Evaluation of bioluminescence-based assays of anti-malarial drug activity

Background Transgenic Plasmodium falciparum expressing luciferase offers an attractive bioluminescence-based assay platform for the investigation of the pharmacological properties of anti-malarial drugs. Here a side-by-side comparison of bioluminescence and fluorescence-based assays, utilizing a luciferase reporter cassette that confers a strong temporal pattern of luciferase expression during the S-phase of intraerythrocytic development, is reported. Methods Key assay parameters for a range of commercially available luminogenic substrates are determined and compared to those measured using a Malaria Sybr Green I fluorescence assay. In addition, the short-term temporal effects of anti-malarial compounds are evaluated using both bioluminescent and fluorescent assay platforms. Results The Z’, % coefficient of variation and 50% inhibition concentrations are essentially the same for bioluminescent and fluorescent assays in transgenic parasites generated in both chloroquine-sensitive and -resistant genetic backgrounds. Bioluminescent assays, irrespective of the luminogenic agent employed, do, however, offer significantly enhanced signal-to-noise ratios. Moreover, the bioluminescent assay is more dynamic in terms of determining temporal effects immediately following drug perturbation. Conclusion This study suggests that opportunities for bioluminescence-based assays lie not in the measurement of 50% inhibition concentrations, where the cheaper fluorescence assay performs excellently and is not restricted by the need to genetically modify the parasite clone under investigation. Instead, assays that use the dynamic response of the luciferase reporter for semi-automated screening of additional pharmacological properties, such as relative rate-of-kill and lethal dose estimation, are a more attractive development opportunity.


Background
Following the emergence and spread of multidrug resistant Plasmodium falciparum and reports of artemisinin treatment failure in South-East Asia, there is an urgent need to identify new small molecule drugs for the treatment of malaria [1,2]. Towards this aim, several recent reports have described the high throughput screening (HTS) of massive chemical libraries and the identification of novel chemical scaffolds to be taken forward for development [3][4][5]. Prioritizing these compounds for further exploration of their pharmacokinetic and pharmacodynamic properties is essential and has been the focus of several recent developments in assay formats [6][7][8][9][10].
Assays of anti-malarial drug activity typically rely on either radiometric or fluorescence-based platforms. Incorporation of tritiated hypoxanthine or ethanolamine has been widely used as a biomarker for growth and offers a sensitive assay with excellent signal/noise (S/N) ratio [11]. There are, however, significant issues around the safe storage and disposal of the radiolabelled material. Fluorescence-based platforms, such as the use of the DNA intercalating agents Sybr Green I or 4 0 ,6 0 -diamidino-2-phenylindole (DAPI), offer a rapid, cheap and sensitive assay but can be problematic due to the low S/ N ratio resulting from high background signals [12][13][14]. That said, the utility of the Malaria Sybr Green I Fluorescence (MSF) assay for single-dose HTS has been readily demonstrated.
A recent innovation has been the adoption of a bioluminescence assay platform that utilizes transgenic P. falciparum expressing the luciferase reporter gene (luc). The utility of bioluminescence has been demonstrated in HTS of medium-sized chemical libraries, exploration of gametocidal and delayed-death drug action, and the development of a murine malaria assay in Plasmodium berghei [15][16][17][18][19][20]. The use of bioluminescence to assay anti-malarial drug activity in P. falciparum was first described in 2008 [17]. Here luc was flanked by the Pfhsp86 5 0 and Pbdhfr-ts 3 0 sequences and integrated as a short concatamer into chromosome 7. The assay provided 50% inhibition concentration (IC 50 ) results comparable to those determined using 3 H-hypoxanthine or MSF, with a subsequent report demonstrating the feasibility of scaling the use of this transgenic parasite into a 384-multiwell format [16]. Screening the Library of Pharmacologically Active Compounds (LOPAC 1280 ) demonstrated the robustness of this assay format with Z' scores >0.7 and S/N ratio of 71. In 2010, a second luc transgenic parasite with a Pfhrp2 promoter was reported for the screening of three chemical libraries, a total of 12,320 compounds, reporting Z' scores of 0.64-0.76 and S/N ratios of 147-430 [20]. The increase in S/N ratio recorded here presumably represents their choice of the Bright-Glo luminogenic substrate in their assays. A final report describes the bxbI integrasemediated integration of a Pfhrp3-luc-Pfhrp2 cassette into chloroquine-resistant (CQR) and -sensitive (CQS) genetic backgrounds [18]. Whilst the bioluminescence assay data correlated well with that generated using 3 H-hypoxanthine and MSF for fast-acting drugs (Z' score of 0.71 and S/N of 51), the MSF assay appeared to provide poor resolving capacity when screening delayed-action drugs such as azithromycin (Z' scores <0.4 and S/N of 2-3).
The generation of a transgenic parasite where luc is under the control of the 5 0 and 3 0 flanking sequences of Pfpcna has previously been described [21,22]. These flanking sequences have been shown to provide the regulatory elements necessary to reconstitute the absolute and temporal control of the endogenous Pfpcna gene on the luc reporter gene. Pfpcna encodes a processivity factor for DNA polymerase δ and, as such, acts as a biomarker for DNA replication in mature trophozoites during intraerythrocytic schizogony [23]. This transgenic parasite shows a strong 20-50,000 fold-increase in luciferase signal in mature trophozoites when compared to ring stage parasites, suggesting that the matched Pfpcna 5 0 and 3 0 flanking sequences confer a strong dynamic temporal control on luc [22]. Here an evaluation of the Pfpcna-luc transgenic parasite as a tool to explore the action of anti-malarial drugs is described along with a side-by-side comparison of this assay with MSF in both CQS and CQR genetic backgrounds. This study offers insights into the opportunities available for bioluminescence assays to explore a range of pharmacological properties of anti-malarial drugs in a format readily scalable for high throughput assays.

