Assessment and Continued Validation of the Malaria SYBR Green I-Based Fluorescence Assay for Use in Malaria Drug Screening

Abstract

Several new fluorescence malaria in vitro drug susceptibility microtiter plate assays that detect the presence of malarial DNA in infected erythrocytes have recently been reported, in contrast to traditional isotopic screens that involve radioactive substrate incorporation to measure in vitro malaria growth inhibition. We have assessed and further characterized the malaria SYBR Green I-based fluorescence (MSF) assay for its ability to monitor drug resistance. In order to use the MSF assay as a drug screen, all assay conditions must be thoroughly examined. In this study we expanded upon the capabilities of this assay by including antibiotics and antifolates in the drug panel and testing folic acid-free growth conditions. To do this, we evaluated a more expansive panel of antimalarials in combination with various drug assay culture conditions commonly used in drug sensitivity screening for their activity against Plasmodium falciparum strains D6 and W2. The detection and quantitation limits of the MSF assay were 0.04 to 0.08% and ∼0.5% parasitemia, respectively. The MSF assay quality was significantly robust, displaying a Z′ range of 0.73 to 0.95. The 50% inhibitory concentrations for each drug and culture condition combination were determined by using the MSF assay. Compared to the standard [³H]hypoxanthine assay, the MSF assay displayed the expected parasite drug resistance patterns with a high degree of global and phenotypic correlation (r² ≥ 0.9238), regardless of which culture condition combination was used. In conclusion, the MSF assay allows for reliable one-plate high-throughput, automated malaria in vitro susceptibility testing without the expense, time consumption, and hazard of other screening assays.

According to the Centers for Disease Control and Prevention (CDC), malaria “is one of the most severe public health problems worldwide” (information can be found at CDC website [http://www.cdc.gov/malaria/impact/index.htm]). In fact, the World Health Organization's (WHO) World Malaria Report indicates that 350 to 500 million clinical malaria infections occur every year, resulting in at least one million deaths per year (38). It is predicted that clinical infections and death will begin to increase due to rapid spread of drug resistance parasites (12, 13, 21, 30, 39). Of the four strains of malaria, Plasmodium falciparum is the most deadly. P. vivax, although it rarely causes death, is a constant annoyance and the leading cause of morbidity due to the dormant phases of the parasite that reside in liver hepatocytes. Both P. vivax and P. falciparum have developed resistance to numerous antimalarial drugs, which has undermined the available options for prophylaxis and treatment (24). Unfortunately, the incidence and specificity of malaria drug resistance is not homogeneous throughout the world. Different geographical locations within the same country can yield malaria parasites with various degrees of sensitivity to commercially available drugs (22). These findings support the need for valid surveillance efforts to predict and determine the level of malaria drug resistance.

Several organizations, including the U.S. Department of Defense, WHO, Wellcome Trust, and CDC, have implemented various drug resistance monitoring efforts. Within the U.S. Department of Defense, the Global Emerging Infections Surveillance and Response System maintains one of the largest networks of malaria drug resistance surveillance sites, including Africa, Southeast Asia, South America, and the Western Pacific islands. The malaria drug screening laboratories within these organizations use both molecular screening techniques to determine drug resistance genotypes and an in vitro cultivation and drug sensitivity assay to determine drug sensitivity levels against a battery of common antimalarial drugs (1, 17, 22, 29). For nearly 40 years, in vitro antimalarial drug sensitivity assays have incorporated the use of radioactive substrates, such as [³H]isoleucine, [³H]hypoxanthine, and [³H]ethanolamine, to measure parasitic growth in the presence known antimalarial drugs or novel test compounds (9, 11, 14). Although these isotopic assays are widely used, accurate, and reliable, they are very expensive, involve multiple processing steps, and require special handling and waste disposal procedures. Thus, the main utility of radioactive assays lies within a fixed research facility and not in the field.

Several in vitro antimalarial drug surveillance microtiter plate assays that involve sensitive nonradioactive procedures have been recently published (3, 7, 10, 28, 33, 37). For example, a few laboratories have described colorimetric enzyme-linked immunosorbent assays (ELISAs) that recognize specific P. falciparum antigens (10, 28). Noedl et al. have demonstrated that histidine-rich protein 2 detection can be used to measure drug sensitivity (27, 28). This assay has since been field tested and yielded results that were similar to those obtained with a modified WHO schizont maturation assay (26). Another ELISA, directed against lactate dehydrogenase (LDH), has been shown to be highly sensitive for determining in vitro drug susceptibility of laboratory strains, as well as field isolates compared to [³H]hypoxanthine incorporation (5, 10, 15). In addition, the enzymatic detection of LDH can be measured to determine parasite chemosensitivity, although at a much lower sensitivity (8, 20). While the ELISAs are accurate and reliable, they are moderately expensive, involve multiple lengthy processing steps, and require monoclonal antibodies that are susceptible to degradation in field environments.

