Colorimetric High-Throughput Screen for Detection of Heme Crystallization Inhibitors

Abstract

Malaria infects 500 million people annually, a number that is likely to rise as drug resistance to currently used antimalarials increases. During its intraerythrocytic stage, the causative parasite, Plasmodium falciparum, metabolizes hemoglobin and releases toxic heme, which is neutralized by a parasite-specific crystallization mechanism to form hemozoin. Evidence suggests that chloroquine, the most successful antimalarial agent in history, acts by disrupting the formation of hemozoin. Here we describe the development of a 384-well microtiter plate screen to detect small molecules that can also disrupt heme crystallization. This assay, which is based on a colorimetric assay developed by Ncokazi and Egan (K. K. Ncokazi and T. J. Egan, Anal. Biochem. 338:306-319, 2005), requires no parasites or parasite-derived reagents and no radioactive materials and is suitable for a high-throughput screening platform. The assay's reproducibility and large dynamic range are reflected by a Z factor of 0.74. A pilot screen of 16,000 small molecules belonging to diverse structural classes was conducted. The results of the target-based assay were compared with a whole-parasite viability assay of the same small molecules to identify small molecules active in both assays.

Malaria poses an enormous public health burden, causing over 1 million fatalities annually worldwide, with the majority of morbidity and mortality attributed to Plasmodium falciparum malaria. Chloroquine (CQ) served as the main chemotherapeutic for several decades, but the emergence and spread of drug resistance has limited its current usefulness. After the introduction of every new antimalarial drug, with the exception of artemisinin, resistant malaria parasites have emerged (25). Hence, the continued development of new antimalarial drugs is necessary to continue to successfully treat malaria infection.

The malaria parasite cycles between two hosts, namely, mosquitoes (Anopheles sp.) and humans. In the human host, after a brief liver stage, P. falciparum resides exclusively inside red blood cells, where it feeds on hemoglobin, reproduces, and then releases progeny, which invade new red blood cells. During this intraerythrocytic stage, proteases digest hemoglobin within the food vacuole, a lysosome-like organelle (9). As hemoglobin is digested, heme molecules, containing redox-active iron centers, are released. The parasite overcomes the oxidative stress thus produced through the crystallization of free heme molecules into hemozoin (20). Hemozoin crystals are formed in an enzyme-independent reaction that is essential for parasite survival and therefore an excellent target for antimalarial chemotherapy. CQ, the most successful antimalarial agent to date, accumulates in the food vacuole, where it inhibits heme crystallization and prevents parasite proliferation (18). Small-molecule disruption of hemozoin formation has been proposed as a mechanism of action of many other antimalarial agents, including mefloquine (MQ), amodiaquine (AMQ), quinine, and quinidine, since each of these drugs successfully inhibits heme crystallization in in vitro assays, and they are all structurally related to CQ.

We believe that the development of antimalarial agents based on the physicochemical process of heme crystallization could identify molecules that are less likely to generate resistance, like CQ. Drug resistance usually entails expression changes or mutations in the target protein or similar changes in pumps in order to expel the antimalarial agent (22, 26). Since heme crystallization inhibitors do not target a protein but a physicochemical process, resistance can occur only through the latter approach. For example, CQ resistance is achieved through multiple mutations in the P. falciparum CQ resistance transporter 1 (pfcrt1) allele, which encodes a transmembrane protein that, when mutated, significantly reduces the concentration of CQ in the food vacuole (8). Since the heme crystallization pathway remains unaltered in resistant parasites, it is still possible to exploit parasite heme crystallization while avoiding cross-resistance with CQ.

In this study, we have adapted the pyridine hemichrome inhibition of β-hematin (Phiβ) assay described by Ncokazi and Egan for high-throughput screening (HTS) (17). This assay recapitulates in vivo heme crystallization by solubilizing hematin and then allowing it to spontaneously form crystalline β-hematin (synthetically identical to hemozoin) within a 384-well plate (1). Pyridine is used as a developing reagent to monitor heme crystallization, as pyridine molecules coordinated with the iron centers of free heme molecules produce a concentration-dependent color change, with a strong absorption at 405 nm. The optimized cell-free heme crystallization screen (CFHCS) was used to verify the mode of action of known antimalarials and to identify new chemical entities that inhibit hemozoin crystal formation.

MATERIALS AND METHODS

CFHCS development.

