ES+): m/z [M+H]+
calcd for C39H53N6O8: 733.3925, found: 733.3936.
s, 1H), 8.14 (br s, 1H), 7.62 (t, J=5.1 Hz, 1H), 7.21 (d, J=1.8 Hz,
1H), 6.96 (d, J=1.8 Hz, 1H), 6.70 (d, J=1.8 Hz, 1H), 6.54 (d, J=
1.9 Hz, 1H), 5.48 (s, 1H), 4.92 (ddd, J=11.2, 6.2, 1.4 Hz, 1H), 3.89 (s,
3H), 3.84 (s, 3H), 3.43 (q, J=5.9 Hz, 2H), 2.56–2.71 (m, 2H), 2.46 (t,
J=6.5 Hz, 2H), 2.44–2.50 (m, 1H), 2.38–2.44 (m, 1H), 2.30 (s, 6H),
1.94–2.09 (m, 2H), 1.73 (quin, J=6.3 Hz, 2H), 1.64–1.84 (m, 3H),
1.32 (s, 3H), 1.21–1.40 (m, 4H), 0.97 (d, J=5.4 Hz, 3H), 0.95–1.00
(m, 1H), 0.90 ppm (d, J=7.5 Hz, 3H); 13C NMR (100 MHz, CDCl3):
d=169.2, 161.8, 159.0, 123.5, 123.1, 121.5, 121.5, 118.9, 118.5,
104.2, 103.1, 103.0, 90.2, 80.8, 70.1, 58.3, 51.7, 45.1, 43.4, 38.3, 37.4,
36.5, 36.4, 36.4, 34.2, 34.2, 30.3, 25.9, 25.8, 24.7, 24.7, 19.9,
12.0 ppm; IR (neat): n˜max =2939, 1647, 1583, 1530, 1437, 1404,
1263, 1207, 1095, 1044, 1013, 814 cmꢁ1; HRMS (ES+): m/z [M+H]+
calcd for C34H51N6O7: 655.3819, found: 655.3822.
1
Compound 4 (138 mg, 28%): H NMR (400 MHz, CDCl3): d=9.23 (br
Computational methods
s, 1H), 8.68 (br s, 1H), 8.33 (br s, 1H), 7.56–7.77 (m, 2H), 7.45 (d,
J=6.9 Hz, 1H), 7.13–7.30 (m, 5H), 7.07 (d, J=7.2 Hz, 1H), 6.55–6.73
(m, 2H), 5.55–5.97 (m, 1H), 5.39 (s, 1H), 3.76–3.89 (m, 9H), 3.33–
3.42 (m, 2H), 2.66–2.85 (m, 1H), 2.39 (t, J=6.1 Hz, 2H), 2.27–2.35
(m, 1H), 2.21 (s, 6H), 1.27 (s, 3H), 1.16–2.06 (m, 11H), 0.99 (d, J=
7.2 Hz, 3H), 0.94 (br s, 3H), 0.87–1.14 ppm (m, 1H); 13C NMR
(100 MHz, CDCl3): d=172.8, 161.9, 159.1, 157.0, 135.9, 130.9, 129.7,
123.4, 123.0, 121.8, 121.5, 120.7, 120.7, 120.1, 120.0, 119.1, 118.8,
118.8, 104.5, 103.5, 103.2, 99.5, 88.4, 80.8, 58.1, 53.4, 52.3, 45.0,
44.2, 38.6, 37.4, 36.5, 36.2, 34.4, 30.7, 26.1, 25.6, 24.5, 24.3, 22.5,
20.2, 12.7, ꢁ21.1 ppm; IR (neat); n˜max =2935, 1647, 1582, 1529,
1434, 1403, 1255, 1096, 1035, 980, 876, 738 cmꢁ1; HRMS (ES+): m/z
[M+H]+ calcd for C45H59N8O9: 855.4405, found: 855.4395.
