Hemozoin inhibiting 2-phenylbenzimidazoles active against malaria parasites

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-(1Hbenzo[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.

Graphical Abstract

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 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 coworkers 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 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 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 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.

Scheme 1.

(i) PPA, 220 °C, 5 h; (ii) pyridine, THF or DMF, −40 °C, 2 – 3 h.

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.

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).

Table 1.

Inhibition of β-hematin formation and parasite growth by compounds 2 – 19.

Compound	β-hematin inhibition IC50 (μM)	NF54 malaria parasite growth inhibition IC50 (μM)a
2	∼3,900b	255 ± 47
3	24.5 ± 4.2	1.3 ± 0.2
4	28.1 ± 1.7	1.34 ± 0.07
5	32.0 ± 1.4	0.41 ± 0.03
6	358 ± 5	2.0 ± 0.1
7	90 ± 19	7.2 ± 0.2
8	38.9 ± 1.5	5.50 ± 0.6
9	16.6 ± 0.3	0.8 ± 0.5
10	28.9 ± 1.6	12.3 ± 2.7
11	13.3 ± 0.5	0.67 ± 0.07
12	16.8 ± 0.4	0.53 ± 0.13
13	38.2 ± 1.4	15.0 ± 0.3
14	25.1 ± 1.1	1.21 ± 0.06
15	∼1,790b	64 ± 8
16	35.6 ± 1.7	11.6 ± 2.6
17	42.4 ± 2.1	7.2 ± 0.7
18	84.1 ± 2.4	5.9 ± 0.8
19	46.6 ± 1.7	1.2 ± 0.4
Chloroquine	31.5 ± 0.5c	0.016 ± 0.006

^aDetermined by the LDH method of Makler et al [36]

^bEstimated based on extrapolation

^cFrom 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 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 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.

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.

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-hematin) + 3.71(HBD) – 1.43(molecular depth) – 8.90. F = 9.81 > F_crit^0.01 = 5.56, r² = 0.68, P < 0.0001. t for the individual parameters > t_crit^0.10 = 1.761 (5.11, 1.88 and 2.56 respectively for the first, second and third coefficients).

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.

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.

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).

Table 2.

Parasite growth inhibition IC₅₀ in chloroquine sensitive (NF54) and resistant (Dd2) strains and cytotoxicity in CHO cells.

Compound	IC50 Dd2 (μM) IC50 NF54 (μM)		RIa	IC50 CHO (μM)	SI (NF54)b	SI (Dd2)b
5	0.96 ± 0.04	0.41 ± 0.03	2.4	189 ± 4	461	197
9	1.10 ± 0.37	0.80 ± 0.50	1.4	35 ± 4	44	32
11	0.31 ± 0.02	0.67 ± 0.07	0.46	206	308	665
12	0.34 ± 0.03	0.53 ± 0.13	0.64	184 ± 9	347	541
Chloroquine	0.15 ± 0.03	0.016 ± 0.006	9.4	−	−	−
Emetine	−	−		0.12 ± 0.06	−	−

^aResistance index = IC₅₀ (Dd2)/IC₅₀ (NF54);

^bSelectivity 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 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	HBD #	HBA #
2	313.35	4.08	2	2
3	392.25	5.05	2	2
4	381.35	5.14	2	2
5	393.24	4.42	2	3
6	369.46	5.77	2	2
7	382.24	4.72	2	2
8	343.38	3.81	2	3
9	382.34	4.65	2	3
10	358.35	3.68	2	2
11	403.09	3.84	2	2
12	383.23	5.17	2	3
13	347.80	4.25	2	2
14	347.80	5.17	2	2
15	347.80	4.53	2	2
16	314.34	3.19	2	3
17	319.38	3.68	2	2
18	303.10	3.84	2	3
19	302.33	2.24	3	2

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 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.

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 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).

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 (E_ads⁰¹¹) 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.

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 β-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 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).

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.

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 NMR and δ 39.52 in ¹³C NMR), chloroform-d (δ 7.26 in ¹H 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 δ 206.26 in ¹³C 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.

Melting points, HRMS data and HPLC purity data for compounds for which syntheses have been previously reported are presented in Table 4. Further details of the syntheses are available in Supplementary Data.

Table 4.

Melting points, HRMS and purity data for compounds for which syntheses and characterization data have previously been reported.

