Tafenoquine Disrupts Intraerythrocytic Stages of Malaria Parasites by Inhibiting Nonclassical Pathways of β-Hematin Nucleation and Growth

Abstract

The rise in malaria resistance to multiple classes of drugs has led to an alarming increase in the number of cases and fatalities worldwide. Tafenoquine (TQ), an antimalarial approved by the Food and Drug Administration in 2018, inhibits multiple stages of malaria parasites’ lifecycle, yet its precise mechanisms of action remain poorly understood. Here we explore how TQ interacts with the main pathway of heme detoxification in malaria parasites residing in erythrocytes: the formation of hemozoin crystals. We employ the synthetic analogue of hemozoin, β-hematin, and show that TQ blocks a fraction of the crystal growth sites and kinks and mildly inhibits the classical growth of β-hematin that occurs through the generation and spreading of layers. We demonstrate that TQ potently suppresses a nonclassical growth mode, which activates at elevated hematin concentrations as well as the nucleation of β-hematin crystals, by impeding the formation of mesoscopic hematin-rich clusters—the shared precursors of both crystal nucleation and nonclassical growth. Measurements of TQ parasite-killing potency reveal that, consistent with the mechanistic insights, this drug exerts a stage-specific inhibitory effect, with greater ring-stage suppression when hemozoin crystals nucleate than trophozoite suppression when the crystals grow. This mechanism highlights TQ’s ability to disrupt parasite survival by inhibiting hemozoin synthesis in addition to the disruption of mitochondrial function.

Graphical Abstract

INTRODUCTION

Malaria is a debilitating disease caused by parasitic infection that remains a major public health challenge, contributing to significant morbidity and mortality across numerous tropical and subtropical regions of the world.^1,2 In 2022, malaria was responsible for an estimated 608,000 deaths across 85 countries.³ There is currently no highly effective malaria vaccine. Plasmodium sporozoites, the parasite species that invade the human bloodstream, have evolved to evade the human immune system, preventing a strong immune response and the production of protective antibodies.^4,5 Accordingly, individuals who recover from malaria do not gain immunity and remain vulnerable to future infections.^4,5 The most advanced available malaria vaccine requires a four-dose regimen and provides only limited protection, with an efficacy of approximately 30–45%.^4,5 Over the past two decades, artemisinin-based combination therapies (ACTs) have contributed to reducing the global burden of malaria, significantly decreasing the total number of cases and fatalities.^6–9 Since 2020, however, there has been a resurgence in malaria cases, primarily attributed to the spread of artemisinin-resistant strains of the parasite, particularly in East Africa.^10–16 More in-depth analyses of the mechanisms of antimalarial drugs and the potential of ACTs are becoming increasingly essential.¹³

The most recent antimalarial approved by the US Food and Drug Administration is tafenoquine (TQ, Figure 1a), which belongs to the 8-aminoquinoline class and targets malaria at both the liver and gametocyte stages.^17,18 Unlike other quinoline antimalarial drugs, TQ (similar to another drug, primaquine) has the unique ability to eliminate Plasmodium vivax hypnozoites in the liver, thereby preventing relapse and progression to the blood stage.^17–19 Furthermore, parent TQ also shows good synergy with other antimalarial drugs in culture assays, such as artemisinin and methylene blue.^17,20 While TQ has been shown to interact with several stages of the malaria parasites’ lifecycles, the exact mechanisms of TQ blood-stage activity remain elusive.¹⁸ Studies involving protozoan parasites, including P. falciparum, pointed out that TQ interfered with their mitochondrial function, leading to apoptotic-like death.²¹ Another study suggested that TQ could operate against the blood-stage forms of Plasmodium falciparum and Plasmodium vivax.²² To date, there have been no studies that examine the mechanism of TQ effects on the physiological processes of blood-stage parasites.

Figure 1.

Suppression of blood-stage parasites by TQ. (a, b) The molecular structures of tafenoquine, in a, and hematin, in b. (c–j) Fractions of P. falciparum parasites killed by 6 h pulses of TQ administered, in c, e, i, immediately after tight synchronization of their lifecycles (early rings, purple squares) and 5 h after synchronization (late rings, blue circles) and d, f, j. 24 h after synchronization (early trophs, purple squares) and 30 h after synchronization (late trophs, blue circles). Strains NF54, in c and d; CAMWT, in e and f; and C580Y, in i and j. Horizontal dashed lines mark 50% inhibition. Error bars denote standard deviation of biologic triplicates with technical triplicates runs. Note the wider ranges of TQ applied to the NF54 parasites.

In the red blood cells, malaria parasites metabolize hemoglobin, releasing its prosthetic group, heme, which quickly converts to hematin (Figure 1b).^23–28 Both heme and hematin are toxic to the parasite; to survive, the parasite neutralizes hematin by inducing it to form benign crystals known as hemozoin.^29–33 Many antimalarial drugs function by inhibiting the formation of hemozoin, allowing hematin to remain soluble, accumulate to levels of high toxicity, and ultimately kill the parasites.^32,34–39

Here we explore whether and how the parent TQ kills blood-stage P. falciparum parasites. We examine the contribution of parent TQ to blood-stage activity independent of its metabolite action on parasite mitochondria. We differentiate between TQ effects on ring-stage parasites, in which hemozoin crystals nucleate, and the trophozoite stage, when they grow by incorporating hematin. For mechanistic insights into the effects of TQ on hematin crystallization, we examine the formation of β-hematin, a synthetic crystal structurally identical to hemozoin.³⁰ Hemozoin and β-hematin crystals belong to the triclinic $\mathrm{P}\overline{1}$ symmetry group and are shaped as bars elongated along the c, or [001], direction.^40–42 We monitored the kinetics of growth and nucleation of β-hematin in the presence of TQ and deduced the relevant mechanisms.

