Targeting a dynamic protein-protein interaction: Fragment screening against the malaria myosin a motor complex

Article abstract of DOI:10.1002/cmdc.201402357

Motility is a vital feature of the complex life cycle of Plasmodium falciparum, the apicomplexan parasite that causes human malaria. Processes such as host cell invasion are thought to be powered by a conserved actomyosin motor (containing myosin A or myoA), correct localization of which is dependent on a tight interaction with myosin A tail domain interacting protein (MTIP) at the inner membrane of the parasite. Although disruption of this protein-protein interaction represents an attractive means to investigate the putative roles of myoA-based motility and to inhibit the parasitic life cycle, no small molecules have been identified that bind to MTIP. Furthermore, it has not been possible to obtain a crystal structure of the free protein, which is highly dynamic and unstable in the absence of its natural myoA tail partner. Herein we report the de novo identification of the first molecules that bind to and stabilize MTIP via a fragment-based, integrated biophysical approach and structural investigations to examine the binding modes of hit compounds. The challenges of targeting such a dynamic system with traditional fragment screening workflows are addressed throughout.

Full text of DOI:10.1002/cmdc.201402357

DOI: 10.1002/cmdc.201402357
Full Papers
Targeting a Dynamic Protein–Protein Interaction:
Fragment Screening against the Malaria Myosin A Motor
Complex
Christopher H. Douse,*[a, b, c] Nina Vrielink,[a] Zhang Wenlin,[a] Ernesto Cota,[b, c] and
Edward W. Tate*[a, c]
Motility is a vital feature of the complex life cycle of Plasmodi-
um falciparum, the apicomplexan parasite that causes human
malaria. Processes such as host cell invasion are thought to be
powered by a conserved actomyosin motor (containing myosi-
n A or myoA), correct localization of which is dependent on
a tight interaction with myosin A tail domain interacting pro-
tein (MTIP) at the inner membrane of the parasite. Although
disruption of this protein–protein interaction represents an at-
tractive means to investigate the putative roles of myoA-based
motility and to inhibit the parasitic life cycle, no small mole-
cules have been identified that bind to MTIP. Furthermore, it
has not been possible to obtain a crystal structure of the free
protein, which is highly dynamic and unstable in the absence
of its natural myoA tail partner. Herein we report the de novo
identification of the first molecules that bind to and stabilize
MTIP via a fragment-based, integrated biophysical approach
and structural investigations to examine the binding modes of
hit compounds. The challenges of targeting such a dynamic
system with traditional fragment screening workflows are
addressed throughout.
Introduction
There is a growing recognition that protein–protein interaction
(PPI) hotspots may be targeted by small molecules, and frag-
ment-based approaches have had success in this area.[1] When
the target is a PPI, in vitro fragment selection relies on a series
of specialized biophysical experiments to detect the weak in-
teractions typically observed between fragments and proteins.
With this in mind, Abell and co-workers recently suggested
a three-stage cascade of experiments suitable for an academic
laboratory, as follows: 1) screening of the initial library by dif-
ferential scanning fluorimetry (DSF), 2) validation of hits by
ligand-observed NMR spectroscopy, and 3) characterization of
binding affinity and mode by isothermal titration calorimetry
(ITC) and X-ray crystallography, respectively.[2] In particular,
high-resolution co-crystal structures of bound hits are usually
essential to inform subsequent elaboration by chemical syn-
thesis.[3] However, the implementation of biophysical assays is
dependent on the target in question.
We sought to employ a fragment-based approach toward
targeting the interaction between the C-terminal “tail” of Plas-
modium falciparum myosin A (PfmyoA) and binding partner
myoA tail domain interacting protein (PfMTIP). This PPI is
thought to be central to the malaria parasite’s ability to glide
and invade host cells in a defined, directional manner, as cor-
rect localization of the myoA motor protein depends on its
interaction with MTIP (called myosin light chain 1, MLC1, in
other apicomplexan parasites such as Toxoplasma gondii).[4] In
the prevailing model of red blood cell invasion by Plasmodium
(Figure 1A),[5] myoA is anchored to a “membranous network of
flattened vesicles”[6] termed the inner membrane complex
(IMC) via interaction with MTIP, which itself is thought to inter-
act with “gliding-associated proteins” (GAP45 and GAP50) and
is palmitoylated at its N terminus, enabling IMC attachment.[7]
The action of the myoA motor is coupled to ATP hydrolysis;
the force produced enables the myosin head domain to tread
along short actin filaments toward the barbed (+) end. This is
thought to translate into overall movement of the parasite via
another set of PPIs. Tetrameric aldolase has been implicated in
joining the filament to the so-called “tight junction” via the cy-
toplasmic tails of thrombospondin-related anonymous protein
(TRAP) adhesins[8] and apical membrane antigen 1 (AMA1); the
latter protein interacts with rhoptry neck protein 2 (RON2),
which is inserted into the host cell plasma membrane.[9] Pepti-
dic and, more recently, small-molecule inhibitors of the AMA1–
RON2 interaction have been shown to result in a block of para-
site invasion, suggesting that PPIs involved in the assembly
may represent new antimalarial drug targets.[10]
[a] Dr. C. H. Douse, N. Vrielink, Z. Wenlin, Prof. E. W. Tate
Department of Chemistry, Imperial College London
South Kensington, London SW7 2AZ (UK)
E-mail: christopher.douse@googlemail.com
e.tate@imperial.ac.uk
[b] Dr. C. H. Douse, Dr. E. Cota
Centre for Structural Biology, Department of Life Sciences
Imperial College London, South Kensington, London SW7 2AZ (UK)
[c] Dr. C. H. Douse, Dr. E. Cota, Prof. E. W. Tate
Institute of Chemical Biology, Imperial College London
South Kensington, London SW7 2AZ (UK)
Supporting information for this article is available on the WWW under
http://dx.doi.org/10.1002/cmdc.201402357.
ꢀ 2014 The Authors. Published by Wiley-VCH Verlag GmbH & Co. KGaA.
This is an open access article under the terms of the Creative Commons
Attribution License, which permits use, distribution and reproduction in
any medium, provided the original work is properly cited.
