Remodeling of the malaria parasite and host human red cell by vesicle amplification that induces artemisinin resistance

Symptoms and pathologies of Plasmodium falciparum malaria are entirely due to parasite stages that infect and remodel host red blood cells. At least a subset of these stages now show resistance to artemisinins: frontline, antimalarial drugs for which we still have no replacements. Emergence and spread of artemisinin resistance threatens worldwide malaria control and elimination.^1-3  PfKelch13 (K13) is a primary marker of artemisinin resistance.^4-7  K13 mutations confer resistance in “ring” stage parasites formed immediately after invasion,^8-10  as measured by the Ring Stage Survival Assay (RSA), an in vitro correlate of in vivo clinical resistance.¹¹  The major mutation K13C580Y diminishes binding to and ubiquitinylation-dependent proteosomal degradation of phosphatidylinositol-3-kinase (PfPI3K) to increase kinase levels.¹⁰  Notably, elevation of PfPI3K’s lipid product phosphatidylinositol-3-phosphate (PI3P) confers artemisinin resistance.¹⁰  Population transcriptomics of over 1000 clinical isolates separately revealed that the parasite endoplasmic reticulum (ER) unfolded protein response (UPR) was associated with K13 mutation.¹²  But where and how PI3P acts and its interaction with UPR (if any)^13,14  in infected red cells remains unidentified. In addition, although clinical artemisinin resistance was first identified as delayed clearance of ring-infected red cells from circulation, its consequences for parasite-induced changes in the host cell are unknown. Here we examine localization and dynamics of parasite PI3P, K13, and proteostasis systems that include UPR to delineate a mechanism that explains why hundreds of parasite determinants and multiple parasite organellar systems are implicated in resistance.^15-23  Furthermore, we provide evidence that drug resistance affects properties of the host red cell linked to immunity.

Pan-“PfEMP1” antibodies were raised to recombinant conserved C-terminal acidic-terminal sequence (ATS) domain of PfEMP1 by the commercial vendor Genscript Inc. Anti-PI3P was from Echelon Biosciences. Anti-human Band 3 was a gift from Phillip S. Low. All other antibodies were from Thermo Scientific Inc. (Rockford, IL).

For immunoelectron microscopy, late trophozoite/schizont stages of the Plasmodium falciparum strain PfNF54K13WT or PfNF54K13C580Y were purified and fixed in 4% paraformaldehyde (Electron Microscopy Sciences, Hatfield, PA) in 0.25 M of N-2-hydroxyethylpiperazine-N′-2-ethanesulfonic acid (HEPES) (pH 7.4) for 1 hour at room temperature and processed for immunoelectron microscopy (IEM).²⁴  Mouse anti-PI3P antibody was used at 1/250 to 1/400 dilution in phosphate-buffered saline/1% fish skin gelatin; rat anti-K13 antibody was used at 1/10 dilution, and guinea pig anti-Pf–binding immunoglobulin protein (BiP) antibody at 1/10 dilution. Samples were viewed in a Philips CM120 Electron Microscope (Eindhoven, The Netherlands) under 80 kV. For fluorescence microscopy, live and fixed cells were imaged by DeltaVision Deconvolution microscopy.²⁵ 

P falciparum laboratory strains and their transgenic counterparts as well as clinical Cambodian isolates (a kind gift from Arjen Dondorp) were prepared, as has been previously described.¹⁰  Clinical strain isolation and culture were approved by the Oxford Tropical Medicine Research Ethical Committee, the Ministry of Health in Cambodia (trial registered under NCT00493363), and the University of Notre Dame. For cytoadherence assays, P falciparum CS2 strains were panned by binding to chondroitin sulfate (CSA), as has been described.²⁶  For K13-membrane association, infected red cells were treated as indicated; soluble and insoluble components were analyzed with sodium dodecyl sulfate (SDS)–polyacrylamide gel electrophoresis and Western blots.

