Epigenetic switches in clag3 genes mediate blasticidin S resistance in malaria parasites

Authors

  • Sofía Mira-Martínez,

    1. Barcelona Centre for International Health Research (CRESIB, Hospital Clínic-Universitat de Barcelona), Barcelona, Catalonia, Spain
    Search for more papers by this author
  • Núria Rovira-Graells,

    1. Barcelona Centre for International Health Research (CRESIB, Hospital Clínic-Universitat de Barcelona), Barcelona, Catalonia, Spain
    Search for more papers by this author
  • Valerie M. Crowley,

    1. Institute for Research in Biomedicine (IRB), Barcelona, Catalonia, Spain
    Current affiliation:
    1. Lewis-Sigler Institute for Integrative Genomics, Princeton University, Princeton, NJ, USA
    Search for more papers by this author
  • Lindsey M. Altenhofen,

    1. Department of Molecular Biology and Lewis-Sigler Institute for Integrative Genomics, Princeton University, Princeton, NJ, USA
    Search for more papers by this author
  • Manuel Llinás,

    1. Department of Molecular Biology and Lewis-Sigler Institute for Integrative Genomics, Princeton University, Princeton, NJ, USA
    Search for more papers by this author
  • Alfred Cortés

    Corresponding author
    1. Catalan Institution for Research and Advanced Studies (ICREA), Barcelona, Catalonia, Spain
    • Barcelona Centre for International Health Research (CRESIB, Hospital Clínic-Universitat de Barcelona), Barcelona, Catalonia, Spain
    Search for more papers by this author

For correspondence. E-mail alfred.cortes@cresib.cat; Tel. (+34) 93 2275400; Fax (+34) 93 3129410.

Summary

Malaria parasites induce changes in the permeability of the infected erythrocyte membrane to numerous solutes, including toxic compounds. In Plasmodium falciparum, this is mainly mediated by PSAC, a broad-selectivity channel that requires the product of parasite clag3 genes for its activity. The two paralogous clag3 genes, clag3.1 and clag3.2, can be silenced by epigenetic mechanisms and show mutually exclusive expression. Here we show that resistance to the antibiotic blasticidin S (BSD) is associated with switches in the expression of these genes that result in altered solute uptake. Low concentrations of the drug selected parasites that switched from clag3.2 to clag3.1 expression, implying that expression of one or the other clag3 gene confers different transport efficiency to PSAC for some solutes. Selection with higher BSD concentrations resulted in simultaneous silencing of both clag3 genes, which severely compromises PSAC formation as demonstrated by blocked uptake of other PSAC substrates. Changes in the expression of clag3 genes were not accompanied by large genetic rearrangements or mutations at the clag3 loci or elsewhere in the genome. These resultsdemonstrate that malaria parasites can become resistant to toxic compounds such as drugs by epigenetic switches in the expression of genes necessary for the formation of solute channels.

Introduction

Plasmodium spp. parasites have a complex life cycle that includes several niches in two different hosts, humans and mosquitoes, but clinical symptoms of malaria disease are almost exclusively associated with cycles of asexual replication inside human erythrocytes. Intracellular parasitism has obvious advantages for many organisms, but it also poses important challenges. In the case of malaria asexual blood stages, the intraerythrocytic niche protects the parasite from immune attack, but this life style also implies that the parasite must develop a transport system to acquire nutrients that are not available inside the erythrocyte. It is well established that the membrane of erythrocytes infected with mature stages of P. falciparum (pigmented trophozoite and schizont stages) is permeable to numerous solutes that are not transported into non-infected erythrocytes, including ions and organic compounds such as sugars and amino acids, among many others. These new transport activities are collectively referred to as the new permeation pathways (NPPs) (Elford et al., 1985; Ginsburg et al., 1985; Saliba and Kirk, 2001; Desai, 2012).

