Nitroimidazole carboxamides as antiparasitic agents targeting Giardia lamblia, Entamoeba histolytica and Trichomonas vaginalis

Diarrhoeal diseases caused by the intestinal parasites Giardia lamblia and Entamoeba histolytica constitute a major global health burden. Nitroimidazoles are first-line drugs for the treatment of giardiasis and amebiasis, with metronidazole 1 being the most commonly used drug worldwide. However, treatment failures in giardiasis occur in up to 20% of cases and development of resistance to metronidazole is of concern. We have re-examined ‘old’ nitroimidazoles as a foundation for the systematic development of next-generation derivatives. Using this approach, derivatisation of the nitroimidazole carboxamide scaffold provided improved antiparasitic agents. Thirty-three novel nitroimidazole carboxamides were synthesised and evaluated for activity against G. lamblia and E. histolytica. Several of the new compounds exhibited potent activity against G. lamblia strains, including metronidazole-resistant strains of G. lamblia (EC50 = 0.1–2.5 μM cf. metronidazole EC50 = 6.1–18 μM). Other compounds showed improved activity against E. histolytica (EC50 = 1.7–5.1 μM cf. metronidazole EC50 = 5.0 μM), potent activity against Trichomonas vaginalis (EC50 = 0.6–1.4 μM cf. metronidazole EC50 = 0.8 μM) and moderate activity against the intestinal bacterial pathogen Clostridium difficile (0.5–2 μg/mL, cf. metronidazole = 0.5 μg/mL). The new compounds had low toxicity against mammalian kidney and liver cells (CC50 > 100 μM), and selected antiparasitic hits were assessed for human plasma protein binding and metabolic stability in liver microsomes to demonstrate their therapeutic potential.