A Multidrop Combi liquid dispenser was used to dispense 30 μl of hemin solution (0.3 mM hemin, 0.1 M sodium hydroxide) into each well of a 384-well clear microtiter plate (Nunc 265196). Candidate compounds were pin transferred (CyBio-Well Vario) to the assay plate to a final compound concentration of approximately 220 μM. Each plate was screened in duplicate, with five library plates typically screened in parallel. AMQ hydrochloride (final concentration of 241 μM; United States Pharmacopeial Convention, Rockville, MD) and dimethyl sulfoxide (DMSO) (Sigma) were used as positive and negative controls, respectively. Following compound addition, 10 μl of 9.7 M sodium acetate solution (pH 4.8) was dispensed into the assay plates. Sealed plates were spun for 10 s at 1,000 rpm and then incubated at 60°C for 120 min. The plates were allowed to cool to room temperature (∼25°C) for 15 min before the addition of 75 μl of 14% (vol/vol) pyridine, 20 mM HEPES to each well. The plates were spun at 2,000 rpm for 1 min to ensure pelleting of heme microcrystals. Following centrifugation, 40 μl of supernatant was transferred to a new 384-well microtiter plate (Nunc 265196), using a RapidPlate 96/384 liquid handler (Caliper Life Sciences, Hopkinton, MA). The microtiter plate was read using a Spectramax Plus 384 plate reader (emission wavelength, 405 nm).

CFHCS dose-response assay.

Stock solutions of AMQ hydrochloride, MQ hydrochloride (Sigma), and atovaquone (United States Pharmacopeial Convention) were prepared in DMSO, while chloroquine diphosphate (Sigma) was dissolved in deionized H₂O. Drug efficacy was evaluated over a concentration range of 0.25 μM to 500 μM. Dose-effect profiles were generated as described above, except that various concentrations of candidate compounds were added manually to the microtiter plate. For each experiment, DMSO controls were added to 96 wells in four rows spaced evenly throughout the plate. The half-maximal inhibitory concentration (IC₅₀) values were measured from two independent experiments performed in quadruplicate. Data were analyzed using GraphPad Prism, version 4.00, for Windows (GraphPad Software, San Diego, CA).

Quantitative assay evaluations and optimization.

To determine the Z factor for the optimized assay, a screening library plate containing 192 wells of DMSO and 192 wells of AMQ hydrochloride (241 μM) was screened in the CFHCS in duplicate in three separate experiments. The following equation was used to determine the Z factor: Z factor = 1 − (3 SD₊ + 3 SD₋)/|Ave₊ − Ave₋|, where Ave₊ is the mean absorbance at 405 nm for wells treated with 241 μM AMQ hydrochloride (positive control), Ave₋ is the mean value for the wells containing DMSO (negative control), SD₊ is the standard deviation for AMQ hydrochloride, and SD₋ is the DMSO standard deviation (28).

Compound libraries screened.

A subset of the screening libraries at the Broad Chemical Biology Platform at the Broad Institute of Harvard and Massachusetts Institute of Technology (MIT), which had previously been screened in a P. falciparum growth inhibition assay, was selected for HTS (2). Compounds in this library were commercially purchased or synthesized in house and were arrayed in 384-well microtiter plates at a concentration of 10 mg/ml in DMSO. The commercially available collections were obtained from ActiMol TimTec, the Biomol ICCB Known Bioactives Collection, ChemDiv, I. F. Lab, Maybridge, Peakdale, MicroSource (National Institute of Neurological Disorders and Stroke Library and Spectrum Collection), and mixed commercial sources. Detailed descriptions of each library can be found in the Harvard Institute for Chemistry and Chemical Biology Screening Compound Database (http://iccb.med.harvard.edu/screening/compound_libraries/index.htm).

P. falciparum growth inhibition assays.

The DAPI (4′,6-diamidino-2-phenylindole) P. falciparum growth assay was carried out as previously described, using the following two strains of P. falciparum in a 384-well microtiter plate format (2): 3D7, a CQ-sensitive strain, and Dd2, a CQ-resistant strain. Compounds were dissolved in DMSO and serially diluted in complete medium to achieve a final concentration of 0.05 μM to 100 μM using a Bravo liquid handling platform (Velocity11, Menlo Park, CA). IC₅₀ values were calculated by nonlinear regression analysis.

Compounds interfering in the fluorescent readout of the DAPI P. falciparum growth assay were assayed for parasite growth inhibition by using the [³H]hypoxanthine incorporation assay as previously described (6). Compounds (CQ as a control; IDI 48) were serially diluted 1:1 in hypoxanthine-free complete medium to achieve a final drug concentration range of 0.05 μM to 100 μM.

RESULTS

High-throughput assay development.