Compound	MP (°C)	HRMSa	Purityb
1 4-(1H-benzo[d]imidazol-2yl)aniline	240–241 lit. 240–241 [47]	Obs: 210.1031 Calc: 210.1031	99.1%
2 N-4-(1H-benzo[d]imidazol-2-yl)phenyl)benzamide	340–341 lit. 335 [48]	Obs: 314.1293 Calc: 314.1293	98.2%
7 N-(4-(1H-benzo[d]imidazole-2-yl)phenyl)-2,4-dichlorobenzamide	273–275 lit. 242 [49]	Obs: 382.0508 Calc: 382.0514	99.6%
11 N-(4-(1H-benzo[d]imidazole-2-yl)phenyl)-3,5-dinitrobenzamide	338–339 lit. 190 [50]	Obs: 404.1004 Calc: 404.0995	99.2%
13 N-(4-(1H-benzo[d]imidazole-2-yl)phenyl)-2-chlorobenzamide	275–276 lit. 311 [49]	Obs: 348.0901 Calc: 348.0904	100.0%
14 N-(4-(1H-benzo[d]imidazole-2-yl)phenyl)-3-chlorobenzamide	320–322 lit. 312 [49]	Obs: 348.0899 Calc: 348.0904	95.9%
15 N-(4-(1H-benzo[d]imidazole-2-yl)phenyl)-4-chlorobenzamide	331–333 lit. 323 [49]	Obs: 348.0898 Calc: 348.0904	100.0%
17 N-(4-(1H-benzo[d]imidazole-2-yl)phenyl)thiophene-2carboxamide	344–346 lit. 349–350 [51]	Obs: 320.0862 Calc: 320.0858	97.3%
18 N-(4-(1H-benzo[d]imidazole-2-yl)phenyl)furan-2-carboxamide	252–253 lit. 301 [49]	Obs: 304.1072 Calc: 304.1086	95.1%
20 N-(4-(1H-benzo[d]imidazole-2-yl)phenyl)acetamide	302–303 lit. 308–309 [33]	Obs: 252.1135 Calc: 252.1137	98.6%

^a[M+H]⁺;

^bHPLC

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 a small round-bottomed flask, which was flushed with N₂(g). SOCl₂ (3 mL, 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 ν_max /cm⁻¹ 3248 (NH), 1652 (C=O amide); m.p. 324 – 325 °C; R_f (MeOH/DCM 5:95) 0.39; δ_H (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); δ_C (DMSO-d₆, 100.6 MHz) 111.1, 118.6, 120.4, 121.5, 121.7, 122.3 (C_quat), 125.6 (C_quat), 126.9, 126.9, 130.3, 130.6, 134.4, 135.0 (C_quat), 136.9 (C_quat), 140.3 (C_quat), 143.9 (C_quat), 151.1 (C_quat), 164.1 (CO); HRMS-ES⁺ Calculated: 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 ν_max /cm⁻¹ 3267 (NH), 1655 (C=O amide); m.p. 302 – 303 °C; R_f (MeOH/DCM 5:95) 0.30; δ_H (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); δ_C (DMSO-d₆, 100.6 MHz) 111.1, 118.6, 120.5, 121.5, 122.3, 124.2, 125.7 (C_quat), 125.3 (C_quat), 126.9, 128.2, 129.2 (q, J = 127.7 Hz, CF₃), 129.7, 131.8, 135.0 (C_quat), 135.6 (C_quat), 140.2 (C_quat), 143.9 (C_quat), 151.0 (C_quat), 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 ν_max /cm⁻¹ 3306 (NH), 1653 (C=O amide); m.p. 325 – 326 °C; R_f (MeOH/DCM 1:9) 0.82; δ_H (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, 1H), 12.83 (s, 1H); δ_C (DMSO-d₆, 100.6 MHz) 111.3, 118.7, 120.0 (C_quat), 120.3, 121.8, 121.8, 125.8 (C_quat), 127.0, 132.0 (C_quat), 135.0 (C_quat), 137.7, 140.0 (C_quat), 143.8 (C_quat), 147.4, 151.0 (C_quat), 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 ν_max /cm⁻¹ 1652 (C=O amide); m.p. 317 – 318 °C; R_f (MeOH/DCM 7:93) 0.92; δ_H (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); δ_C (DMSOd₆, 100.6 MHz) 30.9, 34.7 (C_quat), 113.8, 118.0 (C_quat), 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 ν_max /cm⁻¹ 3331 (NH), 1661 (C=O amide); m.p. 133 – 134 °C; R_f (MeOH/DCM 1:9) 0.93; δ_H (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, 1H); δ_C (DMSO-d₆, 100.6 MHz) 55.6, 112.0, 114.8, 119.7, 120.5, 121.9, 124.7 (C_quat), 125.0 (C_quat), 127.0, 129.7, 132.2, 139.2 (C_quat), 140.5 (C_quat), 151.1 (C_quat), 156.5 (C_quat), 164.6 (CO); HRMS-ES⁺ Calculated: 344.1399 [M+H]⁺ for C₂₁H₁₈N₃O₂, Observed: 344.1393; C18 HPLC, flow rate: 1 mL/min, H₂O / acetonitrile (40:60), 98.3%.