ChemMedChem 2015, 10, 134 – 143
134 ꢀ 2015 The Authors. Published by Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
physiological, “extended” crystal
structure of P. knowlesi MTIP.[12]
Although several of the hit com-
pounds from this screen inhibit-
ed parasite growth with EC50
values on the order of 10À7 m,
and a subset perturbed gliding
motility of Plasmodium liver
stages, all the hit compounds
were highly hydrophobic, which
hampered attempts to obtain
biophysical evidence for MTIP
binding. Furthermore, none of
the inhibitors formed hydrogen
bonds with MTIP residues in the
docked protein–ligand com-
plexes, suggesting that specifici-
ty of the molecules for the in-
tended target may be low. Re-
cently, a small-molecule inhibitor
of T. gondii motility and invasion
called
tachypleginA-2
was
shown to target homologous
TgMLC1, covalently modifying
a cysteine in the disordered N-
terminal extension of the pro-
tein.[13] This cysteine is not con-
served in Plasmodium MTIPs, so
the mode of action is unlikely to
apply to malaria. Nonetheless,
the latter work suggests that
modulation of MTIP with small
molecules may be disruptive to
the parasite.
In our hands it has not been
possible to crystallize the dy-
namic myoA binding construct
of PfMTIP (residues 61–204,
herein termed PfMTIPD60) in the
absence of a myoA tail peptide;
according to a 2013 report by
Khamrui et al., this has also been
the case for full-length PfMTIP
and the CTD in isolation.[15] En-
couragingly, the authors were
able to use a nanobody as a crys-
tallization chaperone and to
obtain a high-resolution struc-
ture of the CTD. However, the
Figure 1. A) Linear model of the Plasmodium motor complex and tight junction in the context of putative roles in
erythrocyte invasion by blood stage parasites (merozoites). A number of the key proteins and protein–protein
interactions in this system are shown. MTRAP=merozoite-specific thrombospondin-related anonymous protein,
AMA1=apical membrane antigen 1, GAP(M)=gliding-associated protein (with multiple membrane spans),
RON=rhoptry neck protein, MTIP=myosin A tail domain interacting protein, IMC=inner membrane complex,
PM=plasma membrane. Co-/post-translational lipid attachments are indicated on MTIP and GAP45 with black
lines. B) Structural and thermodynamic features of the wild-type MTIP–myoA tail interaction.[14] Shown are the an-
notated crystal structure (PfMTIPD60 shown as white cartoon and white surface, shaded by a white–red ramp ac-
cording to chemical shift perturbations induced by binding to PfmyoA(799–818), which is shown as black cartoon
and sticks) and thermodynamic parameters according to ITC experiments.
We and others have developed peptidic inhibitors of the
PfMTIP–myoA complex by mimicking the C-terminal tail of
myoA, which adopts a helical conformation and is “clamped”
by two domains of MTIP (N- and C-terminal domains, NTD and
CTD), burying >2000 ꢁ2 of solvent-exposed protein surface
upon complex formation.[11] However, no small molecules have
been shown experimentally to bind to MTIP. An in silico dock-
ing study was conducted by Kortagere et al. using the non-
accessibility of the myoA binding groove is altered substantial-
ly by contacts between the nanobody and the MTIP construct,
which may compromise subsequent efforts to discover effec-
tive small-molecule ligands. We have taken a different ap-
proach by studying isotopically labeled PfMTIPD60 (containing
both MTIP domains) in solution using multidimensional NMR
1
techniques, and recently reported assignments of the H,15N-
HSQC spectra of both the “free” and “bound” states, along
ChemMedChem 2015, 10, 134 – 143
www.chemmedchem.org
135 ꢀ 2015 The Authors. Published by Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
with a detailed analysis of the conformational dynamics and
thermodynamics of the system.[14] It was found that the free
protein is highly flexible on a range of timescales from milli-
second to sub-nanosecond exchange processes, which may ex-
plain the difficulty in crystallizing the protein in isolation. Our
analysis also included the mapping of MTIP–myoA interaction
hotspots from chemical shift perturbations (CSPs) upon clamp-
ing of the myoA tail (Figure 1B). Interestingly, we found that
a phosphomimetic mutation of Ser108 in the NTD caused
a pronounced structural change in the complex and a substan-
tial decrease in affinity for myoA, even though this residue has
been shown not to be in direct contact with any residue in the
myoA tail. We concluded that the intramolecular clamp be-
tween Ser108 and Asp173 (from the CTD) is broken upon
such a modification. This observation was primarily of biologi-
cal interest, as the residue has been annotated as a phosphory-
lation site in cultured blood-stage malaria parasites,[16] but was
also encouraging in relation to ligand discovery; small mole-
cules will fail to span the large buried PPI surface area, but our
analysis suggested that breaking individual hotspot inter-
actions such as the Ser108–Asp173 clamp may nonetheless be
inhibitory.
The hit rate using this cutoff is remarkably high (26%). It has
been previously recognized that entropy-driven hydrophobic
interactions generally result in a large DTm,[17] and this may be
driven in the present case by exposed hydrophobic patches in
the PfMTIPD60 CTD. In view of the primary data, it was decid-
ed to increase the threshold to +2.08C in defining an initial
hit. Furthermore, the hits were retested using DSF to improve
confidence in the measured values. From 500 compounds, 40
were selected for retesting.
Of these 40 fragments, 15 were discounted upon retesting
as they displayed a DTm <2.08C and/or inconsistent results
(standard deviation >1.08C, n=4), ten gave DTm between 2.0
and 4.08C, and 15 gave DTm >4.08C consistently. The latter 15
fragments (for which DSF data are shown in Figure 2) were se-
lected for further testing. Substituted aniline 1 and 4-phenoxy-
phenol 2 caused the largest thermal shifts of the library, with
remarkable DTm values of +10.38C and +9.08C—just 5–68C
lower than the positive control PfmyoA tail peptide despite
the fragments being an order of magnitude smaller in molecu-
lar weight. Disubstituted furan 8 not only shifted the unfolding
temperature of the protein, but also caused a marked sharpen-
ing of the melting curve, suggesting a significant change in
the dynamics of MTIP unfolding. Of the top 25 fragments,
seven contained one amine predicted to bear a positive
charge in the assay buffer (pH 7.5) with a further two contain-
ing a second amine with pKa >7.5. In contrast, only one frag-
ment hit is predicted to bear negative charge, with the remain-
ing 15 hits neutral at this pH. PfMTIPD60 has a predicted iso-
electric point of 4.4, and the predominance of negatively
charged residues in the sequence suggests an over-representa-
tion of amines may be expected. Otherwise, it is difficult to
pick out clear trends for the hit compounds using the relatively
crude measure of DTm.
