Hemozoin inhibiting 2-phenylbenzimidazoles active against malaria parasites

Article abstract of DOI:10.1016/j.ejmech.2018.09.060

The 2-phenylbenzimidazole scaffold has recently been discovered to inhibit β-hematin (synthetic hemozoin) formation by high throughput screening. Here, a library of 325,728 N-4-(1H-benzo[d]imidazol-2-yl)aryl)benzamides was enumerated, and Bayesian statistics used to predict β-hematin and Plasmodium falciparum growth inhibition. Filtering predicted inactives and compounds with negligible aqueous solubility reduced the library to 35,124. Further narrowing to compounds with terminal aryl ring substituents only, reduced the library to 18, 83% of which were found to inhibit β-hematin formation <100 μM and 50% parasite growth <2 μM. Four compounds showed nanomolar parasite growth inhibition activities, no cross-resistance in a chloroquine resistant strain and low cytotoxicity. QSAR analysis showed a strong association of parasite growth inhibition with inhibition of β-hematin formation and the most active compound inhibited hemozoin formation in P. falciparum, with consequent increasing exchangeable heme. Pioneering use of molecular docking for this system demonstrated predictive ability and could rationalize observed structure activity trends.

Full text of DOI:10.1016/j.ejmech.2018.09.060

Accepted Manuscript
Hemozoin inhibiting 2-phenylbenzimidazoles active against malaria parasites
Fabrizio P. L'abbate, Ronel Müller, Roxanne Openshaw, Jill M. Combrinck, Katherine
A. de Villiers, Roger Hunter, Timothy J. Egan
PII:
S0223-5234(18)30838-9
10.1016/j.ejmech.2018.09.060
EJMECH 10771
DOI:
Reference:
To appear in: European Journal of Medicinal Chemistry
Received Date: 17 August 2018
Revised Date: 22 September 2018
Accepted Date: 24 September 2018
Please cite this article as: F.P. L'abbate, R. Müller, R. Openshaw, J.M. Combrinck, K.A. de Villiers,
R. Hunter, T.J. Egan, Hemozoin inhibiting 2-phenylbenzimidazoles active against malaria parasites,
European Journal of Medicinal Chemistry (2018), doi: https://doi.org/10.1016/j.ejmech.2018.09.060.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

ACCEPTED MANUSCRIPT
Hemozoin inhibiting 2-phenylbenzimidazoles active against malaria parasites
Fabrizio P. L’abbate,a Ronel Müller,b Roxanne Openshaw,a Jill M. Combrinck,c Katherine A. de
Villiers,b Roger Hunter, a and Timothy J. Egan a, d,
*
aDepartment of Chemistry, University of Cape Town, Rondebosch 7701, South Africa
bDepartment of Chemistry and Polymer Science, Stellenbosch University, Private Bag X1,
Matieland 7602, South Africa
cDepartment of Medicine, Division of Clinical Pharmacology, University of Cape Town,
Observatory 7925, South Africa
dInstitute of Infectious Disease and Molecular Medicine, University of Cape Town, Rondebosch
7701, South Africa
* Corresponding author.
Email address: timothy.egan@uct.ac.za.

ACCEPTED MANUSCRIPT
2
ABSTRACT
The 2-phenylbenzimidazole scaffold has recently been discovered to inhibit β-hematin (synthetic
hemozoin) formation by high throughput screening. Here, a library of 325,728 N-4-(1H-
benzo[d]imidazol-2-yl)aryl)benzamides was enumerated, and Bayesian statistics used to predict
β-hematin and Plasmodium falciparum growth inhibition. Filtering predicted inactives and
compounds with negligible aqueous solubility reduced the library to 35,124. Further narrowing
to compounds with terminal aryl ring substituents only, reduced the library to 18, 83% of which
were found to inhibit β-hematin formation <100 µM and 50% parasite growth <2 µM. Four
compounds showed nanomolar parasite growth inhibition activities, no cross-resistance in a
chloroquine resistant strain and low cytotoxicity. QSAR analysis showed a strong association of
parasite growth inhibition with inhibition of β-hematin formation and the most active compound
inhibited hemozoin formation in P. falciparum, with consequent increasing exchangeable heme.
Pioneering use of molecular docking for this system demonstrated predictive ability and could
rationalize observed structure activity trends.
Keywords: malaria; hemozoin; benzimidazoles; docking; Bayesian statistics; QSAR

ACCEPTED MANUSCRIPT
3
1. Introduction
The present WHO recommended use of artemisinin-based combination therapy has
significantly decreased morbidity and mortality rates in malaria stricken countries [1, 2].
Nevertheless, recent reports have demonstrated the evolution of resistance to artemisinin
characterized by phenotypes with delayed parasite clearance [3-5]. This alarming fact, coupled
with resistance to most of the existing clinical antimalarials indicates the importance of
continuing the search for new antimalarials [6]. It is well known that chloroquine targets the
hemozoin formation pathway [7-10], in which chloroquine-resistant strains of Plasmodium
falciparum efflux chloroquine and to varying extents other quinoline-derived drugs via a
transporter, PfCRT, present in the parasite digestive vacuole membrane, which reduces
hemozoin inhibition by these drugs [11]. Evidence suggests that chloroquine-resistant mutants of
PfCRT selectively transport specific molecules, such that hemozoin inhibition remains a viable
target for novel compounds, especially if based on non-quinoline scaffolds [12].
Activity data for inhibition of β-hematin (the synthetic counterpart of hemozoin) formation
for about two hundred thousand compounds (priors) have become available from high
throughput screening (HTS) [13]. Consequently, we recently developed Bayesian statistical
models using HTS data from the GlaxoSmithKline TCAMS [14], St Jude Children’s Research
Hospital [15] and Vanderbilt University libraries [13, 16], as well as compounds synthesized at
the University of Cape Town [17, 18] and Okayama University [19-23] to correctly predict both
β-hematin (the synthetic counterpart of hemozoin) and parasite inhibition activity that is linked
to β-hematin inhibition [24]. The β-hematin inhibition model was found to correctly predict
activity 77% of the time with a ROC score of 0.915, while the parasite growth inhibition activity
model was correct 92% of the time with a ROC score of 0.991. The benzimidazole moiety was