P falciparum purified schizonts/segmenters were lysed in 0.05% saponin, followed by 0.5% NP-40 in 20 mM HEPES (pH 7.9), 10 mM KCl, 1 mM EDTA, 1 mM EGTA, and 1 mM dithiothreitol (with protease inhibitor cocktail; Roche Diagnostics) at 4°C. Protein extracts were solubilized in 1% volume-to-volume ratio Triton X-100 and 1% weight-to-volume ratio sodium deoxycholate and were incubated with anti-PfEMP1 ATS antibodies (or mock treated) and then with protein G agarose. After washing, bound proteins were eluted in 50 mM glycine, pH 2.5, digested with trypsin, and subjected to liquid chromatography–tandem mass spectrometry (MS/MS), and peptides were identified using MaxQuant 1.5.2.8.²⁷  MS/MS spectra were searched against a combined PlasmoDb P falciparum 3D7 (version 24) and UniProt human protein database. Hypergeometric analyses were used to identify enrichment in the clinical transcriptome¹²  or Malaria Parasite Metabolic Pathways 2016 in PlasmoDB (http://plasmodb.org/plasmo/).

The Mann-Whitney U test or 1-way analysis of variance with a Tukey or Bonferroni post hoc analysis was used to compare the mean values between treatment groups. Statistical analysis was performed with GraphPad Prism (version 6.02). P values were 2 sided, with P < .05 being considered significant. All data in this study are included in this article (and the supplemental Methods, tables, and figures, available on the Blood Web site).

Additional details are presented in the supplemental Methods.

PI3P is indistinguishable from phosphatidylinositol-4-phosphate (PI4P) by mass spectrometry because of identical mass and charge. Conventional indirect immunofluorescence assays (IFAs) use detergents (or methanol) to permeabilize cells, which may result in loss of lipids (including PI3P/PI4P, because they are poorly fixed by formaldehyde-based crosslinking used in IFAs). Rather, tagged reporter proteins that bind selectively to PI3P (and PI4P) are used to visualize their sites of concentration in live (and fixed) cells.²⁸  Two PI3P-binding protein reporters (EEA1 and p40Phox, transgenically expressed), carrying a malaria parasite leader signal sequence, are recruited into and accumulate in perinuclear, ER-vesicular domains in early rings²⁹  (data for SS-EEA1 are shown in Figure 1A). When the ring parasite matures and replicates into multinucleated “schizonts,” each daughter inherits PI3P (Figure 1B), which in emerging extracellular “merozoite” parasites appears as a dynamic cluster of vesicles that are transferred to newly formed rings (Figure 1C; supplemental Movie 1). A reporter for cytoplasmic PI3P failed to reveal distinct, prominent sites of concentration in young rings (supplemental Figure 1A-D), but focal regions were seen in later-stage “trophozoites/schizonts” (as has been previously reported³⁰ ; Figure 1D). We were unable to develop parasites dually expressing both secretory and cytoplasmic (EEA1) reporters (not shown). We therefore undertook cryo-IEM to detect endogenous PI3P in infected red cells (using antibodies to PI3P previously validated in IEM studies in yeast³¹ ). As is shown in Figure 1E (and supplemental Figure 1E-G), PI3P was detected at the cytoplasmic and luminal faces (yellow and black arrows) of the food vacuole (fv) and apicoplast. In addition, a high density of PI3P was found in tubules characteristic of the ER in both cytoplasmic and luminal orientations (yellow and black arrows), as well as in clusters of fine vesicles cytoplasmic to ER tubules (yellow double arrows).

