An Epigenetic Antimalarial Resistance Mechanism Involving Parasite Genes Linked to Nutrient Uptake*

Background: Malaria parasites acquire antimalarial resistance through incompletely understood mechanisms. Results: Resistance to blasticidin S results from reversible silencing of parasite clag genes through histone modifications without DNA level changes. Conclusion: Sophisticated epigenetic control of clag genes permits regulated control of nutrient and antimalarial transport at the host membrane. Significance: This resistance mechanism allows rapid parasite adaptation to environmental pressures and is worrisome for drug discovery efforts. Acquired antimalarial drug resistance produces treatment failures and has led to periods of global disease resurgence. In Plasmodium falciparum, resistance is known to arise through genome-level changes such as mutations and gene duplications. We now report an epigenetic resistance mechanism involving genes responsible for the plasmodial surface anion channel, a nutrient channel that also transports ions and antimalarial compounds at the host erythrocyte membrane. Two blasticidin S-resistant lines exhibited markedly reduced expression of clag genes linked to channel activity, but had no genome-level changes. Silencing aborted production of the channel protein and was directly responsible for reduced uptake. Silencing affected clag paralogs on two chromosomes and was mediated by specific histone modifications, allowing a rapidly reversible drug resistance phenotype advantageous to the parasite. These findings implicate a novel epigenetic resistance mechanism that involves reduced host cell uptake and is a worrisome liability for water-soluble antimalarial drugs.

Lost efficacy of chloroquine, an inexpensive and safe drug that once cured malaria with a single oral dose, contributed significantly to failed malaria eradication programs in the 1960s (1). Resistance to subsequently deployed drugs has aggravated malaria control programs, led to use of combination therapies that increase both cost and side effects, and forced increased investment in drug discovery and development. Reports of delayed Plasmodium falciparum clearance with artemisinins (2), the current mainstay of treatment, raise fears that recent gains in malaria control may also be lost. Understanding the resistance mechanisms of the parasite is critical to optimal usage of existing drugs and the development of new, less susceptible therapies.
Antimalarial resistance is known to arise through DNA-level changes in parasite genes that encode the drug target or efflux pumps. For example, specific mutations in the targets of atovaquone and sulfadoxine-pyrimethamine are known to reduce inhibition by these drugs (3,4). pfcrt and pfmdr1, genes for transporters that mediate efflux of chloroquine, mefloquine, and possibly other antimalarials (5)(6)(7), also incur specific causal mutations. In addition to mutations, copy number variation through genomic recombination events have also been implicated in drug resistance (8,9).
Reduced drug uptake by infected erythrocytes represents a distinct and recently proposed mechanism of acquired resistance.
Although not yet described with approved drugs, it has been reported for blasticidin S and leupeptin (10 -12). These agents inhibit parasite ribosomes and intracellular proteases, respectively. In both cases, resistant parasites exhibit changes in the plasmodial surface anion channel (PSAC), 2 an unusual broad selectivity ion channel present only on infected erythrocytes. Nevertheless, the precise mechanism of drug resistance and its molecular basis remain unknown, partly due to possible contributions from putative host transporters.
We now report a novel epigenetic mechanism responsible for blasticidin S resistance. Our studies reveal histone modifications that suppress transcription of parasite clag3 (cytoadherence linked antigen), genes recently linked to PSAC activity (13). Expression of clag2, an uncharacterized paralog, is also curtailed in resistant parasites, suggesting a role for this paralog in transport and revealing sophisticated regulation of this gene family. The ability to acquire and revert from drug-resistant phenotypes via a reversible epigenetic mechanism confers survival advantage under diverse challenges and is worrisome for malaria control programs.

EXPERIMENTAL PROCEDURES
Parasite Culture and in Vitro Selection of Drug-resistant Parasites-P. falciparum laboratory lines were cultivated by standard methods and maintained under 5% O 2 , 5% CO 2 , 90% N 2 at 37°C. Mutant parasite lines resistant to either blasticidin S alone (FCB-br1) or to both blasticidin S and leupeptin (FCB-2mut) were previously generated through in vitro selections applied to the wild-type FCB line (10,11); both are limiting dilution clones. The 2C3 clone was obtained by limiting dilution cloning of a blasticidin S-sensitive line resulting from cultivation of FCB-br1 without drug pressure. 2C3-R1, 2C3-S1, 2C3-R2, and 2C3-S2 were then generated by sequential application and removal of 2.5 g/ml of blasticidin S to continuously propagated cultures; with each cycle, steady-state phenotypes were observed within 6 -28 days after changes in drug pressure.
Computational Analyses-CLAG protein sequences were aligned using MUSCLE (14); the alignment was inspected and improved manually. A Bayesian phylogenetic tree was then calculated using MrBayes version 3.1.2 (15). Convergence was achieved after 2 million generations; the resulting phylogeny was the consensus of 3002 trees. A maximum parsimony phylogeny, calculated using MEGA5 (16), produced the same topology. Ancestral sequences were reconstructed using the PAML version 4.5 maximum likelihood algorithm (17) applied to a calculated Bayesian tree topology with the Jones-Taylor-Thorton substitution model (18). Mean pairwise sequence distance within each paralog clade was calculated using the Jones-Taylor-Thorton substitution model with ␥ distributed rate heterogeneity among sites as implemented in MEGA5.
