In order to work with homogeneous parasite populations, tightly synchronized parasites were used in all experiments, along with an artesunate concentration which irreversibly damages 100% of the parasites within 3 hours. The lethal dose of artesunate, defined here as the minimal concentration needed for a 3 hour exposure to completely prevent reinvasion (see "Materials and methods" section), was 780 nM. Importantly, it corresponds to 100–1000 times the IC₅₀, this dose remains physiological, being in the range of peak plasma concentration reached in patients administered artesunate (1300–8400 nM) 11.

Each experiment consisted in an artesunate-free culture (control) and an artesunate treated culture. For each time point explored, cDNA from parasites with and without the drug were labeled with different cyanines, mixed and hybridized. To analyse dynamic transcriptome alterations, 2 pilot experiments were performed in which RNA was harvested at 30, 90, 180 minutes and 10 hours of incubation with the drug. No significant changes in gene expression could be detected at 30 minutes. In contrast, after 10 hour incubation time, major transcriptome alterations occurred under artesunate, but their analysis was un-interpretable due to drug-induced massive slowing-down of parasite development (data not shown). We thus decided to focus on 2 drug-exposure time, namely 90 minutes and 3 hours after drug addition.

Since artesunate is active on ring stages as well as on older trophozoite stages, we decided to explore different time points along the erythrocytic cycle and search for genes affected at each time point. Therefore, 5 drug treatment experiments, staggered between 20 hours and 30 hours of parasite development, were performed for the comparison of drug versus no drug at 90 minutes and 3 hours.

We looked for genes displaying similar changes across all experiments, with the hypothesis that responses to artesunate would not be stage-dependent within the explored 10 hour window of the developmental cycle. The results of 5 experiments performed at time points staggered between 20–30 hours post-invasion, were analyzed using ANOVA with a Holm-Bonferroni method of p-value adjustment, p < 0.05. This led to a list of genes with similar ratios of expression (i. e. low variance). From this list, only genes differentially expressed at 3 hours under artesunate and not after 3 hours in the absence of the drug were retained (p = 1). It is important to stress that what we refer to, for convenience, as over- or under-expression in fact refers to higher or lower steady state RNA levels.

Due to the high number of microarrays analyzed (32 at 3 hours), power of the statistical analysis was such that expression log ratios as low as +/- 0.3 could be found for genes that displayed significant changes by ANOVA [for distribution of log ratios, see Additional file 1]. This appeared questionable from a biological standpoint and we chose to use a cut-off log-ratio of +/- 0.8. This value is lower than what is usually (and often arbitrarily) chosen in most transcriptome studies, i.e. +/- 1, but we felt it was justified given the large amount of data analyzed here and the low amplitude of fold-changes reported in previous P. falciparum transcriptome studies 812.

In addition, as genes of potential interest could be differentially expressed in the controls but more-so or differently under artesunate, we needed to identify these and filter out the developmentally regulated gene expression profiles. Indeed, the extent to which the drug slowed-down parasite development is unknown. Thus, different expression levels in artesunate-exposed vs. control (unexposed) cultures could reflect the direct effect of artesunate together with a possible difference in growth rate/developmental stage. To identify the developmentally regulated genes in the artesunate-free control culture, the 3 hours RNA was hybridized against the time 0 RNA (3 control experiments). Even within such a short time window, 81 genes showed differential expression in the control culture. Filtering-out this potential confounding "slowing-down effect" established a second list of genes differentially expressed in the artesunate experiments and in the control experiments, but with significantly different expression-ratios under these two conditions (p < 0.05) [see Additional file 2 for the list of differentially expressed genes].

This resulted in the following: out of the 4703 genes analyzed, 398 (approx 8.5%) were differentially expressed after 3 hours in presence of artesunate, 244 over-expressed (approx 5%) and 154 (approx 3.5%) under-expressed. Of these 398 genes differentially expressed at 3 hours, 42 showed significant expression changes at 90 minutes, presumably reflecting an early response to the drug.

Our technically demanding and relatively stringent approach may have generated false negatives, but we wanted to be as confident as possible about the list of genes selected as being differentially expressed. This confidence was supported by the results of qRT-PCR performed with 12 over-expressed and 4 under-expressed genes, which confirmed the microarray data (Spearman correlation coefficient: 0.81) [see Additional file 3 for the comparison between microarray and qRT-PCR results].

