A novel in vitro image-based assay identifies new drug leads for giardiasis.

Giardia duodenalis is an intestinal parasite that causes giardiasis, a widespread human gastrointestinal disease. Treatment of giardiasis relies on a small arsenal of compounds that can suffer from limitations including side-effects, variable treatment efficacy and parasite drug resistance. Thus new anti-Giardia drug leads are required. The search for new compounds with anti-Giardia activity currently depends on assays that can be labour-intensive, expensive and restricted to measuring activity at a single time-point. Here we describe a new in vitro assay to assess anti-Giardia activity. This image-based assay utilizes the Perkin-Elmer Operetta® and permits automated assessment of parasite growth at multiple time points without cell-staining. Using this new approach, we assessed the "Malaria Box" compound set for anti-Giardia activity. Three compounds with sub-μM activity (IC50 0.6-0.9 μM) were identified as potential starting points for giardiasis drug discovery.

Introduction

Giardiasis causes significant worldwide morbidity with an estimated 184 million symptomatic cases annually (Pires et al., 2015) and an associated 171,100 disability-adjusted life years (DALYs) (Kirk et al., 2015). While giardiasis is more prevalent in the developing world it is also a burden in developed countries, with hospital based treatments in the United States of America costing $34.4 million (USD) annually (Collier et al., 2012). Giardiasis is commonly associated with clinical symptoms including nausea, vomiting and acute diarrhoea (Nash et al., 1987, Farthing, 1996). However it can manifest as a chronic disease and cause malabsorption, weight loss and failure to thrive in children (Al-Mekhlafi et al., 2005, Al-Mekhlafi et al., 2013, Bartelt et al., 2013). There is also mounting evidence that Giardia infection may be linked to irritable bowel syndrome, food allergies and obesity (Di Prisco et al., 1998, Hanevik et al., 2009, Guerrant et al., 2013).

As there is no currently available vaccine for humans, the control of giardiasis is dependent on chemotherapy. Current chemotherapeutic options are limited to a small number of compounds which are associated with treatment failures and clinical resistance (reviewed in Ansell et al., 2015). The 5-nitroimidazole class of compounds, typically metronidazole, are the most commonly used treatment agents (Watkins and Eckmann, 2014). However, these compounds have reported clinical failure rates of up to 40% (Oren et al., 1991, Farthing, 1996; reviewed in Watkins and Eckmann, 2014, Nabarro et al., 2015) and can also cause significant side-effects including neurological disorders and sudden death (Escobedo and Cimerman, 2007). Alternative agents including the benzimidazoles, such as albendazole, can also be used. However, the efficacy of these drugs varies widely (e.g. Hall and Nahar, 1993, Escobedo et al., 2003). In addition, the benzimidazole drugs appear particularly susceptible to the development of drug resistance, with data suggesting that parasite resistance can be easily selected in vitro (Gardner and Hill, 2001). New anti-Giardia agents with improved efficacy and toxicity are needed to improve this position.

A number of low to high throughput in vitro assays have been developed to identify new compounds active against Giardia. However, most rely on metabolic indicators or manual cell counting. Activity assays that rely on manual cell counting via microscopy have the advantage of permitting the assessment of growth at multiple time-points and provide useful morphological information, but are time consuming and may be subjective. While the more automated assays that make use of growth indicators including 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide (MTT), 2,3-Bis-(2-Methoxy-4-Nitro-5-Sulfophenyl)-2H-Tetrazolium-5-Carboxanilide (XTT), resazurin (AlamarBlue®) 3H-thymidine, ATP content or the assessment of glucuronidase activity in transgenic parasites (Müller et al., 2009) are more rapid, they inherently increase assay cost, provide limited activity/morphology information and permit only single time-point of assessment. Activity assays reliant on transgene expression are also limited to assessing activity against genetically manipulated parasites.

Efforts to improve current growth assay methods have included combining microscopy with automated image analysis software to decrease time limitations associated with manual enumeration methods (Bonilla-Santiago et al., 2008, Faghiri et al., 2011, Gut et al., 2011). For example, in an approach reported by Gut et al. (2011) parasites are stained with 4′,6-diamidino-2-phenylindole (DAPI) to automatically distinguish and enumerate living trophozoites without bias. While this significantly reduces assay evaluation time, parasites must still be fixed and stained which necessitates extra handling and eliminates the possibility of multiple time-point evaluations.