DNA Sequencing-Candidate genes and their upstream promoter regions were sequenced by standard methods after amplification with specific primers and high fidelity Speed-STAR HS DNA polymerase (Takara Bio, Shiga, Japan). Wholegenome sequencing used 200-bp fragment libraries and HiSeq2000 hardware (Illumina). 101-bp paired reads were mapped to the P. falciparum 3D7 version 8.0 reference assembly.
Quantitative RT-PCR-Gene expression was measured with real-time PCR using total RNA harvested 28 h after sorbitol synchronization of cultures. Contaminating genomic DNA was removed by DNase I treatment (TURBO DNA-free kit, Ambion). RNA was adjusted to 5 g/reaction and reverse transcribed at 50°C using oligo(dT) priming and SuperScriptIII (Invitrogen). The resulting cDNA was used for quantitative real-time PCR to determine expression levels of the indicated genes. PF07_0073, a constitutively expressed gene, was used as a normalization control to allow comparisons between sensitive and resistant parasite lines.
Real-time PCR was performed in single-plex using primers listed in Table 1 and the QuantiTect SYBR Green PCR kit (Qiagen). Primers were selected based on specificity for the desired gene, an amplicon of Յ200 bp, and matched melting temperatures. Amplification was quantified by the iCycler iQ multicolor real-time PCR system (Bio-Rad) using a three-stage ther-mal profile. After heating cDNA to 95°C for 15 min, annealing and extension were performed for 40 cycles consisting of 94°C for 30 s followed by 52°C for 30 s. The final stage used gradual heating from 55 to 95°C, incrementing by 0.5°C/30 s; this dissociation protocol was used to confirm the specificity of primer binding and product formation. Two negative controls, omission of reverse transcriptase or cDNA template, were performed with each experiment to exclude genomic DNA contamination. All reactions were done in triplicate; the average threshold cycle (C T ) was used for data analysis. Each experiment was performed on at least three separate RNA harvests.
Gene expression was then quantified using the 2 Ϫ⌬⌬ C T method (19). Each C T of the target gene was first normalized to the C T of the Pf07_0073 control to obtain a ⌬C T value for each parasite line. The difference between this ⌬C T value and the corresponding measurement for the blasticidin S-sensitive parent yielded the ⌬⌬C T . Because exponential cDNA amplification is expected to double the product with each cycle, the value 2 Ϫ⌬⌬ C T then estimates the fold-change in gene expression in the mutant parasite.
Genome-wide Expression Microarrays-Whole genome expression profiling was performed using an available 70-mer oligonucleotide microarray that represents ϳ6000 open reading frames of the P. falciparum genome (20). Total RNA was isolated from tightly synchronized trophozoite-stage cultures of FCB and FCB-br1, labeled with Cy3 or Cy5-dUTP, and used for two-color hybridization with microarray probes based on the parasite genome sequence 3D7 reference. Dye-swap experiments were performed to exclude possible artifacts. Mean signals from 7 independent hybridization experiments were calculated without filtering. Probes that do not map to the PlasmoDB transcript set were removed. Hits were defined as genes producing a hybridization signal Ն100 absorbance units and a 10-fold change in expression.
Chromatin Immunoprecipitation and qPCR-qCHIP experiments were carried out as described previously (21). Synchronous trophozoite-infected erythrocytes were lysed with 0.05% saponin, washed, and fixed with 1% formaldehyde for 5 min at room temperature. After quenching with 1.25 M glycine, nuclei were isolated (Wizard DNA purification kit, Promega), lysed in 1% SDS, 10 mM EDTA, 50 mM Tris, pH 8.0, with Complete, EDTA-free Protease Inhibitor Mixture (Roche Applied Science) at 4°C, and sonicated to obtain DNA fragments of 500 bp average size. After centrifugation (16,000 ϫ g, 20 min), the lysate was pre-cleared with protein G-agarose/salmon sperm DNA (Millipore) in 167 mM NaCl, 0.01% SDS, 1% Triton X-100, 1 mM EDTA, 20 mM Tris, pH 8.0, for 1 h at 4°C. After reserving a fraction for input qPCR, equal fractions were incubated overnight with 1:100 dilutions of anti-H3, anti-H3K9me3, anti-H3K9ac, anti-H3K4me3, or no antibody (Millipore and Abcam). The immunoprecipitate was recovered on protein G-agarose/salmon sperm DNA after a 4-h incubation, and washed with increasing stringency: the above buffer with 150 and 500 mM NaCl followed by 0.25 M LiCl, 1% Triton X-100, 1% sodium deoxycholate, 1 mM EDTA, 20 mM Tris, pH 8.0, and finally with TE buffer. DNA was eluted with 1% SDS, 0.1 M NaHCO 3 ; cross-links were reversed by overnight incubation at 65°C. DNA was purified and precipitated prior to qPCR with gene-specific primers ( Table 2). Enrichment of histone modifications is presented as the log 2 ratio of each mutant or revertant line relative to the wild-type FCB parent after normalization against the control H3 antibody in each experiment. DNA Transfection to Prevent clag3 Silencing-The fulllength clag3.1 open reading frame along with the native 599-bp 3Ј UTR was cloned from the 3D7A parasite line into the piggyBac integration plasmid pXL-BACII-DHFR (22). DNA sequence encoding the FLAG epitope tag was inserted immediately before the stop codon of the gene. A 1105-bp 5Ј UTR upstream of the msp2 start ATG, previously shown to drive expression of a reporter gene (23), was amplified using 5Ј-CCAACCGTCGACGAATTCTTATTCTTGCCATCC-3Ј and 5Ј-CCGGAACTCGAGTTTGACTAATATAATATGTTA-3Ј, digested with SalI and XhoI, and cloned into the plasmid upstream of the clag3.1 start ATG. Plasmid construction was confirmed by DNA sequencing.
