Antimalarials inhibit hematin crystallization by unique drug–surface site interactions

Significance

Approximately 3.2 billion people are at risk for malaria. The resistance of malaria parasites to current advanced multidrug treatments, recently recorded to spread out of Southeast Asia, has raised concerns of dire public health consequences. We demonstrate that antimalarial drugs suppress heme detoxification in the malaria parasites in a manner that is counter to the prevailing hypothesis in the field. We find that quinoline-class drugs work by specific interactions with β-hematin crystals, which are the by-product of heme detoxification within the digestive vacuole of the parasites. We also identify specific drug adsorption sites on crystal surfaces. These insights may potentially spur development of antimalarial drugs that overcome parasite resistance through a rational approach, far superior to currently used combinatorial methods.

Abstract

In malaria pathophysiology, divergent hypotheses on the inhibition of hematin crystallization posit that drugs act either by the sequestration of soluble hematin or their interaction with crystal surfaces. We use physiologically relevant, time-resolved in situ surface observations and show that quinoline antimalarials inhibit β-hematin crystal surfaces by three distinct modes of action: step pinning, kink blocking, and step bunch induction. Detailed experimental evidence of kink blocking validates classical theory and demonstrates that this mechanism is not the most effective inhibition pathway. Quinolines also form various complexes with soluble hematin, but complexation is insufficient to suppress heme detoxification and is a poor indicator of drug specificity. Collectively, our findings reveal the significance of drug–crystal interactions and open avenues for rationally designing antimalarial compounds.

Many pathological conditions are understood within the realms of molecular biology and biochemistry. An important class of diseases, whose pathology involves the formation of solid or liquid condensate, stands to benefit from the introduction of concepts and mechanisms from physics and materials science (1–3). A prominent example is the formation of crystalline hemozoin, a vital component of malaria parasite physiology (4–7). Malaria infection starts with a mosquito bite introducing Plasmodium sporozoites, which invade the host’s liver cells (Fig. 1). Within 2 wk, merozoite-stage parasites invade the erythrocytes, where they catabolize hemoglobin and release Fe(II) heme (5). The released heme rapidly oxidizes to Fe(III) hematin, which is toxic to the parasite in its free state (6). The main mechanism of heme detoxification (8) implemented by the parasite is sequestration as nontoxic hemozoin crystals (9–12) (Fig. 1C). Hematin crystallization (the synthetic analog of hemozoin has been referred to as β-hematin) (7) has been the most successful molecular target for antimalarial drugs (13, 14). Malaria parasite resistance to current drugs (15–18), including the most advanced line of antimalarial defense, artemisinin (19–24), has renewed the impetus to elucidate the pathways of parasite suppression (25–27).

Fig. 1.

Lifecycle of malaria parasites and hemozoin formation. (A) In human hosts, the lifecycle of malaria parasites consists of liver and red blood cell (RBC) stages. The characteristic times after RBC invasion for parasite growth and division, RBC rupture, and transfer of infection are displayed. (B) Schematic rendering of RBCs. (C) Parasite inhabiting an RBC with a nucleus (N) and a DV containing hemozoin crystals. (D) SEM image of a hematin crystal grown in vitro. (Scale bar, 5 μm.)

It is generally accepted that the quinoline class of antimalarials (quinine, chloroquine, amodiaquine, and others) work by suppressing hematin crystallization (8, 13, 21, 25, 26, 28, 29). The prevailing hypothesis is that drugs decrease the activity of soluble hematin by forming noncrystallizable complexes (13, 30). An alternative inhibition mechanism, involving drug–crystal interaction, has been suggested by experimental evidence (29, 31–38) and theoretical models (39, 40).