PI3P levels are affected by K13 mutations,¹⁰  but the location of K13 in relation to PI3P remains unknown. IFAs localized endogenous K13 in diffuse and focal distribution in trophozoite and schizonts (Figure 2A; supplemental Figure 2A). Cryo-IEM detected PI3P (6 nm gold) in the ER lumen marked by PfBiP (15 nm gold; Figure 2B, black arrow; supplemental Figure 2B) and enriched with K13 on ER tubules and vesicles (Figure 2C, single and double yellow arrows; supplemental Figure 2C). K13 lacks a secretory signal sequence consistent with its association with cytoplasmic PI3P. The gold particles for PI3P look well aligned, suggesting membrane structures beneath, even when tiny vesicles are difficult to see in cryosections. Biochemical fractionation studies suggested that K13 was peripherally associated with membranes (Figure 2D). Stereological analyses of gold particles indicated that ∼75% of K13 localized to ER tubules, vesicles, or both (Figure 2E). K13 was also found in vesicles distal from the ER (13%; they may also be derived from organelles other than the ER), and low levels (11%) were detected all over the parasite. Only low levels of PI3P vesicles free of K13 were observed (Figure 2C, red arrow), and PI3P with a K13 label (yellow arrows) appeared as well-structured clusters. Along the ER, cytoplasmic PI3P (single and double yellow arrows) consistently exceeded luminal lipid (black arrows). Together, these data suggested K13 as a marker for PI3P, concentrated in the ER. Quantitative IEM could not be undertaken at the ring stage of infection because of the small size of rings (1-2 μm), making them easy to miss in thin sections taken through infected red cells. However, IFAs suggested that K13 was assembled in merozoites within schizonts and transferred to newly formed rings (Figure 2F).

Because our prior studies suggested that PI3P targets PfEMP1 to the red cell,²⁹  we examined the distribution of K13 in relation to PfEMP1. A monospecific “pan-PfEMP1” antibody was developed to the conserved acidic C-terminal sequence or ATS present in all members of this antigenic family (supplemental Figure 2D-E). As is shown in Figure 2F-H, in IFAs, PfEMP1 protein appeared closely apposed to K13 (Pearson correlation coefficient of 0.7-0.8; supplemental Figure 2F-G) in merozoites inside segmenters and released extracellularly, suggesting that K13/PI3P vesicles may affect PfEMP1 biology at these stages and new rings (which form minutes after merozoite release).

Pf3D7 lysates prepared from segmenters were optimized for extraction of PfEMP1 (as is summarized in supplemental Figure 3A-D; supplemental Methods). Immunopurification using anti-ATS antibodies (or mock without primary antibodies) was carried out under high (and stringent) concentrations of both nonionic and chaotropic detergents (1% Tx-100 and 1.0% deoxycholate; supplemental Figure 3D, supplemental Methods) to minimize nonspecific protein-protein interactions with beads coated with protein G (Figure 3A; supplemental Figure 3A-D, supplemental Table 1). Two experimental replicates revealed an immunoproteome of 503 shared proteins (Figure 3B), which included K13 and comprised 88% of the total protein mass identified (Figure 3B; supplemental Table 1).

Because the PfEMP1 immunopurification was carried out under stringent conditions, if a protein was represented in each proteome, it was included for analyses in the shared set of 503. To facilitate comparative analyses, we determined the abundance of a protein in each immunoproteome and calculated an abundance ratio (AR; supplemental Methods) as a measure of the fidelity of protein association across replicate proteomes. As is shown in Figure 3C (and supplemental Table 1), over 40% of proteins were accommodated within a twofold range (AR, 0.5 to 1.99), and ∼70% were within threefold range. Importantly, K13, BiP (a major ER chaperone and UPR sensor), and protein disulfide isomerase (PDI) (a second major ER-redox protein) showed AR’s of 0.89, 1.29, and 1.98, respectively.

Hypergeometric analyses suggested that the “503-proteome” was highly enriched in multiple pathways of protein homeostasis or “proteostasis” that control the biogenesis, folding degradation, and trafficking of proteins within the parasite and exported to the host cell, regulated by protein quality control and folding in the ER and cytoplasm (supplemental Figure 3E-F). There was no significant enrichment of apicoplast and mitochondria: significant overlap was seen with the nucleus, but the P values suggested preferential enrichment of ER and cytoplasmic proteostatic pathways. The 503-proteome showed significant intersection with transcriptomic changes associated with artemisinin-resistant clinical isolates¹²  (P ≤ 4.7e-3; Figure 3D; supplemental Figure 3G). This overlap was enriched in 487 (top 9.6%) upregulated transcripts¹²  (P ≤ 3.2e-5; Figure 3D; supplemental Figure 3G) but not with the 511 genes (10.1%) that were downregulated. Finally, in the AR 0.5 to 1.99 core of a 207-proteome, there was no statistically significant intersection with the nucleus or the UPR, but several systems of ER/cytoplasmic proteostasis, including protein translocation export (PTEX) host-export components, continued to be enriched (supplemental Figure 3H-I) and comprised ∼50% of the proteome intensity (Figure 3E; supplemental Table 5). BiP, the key UPR sensor, was a major component of the 207-proteome (Figure 3E; supplemental Table 5; note that hypergeometric analyses do not account for intensity or AR). The data shown in Figures 2 and 3 support the molecular interaction of PfEMP1 with K13: associated PI3P tubules/vesicles of the ER that encompass protein proteostasis functions and augment export to the host.