Introduction
Diarrhoeal diseases caused by intestinal protozoan parasites are a major global health burden. Two of the most common intestinal parasites, Giardia lamblia and Entamoeba histolytica, are responsible for~280 million and~50 million annual infections, respectively [1,2]. Transmission of these parasites occurs by the faecal-oral route through ingestion of cysts in contaminated water or food, or by direct person-to-person contact. G. lamblia may also have animal reservoirs, making the infection a potential zoonotic disease [3].
Upon ingestion of G. lamblia cysts, trophozoites emerge from the cysts and multiply in the lumen of the small intestine, where they can attach to the intestinal mucosa. Symptoms of acute giardiasis include watery diarrhoea, abdominal discomfort, pain and cramps. Chronic disease can result in malabsorption and failure to thrive in children [4]. For amebiasis, trophozoites migrate to the large intestine and can either reside in the lumen or invade the colonic mucosa or other extra-intestinal sites, most prominently the liver [5].
Due to the propensity for spread through contaminated water and food sources and the low infectious dose of G. lamblia and E. histolytica cysts [6], the global disease burden is disproportionately shouldered by developing nations in areas with inadequate sanitation. Protozoan diseases also impact developed nations, often via travellers visiting regions where disease is endemic. The threat to developed nations is recognised by the US National Institute of Allergy and Infectious Diseases as both protozoa are category B bioterrorism threat pathogens [7].
Metronidazole 1 (Fig. 1) is a generic drug for treatment of a range of parasitic and anaerobic bacterial infections. For giardiasis, metronidazole is typically given in 250 mg doses three times a day for 5e7 days, while amebiasis is treated with a higher 750 mg dose three times a day for 5e10 days, often followed by treatment with paromomycin to eradicate cysts from the colon [8,9]. Other 5nitroimidazoles, such as tinidazole 2 and ornidazole 3 ( Fig. 1) have improved dosing schedules with only a single 2 g tablet of either drug for treatment of giardiasis, or 2 g tinidazole once daily for three days for treatment of amebiasis [8e10]. These agents have similar adverse effects such as nausea, vomiting and headaches. Ornidazole is not approved for use in the United States [8e10]. Unfortunately, metronidazole treatment fails in up to 20% of giardiasis cases with metronidazole resistance an ever increasing concern [11,12]. Parasites resistant to metronidazole show crossresistance to tinidazole [1]. Furthermore, resistance of E. histolytica to metronidazole has also been described, as trophozoites can be adapted to grow in the presence of therapeutically relevant levels of metronidazole [13]. Given the sheer number of cases of giardiasis and amebiasis, and treatment failures, development of alternative treatment options remains an important priority.
Re-examination of 'old' nitroimidazoles is a valuable strategy in the development of new drugs for treatment of parasitic diseases. For example, fexinidazole 4 ( Fig. 1), initially discovered in the 1980s, has been "rediscovered" and is in clinical development by the Drugs for Neglected Diseases initiative for treatment of Human African trypanosomiasis (sleeping sickness) and Chagas disease [14]. Metronidazole has been in clinical use for over 50 years, but the expanded potential of metronidazole based agents has recently been demonstrated by modifying metronidazole with a "click chemistry" approach to generate agents with improved potency and activity against metronidazole resistant (MtzR) parasites [15,16].
Nitroimidazole carboxamides (Fig. 1) were originally patented by Merck &. Co. in 1973 for the treatment of infections caused by Histomonas meleagridis and Trichomonas vaginalis [17]. H. meleagridis is a parasite that causes lesions in the cecum and liver of chickens and turkeys, and is commonly known as turkey blackhead disease [18]. In contrast, T. vaginalis infects the genitourinary tract in humans causing inflammation and vaginal discharge in women [19]. The nitroimidazole carboxamides displayed efficacy in in vivo turkey and mouse models of Histomonas maleagridis and T. vaginalis [17], respectively, but no substantial antimicrobial development of this series has since been reported. Given the core 5-nitroimidazole group in the nitroimidazole carboxamides is similar to metronidazole, we hypothesised that these compounds could have therapeutic potential against enteric parasites, including G. lamblia and E. histolytica. In addition, the 2 0 -carboxamide substitution provides a convenient handle to optimise antiparasitic properties. Therefore, we explored the structure activity relationships (SAR) of nitroimidazole carboxamides and conducted preliminary ADME studies to identify improved antiparasitic agents with therapeutic potential against G. lamblia and E. histolytica.

Synthesis of 1-methyl-5-nitroimidazoles
The library of 1-methyl-5-nitroimidazole carboxamides 8a-k was prepared essentially as described by Hoff [17] using the synthetic approach depicted in Scheme 1. Commercially available 1methyl-2-hydroxymethyl-5-nitro imidazole 5 was oxidised with potassium permanganate in acetone to form the potassium carboxylate salt 6, which was isolated in this form to avoid decarboxylation of the free carboxylic acid [17]. The crude carboxylate salt 6, upon reaction with oxalyl chloride and catalytic DMF, provided the acid chloride intermediate 7. The library of 5nitroimidazole carboxamides 8a-k was then prepared by reacting the crude acid chloride 7 with the desired primary or secondary amines in the presence of triethylamine (Scheme 1). Compounds 8a-f and 8i-k are first reported here. Compounds 8g (R ¼ NMe 2 ) and 8h (R ¼ morpholine), originally reported by Hoff [17], were prepared for use as comparators due to their activity in in vivo models of T. vaginalis and H. meleagridis infection. The title compounds 8a-k were all purified by direct or reverse phase chromatography to !95% purity before biological testing. All compounds were characterised by 1 H and 13 C NMR, LCMS and HRMS and detailed experimental procedures and characterisation data are provided in the supplementary information.