Uninfected erythrocytes were loaded with this plasmid and the pHTH helper vector carrying the piggyBac gene. 2C3 para-sites were cultivated in these erythrocytes for 4 days prior to addition of 2.5 nM WR99210 to select for integration. Transfected parasites were detected after 7 weeks of continuous culture. All experiments used limiting dilution clones obtained with the C-SNARF method (24).
Southern Blot Hybridization-Genomic DNA was digested with the XhoI restriction enzyme, separated on a 0.8% agarose gel at 40 volts for 16 -20 h, and transferred to Hybond-N ϩ membrane (GE Healthcare). A DNA probe specific for hdhfr was amplified with primers (5Ј-ATTTCCAGAGAATGACCA-CAAC-3Ј and 5Ј-TTAAGATGGCCTGGGTGATTC-3Ј) and labeled with digoxigenin-dUTP according to the manufacturer's instructions (DIG DNA Labeling and Detection kit, Roche Applied Science). Hybridization was performed overnight at 39°C in DIG-Easy Hyb prior to washing twice in 2ϫ SSC with 0.1% SDS for 10 min at RT and twice in 1ϫ SSC with 0.5% SDS for 15 min at 52°C. The membrane was then incubated with anti-digoxigenin-AP Fab fragments at a dilution of 1:10,000 followed by visualization on HyBlot CL x-ray film using CDP-Star substrate (Roche).
Immunoblots-Total cell lysates were prepared from Percollenriched trophozoite-infected cells by lysis in 7.5 mM Na 2 HPO 4 , 1 mM EDTA, pH 7.5, with HALT protease inhibitor mixture (Thermo Scientific). This lysate was denatured and reduced in LDS Sample Buffer with 100 mM DTT prior to electrophoresis on NuPAGE Novex 4 -12% BisTris gels in MES SDS buffer (Invitrogen) and transfer to nitrocellulose membranes. The nitrocellulose membranes were blocked (3% skim milk in PBS with 0.1% Tween 20) and probed with either anti-CLAG3 or anti-FLAG antibody (Cell Signaling Technology) before detection with HRP-conjugated secondary antibody (Pierce) at a 1:3000 dilution in blocking buffer. The blots were developed using enhanced chemiluminescent substrates (Immobilon, Millipore, or SuperSignal West Pico, Pierce) and visualized on HyBlot CL x-ray film. All blots used matched loading of samples, as confirmed by Coomassie Blue staining of gels or Ponceau S staining of blots for hemoglobin.
Solute Permeability Measurements-Infected erythrocyte permeability to organic solutes was quantified with a light scat- tering assay (25). Trophozoite stage-infected RBCs were enriched to Ͼ95% parasitemia by Percoll-sorbitol density gradient centrifugation, washed in 150 mM NaCl, 20 mM HEPES, 0.1 mg/ml of BSA, pH 7.4, and resuspended at 5% hematocrit. Osmotic lysis due to PSAC-mediated solute uptake was initiated by addition of 40 volumes of buffered lysis solution (280 mM sorbitol or 145 mM PhTMA-Cl with 20 mM Na-HEPES, 0.1 mg/ml BSA, pH 7.4) with inhibitor as indicated at 37°C. Continuous tracking of 700 nm light transmittance through the cell suspension was used to quantify lysis kinetics and estimate solute permeability.
Growth Assays-Parasite growth upon blasticidin S challenge was assessed with microscopic examination of Giemsa-stained smears or SYBR Green I detection of parasite DNA as described previously (26); these methods produced similar results.
Statistical Analyses-One-way analysis of variance (ANOVA) was used to evaluate the hypothesis of no difference between three or more groups. If significance was reached (p Ͻ 0.05), post hoc pairwise comparisons used Student's t tests.

Blasticidin S Resistance Is Not Linked to DNA Level
Changes-A clonal PSAC mutant termed FCB-br1 was previously generated by in vitro selection with blasticidin S (10); subsequent selection for leupeptin resistance followed by cloning yielded a distinct double mutant referred to as FCB-2mut (11). Both mutants exhibit decreased organic solute permeabilities and altered PSAC activity in patch clamp studies. To explore whether genome-level changes account for these mutant channel phenotypes, we first considered the two clag3 genes of these mutants. These genes have recently been implicated in infected erythrocyte permeability (13); how the encoded proteins contribute to PSAC is unclear at present. DNA sequencing of both genes in FCB-br1 and FCB-2mut did not reveal either mutations or indels relative to the blasticidin S-sensitive FCB parental line (not shown).