In addition to ELISAs, assays involving fluorescent nucleic acid intercalating dyes, such as SYBR green I, PicoGreen, and YOYO-1, to measure in vitro malaria growth inhibition have been recently described (3, 4, 7, 16, 31, 33, 37). Since mature erythrocytes lack RNA and DNA, binding of the dye is specific for malarial DNA in any erythrocytic stage of P. falciparum development (4). These dyes also display preferential binding to double-stranded DNA versus single-stranded DNA or RNA and are significantly more sensitive, have better-defined spectral peaks, and are less mutagenic compared to ethidium bromide (product inserts; Molecular Probes, Inc., Eugene, OR). In the studies described above, malaria culture strains were treated with known antimalarial drugs and their respective 50% inhibitory concentrations (IC₅₀s) were determined by using fluorescence malaria drug surveillance assays and directly compared to those obtained from traditional radioactive assays. Smilkstein et al. found that the IC₅₀ values determined by SYBR green I binding to malarial DNA displayed results similar to those obtained with [³H]ethanolamine incorporation (33). A recent publication by Bacon et al. (2) indicated that the SYBR green I assay displayed IC₅₀ values for malaria laboratory strains treated with chloroquine, mefloquine, or quinine comparable to those of the histidine-rich protein 2 capture ELISA method in head-to-head studies. Furthermore, these authors provided the first demonstration of the MSF assay's ability to determine drug IC₅₀s from fresh clinical isolates, even in the presence of white blood cells. Studies performed by Corbett et al. and Quashie et al. demonstrated the ability of PicoGreen dye to similarly detect malarial chemosensitivity compared to [³H]hypoxanthine incorporation and/or microscopic examination (7, 31). Also, Corbett et al. showed the utility of the fluorescence-based assay for the screening of novel antimalarial compounds derived from crude plant extracts in the same report. YOYO-1 was used to search known drugs and bioactive compounds for new antimalarials, resulting in the identification of several novel P. falciparum inhibitors (37). Most recently, Banieki et al. demonstrated the robustness and compatibility of the fluorescent dye DAPI (4′,6′-diamidino-2-phenylindole) for monitoring Plasmodium growth in high-throughput antimalarial drug screening using a 384-well microtiter plate (3). These fluorescence-based assays are accurate, reliable, and less expensive than isotopic substrates and antibodies and may involve only one step in one plate to obtain results but are less characterized than the [3H]hypoxanthine assay, which has been the gold standard in malaria drug surveillance and discovery.

In the present study, we have assessed and further validated the malaria SYBR green I-based fluorescence (MSF) assay in our own laboratory for its use in monitoring drug sensitivity. The robustness and suitability of the assay (Z′) for high-throughput screening was examined. We expanded upon previous work reported in the literature (33) by including antibiotics and antifolates in the drug panel and testing various assay culture conditions commonly used in drug resistance screening, such as folic acid-free growth medium. To do this, P. falciparum strains D6 and W2 were treated with a panel of known antimalarial drugs, including chloroquine, mefloquine, quinine, artemisinin, doxycycline, azithromycin, pyrimethamine, dapsone, and sulfadoxine, and their respective IC₅₀s were determined by using the MSF assay. The results were compared to our [³H]hypoxanthine incorporation assay. The resistance indexes for each assay were determined, as well as the global and phenotypic goodness of fit (r² values) between the two assay systems.

MATERIALS AND METHODS

Reagents.

Chloroquine diphosphate salt, quinine hemisulfate salt, mefloquine hydrochloride, artemisinin, pyrimethamine, and 4-aminophenyl sulfone were purchased from Sigma Chemical Co., St. Louis, MO. Doxycycline, azithromycin, and sulfadoxine were obtained from our in-house Walter Reed Chemical Inventory System (WR-CIS). [³H]hypoxanthine was obtained from American Radiolabeled Chemical, Inc., St. Louis, MO. SYBR green I dye (10,000× in dimethyl sulfoxide) was purchased from Molecular Probes.

P. falciparum culture.

P. falciparum strains D6 (CDC/Sierra Leone) and W2 (CDC/Indochina III) were routinely maintained in continuous long-term cultures in RPMI 1640 medium supplemented with 5% washed human A+ erythrocytes, 11 mM glucose, 25 mM HEPES, 32 nM NaHCO3, 29 μM hypoxanthine, and 10% heat-inactivated A+ human plasma (modified from Trager and Jensen [35]), referred to here as tissue culture medium (TCM). Human erythrocytes and plasma were obtained from Valley Biomedical, Inc., Winchester, VA. The erythrocytes were washed several times with RPMI prior to use. The cultures were incubated at 37°C under an atmosphere of 5% CO₂ and 5% O₂, with a balance of N₂.

Prior to performing the drug assays, the parasites were conditioned to their respective test culture condition for 3 to 4 days. The test culture conditions included: TCM (as described above); TCMA, i.e., TCM with Albumax I (lipid-rich bovine serum albumin at 50 mg/ml; Gibco, Inc., Grand Island, NY) substituted in place of the plasma; folic acid-free and p-aminobenzoic acid-free RPMI 1640 medium with the same additional constituents as TCM except for the absence of hypoxanthine (hereafter referred to as FAF); and FAFA, FAF with Albumax I substituted in place of the plasma.