Synthesis of 4(5)-nitroimidazoles
To examine the corresponding 4(5)-nitroimidazole carboxamide series of 8a-k (i.e no N-methyl substitution) we prepared the analogous series of novel compounds 12a-k. In addition, four alternative novel carboxamides 12l-o were prepared, as shown in Scheme 2. Imidazole-2-carboxylic acid 9 was readily nitrated with conc. HNO 3 /H 2 SO 4 to give 4(5)-nitroimidazole carboxylic acid 10. Carboxamides 12a-o were subsequently prepared by activation of acid 10 (oxalyl chloride/catalytic DMF or PyBOP/DIPEA) followed by coupling of the requisite amine. Amidation via intermediate 11 was the preferred route due to the difficulty of removing the HOBt and tripyrrolidinophosphine oxide by-products formed during the PyBOP mediated coupling. The primary amide 12l was prepared by quenching the acid chloride 11 with concentrated ammonium hydroxide solution. The title compounds 12a-o were all purified and characterised as described for 8a-k.

Synthesis of 4-nitroimidazoles
The novel 4-nitroimidazoles 13a-g and 14a-c were synthesised from the respective 1H-imidazole carboxamides 12g, 12l, 12m and 12p by alkylation with benzyl or alkyl halides under basic conditions (K 2 CO 3 ) (Scheme 3). The hydroxamic acid 14d was prepared by treatment of the ester 14c with hydroxylamine in methanol at 60 C. The hydrazide 14e was prepared from 14f via an acid chloride intermediate and hydrazine. The title compounds 13a-g and 14a-e were purified by chromatography or recrystallisation and characterised as described for 8a-k.   [20,21]), E. histolytica strain HM1:IMSS and T. vaginalis strain F1623 [15] were axenically maintained in TYI-S-33 medium supplemented with penicillin (100 U/mL) and streptomycin (100 mg/mL) [22]. All experiments were performed using trophozoites harvested during the logarithmic phase of growth.

EC 50 assays
Compounds were screened for antiparasitic activity using an ATP-bioluminescence based assay for cell growth and survival [23,24]. Briefly, 2.5 mL of 5 mM stocks were diluted with 17.5 mL sterile water to yield 625 mM working concentration of compounds.
Three-fold serial dilutions were prepared yielding a concentration range of 0.25e625 mM. From this dilution plate, 4 mL volumes were transferred into 96-well microtitre plates followed by addition of 96 mL of trophozoites (5000 parasites) to yield a final 8-point concentration range spanning 0.01e25 mM. Assay plates were incubated for 24e48 h at 37 C in the GasPak™ EZ Anaerobe Gas Generating Pouch Systems (VWR, West Chester, PA) to maintain  anaerobic condition throughout the incubation period. Viable cell numbers were determined in triplicate using the CellTiter-Glo Luminescent Cell Viability Assay [23].

MIC assays
Clostridium difficile strains (630, ATCC BAA-1382 and NAP1/027, ATCC BAA-1803) were maintained as previously described [16]. The minimum inhibition concentration (MIC) was determined according to the CLSI Methods with modifications in broth and inoculum for C. difficile [16,25,26]. Briefly, compounds and control antibiotics were serially diluted two-fold in 96-well plates (Non-binding surface, Corning). The plates were placed in a Coy anaerobic chamber (5% H 2 , 10% CO 2 , 85% N 2 ) overnight to reduce the medium. C. difficile bacteria from BHIS(TA) agar plates were cultured anaerobically in BHIS at 37 C overnight. A sample of culture was then diluted 40fold in BHIS broth and incubated at 37 C for approximately 4.5 h. The resultant mid-log phase culture (OD 600 ¼ 0.4e0.6) was diluted to a final concentration of~1 Â 10 6 CFU/mL, then 50 mL was added to each well of the compound-containing 96-well plates, yielding a final cell concentration of 5 Â 10 5 CFU/mL and final volume of 100 mL with 3% maximum DMSO concentration. Compound concentration ranged from 64 to 0.03 mg/mL. An antibiotic standard, a positive growth control (no compound) and a sterility control (no bacteria) were included on each 96 well plate. Plates were covered and incubated at 37 C for 24 h. MICs for each strain were determined as the lowest concentration without visible growth. Variance between replicates was typically within one 2-fold dilution. Median MICs are reported with a range given when the median MIC was between two tested concentrations.

