Development of a Novel High-Density [³H]Hypoxanthine Scintillation Proximity Assay To Assess Plasmodium falciparum Growth

Abstract

The discovery and development of new antimalarial drugs are becoming imperative because of the spread of resistance to current clinical treatments. The lack of robustly validated antimalarial targets and the difficulties with the building in of whole-cell activity in screening hits are hampering target-based approaches. However, phenotypic screens of structurally diverse molecule libraries are offering new opportunities for the identification of novel antimalarials. Several methodologies can be used to determine the whole-cell in vitro potencies of antimalarial hits. The [³H]hypoxanthine incorporation assay is considered the “gold standard” assay for measurement of the activity of antimalarial compounds against intraerythrocytic forms of Plasmodium falciparum. However, the method has important limitations, as the assay is not amenable for high-throughput screening since it remains associated with the 96-well plate format. We have overcome this drawback by adapting the [³H]hypoxanthine incorporation method to a 384-well high-density format by coupling a homogeneous scintillation proximity assay (SPA) and thus eliminating the limiting filtration step. This SPA has been validated using a diverse set of 1,000 molecules, including both a representative set from the Tres Cantos Antimalarial Set (TCAMS) of compounds and molecules inactive against whole cells. The results were compared with those from the P. falciparum lactate dehydrogenase whole-cell assay, another method that is well established as a surrogate for parasite growth and is amenable for high-throughput screening. The results obtained demonstrate that the SPA-based [³H]hypoxanthine incorporation assay is a suitable design that is adaptable to high-throughput antimalarial drug screening and that maintains the features, robustness, and reliability of the standard filtration hypoxanthine incorporation method.

INTRODUCTION

According to the World Health Organization (WHO), malaria is still one of the most severe diseases worldwide. It affects more than 225 million individuals, resulting in 584,000 estimated deaths annually (1). Malaria is caused by Plasmodium parasites transmitted by infected female Anopheles mosquitoes. Plasmodium falciparum is by far the deadliest of the five human malarial species (P. falciparum, P. malariae, P. ovale, P. vivax, and P. knowlesi). The effectiveness of current antimalarial therapy is under continuous threat, and even the newest antimalarial treatments are now showing evidence of clinical failure. Artemisinin-based combination therapies (ACTs) are recommended by WHO as the first-line treatment for uncomplicated P. falciparum malaria in countries where it is endemic (2). However, along the Cambodia-Thailand border, P. falciparum has become resistant to these treatments (3, 4). As a consequence, there is a significant risk that multidrug resistance can emerge quickly in other regions, thus making urgent the search for alternative (new) antimalarial drugs that could complement or replace those currently in use (5).

Traditionally, the way to measure parasite growth was microscopic examination of blood smears using a modified Wright-Giemsa stain (6). Then, several in vitro methods to determine the potential antiplasmodial activity of drugs against cultured intraerythrocytic asexual forms of P. falciparum were developed, including a [³H]hypoxanthine incorporation assay (7). Recently, new nonradioactive methods which use the intercalation of DNA with a dye, such as SYBR green or PicoGreen, have been used to accurately measure parasite growth inhibition (8, 9). Even though the rate of use of these assays has recently increased because they are cheap and relatively simple, setting up of robust conditions for high-throughput screening (HTS) is challenging because of the low signal/background (S/B) ratio inherent to these assays. Alternative methodologies using imaging as a readout have also been implemented, but they are dependent on complex analysis algorithms and highly expensive equipment (10, 11).

P. falciparum lactate dehydrogenase (PfLDH) activity is also a good surrogate for parasite growth and can be measured in a high-throughput format (12). However, this assay can result in high background levels with slowly acting antimalarials, as the production of significant amounts of PfLDH occurs before full inhibition of parasite growth is achieved.

Nevertheless, the [³H]hypoxanthine incorporation assay is so far the method most widely used and is considered the “gold standard” assay due to its robustness, reliability, and reproducibility. In addition to the fact that only the labeled precursor is incorporated by the parasites, the assay has unique advantages, as it allows the addition of radioactivity 24 h after the start of treatment with the drug. This preincubation step avoids the background due to parasite growth and radioactivity intake before the full inhibitory effects of the drug have taken place. However, it requires the use of radioactive materials, since it is based on the incorporation of [³H]hypoxanthine into parasite nucleic acids.