Assessment of SYBR green I fluorescence linearity.

Experimental conditions and Tecan Genois Plus plate reader settings were verified and/or adjusted by examining the SYBR green I fluorescence linearity of parasitemia values between 0 and 5%, as determined by microscopic examination of Giemsa-stained parasites. This method was previously described by Smilkstein et al. (33) but briefly is as follows. Triplicate wells of D6-parasitized erythrocytes (in early ring or schizont stages) were serially diluted with noninfected erythrocytes at a constant 2% hematocrit in culture medium (100-μl final volume). Next, 100 μl of lysis buffer [20 mM Tris (pH 7.5), 5 mM EDTA, 0.008% (wt/vol) saponin, and 0.08% (vol/vol) Triton X-100] containing SYBR green I (1× final concentration) were added directly to the plates and gently mixed by using the Beckman Coulter Biomek 2000 automated laboratory workstation (Beckman Coulter, Inc., Fullerton, CA). The plates were then incubated for another hour at room temperature in the dark and examined for the relative fluorescence units (RFU) per well using the Tecan Genios Plus (Tecan US, Inc., Durham, NC). The data analysis was performed by MS Excel. The percent parasitemia (x values) was plotted against the RFU (y values, after background subtraction of noninfected erythrocytes) and analyzed by linear regression to determine the goodness of fit (r² value).

Determination of assay quality (Z′).

The statistical test published by Zhang et al. (40) was used to determine assay quality (or Z′). It is calculated as follows: Z′ = 1 − [(3σ₍₊₎ + 3σ₍₋₎)/ μ₍₊₎ − μ₍₋₎], where μ₍₊₎ and σ₍₊₎ are the mean and standard deviation of the infected erythrocytes (positive control), respectively; μ₍₋₎ and σ₍₋₎ are the mean and standard deviation of the noninfected erythrocytes (negative control), respectively, and the denominator value is the absolute value of the difference in the positive and negative control means.

Preparation of predosed microtiter drug plates.

Sterile 96-well tissue culture plates containing 11 twofold serial dilutions of each antimalarial drug, originally suspended in dimethyl sulfoxide or 70% ethanol at various stock concentrations, in test culture medium were produced in advance and frozen at −80°C until use (stored no more than 1 week) or made freshly the day of the assay. No difference was seen in IC₅₀ determination between previously frozen or fresh drug assay plates (data not shown). The Beckman Coulter Biomek 2000 was used to produce all drug assay plates.

[³H]hypoxanthine incorporation assay.

This assay was based on modifications of previously described methods (6, 9, 23). Briefly, P. falciparum strains in late-ring or early-trophozoite stages were cultured in the predosed 96-well microtiter drug assay plates in 200-μl volumes at a starting parasitemia of 0.8% and a hematocrit of 1% for 48-h incubations or at 0.4 and 1% for 72-h incubations, respectively. Microtiter plate wells containing noninfected erythrocytes in the absence of drugs served as negative controls, whereas parasitized erythrocytes in the absence of drugs served as positive controls for parasite growth on each plate. [³H]hypoxanthine in culture medium (25 μl) was added to the culture during the last 24 h of the assay to allow for incorporation into the live parasites. The Beckman Coulter Biomek 2000 was used to dispense the parasites and [³H]hypoxanthine. All drug assays were conducted in a Tri-Mix tissue culture incubator at 37°C under a humidified atmosphere of 5% CO₂ and 5% O₂, with a balance of N₂. After the respective incubation was complete, the assay plates were frozen to lyse the cultures. The parasite DNA was recovered by harvesting the lysate onto glass-fiber filter plates using a Packard FilterMate cell harvester (Packard Instruments Co., Downers Grove, IL), and the radioactivity (in counts per minute [cpm]) was counted on a Packard TopCount microplate scintillation counter (Packard Instruments). The cpm per well (y value) at each drug concentration (x value, transformed to the log[X]) was plotted and analyzed by using nonlinear regression analysis (sigmoidal dose-response/four parameter equation) with an oracle database computer program (DataAspects Corp., California) to determine the IC₅₀ value for each drug tested.

MSF assay.

The MSF assay, adapted from Smilkstein et al. (33), used the same specimen processing, malaria culture techniques, and positive and negative control wells as the standard [³H]hypoxanthine incorporation assay described above. However, 100-μl culture volumes at an optimized starting parasitemia 1% and hematocrit of 2% were used for either 48- or 72-h culture incubations (data not shown). After the plates were cultured for the respective amounts of time, 100 μl of lysis buffer containing SYBR green I was added directly to the plates, followed by gentle mixing and incubation for another hour at room temperature, as described above. The plates were examined for the RFU per well using the Tecan Genios Plus. The drug concentrations (x value) were transformed by using X = Log[X] and plotted against the RFU (y values). The data was then analyzed in GraphPad Prism (GraphPad Software, Inc., San Diego, CA) by nonlinear regression (sigmoidal dose-response/variable slope equation) to yield the IC₅₀, as previously described (33).