Results and Discussion
Initial screening by DSF
With the structural and dynamic features of free MTIP in mind,
and given the lack of small-molecule ligands for the protein,
we used differential scanning fluorimetry (DSF) to screen a li-
brary of 500 fragments for the ability to stabilize PfMTIPD60
with respect to thermal unfolding. We have previously shown
that the protein is thermally unstable in isolation, giving a melt-
ing temperature (Tm) of 38.08C. It was also notable that the
melting transition is rather shallow, interpreted as being due
to extensive conformational flexibility, also reflected in NMR
DSF titrations
It was of interest to investigate whether the stabilization of
PfMTIPD60 observed in the initial DSF screen was dependent
on the fragment concentration. Therefore, “titration” experi-
ments were performed, in which fragment concentrations
were varied from 125 to 2000 mm. Interestingly, several types
of behavior were observed (Figure 3). For some fragment hits,
the stabilizing effect was saturable, that is, DTm increased with
and ITC data.[14]
.
The error associated with the DSF screen was estimated by
taking the standard deviation of the control sample Tm values,
that is, the negative controls (5 mm PfMTIPD60 +1% DMSO)
and the positive controls (5 mm PfMTIPD60 +200 mm PfmyoA
tail peptide). Across the 48 measurements of these samples,
the standard deviations for these measurements were 0.43 and
0.52. Using a two-standard-deviation cutoff reported recently
for the screen against humanized RadA,[1c] we set an initial hit
cutoff value of +1.08C. The results from the screen, run in
duplicate, are shown in Table 1.
fragment concentration and reached
a maximal value
(‘type 1’). Several other hits showed a linear variation (‘type 2’)
and some showed a marked jump to a large DTm at high con-
centrations, but were barely or not at all stabilizing at lower
concentrations (‘type 3’). Data for the 15 fragment hits are
summarized in Table 2.
The effect of ligand concentration on protein stabilization
and its relationship with binding affinity is complex and has
been studied theoretically and experimentally.[18] It has been
concluded that, in general, for 1:1 protein–ligand binding and
a simple two-state approximation of the unfolding transition,
DTm should correlate reasonably with both affinity and ligand
concentration. Furthermore, while the change in Tm may in-
crease sharply as the protein is saturated by the ligand (partic-
Table 1. Results of initial DSF screening of the Maybridge Ro3 500-frag-
ment library against PfMTIPD60.
Set A
Set B
Both
+ve (DTm >1.08C)
134
295
18
126
307
22
112
241
13
no change (1.08CꢀDTm ꢀÀ1.08C)
Àve (DTm <À1.08C)
ambiguous melting curve
53
45
44
ChemMedChem 2015, 10, 134 – 143
www.chemmedchem.org
136 ꢀ 2015 The Authors. Published by Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
significant increase in DTm
should be observed for a ligand
at a concentration well below its
KD, which may lead to type 3 be-
havior for weaker-binding frag-
ments. The highest affinity hits
should, in principle, give the
largest DTm at a given concen-
tration, as is commonly as-
sumed, but may also be ranked
according to titration behavior,
with type 1>2>3.
There are several caveats in re-
lation to these arguments, par-
ticularly for the target we are ad-
dressing. Firstly, the “foldedness”
of PfMTIPD60 at room tempera-
ture is complex because the pro-
tein is highly flexible, and the
unfolding transition is shallow,
suggesting that the protein sam-
ples
a number of conforma-
tions.[14] As a result, the existing
equations for modeling and pre-
dicting the effects of ligand
binding to globular proteins in
the thermal shift assays may be
insufficient for such a target, and
DTm may correlate more poorly
with affinity, particularly be-
tween different classes of mole-
cules. The latter also appears to
be the case for the diverse tar-
gets currently under investiga-
tion in our laboratories (data not
shown). For instance, in the pres-
ent case, it is improbable that
fragment 1 has an affinity ap-
proaching that of the positive
control PfmyoA tail peptide
(KD =351 nm by ITC)[14] despite
similar DTm values (+10.1 and
+15.38C, respectively). In the
case of myoA peptide binding,
Figure 2. Normalized melting curves for PfMTIPD60 (5 mm) in the presence of 1% DMSO (grey) and the top 15
fragment hits (2 mm) and controls.
a
sharpening of the melting
curve is associated with a major
change in the structure and dy-
namics of the protein, that is,
ularly for tight binding) the increase of protein Tm with ligand
concentration may continue after the protein (in its native
state) is fully saturated.[18a] However, limits on ligand solubility
or increased binding of the ligand to the denatured state may
result in a saturation point, that is, type 1 behavior. Indeed, sta-
bilization of the denatured state has been noted as a possible
source of false negatives in DSF screens, as weak binding to
the native state may be masked.[19] Another conclusion—work-
ing under the same two-state unfolding regime—is that no
the unfolding energy landscapes of the two protein–ligand
complexes are likely to be qualitatively different. Furthermore,
the stoichiometry and reversibility of the MTIP unfolding transi-
tion are unknown, as is the character of the unfolded state of
the protein. These concerns underline the particular impor-
tance of the validation stage of the screen, which is described
in the next section.
ChemMedChem 2015, 10, 134 – 143
www.chemmedchem.org
137 ꢀ 2015 The Authors. Published by Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
mostly systematic and very minor such that the published
peak assignments[14] could be adjusted accordingly and used
with confidence. A negative control fragment was also ana-
lyzed in the same way. The pH of each of the samples (15 frag-
ments and controls) was measured before the spectra were ac-
quired, as small variations in pH may affect chemical shifts.
CSPs were noted if the backbone amide resonance was shifted
beyond its linewidth in either the 1H or 15N dimension
(Table 2).
Overall, the CSPs observed in most of the protein–fragment
HSQCs were qualitatively very small, which was surprising
given the large values of DTm obtained, and suggested either
very weak interactions or no interactions with the fragments
under the experimental conditions (Supporting Information
Figure S1). Several residues appeared to be affected in nearly
every sample; indeed, small peak shifts were observed in the
negative control sample (DTm = +0.18C) for Asp76, Ile133,
and Ala149, which are presumed not to be caused by a direct
interaction with the fragment. Chemical shifts are highly sensi-
tive to pH in general—and residues lying close to the histidine
residues in PfMTIPD60 (His136, His150 and the N-terminal His
tag) may be particularly affected due to the pKa of the imida-
zole side chain; changes in the position of the equilibrium be-
tween protonated and unprotonated states will affect the local
chemical environment and thus the chemical shifts of nearby
nuclei. Two of the negative control peak shifts (Ile133 and
Ala149) are thus likely due to pH, as the sample pH was 7.02,
relative to 7.19 for the DMSO control. It is unclear why Asp76,
the amide chemical shift of which is affected in nearly every
sample, should be likewise perturbed. Nonetheless, CSPs ob-
served for these three amides were deemed to be nonspecific.