The digestive vacuole of Plasmodium parasites primarily contains an aqueous medium, within which hemoglobin-degrading proteases function.⁴³ Investigations have revealed that hemozoin crystals develop while suspended within nanospheres composed predominantly of five neutral lipid species, and these lipid assemblies play a critical role in both crystal growth and sequestration of hematin.⁴⁴ Because these lipid nanospheres are in direct contact with the vacuole’s aqueous interior, they unavoidably incorporate small amounts of dissolved water.^45,46 Subsequent research explored whether aqueous-phase crystallization could effectively mediate hematin detoxification and concluded that this route is inefficient.^46,47 The central finding is that hematin’s solubility at the digestive vacuole’s pH is extremely low—approximately 2 nM—such that, given the rate at which hematin is liberated during hemoglobin digestion, the aqueous solution becomes excessively supersaturated (on the order of 10⁶-fold). At these levels of supersaturation, β-hematin crystals accumulate extensive structural imperfections and associated elastic stress, which ultimately halts further crystal growth. Motivated by approaches common in drug development, we emulate the lipid-rich environment of the vacuole using n-octanol saturated with a pH 4.8 citric acid buffer—termed CBSO in prior work.^45–47 Understanding how hematin molecules integrate into crystal lattices from such a medium offers a benchmark for evaluating the efficacy of crystallization inhibitors. Notably, translating these findings into clinical scenarios may present challenges. The complex environment within a human host differs from controlled experimental conditions, which might affect the drug’s efficacy.

RESULTS AND DISCUSSION

TQ Kills Ring- and Trophozoite-Stage Parasites.

We tested whether TQ inhibits blood-stage P. falciparum parasites, where killing by other antimalarials has been attributed to suppression of hematin crystallization and accumulation of toxic soluble hematin.^32,34–39 To differentiate between potential effects of parent TQ on hemozoin crystal nucleation and growth, we synchronized the lifecycles of a parasite population and applied the drug to parasites of four ages: immediately after synchronization, when the parasites are at their early ring stage; five hours later, at the late ring stage; and after 24 and 30 h, when the parasites enter their trophozoite or schizont stage. During the early ring stage, hemoglobin metabolism begins to produce heme, which oxidizes to hematin and spawns nucleation of hemozoin crystals. A drug active at the ring stage could operate by inhibiting crystal nucleation. At the trophozoite stage, hemozoin is abundantly present, and the crystals grow by incorporating hematin; thus, a drug active at this stage likely inhibits the growth of hemozoin crystals.

We investigated stage-specific pulse parent TQ activity on P. falciparum chloroquine-sensitive NF54, chloroquine-resistant but ring-stage artemisinin-sensitive CAMWT, and chloroquine-resistant and ring-stage artemisinin-resistant CAM580Y. We exposed the parasites to 6 h TQ pulses, after which we washed and incubated the parasites in drug-free media.⁴⁸ We measured the fraction of killed parasites using flow cytometry SYBR Green^49–51 and compared the results to those achieved with continuous drug application (Figure S1).

We find that TQ inhibited blood-stage parasites (Figures 1c–j), with greater activity on ring-stage parasites. The drug concentrations that inhibit 50% (IC₅₀) of early ring-stage parasite growth in NF54, CAMWT, and CAMC580Y after a 6 h exposure are 1.6 μM, 0.7 μM, and 0.4 μM (Figure 1c,e,i), rising to 10.7, 2.6, and 0.9 μM at early trophozoite stages (Figure 1d,f,j). These observations suggest that TQ suppresses both the nucleation and, to a lesser extent, the growth of hemozoin crystals. We next explored the parent TQ action on ex vivo heme crystallization.

Classical and Nonclassical Mechanisms of β-Hematin Crystal Growth.

Observations by time-resolved in situ atomic force microscopy (AFM) revealed that at low and moderate hematin concentrations, β-hematin crystals grow by a classical mechanism, whereby new crystal layers are generated by dislocations (Figure 2a and Movie S1) or two-dimensional nucleation (Figure 2b and Movie S2) and advance across the surface.^48,52 The layer thickness h is about 1.1 ± 0.1 nm (Figure 2c, inset), close to the lattice parameter in the [100] direction, a = 1.22 nm⁴⁰. The layers spread owing to the incorporation of hematin molecules from the solution at their edges, called steps. Thus, the velocity of the steps v is a sensitive indicator of the molecular mechanisms of the bimolecular reaction between the solute and the growth sites at the steps, called kinks.⁵³

Figure 2.

β-hematin crystal growth. (a, b, d–g). In situ atomic force microscopy (AFM) images of (100) β-hematin crystal faces at concentrations indicated on the panels. Crystallographic directions are indicated in upper right corners of the images in panels a and b. (a) New layers generated by screw dislocation twist into spirals. (b) New layers are generated by two-dimensional (2D) nucleation. (c) Velocities of steps growing along [001] (black) and [010] (red) directions at different hematin concentrations C_H. Error bars depict the standard deviations from the averages of about 20 measurements. Straight lines display the best-fit linear v(C_H) correlations. Hematin solubility C_e = 0.08 mM. Inset: Height profiles across a 2D island and a particle (highlighted with blue and green, respectively, arrows and contours in panel b). (d–g). The evolution of particles, indicated with green arrows in panel d, landing on the crystal surface. Images were captured at the indicated times. (h) Evolution of a particle height profile along the green lines in panels d–g.

The crystal layers on the (100) face are shaped as ellipses elongated along the [001] direction (Figure 2a,b), indicating that step growth in this direction is faster than in the [010] direction. The ratio between the ellipse axes along the fast and slow directions is maintained at varying hematin concentrations. We measure the step velocity v by monitoring the steady advancement of steps in time.^46,52,54 We observe that v increases linearly with the hematin concentration C_H (Figure 2c). Previous studies have revealed that hematin molecules reach the steps after adsorbing on the terraces between them and diffusing along the surface toward the kinks.^46,55 Constraining the solute supply to two dimensions enforces competition for supply between closely spaced steps.^56–58 With β-hematin, the slowdown of step growth is observed with steps separated by less than 70 nm (Figure S2). Thus, we expect this linear v(C_H) correlation to apply for steps farther apart than this threshold. At all probed C_H, v is about 4-fold faster in the [001] direction than in the [010] direction (Figure 2c), consistent with the concentration-independent elliptical layer shape. The linear v(C_H) correlations cross the C_H axis at the solubility C_e, and when C_H falls below C_e, the steps retreat and the crystals dissolve.⁴⁶