Statistical analysis.

Data were analyzed by analysis of variance and Tukey's multiple comparison test (Sigma Stat; SPSS, Inc., Chicago, IL).

RESULTS

Sensitivity of SYBR green I detection of P. falciparum.

SYBR green I fluorescence staining of parasites, as well as that of PicoGreen, has been demonstrated to show a linear relationship between parasitemia and measured fluorescence over a wide range of parasitemia (7, 33). We verified the fluorescence linearity of the SYBR green I over a range of known parasitemia in different life cycle stages as determined by microscopic examination of Giemsa-stained D6 parasites (Fig. 1), which has not been previously characterized for the SYBR green assay (33). The culture-synchronized parasites used in these experiments were initially sampled while in early-ring stage (>95%), and then the same culture was allowed to progress to schizonts (>95%) and resampled at the same parasitemia. As shown in Fig. 1, a linear relationship between SYBR green I fluorescence and parasitemia was displayed regardless of DNA content of the parasites, which varies so dramatically during their eythrocytic life cycle. In addition, the sensitivity of the assay was examined by determining the detection limit (DL) and quantitation limit (QL) of the assay (34). To do this, the DL was defined as the lowest parasitemia in a well that can be detected above noninfected erythrocyte staining, while the QL was defined as the lowest parasitemia in a well that can be quantitatively determined with suitable precision and accuracy. Therefore, the QL for this method is the equivalent to the lowest parasitemia that can be statistically differentiated from the next lower one tested. The DL varied based on the erythrocytic stage of the parasite examined, such that it was 0.08% for early rings and 0.04% for schizonts. The QL was determined to be ∼0.5% parasitemia for both early rings and schizonts since parasitemias examined for each below this threshold were not statistically different from one another, while those above were.

FIG. 1.

Assessment of fluorescence linearity. The data are presented as the mean of at least duplicate samples ± the standard deviation and are from a representative experiment of three experiments. The data were analyzed by linear regression. The datum points below the QL are not significantly different from one another (P ≥ 0.05).

Assessment of the MSF assay.

We next examined the suitability of the MSF assay (33) for its use in monitoring drug resistance in P. falciparum. Preliminary studies with quinoline family members and artemisinin indicated the occurrence of a significant edge effect in the MSF assay, which was independent of parasite strain used or volumetric changes occurring during drug incubation. Table 1 shows the results from a representative experiment that illustrates the significant edge effect on chloroquine IC₅₀ determination (48-h incubation) in the MSF assay. Similar results were obtained at 72 h (data not shown). Therefore, the microtiter plate edge was excluded in all subsequent experiments for the determination of IC₅₀ values. The quality or robustness of the assay for high-throughput screening was calculated by using the statistical measure of Z′ (40) as described in Materials and Methods. Assays that display a Z′ value of ≥0.5 are generally acceptable for high-throughput screening (HTS) (37). To determine this value, half a plate each of noninfected and D6-infected erythrocytes (starting at late-ring or early-trophozoite stages) in the absence of antimalarial drugs were cultured for 72 h and analyzed by using the MSF assay. The assay quality (Z′) varied depending on the culture medium used, ranging from 0.73 when cultured in folic acid-free medium with Albumax I (FAFA, the most growth suppressive medium used in our drug screening laboratory) to 0.95 when cultured in standard RPMI 1640-based tissue culture medium with plasma (TCM, the most growth permissive medium used). The MSF assay exhibited a signal-to-noise ratio of between 3:1 and 8:1 for FAFA and TCM, respectively, which correlated with the growth permissiveness of the culture medium.

TABLE 1.

Edge effect on MSF assay IC₅₀ determination for chloroquine

Starting row/column	Complete row		Row without edgesa
	IC50 (ng/ml)	r2	IC50 (ng/ml)	r2
A1	124.6	0.8915	97.16	0.9844
B1	148.7	0.9693	141.8	0.9958
C1	153.0	0.9542	144.4	0.9841
D1	151.4	0.9701	144.1	0.9893

^aExcluding columns 1 and 12 of the 96-well microtiter plate.

Validation of the MSF assay.

We next validated the MSF assay (33) for its use in monitoring drug resistance in P. falciparum. For these studies, two well-characterized P falciparum strains, D6 and W2, were cultured in the presence or absence of known antimalarial drugs for 48 or 72 h in various culture conditions routinely used in the [³H]hypoxanthine incorporation assay by our malaria in vitro drug screening laboratory. The drugs tested included chloroquine, mefloquine, quinine, artemisinin, doxycycline, azithromycin, pyrimethamine, dapsone, and sulfadoxine. The complete tissue culture medium used included standard RPMI 1640-based medium with human plasma (TCM), RPMI 1640-based medium with Albumax I (TCMA), folic acid-free and p-aminobenzoic acid-free RPMI 1640-based medium with human plasma (FAF), and folic acid-free and p-aminobenzoic acid-free RPMI 1640-based medium with Albumax I (FAFA). Their respective IC₅₀s were determined by using the SYBR green assay. The results were then compared to the WR-CIS historical IC₅₀ data and/or side-by-side experiments generated using our standard [³H]hypoxanthine incorporation assay. The results of the 72-h experiments are shown in Table 2.