ACCEPTED MANUSCRIPT
4
found to be the most prominent molecular fingerprint in both Bayesian models. It was found that
103 of 155 priors with this fingerprint were active in the β-hematin inhibition model while 194
of 194 priors were active in the parasite growth inhibition activity model [24].
Benzimidazoles are a well-known chemical class that have been shown to have a broad
range of medicinal uses [25]. Since the late 1980s, there have been numerous reports of
benzimidazole compounds inhibiting the growth of Plasmodium parasites [26-28], although most
of these studies have not investigated β-hematin inhibition. Indeed, to date only Chibale and co-
workers have shown weak β-hematin inhibitory activities for benzimidazole compounds [29].
Herein, we describe the use of Bayesian statistical models to guide the synthesis of
benzimidazole analogues, along with SAR and molecular docking of the synthesized derivatives
to rationalize β-hematin inhibition activity as well as SAR and QSAR data for inhibition of
parasite growth.
2. Chemistry
2.1
Compound Selection
One of the many challenges in drug discovery is the selection of a suitable scaffold and
derivatives thereof for further investigation. Our earlier work (Wicht et al.) [24] prompted further
study of benzimidazoles as viable hemozoin inhibiting antimalarials, with the scaffold shown in
Figure 1 exhibiting fingerprints predicted to be strongly favorable for both β-hematin and
parasite growth inhibition (Figure 1A). Using the disconnection approach shown in Figure 1B,
Pipeline Pilot and Discovery Studio were employed to devise a streamlined strategy for selecting
compounds (Figure 1C) [30]. This involved enumeration of compounds based on available
starting materials to create an in silico library, followed by predictions of activities and

ACCEPTED MANUSCRIPT
5
molecular properties to select the most appropriate derivatives to synthesize. Using this
approach, 325,728 compounds were enumerated through a two-step reaction from an input of 26
o-phenylenediamine derivatives, 24 p-aminobenzoic acid derivatives and 522 aromatic
carboxylic acids using available starting materials taken from the Sigma-Aldrich catalogue.
Figure 1. Strategy for selection of compounds for investigation. Leading structural fingerprints
previously identified using Bayesian statistics for both β-hematin inhibition and in vitro P.
falciparum growth inhibition encompassed the benzimidazole scaffold (A) [24]. Disconnection
strategy used for synthesis and enumeration of all possible compounds based on available
starting materials (B). Strategy for filtering enumerated structures based on predicted β-hematin
inhibition, parasite growth inhibition, solubility and generalizations taken from the predicted

ACCEPTED MANUSCRIPT
6
active list used to simplify the final list of target compounds (C). Suitable aryl groups are
discussed in the main text.
After enumerating this in silico library, the next step in the workflow was to predict their
Bayesian score for β-hematin inhibition using a model created by Wicht et al consisting of
64,118 diverse priors and a cut-off activity of 100 µM [24]. All predicted inactive compounds
(19,904), determined by the model cut-off Bayesian score of -4.920, were rejected and the
predicted active compounds (305,824) were then further filtered using a parasite growth
inhibition activity model with a Bayesian cut-off score of -8.580, also created by Wicht et al
based on 41,729 diverse priors and a cut-off activity of 2 µM [24]. While the parasite growth
inhibition model predicts activity against P. falciparum without the need to first predict β-
hematin inhibition itself, conducting both filtering processes allowed two independent
predictions which could be separately experimentally tested. The predicted inactive compounds
(2,148) were again rejected and the predicted active compounds (303,676) were sorted by their
aqueous solubility level, available as part of the Discovery Studio ADMET prediction function.
Those with a cut-off level of 1 or less were excluded. This led to 35,124 compounds that were
predicted to have a moderate to low solubility, while 268,552 were predicted to have little to no
solubility in aqueous medium at 25 °C. Only 68 compounds of the 35,124 compounds were
predicted to have a solubility level of 3 (moderate). When examining the full predicted low to
moderately soluble active set, an absence of substituents on ring B (see Figure 1) was found to be
a defining feature. The only deviation from this generalization was the presence of single nitro
substituents on this ring, or a change from a benzene to pyridine ring. The other dominant
features found in the soluble active compound subset were: the presence of either unsubstituted
nitro- or halogen mono- or disubstituted benzimidazoles (ring A in Figure 1); and the presence of

ACCEPTED MANUSCRIPT
7
5-membered heterocyclic rings, mono or disubstituted benzene or pyridine ring systems to the
right of the amide bond (R in Scheme 1). Based on these predictions, substituents on the central
ring were avoided. Substituents on the benzimidazole ring were also avoided, since single
substituents result in non-equivalent inseparable tautomers in the case of mono-substituted
systems which unnecessarily complicates the system. Consequently, in this study we
strategically decided to concentrate only on varying the arylamide R group on the right of the
molecule (Figure 1). Given a starting total of 522 possible compounds where only R is varied,
this reduced the number of compounds with adequate predicted solubility to warrant synthesis to
the eighteen (2 – 19) shown in Scheme 1.
2.2 Synthesis
Target compounds chosen for investigation are shown in Scheme 1. Of the eighteen
compounds chosen, eight (2, 7, 11, 13 – 15, 17 and 18) have previously been described. For the
remaining ten compounds, no synthesis or characterization data are available in the literature,
despite eight of them being apparently commercially available (mostly via synthesis on demand)
and three having been previously used in biological studies using purchased compound libraries
[16, 31, 32]. The compounds were synthesized in a simple two-step process. This involved
formation of 4-(1H-benzo[d]imidazol-2-yl)aniline (1) by reaction of o-phenylenediamine with p-
aminobenzoic acid catalyzed by polyphosphoric acid (PPA) as described by Chua et al [33].
Reaction of 1 with the appropriate acid chloride using pyridine as an acyl transfer reagent
afforded the products in moderate to excellent yields ranging from 45 – 92%. All products were
fully characterized by 1H and 13C NMR, high resolution mass spectrometry and HPLC and were
at least 95% pure.

