Plasmodium falciparum quinine resistance is multifactorial and includes a role for the drug/metabolite transporters PfCRT and DMT1

Abstract

The genetic basis of Plasmodium falciparum resistance to quinine (QN), a drug used to treat severe malaria, has long been enigmatic. To gain further insight, we used FRG-NOD human liver-chimeric mice to conduct a P. falciparum genetic cross between QN-resistant (Cam3.II) and QN-sensitive (NF54) parasites, which also differ in their susceptibility to chloroquine (CQ). By applying different selective conditions to progeny pools prior to cloning, we recovered 120 unique recombinant progeny. Drug profiling and quantitative trait loci analyses of the progeny revealed predominant peaks on chromosomes 7 and 12 associated with CQ and QN resistance, that is consistent with a multifactorial mechanism of resistance for these compounds. CQ and monodesethyl-CQ (md-CQ) resistance mapped to a chromosome 7 region harboring pfcrt as expected. However, for QN, resistance mapped to a dominant chromosome 7 peak centered 295 kb downstream of pfcrt, with pfcrt showing a smaller peak. We identified the drug/metabolite transporter 1 (DMT1) as the top chromosome 7 candidate due to its structural similarity to PfCRT and proximity to the peak. Deleting DMT1 in QN-resistant Cam3.II parasites significantly sensitized the parasite to QN but not to the other drugs tested, suggesting that DMT1 mediates QN response specifically. We localized DMT1 to structures associated with vesicular trafficking, as well as the parasitophorous vacuolar membrane, lipid bodies, and the digestive vacuole. We also observed that mutant DMT1 transports more QN than the wild-type isoform in vitro. Gene editing confirmed an additional role for mutant PfCRT in mediating QN resistance. In addition, we identified an ATP-dependent zinc metalloprotease (FtsH1) as one of the top candidates in the chromosome 12 locus and confirmed its role as a potential mediator of QN resistance and a modulator of md-CQ resistance using CRISPR/Cas9 SNP-edited lines. Interestingly, this chromosome 12 region mapped to resistance to both CQ and QN and was preferentially co-inherited with pfcrt. Our study demonstrates that DMT1 is a novel marker of QN resistance and that a new chromosome 12 locus associates with CQ and QN response, with ftsh1 as a potential candidate, suggesting these genes in addition to pfcrt should be genotyped in surveillance and clinical settings.

Extended Data 1 |. Quinine inhibits β-hematin formation to a lesser degree than chloroquine but more than mefloquine in a cell-free system.

a, Chemical structures of the antiplasmodial compounds used in this study. Quinine, lumefantrine, and mefloquine are aryl-amino alcohols; chloroquine, monodesethyl-chloroquine, and monodesethyl-amodiaquine are 4-aminoquinolines; doxycycline is a tetracycline-class antibiotic; pyrimethamine is an antifolate; dihydroartemisinin is an active metabolite of artemisinin, a sesquiterpene lactone endoperoxide. Quinine and mefloquine have the aminoquinoline ring found in chloroquine, monodesethyl-chloroquine, and monodesethyl-amodiaquine. Chemical formulas, SMILES, and Tanimoto scores of these compounds are indicated in Table S13. b, Representative concentration-dependent inhibition of β-hematin formation by quinine, chloroquine, md-amodiaquine, mefloquine, two non-heme detoxification pathway-acting negative control antimalarials (doxycycline and pyrimethamine), and DMSO. Mean ± SEM IC₅₀ values from N=3 experiments are listed in Table S12.

Extended Data 2 |. Workflow of the NF54×Cam3.II genetic cross, isolation of progeny, and bulk selection vs. progeny clone-based trait mapping.

a, Overview of the genetic cross performed using Anopheles stephensi mosquitoes and FRG NOD human-liver chimeric mice (N=4). RBCs, red blood cells. b, Drug selection conditions were applied to progeny bulks, followed by a drug-free recovery period after which bulk genomic DNA was harvested and/or the progeny were cloned out. c,d, Genetic determinants of drug resistance were identified using progeny clone-based linkage approaches (c) or bulk selection approaches with bulk segregant analysis (d). Images were created with BioRender.com. The NF54 and Cam3.II parents are described in Fig. 1a, the mosquito and mouse infections are described in Table S1, and the total number of clonal progeny obtained are indicated in Table S15.

Extended Data 3 |. Genome-wide frequencies and pfcrt, dmt1, thzk/ftsh1, and samc allele representation in the recombinant progeny

a, Percentage of NF54 and Cam3.II allele frequencies averaged across the genome in 13,116 SNP positions for each of the parents and 120 recombinant progeny, ordered by haplotype number. b, Number of unique recombinant progeny with the wild-type NF54 or mutant Cam3.II alleles for pfcrt, dmt1, thzk/ftsh1, and samc. The mutant Cam3.II alleles for each gene are listed in Fig. 3a.

Extended Data 4 |. Bulk segregant analyses of progeny pools pressured with quinine or no-drug controls.

a,b, Genetic loci enriched by quinine (QN) in drug-no drug pairwise comparisons. Black lines are G’ values comparing allele frequency between QN-pressured bulk progeny and control non-drug-pressured bulk progeny. The red line indicates the False Discovery Rate at the indicated threshold. Samples are described in Table S5. Quantitative trait locus regions are described in Table S6. Of note, mutant pfcrt/mutant ftsh1, or WT pfcrt/WT ftsh1, continued to be co-inherited in the QN-pressured progeny, as was seen in the non-drug-pressured progeny.

Extended Data 5 |. Phenotypic response of parents and recombinant progeny to chloroquine and monodesethyl-chloroquine closely align.

a, Chloroquine (CQ) response as measured by mean ± SEM IC₅₀ values. b, Logarithm of odds (LOD) plot for geometric mean CQ IC₅₀ values. The red line indicates the 95% probability threshold. c, Overlay of monodesethyl-chloroquine (md-CQ) and CQ LOD plots. A/M, apicoplast/mitochondria (these are always coinherited). The highest LOD peaks for CQ and md-CQ align with pfcrt on chromosome 7 and the segment including ftsh1 on chromosome 12. Note that dmt1 was not within the dominant chromosome 7 peak for CQ or md-CQ, unlike for quinine.

Extended Data 6 |. Clinical isolates harbor mutations in genes of interest within the chromosome 7 QN and chromosome 12 quinine (QN), chloroquine (CQ), and monodesethyl-CQ (md-CQ) quantitative trait locus segments.

a-d, Frequency of the Cam3.II mutations in clinical isolates (Pf3k) of genes in chromosome 7 ((a), dmt1: PF3D7_0715800) and chromosome 12 ((b) ftsh1: PF3D7_1239700, (c) samc: PF3D7_1241600, (d) thzk: PF3D7_1239600)) loci, grouped by sampling country. Yellow indicates isolates with the NF54 wild-type haplotype covered by the indicated SNP positions; green, the Cam3.II mutant haplotype; light grey, mixed call; dark grey, genotype not called; orange or blue, single-amino acid mutants not corresponding to the double-mutant Cam3.II haplotype. (N=50 (Bangladesh), 570 (Cambodia), 85 (Laos), 60 (Myanmar), 148 (Thailand), 97 (Vietnam), 369 (Malawi), 113 (Democratic Republic of the Congo), 617 (Ghana), 100 (Guinea), 96 (Mali), 5 (Nigeria), 137 (Senegal), 65 (The Gambia); total=2,512 samples). These four top candidate genes are further described in Fig. 3a.

Extended Data 7 |. PfDMT1 is structurally similar to another drug metabolite transporter superfamily protein, PfCRT.

Side-by-side comparison of the AlphaFold-predicted NF54 DMT1 structure and the cryo-EM-derived 7G8 PfCRT structure. The predicted template modeling (pTM) is indicated for the AlphaFold NF54 DMT1 structure, which exceeds the 0.5 threshold for predicted fold reliability. The Y107 and S129 positions are indicated on the AlphaFold DMT1 structure. Root mean square deviation (RMSD) scores calculated in ChimeraX MatchMaker are indicated for PfCRT and DMT1.

Extended Data 8 |. DMT1 shows localization to structures associated with vesicular trafficking as well as the digestive vacuole.

a-i, Representative immunofluorescence assays (IFA) showing DMT1–3×HA-tagged parasites stained with α-HA (DMT1, green) and DAPI (nuclear, blue), as well as antibodies or dyes: (a) MitoTracker Red CMXRos (mitochondria, red), (b) α-PfACP (apicoplast), (c) α-Rab5B (early endosome), (d) α-BiP (ER), (e) α-PfEXP2 (parasitophorous vacuolar membrane), (f) α-PfCRT (digestive vacuole membrane), (g) α-PfPM2 (digestive vacuole), (h) Nile Red (lipid bodies), or (i) α-Rab11A (post-Golgi). IFA images are also shown as 3D volume reconstructions (right-most column). Scale bars: 1 μm for IFA (black) or 0.5 μm for 3D volume reconstructions (white). j, Mean ± SEM percent DMT1–3×HA colocalization between DMT1 and the aforementioned organelles. Colocalization p-values (Student’s t-test): *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. Numbers of individual parasites analyzed and mean ± SEM and p-values for two different colocalization metrics (volume and Manders’ coefficient) are described in Table S11.

Fig. 1 |. A P. falciparum NF54×Cam3.II genetic cross yielded 120 independent recombinant progeny with diverse genomic and drug resistance profiles.

a, NF54×Cam3.II genetic cross parent information. ^aDd2 PfCRT haplotype: M74I/N75E/K76T/A220S/Q271E/N326S/I356T/R371I. k13: kelch13 (PF3D7_1343700); pfcrt: chloroquine resistance transporter (PF3D7_0709000); pfmdr1: multidrug resistance protein 1 (PF3D7_0523000). ADQ: amodiaquine; ART: artemisinin; CQ: chloroquine; LMF: lumefantrine; md-ADQ: mono-desethyl amodiaquine; md-CQ: monodesethyl-chloroquine; MFQ: mefloquine; md-CQ: monodesethyl-chloroquine; PPQ: piperaquine; QN: quinine. R, resistant; R^mod, moderate resistance; S, sensitive; S^hi, high sensitivity; S^lo, low sensitivity. Geometric mean IC₅₀ values in nM are listed in the parentheses (N=4–85 with technical duplicates). b, Recombinant progeny representing each of 120 unique haplotypes were hierarchically clustered by similarity based on their NF54 (yellow) or Cam3.II (green) alleles at 13,116 SNP positions within each selection condition: no drug pressure, chloroquine (CQ) 50 nM or 75 nM, or quinine (QN) 75 nM, 95 nM, 140 nM, 180 nM, (95nM+140nM), (95nM+180nM), or (95nM+140nM+240nM), in that order. Some haplotypes were obtained in more than one selection condition (Table S4). NF54 (H001) and Cam3.II (H002) parents are shown on top. H, haplotype; A/M, apicoplast/mitochondria (these were always coinherited); black line, pfcrt; red line, dmt1; grey line, ftsh1; red box, QN IC₉₀, md-CQ IC₉₀, CQ IC₅₀ and IC₉₀ chromosome 12 QTL locus. c-e, Mean ± SEM QN IC₅₀ and IC₉₀ values are shown in c and d, respectively, and md-CQ IC₅₀ values are shown in e (N,n=1–85, with a majority of >4, with technical duplicates). Data were generated with parasites cultured in media with Albumax (no serum).

Fig. 2 |. QTL mapping of QN, CQ, and md-CQ progeny reveals a non-pfcrt-centered chromosome 7 QN IC₉₀ peak and an additional shared peak on chromosome 12.

a-c, LOD plots for mean QN IC₅₀ (a), QN IC₉₀ (b), and md-CQ IC₅₀ (c) values_. The red line indicates the 95% probability threshold. d, md-CQ, CQ, and QN LOD plots overlaid. QTL, quantitative trait locus; LOD, log of the odds; md-CQ, monodesethyl-chloroquine; CQ, chloroquine; QN, quinine; IC₅₀ and IC₉₀, 50% and 90% growth inhibitory concentrations. e, List of significant QTL segments identified by progeny clone-based analysis. chromosome 7: pfcrt (dark green line), dmt1 (pink line). chromosome 12: thzk and ftsh1 (light green line).