Cytotoxicity
Human HEK293 and HepG2 cells were seeded at 3000 and 5000 cells per well in 384-well plates, respectively. Cells were cultured in Dulbecco's modified Eagle's medium with 10% FBS for Scheme 3. Synthesis of 4-nitroimidazoles 13a-g and 14a-f. i) benzyl or alkyl bromide, K 2 CO 3 , DMF, rt / mW 80 C, 7e98%; ii) HNO 3 , H 2 SO 4 , 60 C, 64%; iii) NH 2 OH, MeOH, 60 C, 24 h at 37 C, 5% CO 2 . A dilution series of compounds was added, with the highest concentration of 100 mM. The final concentration of DMSO in culture media was 0.5%, which showed no effect on cell growth. After 24 h incubation with the compounds, 5 mM resazurin was added into each well and incubated at 37 C for 2 h. As a negative control, 1% Triton X-100 was added into the culture media to lyse all of the cells. The fluorescence intensity was read using Polarstar Omega with excitation/emission 560/590 nm. Data were analysed with GraphPad Prism 6 software (La Jolla, California USA) to calculate CC 50 values.

Correlation of compound properties with activity
A correlation matrix between compound activity and physicochemical properties was calculated using Excel correlation analysis (Supplementary Table 2). AlogP, logD, MW, logS and tPSA were calculated from the 2D structure of the compounds, using Pipeline Pilot (Accelrys, Version 9.1.0.13). Antimicrobial activity was expressed as -log 10 values of MIC or EC 50 , using average MIC (mol L À1 ) of MtzS C. difficile ATCC BAA-1382 and ATCC BAA-1803 strains and EC 50 (mol L À1 ) against G. lamblia WB strain, E. histolytica HM1:1MSS strain and T. vaginalis F1623 strain. Correlations of determination (R 2 ) between compound activities and logD, MW or logS were determined by linear regression analysis in GraphPad Prism 6 software (La Jolla, California USA) (Supplementary Figs. 1e3).

Microscopy
The effect of compounds on G. lamblia WB growth and survival was examined by light microscopy. Briefly, stock compounds were diluted in DMSO (100%) to give 400 Â final concentration of com- pound. An aliquot of 2.5 mL of working stock was added to each well of a 24 well tissue culture clear bottom plate (Corning, 3524), followed by trophozoites (1 mL, 50,000 parasites/mL) to yield a final concentration of 3 Â EC 50 . Metronidazole (3 Â EC 50 ) served as a positive control. Media only wells were included as a sterility control, and vehicle only (0.25% DMSO) was included as a control for growth. Assay plates were incubated for 48 h at 37 C in the GasPak™ EZ Anaerobe Gas Generating Pouch Systems (VWR, West Chester, PA) to maintain anaerobic conditions throughout the incubation period. The assays were performed in triplicate (3 wells/ treatment). Growth inhibition was visualised by phase contrast microscopy (200 Â magnification) (Carl Zeiss).

Metabolic stability
Metabolic stability studies were performed with human liver microsomes (HMMC-PL, Lot# PL050B-B, Thermo Fisher Scientific USA) with test compound (5 mM Table 1).
Since nitroimidazole carboxamides contain a 5-nitroimidazole warhead similar to metronidazole, a common treatment for infections caused by both parasites and anaerobic bacteria, we also determined the antimicrobial activity of 8a-k against the anaerobic bacterium C. difficile. C. difficile is an anaerobic Gram-positive bacterium that infects the colon and causes inflammation and diarrhoea [29], similar to symptoms of G. lamblia and E. histolytica infection. Surprisingly, none of the 1-methyl-5-nitroimidazole carboxamides 8a-k had significant activity against the 630 or NAP1/027 strains of C. difficile (MIC ! 32 mg/mL) whereas metronidazole was quite potent (MIC ¼ 0.5 mg/mL) ( Table 1,   Supplementary Table 3). Therefore this compound series, containing the 1-methyl-5-nitroimidazole core, exhibited greater selectivity toward anaerobic protozoan parasites compared to the anaerobic bacteria C. difficile, suggesting differences between the parasitic and bacterial proposed mechanisms of activation of the nitroimidazole carboxamides, or possibly differences in cellular uptake.
The compound series 8a-k was not cytotoxic against human kidney or liver cell lines at the highest concentration tested (CC 50 > 100 mM), so the calculated minimal selectivity indices ranged from >16 to >63 (Table 1).