Molecular docking studies: All molecular docking was performed
using the GOLD docking suite.[25] Ligands 2, 3, 5, 20 and NET were
protonated to physiological pH (pH 7.4) using the “ionize molecule
at pH” component within Pipeline Pilot,[43] 3D structures were
create through minimisation using the Merck molecular force field
(MMFF) 94[44] within the SPARTAN ’08 1.0.0 package.[45] Preparation
of all the input files and run configuration, including setting up the
DNA duplex system, was performed in the Hermes visualise
1.4.1.[46] The d(CGCGAATTCGCG) duplex crystal structure (PDB
code: 1D86[24]) was prepared by removing NET and water mole-
cules to avoid potential interference with the docking. Hydrogen
atoms were added to the DNA duplex according to the default
heuristics. The binding site was defined as being all residues 6 ꢃ
from NET in the 1D86 structure.
1
Compound 5 (311 mg, 71%): H NMR (400 MHz, CDCl3): d=7.97 (br
s, 1H), 7.93 (br s, 1H), 7.68 (t, J=5.0 Hz, 1H), 7.25–7.29 (m, 4H),
7.15 (d, J=1.8 Hz, 1H), 7.05 (d, J=1.8 Hz, 1H), 6.59 (d, J=1.8 Hz,
1H), 6.50 (d, J=1.8 Hz, 1H), 5.46 (s, 1H), 4.88 (d, J=3.0 Hz, 1H),
4.85 (d, J=12.6 Hz, 1H), 4.49 (d, J=12.5 Hz, 1H), 3.85 (s, 3H), 3.84
(s, 3H), 3.65 (s, 2H), 3.41 (q, J=5.9 Hz, 2H), 2.59–2.71 (m, 1H), 2.43
(t, J=6.4 Hz, 2H), 2.36 (dd, J=13.6, 3.9 Hz, 1H), 2.26 (s, 6H), 2.01–
A genetic algorithm (GA) employing GOLDscore as the fitness func-
tion was used for all docking calculations. Ten independent GA
runs were performed for each ligand. Default settings were re-
tained for the “ligand flexibility”, “fitness and search options” and
the “GA settings”, however the “allow early termination” setting in
ꢂ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2013, 8, 709 – 718 716

CHEMMEDCHEM
COMMUNICATIONS
www.chemmedchem.org
the “fitness and search option” was turned off to search a larger
number of ligand conformations. In order to calculate HAC-normal-
ised GOLDscore, the HAC value was calculated using the “element
count (advanced)” component with the Pipeline Pilot.[43]
distance at which clusters are allowed to merge in an agglomera-
tive algorithm. In order to judge the optimal number of clusters,
the clustering quality metrics DBI, pSF and SSR/SST were moni-
tored. The final decision on the optimal number of clusters
involved simultaneous consideration of the clustering quality met-
rics and manual inspection of the superposed cluster centroids.
Molecular dynamics (MD) simulations: Ligand complexes for
compounds 2, 3, 5, 20 and NET were simulated at 300 K using MD.
The starting configurations for each of the ligand–
d(CGCGAATTCGCG) complexes were taken as the best scoring
docking pose for that ligand. All MD simulations were carried out
using the AMBER 11 MD package.[47] All ligands were modelled
using the general amber force field (GAFF),[48] and the DNA duplex
was modelled using the ff99SB force field.[49] Atomic charges for all
ligands were obtained using electrostatic potential calculations at
the Hartree–Fock ab initio level of theory[50] using the locally dense
6–31G* basis set, similar to the previous MD study of NET.[51] Elec-
trostatic potential calculations were performed using Gaussian 09
(revision B.01)[52] as in a previous MD study of NET.[51]
MM/PBSA calculations: MM/PBSA calculations were performed
using the MM/PBSA.py tool within AmberTools, with default pa-
rameters on 1000 snapshots from the production phase of each
simulation sampled at 2 ps intervals to ensure that snapshots were
unrelated.
Acknowledgements
The authors would like to thank the British Engineering and Phys-
ical Sciences Research Council (EPSRC) for a studentship to LP
and the EU FP6 Malaria Initiative ANTIMAL for partial support of
this work.