We previously reported that increasing PI3P levels by transgenic expression of human VPS34 in Pf3D7 parasites to yield Pf3D7VPS34myc confers artemisinin resistance in the absence of the K13 mutation, whereas Pf3D7VPS34AAAmyc expressing a catalytic dead VPS34 (in which active-site DRH was replaced by AAA) showed no increase in PI3P and remained artemisinin sensitive.¹⁰  Therefore, PfEMP1 immunoproteomes were stringently isolated and characterized from Pf3D7VPS34myc and Pf3D7VPS34AAA (Figure 4A-D; supplemental Figure 4A-C; supplemental Tables 7 and 8). The proteomes were similar in composition, with an overlap in 85% of their composite proteins (Figure 4E). Both contained 95% of the (503) PfEMP1 immunoproteome isolated from Pf3D7 but were also larger (705 proteins for Pf3D7VPSmyc and 770 proteins for Pf3D7VPSAAA, primarily due to increased proteosomal components; supplemental Figure 4D-E, supplemental (Tables 7 and 8). Hypergeometric analyses revealed that the Pf3D7VPS34myc immunoproteome was significantly enriched in 487 (top 9.6%) upregulated genes of the clinical transcriptome¹²  (P < 1.6e-7; Figure 4G; supplemental Figure 4F). Notably, this enrichment included 94% of 3D7 immunoproteome, which showed a significant intersection with the clinical proteome. Moreover, as with its 3D7 counterpart, the immunoproteome of Pf3D7VPS34myc was not enriched in the 511 genes (10.1%) downregulated in the clinical transcriptome (Figure 4G; supplemental Figure 4F). Similar results were obtained for the PfEMP1 immunoproteome of Pf3D7VPSAA (as was expected on the basis of its similarity in composition to that of Pf3D7VPS34myc; supplemental Figure 4G). These findings suggested that the major proteostatic characteristics of PfEMP1 immunoproteome of Pf3D7 are preserved in those from the Pf3D7VPS transgenic strains. Moreover, the composition of the vesicle immunoproteome is similar in resistant and sensitive (Pf3D7VPS) strains. However, elevation of PI3P may expand vesicles to confer resistance in Pf3D7VPSmyc, implicating vesicular expansion (of proteostatic mechanisms, including UPR, via chaperone complexes; Figure 4H) in artemisinin resistance.