The precise nature of the channel(s) that mediate solute uptake in infected erythrocytes remains unresolved (Staines et al., 2007; Desai, 2012). Some authors postulated that a single channel type mediates NPPs (Kirk et al., 1994; Desai et al., 2000; Alkhalil et al., 2004) and that this channel is likely to be parasite encoded (Alkhalil et al., 2004; Baumeister et al., 2006). This proposed channel, termed plasmodial surface anion channel (PSAC), combines transport of a broad spectrum of solutes with selectivity for structurally similar solutes (Desai et al., 2000; Alkhalil et al., 2004; Hill and Desai, 2010; Desai, 2012). However, others have suggested that several distinct channels with different electrophysiological properties contribute to the increased permeability of infected erythrocytes. These channels are mainly host-encoded endogenous erythrocyte channels activated by the parasite (Staines et al., 2007; Bouyer et al., 2011; Winterberg et al., 2012). Regardless of whether the actual channel is encoded by the parasite or one or more endogenous channels are activated, a recent study using chemical inhibitors, linkage analysis of a genetic cross and transgenic parasite approaches unambiguously demonstrated that the parasite encoded proteins CLAG3.1 and CLAG3.2 (also named RhopH1/Clag3.1 and RhopH1/Clag3.2, ID PF3D7_0302500 and PF3D7_0302200) play a key role in the formation of active PSAC and are essential for the transport of numerous solutes into infected erythrocytes (Nguitragool et al., 2011). These proteins may constitute the actual channel, be part of the channel, or activate a channel formed by other proteins (Nguitragool et al., 2011; Desai, 2012). Follow up studies have demonstrated that CLAG3 proteins and PSAC play a critical role in nutrient acquisition and are essential for efficient parasite growth when the medium contains physiological concentrations of key nutrients (Pillai et al., 2012).

CLAG3 proteins are encoded by members of the clag gene family, which in P. falciparum consists of 5 different genes. The two clag3 genes (clag3.1 and clag3.2) are separated by only 10kb and share over 95% nucleotide sequence identity, whereas other members of the family (clag2, clag8 and clag9) are more distantly related (Iriko et al., 2008). It remains unknown whether CLAG proteins other than CLAG3 participate in the formation of active PSAC. Of note, clag3 genes show mutually exclusive expression, such that an individual parasite expresses only one of the two genes at a time. Initially described in parasites of 3D7 and HB3 genetic backgrounds (Cortés et al., 2007), mutually exclusive expression of clag3 genes has been later confirmed by different laboratories in parasites of 3D7 genetic background (Comeaux et al., 2011; Crowley et al., 2011) and also in other genetic backgrounds (Nguitragool et al., 2011; Pillai et al., 2012). Together with var, clag3 genes represent the only known example of this type of expression in malaria parasites (Guizetti and Scherf, 2013). The active or repressed state of clag3 genes is regulated at the chromatin level, and clonally transmitted over several generations of asexual growth by epigenetic mechanisms (Cortés et al., 2007; Comeaux et al., 2011; Crowley et al., 2011). Transitions between the two states occur, albeit at low frequency, resulting in switches from expression of one clag3 gene to expression of the other (Cortés et al., 2007; Comeaux et al., 2011; Crowley et al., 2011; Nguitragool et al., 2011; Pillai et al., 2012).

Among the solutes that require NPPs for transport across the infected erythrocyte membrane are some compounds that are toxic for the parasite, including diamidine compounds (Stead et al., 2001), bis-quaternary ammonium compounds (Biagini et al., 2003), the antibiotics fosmidomycin (Baumeister et al., 2011) and BSD (Hill et al., 2007), and the protease leupeptin (Lisk et al., 2008). Interestingly, P. falciparum parasites can acquire resistance to the antimalarial compounds BSD and leupeptin by alterations in PSAC activity, providing support to the idea that PSAC is encoded by the parasite (Hill et al., 2007; Lisk et al., 2008; 2010; Hill and Desai, 2010). These results establish that malaria parasites can acquire drug resistance by changes in the permeability of the infected erythrocyte membrane. While a mutation in clag3.2 associated with leupeptin resistance has been identified (Nguitragool et al., 2011), no mutation associated with BSD resistance has been identified to date. Here we demonstrate that malaria parasites can acquire resistance to BSD by epigenetic changes in the expression of clag3 genes. Parasites acquire resistance to low concentrations of the drug by switching from clag3.2 to clag3.1 expression, whereas resistance to higher drug concentrations involves simultaneous epigenetic silencing of both clag3 genes, an unexpected expression pattern that had not been previously described. Our results imply that expression of alternative clag3 genes results in different transport efficiency of PSAC and add epigenetic alterations to the list of mechanisms by which malaria parasites can become resistant to a drug.