Interesting features were observed in the group of differentially expressed genes. For genes expressed differentially at 3 hours, there was a strong positive bias in the sub-set of over-expressed genes towards polymorphic genes located in sub-telomeric position and carrying PEXEL/HCT motifs 13 used for export beyond the parasitophorous vacuole into the erythrocyte cytoplasm, or to the red blood cell membrane/surface (Figure 1). This was determined after excluding from the analysis the strain-specific var, stevor and rifin multigene families – which are not adequately represented in our case since the parasite strain used differs from 3D7. Concomitant with this positive bias, there was a negative bias towards conserved genes in the over-expressed genes. In contrast, the group of under-expressed genes displayed a positive bias towards conserved genes and a negative bias towards polymorphic genes.

Functions (in most cases putative) were assigned to 213 of the 398 differentially expressed genes 121415161718192021 [see Additional file 2]. Apart from lipid and purine/pyrimidine metabolism, few genes belonged to the metabolic pathways of the parasite (as defined by Ginsburg 16). Most genes with altered expression under artesunate pressure were related to chaperones, transporters, cell cycle, kinases, Zn finger proteins, transcription activating proteins, proteins involved in proteasome degradation, oxidative stress and in cell cycle regulation. A positive bias towards over-expression could be observed for genes related to oxidative stress, kinases, transcription associated proteins and chaperones, and towards under-expression for genes related to proteasome degradation and transporters (Figure 2). Slowing down of parasite development may be linked to the differential expression of three genes involved in cell cycle regulation [see Additional file 4]. Two genes were over-expressed, namely PFF1125 encoding a putative RNA binding protein mei2 homologue and PFL1855 encoding a putative cell cycle control protein, while PFA0345w, encoding a putative centrin, was under-expressed.

Oxidative stress is likely to occur in parasites submitted to artesunate, whether linked to the direct action of the drug or to the reduced capacity of the damaged parasite to inactivate free radicals. The antioxidant defense in P. falciparum involves both the glutathione and the thioredoxin system, the relative contribution of each remaining unknown 22. Four genes involved in the parasite antioxidant defense 2223, showed altered transcription profiles (Table 1) : gamma glutamyl cystein synthetase and glutathione synthetase (both involved in the glutathione system) were over-expressed. In contrast, the genes encoding protein disulfide isomerase related protein (potentially involved in oxidative protein folding, catalyzing both the oxidation and isomerization of disulfides on nascent polypeptides 24) and glutathione peroxidase (shown to rather encode a thioredoxin peroxidase 25 and thus involved in the thioredoxin system) were under-expressed. The consecutive action of gamma glutamyl synthetase and glutathione synthetase leads to the production of reduced glutathione (GSH), which plays a pivotal role in the antioxidant defense through maintenance of the red-ox state of protein -SH moeties, the reduction of the noxious hydrogen and lipid peroxides and the extrusion of toxic compounds (including drugs) 22. The predicted outcome of the over-expression of these 2 genes would be an increased supply of GSH in response to the oxidative stress induced by artesunate.

Genome mining on parasite chaperone analogues has led to the conclusion that approximately 2% of the P. falciparum genes encoded potential or confirmed chaperones 19. Chaperone proteins have diverse functions, such as serving as transcription factors, regulating the cell cycle, being involved in protein degradation, binding damaged proteins, protein transport across membranes and protein folding. As expected in a parasite under severe injury, 13 genes encoding chaperone or chaperone-related proteins (Table 2) were found up-regulated at 3 hours of artesunate pressure, among which PfHsp70, PfHsp90 and 6 genes belonging to the Hsp40 chaperone machinery of P. falciparum 20. The first four were already over-expressed at 90 minutes, along with PFD0080c (a member of the resa-like DnaJ family) and thus can be considered part of the early response to the drug. Interestingly, most (9/13), chaperone-encoding genes up-regulated under artesunate have potential export signals, suggesting possible export of their corresponding proteins from the parasite to the erythrocyte cytoplasm or membrane. Chaperone involvement in trafficking of membrane or exported proteins has been highlighted in a recent analysis of the parasite chaperone network 21. In particular, analysis of the interactome suggested that a member of the Hsp40 family (PF14_0700), which interacts with PfHsp70 (PF08_0054) and PfHsp90 (PF07_0029) could function as a co-chaperone in the Hsp90 complex of the parasite to transport eythrocyte membrane exported proteins such as Antigen 332 (PF11_0507). It is remarkable to note that PF14_0700, PF08_0054, PF07_0029 and PF11_0507 were all over-expressed under artesunate pressure [see Additional file 2]. This suggests artesunate-triggered altered intracellular trafficking, contributing to remodeling of the parasitized erythrocyte membrane, a conclusion further substantiated by the observed increased steady state levels of genes encoding proteins interacting with the erythrocyte cytoskeleton such as MESA (PFE0040c), several members of the RESA family and number of genes carrying PEXEL/HT motifs. Alterations of the intracellular trafficking is also suggested by altered transcription profiles of genes such as signal peptidase (MAL13P1.167) as well as transporters, including organellar importers (see below).