To directly examine vesicle expansion upon K13 mutation, we returned to cryo-IEM. Because our analyses of ER-PI3P (Figures 1 and 2) was initiated in the PfNF54 (wild-type PfNF54WT) strain, we compared findings in the matched-resistant counterpart PfNF54K13C580Y (carrying the major mutation reported in 80% of isolates in southeast Asia and created by CRISPR/Cas9 insertion of a single, chromosomal point mutation in PfNF54K13WT⁸ ). In comparison with their wild-type counterparts, K13C580Y mutants show an intense PI3P label in the ER and all over the parasite (Figure 5A-C; supplemental Figure 5A-C, including inside nuclei; Figure 5B). ER tubules and vesicles of mutants become amplified and heavily labeled with PI3P (Figure 5C). Quantitation revealed a threefold increase in total gold label at schizont stages (Figure 5D: 0-3 hours). The 0- to 3-hour ring stages of PfNF54K13C580Y contain 2.7 times more PI3P than do wild-type parasites,¹⁰  suggesting that mutants show comparable PI3P elevation in both schizont and new rings. There was no change in the labeling of ER marker BiP (Figure 5E) or K13 (Figure 5F; K13 stereological analyses in mutants are not shown). Rather, mutants contained large regions of dense PI3P label, which are devoid of PfK13 (Figure 5F, double red arrows) not seen in wild-type parasites (in which K13-free PI3P are small clusters, shown by the single red arrow in Figure 5F). Stereological analyses suggested that most of the PI3P (∼60%) remains associated with the ER (Figure 5G; although its relative amount was threefold higher in the mutant; Figure 5H). PI3P is seen at the nuclear envelope (an extension of the ER)/nucleus in wild-type of parasites. Nuclei of emerging daughter merozoites concentrated PI3P in mutants (Figure 5B). But the biggest fold change in PI3P in mutants was in the ER (threefold; Figure 5H) plasma membrane and other regions (which include the apicoplast and fv; 10-fold; Figure 5H). RNA sequencing analyses and Western blots confirmed that transcript and total cellular K13 protein levels are not altered between wild-type and mutants (even in 0- to 3-hour parasites; Figure 5I-J; consistent with findings that K13 transcripts were unchanged in clinical isolates¹² ). Together, these data suggest that an increase in PI3P tubules and vesicles in the major K13C580Y mutant closely corresponds to fold changes in cellular levels of this lipid, suggesting that they are parasite major effectors of artemisinin resistance. Furthermore, in addition to its distribution through the entire parasite, PI3P was exported to vesicular structures and “Maurer’s clefts” in the cytoplasm of the infected red cell (Figure 5K-L, red arrows), known to be intermediates of parasite-protein export to the red cell membrane.^32,33 

Dissemination of PI3P to intraerythrocytic vesicles and Maurer’s clefts in Figure 5K-L suggested that PfNF54K13C580Y mutants may affect export of parasite proteins to the host red cell. This led to the idea of tracking parasite protein export as a surrogate for PI3P vesicular amplification. This was particularly important for analyses of 0- to 3-hour rings, the stages of peak artemisinin resistance, but not amenable to quantitative IEM needed to assess quantitative distribution of the lipid PI3P. On the basis of our findings, PfEMP1 is a marker for K13/PI3P vesicles, but little is known about its expression in newly formed rings or modulation by short-term exposure to DHA, a potent PfPI3Kinase inhibitor that suppresses ring PI3P.¹⁰  Therefore (and as summarized in Figure 6A), studies were initiated in which 0- to 3-hour rings were exposed to DHA treatment (6 hours at 4 nM). We detected no significant effect on expression of either the main PfEMP1 (VAR2CSA) or additional minor PfEMP1 genes owing to mutation or drug (Figure 6B; supplemental Figure 6A-B). PfEMP1 export to the red cell was detected in the young rings (Figure 6C). Export was substantially reduced by 4 nM DHA in PfNF54K13WT parasites (Figure 6C-D). Quantitative analyses of 400 optical sections suggested a greater than 50% reduction in export (note that PI3P levels cannot be entirely depleted without compromising cell viability). In contrast, in PfNF54K13C580Y, 4 nM DHA failed to block PfEMP1 export (Figure 6C-D), suggesting sustained (and resistant) vesicular export in a strain in which PI3P is elevated. DHA inhibited PfEMP1 export in a sensitive Cambodian clinical strain (ANL-1) but not the clinically resistant counterparts ANL-2 K13C580Y and ANL-9 R539T (Figure 6E-F; supplemental Figure 6C). Removal of DHA restored export in sensitive ANL1 strains (supplemental Figure 6D). Additional PI3K inhibitors mimicked the effect of DHA on PfEMP1 export in sensitive and resistant clinical strains (supplemental Figure 6E).