Surprisingly, at C_H = 0.25 mM, after about 30 min of growth under steady conditions, we observed not only layers generated by 2D nucleation but also particles that landed on the surface (Figure 2b and Movie S3). Their height profiles revealed that they comprised several layers (Figure 2c, inset), a significant difference from the 2D islands. Their rounded shape suggests that they are amorphous particles of hematin, the only solution component that can form solids and is available in significant amounts. We monitored the evolution of the particles to test whether they partake in the growth of the crystal. Time-resolved imaging and height profile analysis demonstrated that the particles indeed accelerate crystal growth (Figures 2d–h). Over time, the particles transformed into stacks of crystal layers that spread along the surface. The lower layers were exposed to a more abundant solute supply from the crystal surface, which has lower step density, and grew faster (Figure 2f,g).⁵⁹ Steps generated by distinct particles merged (Figure 2g). The seamless integration between the particles and the underlying crystal lattice suggests that the particles are not solid amorphous but, rather, liquid. At longer imaging times, the high step density of the initial particle-generated layer stack dissipated, and steps became indistinguishable from those generated by 2D nucleation (Figure 2g). Layer generation by amorphous particles is a nonclassical mode of crystal growth.^60–64 Our results represent the first observation of nonclassical growth of β-hematin crystals that may have been missed in previous studies limited to short observation times at higher hematin concentrations.

The hematin concentration emerges as the critical factor that determines the growth mechanism of β-hematin. At low C_H, crystal growth is predominantly classical with layers generated by dislocations or 2D nucleation. By contrast, at higher C_H, landed particles contribute significantly to layer generation. The increased role of particles at high C_H suggests that their generation is a sensitive function of C_H.

The Impact of TQ on β-Hematin Step Growth and Layer Generation.

We added TQ to hematin solutions of concentrations, close to the low limit of the ranges of concentration of free hematin in the digestive vacuoles of CamWT parasites, between about 0.2 and 4 mM , and C580Y parasites, for which it is between about 0.5 and 1.2 mM.⁶⁵ This addition does not modify the classical modes of layer generation on the (100) β-hematin crystal surface. Layers are generated by dislocations and, in the absence of dislocations, by 2D nucleation (Figure 3a). The elliptical island shape elongated along the [001] direction is also preserved, suggesting that TQ interacts similarly with the steps growing in distinct directions. The step velocities in the [001] direction exhibit a downward trend as C_TQ increases. At concentrations above $C_{\mathrm{T}\mathrm{Q}}=4\mu \mathrm{M},v$ stabilizes at approximately 70% of the value observed in the absence of TQ, representing a 30% inhibition (Figure 3b). This $v\left(C_{\mathrm{T}\mathrm{Q}}\right)$ trend and the inability of TQ to fully stall step growth align with the characteristics of a kink blocker, which associates with a fraction of the kinks, obstructs solute access to them, and thus partially inhibits step growth (Figure 3b, inset). Complete blockage of all kinks, however, is not achieved even at high $C_{\mathrm{T}\mathrm{Q}}$ owing to the dynamic generation of kinks at characteristic times comparable to those of TQ adsorption.⁵⁴ The kink-blocking mechanism would enforce a $v\left(C_{TQ}\right)$ correlation $v=v_0-\left(v_0-v_{\mathrm{\infty }}\right)\frac{K_{\mathrm{L}}{C}_{\mathrm{T}\mathrm{Q}}}{1+K_{\mathrm{L}}{C}_{\mathrm{T}\mathrm{Q}}}$, where $v_0$ and $v_{\mathrm{\infty }}$ are the step velocities in a pure solution and at the highest inhibition, respectively, and $K_{\mathrm{L}}$ is the Langmuir constant for adsorption at kinks.⁵⁴ This kinetic law linearizes in reciprocal coordinates ${\left(v_0-v\right)}^{-1}={\left(v_0-v_{\mathrm{\infty }}\right)}^{-1}+{\left[\left(v_0-v_{\mathrm{\infty }}\right)K_{\mathrm{L}}{C}_{\mathrm{T}\mathrm{Q}}\right]}^{-1}$.^54,66–69 The $v\left(C_{\mathrm{T}\mathrm{Q}}\right)$ data comply with this signature prediction of the kinkblocking mechanism (Figure 3c). The observed 30% inhibition of step velocity classifies TQ as a weaker kink blocker than other quinoline-based drugs and heme-artemisinin adducts, which achieve inhibition levels of up to 60% and 70%, respectively.^48,54,66,70

Figure 3.

Effect of tafenoquine (TQ) on β-hematin growth at low C_H. (a) AFM images of hematin crystal surfaces at C_H = 0.25 mM in the absence of TQ and in the presence of 10 and 20 μM TQ, as indicated in the images. (b) The velocity of steps v growing in the [001] direction as a function of TQ concentration relative to v₀ in the absence of TQ. Inset: v(C_TQ) data in reciprocal coordinates. Error bars display standard deviations computed from data as in panel b. (c) Schematic of TQ molecules adsorbing at kinks and blocking the access of hematin molecules. (d) Rate of 2D nucleation of new layer relative to that in the absence of TQ as a function of C_TQ. Error bars display standard deviations from the averages of about 10 measurements. (e) Interstep distance l during dislocation growth relative to that in the absence of TQ as a function of C_TQ. Error bars display standard deviations from the averages of about 20 measurements. (f) The normal growth rate V_n of crystals growing by dislocation or 2D nucleation generated layers as a function of C_TQ. Error bars display standard deviations evaluated by propagation of the standard deviations of v, J_2D, and l in panels b, d, and e, respectively. In panels b, d, e, and f, red symbols denote the controls, i.e., measurements of the respective kinetic variables in the absence of TQ, and black symbols, at C_TQ indicated on the horizontal axis.