ACCEPTED MANUSCRIPT
8
Scheme 1. (i) PPA, 220 °C, 5 h; (ii) pyridine, THF or DMF, −40 °C, 2 – 3 h.
3
Results and Discussion
3.1 Physicochemical Properties and Biological Testing
Compounds 2 – 19 were tested for their ability to inhibit β-hematin formation in an NP-40
detergent mediated assay with the use of pyridine to detect unreacted hematin [34, 35], as well as
parasite growth inhibition in the chloroquine-sensitive NF54 strain of P. falciparum cultured in
vitro (Table 1).

ACCEPTED MANUSCRIPT
9
Table 1. Inhibition of
β-hematin formation and parasite growth by compounds 2 – 19.
β
-hematin inhibition IC50
NF54 malaria parasite
growth inhibition IC50 (µM)a
255 ± 47
Compound
(µM)
2
3
4
5
6
3,900b
24.5 ± 4.2
28.1 ± 1.7
32.0 ± 1.4
358 ± 5
1.3 ± 0.2
1.34 ± 0.07
0.41 ± 0.03
2.0 ± 0.1
7.2 ± 0.2
7
90 ± 19
38.9 ± 1.5
16.6 ± 0.3
28.9 ± 1.6
13.3 ± 0.5
16.8 ± 0.4
38.2 ± 1.4
25.1 ± 1.1
1,790b
35.6 ± 1.7
42.4 ± 2.1
84.1 ± 2.4
46.6 ± 1.7
31.5 ± 0.5c
5.50 ± 0.6
0.8 ± 0.5
12.3 ± 2.7
0.67 ± 0.07
0.53 ± 0.13
15.0 ± 0.3
1.21 ± 0.06
64 ± 8
11.6 ± 2.6
7.2 ± 0.7
5.9 ± 0.8
1.2 ± 0.4
8
9
10
11
12
13
14
15
16
17
18
19
Chloroquine
0.016 ± 0.006
a
b
c
Determined by the LDH method of Makler et al [36]; Estimated based on extrapolation;
From Wicht et al [17].
A notable feature of the findings in Table 1 is that 83% of compounds (15/18) were found to
inhibit β-hematin formation below the cut-off of 100 µM, a value that encompasses the clinical
hemozoin inhibiting antimalarials (22.0 ± 0.8 µM for chloroquine to 52.0 ± 0.9 µM for quinine)
[37, 38]. This pleasing hit-rate for the Bayesian method probably reflects the fact that the
benzimidazole scaffold in this study is associated with the most positive fingerprint in the
training set, combined with the high ROC score noted in the introduction above (0.915). SAR
analysis revealed clear trends in the β-hematin inhibition data, which are summarized in Figure
2. Comparison of compounds bearing unsubstituted terminal aryl rings 2, 16 and 17 – 19 showed
that the phenyl group renders the molecule virtually inactive (2), five-membered heteroaromatic
rings result in intermediate activity (17 – 19) and pyridine is most active (16). The trend of the

ACCEPTED MANUSCRIPT
10
pyridine being more active than the phenyl derivative is maintained in compounds 9 versus 4, but
not 5 versus 3 bearing CF3 and Br substituents respectively at the meta position on the aromatic
ring. In the case of compounds with a terminal phenyl ring, substituents at the para position
greatly diminished (6 and 7) or essentially abolished (15) β-hematin inhibitory activity.
Compounds with substituents at the ortho and meta positions were active, but the meta position
was favored (14 versus 13). In the case of compounds with an ortho substituent, more strongly π
electron withdrawing groups favored activity (8 < 13 < 10). Overall, seven of the eight most
active compounds (3, 4, 5, 9, 11, 12 and 14) bear electron-withdrawing substituents at the meta
position, whether on a terminal phenyl or pyridyl ring.
Figure 2. Summary of SAR trends in the R group found for β-hematin inhibition activity (larger
= more active). ERG = electron-releasing group, EWG = electron-withdrawing group.
Malaria parasite growth inhibition measurements revealed 50% of the compounds (9/18) to
be active against the NF54 strain of the malaria parasite at the 2 µM cut-off, while 89% (16/18)
showed activity against parasites below 20 µM (Table 1). Once again, this confirms the
effectiveness of the Bayesian approach, in this case the model for parasite growth inhibition.
SAR analysis of the NF54 parasite growth inhibition IC50 data revealed somewhat different
trends to that of β-hematin inhibition (Figure 3). Here terminal 5-membered heteroaromatic ring

ACCEPTED MANUSCRIPT
11
systems (17 – 19) were preferred to an unsubstituted pyridyl ring (16), while the compound with
a terminal phenyl ring (2) was again virtually inactive. In the case of a terminal phenyl ring,
substituents at the para and ortho positions exhibited similar weak to very weak activity (6 - 8,
10, 13, 15), while those with meta substituents were more active (3, 4, 11, 14). In the case of
terminal phenyl rings bearing substituents at the ortho position, electron releasing groups (8)
were preferred over electron withdrawing groups (10, 13). The most active compounds of the
series were those with terminal pyridyl ring systems bearing electron-withdrawing groups, which
all exhibited sub-micromolar activity (5, 9 and 12). These observations indicate that features
required for strong β-hematin inhibition need to be balanced with properties required for
penetration of the parasite cell.
Figure 3. Summary of SAR trends in the R group found for parasite growth inhibition activity
(larger = more active). ERG = electron-releasing group, EWG = electron-withdrawing group.