Biological activity of 4(5)-nitroimidazoles
The promising activity of the 1-methyl-5-nitrocarboxamide series as antiparasitic agents led us to explore the nitroimidazole carboxamide scaffold further to understand the SAR and determine whether the activity could be improved. Hoff [17] previously reported the influence of alkyl and hydroxyl alkyl groups at the 1 0imidazole nitrogen position on antiparasitic activity but did not describe the corresponding compound series in which the 1 0imidazole nitrogen lacks substitution. Hence we generated an analogous series of 4(5)-nitroimidazole carboxamides (12a-k) with the unsubstituted 1-position and matched carboxamide R groups to 8a-k, and tested them for antiparasitic and antimicrobial activity. Compared to the 1-methyl-5-nitro series, the presence of an acidic imidazole proton in the 4(5)-nitro series permits ring tautomerism, which may alter the reduction potential of the nitro group. Furthermore, the small structural change will influence other physicochemical parameters (e.g. polar surface area and logS) which is likely to be reflected in different SAR profiles between the two series. In general the 4(5)-nitroimidazole carboxamides 12a-k exhibited improved activity against G. lamblia and E. histolytica relative to their 5-nitroimidazole counterparts 8a-k (Table 1). For G. lamblia, the aromatic benzyl amides 12a-d, phenethyl 12e and cyclohexyl 12k groups were very potent (EC 50 ¼ 0.1e0.6 mM). In contrast 12f (R ¼ NHeCH 2 (2-pyridinyl)), 12g (R ¼ NMe 2 ) and 12h (R ¼ morpholine) were 3.5e5.5-fold less active than the respective 1-methyl analogues 8f, 8g and 8h. Compounds 12i (R ¼ pyrrolidine) and 12j (R ¼ NH-cyclopropyl) maintained similar activity (EC 50 ¼ 3.4 and 5 mM, respectively) to the 1-methyl analogues 8i and 8j. A number of compounds with substituted benzyl groups (12a-b and 12d) and the phenethyl derivative 12e also displayed good activity against MtzR G. lamblia (EC 50 2.5 mM).
In contrast to their improved activity against G. lamblia and E. histolytica, compounds 12a-k were not overall more active than 8a-k against T. vaginalis ( Table 1). The SAR was relatively flat: the trend for improved potency with more polar substituents seen with series 8a-k disappeared. The most potent compound was 12d (R ¼ NHCHMe(4-F-Ph)) with EC 50 ¼ 0.6 mM. The other aromatic benzyl compounds 12a-12c, 12e and the pyrrolidine 12i had similar activity (EC 50 ¼ 1.2e2.3 mM), but were generally 2e3 fold less potent than 12d. Interestingly, the absence of N-substitution on the imidazole ring for 12a-k also greatly improved activity against both the 630 and NAP1/027 strains of C. difficile (MIC ¼ 0.5e16 mg/mL), whereas the 1-methyl-5-nitro series were all essentially inactive (!32 mg/mL) ( Table 1, Supplementary Table 3). Small lipophilic and polar 2 0 -carboxamide substituents were preferred in the case of C. difficile. For example, 12j (R ¼ NH-cyclopropyl) was the most active derivative against C. difficile (MIC ¼ 1 mg/mL), although less active than metronidazole (MIC ¼ 0.5 mg/mL), while 12f-i (pyridine, dimethyl, morpholine and pyrrolidine derivatives) had MIC ¼ 2 mg/ mL. In contrast, the aromatic benzyl 12a-d, phenethyl 12e and cyclohexyl 12k compounds were less active (MIC ¼ 4e16 mg/mL). To further understand this preference for activity against C. difficile, additional small, polar amides 12l-o were synthesised. These included 12l (R ¼ NH 2 ), 12m (R ¼ NHMe) and two compounds inspired from the side chain of metronidazole: 12n (R ¼ NHCH 2 CH 2 OH) and 12o (R ¼ NMeCH 2 CH 2 OH). Compounds 12l-m and 12o gave results that supported the previous trend observed against C. difficile (MIC ¼ 0.5e2 mg/mL), while 12n (R ¼ NHCH 2 CH 2 OH) was less active (MIC ¼ 8e16 mg/mL). These additional compounds 12l-o had weak to no activity against the parasites.
The majority of the 4(5)-imidazole series 12a-o were not cytotoxic at the highest concentration tested (CC 50 > 100 mM) against mammalian liver or kidney cell lines. The only compound found to show cytotoxicity was 12b (R ¼ NHCH 2 (4-OCF 3 -Ph)) against the HepG2 liver cell line (CC 50 ¼ 93 mM), but the selectivity index (SI ¼ 465) relative to G. lamblia activity remained excellent.