In the P falciparum CS2 strain,^34,35  the export of VAR2CSA (a ligand for host chondroitin sulfate; CSA) was blocked by DHA and other PfPI3Kinase inhibitors (Figure 7A; supplemental Figure 7A-B). The conservation of the VAR2CSA promoter and gene suggest that its lack of responsiveness to DHA at a transcriptional level in PfNF54K13WT and PfNF54K13C580Y (Figure 6B; supplemental Figure 6A-B) should be conserved across strains.³⁶  Because expression of VAR2CSA at the infected red cell surface confers adherence to chondroitin sulfate, we were interested in testing the effects of K13 mutation on cytoadherence of infected red cells. To do so, we created transgenic CS2 parasites expressing dominant-negative forms of HA-tagged K13C80Y and K13R539T (a strategy shown to be active in Pf3D7¹⁰ ; construct design in supplemental Figure 7). Neither mutation had a significant effect on adhesin levels. Cytoadherence was unaffected by C580Y and may have been slightly increased by R539T (Figure 7B-C). DHA blocked cytoadherence of the parent CS2 strain but not of either K13 mutant and was without effect on VAR2CSA protein levels across all 3 strains. Transgenic expression of a second K13WT gene in the CS2 strain reduced levels of both cytoadherence and VAR2CSA. One explanation may be that K13 increases proteasomal targeting of PfPI3K to affect PI3P turnover, which in turn decreases VAR2CSA levels. DHA further depressed binding of transgenic K13WT strains (Figure 7C, hatched box; supplemental Figure 7) but without additional suppression of the VAR2CSA protein (Figure 7D). VAR2CSA levels in K13C580Y and K13R539T were not elevated in comparison with parental CS2 parasites, suggesting that PfEMP1 is not likely to be a substrate for K13. Mok et al¹²  reported that resistant clinical isolates progress from rings through trophozoite and schizont stages more slowly than do their sensitive counterparts. Because PfEMP1 export and adhesion increase at later stages of growth, the slower progression of mutants does not account for their higher levels of DHA-insensitive export of PfEMP1. Rather, it appears that PfEMP1-export dynamics in mutants with elevated PI3P are no longer suppressed by K13 regulation or PfPI3K inhibition. Together, our findings (shown in Figures 6 and 7) suggest that PfEMP1-export dynamics provide a surrogate for detecting PI3P-amplified vesiculation in early ring stages of K13 mutants across strains: both export of PfEMP1 to the red cell and cytoadherence of the infected red cells (later in life cycle) gain artemisinin resistance.

Our studies reveal several new and unexpected findings on the mechanisms and consequences of the causal marker for artemisinin-resistant K13’s action in P falciparum and its host red blood cell. We show that K13 is a marker for parasite PI3P and associates with the virulence-determinant PfEMP1 exported to the red cell. K13 mutations of artemisinin resistance that increase PI3P also regulate the dynamics of PfEMP1 export to the host cell. The PfEMP1 K13-containing proteome was enriched in host-targeted PTEX machinery,^37,38  and several convey cargo with relevant export signals.^39-41  However, it inexplicably lacked low molecular weight and large antigenic families of rifins and stevors (instead of enriching them), suggesting that there may be additional steps of export selectivity that remain poorly understood.^32,33,42  Importantly, PTEX may not regulate PI3P elaboration within the parasite, a primary resistance survival effect of K13 mutation. Dissemination of PI3P vesicles of proteostasis may also provide mechanisms for mitigating artemisinin damage to the host red cell. Further PfEMP1 export and cytoadherence become DHA resistant in K13 mutants (as detected in static assays, suggesting that even greater differences are expected between wild-type and mutant parasites under circulatory conditions). Therefore, in addition to protecting the infected red cell against proteopathy, resistant mutations may provide better adherence to host receptors and immune evasion, even in the presence of a drug. This may contribute to the persistence of resistant parasites at low and asymptomatic parasitemias despite mass drug administration.^43,44 

We also find that K13 and the parasite PI3P enrich at the ER and to a lesser degree fv and apicoplast. K13 is cytoplasmic, whereas PI3P also locates in the lumen. Because the parasite encodes for a single cytosolic PfPI3K,¹⁰  luminal import of PI3P may result from endovesiculation or yet-to-be-defined ATP-dependent lipid translocases (in yeast, PI3P is detected in the lumen of the autophagosome³¹ ). K13’s physiological function may be to control PI3P levels and turnover in the ER (and secondarily at the fv and apicoplast). PfEMP1, K13, and K13-mutant PI3P-tubovesicular enrichment in schizonts may be transferred via daughter merozoites into newly formed 0- to 3-hour rings, the stage of clinical artemisinin resistance at which PI3P is correspondingly increased and PfEMP1 export provides a protein surrogate to detect amplified vesicular export. K13 mutants at late trophozoite/schizont stages do not manifest artemisinin resistance in the RSA, but their proteostasis pathways are altered, and killing by artemisinins is maximal, suggesting the need to understand properties of K13 mutants at these stages (or correlates thereof) in in vivo human infections.