Plasmodium falciparum cell culture
The transgenic Dd2 P. falciparum clone expressing luciferase under the control of Pfpcna flanking sequences (Dd2 luc ) has been previously described [21]. Dd2 luc were cultured using standard continuous culture conditions (RPMI1640 medium supplemented with 37.5 mM HEPES, 10 mM D-glucose, 2 mM L-glutamine, 100 μM hypoxanthine, 25μgml -1 gentamycin and 8% v/v human serum) at a 2% haematocrit in an atmosphere of 1% O 2 , 3% CO 2 , and 96% N 2 . 5nM WR99210 and 2.5 μg/ml blasticidin S drug selection was applied throughout. Staging and parasitaemia were determined by light microscopy of Giemsastained thin blood smears. Synchronization of cultures was attained using sequential sorbitol lysis treatment [24].
Generation of NF54 luc transgenic parasite NF54 attB and pINTNeo were kindly provided by David Fidock (Columbia University, New York). BxbI integrase-mediated integration of the same Pfpcna-luc reporter construct present in Dd2 luc was performed in NF54 attB as previously described, except that WR99210 selection was not applied as the hdhfr selection cassette is omitted from NF54 attB [21,25]. Integration was confirmed by absence of PCR over the cg6 locus, in which the attB site is located (primers P1, 5 0 ATGAACAAA TACATAAGAGCGC3 0 and P2, 5 0 TCTTTAATTTTA TTTTGGTCATGC3 0 ), and generation of a PCR product over the attL site (primers P1 and P3, 5 0 TAAGGAGAA AATACCGCATCAGG3 0 ) that results from attP × attB recombination.
The reconstituted temporal control of Pfpcna expression over luc was shown by stage-specific northern blot and luciferase assay. NF54 luc was synchronized and samples removed at five points during asexual intraerythrocytic development. These samples represented major stages of morphological development as judged by light microscopy; early rings (0-8 hours post-infection, hpi), late rings (9-16hpi), early trophozoites (17-24hpi), late trophozoites (25-36hpi) and schizonts (37-48hpi), termed T1 to T5, respectively. Luciferase assays on T1 to T5 were carried out as described below. For stage-specific northern blots, total RNA was isolated from samples T1 to T5 and 5 μg of total RNA was size-fractionated, blotted and hybridized to α-32 P dATP random-primer labelled probes to the luc and Pfpcna genes. Images were captured using the Cyclone Storage Phosphor screen apparatus (Packard) and analysed using OptiQuant software (Packard).