A controversy surrounds the environment of hematin crystallization in the parasite digestive vacuole (DV). A group of authors has argued that several neutral lipids are present in the DV as mesoscopic phases that accumulate hematin and serve as crystallization medium (41–43); this scenario has been supported by numerical modeling (44). High-resolution electron microscopy and X-ray tomography, however, have failed to detect lipid phases larger than 25 nm, interpreted in favor of growth from the aqueous environment (10–12), where crystallization may be assisted by a protein complex that catalyzes hematin dimerization (45). In a previous paper, we reconciled these seemingly opposite viewpoints by suggesting that β-hematin crystals grow from a thin shroud of lipid that coats their surface (46), a mechanism that is qualitatively consistent with experimental observations (47). Driven by the evidence favoring neutral lipids as a preferred environment for hematin crystallization in vivo (46), we use a solvent comprising octanol saturated with citric buffer (CBSO) at pH 4.8 as a growth medium.

Recent observations of hematin crystallization from a biomimetic organic medium demonstrated that it strictly follows a classical mechanism of growth (48), where new layers are generated by 2D nucleation and advance by the attachment of solute molecules. This finding allowed identification of the rate of nucleation of new layers, J_2D, and the velocity of advancing steps v as the main quantitative measures of β-hematin crystal growth and suggested criteria to differentiate the specificity and efficacy of growth inhibitors that bind to β-hematin crystal surfaces (48). The physiological relevance of these conclusions is reinforced by the observation that the morphology of the grown crystals is identical to biological hemozoin (28), and their growth rate is similar to those observed in parasites (42, 48).

To establish whether drug–crystal interactions are the dominant mechanism of hematin growth inhibition, we monitor and quantify the processes of hematin crystallization by time-resolved in situ atomic force microscopy (AFM). We use CBSO as the solvent (48). To this growth solution, we introduce six common antimalarial drugs: quinine (QN), chloroquine (CQ), pyronaridine (PY), amodiaquine (AQ), mefloquine (MQ), and artemisinin (ART). We find that in the presence of any of the tested drugs, the basal {100} β-hematin faces grow via a classical layer mechanism, analogous to the one observed in pure CBSO solutions (48).

Results and Discussion

Molecular Mechanisms of Drug Action.

Three of the tested drugs, QN, CQ, and PY, depicted in Fig. 2A, cause a sharp decay of J_2D (Fig. 2B) accompanied by a monotonic decrease in v relative to its value v₀ in solutions of pure hematin with concentration c_H = 0.28 mM, at supersaturation $\sigma =\text{ln}(c_H/c_e)\cong 0.56$, where c_e = 0.16 mM is the solubility at 28 °C, the temperature in the AFM liquid cell (48) (Fig. 2C). AFM observations, presented in Fig. 2D and SI Appendix, Figs. S1–S3, reveal protrusions along advancing steps in the presence of drugs. These corrugated step edges are in contrast to observations in pure solutions (48), indicating a specificity of the drugs for {100} faces; the {100} faces are the largest of the crystal, implying that they are the slowest growing. The concomitant reduction of J_2D and v with increasing drug concentration c_D suggests that the drugs inhibit 2D nucleation of new layers and step motion by a step-pinning mechanism (49), illustrated in Fig. 2F. This mechanism assumes that inhibitors preferentially bind to terraces with a surface coverage governed by the dynamics of adsorption. If the separation between a pair of adsorbed inhibitors Δx is less than the diameter of the critical layer nucleus 2R_crit, the adsorbates enforce a curvature at which the advancing step is undersaturated. The growth of existing steps is arrested (49), whereas newly formed 2D islands are prevented from reaching the critical size and are forced to dissipate. At intermediate inhibitor concentrations, step pinning produces corrugated steps (Fig. 2D) and suppresses 2D layer nucleation. The significance of this mechanism is reflected in the sensitivity of the crystallization rate to the drug concentration (Fig. 2 B and C), which may be a critical factor underlying the increased resistance of Plasmodium falciparum parasite to CQ (50). For instance, resistant parasite strains may have developed means to reduce CQ concentration within the DV to levels that permit effective heme detoxification.

Fig. 2.