The other indicator used to classify CSPs arising from a specific
interaction was whether the affected amides were from a con-
tiguous stretch in the primary MTIP sequence, although it
should be noted that the lack of a structure for free PfMTIPD60
Figure 3. Examples of four different types of behavior in DSF titration experi-
ments. DTm values shown are the mean value of triplicate experiments, with
error bars illustrating Æ one standard deviation. For each fragment concen-
tration, DTm was referenced to the Tm value of a control in which the protein
was incubated with the corresponding concentration of DMSO.
Validation of fragment binding by protein-detected NMR
Given the small size, domain structure and multifarious dynam-
ics of isolated PfMTIPD60, protein-detected experiments were
deemed more appropriate than ligand-detected NMR experi-
ments in the validation of initial fragment hits.[20] Binding of
the 15 hit fragments to PfMTIPD60 was monitored by single-
1
point H,15N-HSQC experiments, in which the spectra of 50 mm
15N-labeled PfMTIPD60 were recorded in the presence of
50 molar equivalents (2.5 mm) of the various fragments. These
were compared with the spectrum of the protein in the pres-
ence of 1.25% DMSO, to ensure that CSPs observed were not
due to the presence of the solvent used to solubilize the frag-
ment stocks. Despite slight changes in chemical shift for the
protein in isolation in the presence of DMSO, these were
Table 2. Summary of DSF and NMR data for MTIP–fragment complexes.
Mean DSF titration “type”
DTm [8C]
NMR
Amide peaks shifted
Comments
sample pH MTIP–myoA hotspots[a]
DMSO
ctrl
Àve
1
2
3
4
5
6
7
–
–
7.19
–
–
+0.1 4 (random)
+10.3 1 (saturated)
+9.0 2 (linear)
7.02
7.05
7.04
7.09
7.07
7.16
7.13
7.08
7.09
7.11
7.14
7.12
7.14
6.94
7.10
7.11
D76, I133, A149
–
D76, I133, A149, D173
I133, N140, V141, G172, D173, E190, I193
L68, E69, D73, E74, S75, D76, S107, D173
D76, A149, D173
A180, C199, D201, I202
D76, C134
nonspecific
small CSPs
validated
nonspecific
small CSPs
nonspecific
nonspecific
+7.2 2 (linear)
+6.6 1 (saturated)
+5.9 1 (saturated)
+5.9 3 (inverted)
+5.8 2 (linear)
+5.6 1 (saturated)
+5.5 3 (inverted)
+5.2 2 (linear)
+4.7 1 (saturated)
+4.6 3 (inverted)
+4.5 2 (linear)
+4.1 3 (inverted)
+4.0 3 (inverted)
D76, I133, A149
8
9
D76, V135, N140, A149, H150, T160, L168, W171, G172, D173, A174, T176, D194, C199, I202 validated
D76
nonspecific
small CSPs
nonspecific
nonspecific
nonspecific
nonspecific
nonspecific
10
11
12
13
14
15
C134, V141, A180
D76, I133, A149, D173
I133, A149, G157
D73, E74, D76, I133, C134, A149, H150, F151, D173
I133
I133
[a] Hotspots according to CSPs upon interaction with the myoA tail, illustrated in Figure 1B.[14]
ChemMedChem 2015, 10, 134 – 143
www.chemmedchem.org
138 ꢀ 2015 The Authors. Published by Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
prevents mapping of shifts in Table 2 onto a bona fide struc-
tural model of the protein.
were observed, suggesting that the interaction was not com-
petitive with the myoA tail (Supporting Information Figure S2).
Taking the above points together, several fragments were
deemed to have insignificant or nonspecific interactions with
MTIP on the basis of the NMR experiments. Notably, none of
the fragments exhibiting type 3 DSF titration behavior were va-
lidated by NMR data. However, several fragments exhibiting
type 1 or type 2 behavior were also not validated. It remains
unclear why many DSF fragment hits fail to interact specifically
with MTIP in the NMR experiment, but still appear to stabilize
the protein significantly, and in a concentration-dependent
manner. It is plausible that the fragments could stabilize the
protein by binding to a conformation of MTIP that is not pres-
ent (or is present as a minor, invisible conformation) in the
NMR tube at 303 K.
Design of peptide–fragment chimeras
Although the affinity of 8 for PfMTIPD60 was in the millimolar
range according to NMR titration data, the clear interaction
with the myoA binding hotspot loop connecting helices a6
and a7 was encouraging. With the aim of obtaining informa-
tion on the mode of fragment binding, we synthesized a small
library of derivatives based on 8 and assayed binding to MTIP
by DSF (Table 3).
Table 3. DSF analysis to establish preliminary structure–activity relation-
ships using derivatives of fragment hit 8.
Of the remaining fragments, by far the largest CSPs in MTIP
were caused by hit 8, a compound which caused not only
a positive DTm by DSF (+5.68C), but also a notable sharpening
of the melting transition, and hit 3. The interaction of 8 was
further confirmed by running an NMR titration experiment, en-
abling measurement of its binding affinity (Figure 4). The CSPs
of 8 localized to the MTIP CTD including several contiguous
MTIP–myoA hotspots, for example, Leu168, Trp171, Gly172,
Asp173, Ala174, and Thr176 that form the loop connecting
helices a6 and a7, involved in the key intramolecular clamp
with Ser108 as described above. To define whether the bind-
ing mode was competitive with the myoA tail, another NMR
experiment was carried out in which aliquots of 8 were added
to a solution containing 15N-labeled PfMTIPD60 bound to
PfmyoA(799–818). No CSPs were observed in the presence of
a large molar excess of the fragment, supporting the notion
that 8 binds weakly to a part of MTIP that is engaged in stron-
ger interactions with the myoA tail. When the same experi-
ment was performed with 3, which was shown in the initial
NMR experiments with free MTIP to cause CSPs in several resi-
dues of the NTD distal to the myoA binding groove, clear CSPs
Compd
Structure
DTm [8C][a]
8
+5.6Æ0.7
8a
+1.2Æ0.6
8b
8c
8d
À1.4Æ0.7
+3.0Æ1.3
À5.7Æ0.9
[a] Values are quoted as the mean ÆSD (n=3).