On the other hand, in order to support the identification of new antimalarial molecules, HTS campaigns are needed to test large compound libraries, which generally involve the testing of thousands or even millions of compounds. They provide a quick and cost-effective way to support the final objective of indentifying new antimalarial molecules by increasing the possibility of finding new chemical entities of interest (13). However, the standard [³H]hypoxanthine incorporation test still has the liability associated with a low-throughput, 96-well radioactive filtration step. This involves multiple labor-intensive steps (filtration, drying, melting of filters on scintillator sheets) and the generation and disposal of radioactive waste, hampering its use in HTS.

To overcome these issues, we have evolved the filtration assay into a high-throughput scintillation proximity assay (SPA) (Fig. 1). Scintillation beads are microspheres containing scintillant. When the radiolabeled molecule is attached or is in proximity to a bead, light is emitted. Upon activation, the scintillator emits red light at 615 nm that is captured by a supercooled charge-coupled-device (CCD) camera, as in an imaging microplate reader. SPA is a well-established technology which requires no separation step, is easy and precise, and allows the use of robotics to minimize manual intervention. Therefore, it is a method amenable to automation, which makes it particularly suited to high-throughput screening applications (14, 15). In the case described here, it permits the rapid and sensitive measurement of parasite growth in a homogeneous system.

FIG 1.

Scheme of the [³H]hypoxanthine-based scintillation proximity assay protocol. P. falciparum cultures (25 μl) with 0.5% initial parasitemia and 2% hematocrit were incubated in plates into which drug had been predispensed and incubated for 24 h under standard conditions at 37°C. After this time, 0.1 μCi/well of [³H]hypoxanthine at 4 μl/well was added and the plates were incubated for an additional 24 h. After a freeze-thaw process which lysed the cells, radiolabeled DNA was captured by the addition of 25 μl of 5 mg/ml poly-Lys YOx SPA beads, which emit luminescence in proportion to the level of parasite growth, as described in the Materials and Methods section. RT, room temperature.

This new SPA [³H]hypoxanthine incorporation assay offers multiple advantages. On the one hand, it offers the possibility to perform the traditional gold standard assay while increasing the throughput by using a 384-well format. On the other hand, it avoids the limiting filtration and washing steps, reducing the time of processing and the manipulation of the radioactive plates.

To validate the assay, we carried out parallel determinations of the 50% inhibitory concentrations (IC₅₀s) of a small set of standard antimalarial drugs with different modes of action against P. falciparum using the classical standard [³H]hypoxanthine filtration assay and the new proposed assay. In addition, several parasites with different genetic backgrounds were tested to ensure that the methodology was useful with a broad range of P. falciparum strains. As we describe below, the IC₅₀ values of these compounds for P. falciparum were nearly identical in both assays. Moreover, the genetic background of the parasites did not interfere with the potencies obtained by both methodologies. These results make the new SPA suitable for use in vitro with a broad diversity of P. falciparum strains.

To further assess the ability of this assay to identify compounds with antimalarial activity, we used a set of drugs with proven whole-cell activity. A structurally diverse set of 1,000 molecules from the Tres Cantos Antimalarial Set (TCAMS), previously identified using a PfLDH assay, was selected (16). The results obtained by both assays showed a good correlation.

In this article, we present the development and validation of an SPA-based [³H]hypoxanthine incorporation assay, and we compare the results generated by this new assay with those produced by the classical filtration assay or PfLDH methods.

MATERIALS AND METHODS

Materials.

Ytrium oxide (YOx) and polystyrene (PS) SPA beads and [³H]hypoxanthine were purchased from PerkinElmer (Waltham, MA, USA). AlbuMAX II, glutamine, and RPMI 1640 were from Thermo Fisher Scientific (Waltham, MA, USA). The rest of the reagents were purchased from Sigma-Aldrich (St. Louis, MO, USA). P. falciparum strains 3D7A, W2, Dd2, and TM90C2B were obtained from the Malaria Research and Reference Reagent Resource Center (MR4). Accurate descriptions of the genetic backgrounds of these strains can be obtained at http://www.beiresources.org. Human erythrocytes were from the Spanish Red Cross Blood Bank.

Culture of P. falciparum strains.

Parasites were cultured using a modification of the method previously described (17). Briefly, parasites were cultured using RPMI 1640, 25 mM HEPES, sodium bicarbonate, and glutamine supplemented with 0.5% AlbuMAX II, 2% d-sucrose, 0.3% glutamine, and 150 μM hypoxanthine. The complete medium was usually prepared fresh just before use and prewarmed to 37°C. Cultures were maintained under an atmosphere of 5% O₂, 5% CO₂, and 90% N₂.