Step-pinning action of three common antimalarials. (A) The structures of QN, CQ, and PY. (B) A decrease in the rate of 2D nucleation of new layers J_2D relative to that in the absence of any drug, J_2D,0, with increasing drug concentration. (C) A decrease in step velocity v relative to that in the absence of any drug, v₀, with increasing drug concentration. In the presence of QN and CQ, v in the $\stackrel{\rightharpoonup }{c}$ and $\stackrel{\rightharpoonup }{b}$ directions maintains a constant ratio of 2.7 (here, only v in the $\stackrel{\rightharpoonup }{c}$ direction is shown). For PY, the ratio increases to ∼8. (D and E) AFM images of islands and steps on a ($\overline{1}00$) face growing in a 0.27 mM hematin solution in CBSO in the presence of QN and PY at concentrations shown in the respective images. The gold arrow in E indicates an area of step bunching. The imaged face is ∼1 × 13 μm². (F) Schematic of the step-pinning mechanism, where Δx is the separation between drug molecules adsorbed on flat crystal terraces, shown in yellow, and R_crit is the critical radius of the 2D nucleus. (G) Schematic of step bunching induced by the putative interaction between PY molecules, shown in purple, and $\stackrel{\rightharpoonup }{b}$ direction steps.

Besides step pinning, which suppresses J_2D and slows down v, the drug PY appears to follow an additional mechanism of action. AFM observations reveal a decrease in v along the $\stackrel{\rightharpoonup }{b}$ crystallographic direction in response to increasing PY concentration that is more pronounced than along the $\stackrel{\rightharpoonup }{c}$ direction. Correspondingly, the islands in Fig. 2E exhibit exaggerated ellipticity, with a c/b ratio of ∼8, compared with 2.7 in pure solutions (51). The sensitivity of the $\stackrel{\rightharpoonup }{b}$-oriented step growth to PY concentration suggests that this drug preferentially attaches to $\stackrel{\rightharpoonup }{b}$ steps, blocking the growth sites available for hematin incorporation. An idealized schematic of this interaction is presented in Fig. 2G. The hindered $\stackrel{\rightharpoonup }{b}$ steps form strips of high-density step bunches, Fig. 2E and SI Appendix, Fig. S3 (52–54), which amplifies the PY inhibition due to competitive supply of solute to closely spaced steps (51).

$v\left(c_D\right)$ profiles similar to those in Fig. 2 B and C are compatible with an alternative mechanism of growth inhibition, whereby inhibitors incorporate in the crystals lattice. The resulting lattice strain increases the chemical potential of the crystal, which raises the solubility and lowers the driving force for growth (55–57). In previous observations of this scenario, the solubility increase was significant if the inhibitor concentration in the solution was comparable to that of the solute (55–57). AFM images of hematin crystals reveal protrusions along advancing steps in the presence of drug (Fig. 2D and SI Appendix, Figs. S1–S3), which seems to indicate that alternative pathways (e.g., step pinning) are the principal routes to inhibit step growth; however, in the absence of data on the thermodynamics of drug incorporation into β-hematin crystals and their elastic properties, the incorporation mechanism remains a feasible pathway of β-hematin growth inhibition by CQ, QN, and PY.

Trends of J_2D and v suggest that the drugs AQ and MQ (Fig. 3A) exhibit mechanisms that differ from those of QN, CQ, and PY (SI Appendix, Fig. S7). Both AQ and MQ suppress v by about half (Fig. 3C), but only AQ inhibits the nucleation of new layers (Fig. 3B). The gradual decrease of v at increasing inhibitor concentration is consistent with a mechanism wherein inhibitor molecules adsorb on kinks at the steps and block hematin molecules from incorporating, as seen with other materials (55, 58) and illustrated in Fig. 3F. To quantify the inhibition of step motion by partial blocking of kinks, we assume that v is proportional to the density of unoccupied kinks, $v=v_0–\left(v_0–v_{\infty }\right)n_D$, where v₀ = v in pure solutions, v_∞ is reached at the maximum surface inhibitor coverage, and n_D is drug adsorption coverage on the kinks, governed by a Langmuir isotherm, $n_D=c_D{\left(A+c_D\right)}^{-1}$, where A is a constant. We obtain, similar to Bliznakov (59), ${\left(v_0–v\right)}^{-1}={\left(v_0–v_{\infty }\right)}^{-1}+\left[A{\left(v_0–v_{\infty }\right)}^{-1}\right]c_D^{-1}$. The scaling between $v^{-1}$ and the $c_D^{-1}$ in the presence of AQ and MQ in Fig. 3 D and E and SI Appendix, Fig. S8 A and B supports that action of this mechanism.