TABLE 2.

Comparison of IC₅₀ values determined using the MSF and [³H]hypoxanthine incorporation assays at 72 h of incubation

Strain and test drug	Mean IC50 (ng/ml) ± SEMa as determined by:
	MSF assay					[3H]HPX assay
	TCM	TCMA	FAF	FAFA	Combined
D6
Chloroquine	9.4 ± 0.8 (6)	10.4 ± 1.1 (4)	8.1 ± 0.8 (12)	8.1 ± 0.6 (13)	8.6 ± 0.4* (35)	4.1 ± 0.1 (136)
Mefloquine	43.7 ± 4.1† (6)	19.9 ± 5.1† (4)	32.8 ± 2.1 (7)	16.0 ± 3.0† (5)	29.6 ± 3.0* (22)	4.5 ± 0.3 (78)
Quinine	15.8 ± 1.6 (6)	11.3 ± 1.7 (4)	14.9 ± 0.5 (7)	11.9 ± 1.2 (5)	13.8 ± 0.7* (22)	5.0 ± 1.0 (3)
Artemisinin	1.25 ± 0.17 (6)	1.7 ± 0.27 (4)	1.7 ± 0.22 (7)	1.6 ± 0.18 (3)	1.5 ± 0.1* (20)	2.4 ± 0.2 (24)
Doxycycline	1,175.3 ± 38.8 (3)	2,371.5 ± 406.1 (4)	856.9 ± 62.6 (6)	1,845.8 ± 484.1 (5)	1,521.3 ± 207.8 (18)	483.3 ± 54.5 (6)
Azithromycin	10,134.3 ± 697.6 (6)	15,277.0 ± 1193.5 (4)	10,073.4 ± 666.4 (10)	14,364.7 ± 640.5 (9)	12,135.5 ± 556.2* (29)	3,746.8 ± 636.7 (30)
Pyrimethamine	ND	ND	0.5 ± 0.2 (6)	0.4 ± 0.1 (8)	0.4 ± 0.1 (14)	0.1 ± 0.1 (14)
Dapsone	ND	ND	3.1 ± 0.4 (4)	10.2 ± 1.5 (4)	6.7 ± 1.5 (8)	2.9 ± 0.4 (8)
Sulfadoxine	ND	ND	10.5 ± 7.2 (5)	2.4 ± 1.1 (4)	6.9 ± 4.1 (9)	16.0 ± 4.7 (15)
W2
Chloroquine	248.5 ± 29.2 (4)	331.5 ± 30.3 (3)	274.6 ± 20.0 (10)	267.7 ± 26.9 (10)	274.5 ± 13.6* (27)	130.4 ± 3.3 (137)
Mefloquine	16.2 ± 2.0 (3)	8.9 ± 1.2 (3)	17.4 ± 3.0 (9)	9.7 ± 0.7 (3)	14.5 ± 1.7* (18)	2.3 ± 0.1 (78)
Quinine	167.5 ± 17.1 (4)	128.5 ± 25.3 (3)	162.3 ± 20.4 (9)	92.3 ± 3.1 (3)	147.0 ± 12.2* (19)	36.0 ± 9.5 (3)
Artemisinin	1.0 ± 0.19 (4)	1.4 ± 0.4 (4)	1.2 ± 0.4 (7)	1.1 ± 0.03 (3)	1.2 ± 0.2 (18)	1.7 ± 0.2 (27)
Doxycycline	2,413.3 ± 438.5 (3)	2,281.3 ± 310.5 (4)	1,863.0 ± 423.4 (4)	2,134.7 ± 343.6 (3)	2,158.6 ± 178.2* (14)	814.0 ± 57.8 (5)
Azithromycin	1,623.3 ± 267.5 (3)	2,733.6 ± 502.4 (5)	1,847.9 ± 375.5 (7)	4,488.8 ± 526.6 (8)	2,929.7 ± 344.0 (23)	5,500.4 ± 1,053.8 (23)
Pyrimethamine	ND	ND	8.3 ± 4.8 (3)	37.6 ± 9.8 (4)	25.0 ± 8.1 (7)	11.9 ± 2.1 (7)
Dapsone	ND	ND	156.2 ± 23.9 (3)	241.6 ± 27.8 (4)	204.9 ± 24.5* (7)	124.1 ± 7.3 (9)
Sulfadoxine	ND	ND	1,642.1 ± 674.0 (3)	2,417.2 ± 888.3 (3)	2,107.1 ± 564.1 (5)	1,112.9 ± 192.1 (10)

^aMeans of experiments run on different days are given. The “Combined” data column represents the means of all of the TCM, TCMA, FAF, and FAFA data together for each respective test drug. *, P < 0.05 compared to the [³H]hypoxanthine (HPX) incorporation assay; †, P < 0.05 compared to the MSF “Combined” data. ND, not determined.