In this study, we developed an automated live-cell digital phase-contrast microscopy assay to assess the activity of compounds against Giardia trophozoites in vitro. The Perkin-Elmer Operetta®, with its associated Harmony® and PhenoLOGIC™ software, was used to exploit the power of automated digital phase-contrast microscopy and image analysis as a mechanism to identify and enumerate parasites based on their morphology without the need for a cell marker. A particular advantage of this approach is the ability to assess parasite growth at multiple time-points. This assay was used to assess the anti-Giardia activity of compounds from the “Malaria Box”. The “Malaria Box”, a set of compounds with known activity against mammalian cells (Kaiser et al., 2015) multiple parasite species including P. falciparum (Spangenberg et al., 2013), Toxoplasma gondii, Entamoeba histolytica (Boyom et al., 2014), Cryptosporidium parvum (Bessoff et al., 2014), Leishmania major (Khraiwesh et al., 2016) and Trypanosoma spp. (Kaiser et al., 2015) has never previously been assessed for anti-Giardia activity.

Materials and methods

Parasites and culture

G. duodenalis (strain BRIS/91/HEPU/1279; metronidazole sensitive; assemblage B (Upcroft et al., 1995, Nolan et al., 2011)) was grown axenically (3% O2 5% CO2, in N2 at 37 °C) in Kiesters-modified TYI-S-33 media in 8 mL borosilicate vials (Pyrex glass, No. 9825; VWR) as previously described (Keister, 1983, Meloni and Thompson, 1987). Media was prepared on a weekly basis and stored at 4 °C. When required for use, aliquots were supplemented with 10% foetal bovine serum, 100 units/mL penicillin and 100 μg/mL streptomycin.

Compounds

Albendazole, metronidazole and furazolidone were obtained from Sigma-Aldrich, USA and prepared in 100% DMSO to stock concentrations of 10–50 mM. Stocks were stored at −20 °C until required. Malaria Box compounds were obtained from the Medicines for Malaria Venture (MMV; www.mmv.org) as 10 mM stocks prepared in 100% DMSO.

Establishing assay conditions

Comparing automated parasite enumeration with manual counting

Giardia parasites were grown in stock 8 mL borosilicate tubes (Section 2.1) to ≤ 80% confluence. Parasites were detached from culture vials by incubating on ice for 30 min. After detachment, parasites were collected, counted using a haemocytometer and seeded in 96-well micro titre plates (Corning Costar 3596; total volume 200 μL; 2 × 104 to 5 × 103 cells/well). Outside wells of plates contained phosphate-buffered saline to reduce evaporation (PBS; 200 μL). Plates were incubated at 37 °C in sealed, activated Anaerocult® C mini bags as per manufacturer instructions as previously described (Upcroft and Upcroft, 2001). Growth of parasites seeded in triplicate wells on two separate occasions was assessed at 24 and 48 h by digital phase-contrast microscopy, enumerated using Harmony® and PhenoLOGIC™ software (Section 2.6) and by the manual counting of bright-field images. Data from all experiments were combined (mean parasite count/1.7 mm2 ± SD) and manual versus automated counts were compared using a student's t-test (Graphpad Prism 7®).

Assessing parasite growth in assay conditions

Giardia parasites were grown and prepared as described above. After detachment, parasites were collected, counted using a haemocytometer and seeded in 96-well micro titre plates (Corning Costar 3596; total volume 200 μL; 6 × 104 to 5 × 103 cells/well). However, as a reliable source of Anaerocult® C mini bags (Merck, Millipore) could not be obtained, microaerophilic conditions were established by incubating plates at 37 °C in air-tight chambers filled with 3% O2 5% CO2 in N2 as previously described (Gut et al., 2011). Growth of parasites was assessed at 24 and 48 h by digital phase-contrast microscopy and enumerated using Harmony® and PhenoLOGIC™ software (Section 2.6). Data are presented as mean trophozoite count ± SD of 4 separate experiments, each carried out in triplicate wells. The average doubling time between 24 and 48 h for each seeding concentration was calculated using the equation, td = (24) × log (2)/log (c2/c1) where td = doubling time, c1 was the average 24 h count and c2 was the average 48 h count.