Fig. 3.

Inhibition of hematin growth by kink blocking. (A) Structures of AQ, MQ, and ART. (B) The rate of 2D nucleation of new layers J_2D relative to that in the absence of any drug, J_2D,0, with increasing drug concentration. (C) The step velocity v relative to that in the absence of any drug, v₀, with increasing drug concentration for $\stackrel{\rightharpoonup }{c}$ direction steps. Dashed and dotted lines denote the reduction of v due to the sequestration of hematin in respective drug–hematin complexes (Fig. 4C). (D and E) The correlation between v and the respective concentrations of AQ and MQ, c_D, in modified reciprocal coordinates. (F) Schematic of drug molecules (shown in green) inhibiting step advancement by partial blocking of kinks. (G) AFM image of islands and steps growing in a 0.27 mM hematin solution in CBSO in the presence of 1.5 μM AQ. (H) Submolecular resolution image of a ($\overline{1}00$) hematin crystal surface. Blue spheres highlight carboxyl, methyl, and vinyl C atoms as seen in the hematin model (Inset). C atoms marked with yellow spheres belong to the same molecule. (I) High-resolution image of a step edge, highlighted with a white contour; higher terrace is on the right. (J) Monitoring the evolution of a step edge with disabled y-axis scans, beginning at time = 4 s (yellow line) reveals temporal displacements of the step due to the attachment or detachment of solute molecules; higher terrace is on the right.

The morphology of advancing layers in the presence of AQ and MQ (Fig. 3G and SI Appendix, Figs. S5 and S6) is similar to that in pure solutions, suggesting that inhibitor adsorption on kinks is independent of step orientation. The independence of J_2D of the MQ concentration suggests that MQ does not adsorb on molecularly flat surfaces or dock to the edges of small, newly nucleated islands. Meanwhile, the addition of ART has no apparent effect on the surface features and the kinetics of layer nucleation and growth (Fig. 3 B and C and SI Appendix, Fig. S4). This result is not surprising given that ART is believed to attain activity after reduction of the endoperoxide bridge by freshly released Fe(II) heme (4, 22, 24, 60).

We observe that drug action by kink blocking does not fully suppress step growth, but rather results in a finite step velocity v_∞ at high drug concentrations (i.e., 0.65v₀ for AQ and 0.4v₀ for MQ in Fig. 3C). To elucidate this phenomenon, we image the structure of advancing steps at submolecular resolution, as shown in Fig. 3H. High-resolution AFM images of the step edges (Fig. 3I) indicate that kinks are separated by distances of ∼1 nm. Monitoring the evolution of a step edge (as in Fig. 3J) reveals that it shifts by about 0.55 nm, consistent with incorporation/removal of a single hematin molecule. The high kink density suggests that the tail of a drug molecule occupying a kink may overlap adjacent sites and thereby protect them from docking of other drug molecules, but may be insufficient to prevent hematin molecules from accessing kinks and thus inhibit further step growth.

Drug Effects on Bulk Crystal Morphology.