Compound 8 is a precursor to histamine receptor antagonist
ranitidine (8a). The latter gave a lower DTm (+1.28C) than 8
(+5.68C), suggesting that the primary free amine is important
to the interaction with MTIP. Re-
placement of this side of the
molecule by an alcohol (8b) se-
verely decreased binding and ac-
tually led to destabilization
(À1.48C) of the protein. Howev-
er, removal of the dimethyla-
mine moiety on the other side
was better tolerated (8c, DTm =
+3.08C). Replacement of the thi-
oether linker by an amide linker
(8d) was strongly destabilizing.
Taken together, these results
suggest that the primary amine
linked to the furan via a thioether
represent key features of 8.
We wished to examine the
1
Figure 4. Measuring the binding affinity of hit 8 using protein-detected NMR. H,15N-HSQC spectra were recorded
with 100 mm 15N-labeled PfMTIPD60 before (black) and after addition of 8 at concentrations of approximately 5
(red), 10 (green), 20 (blue), 30 (yellow), 50 (pink), 80 (cyan) and 120 molar equivalents of the protein. Perturbations
of several resonances were used to derive binding curves; three examples are highlighted. The KD shown is the
mean of the “per-residue” KD values, quoted Æ standard deviation (n=8).
binding mode of
8 in more
detail than afforded by the NMR
validation experiments, which
only gave an approximate bind-
ChemMedChem 2015, 10, 134 – 143
www.chemmedchem.org
139 ꢀ 2015 The Authors. Published by Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
ing site via backbone amide CSPs. However, previous studies
strongly suggest that to crystallize PfMTIPD60, a myoA tail
peptide is required in order to fold the protein into a stable
tertiary structure. In line with these observations, and despite
the large stabilization induced by 8 according to DSF, satura-
tion of free PfMTIPD60 with fragment 8 did not promote crys-
tal growth under any conditions tested (data not shown). How-
ever, given that one end of the molecule was not necessary for
interaction with MTIP, it was hypothesized that attachment of
the key moieties of 8 to a myoA tail peptide would enable an
X-ray structure of the bound fragment to be obtained. Because
the N-terminal part of PfmyoA(799–818) binds to the same site
as the fragment, we systematically truncated the myoA tail at
this end to create a library of chimeric molecules (Figure 5A).
Synthesis of the chimeras (chimera1–3) was achieved via
Fmoc-based solid-phase peptide synthesis, with a glycine
spacer attached to the N terminus of each myoA sequence to
permit flexibility in the orientation of the fragment. Final cou-
pling of the key functionality of 8 using Boc-protected furan-2-
carboxylic acid 8e, followed by TFA-mediated cleavage and de-
protection, yielded the desired products (Figure 5B). Three dif-
ferent myoA tail sequences were prepared (terminating at
Leu805, Leu804, and Ser803); by analysis of residue-specific
CSPs caused by the fragment and the crystal structure of the
wild-type complex with PfmyoA(799–818), it was rationalized
that these positions should provide enough space for the frag-
ment to adopt a binding conformation. Controls (control1–3)
were also prepared in which the peptides were capped with
acetyl groups.
Structures of peptide–fragment chimeras bound to MTIP
Co-crystallization and thermal stability experiments were con-
ducted for the three protein chimera complexes. Single crystals
appeared under a number of conditions in each case within
48 h. Complete datasets were collected for the complexes of
PfMTIPD60 with chimera2 and chimera3; the crystals contain-
ing chimera1 diffracted at best to low (>3.5 ꢁ) resolution and
displayed unusual spot profiles; the dataset could not be fur-
ther processed, and the structure remains unsolved. This may
suggest a drastic change in the protein structure and/or crystal
packing in the presence of this (shortest) chimera. Although
the crystals of the chimera3 and chimera2 complexes ap-
peared morphologically similar, the chimera3 complex crystal-
lized with P212121 symmetry (as for the wild-type complex) and
diffracted to 1.98 ꢁ resolution (Supporting Information
Table S1), whereas the chimera2 complex crystallized with P63
symmetry and diffracted to 2.90 ꢁ resolution (data not shown).
The decreased number of crystal contacts in the latter led to
a very high
(~70%) solvent content.
The conformation of the protein in the X-ray structures is
broadly the same as in that of the wild-type PfMTIP–myoA tail
complex.[11b,14] In the chimera2 complex, the fragment (8)
moiety points toward solvent
and does not appear to engage
in interactions with the protein.
Indeed, there is no electron den-
sity for the flexible thioether-
containing chain that terminates
in the important free primary
amine (data not shown). By con-
trast, in chimera3 the fragment-
derived
component
makes
a number of interactions with
MTIP residues, and clear elec-
tron density is evident for all
non-hydrogen
atoms
(Fig-
ure 6A). Notably, the terminal
free amine, identified as a key
feature of 8, forms a salt bridge
with the side chain of Asp173.
This interaction does not per-
turb the hydrogen bonding
clamp of the latter with Ser108,
which locks the stable protein
conformation. The thioether
moiety fits between Ile109 and
Gly172, although the furan core
does not make substantial inter-
actions with MTIP (Figure 6B).
These results are supported by
the observations that in the
case of chimera3, the Tm of the
Figure 5. A) Design of peptide–fragment chimeras based on analysis of CSPs caused by fragment hit 8 (annotated
red) and the crystal structure of the wild-type MTIP–myoA tail complex (PDB code: 4AOM). B) Scheme showing
the synthesis and generalized structure of peptide–fragment chimera molecules.
ChemMedChem 2015, 10, 134 – 143
www.chemmedchem.org
140 ꢀ 2015 The Authors. Published by Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
were conducted with control3 and chimera3. A two-
fold increase in affinity was measured (106 and
52 nm, respectively), which appeared to be driven by
an additional favorable enthalpic contribution from
the fragment of 6.8 kJmolÀ1 (Figure 6C).
Taking these data in combination with the above
NMR analysis, it is evident that 8 binds to the loop
connecting helices a6 and a7. The extent to which
the crystal structure of the chimera3 complex reflects
the orientation of free fragment 8 in this binding
event remains an open question, as its orientation is
governed to an unknown extent by its covalent at-
tachment to the myoA tail peptide. Nevertheless, the
high-resolution structure of chimera3 opens some in-
teresting possibilities for development of 8. For ex-
ample, rationally growing the fragment into the hot-
spot pockets occupied by side chains of nearby hy-
drophobic myoA residues Leu804 and Val807 may
be an approach to generate a small library of pepti-
domimetic-type inhibitors. Alternatively, further opti-
mization of interactions around the critical Asp173/
Ser108 clamp may present an opportunity to modu-
late MTIP dynamics, and influence phosphorylation
events in the vicinity of these residues that are
thought to be important for control of the motor
complex activity in vivo.[14]
Conclusions
A 500-member fragment library was screened using
DSF for interaction with MTIP, a critical determinant
of malaria parasite actomyosin motor localization
through its tight interaction with the tail domain of
myosin A, and a potential target for novel antimalari-
als or tool compounds. The hit rate was remarkably
high (26%, for DTm >18C), reflecting the low thermal
stability of the highly dynamic free MTIP construct
and the lack of well-defined tertiary structure. This
flexibility is manifest in a large loss of entropy upon
binding its natural myoA partner, and in the shallow
thermal unfolding transition suggesting a complex
pathway that is probably not well-described by exist-
ing models of two-state unfolding.