Assessing the impact of imaging on parasite growth

As assay plates were outside of anaerobic conditions during imaging (Section 2.6; ∼20 min) and then returned to culture post-imaging for further incubation and assessment, the impact of imaging on parasite growth was assessed. In these assays two identical 96-well micro titre plates were prepared. One plate was imaged at 24 and 48 h and the other only at 48 h. In brief trophozoites were seeded into 96-well plates (3 × 104 to 3.75 × 103 parasites/well in 200 μL) and incubated in 3% O2 5% CO2, in N2 at 37 °C until imaging (Section 2.6). The 24 and 48 h imaged plate was returned to culture conditions after imaging at 24 h and re-imaged again at 48 h whereas the 48 h only plate remained in microaerophilic conditions until imaging at 48 h. Each cell seeding concentration was plated in six technical replicates on a single plate and each assay was repeated on three separate occasions. Data are presented as mean parasite count/1.7 mm2 ± SD and cell counts were compared using a student's t-test (Graphpad Prism 7®).

Assessing the activity of control anti-Giardia compounds

The activity of albendazole, metronidazole and furazolidone against Giardia trophozoites was assessed. Each compound was serially diluted in triplicate wells (100 μL; 8 point dilution series for albendazole and furazolidone and 15 point dilution series for metronidazole), and all wells except media only controls, were seeded with 1.5 × 104 Giardia trophozoites (100 μL; 200 μL final volume). Plates were incubated in 3% O2 5% CO2, in N2 at 37 °C until imaging and growth analysis (Section 2.6) at 24 and 48 h. Each assay included no drug with vehicle (0.2% DMSO) and no vehicle controls and in each case three independent assays were carried out. The concentration of DMSO in drug dilutions was kept constant at 0.2% and as previously shown (Johns et al., 1995) had no impact on parasite growth. Mean percentage growth inhibition compared to vehicle (0.2% DMSO) and background controls was determined for each assay. IC50 values were calculated using log-linear interpolation (Huber and Koella, 1993).

Assessing the Malaria Box compounds for activity against G. duodenalis

All Malaria Box compounds were screened for activity against G. duodenalis BRIS/91/HEPU/1279 at a final concentration of 10 μM in singlicate, in two independent experiments. Each plate included, background media, vehicle (0.2% DMSO), no vehicle and albendazole (10 μM) controls. Assays were performed under the same conditions as those used to assess the activity of control anti-Giardia compounds (Section 2.4; 1.5 × 104 parasites/well in final volume 200 μL; imaged at 24 and 48 h). Compounds demonstrating greater than 50% inhibition at this concentration were assessed for activity at 5 μM in duplicate (n = 2). Z-factors were calculated for each plate of each screening assay as previously described (Zhang et al., 1999). Compounds showing ≥50% inhibition at 5 μM were further investigated to determine IC50 values as described for control compounds (Section 2.4; each titration was performed in duplicate on three occasions; 8 point dilution series; compound concentration range 10,000–78 nM).

Digital phase-contrast microscopy, image acquisition and analysis

Individual wells on assay plates were imaged using the PerkinElmer Operetta®. Plates were removed from incubation and each well was imaged using brightfield and phase-contrast microscopy (total area imaged 1.7 mm2; <20 min/plate) before being returned to incubation if required. Brightfield images were taken 1 μm from the base of each well, with exposure set to 100 ms. Digital phase-contrast images were taken with 40 ms exposure between −5 μm and 5 μm, with a speckle scale of 10 μm. Images were automatically analysed and parasites enumerated with Harmony® and Phenologic™ software manually trained to identify and count trophozoites. Manual training was performed within Harmony and Phenologic™ using the “Select Population” building block and the “Linear Classifier” method. After training, an algorithm to identify trophozoites based on properties including their size and morphology was generated and used to assess all subsequent images.

Results

Growth assessment experiments showed that enumeration of Giardia parasites by digital phase-contrast microscopy and automated image analysis is as least as effective as examining parasite growth by manual cell counting (Fig. 1). Trophozoites were reliably identified by trained software (Fig. 1 C and D) and there was no statistically significant difference in parasite numbers when assessed manually or automatically using the Harmony® and Phenologic™ software (Fig. 1 E and F; p > 0.05 in all cases).

Fig. 1

Automatic enumeration of Giardia tropozoites by digital phase-contrast microscopy paired with Harmony® and Phenologic™ automated counting. Giardia trophozoites seeded in 96-well micro-titre plates were imaged using brightfield (A) and digital phase-contrast microscopy (B). Images were automatically assessed by Harmony® and Phenologic™ to identify and count trophozoites (green) amongst other signals (red) (C; brightfield and D; digital phase-contrast images). The effectiveness of the automated counting strategy was determined by comparing automated counts (E and F; green bars) to manually determining parasite numbers (E and F; grey bars). Parasite cultures were initiated at seeding concentrations of 2 × 104 - 5 × 103 cells/well and cell numbers were determined at 24 (E) and 48 h (F). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Data describing trophozoite growth in assay culture conditions suggested that 1.5 × 104 cells/well was the highest initial cell density able to maintain adequate BRIS/91/HEPU/1279 growth for 48 h, the duration of planned screening assays (Fig. 2A). When seeded at higher or lower cell concentrations mean parasite doubling times increased (Fig. 2A insert). Imaging was also found to have no significant impact on parasite growth as growth of parasites on plates imaged at 24 and 48 h was not significantly different to those grown on plates imaged only at 48 h (Fig. 2B; p = 0.68).