Owing to the crystal anisotropy, inhibitor molecules that bind to crystal surfaces exhibit marked differences in inhibition of growth in different crystallographic directions. To assess the anisotropic growth rates, as illustrated in Fig. 4 A and B, we perform bulk crystallization studies to determine the macroscopic dimensions of β-hematin crystals grown for 16 d at varying concentrations of the six drugs (Fig. 4 C and D); these experiments were carried out at c_H = 0.24 mM and at supersaturation $\sigma =\text{ln}(c_H/c_e)\cong 0.7$, where c_e = 0.12 mM is the solubility at 25 °C. The drug ART does not affect the size and morphology of crystals grown in its presence, indicating that ART does not specifically interact with β-hematin surfaces (Fig. 4B), which is consistent with AFM data. In the presence of CQ, QN, PY, and AQ, average crystal lengths l are shorter (Fig. 4C), indicating preferential inhibition of the axial faces, (001) and/or $\left(0\overline{1}\overline{1}\right)$, as illustrated in Fig. 3E. Changes in the average aspect ratio A_sp of crystal length to width (Fig. 4C) is an indication that the drug selectively binds to either the lateral {010} faces or the axial (001) and $\left(0\overline{1}\overline{1}\right)$ faces (Fig. 4 E and F). The data in Fig. 4C indicate that CQ preferentially suppresses growth of the axial faces, whereas QN, PY, and AQ have identical effects on the lateral and axial faces. Similar studies of MQ reveal it to be a weak inhibitor of β-hematin crystal growth in both the lateral and axial directions.

Fig. 4.

Drug effects on crystal size and morphology. (A) Illustration of β-hematin crystal habit. (B) Growth in pure solutions preserves the crystal shape. (C and D) Variations of the average length l and length-to-width, l/w, aspect ratio A_sp of crystals grown in pure CBSO solutions and in the presence of increasing concentrations of five drugs for 16 d at 23 °C. (E and F) Illustrations of modifications to the crystal habit. Drug-induced suppression of crystal l and w by interaction of drugs with axial (E) and lateral (F) crystal faces, respectively. (G) Tapering due to enhanced adsorption of drugs near the crystal edges. (H–L) Scanning electron micrographs of crystals grown in the presence of drugs at concentrations listed in each panel. Tapering due to impurity action on the $\left(0\overline{11}\right)$ face is shown in K, where (K, I) and (K, II) are enlargements of respective dashed boxes.

Fig. 4 C and D reveals that higher inhibitor concentrations are needed to suppress bulk growth than the nucleation and spreading of layers on the {100} faces (Figs. 2 and 3). Discrepancies between the inhibitor concentrations needed to arrest step growth observed by AFM and those that suppress the growth of crystals in a bulk crystallization experiment are common in the crystallization literature (56, 61, 62). Several processes may contribute to this disparity. Bulk crystallization experiments are carried out at σ = 0.70, whereas AFM surface monitoring uses σ = 0.56. The rates of nucleation and growth of layers are significantly faster at the higher σ, leading to shorter times between subsequent layers and lower impurity adsorption. Furthermore, the crystal dimensions evolve owing to the growth of all faces in the crystal habit, whereas Fig. 4 C and D characterizes growth in the [010], [001], and $\left[0\overline{1}\overline{1}\right]$ directions. J_2D and v on the (100) face contribute to growth in the [100] direction. Growth kinetics and their response to inhibitors are anisotropic and may significantly differ for individual crystallographic directions. In addition, it is feasible that in a bulk experiment, in which numerous crystals with significant surface area grow for 16 d, the inhibitor concentration decreases during this time owing to incorporation into the growing crystals, whereas it was maintained at a constant value during short-term AFM monitoring. Lastly, during surface scans the iterative motion of the AFM probe and cantilever stirs the solution and homogenizes the inhibitor concentration throughout the solution volume. In contrast, a significant inhibitor concentration gradient may develop in a steady solution during long-term bulk crystallization experiments, leading to significantly lower inhibitor concentration at the crystal surface than in the solution bulk (63).

Observations of the crystal morphology (Fig. 4 H–L) reveal sharp distinctions between the different drugs. Electron micrographs in Fig. 4H indicate that CQ renders β-hematin crystals short and wide whereas PY induces crystal tapering at both axial termini (Fig. 4 I and J). Tapering has been attributed (64) to blocking of steps on an orthogonal face on approach to a shared edge, induced by enhanced supply of inhibitors at the edge, as illustrated in Fig. 4G. The symmetric tapering in Fig. 4 I and J indicates that PY interacts with both (001) and $\left(0\overline{1}\overline{1}\right)$ faces. In contrast, QN and AQ induce tapering at a single axial face, $\left(0\overline{1}\overline{1}\right)$ for QN in Fig. 4K, and (001) for AQ in Fig. 4L.