The LEaP module within AmberTools was used to prepare each
ligand–DNA duplex system for MD simulation. All ligand–DNA
complexes have a large net negative charge despite the ligands
being cationic or dicationic at pH 7.4; this is due to the presence
ꢁ
of 22PO4 groups in the DNA duplexes. In order to neutralise the
Keywords: artemisinin · binding studies · cytotoxicity · DNA ·
molecular modelling
ligand–duplex complexes, an appropriate number of Na+ counter
ions was added to each system using TIP3P;[53] a water box of 12 ꢃ
distance from the edge atom of the DNA–ligand complex was con-
structed.
[1] Guidelines for the Treatment of Malaria, 2nd Edition, World Health Or-
ganisation (WHO), 2010; http://whqlibdoc.who.int/publications/2010/
9789241547925_eng.pdf.
[2] A. Ghantous, H. Gali-Muhtasib, H. Vuorela, N. A. Saliba, N. Darwiche,
Drug Discovery Today 2010, 15, 668–678.
[3] a) P. M. O’Neill, V. E. Barton, S. A. Ward, Molecules 2010, 15, 1705–1721;
b) P. A. Stocks, P. G. Bray, V. E. Barton, M. Al-Helal, M. Jones, N. V. Araujo,
P. Gibbons, S. A. Ward, R. H. Hughes, G. A. Biagini, J. Davies, R. K.
Amewu, A. E. Mercer, G. Ellis, P. M. O’Neill, Angew. Chem. 2007, 119,
6394–6399; Angew. Chem. Int. Ed. 2007, 46, 6278–6283.
[4] A. E. Mercer, J. L. Maggs, X. Sun, G. M. Cohen, J. Chadwick, P. M. O’Neill,
B. K. Park, J. Biol. Chem. 2007, 282, 9372–9382.
[5] I. Nakase, H. Lai, N. P. Singh, T. Sasaki, Int. J. Pharm. 2008, 354, 28–33.
[6] T. Efferth, Drug Resist. Updates 2005, 8, 85–97.
[7] T. Efferth, Curr. Drug Targets 2006, 7, 407–421.
[8] T. Efferth, F. Oesch, Biochem. Pharmacol. 2004, 68, 3–10.
[9] L. Anfosso, T. Efferth, A. Albini, U. Pfeffer, Pharmacogenomics J. 2006, 6,
269–278.
[10] E. Aldieri, D. Atragene, L. Bergandi, C. Riganti, C. Costamagna, A. Bosia,
D. Ghigo, FEBS Lett. 2003, 552, 141–144.
[11] H. H. Cai, P. H. Yang, J. N. Chen, Z. H. Liang, Q. Y. Chen, J. Y. Cai, Electro-
chim. Acta 2009, 54, 3651–3656.
[12] P. C. H. Li, E. Lam, W. P. Roos, M. Z. Zdzienicka, B. Kaina, T. Efferth, Cancer
Res. 2008, 68, 4347–4351.
[13] D. E. Wemmer, P. B. Dervan, Curr. Opin. Struct. Biol. 1997, 7, 355–361.
[14] P. G. Baraldi, D. Preti, F. Fruttarolo, M. A. Tabrizi, R. Romagnoli, Bioorg.
Med. Chem. 2007, 15, 17–35; L. Banfi, A. Basso, E. Bevilacqua, V. Gandol-
fo, G. Giannini, G. Guanti, L. Musso, M. Paravidino, R. Riva, Bioorg. Med.
Chem. 2008, 16, 3501–3518; H. Tanaka, Y. Tanaka, M. Minoshima, S. Ya-
maguchi, S. Fuse, T. Doi, S. Kawauchi, H. Sugiyama, T. Takahashi, Chem.
Commun. 2010, 46, 5942–5944.
To achieve the most stable and representative simulations, previ-
ously described protocols were adapted to our specific require-
ments.[26,54,55] Prior to MD simulations, each system was subjected
to a staged minimisation using the sander module in AMBER 11.