Eleven genes involved in transport were under-expressed (Table 3). The putative UDP galactose antiporter (PF11_0141) down-regulated at 3 hours, was the only transporter found down-regulated at 90 minutes, suggesting that alteration of its transcript levels is an early event. Four subunits of the vacuolar ATP synthase were under-expressed. This constitutes one of the few cases of concommittant regulation of individual components of a specific machinery or pathway in this study, suggesting that indeed the activity of the vacuolar ATP synthase was severely impaired. This enzyme contributes to acidification of vacuolar and organelle contents.

Only 4 genes encoding transporters were over-expressed, amongst which pfmdr1 (PFE1150w), the P-glycoprotein homologue of P. falciparum, which modulates susceptibility to antimalarial drugs such as quinine, mefloquine and artemisinins 262728. An association has been shown in the field between sensitivity to arylaminoalcohols or endoperoxides, pfmdr1 allelic status and gene copy number 29, suggesting that an over-expressed functional pfmdr1 conferred a multidrug resistance like phenotype 3031. Over-expression of pfmdr1 we show to occur under artesunate pressure may have an impact on susceptibility to the drug(s) associated to artemisinins upon ACT administration. It also suggests that not only gene copy number but also gene transcription rates and mRNA stability should be monitored when exploring mechanisms of field parasite drug resistance.

Eight genes involved in lipid metabolism were differentially expressed (Table 4).

P. falciparum contains an unusually high number of Acyl CoA synthases (ACS) and binding proteins that might play a role in fatty acid salvage and transport from the host cell 3233. Amongst the 13 different ACS genes, a non subtelomeric acyl CoA synthetase gene PFB0685c has expanded into a family of 9 duplicated genes mainly located in the subtelomeric regions of the genome and transcribed during the erythrocytic stage of the parasite. At least two ACS proteins PfACS1 (PF14_0761) and PfACS3 (PFL2570w) were shown to interact with ankyrin through their lysine-rich C-terminal domain 34. Similar domains are only present in the genes of this family, possibly targeting these enzymes to the erythrocyte cytoplasm and/or membrane, an interesting possibility as all the ACS proteins have a signal peptide but are devoid of the PEXEL/HT motif. This would contribute to the capacity of P. falciparum to activate fatty acid scavenged from the plasma or the erythrocyte membrane, a critical metabolic function for the survival of the parasite. All 4 ACS genes over-expressed under artesunate (PFB0685c, PFB695c, MAL13P1_485 and PF14_0761) belong to this family.

Two acyl-CoA binding protein genes, PF08_0099 and PF10_0016, were under-expressed. PF14_0664, an Acetyl CoA carboxylase which generates Malonyl-CoA, likely to be located in the apicoplast 32, was already found over-expressed at 90 minutes. Interestingly, the triose phosphate transporter located in the outermost membrane of the apicoplast, PFE0410w, was under-expressed under artesunate. This transporter, also called PFoTPT, fuels the apicoplast by providing carbon, reducing power and ATP 35, but also quite importantly, dihydroxyacetone phosphate (DHAP), the precursor needed for isoprenoid and for phospholipids biosynthesis 36. Altogether these data suggest an alteration of fatty acid metabolism both intracellularly and in the apicoplast.

The main metabolic function allowed by the active mitochondrial electron transport chain maintained by Plasmodium is regeneration of ubiquinone, which is required as the electron acceptor for dihydroorotate dehydrogenase, an essential enzyme for pyrimidine biosynthesis 37. Indeed, 3 important genes involved in mitochondrial electron transport are under-expressed in presence of artesunate: 2 subunits of the cytochrome c oxydase (downstream of the cytochrome bc1 complex inhibited by atovaquone) coxI and coI and ubiquinol-cytochrome c reductase hinge protein, PF14_0248 (Table 5). This interference with the electron transport chain could in turn affect pyrimidine biosynthesis.

PFC0975c encoding a peptidyl-prolyl cis-trans isomerase cyclophilin-type belonging to the chaperone network of the mitochondrial matrix 16 is under-expressed in response to artesunate. Cyclophilin D/peptidylprolyl isomerase activity is a regulator of mitochondrial permeability transition, a non selective inner-membrane permeabilization occuring in response to increased calcium load and redox stress. These 2 conditions are induced by artesunate. Such a phenomenon can lead to necrosis through activation of phospholipases, proteases and nucleases 38.