Evaluation of P. falciparum growth inhibition using the standard [³H]hypoxanthine incorporation assay by filtration.

Asynchronous cultures (ca. 80% rings) of infected red blood cells (iRBCs) at 0.5% parasitemia and 2% hematocrit were incubated for 24 h in 96- or 384-well plates (100 μl and 25 μl, respectively) with drugs that had previously been dispensed from a dimethyl sulfoxide (DMSO) stock using an Echo liquid handler (Labcyte, Sunnyvale, CA, USA). The incubation culture medium contained 5 μM hypoxanthine. Then, 0.2 μCi (8 μl for the 96-well format) or 0.1 μCi (4 μl for the 384-well format) of a stock solution of [³H]hypoxanthine (0.025 μCi/μl) in RMPI 1640 was added to all the wells and the plates were incubated for an additional 24 h. Then, the plates were frozen. After that period, the plates were thawed and harvested on Filtermat A glass fiber filters (PerkinElmer) using a Cell Harvester96 cell harvester (Tomtec, CT, USA). For the 384-well plates, parasitized cultures were manually transferred to 96-well plates to undergo harvesting. The filters were dried and melted on MeltiLex scintillator sheets (PerkinElmer) to determine the incorporation of [³H]hypoxanthine. Radioactivity was measured using a MicroBeta plate counter (PerkinElmer).

Data were normalized to the level of incorporation of the positive control, consisting of iRBCs without drug (0% inhibition), and the negative control, consisting of iRBCs exposed to 25 μM artesunate (100% inhibition). All solutions were dispensed into plates using Multidrop Combi dispensers (Thermo Scientific, Waltham, MA, USA). Fifty percent inhibitory concentration (IC₅₀) values were determined using Excel software (Microsoft, Redmond, WA, USA) and fitted with Grafit (version 7) software (Erithacus Software, Horley, United Kingdom) with a 2-parameter equation.

Evaluation of P. falciparum growth inhibition by measuring [³H]hypoxanthine incorporation through the scintillation proximity assay.

For SPA bead testing, P. falciparum-parasitized red blood cells (100 μl iRBCs plus 8 μl [³H]hypoxanthine) were cultured in a 96-well microplate as described above. Ten microliters of either RBCs, iRBCs (0.5% initial parasitemia, 2% hematocrit), or iRBCs treated with 50 μM atovaquone and 50 μM artesunate was transferred after freezing-thawing to a 384-well microplate (low volume, white; catalog number 784075; Greiner Bio-One, Frickenhausen, Germany), and 10 μl of seven different types of SPA beads at 10 mg/ml in deionized water was added to each well. The beads were kept under stirring during dispensation to ensure homogeneity. The plates were incubated for at least 1 h at room temperature to allow the beads to settle and avoid any additional centrifugation step and were kept away from the light during this time. Then, the microplate was read in a Viewlux microplate reader (PerkinElmer) by recording the luminescence for 10 min using an emission filter at 613 nm.

To assess assay signal linearity, different P. falciparum inocula from 2-fold serial dilutions with a 1.25% initial parasitemia were used. P. falciparum-parasitized red blood cells (25 μl iRBCs plus 4 μl [³H]hypoxanthine) were cultured in a 384-well microplate (high volume, white; catalog number 781080; Greiner Bio-One). After freezing-thawing of the microplate and by avoidance of any transfer step, 25 μl of 5 mg/ml poly-Lys YOx beads in deionized water was added and the signal was developed as described above. These were the final assay conditions after assay optimization.

Drug sensitivity and validation set screening.

The IC₅₀s of four standard antimalarial drugs for P. falciparum were determined using different parasite genetic backgrounds: 3D7A (a drug-sensitive strain), W2 and Dd2 (chloroquine- and pyrimethamine-resistant strains, respectively), and TM90C2B (a chloroquine-, pyrimethamine-, and atovaquone-resistant strain). All solutions were dispensed into plates containing 3-fold serial dilutions of the drugs using Multidrop Combi dispensers (Thermo Scientific, Waltham, MA, USA). The Z′ factor was used as the main quality control parameter (18).

