Novel Antimalarial Aminoquinolines: Heme Binding and Effects on Normal or Plasmodium falciparum-Parasitized Human Erythrocytes

Abstract

Two new quinolizidinyl-alkyl derivatives of 7-chloro-4-aminoquinoline, named AM-1 and AP4b, which are highly effective in vitro against both the D10 (chloroquine [CQ] susceptible) and W2 (CQ resistant) strains of Plasmodium falciparum and in vivo in the rodent malaria model, have been studied for their ability to bind to and be internalized by normal or parasitized human red blood cells (RBC) and for their effects on RBC membrane stability. In addition, an analysis of the heme binding properties of these compounds and of their ability to inhibit beta-hematin formation in vitro has been performed. Binding of AM1 or AP4b to RBC is rapid, dose dependent, and linearly related to RBC density. Their accumulation in parasitized RBC (pRBC) is increased twofold compared to levels in normal RBC. Binding of AM1 or AP4b to both normal and pRBC is higher than that of CQ, in agreement with the lower pK_a and higher lipophilicity of the compounds. AM1 or AP4b is not hemolytic per se and is less hemolytic than CQ when hemolysis is accelerated (induced) by hematin. Moreover, AM-1 and AP4b bind heme with a stoichiometry of interaction similar to that of CQ (about 1:1.7) but with a lower affinity. They both inhibit dose dependently the formation of beta-hematin in vitro with a 50% inhibitory concentration comparable to that of CQ. Taken together, these results suggest that the antimalarial activity of AM1 or AP4b is likely due to inhibition of hemozoin formation and that the efficacy of these compounds against the CQ-resistant strains can be ascribed to their hydrophobicity and capacity to accumulate in the vacuolar lipid (elevated lipid accumulation ratios).

Chloroquine (CQ) has been effectively used for decades as an antimalarial drug for its efficacy, safety, and stability but is now largely ineffective because of widespread parasite resistance. This has led to the necessity of utilizing artemisinin-based combination therapy as a first-line treatment for uncomplicated malaria (20). However, the recent reports of artemisinin-based combination therapy treatment failure in southeast Asia and the potential emergence of artemisinin resistance (26) indicate that the search for new drugs or new drug combinations is still highly necessary (40, 41). Quinoline-type antimalarials remain an attractive class of compounds because their mechanism of action and resistance are unrelated (14) and resistance has emerged very slowly over time. CQ and quinoline antimalarial drugs are thought to kill parasites by inhibiting the process of crystallization necessary to detoxify ferriprotoporphyrin IX (FP) into hemozoin (HZ) (5, 30, 34). This process is vital for Plasmodium falciparum and can be reproduced in vitro with hematin under acidic conditions, leading to the formation of beta-hematin (BH), a crystal product showing physicochemical properties identical to those of HZ (29).

FP in the food vacuole of the parasite is considered the target of quinoline antimalarials (6, 15). The inhibition of HZ formation leads to the accumulation of free FP, potentially toxic for its prooxidant and lytic activities (7, 27). The resistance mechanism of CQ is based on the reduced drug accumulation at the site of action, the food vacuole, due to mutation of PfCRT (P. falciparum CQ resistance transporter) (5, 16, 39) and does not involve modification of the target.

In the last few years, the research of our group has been directed to the study of new quinolizidinyl and quinolizidinyl-alkyl derivatives of 7-chloro-4-aminoquinoline (33). The terminal quinolizidine ring (octahydro-2H-quinolizine) confers basicity and lipophilicity to the molecules and prevents the metabolic oxidative dealkylation known to limit the usefulness of quinoline derivatives (28). Lead candidates have been obtained that are very stable and highly effective in vitro against both the D10 (CQ susceptible [CQ-S]) and W2 (CQ resistant [CQ-R]) strains of P. falciparum (33) and in vivo in the murine Plasmodium berghei model (23). Two of these compounds, named AM-1 and AP4b, have been selected for further characterization (Fig. 1): AM-1, a pure enantiomer, is semisynthetic and derives from l-lupinine, a natural alkaloid extracted from the seeds of Lupinus luteus; AP4b is synthetic and a racemate.

FIG. 1.

Molecular structures of the compounds used in the study.

For the present article, we evaluated the binding of these compounds to human red blood cells (RBC), normal or parasitized by the P. falciparum strain D10 or W2. In addition, an analysis of the heme binding properties of these compounds and of their ability to affect RBC membrane stability and inhibit FP crystallization in vitro has been performed.

MATERIALS AND METHODS

Chemicals and reagents.

AM-1 and AP4b were synthesized as described previously (3, 33). Human A-positive blood was obtained from healthy donors with citrate-phosphate-dextrose as an anticoagulant and used within 10 days of withdrawal.