Although most plasmodial species have only two or three clag genes in their genomes (27,28), P. falciparum has undergone significant expansion to have five members in this gene family. To explore whether the other paralogs contribute to PSAC, we performed phylogenetic analysis of clag genes from laboratory lines representing Africa, South America, and Southeast Asia and found that each paralog has primary sequence elements that distinguish it from all other clag products, with the clag9 group being the most divergent (Fig. 1A). Notably, even though clag3.1 and clag3.2 undergo mutually exclusive expression and should therefore serve fully overlapping roles, these paralogs also segregate into discrete clades with high confidence (posterior probability ϭ 1.0). These findings suggest relatively ancient expansion and ongoing evolution of the clag gene family in P. falciparum.
CLAG3.1 and CLAG3.2 both carry a ϳ30 residue variable domain located some 1100 residues from the N terminus of the protein (13,29), accounting for the substantial pairwise diversity within this clade (pink circle, Fig. 1A). This domain is exposed at the host cell surface, consistent with selection for variability by host immunity. Our analysis of CLAG2 and CLAG8 sequences also revealed pairwise diversity, although at somewhat lower levels (smaller colored circles, Fig. 1A). Interestingly, the diversity in these proteins was also largely attributable to a variable domain at a similar position in the primary sequence, suggesting that these paralogs may have transmembrane topologies similar to that of CLAG3. In contrast, CLAG9 sequences lack a variable domain and have accrued few polymorphisms in P. falciparum.
Chain termination sequencing of the five clag paralogs did not reveal any changes in either FCB-br1 or FCB-2mut relative to the sensitive FCB parental line. Because unrelated genes may also contribute to PSAC activity (30), we then used next generation sequencing to search for responsible mutations or indels. We analyzed coverage and determined that 95% of coding regions achieved a minimum of 15ϫ coverage; multigene families such as the clag genes were also well covered (data not shown). Only six genes had variants that met our SNP calling threshold in one or both resistant parasites, but these did not confirm with chain termination sequencing. Thus, deep sequencing suggests that blasticidin S resistance does not result from DNA level changes.
Suppressed clag2 and clag3 Expression in Both Mutants-We next used real-time quantitative reverse transcriptase PCR (RT-qPCR) to evaluate possible changes in transcription with primers specific for each open reading frame of the clag genes (Table 1). rhoph2 and rhoph3, genes whose products make stable interactions with CLAG proteins (31,32), were also examined. In synchronous trophozoite-stage cultures, the wild-type FCB parent expressed clag3.1 at levels that were 42 Ϯ 16-fold higher than the constitutively expressed control gene, PF07_0073 (Fig. 1B). Its nearby paralog, clag3.2, was expressed at much lower levels (0.2 Ϯ 0.07 relative to PF07_0073), consistent with monoallelic expression of these two family members (33). Preferential expression of one or the other clag3 gene has been documented for various lines (13,34), but it is unclear whether this reflects differences between the proteins that confer an in vitro survival advantage to parasites.
We found marked reductions in expression of both clag3.1 and clag2 in FCB-br1 cultures grown in the continuous presence of 2.5 g/ml of blasticidin S (140-and 26-fold, respectively; p Ͻ 10 Ϫ4 each, 8 independent trials). FCB-2mut, when cultivated with blasticidin S and leupeptin, exhibited comparably reduced expression of the same paralogs (87-and 18-fold lower than FCB, respectively; p Ͻ 10 Ϫ4 ). Both mutants showed small changes in expression of clag3.2, clag8, and clag9, but these did not reach statistical significance. Although expression of rhoph2 was decreased in FCB-br1, both rhoph2 and rhoph3 exhibited increased expression in FCB-2mut (Fig. 1B, p Ͻ  0.005). Because expression changed in opposite directions for these two mutants and the changes were relatively modest, we did not examine rhoph2 and rhoph3 further in this study. Whole genome expression microarray experiments excluded global changes in parasite gene expression and revealed a small number of candidates for further study (data not shown, GEO Series accession number GSE47579, www.ncbi.nlm.nih.gov/ geo/query/acc.cgi?accϭGSE47579).
clag2 and clag3.1 Silencing Is Reversible in FCB-br1, but Not in FCB-2mut-We previously reported that the FCB-br1 mutant loses its resistance phenotype when cultivated without blasticidin S (10). Reversion was associated with restoration of wild-type PSAC activity in both macroscopic transport measurements and single channel patch-clamp. To examine the significance of clag gene silencing, we next generated a series of blasticidin S-sensitive and resistant lines from FCB-br1 by sequential growth in the absence and then the presence of this antibiotic. Because in vitro selections may generate parasite subpopulations with distinct genotypes and channel phenotypes, we first used limiting dilution cloning to generate the 2C3 line from a FCB-br1 revertant. When challenged with blasticidin S, the 2C3 line acquires resistance more quickly than FCB, as quantified in a later section. 2C3-R1, 2C3-S1, 2C3-R2, and 2C3-S2 were then sequentially generated by applying cycles of blasticidin S pressure to 2C3 cultures.