In order to assess the discriminating power of the SPA to separate active from inactive compounds, a set of nearly 1,000 compounds was selected and the results were compared with previous data obtained using the PfLDH assay. This set contained a comparable number of active and inactive compounds previously assayed using the PfLDH assay. In this way, 357 compounds that showed a minimum inhibition of 80% at 2 μM in a primary screening at a single concentration and a pIC₅₀ of >5.5 in dose-response confirmation experiments were selected from TCAMS (19). All compounds had also shown a pIC₅₀ of >5.5 in the [³H]hypoxanthine incorporation filtration assay and therefore corresponded to compounds confirmed to be active. pIC₅₀ is the negative logarithm of the IC₅₀. On the other hand, 649 compounds were randomly picked by following the distribution of inactive compounds (inhibition, <80% at 2 μM) in the PfLDH assay screening. The final set contained 1,006 compounds in total. According to the published data, the percent inhibition values for the compound set showed a bimodal distribution with modes of about 0 and 100% inhibition, corresponding to the groups of inactive and active compounds, respectively.

Normalization and statistical cutoffs were calculated with ActivityBase screening software (IDBS, Guilford, United Kingdom) using positive and negative controls as described above.

Evaluation of P. falciparum growth inhibition using PfLDH activity.

The PfLDH assay was performed as previously described (12). Briefly, 25 μl of P. falciparum cultures was grown in 384-well microplates in the presence of the compounds. To evaluate Plasmodium LDH enzymatic activity, 70 μl of a freshly made reaction mix containing 143 mM sodium l-lactate, 143 μM acetyl pyridine adenine dinucleotide (APAD), 178 μM nitroblue tetrazolium chloride (NBT), 2.83 U/ml diaphorase, and 0.7% Tween 20 in 100 mM Tris-HCl, pH 8.0, was used. The absorbance at 650 nm was monitored in a plate reader after 10 min of incubation at room temperature. Data were normalized to percent growth inhibition using iRBCs as the positive control (0% inhibition) and iRBCs with 50 μM chloroquine and 50 μM artemisinin as the negative control (100% inhibition).

RESULTS

SPA bead type selection.

Seven different types of imaging beads were tested. There were two different bead matrices suitable for the imaging reader: spherical plastic PS beads and crystalline YOx beads. The plastic beads are larger and stay in suspension longer than the crystalline beads. In both types of beads, the scintillator is europium, which is copolymerized in the polystyrene matrix in the PS beads and trapped within the YOx beads. Upon activation, this scintillator emits red light at 615 nm that is captured on a CCD camera-based detector, which provides an HTS capability by imaging all the wells in a 384-well microplate at one time. Each bead type can be derivatized for use in different applications. We have tested both matrices with different coatings. Poly-lysine (poly-Lys), designed to capture negatively charged cellular membranes in receptor binding assays, and wheat germ agglutinin (WGA), used to investigate receptor-ligand interactions, are both cationic coatings under physiological conditions. Membrane binding YOx beads allow the binding of negatively charged components of membranes, and in positively charged PS beads, polyethylenimine (PEI) blocks potential nonspecific binding sites on the bead surface. All these beads were tested under two different conditions: in 200 mM citrate at pH 3.0 and in distilled water at pH 7.0. Acidic pH has been used in other SPA-based assays to precipitate the nucleic acid and favors the unspecific interaction of the nucleic acid and the beads (19). In this case, the presence of a high concentration of hemoglobin from erythrocytes precluded the use of acidic pH because the hemozoin formed strongly quenches the SPA signal. At neutral pH, YOx beads displayed the highest net signal and signal-to-background ratios when the iRBC signal was compared with either the RBC signal (S/B ratio = 3.5 ± 0.8) or the iRBC signal in the presence of atovaquone and artesunate (S/B ratio = 4.5 ± 0.8) (Fig. 2), allowing the start of assay development. The background signal of RBCs in the absence of parasites was low since these cells do not synthesize either RNA or DNA and, in all cases, was similar to the signal of the iRBCs treated with atovaquone and artesunate. A possible explanation of this result is that at neutral pH the electrostatic interaction between the negative charge of the phosphate groups of the nucleic acids and the positive charge of the poly-Lys beads is favored. Only the [³H]hypoxanthine incorporated into the nucleic acid is in close proximity to the SPA bead and stimulates the scintillant to emit light. This effect would not occur with the free [³H]hypoxanthine that remains in solution. The decay particles from the radioactive molecule are not close enough to excite the SPA bead, and no light is emitted.