All biochemicals were purchased from Sigma (Sigma Italia, Milan, Italy). FP stock solutions were prepared daily by resuspension in 0.02 N NaOH and dilution in phosphate-buffered saline (PBS) and heme equivalents quantified by dissolving an aliquot in 1 N NaOH and reading the absorbance at 385 nm. (ɛ₃₈₅ hematin = 6.1 × 10⁴ M cm⁻¹).

Determination of physicochemical properties.

The pK_a values of the studied compounds were determined potentiometrically at 25°C using a PCA 101 instrument from Sirius Analytical Instruments (East Sussex, United Kingdom). Titrations were done in triplicate in 0.15 M KCl with 0.5 M HCl and 0.5 M KOH, respectively. Partition coefficients of the compounds in n-octanol-0.15 M KCl were determined potentiometrically using the same instrument with two different volume ratios of organic and aqueous phases (0.25 and 0.67). The pKa, log P_Octanol, and log D (distribution coefficient, defined as the ratio of the sum of concentrations for all solute species in the n-octanol phase to the same of the aqueous phase at selected pH) values were calculated from the titration curves using the Refinement Pro software program, v. 1.0 (Sirius Analytical Instruments).

Alternatively, log D values were calculated using the following equation: log D = log P − log[1 + 10^{(pKa1 − pH)} + 10^{(pKa1 + pKa2 − 2pH)}] (35). Vacuolar accumulation ratios (VAR) and lipid accumulation ratios (LAR) were calculated as described previously (37).

Preparation of RBC and RBC ghosts.

Aliquots of blood were centrifuged at 1,850 × g at 4°C for 5 min, the buffy coat was removed, and the erythrocyte pellet was washed three times with 10 volumes of cold (4°C) PBS. Cells were gently resuspended with PBS containing 5 mM glucose and used immediately after isolation. Erythrocyte unsealed ghosts were prepared from washed cells by hypotonic lysis and extensive washing in 20 volumes of 5 mM sodium phosphate buffer (pH 8.0) (27). Parasite cultures and antimalarial assays using both CQ-S (D10) and CQ-R (W-2) P. falciparum strains were carried out as described by Sparatore et al. (33).

Effect of aminoquinolines on human RBC stability.

The effects of AM1 and AP4b on RBC stability were investigated by measuring RBC lysis after incubation of RBC at 5% hematocrit (Htc) in PBS-glucose (2 h, 37°C) with various concentrations of the compounds alone or in the presence of 10 μM FP, which is known to destabilize RBC membranes (8, 27). CQ was used as a control. At the end of incubation, RBC were centrifuged and hemolysis was estimated spectrophotometrically from the absorbance at 412 nm (A₄₁₂) of hemoglobin (Hb) released in the supernatant.

RBC binding and uptake of aminoquinoline compounds.

The uptake of the 4-aminoquinoline compounds by RBC was determined by a fluorimetric method described recently (2). Briefly, human RBC at 15% Htc in PBS-5 mM glucose were incubated at 37°C with serial dilutions (10 to 50 μM) of the test compounds using CQ as a reference. At the end of incubation, cells were pelleted by centrifugation, washed in PBS-glucose to remove the unbound drug, and subjected to an organic extraction in CHCl₃-1 N HCl. Fluorescence of CQ, AM-1, and AP4b was determined in the aqueous phase at an excitation and emission λ corresponding to 258 nm and 380 nm, respectively.

Aminoquinoline-FP interaction.

Interaction of aminoquinolines with FP was investigated by monitoring their effect on the absorption profile of FP. FP (15 μM, final concentration) was diluted in 10 mM sodium phosphate buffer, pH 6.0, containing 40% dimethyl sulfoxide (DMSO) (binding buffer) alone or in the presence of increasing concentrations of the compounds (1 to 200 μM) and absorption spectra were recorded after 2 min (21). In parallel, the spectral changes were determined by a continuous-variation technique (Job's plot) (1). Solutions containing different drug/FP (molar) combinations were prepared. The final combined concentration of hematin plus each compound in the mixtures was 10 μM. Plots were constructed between the difference in the expected FP absorbance at 400 nm and the experimental value versus the drug/FP molar ratio.

Aminoquinoline-induced denaturation of Hb.