Phenotypic effect of 4(5)-nitroimidazole 12a on G. lamblia
Microscopy was used to visually examine the impact of one of the most potent compounds, the 4(5)-nitroimidazole 12a (R ¼ NHCH 2 (4-F-Ph)), on G. lamblia trophozoites. Parasite cell growth was similarly inhibited by treatment with 3 Â EC 50 of either metronidazole (18 mM) or compound 12a (1.5 mM) relative to the vehicle control (which produced a confluent cell layer under the test conditions). The morphology of G. lamblia treated with 12a was altered, while the morphology of the metronidazole-treated cells remained similar to the vehicle control (Fig. 2). A prior study by Tejman-Yarden et al. reported that metronidazole slowed the rate of oscillation of the Giardia flagella, while auranofin, a compound with a proposed different mode of action, caused cell blebbing [24]. The different morphology of the G. lamblia treated with the 4(5)nitroimidazole carboxamide 12a may indicate an additional mode of action compared to metronidazole.

Influence of physicochemical properties on compound activity in the 4(5)-nitromidazole compound series
We observed improved activity profiles of 4(5)-nitroimidazoles relative to the corresponding analogue in the 5-nitroimidazole series against G. lamblia, E. histolytica and C. difficile, but not T. vaginalis. In addition, the 4(5)-nitroimidazoles with the most potent activity against G. lamblia differed significantly from the compounds with the most potent activity against C. difficile. To better understand the relationship between biological activity and physicochemical properties, the correlation coefficients (r) were determined between a range of calculated compound properties (AlogP, logD, molecular weight, logS and topological polar surface area) and biological activity against the different organisms (Supplementary Table 2). G. lamblia activity was positively correlated with AlogP (r ¼ 0.94), logD (r ¼ 0.93) and MW (r ¼ 0.82). A negative correlation with LogS (r ¼ À0.92) was also observed, while there was no meaningful relationship with tPSA (r ¼ 0.06). Nearly identical results were obtained with logP and logD values as only 12f (R ¼ NHCH 2 (2-pyridinyl)) contained an ionisable group. Moderate to weak correlations were observed between E. histolytica or T. vaginalis activity and compound properties (AlogP, logD, MW, logS and tPSA). In contrast, C. difficile activity was positively correlated with LogS (r ¼ 0.72), negatively correlated with AlogP (r ¼ À0.72), logD (r ¼ À0.72) and MW (r ¼ À0.75) and poorly correlated with tPSA (r ¼ À0.23), supporting the qualitative observations made from examination of the SAR.
To quantify the extent that the variability in activity against each organism was dependent on logD, MW and logS, the coefficient of determination (R 2 ) was next calculated (Fig. 3, Supplementary  Figs. 1e3). This analysis supported the correlation between G. lamblia activity and logD, MW and logS properties of the compounds (R 2 ranged from 0.67 to 0.86) (Fig. 3, Supplementary Figs. 1e3). No correlation was found for E. histolytica and T. vaginalis activity and compound properties (R 2 ranged from 0.15 to 0.28) (Fig. 3, Supplementary Figs. 1e3). In contrast, a weak correlation between C. difficile activity and logD, MW and logS was observed (R 2 ranged from 0.47 to 0.56) (Fig. 3, Supplementary  Figs. 1e3), demonstrating greater variability in the data that was not accounted for by changes to logD, MW or logS.
To summarise, while activity against G. lamblia was improved by increasing logD, MW and decreasing logS, this trend was not apparent for E. histolytica or T. vaginalis. In contrast, activity against C. difficile was weakly improved with compounds with lower logD, MW and greater logS.