ACCEPTED MANUSCRIPT
12
3.2 QSAR
A plot of log IC50 (NF54) versus log IC50 (β-hematin) revealed a statistically significant
correlation (Figure 4A), showing that compounds which inhibited β-hematin formation more
strongly were also more active against parasite growth. Although the correlation was significant
(P < 0.0002), the R2 value was only 0.59. Consequently, further analysis was performed.
Quantitative structure activity relationship (QSAR) analysis was performed using the
program MMP Plus [39] to further explore physicochemical factors underlying the biological
activity of these compounds. This revealed three factors that correlated with inhibition of in vitro
P. falciparum growth namely β-hematin inhibitory activity, hydrogen bond donor from charge
calculation (a charge-related property computed in MMP Plus) and molecular depth (Figure 4,
Table S1). The strongest association was found with the IC50 for β-hematin inhibition, indicating
that hemozoin inhibition is central to the activity of this set of compounds.
Comparison of IC50 (β-hematin) and IC50 (NF54) values in Table 1 shows that the former
are always substantially larger than the latter. This is a typical feature of hemozoin inhibitors that
can be ascribed to pH trapping of these compounds in the acidic digestive vacuole (DV) of the
parasite and has previously been shown to correlate with the activities of chloroquine analogues
[40]. The benzimidazoles are all weak bases that are protonated around pH 5 and are thus active
in their protonated form. It can be expected that they will accumulate in acidic organelles such as
the DV to concentrations at which they can inhibit hemozoin formation. This hypothesis can also
account for the much greater activity of chloroquine which has two sites of protonation (the
quinoline N of the 4-aminoquinoline group and tertiary amine in the side chain), both more basic
than the single protonatable imidazole N in the current series.

ACCEPTED MANUSCRIPT
13
Figure 4. Correlation of (A) parasite IC50 and β-hematin inhibition activity and (B) observed
pIC50 plotted against predicted pIC50 for inhibition of P. falciparum growth based on a fitted
multiple correlation equation. In (A) R2 = 0.59 and P = 0.0002. In (B) pIC50 = 28.7(1/IC50 β-
0.01
hematin) + 3.71(HBD) – 1.43(molecular depth) – 8.90. F = 9.81 > Fcrit = 5.56, r2 = 0.68, P <
0.0001. t for the individual parameters > tcrit0.10 = 1.761 (5.11, 1.88 and 2.56 respectively for the
first, second and third coefficients).
To investigate whether the activity of these compounds arises from substructures of these
molecules, or requires the intact molecule, we deconstructed the most active compound by
investigating commercially available derivatives that represent various fragments of compound 5
(Figure 5). None of these fragments showed any significant β-hematin or parasite growth
inhibition activity, confirming that activity does not reside in any single sub-structure.
Interestingly, the Bayesian model correctly predicted that compounds 21 – 25 would be inactive.

ACCEPTED MANUSCRIPT
14
Figure 5. Deconstruction of compound 5. All the compounds representing fragments of 5 (20 –
25) exhibited no significant β-hematin or parasite growth inhibition activity.
3.3 Cross-resistance and cytotoxicity
The compounds most active against the chloroquine-sensitive NF54 strain of P. falciparum
(5, 9, 11 and 12) were then tested for cross-resistance against the chloroquine-resistant Dd2
strain and for cytotoxicity against mammalian cells (Chinese hamster ovarian, CHO, cells). None
of the compounds showed any substantial cross-resistance with chloroquine and all had good to
excellent selectivity against malaria parasites (Table 2).

ACCEPTED MANUSCRIPT
15
Table 2. Parasite growth inhibition IC50 in chloroquine sensitive (NF54) and resistant
(Dd2) strains and cytotoxicity in CHO cells.
SI (NF54)b SI (Dd2)b
Compound IC50 Dd2 (µM) IC50 NF54 (µM) RIa IC50 CHO (µM)
0.96 ± 0.04
1.10 ± 0.37
0.31 ± 0.02
0.34 ± 0.03
0.15 ± 0.03
−
0.41 ± 0.03
0.80 ± 0.50
0.67 ± 0.07
0.53 ± 0.13
0.016 ± 0.006
−
2.4
1.4
189 ± 4
35 ± 4
206
461
44
308
347
−
197
32
665
541
−
5
9
0.46
0.64
9.4
11
12
184 ± 9
−
Chloroquine
Emetine
0.12 ± 0.06
−
−
a
b
Resistance index = IC50 (Dd2)/IC50 (NF54); Selectivity index = IC50 (CHO)/IC50 (NF54 or
Dd2)
3.4 Physico-chemical properties and potential for further development
A major focus of this investigation was to determine whether the Bayesian approach that we
had previously developed could be used to narrow down an enumerated library to active
compounds. If any of these are to be further developed, their drug-like properties and potential
liabilities need to be considered. Table 3 presents key molecular properties. In terms of
Lipinski’s rule of five, none of the compounds have molecular weights above 500, or violate the
hydrogen bond donor or acceptor limits of 5 and 10 respectively. Five compounds do have
calculated LogP values greater that 5 (3, 4, 6, 12 and 14), but only 12 is a frontrunner compound
in terms of activity. Other factors are less satisfactory from a potential drug development point of
view. None of these compounds exhibit adequate aqueous solubility and several contain groups
known to exhibit potential toxic liabilities, including aromatic nitro groups (10 and 11),
thiophenes (17), furans (18) and 1,2-halopyridines (12) [41]. These and other liabilities such as
potential interaction of pyridines with cytochrome P450 and hydrolysis of the amide to produce a