An unexpected consequence of TQ addition was the increase of the number of 2D islands as $C_{\mathrm{T}\mathrm{Q}}$ increased (Figure 3a). We measured the rate of 2D nucleation of new layers $J_{2D}$ by counting the number of islands that form and grow per unit surface area and unit time.^46,54,66 The value of $J_{2\mathrm{D}}$ increased 3-fold in the presence of 20 μM TQ (Figure 3d), indicating that TQ promotes the formation of new layers via 2D nucleation. This behavior has not been observed with previously studied quinoline antimalarials or hematin adducts of artemisinin-class drugs;^48,54,66,70 however, it has been seen during ammonium urate growth in the presence of molecular impostors.⁷¹ We attribute the enhancement of J_2D by TQ to a reduction in the surface free energy of the layer edge γ, caused by TQ adsorption at kinks along the island periphery, similar to the adsorption at kinks at the steps that inhibits step growth (Figure 3c)⁵⁴. Measurements of the two-dimensional critical radii $R_{\mathrm{c}}$ of newly formed 2D nuclei at several hematin concentrations—γ is the scaling parameter in the $R_{\mathrm{c}}\left(C_{\mathrm{H}}\right)$ correlation^46,54,72—in the absence and presence of TQ (Figure S3) revealed that TQ affects γ by less than 5%. Since $J_{2\mathrm{D}}$ is a sensitive function of $\gamma ,J_{2\mathrm{D}}=J_0\mathrm{e}\mathrm{x}\mathrm{p}\left[-\frac{\pi \gamma R_{\mathrm{c}}h}{k_{\mathrm{B}}T}\right]$ (where $J_0$ is a kinetic factor, $k_{\mathrm{B}}$ is the Boltzmann constant, and T is the temperature),⁵⁴ the observed 3-fold acceleration of $J_{2\mathrm{D}}$ by TQ requires a 2% depression of γ, within the limits of experimental measurement.

On the other hand, the separations l between steps generated by dislocations are not affected by TQ (Figure 3e). In contrast to $J_{2\mathrm{D}},l$ is a only a linear function of γ,^61,73 and the weak depression of γ by TQ (Figure S3) insignificantly affects l. With this constant l, the normal growth rate $V_n=hv{l}^{-1}$ for dislocation growth follows the trend of the step velocity v, about 30% inhibition by TQ (Figure 3f). For growth by 2D nucleation of new layers, $V_{\mathrm{n}}=h{\left(v^2{J}_{2\mathrm{D}}\right)}^{1/3}.55$ The increasing $J_{2\mathrm{D}}\left(C_{\mathrm{T}\mathrm{Q}}\right)$ and the decreasing $v\left(C_{\mathrm{T}\mathrm{Q}}\right)$ dependences offset each other, and $V_{\mathrm{n}}$ for 2D nucleation growth is independent of $C_{\mathrm{T}\mathrm{Q}}$ (Figure 3f). The combined $v\left(C_{\mathrm{T}\mathrm{Q}}\right)$, $J_{2\mathrm{D}}\left(C_{\mathrm{T}\mathrm{Q}}\right),l\left(C_{\mathrm{T}\mathrm{Q}}\right)$, and $V_{\mathrm{n}}\left(C_{\mathrm{T}\mathrm{Q}}\right)$ correlations reveal that TQ is a poor inhibitor of the growth of the (100) β-hematin crystal faces at low hematin concentration.

The weak suppression by TQ of V_n (Figure 3f) implies that if the mechanism of action of this drug were constrained to the inhibition of classical crystal growth of β-hematin, parent TQ would be a poor inhibitor of trophozoite-stage parasites by the hemozoin mechanism.

TQ Moderately Inhibits Nucleation but Not Growth of β-Hematin Crystals.

For further insights, we tested parent TQ inhibition on the nucleation and growth (extension) of β-hematin crystals in bulk experiments, in which the heterogeneous environment of the parasite digestive vacuole is partially mimicked by adding monooleylglycerol (MOG) to hematin solutions at a concentration of 50 μM in acetate at pH 4.8;⁷⁴ MOG forms “oil” droplets and vesicles in aqueous media.⁷⁵ In the nucleation assay (Figure 4a), the solution was unseeded, whereas in the extension assay (Figure 4b), preformed crystals were added to initiate crystallization. The results (Figure 4) reveal that TQ inhibits nucleation with IC₅₀ = 45 μM; a positive control with pyronaridine (PY) yielded IC₅₀ = 4 μM (Figure 4a). TQ did not measurably inhibit growth extension at concentrations of up to 60 μM (Figure 4b). Nucleation inhibition was measurable but 10-fold weaker than pyronaridine and twice as weak as chloroquine, whose IC₅₀ for nucleation inhibition is 20 μM.⁷⁷

Figure 4.

TQ inhibition of hematin sequestration into β-hematin crystals. (a) Inhibition of β-hematin nucleation measured in unseeded solutions (nucleation assay). (b) Inhibition of β-hematin crystal growth in solutions, to which preformed crystals were added (extension assay). Both assays were carried out in technical triplicate with biologic replicates of n = 4.

TQ Suppresses β-Hematin Crystal Nucleation.

To understand the mechanism employed by TQ to kill ring-stage parasites (Figure 1c), we tested whether TQ inhibits the bulk nucleation of β-hematin crystals, which predominantly occurs in ring-stage parasites. We employed dynamic light scattering (DLS) (Figure 5a) to detect and monitor the evolution of β-hematin crystals in hematin solutions. The intensity of light scattered by heterogeneities fluctuates as they diffuse in and out of regions of high concentration. Thus, the characteristic time of diffusion of the heterogeneities can be measured as the decay time τ₂ of the intensity autocorrelation function g₂(τ) of scattered light (defined in Figure 5d, inset). We analyzed solutions with C_H = 0.28 and 0.5 mM, both higher than the β-hematin crystal solubility (Figure 2c). We filter the solutions to remove pre-existing particles, load them in the DLS cuvette, and immediately commence data collection. Autocorrelation functions are collected as the solution ages. Immediately after solution preparation, the correlation function g₂ ≈ 0 for all delay times τ (Figure 5b,c) indicating that the solution is homogeneous. The increasing g₂ shoulder at short delay times τ for aged solutions indicates the formation of a condensate, likely of hematin, the only abundant nonsolvent component. Since the solutions are supersaturated with respect to crystals, we propose that the condensates are freshly nucleated β-hematin crystals. Higher g₂ shoulders indicate a greater number of crystals, whereas faster shoulder growth corresponds to accelerated nucleation.