In addition to the possibly impaired respiratory chain interfering with pyrimidine synthesis, expression of other genes related to purine/pyrimidine biosynthesis was also modified by artesunate pressure (Table 6): cytidine and deoxycytidylate deaminase (PF13_0259) hypoxanthine phophorybosyl transferase (PF10_0121), deoxyuridine 5'triphosphate nucleotidohydrolase (PF11_0282) and uridine phosphorylase (PFE0660c) are all under-expressed, while dihydroorotase (PF14_0697) is over-expressed. Interference with such key metabolic processes may readily cause parasite death.

The kinome of P. falciparum is atypical, with an atypical MAP kinase family and a particular R45-FIKK kinase family 39.

Fourteen genes encoding kinases, some of which potentially related to signal transduction, were differentially expressed in presence of artesunate (Table 7). Four of the 11 over-expressed genes were already differentially expressed at 90 minutes, among which the calcium calmodulin dependent protein kinase 2, PfPK2 (PFL1885c). PfPK2 belongs to the CamK group of P. falciparum kinases, a family of kinases which can be activated in response to increased calcium levels, possibly resulting from specific inhibition of the SERCA PfATPase6 by the drug 40. The other 3 genes with increased expression levels at 90 minutes were members of the R45-FIKK multigene family 1841. This Apicomplexa-specific kinase gene family is composed of 20 genes all presenting sub-telomeric positions, scattered over 11 chromosomes. It has recently been shown that these kinases localize in different compartments of the infected erythrocyte, some being associated with the erythrocyte membrane 42. Under artesunate pressure, 3 members of the R45-FIKK family were over-expressed at 90 minutes (PFI0095c, MAL7P1.144 and PFL0040c) and remained at elevated levels subsequently, while two additional R45-FIKK Kinases displayed increased transcript levels at 3 hours, namely PFD1165w and PF11_0510. All 5 present export motifs. PFD1165w and PFL0040c are associated with Maurer's clefts, and PFL0040c with the infected erythrocyte membrane as well 42.

Transcriptome studies of parasites during the erythrocytic stages of development have shown that in many cases, transcription is soon followed by translation, when the parasite "needs" the considered proteins to further develop 4. In contrast, comparison between transcriptomic and proteomic studies have shown the existence of discrepancies in RNA and protein expression for about 50% of the genes expressed during the erythrocytic development 4344. This has been clearly confirmed for certain genes, such as var genes transcribed at the early ring stage and translated several hours later 45 or a series of sexual stage proteins the corresponding genes of which are transcribed long before translation occurs 46. Genome mining revealed that relatively few genes (156, 2/3 less than expected) encode transcription associated proteins 17. Six non-Zn finger transcription associated protein genes, as defined by Coulson 17, were over-expressed under artesunate, 3 were under-expressed (Table 8). This relatively low number suggests that transcriptional regulation may not play a major role in gene-expression modulation under drug-induced stress. Altered decay rates may modulate gene expression in drug-exposed parasites and contribute to the dynamic changes of mRNA levels observed here. In this regard, it is worth noting the recently outlined decreasing RNA decay rate as the parasite matures. A list of genes encoding putative decay components and possibly contributing to such a mechanism of post-transcriptional regulation in P.falciparum has been proposed 47. However, there was no observed effect of artesunate on expression of these genes.

The low number of TAP identified by Coulson 17 contrasted with an over-representation of proteins with Zn finger motifs 48 potentially involved in regulating RNA stability. In the present study, 12 genes identified as putative Zn finger bearing proteins were differentially expressed in the presence of artesunate; 10/12 were over-expressed, two of which (PFI0470w and PFE1245w) already at 90 minutes (Table 9).

The FCR3 strain (FUP/CB line) 62 of P. falciparum was used throughout the experiments. Parasites were cultured using the method described by Trager and Jensen with the modifications described in Ralph et al. 63. Cultures were tightly synchronized by lysing maturing forms through treatment with 0.3 M Alanine at two successive cycles, starting 3 hours after rings appeared in the culture 63. Artesunate treatment was performed the following cycle (see below).

Artesunate powder (Arenco, Belgium) was dissolved in PBS, and the 1 μg/μL aliquoted stock solution was kept at -20°C. To detect possible loss of activity during storage of stock solution, the IC₅₀ of the drug was tested periodically. The IC₅₀ remained stable at around 3.4 nM (2.2–4.6 nM, SD = 1.2). This also confirmed the artesunate sensitivity of the FCR3 strain.