During the first stage, a restrained minimisation was performed in
order to relax the water around each ligand–duplex system. Re-
straints of 500.00 kcalmolꢁ1 ꢃ2 were applied to the DNA duplex
and ligand. The steepest descent algorithm was run for 4000 steps,
followed by the conjugate gradient method for 4000 steps. In the
second minimisation stage, the entire system underwent 2000
steps of steepest descent and 3000 steps of conjugate gradient
minimisation. For all minimisations and MD simulations, a nonbond-
ed cut-off of 10 ꢃ, periodic boundary conditions, and particle mesh
Ewald treatment of long-range electrostatics were used.
For all MD simulations, hydrogen–heavy atom bonds were con-
strained by the SHAKE algorithm, and temperature was controlled
by Langevin dynamics with a collision frequency of 1 psꢁ1 and set
to 300 K. The leapfrog algorithm was used to propagate the
system, with a time step of 0.002 ps (as shown to be appropriate
for a system that employs the SHAKE algorithm). A total of 36 ns
of simulations was performed for each ligand–DNA duplex com-
plex. Initially each system underwent a heating phase of 100 ps in
the canonical ensemble (NVT), thereafter simulations were run in
the constant-pressure canonical ensemble (NPT). Each ligand–DNA
duplex system was equilibrated for 15.9 ns, and then a production
phase of 20 ns was performed.
[15] M. Jones, A. E. Mercer, P. A. Stocks, L. J. I. La Pensꢀe, R. Cosstick, B. K.
Park, M. E. Kennedy, I. Piantanida, S. A. Ward, J. Davies, P. G. Bray, S. L.
Rawe, J. Baird, T. Charidza, O. Janneh, P. M. O’Neill, Bioorg. Med. Chem.
Lett. 2009, 19, 2033–2037.
[16] S. Nishijima, K. Shinohara, T. Bando, M. Minoshima, G. Kashiwazaki, H.
Sugiyama, Bioorg. Med. Chem. 2010, 18, 978–983.
[17] X. Cai, P. J. Gray, Jr., D. D. Von Hoff, Cancer Treat. Rev. 2009, 35, 437–450.
[18] C. Bailly, J. B. Chaires, Bioconjugate Chem. 1998, 9, 513–538.
[19] P. M. O’Neill, L. P. Bishop, R. C. Storr, S. R. Hawley, J. L. Maggs, S. A. Ward,
B. K. Park, J. Med. Chem. 1996, 39, 4511–4514.
Average-linkage clustering of MD trajectories: The average-link-
age clustering algorithm as implemented in “ptraj” with Amber-
Tools was used to find representative binding poses for each
ligand–DNA duplex system, as recommended for systems for
which the optimal number of clusters is not known.[56] The repre-
sentative binding pose is defined as the structure closest to the
centroid of each cluster. In order to find the optimal cluster,
a manual scan for the critical distance was employed from 1.0 to
2.4 ꢃ. The critical distance is defined as the maximum intercluster
ꢂ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2013, 8, 709 – 718 717

CHEMMEDCHEM
COMMUNICATIONS
www.chemmedchem.org
[20] P. M. O’Neill, A. Miller, L. P. D. Bishop, S. Hindley, J. L. Maggs, S. A. Ward,
S. M. Roberts, F. Scheinmann, A. V. Stachulski, G. H. Posner, B. K. Park, J.
Med. Chem. 2001, 44, 58–68.
[43] Pipeline Pilot, Accelrys, Inc., San Diego, USA; http://accelrys.com/
products/pipeline-pilot/.
[44] T. A. Halgren, J. Comput. Chem. 1996, 17, 490–519.
[21] J. Chadwick, A. E. Mercer, B. K. Park, R. Cosstick, P. M. O’Neill, Bioorg.
Med. Chem. 2009, 17, 1325–1338.
[45] Spartan 08, Wavefunction, Inc., Irvine, USA, 2008; http://www.wavefun.
com.