Results

Resistance to BSD is associated with changes in clag3 expression

As part of our ongoing investigations on the rules that govern the mutually exclusive expression of clag3 genes (N. Rovira-Graells, V.M. Crowley and A. Cortés, unpublished), we transfected P. falciparum parasites with the plasmid 3.2-1371-LH-bsdR, which contains a BSD resistance cassette (BSD deaminase gene under the control of a constitutive promoter) and the clag3.2 upstream sequence driving the expression of a luciferase gene reporter (Fig. S1A). Transfected parasites were selected with 2.5 μg ml−1 of BSD to obtain a population of parasites stably maintaining the plasmid as an episome. For these experiments we used the 3D7 subclone 10G (Cortés, 2005), which predominantly expresses clag3.2 and has clag3.1 silenced (Cortés et al., 2007; Crowley et al., 2011; Rovira-Graells et al., 2012). Unexpectedly, we found that in the transfected and BSD-selected population the predominantly expressed endogenous clag3 gene switched from clag3.2 to clag3.1 (Fig. S1B). To determine whether the switch was attributable to the episomal clag3.2 promoter or it was related with BSD selection, we transfected 10G parasites with the BsdR plasmid, which contains the BSD resistance cassette but no gene reporter or clag3 promoter (Fig. S1A). Upon selection of transfected parasites with BSD, expression of endogenous clag3 genes was assessed at different times after transfection. Similar to the results with 3.2-1371-LH-bsdR, BsdR-transfected parasites progressively switched from clag3.2 to clag3.1 expression (Fig. 1A). This switch was not observed in untransfected 10G parasites grown in parallel. These results indicate that BSD selection of transfected parasites can result in switches in the expression of clag3 genes.

Figure 1.

Resistance to BSD is associated with changes in clag3 expression.

A. Relative abundance of clag3.1 and clag3.2 transcripts in untransfected 10G parasite line or the same line transfected with the plasmid BsdR and maintained under BSD selective pressure. Values are the log2 of the expression ratio.

B. Transcript levels of clag3.1 and clag3.2 in unselected control 10G cultures at different times along the experiment (in weeks), in cultures selected with BSD at the concentrations indicated (in μg ml−1) for 2 weeks (0.2 and 0.3) or 4 weeks (0.4 and 0.6), and in cultures maintained for two weeks in the absence of drug after selection with the concentration of BSD indicated. The culture 0.6–2 was sequentially selected with 0.6 μg ml−1 BSD for 4 weeks and then with 2 μg ml−1 for 2 weeks. Transcript levels are normalized against rhoph2, which has a similar time of expression to clag3 genes along the asexual cycle. Values are the average of reactions performed in triplicate, with SD.

C. Relative transcript levels of clag3.1 and clag3.2, expressed as the log2 of the expression ratios, in the same samples described in panel B.

D. Schematic representation of predominant clag3 expression patterns in the parasite lines used in this study. An arrow indicates an active state whereas a cross indicates silencing. Low BSD refers to ≤ 0.4 μg ml−1 whereas high BSD refers to ≥ 0.6 μg ml−1. The schematic is based on data from this study except for 1.2B (Cortés et al., 2007; Crowley et al., 2011).

To address how BSD affects clag3 expression in the absence of exogenous resistance markers, we selected untransfected 10G parasites with different concentrations of BSD ranging from approximately the BSD IC50 in 10G (see below) to a threefold higher concentration. In only 1 week (3–4 generations) we obtained parasite populations adapted to the two lower concentrations (0.2 and 0.3 μg ml−1, 10G-0.2 and 10G-0.3 lines respectively), and in about 2–3 weeks we also obtained cultures adapted to 0.4 and 0.6 μg ml−1 BSD (10G-0.4 and 10G-0.6 lines respectively). Expression patterns of clag3 genes in untreated control 10G cultures remained stable throughout the experiment (Fig. 1B). In contrast, in cultures selected with 0.2 to 0.4 μg ml−1 BSD the majority of parasites expressed clag3.1 instead of clag3.2 (Fig. 1B), similar to the observations with transfected parasites. The ratio of clag3.1 vs clag3.2 expression changed over 100-fold (Fig. 1C). This result indicates that BSD selects for parasites expressing CLAG3.1 rather than the alternative CLAG3.2, implying that in this parasite line CLAG3.2 mediates more efficient uptake of BSD than CLAG3.1.

Selection with 0.6 μg ml−1 BSD not only altered the ratio of clag3.1-to-clag3.2 expression but also resulted in reduced expression of the two clag3 genes (Figs 1B, C and S2A). Next we challenged parasites adapted to grow under 0.6 μg ml−1 BSD with 2 μg ml−1 BSD for 2 weeks. 10G-0.6 did not require an additional adaptation period and was directly able to grow under this high drug concentration (10G-0.6-2 line). Expression of clag3 genes was almost completely abolished in 10G-0.6-2 parasites (40-fold lower clag3 expression than in control cultures, Figs 1B and S2A). All together, these results indicate that switching from clag3.2 expression to clag3.1 expression limits the entry of BSD at low concentrations, but resistance to higher BSD concentrations requires severely reduced expression of the two clag3 genes, presumably compromising PSAC formation (Fig. 1D).