ACCEPTED MANUSCRIPT
16
potentially toxic aniline would need to be taken into consideration for further development. Most
of these liabilities can be addressed by further synthetic modification, including substitution with
bioisosteres and other structural modifications [41].
Table 3. Computed physico-chemical properties of compounds 2 – 19.
Compound
MW
cLogP
4.08
5.05
5.14
4.42
5.77
4.72
3.81
4.65
3.68
3.84
5.17
4.25
5.17
4.53
3.19
3.68
3.84
2.24
HBD #
HBA #
313.35
392.25
381.35
393.24
369.46
382.24
343.38
382.34
358.35
403.09
383.23
347.80
347.80
347.80
314.34
319.38
303.10
302.33
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
3
2
2
2
3
2
2
3
3
2
2
3
2
2
2
3
2
3
2
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
3.5 Cellular investigations
To establish whether the observation that active members of this class of compound inhibit
β-hematin formation in the NP40 detergent based assay translates into hemozoin inhibition in the
parasite, we conducted further investigations of compound 5. This was carried out using a

ACCEPTED MANUSCRIPT
17
cellular heme fractionation assay that we have reported previously [42]. This assay measures the
mass of free heme, hemozoin and undigested hemoglobin in cultured P. falciparum cells.
Compound 5 was found to cause a large decrease in hemozoin, with a corresponding increase in
freely exchangeable heme as well as a much smaller (non-significant) increase in undigested
hemoglobin in the cell, confirming that it inhibits cellular hemozoin formation (Figure 6). This
finding is consistent with the hypothesis that a buildup of free heme is responsible for parasite
death and that hemozoin inhibition is indeed central to the mode of action of these compounds,
although the detailed mechanism of parasite killing remains elusive and continues to be a subject
of investigation. In view of this, molecular docking was used to investigate the possible
molecular basis of β-hematin inhibition activity in these compounds.

ACCEPTED MANUSCRIPT
18
Figure 6. Quantities of undigested hemoglobin (A), haemozoin (B) and freely exchangeable
heme (C) in the NF54 strain of P. falciparum cultured in vivo after 32 h incubation with
compound 5. Parasites were synchronized, and drug added to early ring stage parasites. Amounts
are expressed in term of fg heme iron per cell, * P < 0.05, ** P < 0.01, *** P < 0.001, **** P <
0.0001 versus control.
4. Molecular docking.
In silico methods, especially docking algorithms, are routinely used in drug discovery to
investigate ligand-protein interactions [43]. Biomineralization processes may also be inhibited by
suitable adsorbates, however to our knowledge, few in silico studies have been carried out. We
have now developed a molecular docking method for investigating β-hematin inhibition. Since

ACCEPTED MANUSCRIPT
19
this involves interaction of inhibitors with the solid surfaces of the β-hematin crystal, the method
makes use of BIOVIA Materials Studio [44], a program designed to simulate solid state
materials to model the adsorption of compounds onto surface sites. Using the Adsorption Locator
tool, we were able to simulate the adsorption of inhibitors onto the fastest- and second-fastest
growing faces of β-hematin, (001) and (011) respectively, to identify preferred site(s) of
adsorption and gain insight into structure-activity relationships. Hydrogen bonding and π-π
interactions between the compounds synthesized in this study and the two fastest growing faces
of the β-hematin crystal were identified. Using this approach, it was found that the number of
hydrogen bond contacts in the distance range 2.36 – 3.69 Å between the crystal surface and all
18 compounds remained constant at one. On the other hand, the number of π-π interactions in the
range 3.4 – 3.8 Å varied and the total number of contacts within this range was predictive of
strong inhibition (Figure 7A). All compounds in which the total number of π-π interactions in
this distance range was 2 or fewer in aggregate for both the (001) and (011) faces were virtually
inactive (IC50 > 100 µM) or almost so (IC50 > 80 µM). For the remaining compounds, a direct
proportionality was found between adsorption energy with the (011) face and the experimental β-
hematin inhibition IC50 (Figure 7B).

ACCEPTED MANUSCRIPT
20
Figure 7. A representation of the number of π-π interactions with the (011) and (001) faces of
the β-hematin crystal found by molecular docking and inhibitory activity for compounds 2 - 19
(A). In this representation + refers to IC50 values below 80 µM and sum refers to the number of
π-stacking interactions on both the (001) and (011) faces with a distance less than 3.8 Å. In the
case of the (011) face, a direct proportionality was found between adsorption energy (Eads011) and
IC50 for inhibition of β-hematin formation; r2 = 0.70, P = 0.0004 (B).
Given the approximations made in the docking approach, the predictive ability shown in
Figure 7A and correlation found in Figure 7B is remarkable. In the real system both the β-
hematin surface and the compounds would be hydrated. Furthermore, the surface groups on the
crystal are flexible and experimental evidence from atomic force microscopy reported by
Vekilov and co-workers suggests that hemozoin-inhibiting drugs interact with steps and kinks on
the crystal surface [45, 46]. It is likely that the faces investigated by molecular docking are
sufficiently like the sides of these steps and kinks to account for the predictive ability of this
approach. In addition, the similarity of these molecules probably means that desolvation energies
are almost constant for this series.