The Phiβ assay was successfully adapted into a robust HTS for the discovery of new inhibitors of heme crystallization. The assay uses robotic liquid handling and a 384-well microtiter plate format (Fig. 1). The concentration of sodium acetate solution was lowered to 9.7 M from 12.9 M, as the original, higher concentration was not compatible with robotic liquid handling at room temperature. The lower acetate concentration necessitated a longer incubation time, of 2 hours, to achieve maximum heme crystallization. The concentration and volume of the hemin solution and pyridine solution were systematically modified to improve the dynamic range and robustness of the assay by testing multiple concentrations of each reagent in the assay while holding all other variables constant. The hemin solution volume chosen was optimal for the reproducible addition of drug via pin transfer, while the final pyridine solution volume helped to break up heme aggregates through the mechanical force produced by solution addition, thereby decreasing variability. The fully optimized CFHCS afforded a Z factor of 0.74, using a positive control absorbance of 2.14 ± 0.15 and a negative control absorbance of 0.151 ± 0.02, averaged from three independent experiments. This is a favorable Z factor, as values of >0.5 are considered favorable for HTS due to the high degree of day-to-day reproducibility and large dynamic range that the score reflects (28).

FIG. 1.

CFHCS design. Compounds dissolved in DMSO are pin transferred to a 384-well microtiter plate containing a weakly basic hemin solution. Crystallization is catalyzed by the addition of 9.7 M acetate solution and incubation at 60°C for 2 h. A 14% pyridine solution is added to each well to identify wells in which heme crystallization is inhibited. Pyridine coordinates with the free iron centers of uncrystallized heme molecules, causing an increased absorbance at 405 nm. Supernatant from the assay plate is transferred to a new 384-well microtiter plate to eliminate interference from crystalline β-hematin solids, and absorbance is measured at 405 nm.

Dose-response assay.

Inhibition of heme crystallization by pyrimethamine, AMQ, CQ, and MQ was evaluated using the CFHCS (Table 1). Pyrimethamine, an inhibitor of dihydrofolate reductase, did not affect heme crystallization. The known heme crystallization inhibitors AMQ, CQ, and MQ were active in the CFHCS, and the order of activity for these compounds was maintained between the CFHCS, Phiβ assay, and P. falciparum growth inhibition assay, as previously reported (17).

TABLE 1.

Efficacies of known heme crystallization inhibitors

Compound	IC50 for heme crystallization			IC50 for P. falciparum 3D7 proliferation (nM)
	CFHCS		Phiβ assay (equivalents)a,b
	μM	Equivalentsb
AMQ	67.0 ± 0.67	0.3 ± 0.001	1.45 ± 0.08	6.8 ± 2.3
CQ	374.1 ± 0.29	1.7 ± 0.003	1.91 ± 0.3	12.6 ± 2.3
MQ	476.1 ± 0.88	2.16 ± 0.004	2.90 ± 0.1	12.3 ± 2.0

^aData are from reference 17.

^bIC₅₀ values in the Phiβ assay were reported in molar equivalents of drug to heme, a convention that allows comparisons between heme crystallization assays.

HTS using the CFHCS.

The CFHCS was used to screen a diverse chemical library of 16,000 compounds. This chemical library included a variety of purchased commercial compounds, partially purified natural products, and a collection of known bioactive compounds. Compounds which were screening positives in the CFHCS inhibited crystallization by 50% at a drug concentration of 220 μM and which had a reproducibility between replicates of >0.95. Reproducibility was calculated by the Broad Chemical Biology automated data analysis pipeline, Chembank, as described by Seiler et al. (21). We identified 644 screening positives (3.96% hit rate) meeting these criteria. Several known antimalarials, including the known heme crystallization inhibitors AMQ and quinacrine, were among the compounds that were screening positives in the CFHCS. Additionally, several porphyrins were identified as heme crystallization inhibitors, an activity that has previously been reported for this class of compounds (3, 12, 15).

In silico counterscreen for P. falciparum growth inhibition.

The screening positives from the CFHCS had previously been evaluated for activity in a high-throughput P. falciparum growth inhibition screen (2). We compared the two data sets to identify heme crystallization inhibitors which also inhibited 80% of P. falciparum growth at 30 μM. From this set, 17 compounds were identified which had an IC₅₀ in the heme crystallization screen of <375 μM (the IC₅₀ of CQ in the assay) after retesting and an IC₅₀ of <20 μM against a CQ-resistant strain of P. falciparum (Dd2). The IC₅₀s for hit compounds in the heme crystallization assay ranged from 22 μM to 338 μM, and the IC₅₀s in the P. falciparum growth inhibition assay ranged from 0.2 μM to 19 μM (Table 2). Compound binding to pyridine could potentially cause a color change at 405 nm, resulting in a false-positive result. Under the assay conditions, none of the 17 hit compounds were found to interact with pyridine in this way. In addition to their activity, hits were defined as compounds that had no previously described antimalarial activity. All hits were tested against a CQ-sensitive (3D7) and a CQ-resistant (Dd2) strain of P. falciparum and showed no cross-resistance to drugs that currently target this pathway. Based on structural similarity searches, these classes of compounds also did not have previously reported activity in P. falciparum growth assays and heme crystallization assays. Two groups of structurally related compounds were identified from the 17 hits, namely, pyrimidines and 1,3-benzoxathiol-2-ones (Fig. 2).