FIG 2.

Luminescence signals, obtained using different SPA beads, from nonparasitized red blood cells (RBCs; black columns), P. falciparum-infected red blood cells (iRBCs; dark gray columns), and P. falciparum-infected red blood cells treated with atovaquone and artesunate (iRBCs + At-Ar; light gray columns). Signal-to-background ratios were later improved during assay development. PolyLys, polylysine; WGA, wheat germ agglutinin; MB, membrane binding; PEI, polyethylenimine; YOx, yttrium oxide; PS, polystyrene. Values and error bars are averages and standard deviations, respectively, calculated from 4 technical replicates.

The same poly-Lys coating in the PS beads displayed a raw signal (iRBC) similar to that obtained with the YOx matrix but an S/B ratio of 1.5 ± 0.4 and a net signal 2.1-fold lower. This is because, in the absence of a centrifugation step, the plastic PS beads remain in suspension much longer that the heavier crystalline YOx beads, increasing the nonproximity effect (background due to the stimulation of the beads by unbound [³H]hypoxanthine that is close enough to the bead).

In further experiments, we tried different assay conditions to reduce reagent consumption and increase the S/B ratio and assay quality. The effects of the level of parasitemia, bead volume, concentration, settlement (time, gravity versus centrifugation), and plate format (density, well shape) were studied. After this optimization, a final S/B ratio of >10 was achieved by avoiding the transfer step, decreasing the final bead concentration to 5 mg/ml, and increasing the bead volume (see Fig. S1 in the supplemental material). Lastly, the best results were obtained when equal volumes of cultured parasites and SPA beads were added.

Poly-Lys YOx beads in water at neutral pH under the conditions described above provided a robust method to monitor [³H]hypoxanthine incorporation into the parasite nucleic acids. Interestingly, low backgrounds were observed in all cases, and despite the presence of the deeply colored erythrocytic hemoglobin in the assay, we found a good discrimination between positive and negative controls.

Assessment of assay signal linearity.

As for every assay developed, it is essential to demonstrate a linear correlation between the metabolite or the phenotypic endpoint to be measured and the signal output over the dynamic range of interest. Figure 3 illustrates the relationship between the levels of initial parasitemia and the luminescence signal determined using scintillation beads to measure parasite [³H]hypoxanthine incorporation. Labeled hypoxanthine incorporation was proportional to parasite growth over the range of 0.1 to 1.25% initial parasitemia at 2% hematocrit.

FIG 3.

Relationship between initial parasitemia and [³H]hypoxanthine incorporation by Plasmodium falciparum measured by the poly-Lys YOx bead-based SPA after 48 h of incubation. Plotted values were extracted from 3 independent biological replicates, each of which was performed in duplicate (r² = 0.997).

Drug susceptibility.

The IC₅₀ values of four standard antimalarial inhibitors for P. falciparum were determined using the [³H]hypoxanthine incorporation assay developed through two methodologies: filtration and SPA (Table 1 and Fig. 4). The S/B ratios and Z′ values for the filtration assay and SPA were estimated by use of a robust algorithm for the different strains (see Table S1 in the supplemental material). Although the S/B ratio in the filtration assay was higher than that in the SPA (about 20-fold for strain 3D7A), the coefficient of variation for the negative control providing 100% inhibition in the filtration assay was much higher than that in the SPA (41% versus 5% for strain 3D7A). These two effects compensate for each other, resulting in comparable Z′ values in both assays. The molecules tested displayed similar in vitro potencies against P. falciparum strain 3D7A independently of the methodology used. Moreover, four P. falciparum strains with different genetic backgrounds, including multidrug-resistant strains, were used to test the capacity of the assay to evaluate antimalarial inhibitors. Again, the IC₅₀ values for P. falciparum determined using the SPA methodology were nearly identical to the values obtained using the standard filtration assay. The same level of consistency was achieved independently of the genetic background of the parasite used. These results validate the [³H]hypoxanthine SPA to be a valuable methodology to assess the effects of antimalarial drugs irrespective of the P. falciparum strain used.

TABLE 1.