[22] R. Zhao, J. W. Lown, Heterocycles 1995, 41, 337–344; E. E. Baird, P. B.
Dervan, J. Am. Chem. Soc. 1996, 118, 6141–6146.
[46] Hermes Visualiser, version 1.4.1, Cambridge Crystallographic Database,
Cambridge, UK, 2008.
[23] D. S. Goodsell, M. L. Kopka, R. E. Dickerson, Biochemistry 1995, 34,
4983–4993.
[24] M. Sriram, G. A. Van der Marel, H. L. P. F. Roelen, J. H. Van Boom, A. H. J.
Wang, Biochemistry 1992, 31, 11823–11834.
[25] GOLD docking suite: G. Jones, P. Willett, R. C. Glen, J. Mol. Biol. 1995,
245, 43–53. G. Jones, P. Willett, R. C. Glen, A. R. Leach, R. Taylor, J. Mol.
Biol. 1997, 267, 727–748. M. L. Verdonk, J. C. Cole, M. J. Hartshorn, C. W.
Murray, R. D. Taylor, Proteins: Struct. Funct. Genet. 2003, 52, 609–623.
[26] Minor groove binders study: H. K. Srivastava, M. Chourasia, D. Kumar,
G. N. Sastry, J. Chem. Inf. Model. 2011, 51, 558–571.
[47] D. A. Case, T. A. Darden, T. E. Cheatham, III, 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. Goetz, I. Kolossvꢅry, K. F.
Wong, F. Paesani, 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. Mathews, M. G. Seetin, R. Salomon-Ferrer, C. Sagui, V. Babin,
T. Luchko, S. Gusarov, A. Kovalenko, P. A. Kollman, AMBER 11, University
of California, San Francisco, USA, 2010; http://ambermd.org/.
[48] J. M. Wang, R. M. Wolf, J. W. Caldwell, P. A. Kollman, D. A. Case, J. Comp.
Chem. 2004, 25, 1157–1174.
[27] Y. Li, J. Shen, X. Sun, W. Li, G. Liu, Y. Tang, J. Chem. Inf. Model. 2010, 50,
1134–1146.
[49] V. Hornak, R. Abel, A. Okur, B. Strockbine, A. Roitberg, C. Simmerling,
Proteins: Struct. Funct. Bioinf. 2006, 65, 712–725.
[28] The Pymol Molecular Graphics System, Version 1.3r1, Schrçdinger, LLC,
New York, USA, 2010.
[29] M. L. Kopka, C. Yoon, D. Goodsell, P. Pjura, R. E. Dickerson, J. Mol. Biol.
1985, 183, 553–563.
[30] I. D. Kuntz, K. Chen, K. A. Sharp, P. A. Kollman, Proc. Natl. Acad. Sci. USA
1999, 96, 9997–10002.
[31] A. L. Hopkins, C. R. Groom, A. Alex, Drug Discovery Today 2004, 9, 430–
431.
[32] C. Abad-Zapatero, J. T. Metz, Drug Discovery Today 2005, 10, 464–469.
[33] J. A. Wells, C. L. McClendon, Nature 2007, 450, 1001–1009.
[34] C. Hetꢀnyi, U. Maran, A. T. Garcꢁa-Sosa, M. Karelson, Bioinformatics 2007,
23, 2678–2685.
[35] A. T. Garcꢁa-Sosa, S. Sild, U. Maran, QSAR Comb. Sci. 2009, 28, 815–821.
[36] A. T. Garcꢁa-Sosa, C. Hetꢀnyi, U. Maran, J. Comput. Chem. 2010, 31, 174–
184.
[37] H. Wang, C. A. Laughton, Phys. Chem. Chem. Phys. 2009, 11, 10722–
10728.
[38] D. J. Patel, Proc. Natl. Acad. Sci. USA 1982, 79, 6424–6428.
[39] W. Trager, J. Jensen, Science 1976, 193, 673–675.