Figure 5.

Tafenoquine effects on β-hematin crystal nucleation. (a). Schematic of dynamic light scattering. (b, c) Representative autocorrelation functions g₂(τ) of light scattered from a hematin solution alone (left) and in the presence of 50 μM TQ (right) at two different supersaturations: (b) C_H = 0.28 mM. (c) C_H = 0.5 mM. (d) Evolutions of the amplitudes A₂ of the autocorrelation functions at C_H = 0.28 mM and 0.5 mM, as indicated in the plots, in the absence and presence of 50 μM TQ. Inset: definition of the amplitude A₂ of the autocorrelation function. (e) Schematic of nonclassical crystal nucleation hosted by disordered precursors. Blue spheres, molecules in solution, and in disordered precursors. Red spheres, molecules in crystal nuclei, and crystals.

In the control experiments, no TQ was added. Adding 50 μM TQ led to a substantially slower increase of the g₂ shoulder than in the control runs (Figure 5b,c). The suppression of β-hematin crystal nucleation by TQ is affirmed by the delayed growth of the correlation function amplitude A₂ (defined in Figure 5d, inset, roughly the drop of g₂ between short and long τs) in the presence of the drug.

The suppression of β-hematin crystal nucleation by TQ may be the mechanism by which this drug inhibits ring-stage parasites, the stage at which hemozoin crystals nucleate (Figure 1c). The correlation between the suppression of crystal nucleation and the inhibition of ring-stage parasites is supported by observations with chloroquine (CQ) and pyronaridine. This quinoline-class antimalarial drug effectively suppresses the growth of β-hematin crystals,^46,66 which may be the mechanism by which it kills trophozoite-stage parasites.⁴⁸ We find, however, that CQ is a passive bystander of β-hematin nucleation (Figure S4a,b), and its inhibition of ring-stage parasites is weaker than that of trophozoites (Figure S4c,d). In further support of this correlation, pyronaridine, which inhibits β-hematin nucleation (Figure 4a), was more potent at ring-stage inhibition than at trophozoite stages.⁷⁴

TQ Suppresses Hematin-Rich Clusters.

The suppression of β-hematin nucleation by TQ appears incompatible with classical nucleation theory (CNT), a general framework often employed to understand crystal nucleation. CNT models the thermodynamic barrier to forming new crystals as the balance of two contributions: a free energy loss due to the formation of a new crystalline domain, in which the solute chemical potential is lower than in the solution, and a free energy gain due to the formation of a crystal-solution interface.^72,76,77 This barrier largely determines the nucleation rate. An additive, such as TQ, may adsorb on the nucleus surface and lower its surface free energy or serve as a substrate that deforms the nucleus shape. Importantly, both of these processes would only lower the barrier and thus enhance the nucleation rate.⁷⁸ The hypothetical sequestration of hematin as hematin-TQ complexes—that may lower the hematin chemical potential and suppress crystal nucleation—is incompatible with the independence of the critical radius R_c of newly formed 2D nuclei from C_TQ (Figure S3d).

The discrepancy between the observed TQ effects and the CNT prediction suggests that, similar to numerous other crystalline materials,^78,79 β-hematin crystal nucleation follows a nonclassical pathway. A common nonclassical crystal nucleation scenario is the two-step mechanism. In the first step, precursors form, which, in the second step, enable the nucleation of crystals (Figure 5e).^80–83 TQ may interact with the precursors in a way that lowers the nucleation rate of β-hematin. The nucleation precursors may be identical to the liquid particles seen landing on the crystal surface and generating a stack of layers (Figures 2d–h).

To identify potential precursors of β-hematin crystal nucleation and explore their response to TQ, we employ Brownian microscopy (BM, also called oblique illumination microscopy⁸⁴ or particle tracking⁸⁵). In this method, individual inhomogeneities, whose refractive index is greater than that of the solution, are detected from the light that they scatter (Figure 6a).^84,86 This light is directed by a microscope lens to a camera (Figure 6a), which records the locations of the detected particles and traces their trajectories (Figure 6a,b). From the trajectories of individual particles, we evaluate their diffusion coefficients and, employing the Stokes–Einstein relation,^84,86 their radii R. The numbers of particles N are counted. For experimental statistics, data on R and N, collected for 1 h stretches at 11 different focal depths (Figure 6a) and at five locations in the cuvette, are pooled and averaged.

Figure 6.

Mesoscopic hematin-rich clusters and their response to TQ. (a) Schematic of Brownian microscopy for monitoring solution inhomogeneities. (b,c) Speckle patterns of light scattered from hematin solutions with C_H = 0.05 mM at 25 °C. Bright spots correspond to hematin clusters. (b) C_TQ = 0. (c) C_TQ = 4 μM TQ. (d) The cluster size distributions at C_H = 0.05 mM and C_TQ = 0 at times indicated in the plots. (e) Schematic of the formation mechanism of mesoscopic hematin-rich clusters, highlighted by a blue disk, owing to accumulation of hematin dimers, indicated with an oval. (f) Clusters size distributions at C_H = 0.05 mM and increasing C_TQ as indicated in the plots.

We use solutions with C_H = 0.05 mM, lower than the solubility of β-hematin crystals in CBSO C_e = 0.08 mM, to distinguish any observed heterogeneities from crystals. In these solutions, BM detects particles whose radii R fall within a relatively narrow range, between 10 and 80 nm, with an average about 40 nm (Figure 6d). The observed particles are likely condensates of hematin since in its absence, the solutions are homogeneous. Importantly, the size distributions of the particles remain relatively steady over 60 min (Figures 6d and S4a), which excludes the possibility of the particles being crystallites or dense liquid droplets, whose numbers and sizes should increase over time as they nucleate and grow.^87–89 Steady numbers and sizes of the condensate domains, however, are signature behaviors of mesoscopic solute-rich clusters, observed with compounds ranging from proteins to large organic molecules and inorganic salts.^82,90–96 Analytical models of these clusters attribute the steady cluster size to the dynamics of formation, decay, and diffusion of solute oligomers, e.g., dimers (Figure 6e).^{91–93,97,98} The hematin particles comply remarkably well with the predictions of these models, and we identify them as hematin-rich clusters.