Biological activity of 4-nitroimidazoles
Given the potent activity of the 4(5)-nitroimidazoles relative to the 1-methyl-5-nitroimidazoles, we were interested to determine the activity of 4-nitroimidazole carboxamide analogues, since research by Trunz et al. showed that 4-nitroimidazoles can have potent antiparasitic activity [30]. We therefore prepared a series of 4-nitroimidazole carboxamides 13a-g. Since polar substituents were favourable for activity against G. lamblia in the 5-nitroimidazole series (though not the 4(5)-nitroimidazole series), compounds were synthesised with the 2 0 -substituent as a primary carboxamide group with the 1 0 -ring position substituted with benzyl, phenethyl, heteroaromatic pyridine, cyclopropyl and cyclohexyl groups. The compounds were found to have selective activity against G. lamblia ( Table 2). Several of these compounds,   To determine the relative influence of the 2 0 position on the potency and selectivity for G. lamblia, we next modified the 2 0 position to methyl amide 14a, dimethyl amide 14b, ethyl ester 14c, hydroxamide 14d and hydrazide 14e while maintaining the 1 0 ring position with the preferred CH 2 (4-OCF 3 -Ph) group. Compounds 14a (R ¼ NHMe) and 14b (R ¼ NMe 2 ) were the most active against G. lamblia (EC 50 ¼ 3.4 and 2.7 mM, respectively), slightly more potent than the primary amide 13b (EC 50 ¼ 4.1 mM) and metronidazole (EC 50 ¼ 6.1 mM). Compounds 14c (R ¼ COOEt), 14d (R ¼ NHOH) and 14e (R ¼ NHNH 2 ) had similar or slightly reduced activity relative to 13b (R ¼ NH 2 ). Therefore different 2 0 substituents were tolerated for activity against G. lamblia. Although compound series 14 displayed improved activity compared to compound series 13 against E. histolytica (EC 50 ¼ 10e45 mM vs >50 mM) and C. difficile (MIC ¼ 16 to !64 mg/mL vs >64 mg/mL), the overall activity profile of both series remained considerably inferior to metronidazole ( Table 2, Supplementary Table 4) and compounds within series 8 and 12. These results demonstrate that G. lamblia is selectively sensitive to 4-nitromidazoles, suggesting differences in the nitro-reduction activation and/or uptake of 4nitroimidazoles compared to E. histolytica and C. difficile.

Desnitro and amine derivatives
5-Nitroimidazole antimicrobial agents are pro-drugs that are activated by reduction of the nitro group to reactive intermediates that cause cellular damage [31]. The reduction step is catalysed by organism specific oxidoreductase enzymes, confounding target based drug design and enzymatic assays as approaches to drug development. In G. lamblia, the enzymes pyruvate ferredoxin oxidoreductase, nitroreductase 1 and thioredoxin reductase 1 have been implicated in the reductive activation of metronidazole [32]. Since the nitro group is key to the mode of action of metronidazole, we sought to establish whether this functional group is also important for the activity of these nitroimidazole carboxamides, which are thought to act by similar mechanisms as metronidazole. Thus, we prepared desnitro analogues 17 and 18 and the reduced amine derivative 20 (Scheme 4). As hypothesised, all three compounds displayed no discernable activity against parasites or C. difficile, supporting the importance of the nitro group in the mode of action of nitroimidazole carboxamides (Supplementary Table 5).

Plasma protein binding and microsome stability
Metronidazole is essentially 100% orally absorbed [33], yet exposure of G. lamblia parasites to the drug in the intestinal tract after the initial absorption period continues to occur by biliary excretion and enterohepatic circulation [34]. Oral absorption is also necessary for treatment of invasive amebiasis, underlying the importance of adequate absorption of nitro drugs for in vivo efficacy. To delineate preliminary ADME characteristics of the new nitroimidazole carboxamide compounds, we determined their plasma protein binding and microsome stability, as these properties are likely to influence compound half-life and free drug available at the sites of infection.
Binding to human plasma proteins was measured for several 4(5)-and 5-nitroimidazole carboxamide matched pairs, including 8a and 12a (R ¼ NHCH 2 (4-F-Ph)), 8k and 12k ((R ¼ NH-cyclohexyl)) and 8h and 12h (R ¼ morpholine) (Table 3). Plasma protein binding varied depending on the 1 0 -and 2 0 -substituents. The 4-F-benzylamide (8a and 12a) and cyclohexylamide (8k and 12k) imidazoles were highly bound to plasma proteins (!94%) regardless of the 1 0substituent (H or Me), with the plasma protein binding for 1 0 -H derivatives slightly greater in each instance. In contrast, the morpholine group of 8h ameliorated plasma protein binding (9% bound) for the 5-nitroimidazole but the 4(5)-matched pair 12h displayed high plasma protein binding, while metronidazole was almost completely unbound (<5% bound). The contrast in plasma protein binding between 8h and 12h could be explained by the acidic nature of the imidazole NeH bond observed in the proton NMR (NH d~14.3 ppm) and tendency for plasma proteins such as human serum albumin to bind anionic compounds [27]. While both metronidazole and tinidazole are mostly unbound to plasma proteins [35], tizoxanide, the active metabolite of the prodrug nitazoxanide, is highly plasma protein bound [36]. The influence of plasma protein binding on the free drug concentration at the site of infection is also related to other complex factors, including metabolism, distribution and half-life, and further in vivo efficacy experiments are necessary to determine the impact of high plasma protein binding on in vivo efficacy in this series [37].
The metabolic stability of a compound influences the concentration of drug available in the circulation for treatment of invasive amebiasis, and for prolonged exposure of G. lamblia to drug treatment via enterohepatic recirculation pathways. Therefore, we measured the human liver microsome stability for the 4(5)-and 5nitroimidazoles matched pairs 8a and 12a (R ¼ NHCH 2 (4-F-Ph)) and 8h and 12h (R ¼ morpholine) as these compounds showed good potency and a range of plasma protein binding. All of the compounds were metabolically stable after 2 h incubation with human liver microsomes (Table 3). This result was comparable to metronidazole, suggesting that the compounds have favourable metabolic stability and that different 2 0 substituents were tolerated.

Conclusion
New nitroimidazole carboxamides were identified with activity against the pathogenic parasites G. lamblia, including a metronidazole-resistant strain, and E. histolytica. The most potent derivatives displayed a wide range of plasma protein binding and were metabolically stable, with comparable stability to metronidazole. The rediscovery and derivatisation approach taken in this study could be applied to other 'forgotten' compounds to facilitate rapid research and development of new antiparasitic agents.