To ensure the experimental reproducibility of the assay, MSF data was collected over the course of 6 months from January to September 2006 and side-by-side [³H]hypoxanthine data experiments were collected concurrently. The historical WR-CIS [³H]hypoxanthine data included values dating back to the following for each drug: chloroquine, January 2003; mefloquine, December 2004; quinine, September 1996; artemisinin, August 1994; doxycycline, May 2003; azithromycin, September 2002; pyrimethamine, May 2003; and sulfadoxine, May 2003. For the purposes of the present study, all side-by-side and historical [³H]hypoxanthine data were combined and averaged and, therefore, statistical analyses were performed against combined data from all MSF assay culture conditions tested versus the combined [³H]hypoxanthine data. The IC₅₀ values from the MSF assay showed the expected pattern of drug resistance for both parasitic strains tested and were the same or similar to those from the radioactive assay, although statistical differences do exist (Table 2). In addition, the drug IC₅₀s determined for each parasite strain were insignificant for the various assay culture conditions tested, except for mefloquine-treated D6 parasites (Table 2). The 48 h data displayed the same or similar IC₅₀ values and expected drug resistance pattern as those determined at 72 h (data not shown).

To ensure that the drug resistance patterns observed in antimalarial-treated D6 and W2 were indeed similar, the resistance index for each antimalarial drug was further examined. The resistance indexes are defined as ratios of the IC₅₀ of a drug against W2 to the IC₅₀ of the same drug against D6 (18). The calculated indexes for 72 h experiments using combined MSF and [³H]hypoxanthine data are shown in Table 3. The patterns of resistance between the MSF and radioactive assays were identical or comparable. The index for sulfadoxine appears different, but it is not, based on the following two reasons: (i) the original sulfadoxine IC₅₀ values for the MSF and [³H]hypoxanthine assays are not statistically different from one another so the indexes in turn cannot be, and (ii) the standard error of the mean of the W2 values for both assays was high enough to exaggerate the resistance index value differences seen between the two assays (Table 2). This variability was not observed for the IC₅₀s of sulfadoxine against D6. Similar results were obtained for 48-h experiments (data not shown). The global and phenotypic goodness of fit (r² values) between the two assay systems was additionally plotted to directly determine their correlation to one another for the 72-h experiments. Figure 2A shows the global correlation for all drugs tested with combined data from both parasite strains (r² = 0.9315). Figure 2B and C show the phenotypic correlations for all of the drugs for either D6 or W2, respectively. The values for each individual strain were nearly identical (D6, r² = 0.9288; W2, r² = 0.9238).

TABLE 3.

Comparison of the drug resistance indexes by the MSF and [³H]HPX incorporation assays

Test drug	Drug resistance indexa as determined by:
	MSF assay	[3H]HPX assay
Chloroquine	31.9	31.8
Mefloquine	0.5	0.5
Quinine	10.7	7.2
Artemisinin	0.8	0.7
Doxycycline	1.4	1.7
Azithromycin	0.2	1.5
Pyrimethamine	56.8	99.2
Dapsone	30.6	42.8
Sulfadoxine	305.4	69.6

^aDetermined as The ratio of the W2 IC₅₀ to the D6 IC₅₀ at 72 h of incubation.

FIG. 2.

Scatter plot for the correlation of drug IC₅₀s determined by the MSF assay and the [³H]hypoxanthine incorporation assay. (A) Global analysis of both D6 and W2 strains; (B and C) phenotypic analysis of D6 (B) or W2 (C) alone. Solid shapes, D6 strain; open shapes, W2 strain. Symbols: cross, chloroquine; pentagon, mefloquine; rhombus, quinine; square, artemisinin; triangle, doxycycline; inverted triangle, azithromycin; diamond, pyrimethamine; circle, dapsone; wave, sulfadoxine.

DISCUSSION

This report further establishes the value of SYBR green I for its use in malaria drug surveillance. As shown by Smilkstein et al. (33), there was a linear increase in SYBR green I fluorescence intensity that correlates with the increase in parasitic DNA. We took this one step further and showed that the MSF assay allowed for the linear detection of parasites regardless of their life cycle stage (Fig. 1). The DL for the assay was in the range of 0.04 to 0.08%. The increased DNA content for the schizont forms allowed for more sensitive detection due to their relative increased DNA content compared to the early-ring forms. This agrees well with the confocal microscopic assessment of the P. falciparum stage-specific binding of SYBR green I previously described by Bennett et al. (4). Therefore, the conservative DL for this assay should be considered 0.08%. This is four times more sensitive than that reported for DAPI detection of trophozoite-stage infected erythrocytes using a fluorimeter (3). The quantitative limit was ca. 0.5% (Fig. 1). For our in vitro assays that use malaria laboratory strains, we regularly use a starting parasitemia of 1% since this amount is easily attainable during routine culture and gives maximum assay resolution in our hands.