Comparison of in vitro antimalarial activities of test drugs against sensitive and different multidrug-resistant strains by [³H]hypoxanthine filtration-based assay and [³H]hypoxanthine SPA-based assay

Compound	IC50a for the indicated P. falciparum strains
	Filtration assay				SPA
	3D7A	W2	Dd2	TM90C2B	3D7A	W2	Dd2	TM90C2B
Atovaquone	0.6 ± 0.3 nM	0.3 ± 0.2 nM	0.4 ± 0.3 nM	>35 nM	1.1 ± 0.4 nM	0.5 ± 0.2 nM	1.1 ± 0.3 nM	>35 nM
Artesunate	26 ± 6 nM	16 ± 2 nM	22 ± 6 nM	22 ± 8 nM	9 ± 1 nM	3.3 ± 0.5 nM	5 ± 1 nM	8 ± 2 nM
Chloroquine	38 ± 11 nM	572 ± 96 nM	226 ± 137 nM	550 ± 242 nM	27 ± 4 nM	506 ± 168 nM	496 ± 187 nM	605 ± 127 nM
Pyrimethamine	29 ± 9 nM	>20 μM	>20 μM	>20 μM	29 ± 7 nM	>20 μM	>20 μM	>20 μM

^aAverages and standard deviations were extracted from three independent biological replicates, each of which had technical duplicates.

FIG 4.

Comparison of the dose-response curves determined for test compounds against the 3D7A (A), W2 (B), Dd2 (C), and TM90C2B (D) P. falciparum strains using the standard [³H]hypoxanthine incorporation assay by filtration (open symbols) and SPA (closed symbols).

Validation set screening.

The compound set described above was assayed in duplicate using the SPA and a final concentration of 2 μM. The quality of the poly-Lys YOx bead-based scintillation proximity assay in high-throughput mode was very good, with average Z′ and S/B ratio values of 0.69 ± 0.08 and 17 ± 4, respectively. Coefficients of variation for both the positive control (0% inhibition, iRBCs in the presence of DMSO) and the negative control (100% inhibition, RBCs in the presence of 25 μM artesunate) were in the range of 10%. The correlation with the data obtained using the PfLDH assay was also good (Fig. 5A). The statistical cutoff (3 times the standard deviation for the inactive control population) was similar for both assays: 29% inhibition for the PfLDH assay and 27% inhibition for SPA. A total of 384 compounds displayed percent inhibition values above the cutoff in SPA (Fig. 5A, black), and 372 compounds displayed percent inhibition values above the cutoff in PfLDH assay (Fig. 5A, squares). We found 363 common hits in both assays (Fig. 5A, black squares), 9 hits unique to the LDH assay (Fig. 5A, gray squares), and 16 hits unique to SPA (Fig. 5A, black circles).

FIG 5.

(A) Correlation of average responses (n = 2) of a set of 1,006 selected compounds tested at 2 μM in the poly-Lys YOx bead-based scintillation proximity assay and PfLDH-based colorimetric assay in a high-throughput 384-well format. Black, compounds statistically significantly active in the SPA; gray, compounds statistically significantly inactive in the SPA; squares, compounds statistically significantly active in the PfLDH assay; circles, compounds statistically significantly inactive in the PfLDH assay; horizontal and vertical lines, cutoffs for the respective assays. The compound list comprised active and inactive compounds previously characterized in the PfLDH assay format. (B) Empirical receiver operating characteristic (ROC) curve. The area under the curve (AUC) was 0.986 (95% confidence interval, 0.978 to 0.994, which is very close to 1); classifiers with an AUC of >0.90 are typically considered to have an excellent predictive power.

Pearson's correlation coefficient (r) between the LDH assay and SPA was 0.943, indicating an excellent linear correlation between the two assays. In addition, a receiver operating characteristic (ROC) curve was calculated to assess in an alternative way the statistical association between the two assays. In these curves, the ability of a continuous parameter (in this case, the SPA response) to separate two classes (in this case, an LDH assay response of >80% or not) was evaluated by plotting the true-positive rate against the false-positive rate while varying the SPA response cutoff. A perfect classifier would give a point at coordinate (0,1) of the ROC space (a rate of false positives of 0 and a rate of true positives of 1); for a random classification, the curve would be a diagonal. In order to systematically compare ROC curves, it is customary to summarize them into a single parameter by the area under them (the area under the curve [AUC]): a perfect classifier would have an AUC of 1, while a random classification would give an AUC of 0.5.

Figure 5B displays the empirical ROC curve for this system: the AUC was 0.986 (95% confidence interval, 0.978 to 0.994, which is very close to 1); classifiers with AUCs of >0.90 are typically considered to have an excellent predictive power. This again indicates the exceptional capacity of the SPA to differentiate active and inactive compounds compared with that of the LDH assay.