6.1.2.6. N-(4-(1H-Benzo[d]imidazole-2-yl)phenyl)-3-(trifluoromethyl)nicotinamide (9)

5-(Trifluoromethyl)pyridine-3-carboxylic acid (191 mg, 1.0 mmol) and 1 (105 mg, 0.5 mmol) afforded a white solid. The resulting solid was recrystallized from ethanol to afford white crystals of (9) (116 mg, 68%).

ATR-FTIR ν_max /cm⁻¹ 1661 (C=O amide); m.p. 318 – 320 °C; R_f (MeOH/DCM 1:9) 0.11; δ_H (DMSO-d₆, 400 MHz) 7.20 (m, 2H), 7.59 (m, 2H), 7.97 (d, J = 8.8 Hz, 2H), 8.21 (d, J = 8.8 Hz, 2H), 8.71 (m, 1H), 9.20 (m, 1H), 9.41 (m, 1H), 10.81 (s, 1H); δ_C (DMSO- d₆, 100.6 MHz) 115.1, 120.4, 122.0, 124.9 (q, J = 127.8 Hz, CF₃), 125.8 (C_quat), 127.0, 130.5 (C_quat), 132.7 (C_quat), 132.7, 139.9 (C_quat), 148.6, 148.6 (C_quat), 150.9 (C_quat), 152.6, 162.7 (CO); HRMS-ES⁺ Calculated: 383.1120 [M+H]⁺ for C₂₀H₁₄F₃N₄O, Observed: 383.1117; C18 HPLC, flow rate: 1 mL/min, H₂O / acetonitrile (40:60), 95.0%.

6.1.2.7. N-(4-(1H-Benzo[d]imidazole-2-yl)phenyl)-2-nitrobenzamide (10)

2-Nitrobenzoyl chloride (223 mg, 1.2 mmol) and 1 (126 mg, 0.6 mmol) afforded a yellow solid. The resulting solid was recrystallized from methanol to afford yellow crystals of (10) (145 mg, 67%).

ATR-FTIR ν_max /cm⁻¹ 1652 (C=O amide), 1527 (N-O stretch); m.p. 308 – 309 °C; R_f (MeOH/DCM 5:95) 0.36; δ_H (DMSO-d₆, 400 MHz) 7.16–7.24 (m, 2H), 7.48–7.57 (m, 1H), 7.61–7.69 (m, 1H), 7.74–7.93 (m, 5H), 8.14–8.22 (m, 3H), 10.80 (s, 1H), 12.73 (s, 1H); δ_C (DMSO-d₆, 100.6 MHz) 110.9, 118.5, 119.6, 121.3, 122.1, 124.0, 125.6 (C_quat), 127.0, 129.1, 130.8, 132.4 (C_quat), 133.9, 134.9 (C_quat), 140.0 (C_quat), 143.8 (C_quat), 146.3 (C_quat), 150.9 (C_quat), 164.1 (CO); HRMS-ES⁺ Calculated: 359.1144 [M+H]⁺ for C₂₀H₁₅N₄O₃, Observed: 359.1151; C18 HPLC, flow rate: 1 mL/min, H₂O / acetonitrile (40:60), 99.8%.

6.1.2.8. N-(4-(1H-Benzo[d]imidazole-2-yl)phenyl)-2,6-dichloroisonicotinamide (12)

2,6-Dichloropyridine-4-carboxylic acid (384 mg, 2.0 mmol) and 1 (209 mg, 1.0 mmol) afforded a yellow solid. The resulting solid was recrystallized from methanol and water to afford pale yellow crystals of (12) (264 mg, 69%).