Figure 6. Structural and thermodynamic features of the interaction between MTIP and
chimera3. A) Refined 2FobsÀFcalc electron density map overlaid with the model of the
non-natural functionality in peptide–fragment chimera3. B) Final crystal structure of the
complex between PfMTIPD60 (white surface and cartoon) and chimera3 (grey cartoon;
glycine spacer shown as blue sticks, and fragment functionality shown as red sticks) and
(inset) features of the complex, focusing on the interactions made by the fragment.
C) ITC data for the interaction of PfMTIPD60 with chimera3 and acetylated peptide
control3. Data are quoted Æ standard error (n=1).
As a result of the unusually high hit rate and large
complex (55.88C) is higher than that with the corresponding
acetylated control3 peptide (54.98C), presumably due to addi-
tional interactions observed in the X-ray structure, while the re-
verse is true for the chimera2/control2 pair (53.4 and 54.58C
respectively, Supporting Information Table S2). While the frag-
ment moiety in chimera1 also confers additional thermal sta-
bility (relative to control1) on the complex with MTIP, this
could not be readily rationalized due to the absence of a struc-
ture. The Tm values of the latter complexes were much lower
(47.8 and 45.48C, respectively), suggesting that this length of
peptide sequence may be insufficient to induce the protein to
adopt the stable, clamped fold observed in the other struc-
tures. To further validate the observed effect, ITC experiments
thermal shifts observed, the cutoff was increased to +48C in
the DSF screen, leaving 15 fragments that were further tested.
Surprisingly, despite the large (up to +10.38C) DTm values and
concentration-dependent stabilization given by these mole-
cules, validation of binding by HSQC NMR experiments on
MTIP suggested that the majority of fragments had negligible
specific interactions with the protein. There are several possi-
ble explanations for this result. Firstly, it is possible that the hit
fragments are false positives in the DSF experiment due to
a destabilization of unfolded MTIP by an undetermined mecha-
nism. Alternatively, the fragments bind to one or more minor
conformation(s) of MTIP that are hardly populated at the ex-
perimental temperature (303 K) and are thus invisible in the
ChemMedChem 2015, 10, 134 – 143
www.chemmedchem.org
141 ꢀ 2015 The Authors. Published by Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
ing curves were compared with negative controls containing 1%
DMSO; positive controls contained 200 mm peptide PfmyoA(803–
818), synthesized as described previously.[11d] Where “titration” ex-
periments were performed, the concentration of DMSO in controls
was adjusted correspondingly. Experiments with peptide–fragment
chimeras and controls used 5 mm PfMTIPD60 and 200 mm peptide
under otherwise identical conditions. Fluorescence of dye was
monitored over 15–808C with a Mastercycler ep realplex real-time
PCR instrument (Eppendorf). Melting curves were analyzed by
using a customized MS Excel spreadsheet with a GraFit plug-in
(Erithacus Software) to calculate DTm for each molecule, by fitting
to the Boltzmann equation.[17]
NMR spectra. It is also possible that the affinity of the interac-
tions with the “NMR-visible” protein conformer(s) is so low that
chemical shift perturbations are undetectable. Nonetheless, at
least one genuine “hit” fragment (8) was found that binds to
the protein at the key loop connecting MTIP helices a6 and a7
in the CTD, a region that forms part of the intramolecular
clamp with Ser108 in the NTD and contains a number of
myoA tail interaction hotspots.[14] Notably, this hit was the only
molecule in the library that not only stabilized the protein to
thermal unfolding significantly (DTm = +5.68C) but also caused
a sharpening of the otherwise shallow melting transition. This
suggests that the molecule binds to the protein in such a way
as to quench a specific dynamic process and thus increase the
cooperativity of unfolding. Another hit (3) bound at a site in
the NTD that was independent of the interaction with the
myoA tail. These are the first small molecules to have been
identified and experimentally validated as MTIP binders.
Protein-detected NMR spectroscopy: For single-point fragment
validation experiments, 1H,15N-HSQC spectra were recorded in
3 mm NMR tubes (Norell) on 50 mm 15N-labeled PfMTIPD60 ex-
pressed and purified as previously described[11d,14] in buffer B
(20 mm MOPS (pH 7.0), 50 mm NaCl, 1 mm TCEP), supplemented
with 10% D2O. Fragments were tested at 2.5 mm (added from
a 50 mm stock in the same buffer) and the spectra were compared
with a control containing DMSO (1.25%). For the titration experi-
ment with 8 and the binding experiments with the preformed
MTIP–myoA complex, a 5 mm tube containing 100 mm 15N-labeled
PfMTIPD60 was used, and spectra were recorded before and after
each addition of the ligand. All experiments were conducted at
303 K on a Bruker 600 MHz Avance III spectrometer equipped with
a TCI cryoprobe. Spectra were processed in NMRPipe[21] and ana-
lyzed in NMRView.[22]
High-resolution crystallographic data on fragment binding
modes are often essential to rationally develop the affinity of
hit molecules.[3] In the case of MTIP–myoA, this has proven to
be problematic, as a crystal structure of free MTIP has not
been resolved, and NMR data point to a highly flexible protein
that samples a number of conformations, interconverting on
different timescales. The data reported above highlight a signif-
icant challenge when the target is a protein–protein interac-
tion in which the folding and stability of a protein is defined
to a large extent by the presence of its natural partner protein
or peptide. The screening and subsequent development of hits
for such dynamic targets require innovative strategies; care
must be taken in designing and interpreting the screening
campaign and interpreting the biophysical data. For example,
the high hit rate resulting from a DSF screen against a flexible
protein like MTIP requires some consideration: rather than
judging a hit solely from shift in melting temperature, it could
be more productive to search for fragments that appear to
change the cooperativity or slope of the transition, with the
view that such molecules may be able to trap a dynamic
region of the protein and thus promote particular (in this case,
ideally non-myoA binding) conformations. To obtain necessary
structural data for subsequent development, we first estab-
lished the essential structure–activity relationship of the hit,
and then synthesized novel peptide–fragment chimeras with
flexible linkers both to “trap” a crystallizable conformation, and
to probe the binding mode of the fragment functionality. This
strategy allowed us to obtain a structure that provides an in-
teresting starting point for the development of small mole-
cules targeting key features of the MTIP clamp around the
myoA motor.