DISCUSSION

In the last few decades, very few P. falciparum targets with low sequence similarity or even no sequence similarity to known human orthologues, to avoid selectivity issues, have been demonstrated to be essential for the parasite (20–22). Even when many molecules with the capacity to modify the activity of these targets have been discovered using biochemical assays, hits have had a limited success due to the lack of whole-cell activity (23). However, whole-cell approaches have proven to be very successful, with several chemotypes being identified in phenotypic screens, and these are now in different stages of the drug discovery pipeline or in clinical studies. For example, one of the most promising novel antimalarial classes under development is the spiroindolones (24). A clear advantage of whole-cell screens is that hits show the desired biological response, which is to be active in the cell with a measurable phenotypic effect in a physiologically relevant context. These drugs may act on more than one molecular target (polypharmacology), resulting in favorable clinical profiles, and may exhibit reduced susceptibility to target-associated resistance, in contrast to classical enzyme inhibitors that interdict single targets or a single point in a cellular pathway and display a higher propensity to select for resistance. However, if later in the development process toxicity issues arise with compounds that have progressed without target-associated information, it is much harder to extricate these toxicity issues from the antimalarial effect without the benefit of a target-based or at least a mechanism-based assay. To overcome this issue, many deconvolution approaches in phenotypic profiling have been developed in recent years (25).

The current gold standard assay for measurement of the whole-cell activity of antimalarial lead compounds against cultured intraerythrocytic asexual forms of P. falciparum is a filtration radioactive assay based on the level of incorporation of [³H]hypoxanthine into nucleic acids, which is proportional to the number of parasitized erythrocytes (7). This assay has the limitations associated with a 96-well filtration radioactive assay, such as low throughput; the need to perform multiple labor-intensive steps, such as filtration, drying, and melting of filters on scintillator sheets; and the generation and disposal of large volumes of radioactive waste. The radioactive assay is being used to evaluate compounds during hit and lead optimization programs, as the number of molecules is limited, but it could become a bottleneck when larger collections of compounds need to be screened. Previous hit identification campaigns were performed by measuring by fluorescent techniques PfLDH activity as a marker of P. falciparum growth (12), but this approach also shows the issues inherent to a surrogate assay. Quantitative PCR (qPCR) has been also used to quantify P. falciparum and P. vivax DNA, mainly for malaria diagnosis (26).

This report describes the conversion of the original 96-well, 108-μl heterogeneous filtration assay for measurement of the level of [³H]hypoxanthine incorporation into a 384-well, 55-μl, homogeneous, high-throughput scintillation proximity assay (Fig. 1). This assay incorporates the crucial advantages of eliminating the need for filtration and washing steps with the concomitant reduction of the amounts of radioactive residues and the amounts of parasite culture (from 100 to 25 μl) and radiolabeled reagents (from 0.2 to 0.1 μCi/well) needed, in parallel with an increase in global throughput and productivity. Although the removal of the washing steps has an impact on the S/B ratios, the well-to-well reproducibility of the SPA is usually better than that of the standard assay (15). In particular, this new HTS format has made possible the use of the gold standard assay to determine the in vitro potencies of up to hundreds of thousands of compounds in the last screen campaign at GlaxoSmithKline (unpublished data), increasing the possibilities of finding new chemical entities with antimalarial activity.

Parasite-infected red blood cells can incorporate the [³H]hypoxanthine supplemented in the medium into nucleic acids (39). Lysis of the infected red blood cells releases tritiated DNA into the medium. Negatively charged DNA very likely interacts electrostatically with the positive charges of the poly-Lys-coated SPA beads, generating the proximity needed to allow scintillation between the tritium and the beads. The release of low-energy β particles stimulates the europium contained in the bead, provoking the emission of light from the YOx crystals. The energy from [³H]hypoxanthine not incorporated into the labeled DNA remains in solution, is absorbed as heat, and is not detected. WGA and membrane binding YOx beads are cationic at neutral pH and have also been shown to have some binding capacity in this assay, but the best results were achieved using poly-Lys-coated beads. YOx beads have the advantage of showing less interference from blue compounds, and among the SPA beads, they also show the highest light emissions (15), although they are heavier than the PS beads and require continuous stirring during dispensation. Poly-Lys-coated beads have been described to protect DNA from nuclease attack (27) and cellular membrane isolation (28) by electrostatic interactions.