The addition of increasing amounts of TQ suppresses the number concentration N of the clusters (Figures 6c,f and S4b); the independence of R from C_TQ (Figures 6f and S4b) complies with the prediction of the dimer model for cluster formation.^{91–93,97,98} Although the mechanism is not fully understood, the destabilization of hematin-rich clusters by TQ may be due to the TQ-induced variation in properties such as viscosity. The correlation between the TQ-induced suppression of the hematin-rich clusters and β-hematin nucleation supports the role of the clusters in crystal nucleation and implies that TQ inhibits crystal nucleation by suppressing its precursors. Our observations appear to present a unique example of crystal nucleation control by impacting its nonclassical pathway.

The smaller clusters observed by BM with radii in the 10–20 nm range (Figure 6d) are similar in size to the liquid particles that land on hematin crystal surfaces and support crystal growth by transforming into stacks of new crystal layers (Figure 2b, parts d–h). If we accept the identity of the particles observed by AFM and BM, the AFM observations support the liquid nature of the mesoscopic hematin-rich clusters and exclude their solidification, which is seen with clusters of other compounds, particularly over longer life-times.^61,82 The clusters are likely uniformly distributed in the solution, and with the average $N\approx 16\times {10}^6{\mathrm{c}\mathrm{m}}^{-3}$ (Figure S5a), the mean distance separating them $=N^{-1/3}\cong 40\mu \mathrm{m}$. The diffusivity of a cluster with $R=20\mathrm{n}\mathrm{m}D=k_{\mathrm{B}}T(6\pi \eta R{)}^{-1}\approx 1\times {10}^{-12}{\mathrm{m}}^2{\mathrm{s}}^{-1}$ (where $k_{\mathrm{B}}$ is the Boltzmann constant, $T$ is the temperature, and $\eta =7.36\mathrm{m}\mathrm{P}\mathrm{a}\cdot \mathrm{s}$ is the solvent viscosity⁹⁹) implies their characteristic time of diffusion $\mathrm{\Delta }t\mathrm{\Delta }t=d^2/6D\approx$ 500 s. This long Δt explains why clusters are seen by AFM only after longer observations and why mostly small clusters, which diffuse faster, reach the crystal surface.

TQ Potently Suppresses Normal β-Hematin Growth Rate at Elevated Hematin Concentrations.

The role of hematin-rich clusters in β-hematin growth and their suppression by TQ suggests that TQ may mobilize a potent mechanism to suppress β-hematin growth by reducing the number of amorphous precursors. To test this hypothesis, we measured the growth rates from the shape evolution of crystals growing in a microfluidics channel, where we maintained constant solute concentration (C) by continuously supplying fresh growth solution in CBSO.^61,100 We used concentrations of 0.30 and 0.50 mM to discriminate TQ effects at moderate and high hematin concentrations (Figure 7b,c). The intense brown color of the crystals affords easy microscopic determination of the growth rate V_n of the axial {011} faces from the increase of crystal length (Figure 7a). The length-to-width ratio of the crystals, ca. 10:1, persists after growth at both concentrations, indicating that TQ inhibition of growth in the [001] direction, due to hematin incorporation in the (011) faces, is similar to that of the (010) faces.

Figure 7.

TQ suppression of crystal growth rate. (a) β-Hematin crystals in a microfluidics channel used in growth rate measurements. Arrows indicate the axial [001] direction of fast growth and the lateral [010] direction. (b,c) Growth rates V_n in the axial [001] direction measured in the microfluidics channel at hematin concentrations 0.30 mM, in b, and 0.50 mM, in c, in the absence of TQ (black) and in the presence of 50 μM TQ (red). Each point represents an individual measurement. Short horizontal bars indicate lowest and highest V_n within each data series. Upper, median, and lower horizontal lines exceed 75, 50, and 25% of the data points, respectively.

The normal growth rate measurements reveal that, consistent with the AFM data on the growth of the (100) face at low C_H and the hematin sequestration assay, at moderate hematin concentration, TQ does not inhibit β-hematin crystal growth (Figure 7b). By contrast, at elevated C_H, TQ potently suppresses β-hematin growth (Figure 7c); this suppression may correlate with the nonclassical crystal growth mode (Figures 2d–g) that likely dominates crystal growth in this C_H range and the suppression of the crystallization precursors by TQ (Figure 6f)

The suppression of β-hematin nucleation and growth by TQ sheds light on why TQ exerts inhibitory effects during both the ring and trophozoite stages (Figure 1c,d). During the ring stage, TQ suppresses the nucleation of β-hematin. In the trophozoite stage, TQ inhibits growth driven by dislocations and, more potently, by the deposition of precursors on the crystal surface.

CONCLUSIONS

The insights into the molecular-level pathways of nucleation, growth, and precursor formation of β-hematin reveal that the antimalarial drug tafenoquine employs a dual inhibitory mechanism. Parent TQ suppresses malaria parasites during their intraerythrocytic stages by disrupting the formation of mesoscopic hematin-rich clusters, which are essential precursors for nonclassical β-hematin nucleation, thereby inhibiting heme detoxification during the ring stage. Additionally, the clusters are crucial enablers of nonclassical β-hematin crystal growth in the trophozoite stage. This dual pathway of inhibition of nonclassical nucleation and growth substantially expands the drug’s kink-blocking activity during classical β-hematin growth at moderate hematin concentrations.