Finally, our data support the idea that amplified PI3P tubovesicles encompass protein proteostasis pathways and UPR and confer resistance, presumably by widely disseminating resistance intermediates to mitigate promiscuous artemisinin-mediated “proteopathy” (via protein alkylation and aggregation) throughout the infected red cell.^15,22  In higher eukaryotes, PI3P expansion in the ER stimulates membrane curvature needed for macroautophagy to efficiently remove misfolded/toxic protein aggregates^45-47  as well as unconventional autophagic secretion.⁴⁸  However, ER autophagy remains poorly understood in malaria parasites,^49,50  and its mechanistic analyses need further study, as do the functions of additional phosphoinositide genes recently identified as artemisinin resistance-associated malarial-gene loci.⁵¹  Elaborated PI3P tubules/vesicles reach all destinations in mutant infected red cells (Figure 7E), which may explain the complexity of artemisinin resistance as well as how a single K13-proteostatic determinant may reflect selective pressure in multiple organellar systems and hundreds of parasite genes implicated in resistance.^15-23  Because PI3P elevation is sufficient for inducing resistance, expansion of PI3P vesicles may also account for K13-independent resistance.⁵²  The vesicular resistance mechanism is likely to be effective against new drugs that induce toxicity through protein aggregates/proteopathy^53,54  and may present many targets to facilitate resistance to partner drugs in multidrug-resistant malaria.^55,56  Because PI3P also targets the red cell in K13 mutants, host factors may modulate artemisinin resistance, unlike known resistance mechanisms of major antimalarials presently in clinical use.⁵⁷ 

All parasite gene/protein sequences were obtained from PlasmoDB (www.plasmodb.org). The authors would like to thank the technical competence of Kimberley Zichichi from the Microscopy Facility at Yale University.

This work was supported by grants from the National Institutes of Health, National Heart, Lung, and Blood Institute (HL069630, HL130330 [K.H.]) and National Institute of Allergy and Infectious Diseases (AI103548) (I.C.), the government of India (DST ECR/2015/000387), Department of Biotechnology Ramalingaswami Re-entry Fellowship (BT/HRD/35/02/2006), and University for Potential Excellence–II (Project ID 245) (S.B.). Support of the Wistar Proteomics and Metabolomics Core Facility was provided by a grant from the Cancer Center Support to the Wistar Institute (CA010815).

Contribution: S.B., I.C., A.M., N.S., M.G., Z.S., K.H., H.-Y.T., and D.W.S. designed, performed, and interpreted the experimental work; R.V.S. helped design and interpret the experimental work; K.H., S.B., I.C., A.M., N.S., H.-Y.T., D.W.S., R.V.S., and N.M. wrote the manuscript; I.S. and N.M. provided key intellectual insight into aspects of this study; I.S. conducted and reviewed all of the statistical analyses; and all authors commented on the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

The current affiliation for A.M. is Institut Pasteur of Dakar, Dakar, Senegal.

The current affiliation for M.G. is Wellcome Trust Sanger Institute, Cambridge, United Kingdom.

The current affiliation for Z.S. is Department of Chemical Engineering, University of Chemistry and Technology Prague, Prague, Czech Republic.

The current affiliation for R.V.S. is Department of Medicinal Chemistry and Molecular Pharmacology, Purdue University, West Lafayette, IN.

Correspondence: Kasturi Haldar, Boler-Parseghian Center for Rare and Neglected Diseases, University of Notre Dame, 103 Galvin Life Sciences, Notre Dame, IN 46556; e-mail: khaldar@nd.edu; and Souvik Bhattacharjee, Special Centre for Molecular Medicine, Jawaharlal Nehru University, New Delhi 110067, India; e-mail: souvik@jnu.ac.in.