Synthesis of fragment derivatives 8a–8d: Ranitidine (8a) was pur-
chased from Sigma–Aldrich and was used without further purifica-
tion. The synthesis of derivatives 8b, 8c, and 8d is described in
the Supporting Information.
Synthesis of peptide–fragment chimeras 1–3: Synthesis of chi-
mera1, chimera2, and chimera3 and intermediate 8e is described
in the Supporting Information.
X-ray crystallography: Complexes were formed between
PfMTIPD60 and the three peptide–fragment chimeras as follows:
an aliquot of the protein in buffer A (300 mL, 100 mm) was mixed
with the chimeric molecule (five molar equivalents in the case of
chimera2 and chimera3; ten molar equivalents in the case of chi-
mera1), and the resulting complex was concentrated to 50 mL
(600 mm) using a 3 kDa centrifugal concentrator (GE Healthcare) in
buffer A. PEG/Ion sitting-drop crystallization screens (Hampton Re-
search) were prepared in which the sitting drop contained 100 nL
of protein and 100 nL of reservoir solution. The resulting crystals,
which grew within 48 h at 293 K, were mounted in loops and cryo-
protected using the reservoir solution supplemented with 30%
glycerol, flash frozen in liquid nitrogen and taken to Diamond
Light Source synchrotron, Oxfordshire (UK) for data collection.
Data were collected at 100 K and with 18 oscillations, indexed in
Mosflm and scaled in Scala.[23] Initial phasing by molecular replace-
ment was carried out in Phaser[24] using the structure of wild-type
PfMTIP–myoA tail (PDB code: 4AOM) as the search model. The in-
spection of electron density maps and manual model building was
done using Coot.[25] The non-natural fragment-derived functionality
in chimera3 was built using the PRODRG server.[26] Refinement was
performed in REFMAC,[27] and model validation was done in
Phenix.[28] The coordinates and structure factors for the complex
between PfMTIPD60 and chimera3 were deposited in the RCSB
Protein Data Bank under accession code 4R1E; statistics of data
collection and refinement are shown in Supporting Information
Table S1.
Experimental Section
Differential scanning fluorimetry: Samples (20 mL) were arrayed in
96-well real-time PCR plates (Eppendorf) and contained 5 mm
PfMTIPD60 expressed and purified as previously described[11d,14] in
buffer A (20 mm HEPES (pH 7.5), 50 mm NaCl, 1 mm TCEP), supple-
mented with 10ꢂ SYPRO Orange dye (Sigma–Aldrich). 500 frag-
ments (Maybridge Ro3 library) were tested at 2 mm, and the melt-
ChemMedChem 2015, 10, 134 – 143
www.chemmedchem.org
142 ꢀ 2015 The Authors. Published by Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
[9] S. Besteiro, A. Michelin, J. Poncet, J. F. Dubremetz, M. Lebrun, PLoS
Pathog. 2009, 5, DOI: 10.1371/journal.ppat.1000309.
Isothermal titration calorimetry: Experiments were carried out at
303 K using a MicroCal iTC200 instrument (GE Healthcare). The cell
contained ~200 mL of 20 mm PfMTIPD60 in buffer A; peptide li-
gands control3 and chimera3 were added in 2 mL injections from
a 250 mm solution in buffer A. Data were fit to a single site binding
event using Origin (OriginLab Corp).
[10] a) K. S. Harris, J. L. Casey, A. M. Coley, R. Masciantonio, J. K. Sabo, D. W.
Keizer, E. F. Lee, A. McMahon, R. S. Norton, R. F. Anders, M. Foley, Infect.
Immun. 2005, 73, 6981–6989; b) P. Srinivasan, A. Yasgar, D. K. Luci, W. L.
Beatty, X. Hu, J. Andersen, D. L. Narum, J. K. Moch, H. M. Sun, J. D.
Haynes, D. J. Maloney, A. Jadhav, A. Simeonov, L. H. Miller, Nat.
Commun. 2013, 4, DOI: 10.1038/Ncomms3261.
[11] a) J. Bosch, S. Turley, T. M. Daly, S. M. Bogh, M. L. Villasmil, C. Roach, N.
Zhou, J. M. Morrisey, A. B. Vaidya, L. W. Bergman, W. G. Hol, Proc. Natl.
Acad. Sci. USA 2006, 103, 4852–4857; b) J. Bosch, S. Turley, C. M. Roach,
T. M. Daly, L. W. Bergman, W. G. Hol, J. Mol. Biol. 2007, 372, 77–88;
c) J. L. Green, S. R. Martin, J. Fielden, A. Ksagoni, M. Grainger, B. Y. S. Y.
Lim, J. E. Molloy, A. A. Holder, J. Mol. Biol. 2006, 355, 933–941; d) J. C.
Thomas, J. L. Green, R. I. Howson, P. Simpson, D. K. Moss, S. R. Martin,
A. A. Holder, E. Cota, E. W. Tate, Mol. Biosyst. 2010, 6, 494–498; e) C. H.
Douse, S. J. Maas, J. C. Thomas, J. A. Garnett, Y. Sun, E. Cota, E. W. Tate,
ACS Chem. Biol. 2014, 9, 2204–2209.
[12] S. Kortagere, W. J. Welsh, J. M. Morrisey, T. Daly, I. Ejigiri, P. Sinnis, A. B.
Vaidya, L. W. Bergman, J. Chem. Inf. Model. 2010, 50, 840–849.
[13] J. M. Leung, F. Tran, R. B. Pathak, S. Poupart, A. T. Heaslip, B. A. Ballif,
N. J. Westwood, G. E. Ward, PLoS ONE 2014, 9, DOI: 10.1371/journal.-
pone.0098056.
Acknowledgements
We thank the beamline scientists at Diamond Light Source and
Dr. M. Henry and Dr. J. Moore from the Centre for Structural Biol-
ogy, Imperial College, for assistance with X-ray data collection.