Stability of clag3 expression patterns in the absence of BSD pressure

We maintained the cultures adapted to the different BSD concentrations in the absence of drug for 2 weeks and measured their clag3 expression. The clag3 expression patterns of 10G-0.2, 10G-0.3 and 10G-0.4 did not change (Fig. 1B), in line with previous observations showing that clag3.1 expression, as well as clag3.2 expression, are stably transmitted expression patterns under culture conditions (Cortés et al., 2007). On the other hand, 10G-0.6 parasites grown in the absence of drug resumed normal clag3 expression (Figs 1B and S2A), indicating that severely reduced clag3 expression imposes a growth disadvantage for the parasites. This idea is supported by the observation of a lower growth rate (measured in the absence of drug) in lines selected with higher drug concentrations (growth rate 5.33 ± 1.58 for 10G-0.6-2 compared with 8.99 ± 1.26 for unselected 10G, P = 0.01, n ≥ 4).

clag3 expression patterns are associated with changes in BSD sensitivity

We compared BSD growth inhibition between the original 10G and the BSD-selected lines and found that BSD sensitivity decreased with increasing concentrations of drug used for selection (Fig. 2A and Table 1). Parasites selected with 0.2 to 0.4 μg ml−1 BSD showed progressive modest increases in their BSD IC50. However, in 10G-0.4 substantial growth (almost 10% of growth in untreated cultures) was observed even at the highest BSD concentration tested (2.5 μg ml−1). This may correspond to a subpopulation of highly resistant parasites in this selected culture. 10G-0.6 was highly resistant to the drug and 10G-0.6-2 was essentially insensitive even to high concentrations of BSD (Fig. 2A and Table 1), indicating that silencing of both clag3 genes severely blocks the entry of this drug. Next we tested the BSD sensitivity of the 3D7 subclone 1.2B (Cortés, 2005), which is isogenic with 10G but spontaneously expresses clag3.1 rather than clag3.2 without having ever been selected with BSD (Cortés et al., 2007; Crowley et al., 2011). The subclone 1.2B was more resistant to BSD than 10G (Fig. 2B and Table 1). BSD susceptibility of 1.2B was similar to low BSD-selected 10G lines (10G-0.2 and 10G-0.3) that show the same clag3 expression pattern as 1.2B (Figs 1D and 2 and Table 1). This result supports the idea that clag3 expression, rather than other potential alterations arising during selection, is the main determinant of BSD sensitivity.

Figure 2.

BSD dose–response curves for parasite lines with different clag3 expression patterns.

A. BSD susceptibility of 10G-derived parasite lines previously selected with different concentrations of the drug. Values are the result of a representative experiment performed in duplicate, with range.

B. Comparison of BSD susceptibility between the isogenic 3D7 subclones 10G and 1.2B, which spontaneously show different clag3 expression patterns without having been selected with BSD. Values are the average of two independent experiments performed in duplicate, with range.

Table 1. BSD IC50 of parasite lines used in this study
Parasite lineIC50 (in μg ml−1)
  1. Values are the average of two independent experiments performed in duplicate, with range.
10G0.20 (0.20–0.20)
10G-0.20.34 (0.32–0.35)
10G-0.30.43 (0.38–0.47)
10G-0.40.79 (0.69–0.88)
10G-0.61.64 (1.54–1.73)
10G-0.6-2> 5 (> 5)
1.2B0.31 (0.27–0.35)

10G-0.6 cultures maintained in the absence of drug, which recovered normal clag3 expression levels, only regained BSD sensitivity partially: growth of 10G-0.6 maintained for 3–4 weeks without drug was 61% inhibited by 1 μg ml−1 of BSD (range 59–62%, n = 2), which is higher than the inhibition of 10G-0.6 tested immediately after selection (34%, range 28–40%, n = 2) but similar to inhibition of 10G-0.4 (62%, range 55–69%, n = 2) that expresses normal clag3 levels (clag3.1). However, growth of unselected 10G was invariably > 99% inhibited by this BSD concentration. These results indicate that a proportion of high-BSD resistant parasites remain in the 10G-0.6 population after several weeks without drug.