Fig. 3 |. Identification of the top candidate genes from the QN chromosome 7 and QN, CQ, md-CQ chromosome 12 QTL peaks.

a, Top candidate genes from progeny clone-based QN, CQ, and md-CQ QTL analyses. All genes are expressed in the P. falciparum ABS parasites. QTL, quantitative trait locus; SNP, single-nucleotide polymorphism; QN, quinine; CQ, chloroquine; md, monodesethyl; CDS, coding sequence; KO, knockout; CNV, copy number variants; ND, not determined; N/A, not applicable. ^aDd2 PfCRT haplotype: M74I/N75E/K76T/A220S/Q271E/N326S/I356T/R371I. ^bref. 42, the closer the mutant fitness score is to −4, the higher the predicted fitness cost. ^cref. 101. ^dPersonal communication with Dr. Susan Wyllie. ^eref. 41. ^fref. 40. b-e, QN (b,c) and md-CQ. d,e, IC levels for the profiled unedited cross progeny and two parents, grouped by pfcrt, dmt1, thzk, ftsh1, and samc genotypes. Statistical significance was determined by Mann-Whitney U tests (N,n=1–27, with technical duplicates). **p < 0.01, ***p < 0.001, ***p < 0.0001.

Fig. 4 |. ftsh1 is a candidate gene for the chromosome 12 QTL locus.

a-f, Quinine (a,b), chloroquine (c,d), and md-Chloroquine (e,f) response in samc and ftsh1 SNP-edited Cam3.II as measured by mean ± SEM IC₅₀ or IC₉₀ values. For each parasite, the parental unedited parasite (endogenous haplotype) (dark color) and the SNP-edited parasite (wild-type revertant parasite) (light color) were profiled. p-values (Student’s t-test) are indicated (N,n=6–16,2). *p < 0.05; **p < 0.01; ***p < 0.001.

Fig. 5 |. Cam3.II parasites lacking DMT1 or expressing WT PfCRT are significantly sensitized to QN.

a, Quinine response of DMT1 SNP-edited parents and progeny as measured by mean ± SEM IC₅₀ values. For each parasite, the parental unedited parasite (endogenous DMT1 haplotype) and the SNP-edited parasite (Y107N / S129L mutant or wild-type revertant haplotype) are indicated by dark and light colors, respectively. p values (Student’s t-test) are indicated (N,n=2–8,2). WT, wild-type; YNSL, Y107N / S129L mutation. b, Quinine, mefloquine, lumefantrine, chloroquine, md-Chloroquine, md-Amodiaquine response in Cam3.II DMT1 knockout parasites as measured by mean ± SEM IC₅₀ values. p-values (Student’s t-test) are indicated for the Cam3.II KO strain vs the Cam3.II parent (N=4–21, with technical duplicates). c-d, Quinine (c) and chloroquine (d) response in Cam3.II pfcrt^WT revertant parasites as measured by mean ± SEM IC₅₀ and IC₉₀ values. p-values (Student’s t-test) are indicated for the Cam3.II pfcrt^WT strain vs the Cam3.II parent (N=4–5, with technical duplicates). *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

Fig. 6 |. DMT1 shows localization to structures associated with vesicular trafficking as well as the digestive vacuole and exhibits differential uptake of [³H]QN between the NF54 and Cam3.II isoforms.

a-g, Representative immunofluorescence assays (IFA) showing DMT1–3×HA-tagged parasites stained with antibodies or dyes: α-HA (DMT1, green), DAPI (nuclear, blue). a, α-Rab7 (late endosome), b, α-Rab5B (early endosome), c, α-BiP (ER), d, α-PfK13 (ER, vesicles, near cytostome), e, α-PfEXP2 (parasitophorous vacuolar membrane), f, α-PfCRT (digestive vacuole membrane), or g, Nile Red (lipid bodies). IFA images are shown as 3D volume reconstructions. Scale bars: 1 μm for IFA (black), 0.5 μm for 3D volume reconstructions (white). h,i, Time course of 300 nM [³H]QN and 300 nM [³H]CQ uptake measured with PfCRT (NF54 or Cam3.II), DMT1 (NF54 or Cam3.II), or Band 3-containing proteoliposomes. Data are mean ± SEM of n = 3 independent experiments with technical duplicates. Statistical significance between NF54 and Cam3.II variants for each protein was determined by unpaired Student’s t-tests corrected for multiple testing (h, i). *p < 0.05, **p < 0.01, ****p < 0.0001.