Standard luciferase assay procedure and time course
A standard single-step lysis procedure was used throughout [22]. Forty μl samples of P. falciparum culture (typically at 2% haematocrit, HCT) were transferred to a well on a white 96-multiwell plate (Greiner, UK) and 10 μl of passive lysis buffer (Promega, UK) added and homogenized by shaking the plate. An equal volume, 50 μl, of luminogenic substrate was mixed with the lysed parasites and the bioluminescence (in relative light units, RLU) was measured for 2 sec in a Glomax Multi Detection System (Promega, UK). Four commercially available (Promega, UK) luminogenic substrates were used in this study: standard luciferase (SL) substrate, Bright-Glo ™ , One-Glo ™ and Steady-Glo TM .
For the timecourse assay, a 2% trophozoite-stage Dd2 luc culture at 2% HCT was prepared. Five 40 μl samples were tested with each luminogenic reagent and the bioluminescent signal from each sample monitored over 100 mins immediately thereafter. Mean RLU ± standard deviation (stdev) were plotted against time. To determine the relative decay of samples, the RLU measured at each timepoint was expressed as a fraction of the initial mean RLU with the mean fraction ± stdev plotted against time.

Dynamics of bioluminescence and fluorescence following drug treatment
A 2% early trophozoite-stage (17-24hpi) Dd2 luc culture at 2% HCT was prepared and divided into 2 ml aliquots; one for each drug/assay to be tested and a further nodrug control. At t = 0 hrs, either three 40 μl aliquots were removed to determine the initial bioluminescent signal or three 100 μl samples removed to measure the fluorescent signal (see below). The drug being tested was added at this time. At the timepoint indicated, samples of the appropriate volume were removed and the bioluminescent/fluorescent signal measured. The mean ± stdev fraction of the bioluminescent/fluorescent signal measured from the drug-treated samples was normalized against the mean bioluminescent/fluorescent signal from the no-drug control at the same timepoint.
Bioluminescent signals were measured using the protocol described above with the SL substrate. Fluorescent signals were measured using a standard MSF assay [14]. Specifically, an equal volume of MSF lysis buffer (100 μl of 20 mM Tris (pH 7.5), 5 mM EDTA, 0.008% (w/v) saponin and 0.08% (v/v) Triton X-100) containing SYBR green I (1 × final concentration, from 5000x stock supplied by Invitrogen, UK) was added to 100 μl of Dd2 luc aliquoted onto a black 96-multiwell plate (Greiner, UK). Well contents were homogenized by repeated pipetting and incubated for one hour in the dark at room temperature. The fluorescent signal, in RLU, was measured using the blue fluorescent module (excitation 490 nm: emission 510-570 nm) of a Glomax Multi Detection System (Promega, UK).

Measuring IC 50 using luciferase and MSF assay formats
Trophozoite-stage cultures of the transgenic line being tested (100 μl, 2% parasitaemia, 4% HCT, n = 3) were added to 96-multiwell plates containing 100 μl of predosed (five-fold dilution series) complete medium. On each assay plate, six wells containing 200 μl of 2% parasitaemia cell culture (2% HCT) in the absence of drugs served as a positive control (100%), whereas the same culture mix in the presence of a 1 μM supralethal dose of artemether served as a negative growth control (0%). The outermost wells on each plate contained 200 μl of incomplete medium to minimize edge effects from evaporation during 48 hr incubation in a gassed (1% O 2 , 3% CO 2 , and 96% N 2 ) chamber at 37°C.
For bioluminescent assays, RLU were measured using the SL substrate. Irrespective of the assay, the % growth was calculated as follows: where μ (S) , μ (+) and μ (−) represent the means for the sample in question and 100% and 0% controls, respectively. The % growth was plotted against log 10 -transformed drug concentration and the IC 50 determined using a nonlinear regression (sigmoidal dose-response/variable slope equation) in GraphPad Prism v5.0 (GraphPad Software, Inc., San Diego, CA).

Determination of assay quality parameters
Derivation of the statistical tests of the assay quality parameters; Z' score, %CV max , %CV min and signal/noise ratio was as described [26]. The Z' score was calculated :