Expression of other clag genes in BSD-selected parasites

The participation of clag genes other than clag3 in the formation of PSAC has not been determined. We measured expression of clag2, clag8 and clag9 in all BSD-selected cultures, and only clag2 expression was reduced by selection with BSD, although expression levels did not correlate with the BSD concentration used for selection (Fig. S2A). However, clag2 is the only member of the clag family besides clag3.1 and clag3.2 that shows clonally variant expression, and it is silenced in the 10G subclone (Cortés et al., 2007; Rovira-Graells et al., 2012) (note the different scale in the clag2 panel in Fig. S2A, showing residual clag2 expression only). To further assess the possible association between clag2 silencing and BSD resistance, we selected with 0.3 or 0.6 μg ml−1 BSD the parasite subclone 1.2B, which is isogenic with 10G but has the clag2 gene in an active state (Cortés et al., 2007; Rovira-Graells et al., 2012). clag2 transcript levels were similar between control and BSD-selected 1.2B (Fig. S2B). Hence, since parasites can become resistant to BSD while expressing clag2, clag8 and clag9, we conclude that these genes are unlikely to play a major role in the transport of BSD into infected erythrocytes in parasites of 3D7 genetic background, at least when the drug is present at moderate concentrations.

Resistance to BSD is paralleled by resistance to sorbitol lysis

Sorbitol transport into infected erythrocytes, which results in haemolysis, requires PSAC activity (Wagner et al., 2003; Nguitragool et al., 2011). Ring-stage infected erythrocytes are not lysed by sorbitol because functional PSAC is assembled at the onset of the pigmented trophozoite stage. We treated magnet-synchronized late stages (pigmented trophozoites and schizonts) of control and BSD-selected cultures with 5% sorbitol. The proportion of sorbitol resistant parasites was low in 10G-0.2 and 10G-0.3, but it was 14% in 10G-0.4 (Fig. 3). This is consistent with the existence of a high BSD-resistant subpopulation in 10G-0.4 (Fig. 2A). Strikingly, over 60% of 10G-0.6 and over 70% of 10G-0.6-2 late stages were resistant to sorbitol lysis (Fig. 3), consistent with the large reduction in clag3 expression observed in these parasite lines. Parasites selected with a high BSD concentration were also highly resistant to haemolysis of infected erythrocytes by a structurally unrelated PSAC substrate, the amino acid l-Alanine (data not shown). These results confirm impaired PSAC function in these parasites and the link between BSD resistance, clag3 expression, NPPs and PSAC (Fig. 4). When 10G-0.6 parasites were maintained in the absence of BSD, they progressively lost resistance to sorbitol lysis (data not shown). By 4 weeks after drug removal, less than 20% of late stage parasites were resistant to sorbitol lysis. Together with the moderate increase in BSD sensitivity and the recovery of clag3 expression observed after removing drug pressure, these results indicate that in the absence of BSD parasites with reduced total clag3 expression are selected against. However, this negative selection is slow and a subpopulation of high-BSD and sorbitol resistant parasites remains after several weeks without drug. The expected reduction in clag3 transcript levels associated with the presence of this subpopulation (e.g. 20% reduction) is below the accuracy limits of the transcript quantification method.

Figure 3.

Sensitivity to sorbitol lysis in parasite lines selected with different concentrations of BSD.

A. Values are the proportion of late-stage parasites (pigmented trophozoites and schizonts) resistant to lysis with 5% sorbitol, and are the average of two independent experiments, with range/2.

B. Representative fields of Giemsa-stained smears of 10G and 10G-0.6 treated or not with 5% sorbitol.

Figure 4.

Model for the acquisition of BSD resistance by epigenetic changes in the expression of clag3 genes. The majority of parasites in the unselected 10G population express CLAG3.2 (red cylinder) and not CLAG3.1 (green cylinder). Culturing with BSD at low concentrations (up to 0.4 μg ml−1) selects for pre-existing parasites within the population that had switched from clag3.2 to clag3.1 expression, which results in less efficient BSD transport and allows parasite survival at low BSD concentrations. However, higher concentrations of the drug select parasites with severely reduced clag3 expression, which blocks uptake of the drug even at high concentration and also blocks sorbitol uptake. Whether residual clag3 expression in these parasites allows the formation of a small number of active channels (not shown) is not known. This model represents the predominant expression patterns in each population, but cultures selected with 0.4 or 0.6 μg ml−1 BSD contain a small fraction of individual parasites with the high BSD and low BSD patterns respectively. Removal of drug from cultures selected with low BSD concentrations does not result in additional changes in clag3 expression. However, removal of drug from cultures selected with high BSD concentrations results in recovery of clag3 expression and a concomitant slow increase in sorbitol and BSD sensitivity, indicating that in the absence of drug parasites with both clag3 genes silenced are selected against.