With each cycle, we quantified PSAC activity by measuring the permeability of phenyl-trimethyl ammonium (PhTMA ϩ ), an organic cation with known channel-mediated uptake (35,36) (Fig. 2A). These studies revealed that each sensitive line has similarly high PhTMA ϩ permeability (FCB, 2C3, and 2C3-S1) and that resistant lines have low permeabilities that are indistinguishable (FCB-br1, 2C3-R1, and 2C3-R2). The sensitive 2C3-S2 line appears to have an incomplete recovery of transport after extended culture without blasticidin S, but this was not significant (p ϭ 0.1 relative to FCB, n ϭ 7 trials each).
These lines were then used to examine the biological relevance of suppressed clag gene expression. RT-qPCR revealed that each resistant line had significantly lower expression of both clag2 and clag3.1 than any sensitive line (Fig. 2B), providing evidence for a direct link between gene expression and PSAC activity. At the same time, these transcription levels revealed some interesting discrepancies. For example, each of the blasticidin S-sensitive revertants retained lower expression of these genes than the parental FCB line despite complete recovery of channel-mediated solute uptake (p Ͻ 0.01 for comparison of clag2 in FCB to 2C3 and each subsequent sensitive line; similar comparisons not significant for clag3.1). We also noticed diminishing levels of clag3.1 recovery after each sequential cycle of blasticidin S pressure ( Fig. 2B; p ϭ 0.05 for comparison of 2C3 and 2C3-S2). These findings suggest that memory of prior blasticidin S exposure can be passaged through many parasite generations; similar phenomena have been reported in other organisms (37).  red and blue, respectively). B, transcript abundance for clag, rhoph2, and rhoph3 genes in the indicated parasite lines, calculated according to 2 Ϫ⌬ C T using PF07_0073 as an internal control. The ordinate is shown on a log scale and begins at 0.003 based on a maximum C T value of 32 that could be reproducibly detected. Asterisks indicate significant differences within a group (p Յ 0.005, one-way ANOVA).
We also examined expression of other clag genes and found significantly increased expression of clag3.2 in 2C3-S1 and 2C3-S2 than in the wild-type FCB parasite ( Fig. 2C; p Յ 0.02 in post hoc t-tests, p Յ 10 Ϫ4 for all parasites in one-way ANOVA). Nevertheless, the total clag3 expression, calculated as the sum of the normalized transcript levels for both clag3 genes, remained low in 2C3-S1 and 2C3-S2 (compare Figs. 1B and 2C). One explanation for increased clag3.2 expression is that selection with blasticidin S may have interfered with factors that produce preferential expression of clag3.1 in the wild-type FCB line. Significant changes were not detected in transcription of the other two paralogs, clag8 and clag9 (p ϭ 0.8 and 0.4, respectively, ANOVA); there remains no experimental evidence for a role of these genes in PSAC activity.
The above findings reveal a correlation between PSAC transport rates and expression of clag2 and clag3. One explanation is active suppression of erythrocyte permeability by the parasite to reduce antibiotic uptake; another is that blasticidin S may directly suppress transcription of some parasite genes, possibly via interactions with unidentified DNA-binding proteins. The FCB-2mut line, which retains both in vitro resistance and reduced solute permeability despite prolonged cultivation without blasticidin S and leupeptin (11), permitted us to distinguish these possibilities. Cultivation of this stable channel mutant without antibiotics did not alter expression of any the genes under study (Fig. 2D), excluding a direct effect of either blasticidin S or leupeptin on transcription. These findings suggest a parasite mechanism for regulation of these channel-associated genes.

Gene Silencing Is Associated with Epigenetic Histone
Modifications-Individual wild-type parasites express only one of the two clag3 genes, but occasionally switch to express the other paralog. This monoallelic expression and switching is mediated through post-translational histone modifications (21,34). Notably, this physiological switching does not impact clag2 expression (33). To explore whether epigenetic histone marks mediate silencing in our mutants, we performed chromatin immunoprecipitation and quantified gene-specific histone modifications with quantitative PCR (qChIP, Fig. 3). These experiments revealed a significant decrease in H3K9ac and H3K4me3 modifications on both clag2 and clag3.1 genes in FCB-br1 and FCB-2mut (p Ͻ 10 Ϫ4 for each gene in both mutants, one-way ANOVA). Both types of modifications are associated with active promoters in P. falciparum and other organisms (21,34,38). Both were restored in the revertant 2C3 line (Fig. 3C), consistent with gene silencing due to loss of these activating marks. Interestingly, H3K9me3, a modification strongly associated with repression and heterochromatin formation in malaria (39,40), was not measurably enriched in either resistant line, suggesting differences between this resistance mechanism and conventional switching of clag3 and other variant gene families in malaria (21). clag9, used as a control gene in these qChIP experiments, exhibited modest increases in each of these modifications in the FCB-br1 line.