Results
Evaluating the parameters of the Pfpcna-luc bioluminescence assay The generation of the Pfpcna-luc transgenic parasite in the Dd2 (CQR) genetic background (Dd2 luc ) has been previously described [21]. The haematocrit and parasitaemia starting conditions suitable for a 96-multiwell format assay using trophozoite stage Dd2 luc were established using a range of parasitaemia (0.5-5%) at three different haematocrit (HCT; 1,2 and 4%). The standard luciferase (SL) luminogenic substate was used to measure bioluminescence in relative light units (RLU). Correlating RLU against parasitaemia for each HCT used ( Figure 1) shows a strong linear relationship (all R 2 > 0.97) irrespective of the HCT. Increasing the parasite number through fold-increases in HCT does initially (1 to 2% HCT) result in similar fold-increases in RLU; however, the fold-increase in RLU between 2% and 4% HCT is markedly reduced due to the quenching effect of the additional haemoglobin released at these higher HCT [16,22]. Starting conditions of 2% HCT and 2% trophozoite-stage parasitaemia were selected for all future experiments. The selection of 2% parasitaemia balances the effect of kill (2% and below) and growth (assuming a 2-3 fold increase in parasitaemia) on the linear response curve.
The absolute signal intensity and time-dependent decay of RLU generated from a range of commercially available luminogenic substrates was next explored. The luminogenic substrates were added to lysed Dd2 luc and the mean RLU (n = 5) measured over 100 minutes (Figure 2A). The Bright-Glo substrate provided the highest initial absolute signal, some 60-80% higher than those for the SL and One-Glo substrates and 20-fold higher than that of Steady-Glo. Following the time-dependent decay of each luminogenic substrate, determined in terms of absolute RLU (Figure 2A) or as a fraction of initial RLU (Figure 2B), shows that the signal from the Steady-Glo substrate remains essentially unchanged over the 100 minutes assayed, whilst those of Bright-Glo, SL and One-Glo substrates decay to 10%, 22% and 28% of the original signal, respectively. Importantly, within the first five minutes, a time in which a 96-multiwell plate can be readily measured, there is a maximum loss of 10% of the initial RLU irrespective of the luminogenic substrate chosen.
The effect of choice of luminogenic substrate for the determination of IC 50 values was evaluated using chloroquine (CQ) and artemether (ART). Comparison of the absolute RLU ( Figures 3A and 3C) over the range of drug concentrations tested shows the expected relative relationship. Normalizing the RLU as a fraction of the untreated control for each luminogenic substrate generates overlapping non-linear sigmoidal dose-response curves that provide essentially identical IC 50 values irrespective of the substrate chosen ( Figures 3B and 3D). Ten samples of Dd2 luc exposed for 48 hrs to either no drug (100% growth) or a supralethal dose of ART (1 μM, 0% growth) allow the Z' score, S/N ratio and coefficient of variation (%CV) assay parameters to be determined for each luminogenic substrate in a 96-multiwell format ( Table 1). For comparison, the same assay parameters were determined using the MSF assay (Table 1). These data indicate that whilst the Z' scores are essentially the same for all the assay formats tested, and all formats would be readily suitable for HTS, there were significant  differences in the S/N ratios. Critically, luminogenic substrate provides S/N ratios between 469 and 4042 fold higher than those of the fluorescence-based MSF assay. As all the luminogenic substrates share a similar and very low background signal (50-80 RLU), the fold increase in S/N ratio correlates with the initial absolute RLU generated by each substrate. The low variation in the maximal signal (100% growth, %CV max ) across all the luminogenic substrates demonstrates the robust performance of the bioluminescence assays in generating consistent data. The higher %CV min provided by the bioluminescence assays is a direct result of variation of the very low background signal (50-80 RLU).