TABLE 2.

In vitro and in vivo activities of all CFHCS hits

IDI no.	CFHCS IC50 (μM)	VARa	P. falciparum IC50 (μM)
			3D7	Dd2
1	338	1.3/1	3	6
7	236	1.4/1	3	6
11	182	NA	9	8
13	65	NA	4	6
17	158	NA	0	0
19	279	NA	14	16
30	22	NA	6	8
31	41	NA	12	9
40	30	NA	5	6
43	41	NA	1	2
48	36	NA	10	11
50	30	NA	15	10
52	219	315/40	8	5
57	102	NA	13	10
66	238	NA	0	0
70	111	NA	6	19
71	215	NA	13	16

^aCalculated for a parasite cytosol pH value of 7.3 and food vacuole pH values of 4.8 (10) and 5.7 (5) (values before and after the slash, respectively). NA, no accumulation.

FIG. 2.

Antimalarial chemotypes identified in the CFHCS. (A) Pyrimidine class hit compounds. (B) 1,3-Benzoxathiol-2-one class hit compounds.

DISCUSSION

The CFHCS pilot screening campaign identified novel heme crystallization inhibitors with potent antimalarial activity. To our knowledge, the CFHCS is the first heme crystallization assay to utilize a 384-well microtiter plate format, and thereby the first screen compatible with most HTS facilities used today. The assay uses commercially available, inexpensive reagents and does not require any parasite extract, recombinant protein, or radioactivity, making it executable in a wide range of laboratory environments. A pilot screen against ∼16,000 compounds has shown the assay to be robust, reproducible, technically simple, and suitable for automation. A number of heme crystallization inhibitors (IC₅₀ of <220 μM) were identified, yielding a final hit rate of about 4%, and 3% of those compounds screening positive possessed antimalarial activity, with IC₅₀s of <20 μM. There was a hit rate of 0.1% for compounds that both inhibited heme crystallization and inhibited P. falciparum growth.

The drug concentration screened in the CFHCS is considerably higher than the screening concentration used in P. falciparum growth assays and in many protein-based high-throughput screens. Micromolar compound concentrations have been used in all heme crystallization screens described to date, even when the in vivo activity of the compound is in the nanomolar range, and therefore the CFHCS is within the range published for heme crystallization assays (4, 9, 11, 13, 19). Many known heme crystallization inhibitors are weak bases, which have been shown to reach high local concentrations through accumulation in the acidic food vacuole through a pH-trapping mechanism (27). ChemAxon Marvin software (5.0.4; ChemAxon, Budapest, Hungary) was used to estimate the pK_as of amine groups in each hit compound. Estimated pK_a values were used to calculate the vacuolar accumulation ratio (VAR) for each compound, with either food vacuole and parasite cytosol pH values of 4.8 and 7.3, respectively, from the Kirk group (10), or pH values of 5.7 and 7.3, respectively, from the Roepe group (5, 7), as previously described (24). Three of our hits are predicted to be protonated in the acidic environment of the food vacuole and therefore to accumulate at higher concentrations than the whole organism dose. The highest VAR was around 315-fold, as calculated using the Kirk group pH values (40-fold with the Roepe group values), which is orders of magnitude lower than the VAR of CQ, which is calculated to have a 10⁵- to 10³-fold accumulation. While CQ is a dibasic amine, all three of these compounds are monobasic amines, which results in lower VAR values. We plan to explore how the addition of a second basic amine to these structures affects their activity in the parasite.

This screen to identify novel antimalarial chemotypes targeting heme crystallization yielded compounds unlike those described in a notable previous effort, a radioisotope screen conducted at Roche, and unlike those based on modifying the quinoline scaffold (13, 14, 16, 23). Possibly as a result of being able to conduct an HTS on unbiased libraries, the compounds identified in this screen had comparable antimalarial efficacies in both the CQ-sensitive strain 3D7 and the CQ-resistant strain Dd2 of P. falciparum. We are currently testing hit compounds against multiple-drug-resistant strains of P. falciparum to further confirm that they do not show cross-resistance with currently deployed heme crystallization inhibitors.