SPA is a homogeneous radioisotopic technique which has been widely used during the last 30 years in assay development and biochemical screening, as it enables the analysis of a collection of biological systems, such as enzymatic activity and ligand-receptor, protein-protein, and protein-DNA interactions (29–32), in a rapid, robust, and sensitive way. The main advantage of this technology is that it requires no separation steps, making it well-suited for automation and high-throughput applications, such as screening in drug discovery. It has also been described to be used to quantify nucleic acids (33–35). The SPA imaging beads have an emission maximum peak at 615 nm and can be detected by CCD imagers. The combination of high-density assay microplates and these red-shifted imaging beads offers the ability to simultaneously read all wells with reduced interference from colored compounds. For instance, in our assay, hemoglobin absorbs strongly in the green part of the visible spectrum, but absorption dramatically decreases from 600 nm. YOx SPA imaging beads have their maximum emission peak at 613 nm, allowing the detection of luminescence from the beads even in the presence of the deeply colored hemoglobin from the erythrocyte cultures. This feature allows signal measurement without any separation step, converting the assay into a homogeneous technique ideal for HTS. Moreover, the assay signal is stable for long periods of time (see Fig. S2 in the supplemental material), allowing the testing of large compound sets in single runs.

In this work, the scintillation proximity assay described here for measurement of P. falciparum growth was validated in several ways. First, we demonstrated that the linear relationship between the initial parasitemia and assay luminescence is a straightforward method to measure the level of [³H]hypoxanthine incorporation into the parasite. Second, the IC₅₀ values of well-described compounds with antimalarial activity determined using the SPA methodology were nearly identical to the values obtained using the standard assay by filtration. In particular, four compounds were used in this study: atovaquone, a potent inhibitor of the mitochondrial respiratory chain of the parasite (36); artesunate, an inhibitor of P. falciparum exported protein 1 (EXP1) (40); chloroquine, an inhibitor of erythrocytic heme detoxification (37); and pyrimethamine, an inhibitor of parasitic folate pathway (38). Also, the SPA methodology was compared to the filtration and PfLDH assays for their sensitivity of detection of the inhibitory activities of an annotated compound set from a previous whole-cell GlaxoSmithKline screen, and the correlation and discrimination of active from inactive compounds were also excellent. In terms of throughput and quality, the poly-Lys YOx bead-based scintillation proximity assay was able to generate screening results comparable to those of established surrogate high-throughput assays intended to measure P. falciparum growth, such as the well-established PfLDH assay. The cost per well of this new assay was comparable to that of both the standard filtration and PfLDH assays. When common costs were removed, the cost per well of the standard filtration assay was about 8 cents (0.2 μCi/well hypoxanthine), that of SPA was 7 cents (4 cents for 0.1 μCi/well hypoxanthine and 3 cents for 125 μg beads/well), and that of the PfLDH assay was 7 cents (mostly diaphorase, which cost about 5 cents per well). The net amount of radioactivity used per well was 50% less in the SPA than in the filtration assay. Moreover, the radioactive solution was retained in the assay plate and was disposed of as mixed material; in the filtration assay, the large volumes of radioactivity generated during the washing steps had to be considered and treated as radioactive material.

False-positive results are inherent to the different technologies used to measure parasite growth and thus may appear. In the case of SPA, optical quenchers at the emission wavelength of the beads or displacers of the DNA-bead interaction could result in false-positive results. On the other hand, autofluorescent compounds that fluoresce over the range of the spectrum recorded could provide false-negative results in this technology.

In conclusion, this report demonstrates the feasibility of a simple, rapid, SPA-based assay for use in high-throughput antimalarial drug screening. This assay can be fully automated to be run in a standard HTS platform and test tens of thousands of wells per day (unpublished data).

New antimalarial treatments are urgently required because of the spread of resistance to all classes of antimalarials, including ACTs, which are our last line of defense. Therefore, diverse chemicals with novel antimalarial modes of action and activity against multidrug-resistant strains need to be identified. The use of whole-cell assays in phenotypic screens has been demonstrated to be a reliable starting point for lead optimization programs. The assay described in this article enables the much higher throughput of the gold standard hypoxanthine incorporation assay and will be extremely helpful to identify the next generation of antimalarial drugs.

This article is in loving memory of Emilio Álvarez-Ruiz.