TQ accumulates in erythrocytes to double the concentration found in plasma.^101,102 The peak concentrations in plasma approach 1 μM when dosed for liver-stage activity. TQ accumulates in the liver at levels 40-fold higher than plasma concentrations. This work shows that parent TQ is able to inhibit hemozoin nucleation which predominantly occurs in ring-stage parasites and, to a lesser extent, in trophozoite stages. The peak blood concentrations of 1 μM in plasma and 2 μM in erythrocytes indicate a minor contribution to bloodstage parasite killing, separate from TQ liver-generated metabolites compromising the parasites’ mitochondria. Intra-erythrocytic Babesia, which lacks heme crystal formation, is cleared by tafenoquine most likely through mitochondrial inhibition by liver metabolites.¹⁰³

In contrast to traditional antimalarial agents from the quinoline class and the adducts of artemisinin family drugs, TQ exhibits a distinct and multifaceted mechanism of action. This unique behavior makes it an excellent element for combination therapies, offering the potential to enhance treatment efficacy while reducing the likelihood of drug resistance by targeting alternative pathways.

METHODS

Materials.

The following compounds were purchased from Sigma-Aldrich: citric acid (anhydrous, ≥99.5%), sodium hydroxide (anhydrous, ≥98%), n-octanol (anhydrous, ≥99%), and tafenoquine succinate (≥95%, HPLC). Porcine hematin (≥97%, HPLC) was purchased from Chem-Impex. All reagents were used as received without further purification unless otherwise noted. Deionized (DI) water was produced by a Thermo Fisher Scientific Barnstead Smart2Pure Water Purification System 6 UV/UF.

Preparation of Citric Buffer-Saturated Octanol.

We followed the procedure discussed by KN Olafson et al.¹⁰⁴ Five mL portion of citric buffer stock solution at pH 4.80 was placed in a 40 mL glass vial, and 10–20 mL of n-octanol was added. The resulting two-phase solution was sealed and allowed to equilibrate without stirring at 23 °C for 30 min. The denser citric buffer occupied the bottom layer, and the n-octanol was located on top. The top layer was decanted with a pipet. Fresh CBSO solutions were prepared before every crystal synthesis and in situ atomic force microscopy (AFM) experiment.

Preparation of β-Hematin.

To prepare the β-hematin crystals, 1 mL of 0.1 M NaOH was added to 10 mL of octanol in a 20 mL glass vial. The vial was left uncapped on a heating plate, with the temperature set to 60 °C for 12 h to remove the aqueous proportion. Octanol with NaOH aids in the dissolution of hematin. Following this, 10 mg of hematin was added to the solution, and the vial was sealed to prevent oxidation. The mixture was incubated at 60 °C in an incubator for 8–12 h and then filtered with 0.22 μm nylon filters. The concentration of hematin in the filtered solution was measured by UV–vis spectrometry, using an extinction coefficient of ε = 3.1 ± 0.1 cm⁻¹ mM⁻¹ at λ = 594 nm. The solution was then diluted to approximately 0.20 mM using citric acid-saturated octanol and citric buffer-saturated octanol to adjust the pH to 4.8. A scratched 15 μm-diameter round coverslip was immersed in 5–10 mL of 0.2 mM hematin solution and stored in a sealed vial at room temperature in a dark, stationary environment. After 3–5 days, small crystals could be observed under an optical microscope. The crystals reached their maximum size after approximately 14 days. The coverslips were removed from the vial and dried in air before being monitored by AFM.

In Situ Monitoring of β-Hematin Growth by AFM.

We employed a Multimode atomic force microscope and controller (Nanoscope VIII) from Bruker (Santa Barbara, CA) for in situ AFM. All AFM images were obtained in tapping mode (32 kHz) using BL-AC40TS probes from Oxford Instruments (tip radius = 8 nm and spring constant = 0.09 Nm⁻¹). The scan rate was maintained at 2.1 Hz. The detailed monitoring method was described by Ma et al^.72

Microfluidic Measurement of Crystal Growth Rates.

A semiautomatic inverted light microscope (Leica DMi8 equipped with PL Fluotar 5×, 10×, 20×, and N Plan 50× objectives)¹⁰⁵ was used to examine the growth of hematin crystals. For this study, hematin crystals were prepared on glass disks, and a glass disk was placed inside a polydimethylsiloxane (PDMS) channel. Growth solutions of two concentrations 0.30 and 0.50 mM were prepared, and each solution was pumped inside the channel continuously at a rate of 0.1 mL/min using a syringe pump (Chemyx, Fusion 4000). Images were captured every 5 min at multiple positions along the channel. Optical micrographs of crystals were analyzed by using ImageJ to determine the average length. The growth rate was measured by linear regression of 20 crystal lengths from 5 to 10 crystals.

β-Hematin Nucleation and Extension Assay.

Hematin in 100 mM sodium acetate (pH 4.8) was prepared fresh at 50 μM from a 10 mM hematin stock in DMSO. For each reaction, up to 245 μL of 50 μM hematin and either 1.25 mL of 10 mM 1-oleoyl-rac-glycerol (MOG) (Sigma M7765) for nucleation or 5 μL of 1 nmol/μL β-hematin for extension were added to a 1.5 mL Ependorf tube. For samples with drug, up to 10 μL of TQ was added to each reaction to achieve a range of final drug concentrations from 10 to 200 μM. Positive controls included 50 μM hemin and 5 nmol of β-hematin only to measure extension in the absence of drug inhibition. Negative controls included 5 nmol of β-hematin in sodium acetate only (without hematin) to account for any loss of β-hematin after washing. Each sample was run in triplicate tubes. Tubes were vortexed and incubated in a 37 °C water bath for 8, 16, or 48 h. After incubation, 10 μL of 10% SDS was added to each tube. β-Hematin crystals were isolated with a series of washes: first with 250 μL of 2% SDS/80 mM sodium bicarbonate and sonication, followed by 250 μL of 2% SDS and sonication. The remaining pellets were decrystallized using 250 μL of 2% SDS/50 mM NaOH at room temperature for 30 min, followed by heme quantification by spectroscopy.

Characterization of Hematin Nucleation by Dynamic Light Scattering.