Suppressed Transcription Yields Reduced CLAG3 Protein-Because transcript levels do not necessarily predict steady-state protein abundance, we next used immunoblotting with a specific antibody to quantify CLAG3 production. The antibody,  ). B, mean Ϯ S.E. expression levels for clag2 and clag3.1, normalized to 1.0 for FCB parasites using the 2 (Ϫ⌬⌬ C T ) method. Both genes exhibit significant changes upon blasticidin S cycling (p Յ 10 Ϫ3 , ANOVA). C, transcript levels for clag3.2, clag8, and clag9, presented relative to the internal control PF07_0073 because of inherent differences in expression of these genes. C T values for clag8 were below the detection threshold. clag3.2 expression varies significantly upon blasticidin S cycling (p Ͻ 10 Ϫ4 , ANOVA). D, transcript levels for the indicated genes in the FCB-2mut parasite when cultivated with blasticidin S and leupeptin (black bars) or after continuous cultivation without these agents for 1 month (white bars). Values represent mean Ϯ S.E. determined relative to PF07_0073. Antibiotic removal does not measurably affect transcription in this stable mutant. raised by immunizing mice with a 141-residue C-terminal fragment of CLAG3 (13), may also recognize CLAG2 based on sequence similarity, but low identity with CLAG8 or CLAG9 suggests it will not recognize these paralogs. Immunoblotting with this antibody revealed a markedly reduced signal in each blasticidin S-resistant parasite (Fig. 4A), consistent with reduced CLAG3 abundance due to suppressed transcription.
Peripheral membrane protein extraction with Na 2 CO 3 revealed that CLAG3 is detected as an integral membrane protein in FCB parasites, but is below the detection threshold in FCB-br1 (Fig. 4B). It is possible that CLAG3 delivery to the host membrane is also compromised in blasticidin S-resistant parasites, as reported for ion channel proteins in other systems (41). We were unable to examine such post-translational regulation because the low CLAG3 levels in our mutants prohibited protein trafficking studies.
Resistance Requires Suppressed clag3 Expression-To determine whether clag gene silencing is sufficient to produce blasticidin S resistance and altered PSAC activity, we sought a DNA transfection approach that would prevent silencing under blasticidin S pressure. We reasoned that use of an unrelated promoter to drive expression may allow escape from the above epigenetic mechanisms. At the same time, it is important to preserve the stage specificity of clag gene expression because transcripts made on an incorrect schedule in the parasite life cycle may yield protein that either fails to interact with essential ligands or does not traffic properly (42). We selected the well characterized promoter of merozoite surface protein-2 (msp2, PFB0300c) because it, like that of all clag genes, directs maximal transcription at the schizont stage (23,43). Finally, to avoid silencing mechanisms that depend on the location of the gene within the parasite genome and local heterochromatin formation, we selected the piggyBac transposase system. Stable complementation of genes using this transposase allows random integration at one of numerous sites throughout the parasite genome (44), a potential advantage under our circumstances. Fig. 5A shows the pXL-M2-120w vector, which carries a 1105-bp msp2 promoter followed by the full-length clag3.1 cDNA cloned from 3D7A parasites, a C-terminal FLAG epitope tag, and a 599-bp 3Ј UTR native to clag3.1. This gene and a human dihydrofolate reductase (hDHFR) selectable marker were inserted in tandem between two inverted terminal repeat  sequences, which are recognized by piggyBac transposase. After transfection of 2C3 with this vector and the pHTH helper plasmid encoding piggyBac, we obtained parasites resistant to WR99210, indicating expression of hDHFR. Subsequent limiting dilution cloning yielded two distinct clones, 2C3pB-C2 and 2C3pB-G11. Southern blotting revealed that these clones had integrated the msp2 promoter driven clag3.1 gene and hDHFR cassette into separate sites in the parasite genome (Fig. 5B). Both clones made CLAG3 protein of correct size, as determined by immunoblotting with an antibody against the FLAG epitope tag (Fig. 5C).
We next explored whether this heterologous protein contributes to transport at the host membrane. Although CLAG3 proteins are conserved in divergent parasites (Fig. 1A), each line FIGURE 5. Functional expression of CLAG3 under the msp2 promoter to circumvent silencing. A, plasmid map for pXL-M2-120w, an integration vector that expresses the 3D7A clag3.1 gene under the msp2 promoter; the product carries a C-terminal FLAG epitope tag. The two inverted terminal repeat (ITR) sequences on this plasmid are recognized by piggyBac, expressed from the pHTH helper plasmid. B, Southern blot showing separate sites of integration in the 2C3pB-C2 and 2C3pB-G11 clones. DNA was digested with XhoI; detection used an hDHFR probe. Each transfection clone yielded a single band distinct from the original plasmid. Untransfected 2C3 parasite DNA does not bind this probe. C, immunoblot using anti-FLAG antibody, showing that both clones express full-length CLAG3 of expected size. 2C3 is not recognized, indicating antibody specificity. D, structure of ISPA-2, a PSAC inhibitor that distinguishes between 2C3 and 3D7A channels. E, osmotic lysis kinetics for the indicated parasite lines in buffered sorbitol with 0, 1, 2.5, 5, 7.5, and 10 M ISPA-2 (top to bottom traces in each panel, respectively). Notice the greater efficacy of lysis inhibition in 2C3 and an intermediate effect on the transfected 2C3pB-C2 line. F, tallied dose responses for ISPA-2 inhibition of PSAC, normalized to 1.0 for no inhibitor. Black circles, 2C3; blue circles, 3D7A; red triangles, 2C3pB-C2; yellow triangles, 2C3pB-G11. Solid lines reflect the best fits to y ϭ (a/(1 ϩ (x/c))) ϩ (1-a/ (1 ϩ (x/b))), where a, b, and c are constants. carries a unique polymorphic region near the C terminus. This polymorphic region is exposed at the host erythrocyte surface and interacts with a subset of PSAC inhibitors (13), which are referred to as ISPA compounds (for isolate-specific PSAC antagonists) to highlight their ability to discriminate between channels linked to specific genotypes. Because piggyBac complementation should yield parasites that express clag3.1 products from both 3D7A and 2C3 lineages, we surveyed several ISPA hits from recent high-throughput PSAC inhibitor screens (13,26). This revealed ISPA-2 as a compound with greater potency for inhibition of channels from the 2C3 parasite than from 3D7A (Fig. 5, D-F; p Ͻ 10 Ϫ9 , n ϭ 7-10 trials each). Transport studies revealed that the two transfection clones have identical ISPA-2 inhibition dose responses that are intermediate between those of 2C3 and 3D7A. Although inhibition more closely resembled that of 2C3, the change from this untransfected parent was significant (p Յ 10 Ϫ4 for comparisons of 2C3 to either 2C3pB-C2 or 2C3pB-G11 at 1 M ISPA-2). Because the two clones carry the 3D7A transgene at separate sites in their genomes, their similar inhibition phenotypes suggest that expression level, rather than genomic environment, determines channel block by ISPA-2. Importantly, the intermediate inhibition dose response is consistent with contributions from clag3 products encoded by both 2C3 and 3D7A parasites. One possibility is that these pseudodiploid parasites assemble channels consisting of the two CLAG3 isoforms and therefore exhibit inhibition that is a weighted average of the individual isoforms. The absolute change in ISPA-2 affinity cannot, however, be used to estimate the relative copy number of the two channel isoforms because the precise role and stoichiometry of CLAG3 contribution to PSAC is unknown.
We then challenged both transfected and untransfected lines with blasticidin S (Fig. 6A). When compared with the resistant FCB-br1 clone, each sensitive line was adversely affected by addition of the antibiotic. However, the 2C3 line exhibited a remarkably rapid adaptation as blasticidin S was unable to produce microscopic clearance of the culture and parasitemia recovered fully within 6 days. The fully susceptible FCB parasite did not recover within the time frame of this experiment, and instead required Ն2 months to acquire resistance (10). This difference between FCB and the revertant 2C3 daughter implicates genome-level changes in addition to the epigenetic mechanism studied here.
Importantly, the two transfection clones did not adapt readily to blasticidin S challenge, indicating a specific effect of clag3 complementation. Nearly a month of continuous cultivation under blasticidin S pressure was required before viable parasites could be detected from these cultures (Fig. 6B). Thus, whereas transfection did not abolish acquisition of blasticidin S resistance, complementation with clag3.1 under the msp2 promoter establishes that gene silencing is essential for rapid acquisition of the resistance phenotype.
To explore possible mechanisms for the eventual growth of resistant parasites, we quantified gene expression in the blasticidin S mutant generated by selection of 2C3pB-G11 and confirmed silencing of the endogenous clag3.1 gene (native clag3.1, Fig. 6C). Surprisingly, however, we also found that the 3D7Aderived clag3.1 transgene had also undergone modest, but statistically significant silencing (p ϭ 0.02, n ϭ 4 trials) despite use of an unrelated msp2 promoter. DNA sequencing of the heterologous msp2 promoter did not reveal mutations or indels that could account for reduced expression of the transgene, again suggesting an epigenetic mechanism. Because expression of the endogenous msp2 gene did not undergo similar suppression, the parasite appears to have a remarkable ability to regulate clag3.1 gene expression via mechanisms that are only partially dependent on promoter sequence or chromosomal location. Epigenetic modifications associated with clag coding regions, as detected here (Fig. 3), may account for silencing despite expression under the msp2 promoter.

DISCUSSION
Epigenetic histone modifications regulate cytoadherence, erythrocyte invasion, and developmental progression in malaria parasites (45)(46)(47), but were not known to produce drug resistance. We report this new role for epigenetic marks. Blasticidin S-resistant parasites silence their clag2 and clag3 genes to reduce the encoded protein levels at the host membrane and FIGURE 6. Circumvented clag3 silencing aborts rapid acquisition of blasticidin S resistance. A, rapid recovery of 2C3 cultures, but not of the parental FCB sensitive line or the two 2C3 transfection clones that express CLAG3 under the msp2 promoter. Bars represent mean Ϯ S.E. fold-increase in parasitemia over 6 days after addition of 2.5 g/ml of blasticidin S (n ϭ 3 replicate experiments each). B, number of days required for detection of parasites after blasticidin S challenge in the indicated transfection clones, as determined by microscopic examination of Giemsa-stained smears. C, transcript abundance in the resistant line generated from 2C3pB-G11 by continuous growth with blasticidin S, normalized relative to the unchallenged clone according to 2 (Ϫ⌬⌬ C T ) . Both the native clag3.1 and the transgene driven by the msp2 promoter (clag3.1 3D7 ) are significantly suppressed (asterisk, p Յ 0.02).
curtail channel-mediated uptake of this antibiotic. We confirmed a direct contribution to resistance by expressing clag3 under a heterologous promoter to circumvent epigenetic silencing. Although blasticidin S is not used to treat human malaria, our findings suggest antimalarial drugs that require channel-mediated uptake to reach their intracellular targets may also be prone to resistance and clinical failure by this mechanism. Although chloroquine and artemisinin are lipophilic and do not appear to require channel-mediated uptake (48,49), drug discovery programs increasingly prefer high water solubility to improve bioavailability (50); many such compounds may depend on uptake via PSAC to reach their intracellular parasite targets. Thus, future antimalarial drugs may encounter the resistance mechanism described here.