BSD resistance and switches in clag3 expression are not associated with large genetic rearrangements or mutations

The high level of similarity between clag3.1 and clag3.2 (95% identity in the ORF) can result in recombination events between the two genes (Iriko et al., 2008; Pillai et al., 2012). To exclude the possibility that the changes in clag3 expression observed upon BSD selection are mediated by genetic rearrangements at the clag3 loci, we analysed the genomic DNA of the selected lines by long PCR, but found no difference relative to the unselected line (Fig. S3A). We also excluded the possibility of a deletion affecting one or the two clag3 genes, because qPCR analysis of the two genes revealed identical copy number relative to an essential gene between unselected 10G and 10G-0.6 (Fig. S3B). Furthermore, next-generation sequencing (NGS) of the complete genomes of unselected 10G and two of the adapted lines (10G-0.2 and 10G-0.6) did not reveal any genetic differences associated with BSD selection at the clag3 loci (Fig. S4) or at other clag genes (data not shown). These results indicate that changes in clag3 expression associated with BSD resistance are not mediated by major genetic rearrangements or mutations at the clag3 loci. Furthermore, a genome-wide analysis of the NGS data for 10G, 10G-0.2 and 10G-0.6 did not reveal any significant sequence differences associated with BSD selection (Table S2). Considering that we and others have previously demonstrated that expression of clag3 genes is regulated by chromatin-based epigenetic mechanisms (Comeaux et al., 2011; Crowley et al., 2011), we conclude that epigenetic switches in clag3 expression and not mutations are the main determinants of BSD resistance in our adapted lines.

Discussion

The PSAC plays an important role in the biology of P. falciparum by enabling the uptake of solutes that are necessary for parasite growth (Pillai et al., 2012). However, the presence of this broad-specificity channel in the surface of infected erythrocytes also poses a risk for the parasite, as it can allow the entrance of harmful solutes. Here we demonstrate that P. falciparum can evolve resistance to toxic compounds by altering the expression of clag3 genes, which are necessary for the formation of functional PSAC. While parasites acquire resistance to low concentrations of BSD by switching from expression of one clag3 gene to the other, resistance to high concentrations of the drug requires simultaneous silencing of both genes, resulting in severely affected PSAC activity (Fig. 4). Furthermore, the parasite subclone 1.2B, which is isogenic with 10G but spontaneously expresses CLAG3.1 instead of CLAG3.2 without having been exposed to BSD, is less sensitive to the drug than 10G, linking spontaneous switches in clag3 expression with resistance to toxic compounds.

The results presented here have important implications for our understanding of the regulation and function of clag3 genes: first, they imply that CLAG3.1 or CLAG3.2 confer different transport efficiency to PSAC, such that at least for BSD and in a 3D7 genetic background, CLAG3.2 determines more efficient transport than CLAG3.1. Whether CLAG3.2 generally provides more efficient solute transport, or each CLAG3 protein determines higher transport efficiency for different solutes, remains unknown. Of note, clag3.1 and clag3.2 are highly polymorphic and gene conversion events are common between these genes (Iriko et al., 2008), implying that CLAG3.1 and CLAG3.2 properties likely vary between parasites of different genetic backgrounds. A second important implication of our results is that mutually exclusive expression of clag3 genes, such that in an individual parasite one and only one of the clag3 promoters is active, is not strict and parasites with an unusual expression pattern can be detected under strong selective pressure. In contrast to previous studies that invariably showed one active clag3 promoter in all parasites studied (Cortés et al., 2007; Comeaux et al., 2011; Crowley et al., 2011; Nguitragool et al., 2011; Pillai et al., 2012), here we describe a parasite line with the two clag3 genes simultaneously silenced (10G-0.6-2). A previous study described a transgenic parasite line with a truncated clag3.2 gene that did not express full-length clag3 transcripts, but one of the clag3 promoters was active even if it was controlling a truncated gene (Comeaux et al., 2011). The molecular mechanisms underlying mutually exclusive expression in malaria parasites remain a mystery (Guizetti and Scherf, 2013), but in the case of clag3 genes it is possible that these mechanisms do not actively prevent simultaneous silencing of the two genes. Instead, the fitness cost associated with this clag3 transcriptional state would keep the proportion of parasites that harbour it at very low levels within a population, making them detectable only under selective conditions that confer them a growth advantage. Our observation that parasites with severely reduced clag3 expression regain normal clag3 expression levels when cultured in the absence of drug, and the substantially lower growth rate of 10G-0.6-2 compared with unselected 10G, support this idea (Fig. 4). However, parasites with severely reduced CLAG3 levels still grow, albeit with reduced growth rate; whether residual expression of clag3 genes when they are epigenetically silenced allows the formation of a small number of active channels sufficient for the uptake of some essential solutes is not known. Studies using CLAG3 inhibitors indicate that an essential role for CLAG3 is only revealed when using modified media with restricted concentrations of key nutrients (Pillai et al., 2012). A third important consideration is that our results provide a plausible explanation for the function of mutually exclusive expression in clag3 genes. We propose that mutually exclusive expression of these genes plays a role in preventing access of undesired compounds present in the host plasma. In this scenario, changing metabolic or pharmacological conditions of the host would select parasites with clag3 expression patterns that confer the best balance between efficient transport of necessary solutes and reduced permeability to toxic compounds. In the case of var genes, mutual exclusion is driven by immune evasion (Scherf et al., 2008; Deitsch et al., 2009), but given that mutual exclusion of clag3 genes involves only two genes, we consider immune evasion a less likely driving force in their case.