Bioluminescence is a more dynamic reporter of drug action
The fluorescence and bioluminescence assays of Dd2 luc both monitor DNA replication. MSF directly monitors DNA content and bioluminescence measures the S-phase-linked induction of expression of DNA replication proteins. However, the DNA biomarker is much more stable than the luciferase biomarker, and each is subject to distinct cellular regulation processes. To explore the effect of the stability of these different biomarkers, a lethal drug perturbation was induced and the immediate dynamic responses of the bioluminescence and fluorescent assays followed. For the bioluminescence assays, the SL substrate  was selected for use throughout, its choice reflecting a balance of signal intensity, stability and relative cost. Dd2 luc was exposed to supralethal doses of drug perturbation; kill was induced using 1 μM of the RNA polymerase inhibitor Actinomycin D (ActD) or the ribosome translational elongation inhibitor cyclohexamide (CHX). Over 8 hrs, samples were harvested and the bioluminescent/fluorescent signal measured. In each case, the RLU measured were expressed as a fraction of the mean RLU from an untreated control harvested at the same timepoint to compensate for the normal temporal programme of luc expression over the course of the experiment.
Kill induced by either drug results in decreasing bioluminescent/fluorescent signal over the 8 hours assayed ( Figure 4A). Whereas the bioluminescent signal decreased to between 3 and 22% of the untreated control (square symbols in Figure 4A), the fluorescent signal only decreased to between 57 and 65% of the untreated control over the same period (circles in Figure 4A). Interestingly, the fluorescent signal initially decreased more on ActD treatment, whereas the bioluminescent signal was most affected by CHX. These data suggest that the action of ActD appears to initially reduce Sybr Green I binding, presumably through low-affinity binding to a component of double-stranded RNA, with the inhibition of de novo DNA synthesis by CHX slightly delayed. The residual 65% of fluorescence even after 8 hours of supralethal drug treatment presumably reflects the stability of the nucleic acid biomarker even after cell death. By contrast, supralethal doses of CHX rapidly block the de novo synthesis of luciferase and the induced kill leads to a rapid decline in bioluminescence. Over the 8 hours of treatment, the bioluminescent signal decreases to 3% of the untreated control, with the timeresponse curve indicating that the luciferase protein has a half-life of approximately 1.5 hours in P. falciparum. The action of ActD appears slightly delayed, presumably as luc transcript is available for translation prior to cell death.
Based on the initial dynamic response of the bioluminescence assay to ActD and CHX, the effect of antimalarial drugs was similarly tested. Four drugs were selected; dihydroartemisinin (DHA, 20nM), artemether (ART, 25nM), chloroquine (CQ, 500nM) and azithromycin (AZ, 6 μM) with the concentration of each drug representing the 3xIC 50 dose (see below). The timeresponse curves generated provide relative rate-of-kill data that follows the known relative action of these drugs ( Figure 4B) [10]. The artemisinins ART and DHA provide the fastest rate of kill. CQ represents an intermediate between the artemisinins and AZ, a drug that shows a delayed-death phenotype. The killing-effect of AZ would not be expected to be exerted within the 8 hour time period used here, instead requiring the parasite to complete at least one full cycle of intraerythrocytic development [18].

Side by side comparison of bioluminescence v. fluorescence determination of IC 50
As luciferase expression in Dd2 luc is temporally linked to DNA replication, this parasite offers the most appropriate bioluminescence assay for a side by side comparison with the DNA intercalating action of Sybr Green I in the MSF assay. To extend the comparison of IC 50 derived from bioluminescence and fluorescence assays using the Pfpcna-luc cassette, this cassette was also introduced Figure 4 Exploring the immediate dynamic response of bioluminescence and fluorescence assays of drug activity. (A) Plot of the time-dependent changes in relative RLU (bioluminescent and fluorescence when compared to untreated control) measured from trophozoite stage Dd2 luc (2% parasitaemia, 2% HCT) exposed to a supralethal 1 μM dose of actinomycin D (filled symbol) or cyclohexamide (open symbol). The fraction of mean RLU ± stdev (n = 3) at each timepoint are shown for a luciferase assay (square) or MSF assay (circle). (B) Plot of the time-dependent changes in mean fraction of luciferase RLU ± stdev (n = 3) exposed to 3xIC 50 doses of the indicated anti-malarial drug; artemether (ART), azithromycin (AZ), dihydroartemisinin (DHA) and chloroquine (CQ).
into the CQS genetic background provided by NF54. The Pfpcna-luc cassette was introduced into NF54 attB by bxbI-mediated integration ( Figure 5A) as confirmed by PCR analysis (Figure 5B). To demonstrate that the same temporal control of Pfpcna was reconstituted over the luc gene in NF54 attB , stage-specific northern blots and luciferase assays were performed (Figures 5C and 5D). Taking five timepoints (T1 to T5) that represent major morphological stages during asexual intraerythrocytic development (early ring, late ring, early trophozoite, late trophozoite and schizont, respectively), the expected patterns of temporal steady-state transcript levels and luciferase activity were demonstrated, with the highest levels for both coinciding with S-phase in late trophozoites (sample T4) [22].
Using the luciferase transgenic lines Dd2 luc and NF54 luc , the IC 50 s for eight anti-malarial drugs, representing a range of modes of action, were determined using both bioluminescence and MSF assays ( Table 2). The parental lines Dd2 attB and NF54 attB were included using the MSF assay alone as a control to explore any effect on IC 50 resulting from the genetic modification with the Pfpcna-luc cassette. Irrespective of the assay format, the IC 50 s determined for these drugs were essentially the same. The differences in IC 50 for CQ and quinine (QN) between Dd2 luc and NF54 luc reflect known differences in their resistance profiles for these quinolone drugs. Determining the mean IC 50 ratio for Dd2 luc /NF54 luc for CQ gave values of 8  unchanged from the MSF assays of the related luc transgenic lines.