Additional efforts to increase the sensitivity of the MSF assay, including adjustments in the starting parasite density and hematocrit levels, microtiter plate selection, and plate freezing were undertaken, but did not significantly affect the results obtained (data not shown). To summarize, initial parasitemia between 0.2 and 1% were examined, with the most reliable results occurring with parasitemia above 0.5% (in agreement with our experimental QL determination). No significant difference was seen in tests with hematocrit levels between 1 and 4%. Interestingly, the use of phenol red-free basal medium in our hands did not affect the fluorescence background in the MSF assay as had been previously reported (4). The use of higher dye concentrations and labeling of nonlysed cells may explain the observable differences (4, 33). Using this methodology (4), we found that SYBR green I intercalated initially into the plasma membrane of the erythrocytes and took several hours to specifically label the parasite DNA. Given enough time for DNA binding, the growth inhibition curves using nonlysed cells were similar to those obtained here for the MSF assay with lysed cells. Using the SYBR green I lysis buffer eliminated staining variability, resulting in tighter IC₅₀ value determinations. Black microtiter plates commonly used to enhance fluorescence plate reader applications did not affect test outcome compared to standard transparent tissue culture 96-well microtiter plates, in contrast to what has been previously reported (4). This finding significantly reduces overall assay cost. Freezing the plates before or after the addition of dye minimally enhanced SYBR green I fluorescence signal in our MSF assay, unlike that previously reported (33), since complete disruption of the cultures was obtained by using the Biomek 2000 robotics platform to add the SYBR green I lysis buffer and mix the suspension. In addition, use of the Biomek 2000 reduced plate manipulation time by an average of 60% and significantly increased assay intra- and interplate reproducibility compared to hand pipetting (data not shown).

Preliminary experiments testing the MSF assay indicated that a reproducible edge effect existed for certain drugs, such as chloroquine (Table 1). It resulted in lower IC₅₀ values for those drugs affected and could potentially influence the determination of resistance during screening. Elimination of the edge from IC₅₀ calculation significantly increases the intra- and interplate reproducibility of the MSF assay. We believe that the data indicate that the edge effect is most certainly real. The upper asymptote of our curves are consistently more affected by the edge effect than our lower ones, indicating that the effect likely indicates decreased growth rather than a loss of detection. Furthermore, the effect is exacerbated in the four corner wells, which actually have two edges as the sides of the plate converge. This is important for malaria drug surveillance in which small changes in the IC₅₀ values reported could misleadingly affect the resistance profile for a study population in a region where malaria is endemic. It could also affect patient treatment regimen based on the misdiagnosis of their parasite resistance index. The cause of the edge effect remains unclear but was not due to volumetric differences (i.e., well dehydration) or to intrinsic fluorescence properties of the drugs tested (data not shown). Resting the plates for 15 min at room temperature prior to incubation also did not minimize the edge effect, as has been suggested by others (3, 19). To avoid this issue, the outer wells composing the edge of the microtiter plate were not included in the data sets.

For the industrial high-throughput prescreening for novel antimalarials, the use of the edge may be initially less important and allow for higher per-plate throughput. Most colorimetric and fluorimetric method-based assays display intrinsic background. This background is typically subtracted from experimental values, resulting in the lower asymptote of the dose-response curve resting on or very near to the x axis of the plot. It does not, however, affect the calculation of the IC₅₀ value since the difference between the upper and lower asymptotes does not change with the subtraction of the background in relationship to the log of the drug concentration (x value), the RFU (y value), and the Hill slope. Also, the reduction of background noise is not always possible or advisable. One could fix, permeabilize, and then label the MSF assay plates. This would allow the removal of any unbound dye through centrifugation and lower the intrinsic background of the assay, as was done in the work by Banieki et al. with DAPI. However, this adds numerous steps to fix, permeabilize, and wash the assay plates, including centrifugation—- all of which dramatically increase the assay time and associated labor and material expenses. These additional steps also significantly impair the assay's field and HTS amenability.

For true fully automated HTS, the assay is most preferably performed in one plate with as few steps as possible, so these extra steps are not acceptable. Therefore, many investigators measure the HTS fitness and assay robustness by determining the Z′ since it weighs the error associated the assay-positive and -negative values with the absolute difference between the positive and negative values (40). Importantly, the Z′ had not been previously reported for the MSF assay. We found the assay quality to be more than fit for high-throughput screening efforts (0.73 ≥ Z′ ≤ 0.95), regardless of the culture conditions or parasite strain used. Our value range is similar to that previously described for YOYO-1 (median Z′ of 0.75) (37) and DAPI (Z′ = 0.799) (3). In addition, the assay appears to work very well in both 384- and 1536-well formats (Elizabeth Winzeler, unpublished data).