Drug resistance in malaria parasites is usually mediated by genetic alterations, including point mutations or larger genetic alterations in genes encoding the target enzyme or transporters that pump the drug out of its site of action (Goldberg et al., 2012). However, the mechanism of resistance to several drugs remains largely unknown, including resistance to artemisinin derivatives (Cheng et al., 2012). The results of our studies with BSD, a drug that is not clinically used against malaria because of its high toxicity to human cells, provide a proof of principle demonstrating that in addition to genetic changes, parasites can develop drug resistance by transcriptional alterations transmitted by epigenetic mechanisms. Similar to genetic changes, epigenetic alterations are transmissible, but they provide the additional advantage for the parasite that they are easily reversible and hence confer flexibility to rapidly adapt to fluctuating conditions (Cortés et al., 2012; Rovira-Graells et al., 2012). It is currently unknown if other drugs are susceptible to development of resistance by transcriptional changes in clag3 genes. Small hydrophobic compounds are expected to diffuse through lipid membranes and not require a channel to enter infected erythrocytes, suggesting that molecular size and hydrophobicity indexes such as the logP value could in principle be used to predict if a solute requires PSAC for its uptake (Lisk et al., 2010). However, experience has shown that no simple criteria can predict which solutes are transported via NPPs (Baumeister et al., 2011), as even some very hydrophobic compounds are known to use NPPs to enter infected erythrocytes (Stead et al., 2001; Biagini et al., 2003). Hence, only experimental validation will determine which drugs are susceptible to development of resistance by the mechanisms described here. In this regard, the parasite line 10G-0.6-2, which keeps the two clag3 genes silenced, represents a valuable tool that should be incorporated into large drug screening efforts to determine which drug leads are susceptible to this mode of resistance, as 10G-0.6-2 may show reduced sensitivity to any toxic compounds that require PSAC activity to enter infected erythrocytes.

The pattern of adaptation to BSD observed here is consistent with BSD selecting for pre-existing parasites within the population (Fig. 4). Our previous work has established that genetically homogeneous P. falciparum populations are transcriptionally heterogeneous, such that some genes, including clag3.1 and clag3.2, show clonally variant expression as an intrinsic property and are active in some individual parasites and silenced in others (Cortés et al., 2012; Rovira-Graells et al., 2012). 10G is a recently subcloned line, but it is expected that some individual parasites within the population have spontaneously switched from clag3.2 to clag3.1 expression or have silenced both genes after subcloning. We have previously proposed that epigenetic heterogeneity within isogenic parasite populations plays a role in the adaptation of parasites to changes in their environment by allowing selection of parasites with transcriptional patterns that confer more fitness (Cortés et al., 2012; Rovira-Graells et al., 2012). This bet-hedging adaptive strategy operates for immune evasion (Scherf et al., 2008; Deitsch et al., 2009) and for adaptation to periodical heat-shock mimicking cyclical malaria fever (Rovira-Graells et al., 2012), and here we demonstrate that it also plays a role in the adaptation of malaria parasites to the presence of toxic compounds.

CLAG3 proteins determine which plasma solutes access infected erythrocytes, the niche where malaria parasites spend most of their time and produce human disease. In spite of this key role in parasite biology, clag3 expression dynamics in natural infections remains completely unknown. Understanding how clag3 gene expression is altered in response to different host physiological conditions during the course of natural infections and determining to which drugs the parasite can develop resistance by changes in clag3 expression should be considered urgent research priorities.

Experimental procedures

Parasite cultures

10G and 1.2B subclones of the 3D7-A stock of the clonal P. falciparum line 3D7 have been described and characterized before (Cortés, 2005; Cortés et al., 2007; Crowley et al., 2011; Rovira-Graells et al., 2012). Parasites were cultured in B+ erythrocytes at a 3% haematocrit under standard conditions, with Albumax II and no human serum. Cultures were synchronized by treatment with 5% sorbitol (unselected cultures or cultures selected with up to 0.3 μg ml−1 BSD) or by magnetic separation using Miltenyi Biotec CS columns (cultures selected with higher BSD concentrations containing an important fraction of parasites refractory to sorbitol synchronization). To prepare RNA for transcriptional analysis, synchronized cultures were harvested when the majority of parasites were at the schizont stage and a small proportion of schizonts had already bursted.