ATR-FTIR ν_max /cm⁻¹ 1664 (C=O amide); m.p. 348 – 349 °C; R_f (MeOH/DCM 1:9, 1% Et₃N) 0.32; δ_H (DMSO-d₆, 400 MHz) 7.20 (m, 2H), 7.59 (m, 2H), 7.93 (d, J = 8.8 Hz, 2H), 8.03 (s, 2H), 8.20 (d, J = 8.8 Hz, 2H), 10.80 (s, 1H), 12.80 (s, 1H); δ_C (DMSO-d₆, 100.6 MHz) 111.1, 111.4, 118.3, 120.4, 121.8, 121.8, 126.2 (C_quat), 127.0, 139.5 (C_quat), 144.1 (C_quat), 144.1 (C_quat), 148.0 (C_quat), 149.8 (C_quat), 150.9 (C_quat), 161.3 (CO); HRMS-ES⁺ Calculated: 383.0466 [M+H]⁺ for C₁₉H₁₃Cl₂N₄O, Observed: 383.0468; C18 HPLC, flow rate: 1 mL/min, H₂O / acetonitrile (40:60), 98.7%.

6.1.2.9. N-(4-(1H-Benzo[d]imidazole-2-yl)phenyl)nicotinamide (16)

Nicotinic acid (246 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 light brown crystals of (16) (211 mg, 67%).

ATR-FTIR ν_max /cm⁻¹ 3276 (NH), 1648 (C=O amide); m.p. 332 – 333 °C; R_f (MeOH/DCM 15:85) 0.17; δ_H (DMSO-d₆, 400 MHz) 7.16–7.24 (m, 2H), 7.50–7.72 (m, 3H), 7.97 (m, 2H), 8.12–8.24 (m, 2H), 8.30–8.36 (m, 1H), 8.76–8.81 (m, 1H), 9.14–9.19 (m, 1H), 10.64 (s, 1H), 12.81 (s, 1H); δ_C (DMSO-d₆, 100.6 MHz) 115.1, 120.2, 121.6, 123.2, 125.6 (C_quat), 126.8, 130.3 (C_quat), 135.1, 140.0 (C_quat), 142.9 (C_quat), 148.4, 150.9 (C_quat), 151.9, 164.0 (CO); HRMS-ES⁺ Calculated: 315.1246 [M+H]⁺ for C₁₉H₁₅N₄O, Observed: 315.1249; C18 HPLC, flow rate: 1 mL/min, H₂O / acetonitrile (40:60), 97.7%.

6.1.2.10. N-(4-(1H-Benzo[d]imidazole-2-yl)phenyl)1H-pyrrole-2-carboxamide (19)

Pyrrole-2-carboxylic acid (222 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 a light brown solid of (19) (162 mg, 54%).

ATR-FTIR ν_max /cm⁻¹ 1654 (C=O amide); m.p. 292 – 293 °C; R_f (MeOH/DCM 1:9) 0.61; δ_H (DMSO-d₆, 400 MHz) 6.21 (m, 1H), 7.03 (m, 1H), 7.17 (m, 1H), 7.49–7.54 (m, 2H), 7.77–7.83 (m, 2H), 8.08–8.13 (d, J = 8.9 Hz, 2H), 8.25–8.30 (d, J = 8.9, 2H), 10.28 (s, 1H), 11.84 (s, 1H); δ_C (DMSO-d₆, 100.6 MHz) 109.1, 112.5, 113.8, 117.6 (C_quat), 119.7, 123.3, 125.2, 125.6 (C_quat), 128.6, 132.7 (C_quat), 143.8 (C_quat), 149.1 (C_quat), 159.3 (CO); HRMS-ES⁺ Calculated: 303.1246 [M+H]⁺ for C₁₈H₁₅N₄O, Observed: 303.1233; C18 HPLC, flow rate: 1 mL/min, H₂O / acetonitrile (40:60), 95.0%.

6.1.2.11. N-(4-(1H-Benzoyl-benzo[d]imidazole-2-yl)phenyl)-3-bromobenzamide (26)

Compound 3 (300 mg, 0.76 mmol), was added to a solution of anhydrous pyridine (2 mL) and THF (2 mL) in a 25 mL reaction vessel under N₂(g). Benzoyl chloride (0.8 mL, 6.9 mmol) was added and the solution left to stir for 18 h at room temperature. The reaction mixture was quenched with MeOH, acidified with 1M HCl to pH 5 and the combined organic components extracted, first using MeOH: chloroform (1:3, 20 mL) and then chloroform (2× 20 mL). The organic extracts were dried over anhydrous Na₂SO₄ and the crude product purified using column chromatography with MeOH: DCM (1:99 to 7:93). The resulting solid was further recrystallized from acetone and H₂O which afforded a light brown solid of 26 (146 mg, 39% yield).