The values we obtained for our MSF assay drug resistance studies for the combined data were similar or exactly the same as those obtained using the [³H]hypoxanthine incorporation assay for both 48- and 72-h assays (data not shown and Table 2). Statistically significant differences were admittedly seen between IC₅₀ values generated between the two assays, as shown in Table 2. In spite of these differences, the assays showed the same sensitivity index (Table 3). Historically, W2 is resistant to chloroquine, quinine, and pyrimethamine and susceptible to mefloquine, whereas D6 tends to be more resistant to mefloquine and susceptible to chloroquine, quinine, and pyrimethamine. This drug sensitivity profile is the same, indicating that although statistically significant differences do exist between the two assays, they are likely not biologically significant. Therefore, when the MSF assay is used, it may be necessary to develop new drug resistance threshold IC₅₀ values for certain drugs in order to more precisely compare them to the isotopic assay results. More evidence that the differences seen are not biologically significant comes from the high degree of correlation between the MSF and radioactive assays (global r² = 0.9315, Fig. 2A). Furthermore, no strain specificity in the correlation was evident, as demonstrated by the nearly identical phenotypic IC₅₀ values for D6 (r² = 0.9288) and W2 (r² = 0.9238), respectively (Fig. 2B and C). These values are in agreement with those previously reported for the DELI assay versus the radioactive assay (r² = 0.95) (10) and histidine-rich protein 2 ELISA versus the radioactive assay (r² = 0.892) (28).

The data collection also occurred over 6 months in order to facilitate the true reproducibility of the assay. Parasitic IC₅₀ values can sometimes flux up and down for short periods of time. Our data collection approach and use of historical radioactive data eliminated the chance of this factor skewing our results in the present study compared to other studies that compared only a few experiments within a short period of time. The drug panel used in the present study is also representative of the diversity in chemotype and mechanism of action seen within the portfolio of most malaria drug programs, allowing the present study to serve as a benchmark for future studies. Furthermore, the testing of the various culture conditions was necessary since antifolates are included in screening panels, and their in vitro activity can be greatly diminished in the presence of culture medium containing folic acid and p-aminobenzoic acid (23). This is why known antifolates in our test drug panel were not examined in TCM or TCMA in the present study. With the exception of mefloquine-treated D6, no significant differences were seen between MSF assay IC₅₀s from the different culture conditions. However, plasma-binding trends for certain drugs were noticeable. For example, doxycycline and azithromycin displayed slightly lower IC₅₀ values in plasma-containing medium compared to those containing Albumax I. Mefloquine and quinine showed the opposite trend, with lower IC₅₀s in Albumax I-containing medium. This is supported by previous observations in the literature that document extensive mefloquine binding to plasma proteins (25, 36). Furthermore, many antifolate-like compounds have additional bioactivities in addition to their antifolate properties. One could access the in vitro parasite activities associated with these differing mechanisms of action by assaying the parasites drug sensitivity in both folic acid-free and folic acid-containing media.

The information reported here was important for us to determine since we have more than 30 years of [³H]hypoxanthine incorporation assay data in our WR-CIS. The use of radioactive assays at many research and surveillance outposts is being eliminated due to regulatory and disposal issues. Therefore, we sought to find a suitable replacement for the radioactive assay at these locations. The value of the MSF assay, a fluorescent dye-based system, was clear. First, it is a one-plate assay, unlike the radioactive assays that require costly filter plates, and uses standard, sterile transparent tissue culture microtiter plates. Fewer culture reagents are required, since the culture volume has been scaled to one-half that of the radioactive assays to facilitate one-plate assay conditions. Second, the MSF assay is a one-step assay, meaning that dye and/or lysis buffer solution is added directly to the malaria cultures, briefly incubated, and then read on the fluorometer. The dye is stable enough for actual field use in tropical environmental conditions. Unlike the ELISAs currently available, the assay requires no antibodies or washing steps, which requires centrifugation. Third, any fluorometer model can be used to obtain MSF readouts. The filter set needed for data acquisition is the standard set used for the detection of fluorescein and is readily available worldwide. It is critical that we use the same acquisition and data analysis platforms (i.e., MS Excel and GraphPad Prism) for the purposes of multisite standardization and centralized data capture (i.e., the WR-CIS). Lastly, the cost per MSF assay is 100 times less than the [³H]hypoxanthine assay for the reasons described above and due to the absence of expensive radioactive substrate use and disposal (33). A recent cost analysis breakdown by Bacon et al. also indicated a cost saving of at least 10-fold for the MSF assay compared to the HRP II ELISA (2).

In conclusion, compared to the standard radioactive assay, the MSF assay is more cost-effective, simple, less hazardous, and field amendable, while still allowing for accurate high-throughput, automated drug resistance surveillance and testing at global malaria surveillance sites. We provide evidence that the MSF assay is robust and amenable to HTS as a one-plate/one-step method, in contrast to other multiplate and/or step nonradioactive detection systems such as those involving DAPI, pLDH, and histidine-rich protein 2 (3, 10, 28). These efforts are important, because with several surveillance networks using the same assay system, the malaria drug sensitivity data can be compared and analyzed from various drug resistance programs in addition to our own (38). This in effect will increase the usefulness of the data generated from all of these malaria drug resistance surveillance sites and facilitate the capture of the data into a centralized database (32).