We also thank Prof. R. Leatherbarrow for his help with DSF data
fitting with GraFit, and Dr. P. Simpson for advice on the acquisi-
tion and analysis of NMR data. This work was supported by the
Institute of Chemical Biology EPSRC Centre for Doctoral Training
(studentship to C.H.D.) and by the Erasmus Student Exchange
Programme (bursary to N.V.).
[14] C. H. Douse, J. L. Green, P. S. Salgado, P. J. Simpson, J. C. Thomas, G.
Langsley, A. A. Holder, E. W. Tate, E. Cota, J. Biol. Chem. 2012, 287,
36968–36977.
[15] S. Khamrui, S. Turley, E. Pardon, J. Steyaert, E. K. Fan, C. L. M. J. Verlinde,
L. W. Bergman, W. G. J. Hol, Mol. Biochem. Parasitol. 2013, 190, 87–91.
[16] M. Treeck, J. L. Sanders, J. E. Elias, J. C. Boothroyd, Cell Host Microbe
2011, 10, 410–419.
[17] F. H. Niesen, H. Berglund, M. Vedadi, Nat. Protoc. 2007, 2, 2212–2221.
[18] a) P. Cimmperman, L. Baranauskiene, S. Jachimoviciute, J. Jachno, J. Tor-
resan, V. Michailoviene, J. Matuliene, J. Sereikaite, V. Bumelis, D. Matulis,
Biophys. J. 2008, 95, 3222–3231; b) D. Matulis, J. K. Kranz, F. R. Salemme,
M. J. Todd, Biochemistry 2005, 44, 5258–5266.
[19] T. T. Waldron, K. P. Murphy, Biochemistry 2003, 42, 5058–5064.
[20] M. J. Harner, A. O. Frank, S. W. Fesik, J. Biomol. NMR 2013, 56, 65–75.
[21] F. Delaglio, S. Grzesiek, G. W. Vuister, G. Zhu, J. Pfeifer, A. Bax, J. Biomol.
NMR 1995, 6, 277–293.
Keywords: differential scanning fluorimetry
screening malaria myosins NMR protein–protein
interactions
·
fragment
·
·
·
·
[1] a) A. G. Coyne, D. E. Scott, C. Abell, Curr. Opin. Chem. Biol. 2010, 14,
299–307; b) A. M. Petros, J. Dinges, D. J. Augeri, S. A. Baumeister, D. A.
Betebenner, M. G. Bures, S. W. Elmore, P. J. Hajduk, M. K. Joseph, S. K.
Landis, D. G. Nettesheim, S. H. Rosenberg, W. Shen, S. Thomas, X. Wang,
I. Zanze, H. Zhang, S. W. Fesik, J. Med. Chem. 2006, 49, 656–663; c) D. E.
Scott, M. T. Ehebauer, T. Pukala, M. Marsh, T. L. Blundell, A. R. Venkitara-
man, C. Abell, M. Hyvonen, ChemBioChem 2013, 14, 332–342; d) I. Van
Molle, A. Thomann, D. L. Buckley, E. C. So, S. Lang, C. M. Crews, A. Ciulli,
Chem. Biol. 2012, 19, 1300–1312.
[2] a) E. H. Mashalidis, P. Sledz, S. Lang, C. Abell, Nat. Protoc. 2013, 8, 2309–
2324; b) H. L. Silvestre, T. L. Blundell, C. Abell, A. Ciulli, Proc. Natl. Acad.
Sci. USA 2013, 110, 12984–12989.
[22] B. A. Johnson, Methods Mol. Biol. 2004, 278, 313–352.
[23] P. Evans, Acta Crystallogr. Sect. D 2006, 62, 72–82.
[3] C. W. Murray, M. L. Verdonk, D. C. Rees, Trends Pharmacol. Sci. 2012, 33,
224–232.
[24] A. J. McCoy, R. W. Grosse-Kunstleve, P. D. Adams, M. D. Winn, L. C. Storo-
ni, R. J. Read, J. Appl. Crystallogr. 2007, 40, 658–674.
[4] L. W. Bergman, K. Kaiser, H. Fujioka, I. Coppens, T. M. Daly, S. Fox, K. Ma-
tuschewski, V. Nussenzweig, S. H. Kappe, J. Cell Sci. 2003, 116, 39–49.
[5] R. E. Farrow, J. Green, Z. Katsimitsoulia, W. R. Taylor, A. A. Holder, J. E.
Molloy, Semin. Cell Dev. Biol. 2011, 22, 953–960.
[6] A. F. Cowman, D. Berry, J. Baum, J. Cell Biol. 2012, 198, 961–971.
[7] a) J. Baum, D. Richard, J. Healer, M. Rug, Z. Krnajski, T. W. Gilberger, J. L.
Green, A. A. Holder, A. F. Cowman, J. Biol. Chem. 2006, 281, 5197–5208;
b) K. Frꢃnal, V. Polonais, J. B. Marq, R. Stratmann, J. Limenitakis, D. Solda-
ti-Favre, Cell Host Microbe 2010, 8, 343–357; c) M. L. Jones, M. O. Collins,
D. Goulding, J. S. Choudhary, J. C. Rayner, Cell Host Microbe 2012, 12,
246–258.
[25] P. Emsley, K. Cowtan, Acta Crystallogr. Sect. D 2004, 60, 2126–2132.
[26] A. W. Schꢄttelkopf, D. M. van Aalten, Acta Crystallogr. Sect. D 2004, 60,
1355–1363.
[27] G. N. Murshudov, P. Skubak, A. A. Lebedev, N. S. Pannu, R. A. Steiner,
R. A. Nicholls, M. D. Winn, F. Long, A. A. Vagin, Acta Crystallogr. Sect. D
2011, 67, 355–367.
[28] P. D. Adams, P. V. Afonine, G. Bunkoczi, V. B. Chen, I. W. Davis, N. Echols,
J. J. Headd, L. W. Hung, G. J. Kapral, R. W. Grosse-Kunstleve, A. J. McCoy,
N. W. Moriarty, R. Oeffner, R. J. Read, D. C. Richardson, J. S. Richardson,
T. C. Terwilliger, P. H. Zwart, Acta Crystallogr. Sect. D 2010, 66, 213–221.
[8] J. Bosch, C. A. Buscaglia, B. Krumm, B. P. Ingason, R. Lucas, C. Roach, T.
Cardozo, V. Nussenzweig, W. G. J. Hol, Proc. Natl. Acad. Sci. USA 2007,
104, 7015–7020.
Received: August 18, 2014
Published online on November 3, 2014
ChemMedChem 2015, 10, 134 – 143
www.chemmedchem.org
143 ꢀ 2015 The Authors. Published by Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim