Artemisinin-resistant Plasmodium falciparum Kelch13 mutant proteins display reduced heme-binding affinity and decreased artemisinin activation

Abstract

The potency of frontline antimalarial drug artemisinin (ART) derivatives is triggered by heme-induced cleavage of the endoperoxide bond to form reactive heme-ART alkoxy radicals and covalent heme-ART adducts, which are highly toxic to the parasite. ART-resistant (ART-R) parasites with mutations in the Plasmodium falciparum Kelch-containing protein Kelch13 (PfKekch13) exhibit impaired hemoglobin uptake, reduced yield of hemoglobin-derived heme, and thus decreased ART activation. However, any direct involvement of PfKelch13 in heme-mediated ART activation has not been reported. Here, we show that the purified recombinant PfKelch13 wild-type (WT) protein displays measurable binding affinity for iron and heme, the main effectors for ART activation. The heme-binding property is also exhibited by the native PfKelch13 protein from parasite culture. The two ART-R recombinant PfKelch13 mutants (C580Y and R539T) display weaker heme binding affinities compared to the ART-sensitive WT and A578S mutant proteins, which further translates into reduced yield of heme-ART derivatives when ART is incubated with the heme molecules bound to the mutant PfKelch13 proteins. In conclusion, this study provides the first evidence for ART activation via the heme-binding propensity of PfKelch13. This mechanism may contribute to the modulation of ART-R levels in malaria parasites through a novel function of PfKelch13.

Introduction

Malaria remains a major global health problem, with 249 million cases and 608,000 deaths reported in 2022 in the 84 malaria-endemic countries¹. Artemisinin (ART) derivatives are key components of ART-based combination therapies (ACT), the current frontline treatment for uncomplicated malaria. ART derivatives contain an endoperoxide bridge that is cleaved within the malaria parasite by newly released hemoglobin-derived iron sources (predominantly ferrous iron, ferrous protoporphyrin IX (FPIX)/heme, or both) to form activated oxy-centered and C4-centered radicals². These target the C-C, C-H, and other bonds in parasite proteins/lipids, and the resulting alkylation leads to extensive molecular damage and parasite death^3,4,5,6,7,8. Alkylation of protein targets by ART radicals appears to be non-selective as two different click-chemistry studies using ART probes found little overlap in the set of alkylated parasite proteins^6,9. Other anti-parasitic mechanisms include inhibition of β-hematin crystallization and heme detoxification by heme-drug adducts and, paradoxically, of endocytosis-mediated uptake of hemoglobin, the source of ART activator^10,11. The role of heme-ART adducts is further supported by the detection of those adducts by LC-MS in live malaria parasites grown under culture conditions^12,13. Non-heme (inorganic) iron, heme and hemoglobin are all implicated as key molecules in ART activation in vitro¹⁴. Observations in the parasite support a critical role of hemoglobin digestion and heme, either recycled from hemoglobin or newly synthesized, in ART activation and toxicity^6,7,15,16.

The emergence and spread of resistance to ARTs in Plasmodium falciparum threatens progresses made in reducing the global malaria burden over the past two decades following introduction of ACTs. First reported in the Greater Mekong Subregion in 2008^17,18, P. falciparum parasites with reduced susceptibility to ARTs are now widely reported in several regions of the world^19,20,21. Resistance to ART is defined either clinically by a delayed parasite clearance half-life in vivo or by increased parasite survival in vitro after briefly exposing early ring-stage parasites (0–3 h posterythrocyte invasion; hpi) to a high dose of dihydroartemisinin (DHA) in the ring-stage survival assay (RSA)^18,22. Currently, mutations in the P. falciparum Kelch13 (PfKelch13) protein are confirmed determinants of clinical ART resistance^23,24, although single nucleotide polymorphisms in the P. falciparum actin-binding protein Coronin and Ap2-mu have also been implicated in increased in vitro tolerance to ART via PfKelch13-independent mechanisms^25,26,27. PfKelch13 is a 726 amino acid protein encoded by the pfkelch13 gene (PF3D7_1343700) on chromosome 13 and contains structurally conserved domains: a coiled-coil-containing (CCD) domain, an intervening region (IVR), a broad complex, tramtrack and bric-à-brac (BTB) / poxvirus and zinc finger (POZ) domain, and a kelch domain (also known as Kelch-repeat propeller, KRP domain with six β-propeller blades) which encompasses almost all the clinical ART resistance mutations^23,28. The PfKelch13 C580Y and R539T mutations are the most common in ART-resistant strains in Southeast Asia^29,30, while the ART-R C469Y, R561H, R622I, and A675V mutants have emerged more recently in East Africa^31,32,33. The A578S mutation is also common throughout Africa but does not confer ART resistance^29,30.

Several studies have shown reduced steady-state cellular levels of PfKelch13 proteins in ART-resistant C580Y and R539T mutant parasites^34,35,36,37. Conditional mislocalization of PfKelch13 using the diCre-based system resulted in the developmental arrest of the replicating intraerythrocytic parasite at the late ring-stage, highlighting the essential role of PfKelch13 at a stage where the ART-resistant phenotype also manifests^22,38,39. GFP-tagged and native PfKelch13 were found to localize to membranous compartments proximal to the parasite food vacuole (FV) and to cytostomal structures near the parasite periphery^34,35, both mislocalized in ART-resistant PfKelch13 mutants, which also exhibit reduced endocytic uptake of host cytosol and free heme level at the ring stage^34,36. It is now recognized that PfKelch13, together with other proteins, regulates hemoglobin endocytosis and subsequent degradation, thereby modulating the level of heme-mediated activation of ART into toxic molecules. PfKelch13 also colocalizes with parasite endoplasmic reticulum, Rab-positive vesicles, and partially with parasite mitochondria upon DHA treatment⁴⁰. Of note, the three PfKelch13-associated malarial structures (cytostome, FV, and mitochondria) are associated with either heme-loaded hemoglobin capture, hemoglobin degradation or heme synthesis. Cytostome-derived vesicles loaded with host hemoglobin may also contain newly released free heme from early vesicular degradation of hemoglobin⁴¹, whereas large amounts of heme are released during the digestion of hemoglobin in the FV^42,43,44. The parasite mitochondrion, together with the plastid, are sites of de novo heme synthesis, and mitochondrial metabolism can modulate ART resistance. Indeed, deletion of the malaria mitochondrial protease PfDegP has been reported to reduce heme levels and DHA susceptibility⁴⁵. Thus, the detection of PfKelch13 proximal to heme/hemoglobin-containing cellular structures (either cytosolic or membrane-bound) and the evidence of a labile heme pool (∼1.6 μM) stably maintained in the parasite⁴⁶ raises the possibility of direct interaction(s) between PfKelch13 and heme. In the literature, we identified one KRP domain with iron binding properties from the field pennycress Thlaspi arvense thiocyanate-forming protein (TaTFP) of the Brassicaceae family^47,48. Previous studies have shown increased activity of TaTFP upon Fe²⁺ supplementation as compared to Fe³⁺, while the other metal ions were ineffective^47,48,49,50.

Here, we sought to test the hypothesis that PfKelch13 could interact with iron and heme. Recently, we reported the purification of recombinant PfKelch13 WT and A578S, C580Y, and R539T mutant proteins expressed in Escherichia coli⁵¹. These recombinants led us to investigate the in vitro binding of purified PfKelch13 to iron and heme, as well as the production of heme-ART adducts by PfKelch13-bound heme.

Results

The KRP domain of PfKelch13 shares structural features with the iron-binding KRP domain of Thlaspi arvense TFP

First, we compared the KRP domain of PfKelch13 with that of the enzymatically active, iron-binding KRP of the field pennycress Thlaspi arvense thiocyanate-forming protein (TaTFP) of the Brassicaceae family^47,48 (Fig. 1A). Binding of TaTFP to iron directly involved amino acid positions of the KRP domain (E266, D270 and H274, while W309 supported the positioning of H274)^52,53. By performing pairwise sequence and 3D structure alignments of the KRP domains of PfKelch13 (RCSB-PDB ID 4YY8, 1.8 Å) and TaTFP (RCSB-PDB ID 5A11, 2.5 Å), amino acid E266 of TaTFP aligned both in sequence and structure with E688 of PfKelch13. The side chains of both TaTFP-E266 and PfKelch13-E688 were also oriented towards the shallow pocket of KRP within the respective resolution limits of their 3D structures (Fig. 1B-D and Supplementary Fig. 1). Among the other amino acids, W706 of PfKelch13 only showed alignment in sequence with W309 of TaTFP; while the other amino acids, i.e., E691 and H697 of PfKelch13, did not align in either sequence or structure with D270 and H274 of TaTFP (Fig. 1B-D). We also searched for putative iron (Fe²⁺ or Fe³⁺)-interacting residues in silico within the amino acid sequences of the full-length PfKelch13, the truncated PfKelch13 TrK13-WT(as reported in our previous study⁵¹ with the amino acids encoding the step-tag sequence at the C-terminus) using the MIB server (http://bioinfo.cmu.edu.tw/MIB/)⁵⁴. Typically, 95 and 117 amino acid residues in the full-length PfKelch13, and 97 and 110 different amino acid residues in TrK13-WT were predicted to bind Fe²⁺ and Fe³⁺, respectively, including six and three residues from the strep affinity tag in TrK13-WT (summarized in Supplementary Table 1). Taken together, these results prompted us to biochemically test the binding ability of different metal ions to the recombinant PfKelch13.

Fig. 1: Structural and functional similarities in the iron-binding properties between the KRP domains of PfKelch13 and TaTFP.

A Schematic representation of the domain organization in full-length PfKelch13 (top), recombinant TrK13-WT (middle), and TaTFP (bottom), showing the respective spans of the coiled-coil domain (CCD), intervening region (IVR), bric-à-brac; (BTB)/poxvirus and zinc-finger (POZ) domain (BTB/POZ), and the Kelch-repeat propeller (KRP) domain. The three specific mutations in the full-length PfKelch13: R539T, C580Y and A578S are indicated in red text. The iron binding residues in TaTFP and corresponding homologous amino acids in TrK13 are shown in blue text. B Sequence alignment between a part of the KRP domains of TaTFP and PfKelch13. Residues highlighted in yellow represent positional conservation between PfKelch13 and those in TaTFP involved in Fe²⁺ binding. C–D Structural alignment of the PfKelch13 propeller (blue) and TaTFP (gold) proteins. Protein structures are shown as cartoons. Iron-binding amino acid residues in TaTFP (italic) and homologous sites in PfKelch13 (bold) are shown as sticks (in D). E SDS-PAGE and Western blot profiles showing affinity-purified TrK13 proteins of ~61 kDa (arrowheads) and their specific recognition by antibodies. F–G UV-visible spectroscopy scans showing the increase in absorbance profiles of TrK13-WT only in the presence of FAS and FES, but not with other metal salts (G). The corresponding spectrum of free salts in the absence of any TrK13-WT protein is shown in F. The spectrum profile of TrK13-WT protein without any metal salt is shown as an inset. The graphs shown are best representative of three independent experiments. H. Graph showing the relatively higher binding of the TrK13-WT protein to FAS, as measured by absorbance at 595 nm using the Ferene S assay, compared to the TrK13-QNG mutant protein. Data are shown with their respective standard errors from three independent experiments. I. SDS-PAGE showing higher affinity of TrK13-WT protein (arrowhead) for FAS-, FeSO₄- and FES-NTA beads compared to Ni-NTA beads. Input samples are shown below. J-K. Inhibition of TrK13-WT (J) or TrK13-R539T (K) binding to FAS-NTA or FES-NTA beads with stepwise increase of free FAS or FES concentrations. Molecular weight standards (in kDa) for all SDS-PAGE gels are as shown.

Recombinant PfKelch13 protein shows specific binding to salts of iron but not other divalent cations

The cloning, expression and purifications of recombinant PfKelch13 proteins from E. coli were performed as described in our previous study⁵¹. Similar to our earlier observations, the recombinant purified PfKelch13 was deleted in the Apicomplexa-specific N-terminal segment (TrK13). All the TrK13 variants (WT and R539T, A578S and C580Y mutants) migrated as a ~ 61 kDa band in SDS-PAGE (Fig. 1E and Supplementary Fig. 2). TrK13 proteins were immunodetected in western blots using either commercially available antibodies against the strep tag or custom-generated antibodies against a PfKelch13 peptide⁵¹. We then looked for evidence of iron (and then heme) binding using different methods. Firstly, the TrK13 proteins (5 μM) incubated in solution with 100-times molar excess (500 μM) of individual metal salts (MnCl₂, CuSO₄, MgSO₄, ZnCl₂, NiSO₄, ferrous ammonium sulfate or FAS, and ferric ammonium sulfate or FES) were analyzed by UV-visible spectrophotometry. At these concentrations, the free metal salts alone showed no observable absorbance in the 250–700 nm spectrum (Fig. 1F). Interestingly, TrK13-WT showed a dramatic increase in absorbance at all the wavelengths in presence of the two iron salts FAS and FES (Fig. 1G). At 280 nm (representing the cumulative absorbance by the aromatic amino acids tyrosines, tryptophans, and phenylalanines), we observed a 5.82-times and 12.84-times increase in the absorbance of TrK13-WT in the presence of FAS and FES, respectively, relative to TrK13-WT alone (no salts). The solubility of TrK13-WT remained unchanged in the absence or presence of FAS or FES at these concentrations (Supplementary Fig. 3). This FAS- and FES-dependent increase in absorbance was a proxy for ligand binding and was also detected in TrK13-C580Y, TrK13-R539T and TrK13-A578S mutant proteins (Supplementary Fig. 4). Altogether, these data suggested that TrK13 binds iron salts and that the three tested mutations did not largely affect iron binding.

The host erythrocyte infected by the parasite contains high levels of hemoglobin, which comprises >95% of the cytosolic protein and approaches concentrations of 5 mM (reviewed in^55,56). The concentration of labile and hemoglobin-bound iron in uninfected erythrocytes has been previously reported as 1.37-4.72 µM and 17.92 ± 1.05 µM, respectively⁵⁷. During the parasite intraerythrocytic cycle, 60-80% of the host hemoglobin is internalized^41,58 and proteolytically degraded by plasmepsin and falcipain at the parasite FV^59,60,61. Yet the precise distribution and dynamics of iron in malaria-infected erythrocytes are not well delineated and remain as emerging areas of study⁶². Nonetheless, it is well accepted that low pH and digestive enzymes of the food vacuole render hemoglobin as a major source of ferrous (Fe²⁺) heme^59,60,61,63 for the parasite. Oxygen disassociated from the Fe²⁺ heme oxygenated complex, accepts an electron from the Fe²⁺ heme, to yield to oxidized Fe³⁺ heme and releases O₂^•-, a reactive oxygen species (ROS⁶⁴; which in turn dismutates to other ROS⁶⁵). It is also widely accepted that ART (and other antimalarial endoperoxide pro-drugs) are activated by Fe²⁺ heme released from hemoglobin in the FV. The oxygen free radicals generated induce covalent drug adducts and ROS effects of DNA damage and lipid peroxidation^3,66,67. During this process, Fe²⁺ heme is concomitantly re-oxidized Fe³⁺ heme^3,68. Even in the absence of drug, the FV contains very high levels of a Fe³⁺ heme that exist as microcrystalline cyclic dimeric ferriheme through Fe1-O41 bonds⁶⁹ that stacks as lattices to constitute the black pigment of malaria, namely hemozoin^70,71. Approximately 88 ± 9% of the FV iron has been reported to be recovered as hemozoin pellet from infected erythrocytes⁷². Interestingly, a recent inferred that non-crystalized heme at the FV was most likely associated with undigested hemoglobin⁷¹. Regardless, Fe³⁺ heme does not cleave ART’s endoperoxide bond, suggesting that Fe²⁺ species is the major drug-activating determinant, that may be derived from hemoglobin in the late ring/trophozoite stages or de novo heme synthesis in mitochondria and/or apicoplast in earlier ring stages^73,74,75.

Since it is widely accepted that antimalarial endoperoxide pro-drugs are activated by Fe²⁺ heme to form covalent drug adducts^3,67,76, during which Fe²⁺ heme is reoxidized back to Fe³⁺ heme^3,68, we investigated the binding of TrK13-WT to ferrous iron salt (FAS). This was done using the in-solution Ferene S colorimetric assay. Here, iron was reduced to Fe²⁺ by the addition of ascorbic acid, followed by bidentate chelation of Fe²⁺ by Ferene to form a stable blue complex between pH 3.0 and 6.0⁷⁷. The TrK13-WT protein (5 μM) was incubated with a 30-times molar excess (150 μM) of FAS, and the mixture was then desalted by size-exclusion chromatography using a PD Spintrap™ G-25 desalting column. Approximately 10-fold increase in the absorbance at 595 nm was detected in the eluate for TrK13-WT incubated with FAS, as compared to the FAS alone or TrK13-WT alone controls, which were similarly processed (Fig. 1H). We also included a recombinant TrK13 variant (TrK13-QNG) with E688Q-E691N-H697G point mutations at the three positions homologous to the Fe²⁺ binding positions in TaTFP^52,53, based on the sequence alignment between TaTFP and PfKelch13. We observed ~33% reduction in absorbance at 595 nm in the TrK13-QNG variant compared to TrK13-WT. Taken together, these results indicated that FAS molecules co-purified with the TrK13 proteins, thereby suggestive of iron binding. Since our in silico analysis predicted ~100 Fe²⁺-binding sites distributed across the TrK13-WT sequence (Supplementary Table 1), we did not attempt generating other mutants.

We next evaluated the binding of TrK13-WT to iron by using immobilized-metal affinity chromatography, in which the metal ion was coupled to a resin matrix via nitrilotriacetic acid (NTA). We tested three independently immobilized iron matrices (i.e., FAS-NTA, FeSO₄-NTA, and FES-NTA; collectively referred to as Fe-NTA)^78,79 and Ni-NTA as a control. TrK13-WT (5 µM) was incubated with the different matrices in neutral pH buffer under rotating conditions, and after extensive washing, the matrix-bound proteins were eluted with Laemmli buffer and separated by SDS-PAGE. After Coomassie staining, we observed between 30 and 50% binding of the ~61 kDa TrK13-WT protein to the different Fe-NTA matrices and ~3-5 times better binding as compared to control Ni-NTA beads (Fig. 1I). We also observed a reduction in the bound TrK13-WT protein levels when excess free FAS or FES (0.1 to 10 mM) was added during the bead incubation period, suggestive of specific binding (Fig. 1J). A similar trend was also observed for the ART-R mutant TrK13-R539T (Fig. 1K).

In our earlier study, we reported the homo-hexameric assembly of TrK13-WT by small-angle X-ray scattering (SAXS), which remained stable in the presence of up to 2 M urea concentration. At higher urea concentrations, the hexameric TrK13-WT assembly was found to dissociate into ~61 kDa monomeric units⁵¹. We therefore tested the binding of TrK13-WT to FAS-NTA in the presence of increasing urea concentrations from 0 to 6 M. Compared to the sample without urea, we observed ~ 50% reduction in the amount of eluted TrK13-WT at urea concentrations of 2 M and beyond (Supplementary Fig. 5). Since the monomeric TrK13-WT unit still retained significant FAS-NTA-binding, the results suggested that iron binding was not strictly dependent on the hexameric assembly of the TrK13-WT protein.

The recombinant WT and mutant TrK13 proteins show specific interactions with heme

We subsequently investigated if the TrK13 protein also binds to the iron-containing heme molecule. In specific experiments, we included the ART-R TrK13-R539T and TrK13-C580Y and the ART-S TrK13-A578S recombinant proteins for comparative analysis with the TrK13-WT protein.

At first, we qualitatively evaluated the heme-binding potential of TrK13-WT by in gel staining^80,81,82 using 3,3′,5,5′ tetramethylbenzidine (TMBZ). TrK13-WT (10 µM) was incubated with equimolar heme (hemin chloride reduced to heme in the presence of 2 mM sodium dithionite) for 30 min and resolved by native PAGE. After gel staining with TMBZ, an intense bluish-green band was detected at >180 kDa (Fig. 2A), a molecular mass reminiscent of the TrK13-WT oligomer previously reported by us⁵¹. Without heme co-incubation, no TMBZ staining was seen with either the TrK13-WT protein or the proteins from the molecular mass standards. The light bluish-green smear tail was likely, at least in part, due to the background TMBZ staining⁸⁰. The parallel Coomassie stained replicate gel showed a similar staining pattern (Fig. 2A). These data indicated that heme molecules were predominantly co-migrating with the TrK13-WT protein even under native electrophoretic conditions.

Fig. 2: The recombinant TrK13-WT protein shows specific interactions with heme/FPIX. A.

Native PAGE of TrK13-WT shows a bluish-green TMBZ-positive band (arrow) only in the heme-preincubated sample (right lane) but not in the absence of heme (left lane). Corresponding Coomassie-stained SDS-PAGE of TrK13-WT protein with or without preincubated heme are shown in the left gel. B–C UV-visible spectroscopy scans showing increase in the relative absorbance profiles of G-25 desalted TrK13-WT (red) and positive control BSA (blue), but not negative control lysozyme (green), in the presence of heme (C) compared to their absorbance profiles without heme (B). No soret peak of free heme (black) is detected in C at this concentration (10 μM), indicating effective heme removal by the G25 desalting column. D–E Differential absorption spectra of TrK13-WT protein (10 µM) titrated with heme (0-26 µM) in the wavelength region 250–700 nm (D) and double reciprocal plot of 1/ΔA versus 1/[heme] (E), showing the variation of difference in absorbance of TrK13-WT-heme complex with heme concentrations. The K_d and R² values are as indicated. F Far-UV circular dichroism (CD) spectra of TrK13-WT protein (10 μM) titrated with heme (0–30 μM) in the wavelength region 200-260 nm. G UV-visible spectroscopy scans showing the increase in the relative absorbance profiles of G-25 desalted TrK13-WT protein (10 μM) in the presence of 10 μM heme/FPIX (solid red line) or its analogs such as Zn-PPIX (solid blue line) or PPIX (solid green line) compared to the control (solid black line). The spectral profiles of the remaining unbound heme variants after the G-25 desalting step are shown as corresponding colored dotted lines and indicate efficient removal as no soret peak was visible. H-I. SDS-PAGE (H) and graph (I) showing the relative abundance of TrK13-WT protein (arrow) pulled down with hemin-agarose beads in the absence or presence of excess free heme, PPIX, or Zn-PPIX. Molecular weight standards (in kDa) for all SDS-PAGE images are as shown.

We then evaluated the binding of heme to the WT and three mutant TrK13 proteins by using immobilized heme affinity chromatography with hemin-agarose beads⁸³. After extensive washing, the bead-bound proteins were eluted with Laemmli buffer and resolved by SDS-PAGE. All the four TrK13 proteins were detected in the Coomassie-stained gel (Supplementary Fig. 6). The elution yield was ~30–40% of the initial input for each TrK13 variant, suggesting no major effect of these mutations on the hemin-agarose binding property of the TrK13 proteins. Somewhat unexpectedly, the TrK13-WT protein also showed no change in hemin-agarose binding affinity in the presence of urea at concentrations up to 6 M (Supplementary Fig. 5). It is important to note that the qualitative hemin-agarose experiments were limited by fixed heme concentration in the beads and the possibility of steric constraints on heme accessibility by the proteins, and thereby do not likely reflect the relative binding affinities of individual TrK13 proteins.

To overcome these limitations, the TrK13-WT protein (10 µM) was incubated with an equimolar concentration of heme (reduced with sodium dithionite) for 30 min and the mixture was desalted using the G-25 columns. The UV-visible spectral profile of the desalted mixture showed a significant increase in the absorbance spectrum throughout the 250-700 nm wavelength, as compared to the desalted TrK13-WT alone control (Fig. 2B), with a prominent Soret peak at 380 nm, typical for heme-containing molecules (Fig. 2C). Our desalting procedure using a G-25 column efficiently removed free heme molecules, as no Soret peak was visible after desalting the heme only sample (Fig. 2B). The TrK13-WT-heme mixture showed a hypsochromic or blue shift from 410 nm to 380 nm, corresponding to the Soret peak. Of the two control proteins incubated with heme and processed similarly, BSA (a confirmed heme-binding protein^84,85,86) showed a larger Soret peak, whereas lysozyme (with no heme-binding affinity⁸⁷) showed no increase in absorbance (Fig. 2C). To calculate the dissociation constant (K_d) for the TrK13-WT heme complex, the protein was incubated with increasing concentrations (0-26 µM) of heme, desalted using G25 columns, and spectrophotometrically analyzed (Fig. 2D). The double reciprocal plot of 1/ΔA versus 1/[heme] was a straight line with K_d value of 36.22 ± 0.1 nM (Fig. 2E).

The effect of increasing concentrations of FPIX/heme on the secondary structures of TrK13-WT protein and BSA (as a control) were also measured by circular dichroism (CD) spectroscopy (Fig. 2F and Supplementary Fig. 7). In the absence of heme, the CD profiles revealed primarily β-pleated secondary structures with traces of α-helical regions. In the presence of 10 μM heme, changes in the secondary structures were observed with increase in molar ellipticity, particularly affecting the 222 nm minimum, suggesting an enhancement in helical content⁸⁸. At higher heme concentrations (20 and 30 μM), no major changes were evident. In summary, these studies suggested that changes occurred in the helical content, but not in the overall secondary structure of the TrK13-WT protein or BSA (Supplementary Fig. 7) on heme-binding, which could also be due to the location of heme binding sites far from the aromatic amino side chains in the protein, thereby not causing a large variability in the secondary structure.

The TrK13-WT-heme interaction involves both the Fe²⁺ ion and the PPIX porphine core

The heme molecule consists of protoporphyrin IX (PPIX), a tetrapyrrole macrocyclic porphine core with a Fe²⁺ cation in the center. To investigate whether the TrK13-WT heme interaction involves either or both the components, the TrK13-WT protein (10 μM) was incubated with equimolar concentrations of heme or heme analogs such as PPIX (without the Fe²⁺ ion) or Zn-PPIX (with Zn²⁺ replacing the Fe²⁺ cation), desalted through G-25 columns, and analyzed by UV-visible spectrophotometry. As previously observed for heme, there was an increase in the absorbance spectrum of TrK13-WT regardless of the heme analog tested. Specifically, distinct peaks with absorbance maxima between 300-450 nm wavelengths were visible in all cases and characteristic of the heme analog used (Fig. 2G; and absent in the protein alone control). Efficient desalting was confirmed by the absence of such maxima peaks in the control samples: this excluded any contributions from free heme or its analogs. To test the binding of heme analogs, the TrK13-WT protein (10 µM) was incubated with hemin-agarose beads in the presence of 10-fold molar excess of each heme analogs. After extensive washing, co-incubation with heme or analogs decreased the amount of TrK13-WT eluted (Fig. 2H-I). Heme displayed the greatest inhibition ( ~ 90%) compared to PPIX ( ~ 50%) and Zn-PPIX ( ~ 65%). Taken together, these results strongly suggested the involvement of both the Fe²⁺ metal ion and the PPIX moiety in TrK13-WT-heme interactions.

Kinetics of TrK13 tryptophan fluorescence quenching by heme

In tryptophan fluorescence quenching, the tryptophan residues within a protein are selectively excited at a wavelength of 295 nm and the emission spectrum, which peaks at 300-350 nm, is measured in the absence or presence of a ligand^89,90. The full-length PfKelch13 has seven tryptophan residues distributed along the entire stretch of the TrK13 sequence at positions 229, 490, 518, 565, 611, 660 and 706, while the recombinant TrK13 proteins (WT and mutants) have nine tryptophan residues (two additional tryptophan residues are contributed by the one-strep tag fusion at the C-terminus; Supplementary Fig. 8). The fluorescence spectra of TrK13-WT, -R539T, -C580Y, and -A578S proteins exhibited peak maxima between 332–337, 338–341, 336–338, and 326–333 nm, respectively, and were gradually quenched by increasing concentrations of heme (up to 5 µM; Fig. 3A–D). Blue shifts to lower wavelengths (i.e., 329 nm) upon heme binding were observed for all TrK13 proteins, which could be attributed to the formation of a heme-protein complex accompanied by increased hydrophobicity around the tryptophan sites^84,91. Interestingly, 50% quenching of tryptophan fluorescence was achieved at heme concentrations between 2-2.40 μM for the ART-sensitive TrK13-WT and -A578S proteins, while it was achieved at lower concentrations (0.90-1 μM) of heme for the ART-resistant mutants TrK13-R539T and -C580Y. The Stern-Volmer plots showed an upward curvature for all TrK13 proteins, especially at higher heme concentrations, suggesting either combined fluorescence quenching or the existence of >1 heme-binding sites with different accessibility to the tryptophan residues in TrK13 proteins⁸⁵ (Fig. 3E). Values for the Stern-Volmer quenching constant K_SV (determined only for the linear part of the Stern-Volmer plot⁹²) excluded the effect of collisional quenching in water. Similarly, the values for the quenching rate constants (k_q) were in the order of 10¹³, which is generally considered to be well above the maximum k_q due to collisional quenching⁹³. These results further suggest the possibility of binding interactions between the TrK13 proteins and heme molecule to form the ground-state complexes with the tryptophan indole rings and leading to the fluorescence quenching of the proteins. Interestingly, we observed marked differences in the Stern-Volmer (K_SV) and association constants (K_a) between the TrK13 variants (Fig. 3E-G). The two mutants associated with ART-R in Southeast Asia (TrK13-R539T and -C580Y) showed higher K_SV values (1.17 and 1.75 μM^-1, respectively) compared to those of TrK13-WT and the ART-sensitive -A578S (0.59 and 0.66 μM^-1, respectively). Similarly, the K_a values for TrK13-R539T and -C580Y (7.86E + 07 and 1.47E + 07, respectively) were approximately three-fold lower than those of TrK13-WT and -A578S (at 1.35E + 11 and 2.88E + 11, respectively). The quenching mechanism can be attributed to FRET resulting from a significant overlap between the tryptophan emission and heme absorption. TrK13-WT showed the least quenching, while there was an overall increase in quenching for the mutants, with TrK13-C580Y showing the maximum. Since the extent of energy transfer depends on the distance between the donor (tryptophan residue) and the acceptor (heme), the quenching data suggest that heme binding induces distinct changes in the protein structure/conformation that lead to changes in the distance between tryptophan residues and heme.

Fig. 3: Heme-induced fluorescence quenching of tryptophan residues in the four variants of TrK13 proteins without profound changes in the circular dichroism profile.

A–D Fluorescence spectra of TrK13-WT (A), TrK13-R539T (B), TrK13-C580Y (C), and TrK13-A578S (D) incubated in the absence (dotted lines) or presence of heme at final concentrations between 0.5-5.0 μM heme (gradual light to dark shading in each color) and showing a successive decrease in fluorescence intensity. E Changes in tryptophan fluorescence quenching in the four variants of TrK13 as deduced from A to D. F Plots of log[(F₀-F)/F] versus log[heme] for the interaction of TrK13-WT protein and heme deduced from A to D. The Stern-Volmer quenching constants (K_SV), association constants (K_a) and the number of binding sites (n) were calculated from the equations F₀/F = K_sv [Q] + 1 and log [(F₀-F/F] = logK_a + nLog[Q], where F₀ = fluorescence intensity in the absence of heme, F = fluorescence intensity in the presence of heme, [Q] = concentration of quencher (heme). Curves are the average of three replicates. G Tabular representation of K_sv, K_a and n.

These results were also in good agreement with our earlier structural studies of TrK13 proteins using SAXS, which indicated an increased length of the oligomeric assembly and a more skewed orientation of the KRP domains away from the long axis of the protein structures in TrK13-R539T and -C580Y mutants compared to TrK13-WT and -A578S mutant⁵¹. However, quantitative assessment of the true heme-binding affinities of these recombinants was further estimated by more sensitive assays.

Quantitative evaluation of TrK13-heme interactions using microscale thermophoresis

We quantified the true affinity between heme and TrK13 using microscale thermophoresis (MST) in small glass capillaries with a solution of fluorescent molecules^94,95. The infrared laser-induced temperature gradient allowed excitation of the fluorophores and the emitted fluorescence was collected, which allowed measurement of thermophoresis-dependent depletion of fluorophores within the temperature gradient.

In all our MST experiments, tween 20 (0.05% v/v) was added to help mitigate protein adsorption and aggregation in the narrow uncoated capillaries, and to act as a surfactant to improve protein solubility and limit aggregation⁹⁴. Our preliminary analyses indicated ~50% increase in the heme binding efficiency of TrK13-WT protein (as measured by the characteristic soret peak at 380 nm) in the presence of 0.05% tween 20 using UV-visible spectrophotometry (Supplementary Fig. 9). However, we observed no change in the oligomerization status of the recombinant TrK13-WT protein in the presence of tween 20 (up to the 0.10% v/v), thereby implying that the increased heme binding was not as a result of better accessibility of previously hidden epitopes, exposed upon the disruption of the oligomeric assembly of TrK13-WT protein. We speculated that the effect of tween 20 would essentially resonate with similar relative effects across all the four TrK13 protein-heme interactions, thus we retained tween 20 in our MST studies. For deriving binding constants in MST, we used multiple capillaries with constant concentrations of the fluorescent molecule. When titrated against increasing concentrations of heme, the thermophoretic mobility of TrK13 changed, indicating effective interactions. The dose-response curves showed a negative response amplitude for all TrK13 variants, indicating a lower MST signal of the heme-protein complex compared to TrK13 alone (Fig. 4A–D). Furthermore, equilibrium dissociation constant (K_d) calculations revealed values of 61.52 ± 12.41 nM and 66 ± 17.87 nM for the ART-sensitive TrK13-WT and TrK13-A578S mutant respectively (Fig. 4E-F). Interestingly, higher K_d values were calculated for the TrK13-R539T and TrK13-C580Y mutants (131 ± 14.98 nM and 204 ± 20.44 nM, respectively, Fig. 4F). These results indicated comparatively weaker association of heme with the two ART-R mutants compared to the ART-sensitive proteins.

Fig. 4: Quantitative evaluation of TrK13- heme binding by MST.

A–D Dose-response curves showing negative amplitude of response for TrK13-WT (A), TrK13-R539T (B), TrK13-C580Y (C) and TrK13-A578S (D) titrated against increasing concentrations of heme. MST experiments were performed at medium MST power at 25 °C. LED power was set to 60% (A and D) or 40% (B and C) excitation. E Comparative dose-response curves for the binding interaction between the four TrK13 variants and heme, as indicated. The protein concentration was kept constant at either 50 nM (TrK13-WT or TrK13-A578S) or 25 nM (TrK13-R539T or TrK13-C580Y), while the ligand heme concentration varied from 0-5 µM. F Table showing the decreasing affinity scale: TrK13-WT > TrK13-A578S > TrK13-R539T > TrK13-C580Y and signal-to-noise (S/N) ratio.

In silico docking of heme, oxy-heme, and ART on the models of hexameric TrK13 protein and mutants

Our preliminary studies revealed that the disruption of oligomerization status of TrK13-WT (with gradual increments in urea concentrations) led to only 50% reduction in its Fe-NTA binding property, while the hemin-agarose binding characteristics largely remained unaffected (Supplementary Fig. 5A-C). We then used the SAXS data-driven assembly structures submitted to the SASBDB⁵¹ database as templates for visualization of interactions between TrK13 proteins and heme. It is to be noted that these structures were energy minimized in implicit water using United atom OPLS forcefield prior to submission at the database. Although the resolution of these dummy residue models was between 5-6 Å, they served as best available templates for in silico docking of heme and ART moieties since no crystal structures for TrK13-WT or mutant proteins were reported in the RCSB-PDB databank. In addition, no refined maps from cryo-electron microscopy, restraints from multi-dimensional NMR data, or even DeepMind-based predictions were available. For this study, we considered the coordinates of the residue detail models generated using dummy atom model or SAXS data as reference for docking experiments. To further the stringency, we also applied YASARA Program v6^96,97,98 and YAMBER2 forcefield to the protein assemblies with added hydrogen atoms after energy minimization to a cut-off gradient of 0.01 Å to assess the best-fit model within conformity of atomic geometry. Mostly, hydrogen bonds (H-bonds) and atomic interactions were considered within distances up to 3-4 Å, and these all were within the input resolution of the SAXS data based dummy atom models.

We also envisioned that in the environment accessible to the native PfKelch13, both oxygenated (oxy-heme) and deoxygenated heme (hereafter referred as ‘heme’ in this study) should be available for interaction and binding. Thus, for docking runs, both oxygenated and deoxygenated were considered. Oxy-heme had O₂ molecule in one of the axial positions of Fe³⁺ iron and in heme, Fe²⁺ iron was considered⁹⁹. As mentioned in the methods section, previously reported hexameric assemblies of TrK13-WT and mutants, and SDF files of oxy-heme and heme were used for docking experiments. The respective protein assemblies were considered as static receptors and flexible versions of ligands were compiled as clusters, with poses of ligands within 0.5 Å considered as a part of the same cluster (Fig. 5). Clusters were considered instead of individual poses of ligands as those provided lesser bias on pose-specific values which would likely change on solvation as in wet experiments. Docked clusters were parsed based on calculated theoretical binding energies and their targeted residues and chains are detailed in the log files (Supplementary Datas 1 and 2). At the foremost, we observed no direct interaction (bond formation) between the C580, R539 or A578 residues (corresponding to R351, C392 and A390 residues in the TrK13-protein sequence; Supplementary Fig. 8) in any of the six individual chains of the TrK13-WT structure and the oxy-/heme moieties present in any of the clusters (Fig. 5; top left image). Using the same TrK13-WT structure, we also found no interactions between the oxy-heme moieties and shortlisted amino acid mutants in PfKelch13 that are known to elicit medium/high ART-R indices (Supplementary Table 2). Surprisingly, the predicted Fe²⁺-binding residues in TrK13-WT (based on its sequence similarity with TaTFP; Fig. 1A-B) also showed no interactions with the Fe²⁺/Fe³⁺ moiety in oxy-/heme molecules. We speculate that this could either be due to restrictions in the orientation of iron moiety within the heme molecule, or limitations in the accessibility of corresponding residues within the KRP region of TrK13-WT due to its hexameric association, or from both. We then analyzed the top ten oxy-heme clusters based on their binding energy spread (in kcal/mol; more positive energies indicative of stronger binding). The average dissociation constants (K_d) indicated strong binding affinity (in the nanomolar range) for the ten clusters in all the TrK13 proteins. While the oxy-heme and heme clusters displayed K_d values of 7.90 nM and 5.39 nM, respectively for TrK13-WT, the corresponding K_d values in the ART-R mutants were 16.54 nM and 7.61 nM (TrK13-R539T) and 6.12 nM and 4.94 nM (TrK13-C580Y), respectively. Notably, the average K_d values for the oxy-heme clusters in the ART-S mutant TrK13-A578S were not identical to those exhibited in TrK13-WT. Instead, the TrK13-A578S-oxy-heme and - heme clusters displayed K_d values of 8.96 nM and 7.33 nM, respectively. Remarkably, we observed clear differences in the distribution profile of these oxy-heme clusters between the four TrK13 hexamers. In TrK13-WT, 90% (9/10) clusters were limited to the KRP domains (amino acids 261-537), and only a single (1/10 or 10%) oxy-heme cluster was found at the coiled-coil domain (CCD; amino acids 1-120). In contrast, the oxy-heme clusters were more inclined for regions other than the KRP domains in the ART-R TrK13-R539T and TrK13-C580Y variants. Specifically, these clusters primarily favored (90% for oxy-heme and 100% for heme) docking at the interface between the CCD and BTB-POZ domains (amino acids 162-260) in TrK13-R539T. Similarly, 100% of the top ten clusters in the TrK13-C580Y structure were restricted to CCD domains. We found no oxy-heme clusters at the IVR (amino acids 121-161) region in all the TrK13 variants. Cumulatively, these results largely uphold our prediction that there are multiple interaction sites possible for both versions of heme used on the solution structural assembly of TrK13 and its mutants with significant variation in the preferred epitopes across the same chain or adjacent ones in the TrK13-WT and mutants.

Fig. 5: Docking clusters of oxygenated (oxy-heme) and deoxygenated heme in the hexameric assemblies of the TrK13-WT and mutant structures.

SAXS-derived hexameric assemblies (grey ribbon; derived from https://www.sasbdb.org) of the ART-sensitive (TrK13-WT and TrK13-A578S), and ART-resistant (TrK13-R539T and TrK13-C580Y) variants showing the 38 docked oxy-heme (top panel) and heme (bottom panel) clusters (red). Dotted lines indicate separation between the KRP/β-propeller domain and the CCD + IVR + BTB/POZ domains in the TrK13 structures. Theoretical average theoretical docking scores and other characteristics for the top 10 clusters are shown in the table below. The CCD (purple), IVR (green), BTB/POZ (black) and the KRP domain with six β-propeller regions (orange) are colored accordingly and shown as a schematic above the table. In cases where the oxy-/heme cluster contacted interface between two domains, numbers were assigned to both the domains of TrK13.

Since ART colocalizes with heme to enable adduct formation, we attempted to visualize whether ART favored TrK13-oxy-heme clusters. We thus docked ART to TrK13-oxy-heme complexes using similar constraints, i.e., the protein-oxy-heme assemblies were considered as static receptors and flexible versions of ART as clusters (with poses within 0.5 Å as a part of the same cluster). As a control, we first docked ART to the TrK13 structures (Supplementary Fig. 10). Although ART exhibited a preference for the KRP domains in all the four structures, the respective theoretical binding affinities (K_d values) were largely in the millimolar range, thereby implying much weaker direct interactions. Indeed, recent studies with an alkyne-labeled ART analog AP1 coupled to biotin failed to alkylate PfKelch13^6,9. Interestingly, ART docking onto the TrK13-oxy-heme complex favored the pre-docked oxy-heme clusters (Fig. 6; also see the comprehensive log files provided in Supplementary Data 3 and 4). Thus, the top 10 ART clusters (with theoretical K_d values in the nanomolar range) were largely localized at regions in the KRP domains of TrK13-WT and -A578S where oxy-heme clusters were already present. Similarly, the ART clusters favored the CCD (100% or 80% for pre-docked oxy-heme or heme, respectively) or the CCD-BTB/POZ interface (20% for pre-docked heme) in TrK13-R539T. For TrK13-C580Y, 100% of the ART clusters were in the oxy-heme bound CCD domains.

Fig. 6: Docking clusters of ART in the pre-docked oxy-/heme assemblies of TrK13-WT and mutant structures.

Oxy-heme (red; top panel) and heme (red; bottom panel) clusters (red) complexes with the SAXS-derived hexameric assemblies (grey ribbon; derived from https://www.sasbdb.org) of the ART-sensitive (TrK13-WT and TrK13-A578S), and ART-resistant (TrK13-R539T and TrK13-C580Y) variants showing the 38 docked ART clusters (blue). Dotted lines indicate separation between the KRP/β-propeller domain and the CCD + IVR + BTB/POZ domains in the TrK13 structures. Theoretical average docking scores and other characteristics for the top 10 clusters are shown in the table below. The CCD (purple), IVR (green), BTB/POZ (black) and the KRP domain with six β-propeller regions (orange) are colored accordingly and shown as a schematic above the table. In cases where the ART cluster contacted interface between two domains, numbers were assigned to both the domains of TrK13.

Based on the cumulative observations from our in silico docking studies, we interpret that overall 3D conformation of individual TrK13 hexamers (and their changes due to single point mutations, as previously reported by us⁵¹), likely influenced the docking of oxy-/heme and subsequently prompted ART docking. Currently, we are unable to speculate how the relative preference of oxy-/heme to dock at the KRP regions in ART-S TrK13 variants but not in the ART-R variants play a role in the downstream events leading to ART-R phenotype in parasites with PfKelch13 R539T and C580Y mutations. Further experiments with a larger scope of PfKelch13 mutants will be necessary to offer better interpretations. Nonetheless, these studies offered a good visualization of our biochemical data and paved the way towards subsequent experiments.

The TrK13-bound heme forms heme-ART activated species

A general scheme for endoperoxide bond cleavage in ART derivatives by reduced heme to generate either oxy-radicals or C4-centered ART adducts has been reported previously^3,13,67,100. Because we were unable to distinguish between the peaks corresponding to either oxy-radicals or C4-centered ART adducts in the downstream assays, we chose to use the term ‘ART activated species (AAS)’ to collectively represent the products. We investigated the formation of AAS under reducing conditions by high-performance liquid chromatography (HPLC) after 30 min incubation of heme with ART or analogs (Supplementary Figs. 11 and 12). The time-dependent formation of AAS in our experimental setup is shown in Supplementary Fig. 12B. Heme-containing molecules were detected based on the presence of porphyrins excited at 390–410 nm (Soret peak) whereas ART and DHA have been previously reported to absorb only in the UV region between 190-210 nm¹⁰¹. At a molar ratio of 1:1.20 (heme: ART or respective derivatives), additional peaks with 5 to 10 times lower intensity and slightly delayed elution times were visible in the chromatogram at 410 nm with either ART or DHA compared to heme alone or heme with an inactive analog of ART such as deoxyartemisinin (DOA) (Fig. 7A). The elution time and profile of these additional delayed peaks varied with the derivatives tested, namely DHA or ART (Fig. 7A) or artesunate (AS, Supplementary Fig. 11), and likely corresponded to AAS as reported in previous studies^10,12,13. The formation of AAS was also verified by LC-MS in the in vitro reaction mixtures of heme with or without ART. As shown in other studies, prominent peaks at m/z 529.4, 543.4, 557.2, and 616.2 characteristic of heme and common degradation products generated by in-source fragmentation of heme and oxidation-reduction processes during electrospray ionization were observed in the heme alone sample (Fig. 7B)^12,102. When heme (under reducing conditions) was reacted in vitro with ART (mass of 282.3 Da), additional peaks at larger m/z 801.6, 838.3, and 851.4 were visible and indicative of the AAS (Fig. 7B-C).

Fig. 7: Detection of AAS by HPLC and mass spectrometry.

A HPLC elution profiles at 410 nm for the total reaction mixtures of heme and the corresponding peaks for free heme (tall peaks at 6–7 min) and AAS (smaller peaks at 8–10 min elution time; boxed and magnified in the inset) in the presence of either ART (blue) or DHA (red) or the inactive DOA analog (green). The elution profile of free heme is indicated by the dotted black line. B MS data of the eluted samples from A (dotted box) showing the characteristic heme intensity peaks (red text) at m/z 529.4, 543.4, 616.2 and 557.2 and several other peaks generated by in-source fragmentation of heme and oxidation-reduction processes during electrospray ionization (black text). C Mass spectra of the two-component mixture of heme-ART from the eluted samples of A (dotted box), showing additional AAS peaks at m/z 801.6, 838.3, and 851.4 (blue text). Peaks at these m/z were absent in the heme alone sample shown in B. D–E HPLC elution profiles showing the heme (tall peaks at 5–7 min elution time) and AAS peaks (short peaks at 8–10 min elution time) in TrK13-WT (red) or TrK13-R539T (blue) containing reaction mixtures with heme (D) or in the protein-bound heme (after G-25 desalting; E). The AAS peaks are framed by dotted lines and magnified in the inset. F Percentage of AAS versus heme, as measured by area under the curve (AUC) for the respective sample peaks and expressed as a percentage. Data are mean ± SD of three independent experiments, comparison was performed by unpaired t-test, * p < 0.05, ** p < 0.005, ns not significant (p > 0.05).

Next, we investigated whether AAS formation also occurred in the presence of TrK13-WT and TrK13-R539T proteins. TrK13 was first preincubated with heme for 60 min and then ART was added (at a molar ratio of protein: heme: ART of 1: 6: 30, i.e., 10: 60: 300 μM). The delayed AAS peaks were detected with both WT and mutant TrK13 proteins (Fig. 7D), indicating that their formation was not affected by the presence of TrK13. We noticed that both the larger and smaller delayed peaks eluted at slightly different times depending on which variant of TrK13 was present.

The above assay lacked the ability to distinguish between the AAS formed by TrK13-bound heme and excess-free heme. To address this issue, TrK13 was first incubated with equimolar concentrations of heme and desalted using a G-25 column to remove excess unbound free heme. Finally, ART was added to the desalted heme-TrK13 mixture. The resulting HPLC profile showed an overall 75% reduction in total peak heights compared to the non-desalted reaction conditions (compare peak heights between Fig. 7D and Fig. 7E). Interestingly, our quantitative evaluations based on the relative HPLC-resolved area-under-the-curve for the heme and AAS peak profiles suggested an enhanced AAS-forming ability of the heme molecule(s) when bound to the TrK13-WT protein ( ~ 50%) compared to that of TrK13-R539T ( ~ 30%) or free heme (33%) (Fig. 7F).

We also used UV-visible spectrophotometry to quantitatively visualize and compare the ability of TrK13-bound heme to form AAS under reducing conditions^12,13. The spectral profile of reduced heme alone remained unchanged over the 40-min recording period (Supplementary Fig. 12A). The three combinations of two components (heme + ART, heme + TrK13, and ART + TrK13) were then tested individually as controls. When 2.5 µM heme was incubated with excess ART (50 µM), a dramatic broadening and redshift of the heme Soret band from 400 nm to 412 nm was observed and remained stable during the 60-min experiment (Supplementary Fig. 12B). When heme was incubated with TrK13-WT, we observed immediate hyperchromic and bathochromic/red shifts in the spectral profile of TrK13-WT, with no further changes visible during the 60-min experiment, as in our previous observations (Fig. 2C and Supplementary Fig. 12C). Finally, in the presence of excess ART, there were no visible changes in the spectral profile of TrK13-WT at 250–700 nm compared to the protein alone, suggesting no direct interactions between ART and the protein (Supplementary Fig. 12D).

We then prepared the heme-TrK13-WT mixture and, after desalting with G-25 column, added ART. After 30 min of incubation, we observed a shift in the absorbance spectrum across the range, with the most pronounced changes between 300 and 450 nm (Fig. 8A). Specifically, we observed a hypsochromic/blue shift of the Soret peak from 410 nm to 380 nm (as previously shown in Fig. 2C). In addition, there was a ~ 42% increase in the intensity of the 380 nm peak in the presence of ART compared to heme-bound TrK13-WT without ART (Fig. 8A). The increase in the absorbance peak at 380 nm was likely due to AAS formation, since no increase in the corresponding absorbance was detected in the control samples which were similarly treated with either zinc-containing heme analog (Zn-PPIX) instead of heme (Supplementary Fig. 13A) or the inactive analog DOA instead of ART (Fig. 8B). No increment in the 380 nm absorbance peak was evident after the addition of ART under conditions when the Fe³⁺ hemin was not pre-reduced to Fe²⁺ heme with sodium dithionate, thereby implying that the reactivity of ART was only towards the reduced Fe²⁺ heme (Supplementary Fig. 13B). In addition, we also considered the scenario wherein the increase in absorbance at 380 nm was due to the subsequent oxidation of Fe²⁺ to Fe³⁺ (since TrK13-WT showed higher absorbance when bound to FES as compared to FAS; Fig. 1G). However, such an oxidation event would likely occur during the generation of AAS intermediates, as shown in the schematic for AAS formation previously¹³). The formation of AAS in the same samples was then tested by LC-MS. Discrete heme peaks at m/z 512.5, 540.5, 647.5, and 664.5 were detected in the TrK13-WT-heme sample without ART, whereas additional peaks at 774.6, 801.6, and 851.4 were detected only in the sample additionally containing ART, again suggesting the formation of AAS (Fig. 8C-D).

Fig. 8: Reduced AAS formation in ART-R TrK13 mutants compared to ART-S variants. A, E-G.

UV-visible spectroscopy scans showing further hyperchromic shift (red lines; increase indicated by blue to red gradient dotted arrow) in the absorbance profiles at 380 nm in the heme-bound ART-S TrK13-WT (A) and TrK13-A578S (G) proteins, but not in the ART-R TrK13-R539T (E) and TrK13-C580Y (F) proteins, when incubated in the presence of ART for 30 min. The Soret peaks at 380 nm (red lines) and absorption profiles of the respective proteins in the absence of heme (dotted black lines) are as indicated. B UV-visible scan showing no hyperchromic shift on the Soret peak of heme-bound TrK13-WT in the presence of the inactive ART analog DOA. C–D MS data of the heme-bound TrK13-WT samples (after G-25 desalting) incubated in the absence (C) or presence (D) of ART. While the peaks corresponding to heme (in red text) were visible in both samples, the peaks at m/z 801.6 and 851.4 (blue text) were only visible in the presence of ART (D) and are indicative of AAS. All samples were treated as described earlier¹³. The inset shows the expanded m/z region between 750 and 900. H–I UV-visible spectroscopy scan showing no AAS formation by the heme-binding proteins BSA and HRP2. J Graphical representation of the percentage increase in the Soret peak at 380 nm for the data in A, B, E–I Error bars are from three independent experimental replicates.

A profile similar to that of TrK13-WT was observed for the ART-sensitive TrK13-A578S protein, but with an even greater ~176% increase in peak intensity at 380 nm (Fig. 8G). Surprisingly, the heme-TrK13-R539T and heme-TrK13-C580Y complexes showed barely detectable increase in their respective peak intensities at 380 nm in the presence of ART (Fig. 8E-F), suggesting much lower AAS formation. Neither BSA (a validated heme-binding protein) nor P. falciparum Histidine-rich protein 2 (PfHRP2, a well-established heme-binding parasite protein¹⁰³) showed an increase in peak absorbance at 380 nm, indicating that the ability of proteins to bind heme does not necessarily translate into AAS formation (Fig. 8H-J; Supplementary Fig. 14). In summary, the ART-R TrK13-R539T and TrK13-C580Y proteins exhibited reduced affinity for heme (as also observed in the MST experiments; Fig. 4) and were associated with reduced AAS formation in vitro, as compared to the TrK13-WT protein.

Our in silico docking data suggested noncovalent association between the TrK13 proteins and heme, as also reported for hemopexin-heme interactions¹⁰⁴. Thus, we tested whether some of the AAS formed by the heme-TrK13 complex were released from TrK13 as soluble molecules. For this, the desalted TrK13-WT-bound heme samples were incubated in the absence (control) or presence of ART for 30 min. The reacted samples were then analyzed spectrophotometrically between 250-700 nm wavelengths both before and after a second (final) desalting step. We observed that the spectral profiles of the TrK13-heme sample incubated without ART were qualitatively similar both before and after the final desalting, showing no change in the characteristic soret peak at 380 nm (compare solid and dotted blue lines in Fig. 9A) and essentially indicating efficient retention of heme molecule(s) with no leaching or release of bound heme during the incubation period. Interestingly, UV-visible scanning of the TrK13-WT-heme sample incubated in the presence of ART showed a dramatic decrease in the height of the 380 nm peak after the second desalting step, suggesting release and removal of low molecular weight AAS from their protein-bound state following desalting (compare solid and dotted red lines in Fig. 9A). The height of the 380 nm peak decreased to a level even lower than that observed for the heme-TrK13-WT complex incubated in the absence of ART (compare dotted blue line versus dotted red line). Thus, the residual 380 nm absorbance peak in the ART-treated sample likely represented the retained heme fraction in the TrK13-WT-heme complex that was not involved in AAS formation (Fig. 9A-B).

Fig. 9: Heme bound to TrK13-WT protein is released after AAS formation.

A UV-visible spectroscopy scans showing the heme absorbance peak at 378 nm due to the TrK13-WT-bound heme (solid blue line) after desalting step #1 and its hyperchromic shift upon addition of ART (solid red line). The scan profile of TrK13-WT without heme is shown as a dotted black line. After desalting step #2, a steep decrease (green dotted arrow) of the heme absorbance maximum peak at 380 nm is observed in the ART-treated sample (dotted red line), indicating the release of AAS from the TrK13-WT protein and its removal by the second desalting step. The resultant peak at 380 nm could be due to residual bound heme that did not form AAS. No decrease in λ_max at 380 nm is observed in the TrK13-WT-heme-bound sample processed similarly in the absence of ART (dotted blue line), thus excluding the possibility of random heme release during the incubation period. B Schematic representation of the molecules at various stages of the experiment. Recombinant ART-S (TrK13-WT; black pac-man) or ART-R (TrK13-R539T or –C580Y; blue pac-man) proteins with higher or lower heme-binding affinity (solid red square), respectively were desalted (desalting step #1) to remove the excess free heme or loosely-bound heme. Following incubation with 20-times excess ART (green ball pin), desalting step #2 was performed to remove any AAS released from its TrK13-bound state (green question mark). The orange dotted arrow symbolizes the proposed role of liberated AAS in alkylating parasite protein targets or inhibiting hemozoin formation during the intra-erythrocytic stages of infection. Reaction mixtures at each step were analyzed by UV-visible spectrophotometry spanning 200–700 nm wavelengths.

The native PfKelch13 protein is pulled down from P. falciparum lysate using hemin-agarose beads

Finally, to test the putative relevance of our in vitro observations with respect to the parasite-derived proteins, the RIPA-solubilized extracts from asynchronous P. falciparum 3D7 parasites were incubated with either hemin-agarose beads or glutathione-agarose beads (as negative control) under reducing conditions. After extensive washing, the bound proteins were eluted with Laemmli buffer and resolved by SDS-PAGE. PfKelch13, was detected as a distinct ~80–90 kDa band in western blots in the eluates from hemin-agarose beads by custom-generated anti-PfKelch13 polyclonal antibodies⁵¹ (Fig. 10). In contrast, PfKelch13 was quantitatively reduced in the similarly processed samples from the control glutathione-agarose beads (Fig. 10). Membranes were similarly probed with antibodies against either the parasite P. falciparum histidine-rich protein 2 (PfHRP2, a known heme-binding protein¹⁰⁵) as a positive control). The results revealed a prominent PfHRP2 band in eluates from hemin-agarose beads and not in control glutathione-agarose beads. Altogether, these data indicate that the native PfKelch13 was specifically recovered from the parasite solubilized-proteome through its interaction(s) with hemin-agarose beads. These results were consistent with the hypothesis that native PfKelch13 from the parasite interact with heme.

Fig. 10: Pulldown of the native PfKelch13 protein from P. falciparum total protein extracts using hemin-agarose beads.

3D7 parasites were isolated by saponin treatment from asynchronized culture and proteins were solubilized with RIPA buffer. Protein extracts were incubated with hemin-agarose or glutathione (GSH)-agarose beads. After extensive washing, bound proteins were eluted with Laemmli buffer, separated by SDS-PAGE, and blotted against PfKelch13 antibodies. PfKelch13 antibodies were custom-generated using a PfKelch13 peptide⁵¹. Molecular weight standards (in kDa) are as indicated.

Discussion

Multiple cellular functions, localizations, and protein-interacting partners have been assigned to the Kelch13 protein in the apicomplexan parasites P. falciparum^34,40,106 and Toxoplasma gondii^107,108. Unraveling the molecular consequences of ART-R mutations in PfKelch13 remains an ongoing area of investigation. Overall, PfKelch13 modulates ART sensitivity in the (late) ring stage primarily by regulating the level of hemoglobin uptake at the parasite periphery, which in PfKelch13 mutants translates into decreased hemoglobin degradation into ART activator molecules^34,36 and lower levels of heme-DHA adducts^38,109,110. The overall resistant phenotype (reduced hemoglobin uptake and subsequent ART activation) is primarily controlled by decreased abundance of the mutant PfKelch13 at the cytostome collar ring where hemoglobin uptake is initiated^{34,35,36,78,111}, thereby resulting in atypical invaginations lacking the constricted necks characteristic of cytostomes as reported in PfKelch13-mislocalized parasites³⁶. Multiple alterations in the cellular pathways have also been associated with PfKelch13 resistance mutations^{111,112,113,114,115}. However, precise function of PfKelch13 at its diverse cellular locations remains to be elucidated.

In this study, we used purified recombinant versions of PfKelch13 with the KRP domain along with BTB/POZ and coiled-coil domain (TrK13)⁵¹ and identified several novel biochemical properties of PfKelch13 in vitro. The purified TrK13 protein bound iron salts and heme, an observation that was repeatedly verified using a variety of techniques (Figs. 1–4). Further, when incubated with ART, heme bound to TrK13 was associated with AAS formation. Remarkably, the binding to heme and the production of AAS were lower for the TrK13-R539T and -C580Y ART-R mutants compared to TrK13-WT and the ART-sensitive mutant -A578S proteins (Fig. 8). Finally, the AAS formed through TrK13-WT-bound heme were likely released as soluble molecules.

At this stage, the function of PfKelch13 and heme association has not yet been elucidated in young ring-infected erythrocytes. Interestingly, in our experimental setup, the native wild-type PfKelch13 was recovered from the detergent-solubilized parasite proteome by binding to hemin-agarose beads. This is consistent with the hypothesis that the native PfKelch13 protein binds heme, although we cannot formally exclude the possibility that PfKelch13 could interact with another protein pulled down by hemin-agarose under non-denaturing conditions. We note that in a previous study using mass spectrometry, PfKelch13 was not found among proteins that bind to ferriprotoporphyrin IX immobilized on agarose¹¹⁶. This discrepancy could be due to the different proteome solubilization procedures (detergent versus sonication), which might be critical to solubilize PfKelch13 that is known to be associated with membranes¹¹⁷.

Using the structure of the hexameric assemblies reported earlier for TrK13 and three mutants, in silico docking of oxy-/heme moieties showed multiple interaction sites on the protein albeit with stark differences in the preferred domains in ART-S versus ART-R assemblies. Cluster over cluster docking of ART moiety revealed that ART has a clear preference for the bound heme to TrK13 assembly. Low K_d values of the best poses imply that ART will bind tightly to TrK13 bound heme. In other words, ART will spend significant time bound to heme immobilized on TrK13 protein, enabling the rearrangements required to form AAS, as seen in our experiments. We still think that the differences in domains preferred by heme that may or may not dock ART, cumulatively regulate AAS formation. This explains why there is a differential profile and rates of adduct formation in ART-S versus ART-R mutants explored here in this work and warrants extension of similar work on other PfKelch13 mutants.

Our tryptophan fluorescence quenching (Fig. 3) and MST studies (Fig. 4) independently revealed comparatively weaker heme binding affinities in the TrK13-R539T and TrK13-C580Y mutants as compared to the ART-sensitive proteins. We observed molar variations in K_a or K_d values contingent on the unique detection parameter harnessed by each method. While, MST technique harnessed the near uniform labeling by the extrinsic fluorophore and their displacement (independent of tryptophan residues in TrK13), the tryptophan fluorescence quenching relied heavily on the fluorescence resonance energy transfer (FRET) between the tryptophan (donor) emission and heme (acceptor) absorption spectrum, and hence was directly influenced by the spread of tryptophan residues across TrK13 regions (tryptophan-deficient regions silent to FRET), site of heme docking, and Förster distances (reduced FRET efficiency in distances >25 Å). Considering all these factors, it is understandable why we observed lower units in binding affinity by fluorescence quenching as compared to MST. In spite of these, our experiments, being extensive in their own rights, have provided conclusive evidence about the binding of heme, and more so, variations in the binding affinity between the proteins, follow similar trends in both the techniques, thereby further validating our analyses and conclusion thus derived.

The HPLC analyses used in this study revealed distinct elution profiles of heme and AAS with the expected m/z as reported previously^3,13 (Fig. 7). We then provided the first evidence that AAS could be successfully generated by heme molecule(s) bound to TrK13 protein(s). Unfortunately, the amount of AAS was below the detection limit when using LC-MS, we were unable to quantify the AAS formed by heme-bound TrK13-WT and TrK13-R539T when we isolated the AAS fraction corresponding to fractions with 8-10 minutes elution time. Another interesting result of this study was the absence of detectable AAS formed by heme-BSA and heme-PfHRP2 complexes. Since BSA and PfHRP2 are two well-established heme-binding proteins^105,118,119, our results suggest that the heme-binding property of a protein per se does not translate into the propensity to form AAS (Fig. 8). We speculate that it could be due to different heme-protein binding modes that alter the accessibility and/or chemical properties of the bound heme or iron in the heme molecule. Further experiments are needed to characterize the fate of heme-TrK13 complex and to test whether it is also detected in the parasite.

In the context of the intraerythrocytic malaria parasite, we speculate that there could be multiple sites of PfKelch13-heme encounters. PfKelch13 is located as cytosolic foci to various cellular compartments: at the parasite periphery, proximal or associated with the plasma membrane; close to the FV, the apicoplast and the mitochondria; near parts of the endoplasmic reticulum (ER); and in or close to compartments associated with Rab, ATG18, and PI3P^34,40,117. Several of these compartments are involved in hemoglobin uptake, hemoglobin degradation, and heme synthesis, respectively^34,35,39,117. Thus, we speculate that, by yet unknown pathways, they may provide heme to the parasite cytosol. This is supported by the evidence of a labile heme pool (∼1.6 μM) that is stably maintained throughout the intraerythrocytic development of the parasite⁴⁶ although the dynamic relationship of this pool (and that in the cytostome) with PfKelch13 is yet to be understood.

Our results also indicate that some of the AAS generated by the TrK13-heme complex are likely to be soluble, as evidenced by release experiments performed with the TrK13-WT (Fig. 9). These results indirectly suggest that there may be limited proteotoxicity of these TrK13-directed AAS. This is consistent with two reports that did not identify PfKelch13 as an alkylation target of artesunate or the alkyne-labeled ART analog AP1 coupled to biotin^6,9. Nonetheless, the existence of putative TrK13-AAS adducts in our in vitro system should be carefully investigated by appropriate methods.

To date, the reduced abundance of PfKelch13 mutant proteins at the cellular level is considered to be the sole PfKelch13 feature that distinguishes ART-S and ART-R parasites^28,34,37,109. A recent study comparatively measured the relative abundance of PfKelch13 protein in a series of PfKelch13 mutants versus their resistance index (RSA) within the same isogenic 3D7-derived parasites by direct fluorescence-based microscopy of N-terminus-labeled GFP-PfKelch13 variants¹²⁰. While three PfKelch13 mutants C580Y, R539T and R561H had similar PfKelch13 abundances of 51-52% compared to the WT protein, they differed in their percentage of RSA survival (of about 16%, 38% and 11%, respectively). These results suggest that in addition to protein abundance, other changes in PfKelch13 properties due specific mutations may influence ART-R. The results of our study also provide a plausible explanation for the previously reported ART-R to ART-S revertant phenotype in the NF54^C580YattB parasites (initial RSA: 4.8%) that was slightly more pronounced upon in trans co-expression of 3HA-PfKelch13^WT (RSA: 1%) than of GFP-PfKelch13^C580Y (RSA: 1.6%). Although this small difference could be due to different protein levels of the co-expressed PfKelch13 (because of different tags and/or genotypes), we speculate that it could also be due to different functions of PfKelch13. The in trans co-expression of 3HA-PfKelch13^WT in the NF54^C580YattB parasites (which originally possessed a native low-affinity heme binder PfKelch13^C580Y protein) effectively replenished these parasites with the high-affinity PfKelch13^WT protein, allowing for enhanced AAS generation and greater reversion to the ART-S phenotype.

The data obtained in the parasite is altogether consistent with our in vitro findings that show altered biochemical properties in recombinant TrK13 mutants. We propose here that ART-R mutations in PfKelch13 would confer two distinct properties at the PfKelch13 protein level: a decreased protein abundance^36,40 and a decreased heme-binding activity. Both properties would contribute to reduced ART-heme AAS formation but through distinct heme pools, free heme versus PfKelch13-bound heme. Since the parasite cannot reduce the hemoglobin uptake pathway too much without compromising its own survival, a direct effect of mutant PfKelch13 on reducing the production of AAS would be beneficial, complementing the ART resistance phenotype associated with reduced heme supply. However, this scenario assumes that heme bound to PfKelch13 would act as a better trigger for ART activation compared to cytosolic-free heme, a strong hypothesis that needs to be tested experimentally. Of note, heme is also known to be an important regulatory molecule in some other eukaryotic cells^121,122,123. It would then be interesting to explore whether heme regulates some native function(s) or pathway(s) associated with PfKelch13.

In conclusion, while heme has already been validated as the ‘spark’ for ART activation, the parasite PfKelch13 protein could serve as a double ‘valve’ to regulate the level of ART activation.

Materials and Methods

Reagents and chemicals

All reagents and chemicals used in this study were of analytical grade or higher and were purchased from various sources. Hemin chloride (51280), Ferene S (82940), 3,3′,5,5′ tetramethylbenzidine (TMBZ; T0440), hemin-agarose (H6390), artemisinin (361593) and artesunate (A3731) were purchased from Merck Millipore (Germany). StrepTactin XT beads (2-5010-025) and biotin (0-1016-002) were purchased from IBA Lifesciences (Germany). Dihydroartemisinin (DHA; 19846) and deoxyartemisinin (DOA; 20426) were purchased from Cayman Chemicals (USA).

Cloning, expression, and purification of recombinant TrK13 proteins from E. coli

The E. coli expression constructs pPR-IBA101 (PfKelch13-WT), pPR-IBA101 (PfKelch13-C580Y), pPR-IBA101 (PfKelch13-R539T) and pPR-IBA101 (PfKelch13-A578S) were generated as described in our previous study⁵¹. The recombinant proteins TrK13-WT, TrK13-C580Y, TrK13-R539T and Trk13-A578S with a C-terminus-fused strep tag were purified from isopropyl-1-thio-β-D-galactopyranoside (IPTG) and arabinose (0.2%) induced culture lysate using StrepTactin XT Superflow resin (IBA, Germany). The purified proteins were dialyzed overnight at 4°C against MOPS buffer, pH 7.4 with at least three periodic changes. Protein samples were then concentrated using Amicon 3-kDa concentrators (Millipore), and the purity of each protein was assessed by SDS-PAGE and further confirmed by Western blotting using commercial antibodies against the strep tag (Biobharati Life Sciences) or custom-generated anti-PfKelch13⁵¹.

The pPR-IBA101 (PfKelch13-QNG) plasmid was generated using an overlapping PCR strategy. Briefly, the pPR-IBA101 (PfKelch13-WT) plasmid was used as a template for two independent PCRs (PCR1 and PCR2) with PfKelch13-BsaIF (5’-CAAATGGGAGACCTTATGGAAGGAGAAAAAGTAAAAACAAAAGCAAATAGTATCTCG-3’)/ K13-EEHtoQNG-R (5’-AAAGAATCCACATGAATTTAGAACATTGCCATTTTGTCCTCCTGTAATTATATAAGAATCTGACAATGTGGC-3’) and K13-EEHtoQNG-F (5’-GGAGGACAAAATGGCAATGTTCTAAATTCATGTGGATTCTTTTCACCAGATACAAATGAATGGC-3’)/PfKelch13-BsaIR (5’-CCAAGCGCTGAGACCAGCAGCTATATTTGCTATTAAAACGGAGTGACCAAATCTGGG-3’) primer pairs. These PCR products were used as templates in PCR3 to reconstitute pfkelch13-qng, which was then digested with BsaI and cloned into the appropriate site in pPR-IBA101. The resulting E688Q, E691N and H697G mutations in pPR-IBA101 (PfKelch13-QNG) were confirmed by Sanger sequencing and the recombinant protein TrK13-QNG was purified in a similar manner.

The pPR-IBA101 (PfHRP2) plasmid was prepared by amplifying pfhrp2 (total length 1064 base pairs) from nucleotide positions 228-1061 (excluding the signal sequence coding region) by PCR using P. falciparum genomic DNA as template and the primers NoSS-HRPII-BsaIF (5’- CAAATGGGAGACCTTGCATTTAATAATAACTTGTGTAGCAAAAATGCAAAAGG-3’) and HRPII-NS-BsaIR (5’- CCAAGCGCTGAGACCAGCAGCATGGCGTAGGCAATGTGTGG-3’). The PCR product was then digested with BsaI and cloned into the appropriate site in pPR-IBA101 and confirmed by Sanger sequencing. The recombinant PfHRP2 was purified as described above for TrK13 proteins.

In silico prediction of Fe²⁺ and Fe³⁺ binding sites in full-length PfKelch13 and TrK13-WT with C-terminal strep tag fusion

The binding sites for Fe²⁺ (ferrous) and Fe³⁺ (ferric) were predicted using the MIB server (http://bioinfo.cmu.edu.tw/MIB/)^54,124. MIB is a metal ion binding site prediction and docking server that provides an integrated approach to searching for residues involved in metal ion-binding sites using the fragment transformation method. The PfKelch13 amino acid sequence was compared with Fe²⁺ (ferrous) and Fe³⁺ (ferric) metal-binding templates in the database to locate putative metal-binding residues. Each residue of the query protein is assigned a binding score, which is a combination of sequence and structure conservation measures. If a residue’s binding score is higher than a specified threshold, that residue is predicted to be a metal-binding residue. Based on the local 3D structure alignment between the query protein and the metal ion-binding template, the metal ion in the metal-binding template can be transformed into the structure of the query protein.

UV-visible spectrophotometry-based detection of interactions between recombinant TrK13-WT protein and metal salts

The absorbance and peak profile of purified recombinant TrK13-WT protein in the absence or presence of excess metal salts were measured at a wavelength of 250-700 nm using a JASCO UV600 spectrophotometer with a path length of 1 cm, and changes therein were considered indicative of protein-ligand interactions. Briefly, 500 µM of individual salt solutions of different metals [ferrous ammonium sulfate (FAS), ferric ammonium sulfate (FES), magnesium sulfate (MgSO₄), manganese chloride (MnCl₂), zinc chloride (ZnCl₂), and copper sulfate (CuSO₄)] in 50 mM MOPS buffer, pH 7.4 were scanned at a wavelength of 250-700 nm. As a control, purified recombinant TrK13-WT protein (5 µM in 50 mM MOPS buffer, pH 7.4) was also scanned at these wavelengths. Reaction mixtures containing 5 µM TrK13-WT (or mutants) and 500 µM metal salts were then incubated for 30 min and scanned. Experiments with TrK13 mutants were performed similarly using FAS and FES alone.

Estimation of iron-binding by Ferene S colorimetric assay

The Ferene S [3-(2-pyridyl)-5,6-bis (2-(5-furylsulfonic acid))-1,2,4-triazine, disodium salt] colorimetric assay was used as a specific stain to detect iron molecules bound to the recombinant TrK13-WT protein according to published protocol¹²⁵. All glassware used in the assay was previously soaked in 1% (v/v) HCl to remove iron ions. Purified recombinant TrK13-WT protein (5 μM) was incubated for 60 min with 450 μM FAS in 50 mM MOPS buffer. As a control, a sample with 450 μM FAS alone (without TrK13) in 50 mM MOPS buffer was similarly processed. After incubation, free unbound salt was removed by size-exclusion chromatography using a PD Spintrap™ G-25 desalting column, and the eluate-containing biomolecules larger than 5 kDa was incubated with 30 μl of 10 M HCl for 10 min under rotating conditions. Then 25 μl of 80% trichloroacetic acid (TCA) was added and kept on ice for 10 min. The solution was then centrifuged at 10,000× g for 10 min and the supernatant was transferred to a cuvette. Then, 30 μl of 45% acetic acid and 100 μl of freshly prepared Ferene S reagent (containing 45% w/v sodium acetate, 10 mM ascorbic acid, and 0.75 mM Ferene S reagent) were added and the sample was immediately mixed. The absorbance was then measured at 595 nm. In this assay, iron-containing solutions turn blue.

Preparation of Fe-NTA beads and pull-down assays with recombinant TrK13-WT protein

Fe-NTA agarose beads were prepared from commercially available Ni-NTA beads (Qiagen) according to previously described protocols^78,79. Briefly, 1 ml of a 50% v/v suspension of Ni-NTA was washed with 20 ml of 100 mM EDTA to remove any bound Ni²⁺ ions. The beads were then sequentially washed with 30 ml of double-distilled water, 30 ml of 0.1 M acetic acid, and 30 ml of 100 mM iron salts (FAS, FES, or FeSO4) to load with the appropriate metal ions. After binding, the beads were washed again with 30 mL of water followed by 30 ml of 0.1 M acetic acid to remove any unbound metal ions. Finally, the Fe-NTA beads were equilibrated with buffer W (10 mM Tris/HCl pH 8.0, 1 mM EDTA) and used for pull-down assays. The loading conditions assumed 100% replacement of Ni²⁺ by iron salts. For pull-down assays, 5 µM affinity-purified TrK13-WT or TrK13-R539T protein were co-incubated with Fe-NTA or Ni-NTA beads in buffer W in the absence or presence of excess (0.1 or 10 mM) free FAS or FES salts for 1 h at room temperature. The beads were thoroughly washed with buffer W, and Fe-NTA-bound proteins were resolved by SDS-PAGE followed by Coomassie staining.

To analyze the effect of TrK13-WT oligomerization status on Fe-NTA agarose binding, the affinity-purified protein (in 50 mM MOPS buffer, pH 7.4) was incubated either without or with increasing concentrations of urea (0–6 M) for 30 min, followed by cross-linking with 0.1% (v/v) glutaraldehyde (with spacer length of 5 Å). The samples were then incubated with 30 µl of 50% v/v resuspended Fe-NTA beads at RT for 1 h, washed with the same buffer to remove unbound proteins, resolved by SDS-PAGE, and stained with Coomassie.

Sequence and structural overlap between the KRP domains of TaTFP and PfKelch13

The pairwise alignments between the amino acids of the KRP domains of TaTFP and PfKelch13 was performed using Clustal Omega (https://www.ebi.ac.uk/Tools/msa/clustalo)¹²⁶. The PDB structures of TaTFP (RCSB PDB ID: 5A11) and PfKelch13 (RCSB PDB ID: 4YY8) were retrieved from the RCSB database (https://www.rcsb.org)¹²⁷ and aligned using PyMOL software (https://pymol.org).

In gel staining with TMBZ to detect heme-binding to the TrK13-WT protein

In gel staining of TrK13-WT protein with 3, 3′, 5, 5′ tetramethylbenzidine (TMBZ) was performed according to a published protocol^80,81. Briefly, 10 µM of affinity-purified recombinant TrK13-WT was preincubated in the absence or presence of an equimolar (10 µM) concentration of heme and resolved by PAGE under native conditions. The gel was incubated for 2 h in the dark with freshly prepared TMBZ solution (6.3 mM TMBZ was dissolved in methanol and mixed with 0.25 mM sodium acetate pH 5.0 in a 3:7 ratio). The bands were developed by adding 30 mM H₂O₂ for 30 min. Finally, the gel was fixed by three washes with isopropanol and 0.25 mM sodium acetate pH 5.0 in a 3:7 ratio.

UV-visible spectrophotometry for salt and heme-binding to the recombinant proteins

All spectrophotometric experiments were performed in an 8-well microcuvette using a JASCO UV660 spectrophotometer with 1 cm path length. The total reaction volume was 100 µl in all cases. For salt-binding assays, 5 µM concentrations of affinity-purified recombinant proteins (TrK13-WT or the corresponding -C580Y, -R539T, and -A578S) were incubated with 100-times excess (500 µM) concentrations of FAS, FES, MgSO₄, MnCl₂, ZnCl₂, or CuSO₄. Reaction mixtures containing the protein and/or the respective salts were prepared in MOPS buffer, pH 7.4. After incubation for 1 h, the samples were scanned between 250-700 nm.

For heme-binding assays, a stock solution of reduced heme was prepared by dissolving hemin chloride in 0.1 M NaOH, followed by dilution to the respective working concentrations in MOPS buffer, pH 7.4, in the presence of 2 mM sodium dithionite (to reduce hemin to heme). Affinity-purified recombinant TrK13 proteins (WT, C580Y, R539T and C580Y) or commercially purchased BSA (A7030) or lysozyme (L6876) (Merck, Germany), each at 10 µM concentration, were incubated in the absence or presence of an equimolar concentration of reduced heme for 1 h at room temperature (RT) and immediately desalted using PD Spintrap™ G-25 columns (Cytiva) to remove excess unbound heme. The eluted fractions were then analyzed by spectrophotometric scanning between 250-700 nm wavelengths. Protein samples incubated in the absence of heme were processed similarly and served as appropriate controls. Experiments with the heme analogs PPIX and Zn-PPIX were also performed according to the same protocols.

To calculate the dissociation constant (K_d) for heme binding by the TrK13-WT protein, increasing concentrations (0-20 µM) of the recombinant protein were incubated with a fixed concentration (10 µM) of heme and processed similarly. The values for the absorbance maxima at 380 nm (y-axis) were plotted against the respective protein concentrations (x-axis) as previously described¹²⁸. In all cases, the composite data were plotted using Excel software.

Binding of recombinant TrK13 proteins to hemin-agarose beads and competitive inhibition by excess heme and analogs

The affinity-purified recombinant TrK13 proteins (10 µM) were incubated with 30 µl suspensions of 50% v/v hemin-agarose beads for 1 h at RT in MOPS buffer, pH 7.4, under rotating conditions, as described elsewhere^129,130. Excess unbound proteins were removed by successive five washes with ten volumes of MOPS buffer, pH 7.4, and bound fractions were solubilized in Laemmli’s sample buffer, resolved by SDS-PAGE, and visualized by Coomassie staining. For competitive inhibition, assays were performed under similar conditions in the absence and presence of excess heme or heme analogs. Briefly, 10 µM TrK13-WT protein was preincubated with 10-timesmolar excess (100 µM) concentrations of reduced heme, PPIX, or Zn-PPIX for 1 h at RT, followed by the addition of hemin-agarose beads. Percentage inhibition was calculated based on the band intensity of hemin-agarose bound TrK13-WT protein without any heme or analogs and considered as 100%. The pixel intensity of each band was calculated and plotted using ImageJ software.

Tryptophan fluorescence quenching

Tryptophan fluorescence quenching experiments were performed using an FLS920 fluorescence spectrophotometer (Edinburgh Instruments, UK). Briefly, 10 µM of TrK13-WT protein or corresponding mutants were incubated with stepwise increasing (0–5 µM) concentrations of heme as a quencher. The excitation wavelength was 295 nm and the emission wavelength was 300–500 nm. The fluorescence quenching constant (K_SV) was calculated from the Stern-Volmer equation as described in Edgar et al. (2016)⁸¹; F₀/F = 1 + k_qζ₀[Q]) 1 + K_SV[Q](1), where F₀ and F are the fluorescence intensities before and after quencher addition, respectively, k_q is the bimolecular quenching constant, ζ₀ is the lifetime of the fluorophore in the absence of quencher, and [Q] is the concentration of quencher. The dissociation constant (K_d) was calculated from the change in fluorescence intensity (ΔF) at 334 nm and the concentration of heme using SigmaPlot v12.3 and the following equation: ΔF = (ΔF_max × [heme])/ (K_d + [heme]).

Circular dichroism (CD) of TrK13-WT protein and BSA in the absence or presence of heme

CD data of recombinant TrK13-WT protein and BSA were collected in the absence or presence of heme at concentrations of 0-30 µM using an AVIV circular dichroism spectrophotometer (model 420SF, Lakewood, NJ, USA). Spectra were recorded in a 0.2 cm path-length quartz cuvette with a bandwidth of 1.0 nm and a resolution of 200-250 nm. A total of five scans were recorded and then averaged to obtain a better signal-to-noise ratio. Experiments were performed in a selection buffer (150 mM NaCl and 20 mM Tris. HCl pH 8) in a concentration range of 2-5 µM protein at a scan speed of 20 nm/min. The data were processed to estimate the secondary structure content of the protein using the K2D2 server⁸⁸.

Microscale Thermophoresis (MST)

The binding affinity of each recombinant TrK13 protein for heme was evaluated by MST analysis using the Monolith NT.115 instrument (NanoTemper Technologies). The recombinant TrK13 proteins (50 nM) were diluted in phosphate-buffered saline (PBS; pH 7.4) supplemented with 0.05% Tween 20 (PBS-T) to prevent sample aggregation, followed by labeling with cysteine-reactive dye (30 mM; Monolith Protein Labelling Kit Red-Maleimide 2nd Generation, NanoTemper) and incubated for 30 min at room temperature (RT) in the dark. The labeled PfKelch13 protein and buffer were added to an equilibrated column and the elution fractions were collected. Elutes with less than 1,000 fluorescence counts were used for interaction analysis. Labeled TrK13-WT was titrated with hemin chloride (0.01 nM to 1 mM). Samples were premixed and incubated for 10 min at RT in the dark before loading onto standard treated capillaries (K002 Monolith NT.115). For interaction analysis, the change in thermophoresis was expressed as the fluorescence change in the MST signal, defined as F_hot/F_cold (F_hot as the hot region after IR laser heating and F_cold as the cold region at 0 s). Titration of the non-fluorescent ligand results in a gradual change in thermophoresis, which was plotted as DF_norm to obtain a dose-response (binding) curve that can be fitted to derive binding constants. Data evaluation was performed using Monolith software (Nano Temper). Measurements were performed using standard capillaries from 25°C at 40% MST power. The thermographs showed no aggregation or molecular adsorption to the capillaries (data not shown). The smooth binding curve relating heme concentration to normalized fluorescence was fitted using a K_d model. Each point represents the average of three sets of individual measurements.

Docking of Heme and Oxy-heme on models of hexameric TrK13 protein and its mutants

Structures of TrK13 WT, and mutants were downloaded from the SASBDB database (https://www.sasbdb.org)⁵¹. The resolution of SAXS is the overall resolution based on the shortest input vector and is different from best resolution of crystal structure, which could be local. It is to be noted that SAXS data-driven assembly structures as submitted to the database were energy minimized in implicit water using the United atom OPLS forcefield. Mostly, hydrogen bonds (H-bonds) and atomic interactions were considered within distances up to 3-4 Å or 0.3-0.4 nm, and these all were within the input resolution of the SAXS data-based dummy atom models. However, for this study, the coordinates of the residue detail models generated using dummy atom model or SAXS data as reference were considered for docking experiments. In this study, using YASARA Program v6^96,97,98 and YAMBER2 forcefield, the protein assemblies with added hydrogen atoms were energy minimized to a cut-off gradient of 0.01. The coordinates of heme and oxy-heme were taken from the RCSB ligand database in SDF format. They were parameterized and energy minimized using same forcefield. Protein assemblies were considered as static receptors and flexible versions of ligands were used for docking runs. Using a 16-core system supported by four GPUs, 50 runs of docking using VINA protocol in YASARA were performed. The results were compiled as clusters were poses of ligands where positions of ligand within 0.5 Å were considered to be part of same cluster. Additional results were theoretical binding energies and the location of clusters in form of residues and chains. Visual analysis of the resultant PDB files were done by PyMOL program.

Detection of AAS by High-Performance Liquid Chromatography (HPLC)

HPLC analyses were performed using a GSK Gel Protein 5 µm C4-300 4.6 × 150 mm semi-preparative column and a flow rate of 1 mL/min by modification of previously described methods¹³¹. The reaction mixture containing 300 µM heme in the absence or presence of 360 µM ART or ART derivatives (DHA, AS) or the metabolically inactive analog DOA (1:1.2 molar ratio) was incubated for 30 min at 37°C under reducing conditions (2 mM sodium dithionate). The reaction products were separated by HPLC using the following order of eluents: (A) 0.05% v/v trifluoroacetic acid in water, (B) 0.05% v/v trifluoroacetic acid in acetonitrile. The elution gradient was as follows: from A/B = 63/37 to A/B = 52/48 in 15 min, then A/B = 52/48 for 15 min. The flow rate was 1 mL/min and the UV detection was set at 410 nm for heme detection (Soret peak).

For assays containing the purified TrK13 proteins, 10 µM TrK13 were incubated with 60 µM reduced heme for 30 min at 37°C, followed by the addition of 72 µM ART. The reaction mixtures were incubated again for 30 min, loaded onto the HPLC column, and eluted as described above. In some cases, the heme-TrK13 mixtures were desalted using G-25 columns after the 30 min incubation to remove unbound heme and before the addition of ART.

Identification of AAS by Liquid Chromatography-Mass Spectrometry (LC-MS)

Reaction mixtures containing heme incubated in the absence or presence of ART were extracted with 1.5 ml CHCl₃: H₂O (2:1) as described elsewhere¹³. The organic layer was separated, dried under vacuum, and redissolved in MS-grade methanol and 0.1% formic acid for LC-MS analysis. Data were collected in positive ion mode using an Agilent Q-TOF G6550A mass spectrometer equipped with a dual ASI ion source. The electrospray ionization source was operated at a nozzle voltage of 0.5 kV with N2 as the nebulizer gas. The m/z acquisition range was 100-1000 and the scan rate was 1 spectra/sec. A similar protocol was followed with slight modifications for the LC-MS analysis of samples containing TrK13. Briefly, 10 µM TrK13-WT was pre-incubated with 60 µM heme and PD Spintrap™ G-25 columns were desalted to remove free unbound heme, followed by the addition of 72 µM ART.

Detection of AAS by UV-visible spectrophotometry

Ten micromolar purified TrK13 or commercial BSA (A7030, Merck, Germany) was incubated in the absence or presence of an equimolar concentration of reduced heme or the heme analog Zn-PPIX for 1 h at room temperature and immediately desalted using PD Spintrap™ G-25 columns (Cytiva) to remove excess unbound heme according to the manufacturer’s instructions. The G-25 column eluate, containing biomolecules larger than 5 kDa (i.e., TrK13 and heme-TrK13 complex), was then incubated with 200 µM ART or DOA for 30 min and analyzed by spectrophotometric scanning between 250-700 nm wavelengths to detect the presence of a Soret peak from heme or Zn-PPIX. Control samples without heme and/or without ART were included for comparative analyses in specific experiments.

To determine whether the AAS were released from the TrK13 protein (or from other control proteins), the same protocol as above was followed except that the final reaction mixture was desalted using PD Spintrap™ G-25 columns and the eluate containing low molecular weight compounds was analyzed by spectrophotometric scanning between 250-700 nm wavelengths.

Parasite culture and pulldown assays

P. falciparum 3D7 parasites were grown in human erythrocytes supplemented with 2 mM L-glutamine, 50 mg/l hypoxanthine, 25 mM HEPES, 0.21% NaHCO₃, 0.1 mg/l penicillin-streptomycin, and 0.5% w/v Albumax II (Invitrogen) under 5% CO₂ and 5% O₂, at 37°C. Approximately 2 × 10⁹ 3D7 parasites were isolated by saponin treatment from asynchronized culture (4 flasks of 25 mL culture at 4% parasitemia) and immediately frozen at -80°C as dry pellets (total parasite pellet volume was approximately 100 µl). Total parasite proteins were solubilized with RIPA buffer as follows: 1 volume of dry parasite pellet was resuspended in 10 volumes of ice-cold RIPA buffer complemented with 1X protease inhibitor (cOmplete, Roche). After one hour of incubation on ice with regular mixing, the sample was passed 4 times through a 21 G needle. After centrifugation (15 min, 21,100× g, 4°C) the supernatant containing soluble proteins was collected. 48% of this soluble protein extract was incubated with 35 µl of hemin-agarose or glutathione (GSH)-agarose bead pellets prewashed three times with 1X PBS and once with RIPA buffer. The remaining 4% of protein extract was kept frozen at -80°C. The protein extract/beads mixture was incubated overnight at +4°C on a rotating wheel and then centrifuged (5 min, 500× g, 4°C). The supernatant was stored as flow through and the beads were washed 6 times with 1 ml RIPA buffer complemented with 0.25 × protease inhibitor (cOmplete, Roche). Bound proteins were eluted from agarose beads with 1 bead volume of Laemmli buffer heated at 95°C for 5 min. After a brief centrifugation (30 sec, 2000 × g, room temperature), proteins were electrophoresed on precast 4–20% Tris-Glycine gels (Bio-Rad) and transferred onto nitrocellulose membranes which were blotted against PfKelch13 antibody generated in rabbits against the peptide CEHRKRFDEERLRFL corresponding to amino acids 297-310 of the PfKelch13 sequence⁵¹, PfHRP2¹³² antibodies and anti-rabbit (1:5000, StarBright Blue 520, Bio-Rad) secondary antibodies were used and western blots were imaged on a ChemiDoc system (Bio-Rad).

Reporting summary

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