ACCEPTED MANUSCRIPT
21
Closer examination of the docking of compound 5 with the (011) and (001) faces of β-
hematin also provides more insight into the structure activity relationships discussed earlier. On
both the (011) and (001) faces the amide group acts as a hydrogen bond donor to the carbonyl
oxygen of the coordinated propionate group of β-hematin (Figure 8A and B). The terminal (C)
ring forms a π-stacking interaction with pyrrole rings of the heme molecule on both the (011)
and (001) faces, while the B ring π-stacks with a pyrrole ring on the (011) face (Figure 8A) and
the A ring with the (001) face (Figure 8B). This probably explains the need for the full structure
for β-hematin inhibition, accounting for the inactivity of all the fragment derivatives in Figure 5.
A proposed structure activity relationship is shown in Figure 8C for compound 5. A
complicating factor is that both surfaces can accommodate this series of compounds in different
orientations and so, for example, compound 3 which is also active makes a similar interaction
with the (001) face, but rings A and C form π-stacking interactions with the (011) face, rather
than rings A and B. In the case of the much less active compounds 2, 6, 7, 15 and 18 several
factors appear to hinder the interaction. In the case of 2, interaction with the (011) face is little
changed, but it enters the groove on the (001) face with the C ring, rather than the A ring and
makes much longer π-stacking contacts in the groove (>4 Å) and is unable to make any such
contacts outside the groove. Similar interactions with the (001) face were seen with 7 and 15,
while the bulky tert-butyl group prevented 6 from entering the groove at all. Compound 18
entered the groove side-on and did not interact with the (001) face outside the groove at all.
Thus, the ability of these compounds to inhibit β-hematin formation appears to stem from a
complex interplay of molecular shape and electronic factors in which the relatively less electron
dense pyridine ring, or phenyl rings bearing electron withdrawing groups can strongly interact
with the porphyrin rings via π-stacking, accounting for the observed trends in activity discussed

ACCEPTED MANUSCRIPT
22
earlier. An interesting observation based on molecular docking is that the NH group of the
benzimidazole moiety does not hydrogen bond with the hemozoin surface, probably because of a
preference for the flat benzimidazole ring to lie approximately parallel with the heme rings. The
docking model suggests that the imidazole NH group ought to represent a suitable site for
derivatization. To test this, we introduced a phenyl amide group at the imidazole N atom by
reacting compound 3 with benzoyl chloride to produce 26 (Figure 8D). This compound showed
very similar activity to 3, with slightly stronger β-hematin inhibitory activity (IC50 15.6 ± 0.5
versus 24.5 ± 4.2 µM) and slightly weaker parasite growth inhibition activity (IC50 2.6 ± 0.6
versus 1.3 ± 0.2 µM), confirming that modification at this site is indeed tolerated, consistent with
the docking model.
Figure 8. Molecular docking of compound 5 to the (011) face (A) and (001) face (B) of
hemozoin. Interactions 1 and 2 represent π-stacking and 3 hydrogen bonding. A structure activity
relationship hypothesis for this pharmacophore is presented in C. Modification of the imidazole
NH group in compound 26 had relatively little influence on activity (D).

ACCEPTED MANUSCRIPT
23
5. Conclusions
This study has demonstrated the utility of in silico methods in the search for new hemozoin
inhibiting antimalarials. Bayesian models proved to be very effective for pre-selection of β-
hematin inhibiting compounds from a large enumerated in silico library based on synthesis from
commercially readily available starting materials. This allowed the search for hit compounds to
be dramatically narrowed down by predicting activities before undertaking time-consuming
synthesis. A library of 325,728 compounds was narrowed down to just 18 for experimental
investigation. Of these, 83% were found to indeed inhibit β-hematin formation below the model
cut-off of 100 µM and 50% were active against the NF54 strain of P. falciparum cultured in vitro
below a cut-off of 2 µM. QSAR analysis revealed that parasite growth inhibition activity
increased with an increase in β-hematin inhibition activity, number of hydrogen bond donors and
molecular depth, strongly indicating hemozoin inhibition is key to their activity. In the case of
compound 5, this was confirmed in the parasite cell itself, where hemozoin formation was
decreased and freely exchangeable heme increased as a function of dose. Molecular docking
could account for β-hematin inhibition activity based on the number of π-stacking interactions
with the two fastest growing faces of the crystal as well as the calculated adsorption energy with
the second fastest growing face (011). This further demonstrates the utility of these in silico
approaches in rationalizing and predicting hemozoin inhibition. To our knowledge, this is one of
the first studies in which such methods have been applied to hemozoin inhibitors to both predict
activity and rationalize SARs. Finally, four compounds were identified with nanomolar activity.
They exhibited no cross-resistance with chloroquine in the Dd2 strain of parasite and had low
cytotoxicity in the mammalian CHO cell line with excellent selectivity against malaria parasites,
warranting further investigation and derivatization.

ACCEPTED MANUSCRIPT
24
6. Experimental Section
6.1 Chemistry
All reagents were purchased from commercial sources (Sigma-Aldrich, Kimix, and Protea
chemicals). Solvents other than anhydrous pyridine were distilled before use. Tetrahydrofuran
was distilled under N2(g) using Na wire and benzophenone. CaCl2 was used as a drying reagent
in drying tubes. Thin layer chromatography was performed using aluminum-backed silica gel 60
F254 plates, which were purchased from Merck and visualized using a UV lamp, anisaldehyde
spray (a mix in a 1:1 (v/v) ratio of 10% H2SO4 in ethanol solution and 5% anisaldehyde in
ethanol solution) or a ninhydrin spray (300 mg ninhydrin in 97 mL absolute ethanol and 3 mL of
glacial acetic acid). Column chromatography was carried out using silica gel 60 (mesh 68 – 280
µm) purchased from Fluka and a Biotage Isolera One flash chromatography system. Nuclear
magnetic resonance experiments were recorded using Bruker or Varian Unity 300, 400 or 600
MHz instruments. Chemical shifts (δ) were recorded relative to residual DMSO-d6 (δ 2.50 in 1H
1
NMR and δ 39.52 in 13C NMR), chloroform-d (δ 7.26 in H NMR and δ 77.16 in 13C NMR),
methanol-d4 (δ 4.87 in 1H NMR and δ 49.00 in 13C NMR) or acetone-d6 (δ 2.05 in 1H NMR and
13
δ 206.26 in C NMR). All chemical shifts are reported in ppm and all J values are reported in
Hz. Melting points were obtained using a Reichert-Jung Thermovar hot stage microscope. HPLC
experiments were carried out using an Agilent Technologies 1220 Infinity LC with a reverse
phase C18 column using a solvent system of deionized Millipore© Direct-Q H2O and HPLC
grade CH3CN. High resolution mass spectrometry was performed on a Waters Synapt G2
instrument. All mass spectra were recorded using the electrospray positive (ES+) technique and
the sample was introduced via an ESI probe injected into a stream of CH3CN.