Fig. 2

Parasite growth assessment. The growth of Giardia trophozoites in assay conditions was assessed by digital phase-contrast microscopy (A). Parasite cultures were established in 96-well micro-titre plates at different seeding concentrations and growth was assesed at 24 (dark grey) and 48 h (light grey). Data are presented as mean parasite count ± SD of three independent experiments, each carried out in triplicate wells. In separate assays, the impact of parasite imaging on growth was assessed by comparing growth in plates imaged at both 24 and 48 h (B; dark grey) to those imaged only at 48 h (B; light grey). Data are presented as mean parasite count ± SD of three independent experiments. There was no significant difference in the growth of parasites grown on plates imaged once at 48 h or at both time points (p = 0.68).

The in vitro anti-Giardia activity of albendazole, metronidazole and furazolidone was assessed using digital phase-contrast microscopy and automated enumeration (Table 1). None of the IC50 values determined using the assay described here fell further than a single standard deviation from the published range at either 24 h or 48 h (Table 1).

Table 1

In vitro anti-Giardia activity of control compounds.

Compound	IC50 24 h		IC50 48 h
	Operetta (Mean ± SE)	Published range	Operetta (Mean ± SE)	Published range
Metronidazole	74 ± 3 μM	2a–75b μM	3 ± 1 μM	1c - 9d μM
Albendazole	93 ± 15 nM	27a–9600b nM	89 ± 9 nM	38e - 377f nM
Furazolidone	0.40 ± 0.06 μM	0.43g–1.4h μM	0.20 ± 0.04 μM	0.4i–1.1c μM

Cruz et al., 2003.

Arguello-Garcia et al., 2004.

Hounkong et al., 2011.

Edlind et al., 1990.

Cedillo-Rivera et al., 2002.

Cedillo-Rivera and Munoz, 1992.

Boreham et al., 1984.

Townson et al., 1992.

Tejman-Yarden et al., 2011.

Assessing the anti-Giardia activity of Malaria Box compounds

Preliminary screens of the Malaria Box compound set identified 122 compounds with >50% inhibition at 24 or 48 h when assessed at 10 μM (Fig. 3A and B; Table S1). The Z factor of all assays plates in the 10 μM screen was >0.5 (average ± SD; 0.74 + 0.11). Further analysis of the 122 compounds identified 22 with >50% growth inhibition at 5 μM (Fig. 3C and D; Table S1). The Z factor of all assays plates in 5 μM assays was also >0.5 (average ± SD; 0.73 + 0.10). Further dose response analysis of the 22 compound with >50% inhibition at 5 μM identified three compounds (MMV007384, MMV019690 and MMV006203) with sub-μM IC50 values (Table 2, Table S1 & Fig. S1) at either time-point.

Fig. 3

Anti-Giardia activity of “Malaria Box” compounds. The antigiardial activity of compounds in the ‘Malaria Box’ was determined using a primary 10 μM screen for 24 (A) and 48 h (B) and a secondary 5 μM screen for 24 (C) and 48 h (D) of compounds active (>50% inhibition) at 10 μM. All data are presented as mean % inhibition with standard error of the mean provided for 5 μM data (C & D).

Table 2

In vitro activity of Malaria Box compounds with sub μM IC50 against Giardia BRIS/91/HEPU/1279 parasites.

Compound Structure	Compound Name	IC50 (μM; mean ± SE)		SIa
		24 h	48 h
Image 2	MMV006203	3.1 ± 1.1	0.7 ± 0.2	25.7
Image 3	MMV007384	0.8 ± 0.2	0.6 ± 0.2	8.7
Image 4	MMV019690	2.8 ± 0.2	0.9 ± 0.1	4.8

SI determined by comparing 48 h Giardia IC50 to existing MRC-5 fibroblast IC50 data (Kaiser et al., 2015).