ATR-FTIR ν_max/cm⁻¹ 1657 (C=O amide); m.p. 222 – 223 °C; R_f (MeOH/DCM 1:9) 0.90; δ_H (DMSO-d₆, 400 MHz) 7.30–7.33 (m, 2H), 7.35–7.41 (m, 1H), 7.44–7.52 (m, 3H), 7.60–7.65 (m, 3H), 7.75–7.80 (m, 5H), 7.84 (dt, J = 0.9, 7.9 Hz, 1H), 7.94 (ddd, J = 1.0, 1.6, 7.8 Hz, 1H), 8.13 (t, J = 1.8 Hz, 1H), 10.45 (s, 1H); δC (DMSO-d₆, 100.6 MHz) 112.7, 119.6, 119.7, 121.6 (C_quat), 124.2, 124.4, 125.3 (C_quat), 126.8, 128.9, 129.6, 130.2, 130.3, 130.6, 132.7 (C_quat), 134.3, 134.4, 134.6 (C_quat), 136.7 (C_quat), 140.2 (C_quat), 142.6 (C_quat), 153.2 (C_quat), 164.1 (CO), 168.9 (CO); HRMS-ES⁺ Calculated: 496.0661 [M+H]⁺ for C₂₇H₁₉Br⁷⁹N₃O₂, Observed: 496.0654; C18 HPLC, flow rate: 1 mL/min, H₂O/acetonitrile (40:60), 96.6%.

6.2. Molecular Docking

6.2.1. Force field development.

All simulations were performed using the BIOVIA MS software package. Additional parameters were added to the cvff force field to account for the Fe(III) center in Fe(III)PPIX. The crystal structures of β-hematin [52] and eight related high-spin Fe(III) porphyrin complexes [53–58] were used to determine average values of bond lengths (including the axial Fe–O bond), bond angles and dihedral angles around the porphyrin core. The crystal structures were first minimized using the BIOVIA MS DMol³ tool, which uses a density functional theory (DFT) calculation together with the generalized gradient approximation (GGA, PW91) [59], and a double numerical basis set with polarization (DNP) [60]. All structures were then charged using the charge equilibration method (QEq) [61]. The force field was considered optimized when these specific bond lengths, angles, and dihedral angles did not deviate by more than 0.015 Å, 1.5° and 4° respectively for a calculated structure compared to the averaged reference structure [62]. The parameters for the force field parameterization and structure optimizations are reported in Tables S2 and S3, respectively.

6.2.2. Adsorption

The optimized β-hematin μ-propionato dimer was used to build a super cell (3×3×3 unit cells) using the MS Super Cell building function. The prepared crystal lattice, including surface functional groups was constrained during simulations. A vacuum slab was created in front of the face of interest; its purpose is to terminate the periodicity of the crystal system in the direction perpendicular to it, while also demarcating the region within which an inhibitor is permitted to move during the adsorption calculation. Calculations were performed with derivatives containing both neutral and monoprotonated forms of the benzimidazole core. The latter were found to better predict activities and were then used in further simulations. The 18 compounds synthesized in this study were built and optimized in vacuo using geometry and energy optimization tools in BIOVIA MS. The Adsorption Locator tool was then applied in order to identify the preferred adsorption site(s) of the inhibitors on a particular face.[44] Specifically, a series of Monte Carlo searches was carried out together with simulated annealing dynamics in order to survey the available space of both inhibitor and target. Once adsorbed, a second simulated annealing calculation was required to establish the global minimum configuration of the inhibitor within the (rigid) binding site. The adsorption and simulated annealing parameters are reported in Tables S4 and S5, respectively. The E_ads value is the total change in energy of the system upon adsorption of the inhibitor to the crystal surface, which considers the interaction energy between the inhibitor and the crystal surface, as well as the change in internal energy of the inhibitor. With the crystal lattice kept rigid, it does not contribute to the total change in energy. Thus, E_ads is determined from Equation 1, where E_sys is final energy of the system, and E_c and E_i are the initial energies of the crystal and inhibitor, respectively. Values of E_ads are expected to be negative owing to the stabilization of an inhibitor following adsorption; the more negative the value, the more favorable the adsorption of an inhibitor to the crystal surface.

$$E_{ads}=E_{sys}-\left(E_c+E_i\right)$$ (1)

Highlights.

• Bayesian statistics predicted β-hematin inhibition with a hit rate of 83%

• Bayesian statistics predicted parasite growth inhibition with a hit rate of 50%

• Malaria parasite growth inhibition correlated with β-hematin inhibition activity

• 2-phenylbenzimidazoles were selective with no cross-resistance with chloroquine

• Molecular docking with the hemozoin surface predicted β-hematin inhibition activity

Appendix A. Supplementary data

Supplementary data related to this article can be found at…