[40] R. E. Desjardins, C. J. Canfield, J. D. Haynes, J. D. Chulay, Antimicrob.
Agents Chemother. 1979, 16, 710–718.
[50] a) D. R. Hartree, Proc. Cambridge Philos. Soc. 1928, 24, 89–110; b) V.
Fock, Z. Phys. 1930, 61, 126–148; c) U. C. Singh, P. A. Kollman, J.
Comput. Chem. 1984, 5, 129–145.
[51] W. L. Jorgensen, J. D. Madura, Mol. Phys. 1985, 56, 1381–1392.
[52] Gaussian 09, Revision A1, M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E.
Scuseria, M. A. Robb, J. R. Cheeseman, G. Scalmani, V. Barone, B. Men-
nucci, G. A. Petersson, H. Nakatsuji, M. Caricato, X. Li, H. P. Hratchian,
A. F. Izmaylov, J. Bloino, G. Zheng, J. L. Sonnenberg, M. Hada, M. Ehara,
K. Toyota, R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y. Honda, O.
Kitao, H. Nakai, T. Vreven, J. A. Montgomery, Jr., J. E. Peralta, F. Ogliaro,
M. Bearpark, J. J. Heyd, E. Brothers, K. N. Kudin, V. N. Staroverov, R. Ko-
bayashi, J. Normand, K. Raghavachari, A. Rendell, J. C. Burant, S. S. Iyen-
gar, J. Tomasi, M. Cossi, N. Rega, J. M. Millam, M. Klene, J. E. Knox, J. B.
Cross, V. Bakken, C. Adamo, J. Jaramillo, R. Gomperts, R. E. Stratmann,
O. Yazyev, A. J. Austin, R. Cammi, C. Pomelli, J. W. Ochterski, R. L. Martin,
K. Morokuma, V. G. Zakrzewski, G. A. Voth, P. Salvador, J. J. Dannenberg,
S. Dapprich, A. D. Daniels, ꢆ. Farkas, J. B. Foresman, J. V. Ortiz, J. Cio-
slowski, D. J. Fox, Gaussian Inc., Wallingford, USA, 2009.
[53] B. Wellenzohn, R. H. Winger, A. Hallbrucker, E. Mayer, K. R. Liedl, J. Am.
Chem. Soc. 2000, 122, 3927–3931.
[54] M. A. Young, G. Ravishanker, D. L. Beveridge, Biophys. J. 1997, 73, 2313–
2336.
[55] M. A. Young, B. Jayaram, D. L. Beveridge, J. Am. Chem. Soc. 1997, 119,
59–69.
[56] J. Shao, S. W. Tanner, N. Thompson, T. E. Cheatham III, J. Chem. Theory
Comput. 2007, 3, 2312–2334.
[41] A. E. Mercer, I. M. Copple, J. L. Maggs, P. M. O’Neill, B. K. Park, J. Biol.
Chem. 2011, 286, 987–996.
[42] M. J. Gardner, N. Hall, E. Fung, O. White, M. Berriman, R. W. Hyman, J. M.
Carlton, A. Pain, K. E. Nelson, S. Bowman, I. T. Paulsen, K. James, J. A.
Eisen, K. Rutherford, S. L. Salzberg, A. Craig, S. Kyes, M.-S. Chan, V. Nene,
S. J. Shallom, B. Suh, J. Peterson, S. Angiuoli, M. Pertea, J. Allen, J. Selen-
gut, D. Haft, M. W. Mather, A. B. Vaidya, D. M. A. Martin, A. H. Fairlamb,
M. J. Fraunholz, D. S. Roos, S. A. Ralph, G. I. McFadden, L. M. Cummings,
G. M. Subramanian, C. Mungall, J. C. Venter, D. J. Carucci, S. L. Hoffman,
C. Newbold, R. W. Davis, C. M. Fraser, B. Barrell, Nature 2002, 419, 498–
511.
Received: November 19, 2012
Published online on March 11, 2013
ꢂ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2013, 8, 709 – 718 718