How Important is the Contribution of Drug–Hematin Complexation?

Our findings refute the hypothesis that drug–hematin binding, which sequesters hematin into soluble complexes, is the main mode of hemozoin growth inhibition (13). Sequestration would completely suppress the growth of β-hematin crystals if the residual concentration of unliganded hematin $\left[H\right]$ is lowered to values at or below its solubility c_e. To compare the relative efficacy of complexation and crystal surface interaction as inhibition pathways and to test if complexation dictates the selection of the drug mode of action, we characterized drug (D)–hematin (H) binding equilibria in CBSO, using a method based on UV-visible spectroscopy, previously used for predominantly aqueous solvents (65–68). The UV-visible spectra of hematin in the presence of six drugs reveal that PY and ART do not form complexes with hematin in CBSO (SI Appendix, Figs. S10 and S11). Using the spectroscopic data obtained in the presence of QN, CQ, AQ, and MQ (Fig. 5A), we discriminate between several different stoichiometries and complexation steps and evaluate the respective binding constants (Fig. 5B and SI Appendix).

Fig. 5.

Drug–hematin complexation and speciation. (A) UV-visible spectra of hematin at three concentrations c_H in the presence of QN, CQ, AQ, and MQ where the inhibitor concentration increases in the direction of the arrows (note that PY and ART data revealing the lack of complexation are displayed in SI Appendix, Figs. S10 and S11). (B) Relative decrease of the absorbance of a solution with c_H = 0.3 mM at 594 nm as a function of the concentration of the respective drug. Symbols are absorbance data and lines are the complexation model fits (H, hematin; D, drug; HD and H₂D, complexes). The relevant binding constants are listed; units of K₁ and K₂ for QN are mM⁻¹, whereas the unit of K for the other drugs is mM⁻². (C) Distribution of hematin, drug, and complex species as a function of total drug concentration at c_H = 0.3 mM. Vertical dotted line marks the QN concentration at which the concentration of unliganded hematin $\left[H\right]$ decreases below the solubility c_e = 0.16 mM, marked with horizontal dashed lines in each plot. Purple shading highlights the concentration range of each drug tested in Figs. 1 and 2.

Drug–hematin complexation lowers $\left[H\right]$ below c_e = 0.16 mM (48) when the drug concentrations exceed 0.09 and 0.07 mM for QN and CQ, respectively. This is more than 20-fold higher than the concentrations that induce complete growth cessation in AFM studies (∼1 μM in Fig. 2 C and D). These considerations suggest that drug interaction with the β-hematin crystal surface is a significantly more efficient pathway to inhibit hematin crystallization than complexation of free hematin in the solution. Our findings indicate that the formation of complexes does not correlate with the selection of specific inhibition mechanism; however, for drugs such as AQ and MQ, the combination of relatively strong drug–hematin binding and the limited growth suppression (Fig. 3 C, I, and J) increases the significance of drug complexation as a contributing inhibition pathway.

Conclusions

We have demonstrated that five antimalarials exhibit distinct mechanisms of β-hematin crystal growth suppression based on their specific interactions with crystal terraces, step edges, kinks, or newly nucleated islands. We show that drug–crystal interactions are a significantly more efficient pathway to inhibition of hematin crystallization than sequestration of soluble hematin into drug–hematin complexes; it is feasible that the species that absorbs on a specific surface site is a drug–hematin complex. Through physiologically relevant techniques, we find that kink blocking is implemented by two antimalarials. We find that drug–crystal interactions do not correlate with the propensity of drugs to form complexes with soluble hematin, but are predominantly driven by chemical recognition between drug molecules and the topological features of β-hematin crystal surfaces. These findings may prove influential for the design of new antimalarials and demonstrate, in the broader context of condensation diseases, that the application of physical science approaches to crystallization in living organisms can provide valuable insight into parasite physiology.