ACCEPTED MANUSCRIPT
25
Melting points, HRMS data and HPLC purity data for compounds for which syntheses have
been previously reported are presented in Table 3. Further details of the syntheses are available
in Supplementary Data.
Table 3. Melting points, HRMS and purity data for compounds for which syntheses and
characterization data have previously been reported.
HRMSa
Purityb
MP (°C)
Compound
240-241
lit. 240-241
[47]
Obs: 210.1031
Calc: 210.1031
1 4-(1H-benzo[d]imidazol-2yl)aniline
99.1%
340-341
lit. 335 [48]
273-275
lit. 242 [49]
338-339
lit. 190 [50]
275-276
lit. 311 [49]
320-322
lit. 312 [49]
331-333
lit. 323 [49]
344-346
lit. 349-350
[51]
Obs: 314.1293
Calc: 314.1293
Obs: 382.0508
Calc: 382.0514
Obs: 404.1004
Calc: 404.0995
Obs: 348.0901
Calc: 348.0904
Obs: 348.0899
Calc: 348.0904
Obs: 348.0898
Calc: 348.0904
2 N-4-(1H-benzo[d]imidazol-2-yl)phenyl)benzamide
7 N-(4-(1H-benzo[d]imidazole-2-yl)phenyl)-2,4-dichlorobenzamide
11 N-(4-(1H-benzo[d]imidazole-2-yl)phenyl)-3,5-dinitrobenzamide
13 N-(4-(1H-benzo[d]imidazole-2-yl)phenyl)-2-chlorobenzamide
14 N-(4-(1H-benzo[d]imidazole-2-yl)phenyl)-3-chlorobenzamide
15 N-(4-(1H-benzo[d]imidazole-2-yl)phenyl)-4-chlorobenzamide
98.2%
99.6%
99.2%
100.0%
95.9%
100.0%
17 N-(4-(1H-benzo[d]imidazole-2-yl)phenyl)thiophene-2-
Obs: 320.0862
Calc: 320.0858
97.3%
95.1%
98.6%
carboxamide
252-253
lit. 301 [49]
302-303
lit. 308-309
[33]
Obs: 304.1072
Calc: 304.1086
18 N-(4-(1H-benzo[d]imidazole-2-yl)phenyl)furan-2-carboxamide
Obs: 252.1135
Calc: 252.1137
20 N-(4-(1H-benzo[d]imidazole-2-yl)phenyl)acetamide
a [M+H]+; b HPLC
6.1.1 Preparation of acid chlorides
Using a typical scale of 2.0 mmol, the appropriate aryl carboxylic acid (2.0 mmol) was added to
small round-bottomed flask, which was flushed with N2(g). SOCl2 (3 mL,
a

ACCEPTED MANUSCRIPT
26
41.4 mmol) was added after which the reaction mixture was refluxed, with stirring, at 80 °C and
left overnight to ensure complete conversion. The reaction mixture was cooled and excess SOCl2
was completely removed under reduced pressure. The crude acid chloride was then used as a
reagent for subsequent reactions.
6.1.2 Synthetic
procedure
for
preparation
of
N-(4-(1H-benzo[d]imidazol-2-
yl)phenyl)(hetero)arylcarboxamides
Using a typical scale of 1.0 mmol, 1 (1.0 mmol) was dissolved in anhydrous pyridine (2 mL) in a
small round-bottomed flask and the mixture was cooled to -40 °C after which the acid chloride
(2.0 mmol; purchased or prepared as above) in dry THF or DMF (2 mL) was added dropwise
over 20 min with vigorous stirring. After 2-6 h the reaction mixture was warmed to room
temperature and the solvent reduced under pressure. Without using a work-up the crude product
was purified by column chromatography directly using MeOH/DCM mixtures (1:99 to 2:8).
6.1.2.1 N-(4-(1H-benzo[d]imidazole-2-yl)phenyl)-3-bromobenzamide (3)
3-Bromobenzoic acid (804 mg, 4.0 mmol) and 1 (418 mg, 2.0 mmol) afforded a brown solid.
The resulting solid was recrystallized from methanol to afford light-brown crystals of (3) (668
mg, 85%).
ATR-FTIR νmax /cm-1 3248 (NH), 1652 (C=O amide); m.p. 324 - 325 °C; Rf (MeOH/DCM 5:95)
0.39; δH (DMSO-d6, 400 MHz) 7.20 (m, 2H), 7.63 – 7.79 (m, 3H), 7.81 (d, J = 7.8 Hz, 1H), 7.98
(m, 3H), 8.19 (m, 3H), 10.54 (s, 1 H), 12.80 (s, 1 H); δC (DMSO-d6, 100.6 MHz) 111.1, 118.6,
120.4, 121.5, 121.7, 122.3 (Cquat), 125.6 (Cquat), 126.9, 126.9, 130.3, 130.6, 134.4, 135.0 (Cquat),
136.9 (Cquat), 140.3 (Cquat), 143.9 (Cquat), 151.1 (Cquat), 164.1 (CO); HRMS-ES+ Calculated:

ACCEPTED MANUSCRIPT
27
392.0398 [M+H]+ for C20H15Br79N3O, Observed: 392.0399; C18 HPLC, flow rate: 1 mL/min,
H2O / acetonitrile (40:60), 99.8%.
6.1.2.2 N-(4-(1H-Benzo[d]imidazole-2-yl)phenyl)-3-(trifluoromethyl)benzamide (4)
3-(Trifluoromethyl)benzoic acid (266 mg, 1.4 mmol) and 1 (146 mg, 0.7 mmol) afforded a white
solid. The resulting solid was recrystallized from ethanol to afford white crystals of (4) (119 mg,
45%).
ATR-FTIR νmax /cm-1 3267 (NH), 1655 (C=O amide); m.p. 302 - 303 °C; Rf (MeOH/DCM 5:95)
0.30; δH (DMSO-d6, 400 MHz) 7.16-7.23 (m, 2H), 7.52 (d, J = 7.4 Hz, 1H), 7.65 (d, J = 7.1 Hz,
1H), 7.81 (t, J = 7.8 Hz, 1H), 7.96-8.00 (m, 3H), 8.20 (d, J = 8.8 Hz, 2H), 8.30 (d, J = 7.8 Hz,
1H), 8.34 (s, 1H), 10.66 (s, 1H), 12.81 (s, 1H); δC (DMSO-d6, 100.6 MHz) 111.1, 118.6, 120.5,
121.5, 122.3, 124.2, 125.7 (Cquat), 125.3 (Cquat), 126.9, 128.2, 129.2 (q, J = 127.7 Hz, CF3),
129.7, 131.8, 135.0 (Cquat), 135.6 (Cquat), 140.2 (Cquat), 143.9 (Cquat), 151.0 (Cquat), 164.2 (CO);
HRMS-ES+ Calculated: 382.1167 [M+H]+ for C21H15N3OF3, Observed: 382.1166; C18 HPLC,
flow rate: 1 mL/min, H2O / acetonitrile (40:60), 99.6%.
6.1.2.3 N-(4-(1H-Benzo[d]imidazole-2-yl)phenyl)-5-bromonicotinamide (5)
5-Bromopyridine-3-carboxylic acid (283 mg, 1.4 mmol) and 1 (146 mg, 0.7 mmol) afforded a
white solid. The resulting solid was recrystallized from ethanol to afford white crystals of (5)
(211 mg, 77%).
ATR-FTIR νmax /cm-1 3306 (NH), 1653 (C=O amide); m.p. 325 - 326 °C; Rf (MeOH/DCM 1:9)
0.82; δH (DMSO-d6, 400 MHz) 7.20 (m, 2H), 7.58 (m, 2H), 7.95 (d, J = 8.8 Hz, 2H), 8.19 (d, J =
8.8 Hz), 8.58 (t, J = 2.1 Hz, 1H), 8.92 (d, J = 2.2 Hz , 1H), 9.10 (d, J = 1.9 Hz, 1H), 10.69 (s,

ACCEPTED MANUSCRIPT
28
1H), 12.83 (s, 1H); δC (DMSO-d6, 100.6 MHz) 111.3, 118.7, 120.0 (Cquat), 120.3, 121.8, 121.8,
125.8 (Cquat), 127.0, 132.0 (Cquat), 135.0 (Cquat), 137.7, 140.0 (Cquat), 143.8 (Cquat), 147.4, 151.0
(Cquat), 152.8, 162.7 (CO); HRMS-ES+ Calculated: 393.0351 [M+H]+ for C19H14Br79N4O,
Observed: 393.0354; C18 HPLC, flow rate: 1 mL/min, H2O / acetonitrile (40:60), 98.4%.
6.1.2.4 N-(4-(1H-Benzo[d]imidazole-2-yl)phenyl)-4-(tert-butyl)benzamide (6)
4-tert-Butylbenzoic acid (357 mg, 2.0 mmol) and 1 (209 mg, 1.0 mmol) afforded a white solid.
The resulting solid was recrystallized from ethanol to afford white crystals of (6) (297 mg, 87%).
ATR-FTIR νmax /cm-1 1652 (C=O amide); m.p. 317 - 318 °C; Rf (MeOH/DCM 7:93) 0.92; δH
(DMSO-d6, 400 MHz) 1.34 (s, 9H), 7.52 (m, 2H), 7.57 (d, J = 8.6 Hz, 2H), 7.81 (m, 2H), 7.95
(d, J = 8.6 Hz, 2H), 8.13 (d, J = 8.9 Hz, 2H), 8.36 (d, J = 8.9 Hz, 2H), 10.67 (s, 1H); δC (DMSO-
d6, 100.6 MHz) 30.9, 34.7 (Cquat), 113.8, 118.0 (Cquat), 120.2, 125.2, 125.4, 127.7, 128.8, 131.7
(Cquat), 132.3 (Cquat), 143.7 (Cquat), 148.7 (Cquat), 154.9 (Cquat), 166.0 (CO); HRMS-ES+
Calculated: 370.1919 [M+H]+ for C24H24N3O, Observed: 370.1914; C18 HPLC, flow rate: 1
mL/min, H2O / acetonitrile (40:60), 99.7%.
6.1.2.5 N-(4-(1H-Benzo[d]imidazole-2-yl)phenyl)-2-methoxybenzamide (8)
2-Methoxybenzoic acid (304 mg, 2.0 mmol) and 1 (209 mg, 1.0 mmol) afforded a brown solid.
The resulting solid was recrystallized from methanol and water to afford dark brown crystals of
(8) (280 mg, 82%).
ATR-FTIR νmax /cm-1 3331 (NH), 1661 (C=O amide); m.p. 133 - 134 °C; Rf (MeOH/DCM 1:9)
0.93; δH (DMSO-d6, 400 MHz) 3.93 (s, 3H), 7.09 (t, J = 7.4 Hz, 1H), 7.18-7.23 (m, 3H), 7.52 (m,
1H), 7.59 (m, 2H), 7.69 (m, 1H), 7.93 (d, J = 8.4 Hz, 2H), 8.16 (d, J = 8.4 Hz, 2H), 10.32 (s,