Collateral hypersensitivity between ZY19489 and piperaquine neutralizes PfCRT-mediated drug efflux and Plasmodium falciparum resistance

Abstract

New antimalarial drugs are needed to combat the current emergence and spread of Plasmodium falciparum parasite resistance to artemisinin-based combination therapies. Here, we characterize ZY19489, a triaminopyrimidine presently in a Phase Ib clinical trial. Asexual blood-stage parasites pressured with ZY19489 acquire low-grade resistance, mediated by a novel mutation in the P. falciparum chloroquine resistance transporter (PfCRT) that causes slow growth rates and a substantial fitness cost. ZY19489-resistant parasites lose their chloroquine resistance status and become hypersusceptible to piperaquine (PPQ), an artemisinin-based combination partner drug. Uptake studies in proteoliposomes loaded with drug-resistant PfCRT isoforms demonstrate that ZY19489 can block mutant PfCRT-mediated PPQ and chloroquine transport. In parasites, PfCRT mutant variants can mediate PPQ and chloroquine resistance via their efflux out of the digestive vacuole. Our findings evoke a scenario of an evolutionary trap whereby resistance to ZY19489 can block PPQ and chloroquine efflux and thereby restore their activity. Metabolomic studies show that ZY19489 leads to significantly reduced intracellular levels of short hemoglobin-derived peptides (a natural substrate of PfCRT) and accumulation of pyrimidine deoxynucleotides. Our results present a possible marker for tracking the evolution of clinical resistance to ZY19489 and a rationale for pairing this molecule with PPQ to generate a resistance-refractory combination.

ZY19489 is a potent antimalarial candidate, which shows a low resistance liability, mediated by a fitness-compromising mutation in PfCRT that restores Plasmodium falciparum susceptibility to the licensed drug piperaquine by blocking its efflux.

Introduction

The World Health Organization recently reported an estimated 282 million cases and 610,000 malaria-related deaths globally in 2024¹, highlighting the continuing neutral to upward trend in malaria infections since 2016. This upsurge in disease burden has been, in part, due to the spread of Plasmodium falciparum alleles associated with loss of parasite sensitivity to artemisinin-based combination therapies (ACTs) and, more recently, the emergence of parasites with deletions in histidine-rich proteins 2 and 3 (pfhrp2/3) that confound HRP2-based case detection^1–4. Currently approved malaria vaccines, which offer moderate protection, are not yet widely available^5,6. Therefore, there is an urgent need for new, safe and efficacious antimalarial drugs endowed with distinct modes of action (MoA) to complement current treatment options.

Recent antimalarial drug discovery efforts aim to develop new medicines that can circumvent emerging drug resistance, whilst ensuring longer efficacy and improved dosing convenience. These efforts have yielded several new molecules with novel mechanistic profiles that are currently in various phases of clinical development (https://www.mmv.org/mmv-pipeline-antimalarial-drugs). One such candidate is ZY19489 (Fig. 1A), a fast-killing and long-acting triaminopyrimidine identified from a high-throughput screen of an AstraZeneca library comprising 500,000 compounds tested against P. falciparum asexual blood-stage (ABS) parasites using high-content imaging⁷. This compound (also known as Zintrodiazine, MMV674253, AZ412, or TAP12) showed good safety, pharmacokinetic, and efficacy results in a first-in-human clinical trial⁸. ZY19489 also retains potency against P. falciparum field isolates expressing mutant kelch13 alleles that confer partial resistance to artemisinin⁹. Currently, ZY19489 is in a Phase Ib trial in combination with the 4-aminoquinoline, ferroquine, for treatment of uncomplicated P. falciparum malaria (https://classic.clinicaltrials.gov/ct2/show/NCT05911828). If approved, this combination could constitute an important non-ACT treatment option in areas where artemisinin partial resistance has emerged¹⁰.

Fig. 1. Plasmodium falciparum chloroquine resistance transporter (PfCRT) mediates in vitro resistance to ZY19489.

A Chemical structures of ZY19489, a structurally-analogous triaminopyrimidine derivative referred to as compound 6, and the 4-aminoquinoline, piperaquine (PPQ). B Drug susceptibility profiles of a ZY19489-selected bulk culture (gray) and individual clones (C8, E3, F7 and G2) compared to their Dd2-Polδ parent (black) presented as mean ± SEM IC₅₀ values calculated from 5 to 6 independent experiments with technical duplicates (N, n = 5 − 6, 2). Statistical significance was determined using two-tailed Mann-Whitney U tests where **p = 0.0043. C Mean ± SEM IC₅₀ values of pfcrt- and pfapc10-modified lines compared to their isogenic parents. Dd2^Dd2crt+N246H (red) and Dd2^{Dd2crt+apc10 D233N} (green) clones were compared to Dd2^Dd2crt (black) while Dd2^GB4crt+N246H (deep blue) and Dd2^FCBcrt+N246H (deep purple) were compared to Dd2^GB4crt (light blue) and Dd2^FCBcrt (light purple), respectively. These IC₅₀ values were calculated from 4 to 6 independent experiments with technical duplicates (N, n = 4 − 6, 2) and statistical significance was determined using two-tailed Mann-Whitney U tests where *p = 0.0159 and **p = 0.0079. D Mean ± SEM IC₅₀ values for ZY19489, mefloquine (MFQ) and lumefantrine (LMF) against a FCB line (black) that expresses two copies of pfmdr1 and the isogenic FCB-KD^mdr1 (blue) line that expresses a single copy of the gene due to targeted disruption of the second (Supplementary Table 5). 72-h susceptibility assays were performed in technical duplicate on 4 to 5 independent experiments (N, n = 4 − 5, 2) over a range of drug concentrations and statistical significance was calculated using two-tailed Mann-Whitney U tests; *p = 0.0286; **p = 0.0079. E ZY19489 susceptibility of NF54 (blue), RF7 (red) and Cam3.II (green) parental lines and the 34 unique NF54×RF7 recombinant progeny with inherited NF54 wildtype (same as 3D7) or RF7 mutant pfcrt alleles. The dotted lines denote the baseline mean ZY19489 IC₅₀ values against NF54 (5.8 nM) and RF7 (27.3 nM). The mean IC₅₀ ± SEM values of NF54 and RF7 were obtained from 14 independent biological repeats conducted in technical duplicates, while those of Cam3.II and the recombinant progeny were obtained from 2 and 4 biological repeats, respectively. The error bars show the range of variability referring to the uppermost and lowermost IC₅₀ values obtained for each sample. F LOD (Logarithm of the Odds score) plots showing associations between ZY19489 IC₅₀ values and segments across the P. falciparum genome (apicoplast, API). The red line signifies the 95% probability threshold. A locus on chromosome 7, which includes pfcrt, shows the highest association with ZY19489 in vitro activity.

The putative MoA and molecular mechanisms that potentially underlie resistance to ZY19489 remain unknown. In prior in vitro selection experiments using several triaminopyrimidines structurally analogous to ZY19489, one derivative (compound 6; Fig. 1A) yielded resistant parasites that exhibited a 3-fold increase in the half-maximal growth inhibition concentration (IC₅₀) compared to parental Dd2⁷. This low-level resistance was initially associated with a G29V mutation in subunit D of the P. falciparum vacuolar proton-transporting V-type ATPase complex (PfV1-D; PF3D7_1341900), suggesting possible involvement of this protein complex in conferring ZY19489 resistance. However, in a recent study using a “pH fingerprint” assay to elucidate the MoA of candidate antimalarials, ZY19489 did not affect the ability of the V-type H⁺ ATPase to regulate digestive vacuole (DV) or cytoplasmic pH. In addition, ZY19489 was unable to significantly inhibit ATP hydrolysis via the V-type H⁺ ATPase pump¹¹. These results suggested that the compound has a different MoA in the parasite.

In this report, we explored ZY19489 by leveraging in vitro resistance selection and gene editing, genetic cross trait mapping, metabolomics, drug transport experiments, parasite growth fitness assays, and molecular dynamics simulations. Our study identified a novel point mutation in the P. falciparum chloroquine resistance transporter (PfCRT; PF3D7_0709000) that confers low-grade resistance to this inhibitor. We provide evidence of a high resistance barrier for ZY19489 based on combination treatments in mice and highlight its efficacy against resistance-conferring mutant pfcrt alleles currently circulating in the field. Metabolomic results offer a window into the mechanistic basis of ZY19489 anti-ABS potency. These findings yield a locus for surveilling possible evolution of clinical resistance to this antimalarial drug and a rationale for potential pairing of this compound with the 4-aminoquinoline piperaquine (PPQ; Fig. 1A) to generate a resistance-refractory combination.

Results

In vitro selection and genome sequencing identify mutant parasites with low-grade resistance to ZY19489

To test whether the prior association between ZY19489 resistance and a mutated form of the P. falciparum V-type ATPase complex⁷ was causal (see Introduction), we generated a conditional knockdown (cKD) line, NF54^{V1-D cKD} (Supplementary Table 1; and Supplementary Fig. 1A–C), with PfV1-D expression regulated by anhydrotetracycline (aTc). Assays with the control V-type ATPase inhibitor bafilomycin A showed the expected reduction in IC₅₀ in parasites with reduced expression of PfV1-D. In contrast, we observed no difference between ZY19489 activity against PfV1-D cKD parasites cultured in high or low concentrations of aTc and control lines under no aTc, further discounting a causal role for the V-type ATPase complex (Supplementary Table 2).

To identify a genetic basis of in vitro resistance to ZY19489, we used a high parasite inoculum strategy and the hypermutable Dd2-Polδ line that has a higher basal mutation rate¹² (Supplementary Fig. 2). Parasites were intermittently pressured with 100 nM ZY19489, corresponding to ~10× its IC₅₀. Recrudescent parasites obtained from one of the three flasks after 136 days of selection pressure exhibited a 12-fold IC₅₀ increase compared to Dd2-Polδ. Resistance was confirmed in four parasites cloned by limiting dilution (Fig. 1B; and Supplementary Table 3).

Whole-genome sequence (WGS) analysis of the four ZY19489-pressured clones identified single nucleotide polymorphisms (SNPs) in four different genes (Table 1). Of these, the N246H mutation in exon 5 of pfcrt and D233N in the anaphase-promoting complex subunit 10 (pfapc10; PF3D7_1217600) were present in all the clones. Mutations in two other genes were observed in a single clone (C8; Table 1), which also showed amplification of an ~80 kb chromosome 5 segment spanning 15 genes including the P. falciparum multidrug resistance protein 1 (PfMDR1; PF3D7_0523000; Supplementary Table 4).

Table 1.

Mutations identified from whole-genome sequencing of ZY19489-selected Dd2-Polδ clones

Gene ID	PlasmoDB annotation	Amino acid change	Clones with mutation
PF3D7_0709000	P. falciparum chloroquine resistance transporter	N246H	C8, E3, F7, G2
PF3D7_1217600	anaphase-promoting complex subunit 10, putative	D233N	C8, E3, F7, G2
PF3D7_1321300	conserved Plasmodium membrane protein, unknown function	T1776I	C8
PF3D7_1320800	dihydrolipoyllysine-residue succinyltransferase component of 2-oxoglutarate dehydrogenase complex	K242T	C8

Genetic validation confirms PfCRT N246H as the primary determinant of ZY19489 resistance

To test the role of pfcrt N246H and pfapc10 D233N, we genetically engineered these mutations into ZY19489-sensitive lines (Supplementary Table 1). N246H was edited into pfcrt in recombinant Dd2 lines expressing either the dominant Southeast Asian (Dd2^Dd2crt) or two African (Dd2^GB4crt and Dd2^FCBcrt) alleles, each of which confer chloroquine (CQ) resistance. The pfcrt Dd2 allele has eight point mutations that differ from the CQ-sensitive wildtype (3D7) allele, whereas GB4 and FCB have six and seven mutations, respectively (lacking N326S in GB4 and I356T in both GB4 and FCB)¹³. The D233N mutation was edited into pfapc10 in the Dd2^Dd2crt line.

Gene editing used zinc-finger nucleases for pfcrt¹⁴ and a CRISPR/Cas9 system for pfapc10¹² (Supplementary Fig. 3A, B). Successful editing yielded Dd2^Dd2crt+N246H, Dd2^GB4crt+N246H, Dd2^FCBcrt+N246H, and Dd2^{Dd2crt+apc10 D233N}, as confirmed by Sanger sequencing. We observed an 11-, 8- and 5-fold increase in ZY19489 IC₅₀ values in the gene-edited Dd2^Dd2crt+N246H, Dd2^GB4crt+N246H and Dd2^FCBcrt+N246H lines, respectively, compared to their isogenic parental lines (Fig. 1C; and Supplementary Table 5). These shifts reproducibly phenocopied the results from the drug-selected Dd2-Polδ clones and indicated a marginally lower fold change in Dd2^GB4crt+N246H and Dd2^FCBcrt+N246H. On the other hand, two PfAPC10 mutant clones (with mean ± SEM ZY19489 IC₅₀ values of 7.7 ± 0.7 nM and 8.8 ± 1.0 nM) showed comparable susceptibilities to Dd2^Dd2crt (mean ± SEM IC₅₀; 8.5 ± 0.4 nM) (Fig. 1C; Supplementary Table 5). Commercially sourced APC10 and ubiquitin ligase inhibitors tested against Dd2^{Dd2crt+apc10 D233N} and its isogenic parent showed no difference in activity (Supplementary Table 6), further arguing against a modulatory role for the pfapc10 D233N mutation. The K242T and T1776I mutations observed in PF3D7_1320800 and PF3D7_1321300, respectively, were not studied further since they were present in only one clone.

Because the WGS analysis had also flagged copy number amplification on chromosome 5 that includes pfmdr1, we also examined the influence of differential copies of this pleotropic efflux transporter on ZY19489 activity by testing the susceptibility of two isogenic FCB lines that differ in their pfmdr1 copy number and expression levels¹⁵. These lines showed no significant difference in susceptibility to ZY19489. The aryl amino alcohols, mefloquine and lumefantrine (Supplementary Fig. 4), both showed a ~ 2-fold IC₅₀ decrease against the single-copy FCB-KD^mdr1 relative to the parental line that expresses two copies, as expected¹⁵ (Fig. 1D; and Supplementary Table 7). These findings establish the PfCRT N246H mutation as the primary driver of in vitro resistance to ZY19489.

Quantitative trait loci analysis map in vitro ZY19489 resistance to a chromosome 7 segment containing pfcrt

We tested the activity of ZY19489 against two contemporary Cambodian multidrug-resistant clinical isolates and observed 2.3- and 4.7-fold higher mean IC₅₀ ± SEM values against Cam3.II (14.9 ± 3.3 nM) and RF7 (27.3 ± 0.7 nM), respectively, compared to the drug-sensitive African reference line NF54 (5.8 ± 0.2 nM). This suggests the presence of naturally occurring field mutations in Cam3.II and RF7 that might modulate parasite response to this compound.

We therefore analyzed a recent genetic cross¹⁶ conducted between the PPQ- and dihydroartemisinin (DHA)-resistant RF7 parental line and the drug-sensitive NF54 parental line and performed quantitative trait locus (QTL) mapping to localize candidate determinants of ZY19489 susceptibility. Phenotype profiling of the 34 recombinant progeny clones recovered from that cross revealed ZY19489 mean IC₅₀ values ranging from 8 to 30 nM (Fig. 1E), which allowed for a phenotype-genotype linkage analysis to uncover QTLs associated with partial in vitro resistance to ZY19489. QTL analysis revealed a significant peak on chromosome 7 that spanned 300 kb and contained 34 genes with non-synonymous mutations between NF54 and RF7 (Supplementary Table 8). The maximum LOD (Logarithm of the Odds) score of 5.3 within this segment mapped to pfcrt, providing evidence of an association between ZY19489 in vitro activity and the pfcrt Dd2 + M343L mutant allele present in RF7 (Fig. 1F). These observations, along with the WGS data from the in vitro selected clones and genetic validation experiments, confirm that PfCRT mutations can modulate ZY19489 in vitro activity.

Structural evidence suggests that the N246H mutation alters ZY19489 interactions with the PfCRT central cavity

PfCRT localizes to the ABS parasite’s DV membrane and comprises 10 transmembrane (TM) helices, including four antiparallel pairs that surround a large negatively charged central cavity¹⁷. Distinct amino acid substitutions, including those that confer CQ or PPQ resistance, line this cavity and mediate the conversion of drug binding events into transport across the DV membrane¹⁷. To explore the effect of the N246H mutation on the PfCRT structure and elucidate the protein-ligand interactions that govern ZY19489 resistance, we performed molecular dynamics simulations with the cryo-EM elucidated 7G8 structure and a separate version into which we introduced the N246H mutation (Supplementary Table 9). This residue maps to TM7 and lines the central cavity close to the parasite cytosol (Fig. 2A). In the open-to-DV conformation, ZY19489 appeared to preferentially reside near the CQ resistance-conferring 76T residue in both the parental (7G8) and N246H-mutated proteins, with no direct interaction with residue 246 (Fig. 2A).With the 7G8 isoform, ZY19489 adopted orientations that facilitated hydrogen bonding with residues such as S94 and the 326D variant, unlike in the 7G8 + N246H mutant where no specific ligand-protein hydrogen bonds persisted for more than 15% of the simulation time. This suggests a weaker ZY19489 interaction with the mutated protein than with the 7G8 isoform. Additionally, ZY19489 appeared to disrupt the residue interaction network around 76T (Y68, S72, N75, 76T, H97, 326D, D329, Q352) more substantially in 7G8 than in the 7G8 + N246H isoform. Interestingly, the 7G8 + N246H protein was predicted to form a salt bridge between K80 and E207, or other nearby residues, particularly in the regions between TM1-TM6, TM5-TM9, and TM6-TM9 (Supplementary Fig. 5). This potential salt bridge formation suggests that the mutant protein might transiently adopt an occluded (partially closed-to-DV) conformation, which might contribute to ZY19489 resistance. Conversely, the 7G8 protein did not appear to adopt this occluded conformation (Supplementary Fig. 5; Supplementary Fig. 6).

Fig. 2. PfCRT interacts with ZY19489, with this drug interaction interfering with PfCRT-mediated CQ and PPQ efflux.

A PfCRT in complex with ZY19489 (in gray and blue stick form) in the 7G8 parental (N246) and mutant (246H) protein. N246H localizes to transmembrane 7 (green) in the central cavity, close to the parasite cytosol. In the open-to-DV conformation, the ligand preferentially resides near the 76T region but does not interact with either N246 or 246H. B Hydrogen bonding patterns between ZY19489 and 7G8 PfCRT in the open-to-cytosol conformation implicate the side chain of E232 and the backbone carbonyl group of L245 as important for stabilizing drug-transporter interactions. The bonding pattern with 7G8 + N246H PfCRT, illustrating a single significant hydrogen bond between the carbonyl oxygen of the Q161 side chain and the ligand that persisted for more than 30% of the simulation period in one of the three molecular dynamics simulations replicates. Hydrogen bonds are represented as dashed blue lines. C Parasite susceptibility profiles for chloroquine (CQ), mono-desethylamodiaquine (mono-DEAQ), ferroquine (FQ), piperaquine (PPQ), atovaquone (ATQ), lumefantrine (LMF), and dihydroartemisinin (DHA), tested against Dd2^Dd2crt (black), Dd2^Dd2crt+N246H (orange), Dd2^GB4crt (blue) and Dd2^GB4crt+N246H (green). Data show mean IC₅₀ ± SEM values calculated from 4 to 5 independent experiments conducted in technical duplicates (N, n = 4 − 5, 2). Statistical significance was calculated using two-tailed Mann-Whitney U tests: *p = 0.0286 and **p = 0.0079. Uptake of (D) 100 nM [³H]CQ or (E) [³H]PPQ in proteoliposomes containing (D) 7G8^7G8crt or (E) 7G8^7G8+F145Icrt variants normalized to control liposomes. Data represent mean ± SEM of 6 to 13 independent experiments conducted in technical duplicates (N, n = 6 - 13, 2). In D, E statistical significance was calculated using two-tailed Mann-Whitney U tests: ****p < 0.0001.

In the open-to-cytosol model, our simulations predicted significant differences in the hydrogen bonding patterns, favoring stronger interactions and stabilization of ZY19489 with 7G8 than with the mutated protein. Specifically, hydrogen bonds with the side chain of E232 and the backbone carbonyl group of L245 appeared to be important for stabilizing the ligand in the 7G8 isoform (Fig. 2B; and Supplementary Fig. 6). In the 7G8 + N246H variant, we observed only one hydrogen bond interaction with Q161 that persisted for more than 30% of the simulation period. This hydrogen bond was not consistently detected in all trajectories, appearing exclusively in only one of the three independent molecular dynamics replicate simulations. This absence of consistent hydrogen-bonding in the 7G8 + N246H mutant evokes possibly weaker ligand interactions compared to the 7G8 isoform. These findings suggest that the N246H mutation may significantly alter the binding dynamics of ZY19489 in the open-to-cytosol conformation, potentially reducing the interaction stability of the ligand and subsequently facilitating its transport via the mutated protein, which could result in resistance.

PfCRT N246H confers sensitivity to DV-acting aminoquinoline antimalarials

Certain PfCRT mutant isoforms can mediate reduced parasite sensitivity to CQ and PPQ via their transport-mediated efflux out of the DV^17,18. To relate the molecular dynamics simulation results to experimentally derived influence of PfCRT N246H on various antimalarials, we tested the susceptibility of Dd2^Dd2crt+N246H, Dd2^FCBcrt+N246H and Dd2^GB4crt+N246H to clinically important drugs including CQ, mono-desethylamodiaquine (the primary active metabolite of the ACT partner drug amodiaquine), ferroquine, PPQ, atovaquone (ATQ), lumefantrine, and DHA (Supplementary Fig. 4). This antimalarial set comprises 4-aminoquinoline ring-containing drugs that are believed to inhibit heme detoxification in the DV (the first four listed) as well as other drugs whose primary targets lie elsewhere. Dd2^Dd2crt+N246H and Dd2^GB4crt+N246H were 9- and 11-times significantly more susceptible to CQ, respectively, than their parental lines, consistent with ablation of CQ resistance (Fig. 2C; and Supplementary Table 10). These lines also exhibited a 4- to 6-fold increased sensitivity to mono-desethylamodiaquine as well as a 2- to 3-fold sensitization to ferroquine and PPQ. Although the same trend was also observed for lumefantrine, the drop in IC₅₀ was modest (20%) and not statistically significant (Supplementary Table 10). There was no significant change in susceptibility towards DHA or ATQ (Fig. 2C; and Supplementary Table 10), alluding to a collateral sensitization of the N246H lines that was specific to DV-acting 4-aminoquinolines.

ZY19489 competes with CQ and PPQ in PfCRT-mediated drug efflux

To interrogate whether ZY19489 might influence PfCRT-mediated drug transport across the DV membrane, we leveraged an experimental physiological model that mimics an “inside-out” parasite DV whereby [³H]CQ or [³H]PPQ uptake can be measured in PfCRT-containing proteoliposomes¹⁷. We observed a significant 72% reduction in [³H]CQ transport by the CQ-resistant 7G8 isoform upon addition of 1 μM ZY19489 (Mann-Whitney U tests; p < 0.0001), suggesting that this compound can directly impede CQ transport (Fig. 2D). Similar interference was also observed against [³H]PPQ transport by the PPQ-resistant isoform 7G8 + F145I, with a significant 79% inhibition of PPQ transport achieved with 1 μM ZY19489 (Mann-Whitney U tests; p < 0.0001, Fig. 2E). Transport of [³H]CQ and [³H]PPQ by the PfCRT isoforms 7G8 and 7G8 + F145I, respectively, was significantly inhibited by the addition of the reciprocal agents PPQ and CQ, to levels comparable to those observed with ZY19489 (Mann-Whitney U tests; p < 0.0001, Figs. 2D, E). No competitive inhibition of transport was observed with pyrimethamine or ATQ whose resistance mechanisms (mutations in dihydrofolate reductase and cytochrome bc₁, respectively) are unrelated to PfCRT-mediated transport. These results show that ZY19489 can antagonize CQ and PPQ transport and offer evidence of competition for the protein’s binding cavity. They also provide compelling evidence favoring ZY19489 as an agent that can restore CQ or PPQ potency by inhibiting their mutant PfCRT-mediated efflux away from their heme target in the DV.

PfCRT N246H confers a slow growth rate and high fitness cost on ABS parasites

During routine culturing of the ZY19489-pressured, resistant Dd2-Polδ mutants, we noticed a unique parasite morphology characterized by translucent distended DVs during the development from mid-trophozoites to mid-schizonts (Fig. 3A). This aberrant phenotype was retained in all N246H pfcrt-edited lines but absent in the Dd2-Polδ, Dd2^Dd2crt and D233N pfapc10-edited clones, suggesting an interplay between this variant PfCRT isoform and parasite DV physiology.

Fig. 3. ZY19489 retains potency against parasites with contemporary pfcrt and pfmdr1 mutations and parasites resistant to this compound display aberrant morphology and a substantial fitness cost.

A Light microscopy images reveal distended, translucent DVs in Dd2 parasites expressing the N246H mutation in three distinct CQ-resistant mutant pfcrt allelic backgrounds. Distended DVs were not observed in Dd2-Polδ, Dd2^Dd2crt or Dd2^GB4crt parental lines lacking the PfCRT N246H mutation. These morphologies were observed on 3 independent occasions but only representative images from one experiment are shown. The scale bar measures 8 μm. B In vitro growth characteristics of PfCRT N246H mutant lines and their isogenic parents. Mean ± SEM growth rates per generation were defined as the average rate of parasite growth per generation over eight generations, conducted on 3 separate occasions in technical duplicates. Statistical significance was calculated using two-tailed Student’s t tests with Welch’s correction; **p = 0.0064; ****p < 0.0001. C In vitro 72-h activity of ZY19489 against naturally occurring or drug-selected contemporary pfcrt and pfmdr1 alleles. The susceptibility of each mutant line was compared to their respective isogenic parental edited or unedited lines. Data are presented as mean ± SEM IC₅₀ values calculated from 4 to 5 independent experiments conducted in technical duplicates (N, n = 4 − 5, 2). Statistical significance was calculated using two-tailed Mann-Whitney U tests; **p = 0.0079.

To assess the impact of the PfCRT N246H mutation on growth rate, we monitored the replication rates of synchronized ring stage parasites over several replication cycles and used it as a proxy for ABS fitness. Cultures of Dd2^Dd2crt+N246H, Dd2^GB4crt+N246H, Dd2^FCBcrt+N246H and their respective parental lines were each seeded at an inoculum equivalent to 8 × 10⁵ parasites per well and tracked over 8 generations (16 days). Parasitemia measurements and media changes were performed on days 3 and 4 and every two days thereafter. Parasite cultures with >2% parasitemia were diluted to 0.2% to reinitiate growth. Dd2^Dd2crt, Dd2^FCBcrt and Dd2^GB4crt exhibited a 6- to 7-fold mean growth rate per generation compared to 2- to 3-fold rates observed in the mutants over the same period. From these data, we quantified the fitness cost of each mutant as a percent reduction in growth rate per generation relative to the isogenic parent (Fig. 3B). Dd2^Dd2crt+N246H, Dd2^GB4crt+N246H and Dd2^FCBcrt+N246H exhibited a mean fitness cost of 75%, 73% and 61%, respectively, consistent with the slow expansion rates in routine cultures. These results suggest that PfCRT N246H confers a severe fitness cost to ABS parasites. We hypothesize that parasites with this mutation would be likely to have a major competitive disadvantage were they to emerge in the field.

ZY19489 retains potency against contemporary mutant pfcrt and pfmdr1 alleles

We next investigated whether parasites with contemporary pfcrt and pfmdr1 alleles known to confer resistance to PPQ or other antiplasmodial agents could also be cross-resistant to ZY19489. Assays included CQ-resistant lines harboring different PfCRT mutations at residues 72-76 as well as PPQ-resistant lines expressing causal PfCRT mutations including C350R, H97Y, F145I, M343L, or G353V¹⁸. We extended this screening to naturally occurring or drug-selected pfmdr1 mutants on NF10, NF54, Cam3.II and Dd2 backgrounds (Supplementary Table 1). A750T and S784L are naturally occurring PfMDR1 mutations previously identified in field isolates from Western Cambodia and the Thai-Myanmar border^19,20 while the M841I + M924I double mutant was obtained from in vitro drug selection using the piperazine-containing ACT-451840²¹. The F1072L and S1075I PfMDR1 mutants were obtained from selection with hexahydroquinoline derivatives²².

ZY19489 was consistently potent against all CQ- or PPQ-resistant pfcrt mutant alleles (of diverse geographic origins) expressed in the SE Asian Dd2 parasite background, with mean IC₅₀ values ranging from 7.1 nM (Dd2^3D7crt) to 13.7 nM (Dd2^Dd2crt+G353V; Fig. 3C; and Supplementary Table 10). Against South American 7G8 parasites expressing either the CQ-resistant 7G8 or PPQ-resistant 7G8 + C350R pfcrt alleles (the 7G8^7G8crt and PPQ-resistant 7G8^7G8crt+C350R lines, respectively), the ZY19489 mean IC₅₀ was 13.6 nM and 14 nM, respectively, highlighting the equipotency of this compound against mutant pfcrt alleles presently circulating in the field. The sensitivity of mutant PfMDR1 parasites, on the other hand, was more varied. There was no difference in potency against NF54 or NF54^{M841I+M924Imdr1} parasites (mean IC₅₀ values of 8.3 nM and 9.9 nM, respectively). Similarly, ZY19489 was equipotent against NF10^A750Tmdr1 and Cam3.II C580Y^S784Lmdr1 parasites compared with their NF10 and Cam3.II C580Y parental lines. In contrast, we observed a 3-fold decrease in ZY19489 activity against Dd2^F1072Lmdr1 (mean IC₅₀ of 31.3 nM) and Dd2^S1075Imdr1 (mean IC₅₀ of 29.6 nM) compared to the parental Dd2 line (mean IC₅₀ of 10.2 nM) (Fig. 3C; and Supplementary Table 11).

ZY19489 exhibits a high in vivo barrier to resistance in mice treated in combination with other candidate antimalarials

To assess whether ZY19489-resistant parasites emerge in vivo following combination treatment, we employed a validated humanized mouse model of P. falciparum ABS infection using NOD-scid IL-2Rγ^null (NSG) mice engrafted with human erythrocytes (TADhuMouse®)²³. In three independent experiments, humanized NSG mice infected with 3D7^0087/N9 (a standard P. falciparum strain adapted to grow reproducibly in humanized NSG mice²⁴) were treated for 1 to 4 days with varying doses of ZY19489, the quinoline-carboxamide M5717²⁵, the pantothenamide MMV693183²⁶, or ZY19489-based combinations including MMV693183, ferroquine, or M5717 (Table 2; and Supplementary Fig. 7).

Table 2.

Phenotypic and genotypic profiles of recrudescent P. falciparum 3D7^0087/N9 parasites from humanized NSG mice treated with ZY19489, MMV693183, M5717, or ZY19489-based combination therapies

Parasite Line IDa	NSG mouse treatment group and dosing regimena	DoRb	IC50 ± SEM (nM)c		Genetic changesd
	ZY19489 + MMV693183		ZY19489	MMV693183
TAD_022	None (control)	NA	11.1 ± 0.6	2.6 ± 0.3	Parental reference
TAD_746	1 × 40 mg/kg ZY19489 + 1 × 1.5 mg/kg MMV693183	15.6	11.1 ± 0.5	2.3 ± 0.2	None
TAD_752	1 × 20 mg/kg ZY19489 + 1 × 5.0 mg/kg MMV693183	17.0	11.2 ± 0.7	2.6 ± 0.2	None
TAD_755	1 × 20 mg/kg ZY19489 + 1 × 5.0 mg/kg MMV693183	16.6	12.3 ± 0.8	2.9 ± 0.1	None
TAD_758	1 × 40 mg/kg ZY19489 + 1 × 1.5 mg/kg MMV693183	17.5	11.2 ± 0.2	2.8 ± 0.2	None
TAD_761	1 × 40 mg/kg ZY19489 + 1 × 5.0 mg/kg MMV693183	18.0	11.8 ± 0.8	2.8 ± 0.3	None
TAD_764	2 × 40 mg/kg ZY19489 + 1 × 1.5 mg/kg MMV693183	22.3	11.7 ± 0.3	2.6 ± 0.3	None
TAD_767	1 × 40 mg/kg ZY19489 + 1 × 5.0 mg/kg MMV693183	20.9	11.9 ± 0.9	2.6 ± 0.2	None
TAD_770	3 × 2.5 mg/kg MMV693183	26.2	11.2 ± 0.5	2.6 ± 0.2	None
TAD_773	1 × 20 mg/kg ZY19489 + 3 × 2.5 mg/kg MMV693183	26.5	11.6 ± 0.8	2.9 ± 0.4	None
TAD_779	2 × 40 mg/kg ZY19489 + 3 × 2.5 mg/kg MMV693183	30.7	11.0 ± 0.3	2.5 ± 0.3	Y1020H in PF3D7_0825000, N636Y in PF3D7_1416100
TAD_782	2 × 40 mg/kg ZY19489 + 3 × 2.5 mg/kg MMV693183	30.7	11.7 ± 1.2	2.9 ± 0.2	None
	ZY19489 + Ferroquine		ZY19489	Ferroquine
TAD_022	None (control)	NA	6.7 ± 0.6	10.3 ± 1.6	Parental reference
TAD_494	4 × 40 mg/kg ZY19489	25.5	7.2 ± 0.5	8.3 ± 0.2	None
TAD_496	4 × 40 mg/kg ZY19489 + 1 × 15 mg/kg Ferroquine	28.3	6.9 ± 1.2	9.5 ± 1.1	D1698N in PF3D7_0210200, PF3D7_1224000 (loss of 3.5 copies)
TAD_497	4 × 40 mg/kg ZY19489 + 1 × 15 mg/kg Ferroquine	27.5	7.1 ± 0.9	8.9 ± 0.6	N1580S in PF3D7_1329100, G48A in PF3D7_1408600
	ZY19489 + M5717
TAD_030	None (control)	NA	-	-	Parental reference
TAD_377	1 × 3 mg/kg M5717	14.1	-	-	No mutations in PF3D7_1451100
TAD_382	1 × 20 mg/kg M5717	20.2	-	-	G373S in PF3D7_0320500, P490L in PF3D7_1032000, R149T in PF3D7_1312600, Y186N in PF3D7_1451100*
TAD_384	1 × 40 mg/kg M5717	21.6	-	-	Y186N in PF3D7_1451100*
TAD_378	1 × 20 mg/kg ZY19489 + 1 × 3 mg/kg M5717	14.2	-	-	G373S in PF3D7_0320500, P490L in PF3D7_1032000
TAD_379	1 × 20 mg/kg ZY19489 + 1 × 3 mg/kg M5717	14.5	-	-	P490L in PF3D7_1032000
TAD_386	1 × 20 mg/kg ZY19489 + 1 × 20 mg/kg M5717	23.7	-	-	P490L in PF3D7_1032000
TAD_400	1 × 20 mg/kg ZY19489 + 1 × 20 mg/kg M5717	28.0	-	-	P490L in PF3D7_1032000, R149T in PF3D7_1312600, Y186N in PF3D7_1451100*

^aRecrudescent parasite lines were recovered from humanized NSG mice infected with the P. falciparum 3D7^0087/N9 strain and treated with ZY19489, MMV693183, M5717, or ZY19489-based combinations (ZY19489 + MMV693183, ZY19489+Ferroquine, or ZY19489 + M5717). Parental control lines (TAD_022 and TAD_030) were sampled from untreated mice.

^bDoR, Day of Recrudescence: Day of experiment when parasites reappear in circulating blood after initial complete clearance. Day 1 corresponds to the first day of treatment administration. NA, not applicable.

^cIC₅₀ values (nM) were determined for the parental controls and drug-exposed recrudescent parasites using standard 72-h in vitro dose-response assays following culture adaptation. Values represent means ± SEM from 3 to 5 independent experiments, each performed in duplicate. Recrudescent parasites from the ZY19489 + M5717 treatment group were not available for phenotypic analysis. Statistical significance was assessed relative to parental controls (TAD-022 or TAD_030) using two-tailed unpaired t-tests. No significant differences in susceptibility to ZY19489, MMV693183, or ferroquine were observed in any tested recrudescent line.

^dSingle nucleotide polymorphisms (SNPs) and copy number variations (CNVs) were identified by whole-genome sequencing of all recrudescent lines relative to the respective 3D7^0087/N9 parental genomes (TAD-022 or TAD-030), except for TAD_377 and TAD_384 that were not whole-genome sequenced. All samples from the ZY19489 + M5717 group, including TAD_377 and TAD_384, were analyzed by targeted amplicon sequencing of the PfeEF2 (PF3D7_1451100) locus. The following genes were affected: PF3D7_0825000 (conserved Plasmodium protein of unknown function); PF3D7_1416100 (SEY1, a putative Plasmodium endoplasmic reticulum-shaping protein); PF3D7_0210200 (conserved Plasmodium protein of unknown function); PF3D7_1329100 (Myosin F, putative); PF3D7_1408600 (40S ribosomal protein S8e, putative); PF3D7_0320500 (nicotinamidase, putative); PF3D7_1032000 (ribosome maturation factor RimM, putative); PF3D7_1312600 (2-oxoisovalerate dehydrogenase subunit alpha, mitochondrial, putative); PF3D7_1451100 (P. falciparum elongation factor 2, PfeEF2) – the known target of M5717. Additionally, a 3.5-fold reduction in copy number was observed in a ~ 1.8 kb segment of chromosome 12 harboring the gch1 gene (PF3D7_1224000) in the TAD-496 genome. None of the observed genomic differences were associated with phenotypic resistance to ZY19489, MMV693183, or ferroquine.

*The PfeEF2 Y186N mutation identified in M5717-exposed recrudescent lines TAD_382, TAD_384, and TAD_400 is known to confer high-grade resistance to M5717, with an ~16,000-fold increase in IC₅₀.

In the ZY19489 + MMV693183 treatment group, 11 recrudescent parasite lines (10 from various dosing regimens of the combination and 1 from a 3-day MMV693183 monotherapy at 2.5 mg/kg) were culture-adapted for phenotypic screening and WGS. No significant differences in parasite susceptibility to ZY19489 or MMV693183 were observed between the parental 3D7^0087/N9 strain and any of the recrudescent lines (Table 2). WGS identified genetic changes in only one line, TAD_779, which carried a Y1020H mutation in PF3D7_0825000, encoding a conserved Plasmodium protein of unknown function, and a N636Y mutation in PF3D7_1416100, encoding SEY1, a putative endoplasmic reticulum-shaping protein. Neither mutation was associated with altered compound susceptibility. Notably, no mutations were observed in PF3D7_0627800 encoding the CoA-binding enzyme acetyl-CoA synthetase, a known resistance marker for MMV693183^26,27. These findings suggest that recrudescence was likely due to sub-optimal drug exposure rather than acquired resistance to ZY19489 or MMV693183.

Similarly, no significant shifts in IC₅₀ or IC₉₀ values for ZY19489 or ferroquine were observed in lines recrudescing after exposure to 4 × 40 mg/kg ZY19489 alone or combined with one dose of 15 mg/kg ferroquine (Table 2). WGS revealed no consistent mutations across ZY19489- or ZY19489 + ferroquine-treated lines, indicating no evidence of acquired resistance to either compound.

In the group treated with either M5717 alone or in combination with ZY19489, seven recrudescent samples were obtained (three from M5717 monotherapy and four from combination regimens). These samples, along with the untreated parental strain, were not available for in vitro phenotypic testing but were genotyped at the PfeEF2 (PF3D7_1451100) locus that encodes P. falciparum translation eukaryotic elongation factor 2, a known target of M5717. Three lines, TAD_382, TAD_384, and TAD_400, harbored the PfeEF2 Y186N mutation, which confers high-grade M5717 resistance (Table 2)^25,28. Two of these (TAD_382 and TAD_384) were from M5717 monotherapies; the third (TAD_400) emerged following 20 mg/kg ZY19489 + 20 mg/kg M5717 combination treatment. WGS was performed on five lines  (TAD_382 and all four from the combination treatments) to identify additional mutations. The PfeEF2 Y186N mutation was confirmed in TAD_382 and TAD_400, consistent with genotyping results from amplicon sequencing. Three additional missense mutations were identified in TAD_382, including a P490L substitution in PF3D7_1032000 (ribosome maturation factor RimM), which was present in all five M5717-exposed lines (Table 2). However, no novel SNPs or CNVs were observed in samples exclusively exposed to the ZY19489 + M5717 combination, confirming that PfeEF2 Y186N likely arose from M5717 pressure.

To assess potential cross-resistance between M5717 and ZY19489, we tested three M5717-resistant lines (TAD_470, TAD_462, and TAD_464), which harbored distinct PfeEF2 mutations (I183M, P754S, and I182T, respectively), for susceptibility to ZY19489. No significant differences in IC₅₀ or IC₉₀ values were observed compared with the drug-sensitive parental line (TAD_022), indicating that PfeEF2-mediated M5717 resistance does not affect parasite sensitivity to ZY19489 (Supplementary Table 12). M5717 susceptibility testing confirmed the previously reported resistance phenotypes of these lines, with IC₅₀ shifts of 4.5-fold (TAD_470), 151-fold (TAD_462), and ~260-fold (TAD_464) relative to TAD_022 (Supplementary Table 12)²⁸. Together, these data underscore the high in vivo barrier to resistance against ZY19489.

ZY19489 in vitro ABS potency is maximal against rings and schizonts

We assessed the timing of ZY19489’s anti-ABS action by measuring its potency against synchronized rings, trophozoites, and schizonts that had been treated separately for 12 h²⁹. For unbiased interpretation of optimal drug activity, we used the drug-sensitive 3D7 strain that expresses wildtype isoforms of most of the proteins associated with antimalarial drug resistance. We included DHA, CQ and the 2-aminopyridine MMV390048 (a phosphatidylinositol 4-kinase inhibitor) as control compounds with established peak activity against early rings, trophozoites and schizonts, respectively²⁹. These compounds yielded mean IC₅₀ values of 1.7 nM, 12.3 nM and 24.4 nM, respectively (Supplementary Table 13). ZY19489 exerted optimal activity against rings and schizonts (mean IC₅₀ values of 9 nM and 12.8 nM, respectively, compared to 28.3 nM against trophozoites; Fig. 4A; Supplementary Table 13), a profile distinct from our control compounds.

Fig. 4. ZY19489 is active against rings and schizont and perturbs Hb catabolism.

A Dot plots showing the stage-specificity profile of ZY19489 against synchronized 3D7 rings (blue), trophozoites (red) and schizonts (green) exposed to drug for 12 h before drug wash-off. IC₅₀ values were assessed at 60 h and data are presented as means ± SEM from six independent experiments. B Metabolomic profiles associated with synchronized 3D7 and Dd2 trophozoites following treatment with 10× ZY19489 for 2.5 h. Data are displayed as supra-hexagonal meta-prints determined from the log₂ fold change values of targeted metabolites following drug treatment relative to no-drug control. Metabolite clusters were associated with eight generalized metabolic pathways using the KEGG database and color coded within the supra-hexagon base map. The red grids highlight the hexagon panels associated with Hb-derived peptides. C Plot of inhibition of β-hematin formation versus concentration profiles for ZY19489 relative to known antimalarial standards. Data were fitted to the sigmoidal concentration response (variable slope) equation in GraphPad Prism to determine the IC₅₀, represented as the mean ± SEM of 4 independent experiments performed in technical duplicates (N, n = 4, 2). The error bars show the range of variability referring to the uppermost and lowermost IC₅₀ values for each compound. (D, F, G and I) Heme fractionation of synchronized NF54 parasites treated for 30 h with different drug concentrations of ZY19489 or CQ. Heme species are presented as ‘free’ heme (left y-axis) and hemozoin (right y-axis) Fe in femtogram isolated per cell (fg/cell) following exposure to different concentrations of (D) CQ or (G) ZY19489. Amounts of hemoglobin Fe (fg/cell) are shown in panels (F, I). Data are presented as means ± SEM from 4 to 8 repeats conducted in technical duplicates (N, n = 4 − 8, 2) and statistical comparisons of the drug-treated parasites to their untreated controls were performed using two-tailed Mann Whitney U tests: *p = 0.0286, **p = 0.0011 and ***p = 0.0002. E, H Representative overlay plots of parasite survival (left y-axis) with ‘free’ heme amounts (right y-axis) as a function of (E) CQ or (H) ZY19489 concentrations. Data were fitted to the sigmoidal concentration response (variable slope) equation in GraphPad Prism where the error bars represent the range of the highest and lowest values of each data point.

ZY19489 inhibition of parasite growth is associated with hemoglobin catabolism inhibition

Our observation of swollen DVs in ZY19489-resistant parasites (Fig. 3A) was reminiscent of a similar morphological feature in PPQ-resistant parasites that was previously associated with defective hemoglobin (Hb) breakdown³⁰. To test whether the ZY19489 MoA might also be linked to Hb catabolism, we incubated synchronized 3D7 and Dd2 trophozoites with 10× IC₅₀ of ZY19489 for 2.5 h and compared the ensuing metabolic responses with no-drug controls³¹. Parasites treated with the reference mitochondrial inhibitor ATQ exhibited significant accumulation of N-carbamoyl-aspartate and dihydroorotate, two pyrimidine biosynthetic intermediates upstream of its target, P. falciparum dihydroorotate dehydrogenase (PfDHODH; PF3D7_0603300)(Supplementary Table 14; Supplementary Fig. 8). These changes were accompanied by reduced levels of intracellular uridine monophosphate (UMP), diphosphate (UDP) and triphosphate (UTP), consistent with ATQ-mediated inhibition of the mitochondrial electron transport chain complex and concomitant loss of PfDHODH activity³². ZY19489-treated 3D7 and Dd2 parasites, on the other hand, showed significant reductions in short Hb-derived peptides, commonly linked with inhibition of Hb endocytosis and/or catabolism (Fig. 4B; and Supplementary Table 14). These short peptides ranged from 2 to 6 amino acids in length, had isoelectric points of four to six, and were negatively charged at pH 5.5 or pH 7.4 (reflecting conditions in the DV lumen or cytosol, respectively) (Supplementary Table 15). We also observed significantly elevated levels of deoxyuridine monophosphate (dUMP), a crucial precursor in the production of thymidine required for DNA replication (Supplementary Table 14; and Supplementary Fig. 8). Separately, we explored whether ZY19489 might target cGMP dependent protein kinase (PfPKG; PF3D7_1436600), given this compound’s structural similarity with the aminopyrimidine PfPKG inhibitor ML-10³³. However, ZY19489 was inactive against recombinant purified PfPKG (Supplementary Fig. 9).

Evidence that ZY19489 inhibits hemoglobin breakdown but not heme detoxification

Hb degradation is initiated by hemoglobinases and leads to the generation of amino acids for growth, with ‘free’ heme released as a toxic byproduct. In its labile form, ‘free’ heme can cause oxidative damage to the parasite and is promptly sequestered into inert hemozoin (Hz) in a process that may be accelerated by P. falciparum heme detoxification protein (PfHDP; PF3D7_1446800) in complex with hemoglobinases including falcipains³⁴. To understand whether ZY19489 inhibits this phase of Hb catabolism, we tested for possible interactions with PfHDP and/or falcipain 2. We generated an aTc-regulatable PfHDP cKD line (designated herein as Dd2^{HDP cKD}) and observed no difference in ZY19489 susceptibility at various aTc concentrations compared to control lines (Supplementary Fig. 10). These data argue against any link between ZY19489 anti-ABS activity and inhibition of PfHDP. On the other hand, quinine displayed a significant ∼2-fold IC₅₀ shift in PfHDP cKD parasites under no aTc relative to parasites cultured at high aTc concentration. We found no significant difference in ZY19489 activity against 3D7 falcipain 2A or 2B knockout lines³⁵ compared to the parental 3D7 line (Supplementary Fig. 11). As a control, we included E64d, which is a membrane-permeable derivative of the epoxysuccinate inhibitor E64, whose broad-spectrum antiprotease activity includes falcipain inhibition³⁵. E64d exhibited high IC₅₀ values against parasites lacking falcipain 2A or 2B (Supplementary Fig. 11).

We then tested for direct interference of Hz formation by first assessing inhibition of the conversion of hematin to β-hematin in a cell-free set-up³⁶. ZY19489 failed to block β-hematin formation (IC₅₀ > 500 μM) in contrast to CQ and amodiaquine (AQ) that were potent inhibitors in this cell-free assay with mean IC₅₀ values of 19 μM and 6 μM, respectively (Fig. 4C). To examine whether the effect of ZY19489 on the pathway was only discernible in live parasites, we then tested for a concentration-dependent effect of the drug on three heme species, namely Hb (undigested Hb species), ‘free’ heme (potentially oxidative labile species liberated from Hb proteolysis), and Hz (inert crystalline species) fractioned from mature trophozoites³⁷. Inhibitors that complex with heme customarily show a concentration-dependent decrease in Hz levels accompanied by a corresponding increase in ‘free’ heme, which correlates with parasite killing³⁷.

In CQ-treated NF54 parasites, the mean ± SEM of ‘free’ heme iron (Fe) present in femtogram per cell (fg/cell) was significantly higher in the highest treatment concentration compared to the untreated control (8.3 ± 0.3 fg/cell vs 5.5 ± 0.2 fg/cell; p < 0.05; Mann-Whitney U tests) (Fig. 4D; and Supplementary Table 16). This increase in ‘free’ heme was concentration-dependent and corresponded with a significant decrease in Hz (Fig. 4D; and Supplementary Table 16). To interpret the physiological relevance of these observations, we superimposed 72-h parasite growth inhibition curves on the ‘free’ heme levels and observed a correlation between heme concentrations and parasite growth inhibition (Fig. 4E). These data suggest that CQ-mediated parasite death was attributable to the buildup of toxic heme. We also observed modest but significant elevation in Hb levels that plateaued between CQ concentrations corresponding to 1.5× and 2.5× IC₅₀ (Fig. 4F; and Supplementary Table 16).

In contrast, the levels of ‘free’ heme in ZY19489-treated parasites showed no concentration-dependent pattern (5.6 ± 0.2 fg/cell in untreated control vs 6.0 ± 0.1 fg/cell in parasites treated with 2.5× IC₅₀) (Fig. 4G). Decreasing amounts of Hz in ZY19489-treated parasites were only observed at higher concentrations and showed no association with ‘free’ heme levels (Fig. 4G; and Supplementary Table 16). Additionally, levels of ‘free’ heme did not correlate with 72-h parasite survival (Figs. 4H, I). Similar lack of inhibition was observed in parasites treated with the antifolate, pyrimethamine (Supplementary Table 16). Nonetheless, treatment with ZY19489 caused significant increases in Hb levels at 1.5× and 2.0× IC₅₀ concentrations but not at the more inhibitory concentration of 2.5× IC₅₀ (Fig. 4I), suggesting some activity of this compound during earlier steps of Hb catabolism in a concentration- and stage-specific manner.

Discussion

Based on its potency, safety and pharmacokinetics, ZY19489 has advanced into Phase Ib clinical evaluation for the treatment of uncomplicated P. falciparum malaria. Its potential incorporation into non-ACTs would be timely given the rapidly expanding list of countries in Africa where mutant k13 alleles associated with artemisinin partial resistance are present³. Here, we reveal key insights into pathways to ZY19489 resistance and its MoA. In vitro drug pressure experiments and WGS revealed a novel PfCRT N246H mutation, which mediated low-grade resistance as confirmed using gene editing. Genetic cross analysis further implicated PfCRT as a primary modulator of susceptibility to ZY19489. Mutant PfCRT N246H parasites had swollen DVs and heightened susceptibility to PPQ and other DV-acting aminoquinolines, including the former gold-standard CQ and the ACT partner drug amodiaquine. Mutant parasites demonstrated a substantial fitness cost, with ABS growth rates reduced by 60-75% compared to isogenic parasite lines lacking this mutation. Encouragingly, ZY19489 maintained potency against parasites carrying mutant pfcrt and pfmdr1 alleles present in the field. This compound also did not select for in vivo resistance in combination studies with multiple antimalarials conducted in P. falciparum-infected humanized mice.

Using metabolomics and Hb degradation assays, we also observed ZY19489-mediated perturbations of Hb catabolism, without a direct effect on downstream inhibition of heme detoxification (the primary MoA of CQ, PPQ and amodiaquine). These data suggest that a major MoA of ZY194899 is related to inhibition of Hb catabolism, which we propose is indicative of a primary action outside of the DV. Our studies also detected the accumulation of pyrimidine precursors. This effect on pyrimidine biosynthesis could relate to the aminopyrimidine substructure in ZY19489 that is also present in the antifolate inhibitor pyrimethamine and P218 (Supplementary Fig. 4). It remains unclear whether this evidence of an effect of ZY19489 on Hb catabolism and levels of pyrimidine precursors suggests direct inhibition of both pathways or is indirectly related. Nonetheless, our results offer insight into the potential MoA of ZY19489 and identify a genetic marker to screen for potential resistance in the field.

Mutations in pfcrt have been shown to modulate the activity of antimalarial drugs in clinical use and under development. The N246H mutation conferred an 11-, 8- and 5-fold increase in mean ZY19489 IC₅₀ values against parasites with Dd2, GB4 and FCB pfcrt alleles, respectively, highlighting the influence of different PfCRT haplotypes in modulating resistance. The low-level resistance observed upon introduction of N246H into GB4 and FCB was an important result as these pfcrt alleles are common in African isolates, unlike Dd2 that is only present in Southeast Asia³⁸. By QTL mapping of genetic cross progeny, PfCRT was also associated with decreased susceptibility to PPQ and the structurally distinct compounds MMV665939 and MMV675939¹⁶, highlighting the influence of this transporter on the activity of diverse antimalarial scaffolds. Of note, the chromosome 7 segment identified in our QTL mapping also included HECT-type E3 ubiquitin ligase UT (PF3D7_0704600), which was previously associated with altered responses to quinine and quinidine and was co-selected with mutant pfcrt³⁹.

Our molecular dynamics analysis mapped PfCRT N246H within the negatively charged central cavity close to the parasite cytosol. CQ and PPQ resistance-conferring mutations localize to specific helices that line this cavity, suggesting this as a primary site of interactions with these drugs^17,40,41. Intriguingly, parasites expressing the N246H mutation were sensitized to DV-acting antimalarials, including PPQ and CQ. In our simulations with the solved 7G8 isoform structure and its 7G8 + N246H variant, ZY19489 consistently sampled the same T76-adjacent region of the cavity, suggesting that this molecule, CQ, and PPQ might partially overlap in their binding to these two PfCRT isoforms. This interpretation aligns with our proteoliposomes data showing that ZY19489 inhibited transport of CQ and PPQ by their respective drug-resistant 7G8 and 7G8 + F145I PfCRT variants. One explanation could be mutual exclusion which, in the context of these PfCRT isoforms, may suggest a single-substrate feature in which only one ligand could occupy the central cavity at a time and preclude binding or transport of another. Our proteoliposomes competition data suggest that all three compounds have at least partially overlapping cavity-binding properties. However, demonstrating true competitive inhibition and, by extension, confirmation of a single binding-site model would require more detailed transport kinetic analyses^40,42,43. It is important to recall that our simulations and transport studies were conducted on 7G8 isoforms and as such this specific interpretation cannot be extended as a generalization to other geographically distinct PfCRT variants. In fact, variations in CQ and PPQ transport phenotypes have been observed between PfCRT isoforms expressing mutant alleles from different geographic regions^41,44, suggesting disparate interactions of these drugs with residues within the central cavity. Additionally, various PfCRT isoforms appear to possess distinct substrate-binding cavity architectures^42,43. Whereas our data suggest that 7G8 PfCRT may accommodate only a single ligand (either CQ or PPQ) at a time, transport kinetics and complementary computational studies have shown Dd2 PfCRT to bind both substrates simultaneously under a partial noncompetitive inhibition model^40,42,43, suggesting the presence of at least two distinct interaction sites within the cavity of this isoform. We also note that isogenic parasites expressing Dd2, GB4 and FCB pfcrt alleles are known to differ in their CQ resistance profiles¹³, which also suggests differences in how these variants interact with antimalarial drugs.

If appropriately matched for pharmacokinetics, we envisage that this collateral sensitivity and reciprocal drug-transport phenomenon could be exploited to develop a resistance-refractory combination of ZY19489 + PPQ. In such a scenario, our data suggest the two drugs could theoretically constitute an effective combination that would force pfcrt into an “evolutionary trap” and prevent the emergence of high-grade resistance to either. This concept remains speculative, as the rarity of selecting for resistance to these compounds alone or in combination precludes experimental validation. Nevertheless, this hypothesis has been validated for other antimalarials, including PfDHODH, Pf20S proteasome and mitochondrial cytochrome bc₁ complex inhibitors¹⁰, and is one of the rationales behind current triple ACT therapies. For instance, amodiaquine paired with artemether + lumefantrine or mefloquine partnered with DHA + PPQ, exert reciprocal selective pressures associated with opposite shifts in the prevalence of mutant versus wildtype isoforms of PfCRT and PfMDR1^45–47. To further consolidate this line of argument, our results dispel concerns of cross resistance between ZY19489 and PPQ or other antimalarials whose clinical efficacy might be compromised by currently circulating mutant pfcrt and pfmdr1 alleles. This is instructive to the ongoing Phase Ib ZY19489 + ferroquine trials in Gabon where mutant pfcrt, which could provide a potential background for N246H to emerge, constituted an estimated 53% of all samples in a recent report⁴⁸. It is also important to point out that this inverse susceptibility phenomenon could still be overcome by cross-resistance, especially in proteins with high mutational plasticity like PfCRT and PfDHODH. For instance, in vitro selection experiments with DSM265 in combination with TCMDC-125334 (active against all PfDHODH mutant lines) yielded parasites with a PfDHODH V532A mutation that were cross-resistant to both compounds and as fit as the wildtype parent⁴⁹. Additionally, a mathematical quantification and experimental evolution study in Escherichia coli highlighted that a second drug can stochastically select for increased susceptibility or increased resistance when following a first⁵⁰. Nonetheless, the absence of overt resistance under the ZY19489-containing regimens tested and the steep fitness cost associated with PfCRT N246H suggest that clinical resistance to this antimalarial is unlikely to arise as mutant parasites would likely be rapidly outcompeted in the field. We, however, urge caution in this interpretation since longer selection or different dosing might still identify additional resistance pathways. We note that this mutation has not been observed in any of the ~20,000 genomes present in the MalariaGEN Pf7 data.

The significant attenuation of short Hb-derived peptide levels in ZY19489-treated parasites suggests that perturbation of Hb endocytosis and/or catabolism constitutes one mechanism through which this compound exerts its antiplasmodial activity. This metabolic fingerprint corroborates a prior clustering of ZY19489 with CQ³¹. Further, our results are consistent with an important role of PfCRT mutations in dictating levels of short oligopeptides in parasites. Aberrant accumulation of peptides has been observed in variant PfCRT isoforms and is predicted to impact the pool of essential amino acids available for protein synthesis, osmotic integrity, and the number of merozoites that can develop within the confines of infected RBCs^30,51,52. This suboptimal physiological context agrees with the lower parasite multiplication rate and reduced parasite fitness observed in the PfCRT N246H mutant in our study and the PPQ resistance-conferring F145I variant earlier studied^53,54. The anomalous peptide accumulation in the N246H mutant also suggests that one of the native roles of PfCRT involves shuttling of peptides from the DV lumen to the cytosol for subsequent liberation of amino acids and alleviation of osmotic stress from the DV, with this function being impaired in drug-resistant PfCRT mutants. This interpretation is supported by a study of PfCRT isoforms expressed in Xenopus laevis oocytes that demonstrated PfCRT-mediated transport of peptides ranging from 4 to 11 amino acids, with the 3D7 isoform (CQ-sensitive) permitting a greater quantity and broader range of peptides than the Dd2 (CQ-resistant) isoform⁵⁵. Our analysis also identified HVDD (a truncated version of HVDDM) and VDPVNF, both of which were suggested to be putative PfCRT substrates in [³H]CQ cis-inhibition experiments in the X. laevis study⁵⁵.

The absence of direct interaction between ZY19489 and ‘free’ heme, or the putative Hz nucleation factor PfHDP, or the hemoglobinases falcipains 2A and 2B, suggests possible inhibition of other targets within the Hb catabolism pathway. The ZY19489 potency observed against rings and schizonts in the stage-specificity experiments offers a clue. Fluorescence and electron tomography studies have shown that, by the mid-ring stage, parasites have detectable acidified compartments that contain small Hz crystals⁵⁶. Additionally, Hb uptake and degradation has been shown to begin in early rings⁵⁷. The assembly of an active Hb degradation machinery so early in the parasite’s lifecycle could explain, at least in part, the high susceptibility of the ring stage to ZY19489. On the other hand, pyrimidine biosynthesis peaks during the trophozoite to schizont transition⁵⁸. Our metabolomic analysis also revealed an accumulation of deoxyuridine monophosphate, implying a potential impact on pyrimidine biosynthesis and corroborating the observed stage-specific potency against schizonts. This likelihood of an additional MoA could help account for the compound’s high barrier to resistance in the conditions tested. Our combined molecular dynamics simulations, heme fractionation and metabolomics studies lead us to propose that ZY19489 might act primarily in the parasite cytosol, and that the N246H mutation might enable mutant PfCRT to sequester this compound in the DV, away from its cytosolic site of action. In contrast, evidence suggests that CQ and PPQ act primarily in the DV by inhibiting Hz formation, with mutant PfCRT mediating resistance by effluxing these drugs out of the DV away from their heme target¹⁸.

In summary, we have provided evidence of a novel PfCRT mutation that mediates low-level resistance to ZY19489. Importantly, the resultant mutants have significant growth defects and are sensitized to DV-acting 4-aminoquinolines. Inhibition of mutant PfCRT-mediated PPQ transport by ZY19489 opens the exciting possibility of partnering these two drugs as a strategy to restrict pfcrt from mutating to confer multidrug resistance, potentially producing an effective resistance-refractory non-ACT combination.

Methods

Inclusion and Ethics Statements

The studies were approved by The Art of Discovery Institutional Animal Care and Use Committee (TAD-IACUC), certified by the Biscay County Government (Bizkaiko Foru Aldundia, Basque Country, Spain) to evaluate animal research projects from Spanish institutions according to point 43.3 from Royal Decree 53/2013, from the 1st of February (BOE-A-2013 − 1337). All experiments were carried out in accordance with European Directive 2010/63/EU. Animal experiment results were reported following ARRIVE guidelines (https://www.nc3rs.org.uk/arrive-guidelines), except for disclosure of business trade confidential information. The human biological samples were sourced ethically, and their research use was in accord with the terms of the informed consents. For routine P. falciparum culturing, RBCs were purchased from the Interstate Blood Bank (Memphis, TN) as pooled, de-identified, anonymized blood that was washed to remove any residual leukocytes. Approval for this protocol (AAAU3761) was provided on 28 October 2022 by the Columbia University Institutional Review Board, which deemed this work to be Not Human Subjects Research under 45 CFR 46.

Generation of P. falciparum V-type ATPase subunit D conditional knockdown parasite line

This cKD line was generated by fusing the coding sequence and non-coding RNA aptamer sequences in the 5’ and 3’ untranslated regions (UTRs), permitting translational regulation using the TetR-DOZI system^59,60. Gene editing was achieved by CRISPR/SpCas9 using the linear pSN054 vector that contains cloning sites for the left homology region (LHR) and the right homology region (RHR) as well a gene-specific guide RNA under control of the T7 promoter. Cloning into the pSN054 donor vector was carried out as described previously^59,60. The vector includes preinstalled V5-2×hemagglutinin (HA) epitope tags, a 10× tandem array of TetR aptamers upstream of an Hsp86 3’UTR, and a multicistronic cassette for expression of TetR-DOZI (regulation), blasticidin S-deaminase (the selection marker) and a Renilla luciferase reporter. The LHR and recoded region were installed in-frame with tandem V5-2×HA tags to generate a C-terminal epitope-tagged V1-A coding sequence, upstream of the regulatory aptamer array. The final construct was sequence-verified and further confirmed by restriction digests. Transfection into Cas9- and T7 RNA polymerase-expressing NF54 parasites was carried out by pre-loading erythrocytes with the donor vector⁶¹. Parasite culture was maintained continuously in 500 nM anhydrotetracycline (Sigma-Aldrich 37919). Drug selection with 2.5 μg/mL of Blasticidin hydrochloride (RPI Corp B12150-0.1) was initiated four days after transfection. Cultures were monitored by Giemsa smears and Renilla luciferase measurements.

Verification of conditional knockdown cKD strategy

To assess regulation of the PfV1-D protein expression, cKD parasites were cultured in the presence (500 nM) or absence of aTc. Protein samples were extracted after 72 h via saponin lysis and resuspended in parasite lysis buffer that consists of 4% SDS and 0.5% Triton X-114 in PBS. Proteins were separated on a Mini-PROTEAN® TGX™ Precast Gels (4–15% gradient) in Tris-glycine buffer, transferred to a polyvinylidene fluoride (PVDF) membrane using the Mini Trans-Blot Electrophoretic Transfer Cell system according to the manufacturer’s instructions, and blocked with 100 mg/mL skim milk in TBS/Tween. Membrane-bound proteins were probed with mouse anti-HA (1:3000; Sigma H3663) and rabbit anti-GAPDH (1:5000; Abcam AB9485) primary antibodies, and anti-mouse (1:5000; Thermo Fisher Scientific 62-6520) and anti-rabbit (1:5000; Cell signaling 7074S) horseradish peroxidase (HRP)-conjugated secondary antibodies. Following incubation in SuperSignal® West Pico Chemiluminescent substrate (Thermo Fisher Scientific PI34080), protein blots were imaged and analyzed using the ChemiDoc™ MP System and Image Lab 5.2.0 (Bio-Rad).

Assessment of parasite growth and compound susceptibility assays using cKD lines

Assessment of parasite proliferation rate upon protein perturbations was carried out using luminescence as a readout of growth. Synchronous ring-stage parasites cultured in the presence (500 nM) or absence of aTc were set up in triplicate wells in 96-well U-bottom plates (Corning 62406-121). Luminescence signals were quantified at 0 and 72 h after incubation using the Renilla-Glo(R) Luciferase Assay System (Promega E2750) and the GloMax® Discover Multimode Microplate Reader (Promega). The luminescence values in the knockdown condition (6 nM aTc) were normalized to 500 nM aTc-treated (100% growth) and 200 nM DHA-treated (no growth) samples. Results were visualized on bar graphs using GraphPad Prism (version 10).

To assess compound activity, ZY19489 and bafilomycin A (a V-type ATPase inhibitor) were serially diluted in complete tissue culture medium. Synchronous ring-stage PfV1-D cKD and control parasites were maintained in high (500 nM) and low (6 nM) aTc or no aTc (for the control line) and were distributed into the drug plate. DMSO and 200 nM DHA treatments served as reference controls. Luminescence was measured after 72 h as described above on four different occasions, and IC₅₀ values were obtained from dose-response curves using GraphPad Prism.

P. falciparum resistance selections with ZY19489

In vitro selection was conducted using a pulsing (intermittent drug pressure) method on Dd2-Polδ (Supplementary Fig. 2). This is a Dd2 parasite line with defective proof-reading activity following CRISPR/Cas9-based replacement of D308 and E310 - the two conserved catalytic residues of P. falciparum DNA polymerase δ, with alanine. This impaired proof-reading activity endows Dd2-Polδ with a ~ 5 to 10-fold higher mutability rate in the presence of a drug compared to Dd2¹². Briefly, 10⁹ parasites in triplicate flasks were exposed to 10× ZY19489 (100 nM) for 11 days until all parasites cleared. The cultures were thereafter passaged in drug-free media with weekly introduction of fresh red blood cells (RBCs) until day 34 when parasites reappeared in two flasks. In vitro drug sensitivity testing confirmed these recrudescent parasites were still susceptible to ZY19489 and subsequently pressured again with 100 nM ZY19489 for another 16 days. During this period, parasites in the third flask also recrudesced but were also exposed to drug since there was no shift in their drug response. Upon clearance on day 50, cultures were again passaged in drug-free media until day 71 when they reappeared. All flasks still had drug-sensitive Dd2-Polδ parasites and were exposed to another round of 100 nM ZY19489 for 19 days until all parasites cleared. Drug pressure was then removed until day 112 when only one flask recrudesced. We further pressured the parasites from this flask for another 12 days before removing drug. On day 136, recrudescent parasites appeared in this flask, which upon testing showed a decrease in drug sensitivity compared to untreated Dd2-Polδ parasites. No parasites were recovered from the other two flasks after day 160 and those selections were halted. Four clones were obtained from the resistant flask by limiting dilution. IC₅₀ values were determined in 72-h drug assays, with parasitemia measured by flow cytometry of cultures stained with SYBR Green and MitoTracker Deep Red (Life Technologies), with ~10,000 cells analyzed per well using an iQue Plus (Sartorius).

72-h in vitro drug susceptibility assays

Susceptibility to all drugs tested in this study was measured by incubating asynchronous, ABS parasites plated at 0.2% to 0.3% parasitemia (or 0.8% for the poorly growing ZY19489-resistant lines harboring the pfcrt N246H allele) and 1% hematocrit at 37 °C in 96-well plates. The assays were set up across a range of drug concentrations with two-fold dilutions. For the experiments testing for susceptibility of recrudescent parasite lines that emerged in humanized NSG mice to ZY19489, M5717, MMV693183, or ferroquine, parasites were cultured at 2% hematocrit in human O⁺ or A⁺ RBCs in RPMI-1640 media, supplemented with 25 mM HEPES (Fisher), 50 mg/L hypoxanthine (Sigma Aldrich), 2 mM L-glutamine (Cambridge Isotope Laboratories, Inc.), 0.21% sodium bicarbonate (Sigma Aldrich), 0.5% (w/v) Albumax (Invitrogen), 8% filtered, heat-inactivated, pooled off-the-clot AB⁺ human serum (Interstate Blood Bank or New York Blood Center), and 10 μg/mL gentamicin (Fisher). The parental P. falciparum Pf3D7^0087/N9 strain that has been adapted to proliferate in this humanized NSG mouse model was cultured in parallel as the drug-sensitive control strain. All compound susceptibility assays for the recrudescent lines and parental control strain were performed using 8% serum-containing medium. Parasite growth in each well was assessed after 72 h using flow cytometry of cultures stained with SYBR Green I and MitoTracker Deep Red (Supplementary Fig. S12). IC₅₀ and IC₉₀ values were derived from growth inhibition data using nonlinear regression analysis (GraphPad Prism 7) or linear interpolation (assays for recrudescent lines from humanized mice). Unless otherwise stated, assays were performed on four independent occasions with technical duplicates, and statistical comparisons were made using two-tailed Mann–Whitney U tests.

Whole genome sequence analysis

DNA from infected RBCs was extracted using the QiAmp DNA Blood Mini kit (Qiagen). The samples were then pooled and sequenced on an Illumina MiSeq flow cell to obtain 300 bp paired end reads. The sequence data were aligned to the Pf3D7 reference genome (PlasmoDB version 48_Pfalciparum3D7; https://plasmodb.org/plasmo/app/downloads/release-48/Pfalciparum3D7/fasta/) using the Burrows-Wheeler Alignment (BWA version 0.7.17). PCR duplicates and unmapped reads were filtered using Samtools (version 1.13) and Picard MarkDuplicates (GATK version 4.2.2). The reads were realigned around indels using GATK RealignerTargetCreator and base quality scores were recalibrated using GATK Table-Recalibration. GATK HaplotypeCaller (version 3.8) was used to identify all possible variants in clones which were filtered based on quality scores (variant quality as function of depth QD > 1.5, mapping quality > 40) and read depth (depth of read > 5) to obtain high-quality single nucleotide polymorphisms that were annotated using snpEFF. The list of variants from resistant clones were compared against the Dd2-Polδ parental clone to obtain homozygous single nucleotide polymorphisms present exclusively in the resistant clones. Integrated Genome Viewer was used to confirm polymorphisms present in resistant clones. BIC-Seq version 1.1.2 was used to discover copy number variants against the parental strain using the Bayesian statistical model.

Targeted sequencing of P. falciparum eukaryotic elongation factor 2 (PfeEF2) and analysis of mutations

The gene encoding PfeEF2 (PF3D7_1451100), which mediates resistance to M5717, was PCR-amplified and sequenced as previously described²⁸. Sequences were aligned to wildtype PfeEF2 from the 3D7 genome reference strain and analyzed using Geneious 9.1.8. Electropherograms were visually inspected to identify mixed sequences indicating multiple subpopulations.

Gene editing using zinc-finger nucleases and CRISPR/Cas9 strategies

pfcrt was edited using a customized two-plasmid zinc-finger nuclease approach that replaces the endogenous allele with a recombinant allele containing the mutations of interest¹⁴. Mutations were introduced into pfcrt^Dd2, pfcrt^GB4 and pfcrt^FCB donor plasmids by site-directed mutagenesis. The D233N mutation was inserted through CRISPR/Cas9 editing, using the previously described “all-in-one” method⁶², with a 610 bp repair template. Two gRNAs were selected based on their selectivity and their proximity to D233, and gRNA primers were annealed and cloned into the pDC2 vector at the BbsI restriction site as previously described⁶². Repair templates containing the D233N mutation and silent shield mutations were synthesized and cloned into the pDC2 vector by In-Fusion cloning at the EcoRI/AatII sites. Dd2 P. falciparum ABS parasites were cultured in RBCs in RPMI 1640-based culture media containing 0.5% w/v Albumax. Parasites were cultured at 37 °C in an airtight chamber with 5% O₂/ 5% CO₂/ 90% N₂. Ring-stage parasites were electroporated with 50–100 µg of purified plasmid DNA in Cytomix. The donor plasmids carrying the human dhfr marker were selected with 2.5 nM WR99210 (Jacobus Pharmaceuticals), and the zinc-finger nuclease plasmid harboring blasticidin S-deaminase was selected with a 6-day pulse of 2 µg/ml blastidicin hydrochloride. Editing was confirmed using PCR and Sanger sequencing, and clones were obtained by limiting dilution. The primer codes and sequences are provided in Supplementary Table 17).

Quantitative trait locus (QTL) analysis and mapping

The generation of the genetic cross between NF54 and RF7, cloning and WGS analysis of the progeny were conducted as described previously¹⁶. Unique recombinants were identified using WGS. The R package (version 2) was used to map QTL peaks. To identify significant QTLs for each drug response phenotype, 1000 permutations of phenotypic data (IC₅₀) were performed to obtain a distribution of maximum log of the odds (LOD) scores. These scores were then used to calculate the LOD threshold at 95% probability. Fine mapping of the QTL segments was performed using Bayesian interval mapping at a 95% confidence level.

Modeling PfCRT in the open-to-DV and open-to-cytosol conformations

The cryo-EM structure of PfCRT strain 7G8 (PDB: 6UKJ)¹⁷ was used as the reference protein template. To include the missing residues (114-122), we employed AlphaFold2, with the original PDB structure serving as the template for modeling. The same approach was used for the N246H mutant, substituting asparagine at position 246 for histidine in the primary sequence. The resulting 7G8 and 7G8 + N246H structures were highly similar, with a root-mean-square deviation (RMSD) of less than 1 Å across all residues. The open-to-cytosol conformation of PfCRT was obtained from previous work¹⁷. The missing segments and residues (81-88, 110-125, 148-154, 169-180, 198-214, 235-251, 269-322, 336-344, 364-381) were modeled using SwissModel⁶³, except for the 277-315 region, which was truncated due to the presence of two disulfide bridges. The final model exhibited an RMSD of 0.9 Å, calculated over 143 pruned atom pairs, when compared to the open-to-cytosol structure from Berger and colleagues⁶⁴, indicating a high degree of structural similarity between the two models. The same procedure was employed to model the mutated 7G8 + N246H isoform.

Molecular dynamics simulations

The protonation states of the amino acids in all models were assigned using PDB2PQR⁶⁵, with pH values of 5.5 for the open-to-DV conformation and 7.0 for the open-to-cytosol conformation. The ionization state of ZY19489 was checked using MolGpKa⁶⁶. To obtain an initial pose of the ligand in each system, molecular docking was performed using GOLD, defining a 12 Å radius around the alpha carbon of residue 246 as the binding site. The initial 3D structure of the ligand was generated with Omega⁶⁷. The 7G8 and 7G8 + N246H docking poses obtained for the open-to-DV and open-to-cytosol systems were then embedded into an explicit lipid bilayer using CHARMM-GUI Membrane Builder⁶⁸. The membrane composition included 1-palmitoyl-2-oleoyl-sn-glycerol-3-phosphocholine (POPC), 1-stearoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine (SOPE) and cholesterol (CHL) in a 4:3:3 ratio, based on the approach from Berger and colleagues⁶⁴. A 150 mM NaCl concentration was added to neutralize the system. The systems were solvated with TIP3P (transferable intermolecular potential with 3 points) water molecules, resulting in ~131k atoms per system.

For the force fields, ff14SB was used for the protein, Lipid21 for the lipids, and GAFF2 for the ligand, with atomic charges derived using AM1-BCC. All-atom molecular dynamics simulations were performed in Amber24 using pmemd.cuda. The systems were initially energy-minimized using the steepest descent method for 10,000 steps. For equilibration, we followed the six-step protocol from CharmmGUI, which includes two NVT simulations with the Langevin thermostat and four NPT simulations at 303.15 K. Finally, a 300 ns production run was performed in triplicate for each system, totaling 3.6 µs of simulation time. To analyze the results from the molecular dynamics simulations, we used cpptraj and MDAnalysis⁶⁹. Representative conformations of each system were obtained through k-means clustering, generating 10 clusters over 500 iterations, based on the ligand-protein RMSD for heavy atoms only.

Drug transport assays

PfCRT 7G8 and 7G8 + F145I variants were purified and inhibition of CQ and PPQ uptake was measured using PfCRT-containing proteoliposomes as described previously¹⁷ with the following modifications. Purified PfCRT variants were reconstituted in preformed liposomes made of E. coli total lipids:cholesteryl hemisuccinate in a ratio of 97:3 (w/w) and the lumen of the proteoliposomes was composed of 100 mM KPi, pH 7.5 and 2 mM β-mercaptoethanol. Uptake of 100 nM [³H]CQ (1 Ci/mmol) or 100 nM [³H]PPQ (1 Ci/mmol) was performed by diluting PfCRT-containing proteoliposomes (30 ng of PfCRT per reaction) in an uptake buffer containing 100 mM Tris/MES, pH 5.5, in the presence or absence of 1 μM of the test compounds. In addition, 1 μM valinomycin was added to the reaction to generate a K⁺ diffusion potential-driven membrane potential (ΔΨ). Reactions were stopped after 30 seconds by the addition of ice-cold 100 mM KPi, pH 6.0, and 100 mM LiCl, and filtered through 0.45 μm nitrocellulose filters (Millipore). Filters were dried and incubated in a scintillation cocktail, and the radioactivity captured on the filters was counted in a Hidex SL300 scintillation counter. The efficiency of detection was calculated with a standard curve of known concentrations of each radiolabeled compound, and this was used to transform decays per minute (dpm) into pmol. The nonspecific interaction of each compound with nitrocellulose filters was determined by measuring mock uptake in the absence of liposomes or proteoliposomes. These values (determined for each experiment) were used to calculate background uptake in liposomes or proteoliposomes. Drug-specific uptake was determined by subtracting the time-dependent accumulation of the tested compounds in control liposomes (lacking PfCRT) from that measured in PfCRT-containing proteoliposomes¹⁷.

In vitro growth rates and fitness assays

We measured relative growth rates of the pfcrt mutant parasite lines by tracking the growth of the parasites and their isogenic parents over eight generations. Briefly, all parasites were seeded at 0.2% starting parasitemia and 2% hematocrit on day 0 with fresh RBCs. The parasites were then allowed to grow undisturbed for the first 3 days, after which parasitemia was assessed by flow cytometry. Media was changed without disturbing the settled culture and culturing was continued until day 4 when parasitemia was measured again. The parasite cultures were cut back to 0.2% parasitemia on day 4, with the dilution ratio recorded. Parasitemia measurements were subsequently done on day 6, 8, 10, 12, 14 and 16, with dilutions made back to 0.2% parasitemia for any cultures that had expanded to >2% parasitemia. These experiments were conducted on three different occasions with technical duplicates. Parasite growth rate was computed as the square root of parasitemia on the day of dilution divided by 0.2% (starting parasitemia) for days 4, 8, 12 and 16. Mean growth rate was obtained from the average of this analysis over three biological repeats. The mean growth rate for the mutant lines was then computed as a proportion of the isogenic parent. The mean fitness cost associated with PfCRT N246H mutation was calculated relative to its isogenic parent using the following formula:

$${\%\mathbf{Mean}\mathbf{FC}}^{\mathit{pfcrt}\mathbf{N}\mathbf{246}\mathbf{H}}=\mathbf{100}\times \left(\mathbf{1}\text{-}\left({\mathbf{MGR}}^{\mathbf{mutant}}÷{\mathbf{MGR}}^{\mathbf{parent}}\right)\right)$$

where FC ^pfcrtN246H is the fitness cost associated with the N246H mutant relative to the parental line and MGR is the mean growth rate per two-day cycle.

Stage-specific profiling of antimalarial activity

Determination of the stage-specific activity of ZY19489 and control compounds was conducted as reported elsewhere with minor modifications²⁹. Briefly, highly synchronized early ring-stage parasites were obtained via 70% Percoll gradient purification of mature schizonts followed 5 h later by sorbitol treatment of young rings upon reinvasion of RBCs. The 40–42 h ABS cycle duration of 3D7 was exploited to allow for three different parasite incubation phases corresponding to rings (0–12 h), trophozoites (18–30 h) and schizonts (30–42 h). Culture synchronicity was confirmed by microscopic observation. After each incubation period (12 h), drugs were removed from the culture through three rounds of washing using pre-warmed media followed by plate change. All pipetting and washing steps were performed using a TECAN Freedom Evo 100 to increase throughput and accuracy and avoid cross-contamination of wells. Each set of plates per time-point was placed in a separate humidified chamber to avoid any delay in growth rate due to temperature variations. For the three different stages, growth inhibition was assessed at the 60-hour time point, when parasites had expanded, reinvaded new RBCs, and developed into trophozoites to allow for accurate quantification by flow cytometry. Parasite survival for both the 72 h and stage-specific 12-h exposures was assessed by SYBR Green and MitoTracker Deep Red FM staining (Life Technologies) and subsequent flow-cytometric analysis (IntelliCyt iQue3, Sartorius). IC₅₀ values were derived from growth inhibition data using nonlinear regression (Prism 7, GraphPad) following six independent biological repeats with two technical replicates.

In vivo efficacy studies in the TADhuMouse-Pfalc model

The in vivo efficacy studies were performed in a standardized commercial model of P. falciparum infection in NOD-SCID IL-2Rγnull (NSG) mice engrafted with human erythrocytes (TADhuMouse®-Pfalc), as described previously²³. Briefly, 22–28 g female NSG mice (Charles River, France) were engrafted with human erythrocytes obtained from Centro Vasco de Transfusiones y Tejidos Humanos (Galdakao, Basque Country, Spain), Centro de Transfusiones de la Comunidad de Castilla y León, (Valladolid, Spain), Centro de Transfusión de la Comunidad de Madrid, (Madrid, Spain) and the Banc de Sang I Teixits (Barcelona, Spain). The erythrocytes were injected by intraperitoneal daily inoculation of 1 mL of 50 - 75% hematocrit cell suspensions in RPMI 1640 medium, 25% (v/v) decomplemented human serum, and 3.1 mM hypoxanthine. When the chimerism of human RBCs in peripheral blood was > 40%, TADhuMouse® were infected by intravenous inoculation via the lateral vein with 0.3 mL of a suspension containing 1.17 × 10⁸ RBCs parasitized with the competent drug-sensitive strain P. falciparum Pf3D7^0087/N9 prepared from peripheral blood of CO₂-euthanized donor mice harboring 5–10% parasitemia²⁴. After random assignment of mice to their corresponding group, drug treatments started when TADhuMouse® had ∼1.3% patent parasitemia (Day 1) in peripheral blood. Treatments were administered by oral gavage with 20 G straight reusable feeding needles (Fine Science Tools GmbH) at 10 mL·kg_bodyweight⁻¹ unless otherwise stated. The drugs administered in combination were co-formulated as fine homogeneous suspensions.

Parasitemia was measured by flow cytometry in serial 2 μL samples of peripheral blood collected from the tail lateral vein and expressed as the % of parasitized RBCs with respect to the total RBCs in circulation. A qualitative analysis of the effect of treatment on P. falciparum Pf3D7^0087/N9 was assessed by microscopy with Giemsa-stained blood smears and flow cytometry by staining with TER-119-Phycoerythrine (a marker of murine erythrocytes) and SYTO-16 (nucleic acid dye) and acquisition in an Attune NxT Acoustic Focusing Flow Cytometer (Invitrogen)⁷⁰.

The concentrations of drugs were measured in samples of peripheral blood (25 μL) taken at different times during dosing, mixed with 25 μL of Milli-Q H₂O, and immediately frozen on a thermal block at −80 °C. The frozen samples were stored at −80 °C until analysis. Blood from control humanized mice was used for the preparation of standard curves, calibration, and quality control purposes. The drugs were extracted from 10 μL of lysates obtained by protein precipitation of blood samples using standard liquid-liquid extraction methods. The samples were analyzed by LC-MS/MS for quantification in a Waters Micromass UPLC-TQD (Waters, Manchester, U.K.). Blood concentration versus time was analyzed by non-compartmental analysis using Phoenix WinNonlin version 8.2 (Certara) or R or Excel (Microsoft), from which exposure-related values (t_max, C_max, and AUC_0–t) were estimated. The clearance of parasitized RBCs from the peripheral blood of mice was assessed by measuring the parasite reduction ratio calculated as the ratio between parasitemia at Day n + 2 divided by parasitemia at Day n for each individual of the study. The parasiticidal effect of drugs in vivo was assessed by measuring the day of recrudescence as the day at which parasitemia after drug treatment reached the % parasitemia at treatment inception⁷¹. Drug-treated mice were deemed cured if there was no detectable parasitemia 60 days after infection.

Sample preparation for untargeted LC-MS metabolomics

3D7 and Dd2 P. falciparum parasite strains were cultured under standard conditions at 2% hematocrit and tested for Mycoplasma using a MycoAlert PLUS Mycoplasma Detection Kit (Lonza). Excluding the cycle preceding sample collection, the Mycoplasma-free parasites were sorbitol-synchronized in each generation for at least two generations and subsequently cultivated to the trophozoite stage with a parasitemia of ~10%. Infected RBCs were then magnetically purified using MACS CS columns and VarioMacs magnets. Metabolite extracts were obtained from the purified infected RBCs treated for 2.5 h with concentrations corresponding to 10× IC₅₀ of atovaquone or ZY19489, resulting in a 1 mL extraction composed of 90% MeOH. For each strain, one biological repeat extraction was run in technical triplicate consistent with how previous metabolomics datasets have been obtained for 3D7 and Dd2 strains of P. falciparum in our laboratory³¹. Extracts were spiked with a 0.5 µM [¹³C₄,¹⁵N₁]-labeled aspartate standard³¹, vortexed, and centrifuged for 10 minutes at 4 °C at 21,130 × g. The supernatant was then transferred into 1.5 mL tubes, dried under a stream of nitrogen, and stored at -80 °C.

Metabolomic profiling of drug-treated parasites via mass spectrometry

Upon thawing on ice, extracts were resuspended in ice-cold HPLC-grade water containing 1 µM chlorpropamide to reach a concentration of 10⁶ cells/µL. After vortexing and centrifugation for 10 minutes at 4 °C at 21, 130 × g, supernatants were transferred into 800 µL CRIMP vials in preparation for mass spectrometry. Samples were fractionated using an XSelect HSS T3 2.5 µm C18 Waters column with a 25-minute gradient of 3% aqueous methanol, 15 mM acetic acid, and 10 mM tributylamine and were processed in negative ionization mode on a Thermo Exactive Plus orbitrap.

Data acquired from the mass spectrometry were imported into the El-MAVEN software package for peak picking⁷² and were standardized using the [¹³C₄,¹⁵N₁] aspartate standard. The metabolites were confirmed by comparing their retention times with those of pure molecular standards run on the same mass spectrometer. Fold changes were computed and uploaded onto Rstudio (http://www.rstudio.com/) employing the Hyperspec and Suprahex packages⁷³. Metabolic profiles, as represented by log₂ fold changes, were transferred onto a hexagonal map featuring 113 metabolites.

Inhibition of recombinant P. falciparum cGMP-dependent protein kinase (PfPKG)

Full length PfPKG (PF3D7_1436600) was expressed in OverExpress™ C41(DE3) pLysS Chemically Competent Cells (Sigma Aldrich, CMC0018) using established methodology^33,74. Briefly, the N-terminal His-tagged recombinant PfPKG protein was purified using a TALON column (GE Healthcare). The final buffer composition of purified protein was 25 mM HEPES pH 7.5, 20 mM NaCl, 120 mM KCl, and 5% glycerol. Purified protein was concentrated using a 15-mL Amicon Ultra 50 KDa MWCO concentrator (Merck Millipore, UFC905024). PfPKG IC₅₀ assays were performed based on previously described methods using the ADP-Glo Kinase Assay (Promega) to measure ADP formation⁷⁵. Briefly, 3-fold serial dilutions of each inhibitor were prepared in 100% DMSO, and inhibitors were subsequently diluted into assay buffer (25 mM HEPES pH 7.4, 0.1 mg/mL BSA, 0.01% (v/v) Triton-X 100, 20 mM MgCl₂, 2 mM DTT, 10 μM cGMP) to 10× the final required concentration. 0.5 μL of each inhibitor dilution was transferred into a white 384-well plate (Greiner 781075). 4.5 µL of a mix containing ATP, GRTGRRNSI-NH2 and PfPKG in assay buffer was then added to each well. The final 5 μL kinase reaction contained 3 nM PfPKG protein, 10 μM ATP, 20 μM GRTGRRNSI-NH2, 1% (v/v) DMSO and inhibitor in assay buffer. Reactions were incubated for 30 minutes at 22 °C.

ADP formation was measured using the ADP-Glo Kinase Kit (Promega). Briefly, 5 μL ADP-Glo reagent was added to each well and incubated for 30 minutes at 22 °C to deplete the remaining ATP. 10 μL of Kinase Detection Reagent was then added and the reaction incubated for a further 30 minutes at 22 °C. The plate was sealed with an adhesive foil seal for all incubation steps. Luminescent signal was measured using a Spectromax Plate Reader. Data were normalized based on the 100% activity controls (1% DMSO only) and the 100% inhibition controls (10 μM kinase inhibitor Staurosporine). Mean IC₅₀ values were calculated from N, n = 3,4 independent experiments.

Generation and drug susceptibility testing of P. falciparum heme detoxification protein (PfHDP) cKD lines

PfHDP cKD parasites were generated using a CRISPR/Cas9-based TetR-DOZI aptamer system⁶⁰. Briefly, the left homology region fused with recodonized 3’ end of PfHDP and right homology region as well as guide sequence were cloned into a pSN054 donor plasmid containing a 3×HA epitope tag at the 3’ end of the coding sequence. The final construct together with the pDC2-cam-coSpCas9-U6-gRNA-hDHFR plasmid carrying the guide sequence was transfected into Dd2 parasites and maintained in 500 nM aTc and 2.5 μg/mL blasticidin hydrochloride. Edited parasites were confirmed by site-specific integration PCR.

In vitro drug sensitivities of ABS parasites were determined as described above with minor modifications. Asynchronous Dd2 and PfHDP cKD parasite lines were cultured in the presence of 500 nM aTc and then washed to remove aTc. Asynchronous ABS parasites were plated at 0.3% parasitemia and 1% hematocrit in 96-well plates and were incubated with a ten-point, three-fold range of drug concentrations in duplicates with either 30 nM, 3 nM and 0 nM aTc. Plates were incubated at 37 °C for 72 hours and parasite survival was assessed by flow cytometry using SYBR Green (Invitrogen) and MitoTracker Deep Red FM (Life Technologies) as nuclear stain and vital dyes, respectively. IC₅₀ and IC₉₀ values were derived from growth inhibition data using nonlinear regression or linear interpolation (Prism 10, GraphPad) as means ± SEM from four independent biological repeats with two technical replicates.

Inhibition of P. falciparum falcipain 2 A and 2B knockout parasites

Falcipain knockout lines were generated as described previously³⁵. Susceptibility to ZY19489 and E64d was measured against the 3D7-A10 and 3D7 FP2A/B KO lines by incubating asynchronous, ABS parasites plated at 0.3% (or 0.5% for 3D7 FP2A/B KO) parasitemia and 1% hematocrit at 37 °C in 96-well plates. The assays were set up across a range of drug concentrations with two-fold dilutions. Parasite growth in each well was assessed after 72 h using flow cytometric analysis of cultures stained with SYBR Green I and MitoTracker Deep Red. IC₅₀ and IC₉₀ values were determined by nonlinear regression analysis. Unless otherwise stated, assays were repeated on four independent occasions in technical duplicates and statistical comparisons were made using two-tailed Mann-Whitney U tests.

In vitro cell-free beta-hematin inhibition assay

Inhibition of lipid-mediated hemozoin formation was quantified by measuring the formation of β-hematin dimers from hemin chloride, in a detergent-mediated assay that substitutes neutral lipids for the commercially available lipophilic detergent Nonidet P-40³⁶. Unreacted hematin was detected through the formation of bis-pyridyl-Fe(III)PPIX complex which absorbs at a wavelength of 405 nm.

Compound stocks (20 mM) of ZY19489, amodiaquine and pyrimethamine were prepared in DMSO, while CQ was prepared in water. These stocks were then diluted to 2 mM with a water/NP40 detergent solution, resulting in a final solution of compound in 61.1 mM NP40/10% DMSO. Serial dilutions into this same buffer were then made for each compound. A 25 mM hemin stock solution was prepared by sonicating hemin in DMSO for one minute and then suspending 179 μL of this stock in a 1 M acetate buffer (20 mL, pH 4.8). The homogenous suspension was then added to the wells to give final buffer and hemin concentrations of 0.5 M and 100 mM, respectively. Plates were covered and incubated at 37 °C for 5 h to enable β-hematin formation. A solution of 50% (v/v) pyridine, 30% (v/v) H₂O, 20% (v/v) acetone and 2 M HEPES buffer (pH 7.4) was prepared and 32 μL added to each well to give a final pyridine concentration of 5% (v/v). Acetone (60 μL) was then added to each well to assist with β-hematin dispersion. The ultraviolet-visible (UV-vis) absorbance of the plate wells was read on a SpectraMax P340 plate reader. Sigmoidal concentration-response curves were fitted to the absorbance data using GraphPad Prism to obtain an IC₅₀ value for each compound.

Cellular heme fractionation assay

The IC₅₀ values used here for ZY19489, CQ and pyrimethamine (PYR) were obtained from drug susceptibility experiments using the 72-h SYBR Green assay described above. The heme profiles of NF54 parasites treated separately with the three antimalarials were determined as described previously³⁷. Briefly, NF54 cultures were synchronized at 48-h intervals with 5% (w/v) sorbitol, and ring-stage parasites at 5% parasitemia and 2% hematocrit were incubated with drug concentrations corresponding to 0.5×, 1×, 1.5×, 2× and 2.5× their IC₅₀ values. Untreated infected RBCs were included as control. Parasites were harvested 30 h post incubation, and the mature trophozoites were isolated with 0.05% (w/v) saponin and washed with 1× PBS (pH 7.5) to eliminate traces of residual RBC-derived Hb. The number of trophozoites in these samples were quantified using flow cytometry. The contents of the trophozoite pellet were then released by hypotonic lysis and sonication. Following centrifugation, the supernatants (first fraction) corresponding to membrane-soluble Hb were treated with 4% (w/v) SDS and 2.5% (v/v) pyridine. Pellets were again treated with 4% SDS, 2.5% pyridine, sonicated, and centrifuged. Supernatants corresponding to the ‘free’ heme fraction (second fraction) were then recovered. The remaining hemozoin pellets (third fraction) were then solubilized in 4% SDS and 0.3 M NaOH, neutralized with 0.3 M HCl, sonicated, and treated with 25% pyridine. The UV-vis spectrum of each heme fraction as a Fe(III) heme-pyridine complex was measured using a multiwell plate reader (Spectramax 340PC; Molecular Devices). The total amount of each heme species was quantified using a heme standard curve³⁷. The mass of each heme Fe species per trophozoite was calculated by dividing the total amount of each heme species by the corresponding number of parasites in that fraction³⁷. Statistical comparisons were made using two-tailed Mann-Whitney U tests using GraphPad Prism.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

J.O., I.A-B., J.C.N., F.M., A-C.U., S.M., M.Q., E.A.W., D.L., M.L., V.T and D.A.F. conceived and designed the experiments. T.N., L.H and K.E.W. generated the edited PfCRT and PfAPC10 mutant lines. H.P. conducted whole-genome sequencing, and T.Y. analyzed the raw sequence data. T.Q. generated all metabolomics data and plots. S.K.D and C.F.A.P. generated conditional knockdown lines. J.K. and E.G-I generated purified PfCRT proteins and conducted drug transport experiments. R.V.C.G and V.B. performed molecular dynamics simulations. J.L.B and S.M. performed QTL analysis on the genetic cross. M.B.J-D and I.A-B. conducted the in vivo studies in mice while V.T. conducted the subsequent in vitro screening and genotyping of recrudescent parasites. I.M.R.M, S.K.N, T.N. and L.V.O. performed drug phenotypic analysis on the edited and conditional knockdown lines. J.O and I.A. generated all the manuscript figures. J.O and D.A.F. wrote the manuscript with contributions from other authors. All authors approved the final manuscript.

Peer review

Peer review information

Nature Communications thanks Daniel Inaoka, and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The metabolomic data generated in this study are publicly accessible through the NCBI Metabolomics Workbench under the Project ID: PR002514. The raw reads of WGS data generated in this study have been deposited in the NCBI Sequenced Read Archive (SRA) under BioProject accession number PRJNA1376518 and are accessible via the link (https://www.ncbi.nlm.nih.gov/bioproject/1376518). The 3D7 reference genomes used in the analyses are available in the PlasmoDB database. Other datasets analyzed during the current study are provided with this paper as Supplementary Material (Supplementary Fig. 1–12 and Supplementary Tables 1–17) and Source Data File. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-70914-1.