Discussion
The utility of bioluminescence in single-dose HTS of small-to medium-sized chemical libraries has been readily demonstrated, providing evidence for the luciferase reporter assay as a robust and reliable platform for antimalarial drug discovery [16,18,20]. An evaluation of a range of commercially-available luminogenic substrates suitable for scale-up of this assay platform are evaluated here. This evaluations shows that key assay parameters such as the Z'-score and % coefficient of variation are essentially the same irrespective of the luminogenic substrate selected. Whilst Steady-Glo appears to offer a significant advantage due to its relative signal stability, particularly when considering use of 384 or even 1536-microwell plates, this is simply ameliorated through the use of microinjection devices that deliver substrate immediately prior to detection of the bioluminescent signal. Thus, the standard luciferase substrate, the only commercial luminogenic substrate investigated here that is not co-formulated with lysis reagent (and thus less prone to foaming on microinjection) would perhaps represent the best choice for scale-up. Critically, however, bioluminescence would appear to offer few advantages over the current choice of Sybr-Green I fluorescence assay for single-dose HTS. The advantages offered by the higher S/N ratio and very low background of bioluminescence, whilst extremely helpful for ready analysis of the data, are significantly outweighed by the increased cost of luminogenic substrate and the restriction of study to only those P. falciparum lines that are genetically modified to express luciferase. The opportunities for bioluminescence would thus appear to lie elsewhere. One such area is the exploration of stage-specific activity of anti-malarial drug action. Currently, in addition to the late trophozoite-specific expression of luciferase reported here, P. falciparum transgenic lines that express peak luciferase levels in ring, schizont and different gametocyte stages exist [15,17,20]. The advantages in very high S/N ratio reported here, however, are not as apparent in the studies that describe these other transgenic parasites. For example, whilst this study reports S/N ratios between 15667 and 17746 using the Bright-Glo substrate, the same substrate only provided S/N ratios up to 430 in a transgenic parasite where the luc gene is under the control of a Pfhrp2 promoter [20]. These findings may result from technical issues associated with choice of luminometer or perhaps, more likely, from a need to better understand the role of luc flanking sequences. Using matched Pfpcna 5 0 and 3 0 flanking sequences, a complete reconstitution of the absolute and temporal control of the endogenous gene over the luc reporter can be done [21]. This same observation is similarly true using a second gene, PF0660w, which is expressed in trophozoite-stage parasites (Hasenkamp et al. submitted). These findings would suggest that use of matched flanking sequences from genes subject to high levels of temporallylinked expression may offer the opportunity to develop a library of transgenic parasites suitable for the screening of stage-specific anti-malarial drug activity.
The high rate of luciferase turnover, however, may offer the most attractive opportunity for assay development. Inducing kill using supralethal doses of the translation inhibitor CHX indicates that luciferase has an approximate 1.5 hr half-life in P. falciparum. This high rate of reporter turnover, accompanied by high quantum yield of light from strong matched regulatory sequences, may be exploited to facilitate the in vitro investigation of the immediate dynamics of drug treatment. Rate-of-kill is a critical determinant of anti-malarial drug action and ideally this should be as fast as possible [1]. Rapid rateof-kill is a specific stated attribute in the target product profile for a single-exposure radical cure drug, with the intention of rapidly reducing the parasite burden, and thus clinical symptoms of disease (www.mmv.org). Rateof-kill is typically first characterized during phase IIa clinical trials but an in vitro assay of rate-of-kill has been recently described [10]. As this recrudescence-based in vitro assay of rate-of-kill assay requires some 21-28 days post-drug selection to develop the data, the utility of a more rapid in vitro assay to explore the rate-of-kill of antimalarial drugs is potentially of interest in establishing priorities for anti-malarial drug development programmes.

Conclusion
This study provides proof-of-principle data supporting the application of bioluminescent assays to determine the relative rate-of-kill of four anti-malarial drugs. The relative rate-of-kill as judged from bioluminescent assays is the same as that previously reported from in vivo PRR and in vitro recrudescence rate-of-kill assays [10]. Critically, drugs that exhibit the same mode of action show similar relative rates-of-kill in this novel assay format. Importantly, these assays were performed in a 96-multiwell plate format over a six-hour window which would suggest some potential for development into a rapid in vitro rateof-kill assay more amenable for scale up.