Determination of minimal drug concentration irreversibly damaging 100% of the parasites: The lethal artesunate dose to be used throughout this investigation was determined using synchronized cultures incubated with increasing concentrations of artesunate (30-100-300-500-3000 ng/mL) for 3 hours, washed twice with RPMI and reincubated at 37°C for 24–30 hours, until next reinvasion. Reinvasion was assessed by examining Giemsa stained blood smears. The lethal dose was the minimal artesunate concentration needed to prevent 100% of reinvasion.

Parasite treatment for studies of the transcriptome: artesunate was added at five different time points during the cycle following parasite synchronization, staggered between 20 hours and 30 hours after appearance of the first rings in the culture (at approximately 20, 22, 26, 28 and 30 hours), while untreated cultures were cultured in parallel in the absence of drug. Parasites were extracted in Trizol after 90 minutes and 3 hours of incubation.

Parasite cultures were harvested by centrifugation and lysed in 5 pellet volumes of Trizol (Gibco) before freezing at -80°C. Total RNA was prepared from thawed samples following the manufacturer's instructions. RNA Quality was assessed with an Agilent 2100 Bioanalyser.

The microarrays used here have been described elsewhere 63. Briefly, glass slides were spotted with 7392 70-mer oligonucleotides originating from the Malaria Oligo Set (Qiagen-Operon) and custom design oligonucleotides, covering most of all P. falciparum genes. Of the 5542 identified P. falciparum genes, the array covers 4741 genes (85%). RNA labelling and hybridization were performed as described 63. For the comparison of drug versus no drug at each time point, dye swaps with two technical replicates were performed to compensate dye effect and to assess technical reproducibility, leading to 4 hybridized slides per experiment.

In order to determine the genes differentially expressed between time 0 and 3 hours due to parasite development, three series of hybridizations were performed with cDNAs prepared at times 20/23 hours, 23/26 hours and 25/28 hours from the control cultures left without artesunate. Microarray data were deposited in ArrayExpress [ArrayExpress:E-MEXP-1435].

All the slides were analyzed pooled together using R 64 and Bioconductor 65 with the limma package 6667. After logarithm transformation of ratio of median of the intensities (without background substraction) in the two channels, data were weighted to penalize spots of poor quality, flagged, or spots for which intensities in the two channels were close to background (median of intensity less than median of background plus 2 standard deviations for both Cy3 and Cy5 signals). In a second step, an intensity-dependent normalization was applied to each slide (loess), followed by quantile normalization 68 applied across the arrays to ensure that the log ratios had the same empirical distribution across arrays and across channels. After fitting a linear model to the expression data for each probe using the least square method, comparisons of interest were extracted with a contrast matrix. A moderated t-statistic and a log-odds of differential expression were computed for each contrast for each gene. Finally, the raw p-values were adjusted using the Holm procedure and a type I error rate of 0.05 was applied. When for a given gene significant log ratios were obtained for several oligos, the retained log ratio was one with the maximum absolute value.

RNA samples were treated with DNAse I (Invitrogen) for 15 minutes at 37°C in presence of RNAseOUT™ inhibitor (Invitrogen). RNA was reverse transcribed with Superscript II RT (Invitrogen) using random hexamers with pre-extension step of 10 minutes at 25°C followed by extension for 50 minutes at 42°C. Then, cDNA was treated with RNAse H for 20 minutes at 37°C.

Quantitative real time PCR was performed on cDNA using an ABI Prism 7900 HT thermal cycler system (Applied Biosystems) for 40 cycles (95°C for 15 s, 55°C for 15 s and 60°C for 45 s). Reactions were done in 20 μL volumes using SYBR Green PCR master Mix (Applied Biosystems) and 0.5 μM primers in triplicate with 6 concentrations of cDNA (2500 pg/μL, 625 pg/μL, 156 pg/μL, 39 pg/μL and 10 pg/μL) to assess primer amplification efficiency. Absence of DNA contamination was check using RNA sample treated as cDNA without RT (2500 pg/μL).

Primers were designed using eprimer3 69 [See Additional file 5 for the primer sequences]. Transcript abundance was compared using mean of ΔΔCt values calculated for all cDNA dilutions using PFI0425w (putative transporter) as endogenous normalizer and the control condition as reference. This gene was chosen as it remained non differentially expressed in all our experiments and showed very moderate levels of variation in whole-cycle transcriptome analysis 4.