BSD selection, growth inhibition assays and determination of growth rates

BSD (Blasticidin S HCl) was obtained from Invitrogen. To select cultures for BSD resistance, the drug was initially applied to cultures at the ring stage. To determine BSD growth inhibition and IC50, the parasitemia of synchronized ring stage cultures was measured by FACS and adjusted to 1%. In drug selected cultures, drug was removed 4 h before starting the assay. Cultures were grown in the presence of different concentrations of drug in duplicate wells of 96-well plates. After ∼ 53 h, parasitemia was determined on Giemsa-stained smears by microscopy, blind-counting the number of infected erythrocytes in 1000 to 10 000 erythrocytes (depending on parasitemia). All experiments included a BSD dose–response assay with unselected 10G that demonstrated stable drug potency and consistency between assays performed on different dates. After LOG-transforming drug concentrations, data was fit to sigmoidal dose–response curves using GraphPad Prism (version 3) setting the maximum to 100 and the minimum to 0, with no weighting. To calculate growth rates, drug was removed from synchronized cultures at the ring stage 4 h before starting the assay. Parasitemia was adjusted to 1% and measured by FACS (initial parasitemia). After ∼ 53 h, parasitemia was again determined by FACS or microscopy (final parasitemia) and the growth rate calculated as the ratio of final parasitemia/initial parasitemia. Growth rates are expressed as average ± SD and compared using unpaired Student's t-test.

Sorbitol sensitivity assays

To determine sorbitol sensitivity, late stage parasites were purified using magnetic columns. After adjusting parasitemia to approximately 10% by addition of uninfected erythrocytes, or after one additional complete asexual blood cycle, cultures were treated in parallel with 5% sorbitol in H2O or with RPMI-HEPES (control) for 7 min at 37°C. After washing with RPMI-HEPES and resuspending in complete parasite culture medium, Giemsa-stained smears were prepared to determine parasitemia by microscopy. The proportion of sorbitol-resistant parasites was calculated by dividing parasitemia in sorbitol-treated samples by parasitemia in controls.

Genetic and transcriptional analysis

The clag3 loci were analysed by long PCR of gDNA as previously described (Iriko et al., 2008). For RNA purification, culture pellets were collected in Trizol and RNA purified, DNase treated and reverse transcribed as described (Cortés et al., 2007). To exclude gDNA contaminations, parallel reactions were performed in the absence of reverse transcriptase. cDNAs were analysed by quantitative PCR in triplicate wells using PowerSYBR Green Master Mix (Applied Biosystems) as described (Crowley et al., 2011). Expression values, in arbitrary units, were calculated using the standard curve method (each 96-well plate contained an identical standard curve made with serial dilutions of 3D7 gDNA). The primers used are described in Table S1.

Next-generation sequencing

Genomic DNA libraries were prepared for multiplexed single-end Illumina TruSeq sequencing as previously described (Straimer et al., 2012). In brief, gDNA was sheared, size selected on a gel, and purified. NEBNext DNA Library Preparation reagents (NEB) were used to end repair, dA-tail, and ligate on NEXTflex DNA Barcodes (Bioo Scientific). Library quality and ligation efficiency was analysed using an Agilent 2100 Bioanalyzer (Agilent Technologies) and a Quant-iT dsDNA High-Sensitivity Assay Kit (Life Technologies). All libraries were determined to be of sufficient quality and quantity with no need for PCR amplification. The three individually barcoded libraries were multiplexed along with 25% PhiX control DNA and run on a single lane using the Illumina HiSeq 2500 Rapid Run (141 bp) system. For details on data analysis, please see Table S2 and Fig. S4.

Acknowledgements

We thank Cristina Bancells, Carmen Fernández-Becerra and Hernando del Portillo for useful discussion and comments on the manuscript. This work was funded by a Spanish Ministry of Science and Innovation Grant (SAF2010-20111) to AC, and ML receives support from the Centre for Quantitative Biology (P50GM071508).

Note added in proof

During the revision of our manuscript an article was published by the Desai lab that provides additional support for some of our conclusions (Sharma et al., 2013). In that work, transcriptional analysis of FCB clone parasites selected with a high concentration of BSD (2.5 μg ml−1) revealed that these parasites silenced expression of both clag3 genes at the epigenetic level, as we have observed in our lines selected with ≥ 0.6 μg ml−1 of the drug.

Ancillary