Our findings implicate a new level of regulated expression for clag genes. Although epigenetic switching between the two clag3 genes is established (13,33,34), the silencing reported here has some distinctive features. First, blasticidin S-induced silencing affects clag2 and both clag3 genes, whereas monoallelic expression suppresses individual clag3 genes without reducing total transcript levels. Second, this resistance mechanism uses a distinct pattern of epigenetic marks as it does not involve H3K9me3, a modification linked to silencing and monoallelic expression in parasite gene families (Fig. 3)(39). Third, this new layer of regulation appears to use epigenetic modifications that can be more rapidly applied or removed. Although switching between the two clag3 genes occurs with a half-time of ϳ2 months after perturbation (13), we found that sensitive 2C3 parasites recover and expand within only 6 days after blasticidin S challenge (Fig. 6A). Reversion to the sensitive phenotype also occurs rapidly as transport was restored in less than 1 month upon removal of blasticidin S ( Fig. 2A). Finally, whereas conventional silencing and switching between the two clag3 genes has only modest effects on channel properties (13), the epigenetic changes described here yield marked reductions in solute transport (Fig. 2). We propose that silencing of individual clag3 genes in sensitive parasites serves to protect the silenced paralog from host immunity (51,52), but that the mechanism reported here functions to regulate host cell permeability in response to nutrient availability and drug pressure (53). For example, in vivo suppression of PSAC may contribute to development of dormant bloodstream parasites, as observed after drug exposure (54). Regulated host cell permeability may also be important for fine-tuning changes in erythrocyte cation concentrations (55).
Gene silencing in resistant parasites represents independent evidence for a role of clag genes in nutrient uptake at the host membrane. Nevertheless, it is not safe to conclude that CLAG proteins form ion channels on their own. They lack sequence homology to known channels and have fewer predicted transmembrane domains than most pore-forming proteins. CLAG proteins also have proposed roles in cytoadherence and protein trafficking (56 -58), raising additional questions about how the protein may alter host cell permeability. Conclusive answers will require multiple experimental approaches including heterologous expression and DNA transfections in malaria parasites.
Our new evidence linking the uncharacterized CLAG2 paralog to PSAC activity provides a framework for examining the elusive composition of these channels. If future studies reveal that CLAG proteins form aqueous pores directly, individual paralogs may define distinct parts of the functional channel. This model is reminiscent of the acetylcholine receptor, which is composed of four related subunits that undergo controlled assembly into a gated ion channel (59). If these parasite proteins function as transport activators rather than as stable channel subunits, then our findings predict that CLAG2 will also exhibit similar enzymatic activity.
A recently generated clag3 knockdown parasite may also help determine the precise role of CLAG proteins (34). There, transfection was used to replace the clag3.2 gene with an hDHFR selectable marker cassette. Subsequent selection for hDHFR expression reportedly silenced clag3.1 and yielded undetectable expression of both clag3 genes. The resulting knockdown parasite may therefore resemble our blasticidin S-resistant lines, which exhibit ϳ100-fold reduced expression of clag3. The transport properties and drug sensitivities of the knockdown parasite were not studied there, but should be pursued. One possibility is that the newly implicated CLAG2 plays a compensatory role to sustain transport.
Although silencing of clag2 and clag3 are tightly associated with blasticidin S resistance, several findings suggest that silencing alone is not sufficient to account for the observed transport phenotypes. 1) Whole cell patch clamp measurements reveal that FCB-br1 has a 7-fold reduction in PSACspecific Cl Ϫ currents (10), which may be less than expected by the measured reductions in transcript and CLAG3 protein (Figs. 2 and 4). 2) Altered single channel gating and solute selectivity in these mutants is also inconsistent with the conservative prediction that transcriptional down-regulation yields a reduced number of otherwise unchanged channels (10,11). 3) Although FCB-br1 and FCB-2mut exhibit similar levels of clag gene silencing, these mutants have very different channel properties and responses to removal of selective pressure. 4) FCB and the revertant 2C3 lines switch to the resistant phenotype at different rates (Fig. 6A). One explanation for these findings is that other proteins are also required to form functional channels at the host membrane. Then, the slow initial acquisition of blasticidin S resistance in FCB may reflect one or more DNA level changes in an unidentified channel component. Such changes may then be permissive for clag2 and clag3 silencing, which appears to be the final step in reducing host cell permeability and antibiotic sensitivity. Another possibility is that PSAC is formed exclusively by CLAG paralogs assembled as a heteromeric complex. In this scenario, differential silencing of specific subunits may lead to assembly of altered channels. The availability of mutants with altered channel phenotypes, robust transport methodologies, and new molecular approaches described here should help reveal the structure and function of this unusual ion channel.
Acknowledgments-We thank Godfrey Lisk, Lanxuan Doan, and Tetsuya Furuya for technical help as well as Dyann Wirth, Sarah Volkman, and members of the Desai laboratory for helpful discussions.