Discussion

We have developed an in vitro medium throughput assay that permits Giardia drug susceptibility testing in real-time without any need to stain parasites. This assay is unique in that it harnesses the power of digital phase-contrast microscopy and dedicated analysis software to identify and count parasites thereby permitting speedy, multi-time point analysis of live parasite numbers. A comparison of automatically generated parasite counts with manual counts demonstrates that the system can quickly and reliably assess trophozoite numbers (Fig. 1). As this assay permits the activity of compounds to be assessed at multiple time-points without any impact on parasite growth (Fig. 2) it also provides an opportunity to optimize data generation and limit cost. Additional information regarding the time course of compound activity and morphological effects, which can be derived from acquired images, may aid in compound triage and mechanism of action studies. Further reductions in cost and additional data acquisition may also be possible given that the assay is likely to be amenable to miniaturization and longer assessment periods (up to 72 h (Upcroft and Upcroft, 2001, Kulakova et al., 2014)).

While a potential limitation of the current assay may be in its assessment of parasite number rather than a metabolic parameter linked to viability, compounds with static activity can be of use therapeutically (Pankey and Sabath, 2004). In addition, metabolic assays can be associated with the same liability in the case of dormancy or when the compounds assessed interfere with the metabolic process used to quantitate inhibition (Collier and Pritsos, 2003, Ulukaya et al., 2004). Indeed, as a result of continued growth and the enhanced metabolic activity of controls over time, both assay types are likely to identify compounds with static activity as inhibitors. More specialized methods designed to assess mode of action are therefore more adequately placed to examine the nature of compound activity, post-identification. An additional limitation of the current assay that should be considered is its assessment of parasite number based on adherence. While this is an inherent limitation of other assays including those that require the removal of culture media and well-washing prior to activity assessment, this would mean that the assay is likely to identify compounds that effect attachment in addition to compounds that effect replication. Although the consequences of this anti-attachment activity in the in vivo setting may be limited, more specialized assays would be required to discriminate between compounds that effect attachment versus those that inhibit replication. Nevertheless, the ability of the current image-based assay to effectively examine the activity of compounds against Giardia parasites was demonstrated by assessing the activity of control anti-Giardia compounds albendazole, metronidazole and furazolidone, with IC50 values generated by the automated imaging and enumeration system being within the range of previously published studies (Table 1). The suitability of the assay as a mechanism to identify compounds with activity against Giardia parasites was also demonstrated by assessing the Malaria Box compound set for potential anti-Giardia activity. The mean Z factor for all plates in these assays (0.74 in 10 μM and 0.73 in 5 μM assays) suggest that the assay is of excellent quality (Zhang et al., 1999) and a promising new tool for Giardia parasite drug discovery. Of interest a previously described image-based Giardia assay which is dependent on parasite staining and hence an end-point assay, reported a Z Factor of 0.54 (Gut et al., 2011). The identification of anti-Giardia compounds within the Malaria Box set that have structural similarities to known anti-Giardia compounds provides additional evidence that the current assay is suitable for compound activity assessment. MMV007384, the most potent of the anti-Giardia hits identified (Fig. S1, Table 2; 24 h IC50 0.8 μM and 48 h IC50 0.6 μM) is a benzimidazole. In addition MMV667492 (Table S1; 24 h IC50 3.7 μM and 48 h IC50 2.6 μM) is a napthoquinone similar to menadione that has been shown to have promising efficacy against G. duodenalis trophozoites and cysts in vitro (Paget et al., 2004).

Two Malaria Box compounds, in addition to MMV007384 were identified to have sub μM IC50 values against Giardia parasites in the current study. These compounds were MMV019690 (Table 2; 24 h IC50 2.8 μM and 48 h IC50 0.9 μM) and MMV006203 (Table 2; 24 h IC50 3.1 μM and 48 h IC50 0.7 μM). While the selectivity index for MMV019690 (4.8; Table 2), generated using IC50 data against MRC-5 fibroblasts (Kaiser et al., 2015) suggest this compound may be associated with toxicity, the selectivity index of MMV006203, (25.7; Table 2), was more favourable, falling within recently published lead criteria range (Katsuno et al., 2015). Importantly, the identification of cell debris in images acquired during the assessment of MMV006203 and MMV019690 (Fig. S1) suggest that these compounds are cidal and warrant further investigation.

The current study has validated digital-phase contrast microscopy and automated parasite enumeration as a method to investigate Giardia drug susceptibility and identified new chemical scaffolds with anti-Giardia activity that may warrant further investigation. Unlike previously published image-based assessment Giardia assays, the method described in this study negates the need for cell staining and permits multiple-time point activity assessment which can improve screening costs and only add value to current drug discovery efforts.