Dynamic light scattering (DLS) measurements were performed using an ALV instrument (ALV-GmbH, Germany) equipped with a Cobolt Flamenco solid-state laser operating at a wavelength of 660 nm. The setup included the ALV/CGS-3 Compact Goniometer System and an ALV/LSE-5004 correlator. Measurements were conducted at a scattering angle of 90°, with each acquisition lasting 30 s. The solvent viscosity at the temperature of the measurement was used as specified. Approximately 2 mL of the solution was transferred into a clean, prewiped glass cuvette and placed in the sample holder. Toluene was used as the index matching fluid—as its refractive index is similar to that of glass—into which the cuvette containing the tested solution was held. To minimize time-dependent effects and enable accurate monitoring of nucleation, measurements were commenced immediately after sample preparation. The autocorrelation functions were analyzed using the G₂(τ) exponential model.

Characterization of the Mesoscopic Hematin-Rich Clusters by Brownian Microscopy.

We used a ZetaView Instrument SN 18-410 Mono (Particle Metrix GmbH, Germany). The instrument features a 520 nm laser and an 11-layer focus system, allowing the collection of more data compared with conventional single-layer particle analysis setups. The depth of focus is 25 μm, and the sample cell channel dimensions are 2 × 5 × 50 mm.

The laser beam interacts with heterogeneities in the solution, causing the particles to scatter light. This scattered light is detected by a microscope objective and recorded by a digital camera, enabling the visualization of particle motion. The data are analyzed by specialized software, which tracks the Brownian motion of individual particles. From this tracking, the mean squared displacement of each particle is determined, and the corresponding diffusion coefficient is evaluated. The Stokes-Einstein equation is employed to estimate the hydrodynamic radius of the particles.

The temperature is maintained at 25 °C throughout the measurements. To ensure accuracy, at least 3 mL of solution is pumped into the sample cell, taking care to remove air bubbles. Measurements are initiated only after confirming that convection is minimal or the drift is within the threshold for software autocorrection. Each measurement consists of capturing 90 frames at a frame rate of 30 frames per second. The trace length, defined as the minimum number of consecutive frames required for valid particle tracking, is set to 15 frames. Five independent measurements are performed for each time point, and the results are compiled and visualized in the corresponding figures.

In Vitro P. falciparum Inhibitory Drug Assay.

For this study, we obtained blood from living donors who are not identifiable to the researchers. Human erythrocytes are obtained from healthy volunteers under Johns Hopkins IRB-approved protocols.

Tafenoquine was dissolved in DMSO (Thermo Fisher). For use in parasite assays, we further diluted all tested compounds and adducts in inhibitor-free culture media, ensuring that DMSO or H₂O levels never exceeded 1% final concentration when introduced into parasite cultures.

In vitro experiments involved P. falciparum NF54, CAMWT, and CAM580Y, which were obtained through BEI Resources. Parasite cultures were maintained under a modification of the Trager and Jensen method.¹⁰⁶ Specifically, parasites were cultured at 2% hematocrit in RPMI 1640 media supplemented with l-glutamine, 25 mM HEPES, 0.25% NaHCO₃, 0.37 mM hypoxanthine, 50 μL of 50 mg/mL gentamicin, and 10% human serum. Cultures were incubated at 37 °C in 5%CO₂/5%O₂/balanced N₂.

P. falciparum cultures were synchronized to the ring stage by incubation in 5% sorbitol 48 h before, and immediately before, young ring pulse initiation. TQ was pulsed at concentrations ranging from 0 to 10000 nM on synchronized young ring-stage cultures for 6 h, washed three times in culture media without hypoxanthine, and then returned to complete culture media and incubated at 37 °C. The old ring pulse was initiated 6 h after the young ring pulse, the young trophozoite pulse was initiated 24 h after the young ring pulse, and the old trophozoite pulse was initiated 30 h after the young ring pulse. The full method is depicted graphically in ref.⁴⁸ Negative growth control wells contained 10 μM chloroquine. Positive growth control wells contained metabolite- or drug-free culture media. Parasite growth was determined by staining with SYBR Green (10 μL of the 1:10 dilution added to a 96-well plate containing 100 μL of 1× SYBR Green in PBS, and incubated for 30 min while shaking) and analyzed via flow cytometry on the Attune NxT Flow Cytometer (Thermo Fisher Scientific) with a 50 μL acquisition volume and a running speed of 25 μL/min, with 10,000 total events collected.

In vitro P. falciparum experiments (Figure 1c,d) were performed in a minimum of biological triplicates on different dates with technical triplicate wells and noted drug dilutions. The Prism software package was used to generate dose-response versus inhibition curves with a nonlinear fit.

Supplementary Material

Mesoscopic hematin-rich clusters participating in Brownian motion in the presence of tafenoquine

Movie S1: Classical growth of β-hematin with layers generated by dislocations (MP4)

Mesoscopic hematin-rich clusters participating in Brownian motion

Movie S2: Classical growth of β-hematin with layers generated by 2D nucleation (MP4)

Classical growth of β-hematin with layers generated by 2D nucleation

Movie S3: Nonclassical growth of β-hematin with mesoscopic clusters landing on the crystal surface (MP4)

supplement 2 tables and 4 figures

Table S1: Measurement of 2D critical radii R_c and surface free energy of the steps at different concentrations of hematin; Table S2: ANOVA test of surface free energy with and without TQ; Figure S1: fractions of P. falciparumparasites killed by continuous application of TQ for 72 h;Figure S2: the correlation between step separation and step velocity; Figure S3: measurement of critical radii and Turnbull rule fitting; Figure S4: the effects of chloroquine (CQ) on β-hematin nucleation and suppression of ring- and trophozoite-stage parasites; Figure S5: time evolution of mesoscopic hematin-rich clusters and their response to TQ (PDF)

Classical growth of β-hematin with layers generated by dislocations

Movie S4: Mesoscopic hematin-rich clusters participating in Brownian motion in hematin solution (MP4)

Nonclassical growth of β-hematin with mesoscopic clusters

Movie S5: Mesoscopic hematin-rich clusters participating in Brownian motion in the presence of tafenoquine (MP4)

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.biochem.5c00164.