The capability of aminoquinoline compounds of accelerating Hb denaturation in acidic conditions was evaluated according to the method of Fitch and Russell (17). Briefly, suspensions of intact RBC prepared from 50 μl of packed RBC were diluted to 10 ml with 150 mM sodium acetate buffer (pH 5.0) and incubated for 90 min at 39°C alone or in the presence of test compounds (100 μM). At the end of incubation, samples were centrifuged at 3,500 × g for 5 min and the pellets washed three times with sodium acetate to remove nondenatured Hb and other proteins. The washed pellets were suspended in Tris-HCl (10 mM) containing sodium dodecyl sulfate (2.5%) and 0.2 N NaOH to solubilize the FP. The amount of denatured Hb was quantified as FP by measuring the absorbance of the solution at 412 nm (Soret band) and subtracting the baseline absorbance at 700 nm.

BHIA.

The quantitative beta-hematin inhibition assay (BHIA) evaluates the ability of a compound to inhibit the formation of β-hematin in vitro. The drug concentration required to inhibit β-hematin formation by 50% (IC₅₀) was determined for each compound, as described previously (31).

RESULTS AND DISCUSSION

To determine whether the compounds under study were able to damage the cell membrane, we examined their ability to lyse human RBC. Similar to the case with CQ, AP4b or AM-1 was not hemolytic per se up to the concentration of 50 μM (not shown), but the two compounds each enhanced the hemolytic effect of FP, with a peak activity at 50 μM (Fig. 2). Under normal conditions, most of the FP released intravascularly is removed by heme-binding proteins such as hemopexin, but in situations of increased hemolysis, as in severe malaria, high plasma levels of heme, up to 20 μM, are thought to be present (25). As previously shown for CQ (11), the bell-shaped dose-response curves indicate that all the compounds at high concentration inhibit their own potentiating effect. For each concentration, the extent of hemolysis induced by the AP4b-FP or AM-1-FP complex was significantly lower than that of CQ-FP. Similar results were obtained when hemolytic activity was assayed at 40°C (50% or 28% hemolysis for AP4b-FP or AM1-FP, respectively, versus 84% for CQ-FP; data not shown). Clinically, hemolysis of normal RBC has not been a problem during CQ treatment, even in patients seriously ill. The fact that our novel compounds are even less hemolytic in vitro than CQ in the presence of FP adds new value to their safety profile.

FIG. 2.

Concentration dependence of the effects of CQ, AM-1, and AP4b on FP-induced hemolysis. RBC suspensions were incubated for 2 h at 37°C with various concentrations of the compounds in the presence of 10 μM FP. Data represent the mean of five experiments run in triplicate.

In subsequent experiments, RBC were preexposed for 2 h to 40 μM FP, washed, and treated for 30 min with the drugs (50 μM) or vice versa, preexposed for 30 min to 50 μM drugs, washed, and incubated for 2 h with 40 μM FP. In the latter cases, the extent of hemolysis was very low and was identical for all the aminoquinolines, suggesting that structural changes leading to hemolysis require a preformed FP-drug complex, allowing a more efficient transfer of hemin into the RBC membrane (19). The lack of correlation between membrane activity (the capability of inducing lysis of RBC) and antimalarial activity (33) indicates that the efficacy of compounds is not simply related to a nonspecific effect of RBC membrane perturbation.

A new fluorimetric method based on the high fluorescence of the quinoline ring (2) allowed the measurement of binding and internalization of the quinolines into RBC. AP4b and AM-1 are cationic amphiphiles, partially present, at neutral pH, in the monoprotonated form. Therefore, they can interact with membranes by both electrostatic and hydrophobic forces, which primarily dictate the affinity of the binding (24). Similar to that of CQ, the uptake of AP4b and AM-1 by normal RBC is rapid, reaching the plateau after 30 min of incubation (not shown). As reported in Fig. 3, the binding of compounds to RBC is higher than that of CQ, dose dependent (panel a), and linearly related to RBC density (panel b). Both compounds readily enter parasitized RBC, as well (panel c). As shown, binding of AP4b and AM-1 to RBC parasitized by the D10 (CQ sensitive) or W2 (CQ susceptible) strain (4% parasitemia) is 2- or 1.5-fold higher, respectively, than that to normal RBC. Compared to results with CQ, binding of AM-1 and AP4b to D10- and W2-parasitized RBC (pRBC) is significantly higher. These findings can be explained on the basis of the physicochemical parameters reported in Table 1. The determined pK_a1 and pK_a2 log P and log D values obtained for CQ are in good agreement with the data previously reported (38). As shown, log P values of the three compounds are very close to each other, while the distribution coefficient at pH 7.4 is quite different, with log D of AP4b > log D of AM-1 ≫ log D of CQ. The lower basicity of the quinolizidine compounds (pK_a1 = 9.11 for AM-1; pK_a1 = 8.87 for AP4b) than of CQ (pK_a1 = 10.21) implies that at physiological pH they are less protonated. These features, attributable to the bicyclic substituent, might improve the cellular permeation of AP4b and AM-1 at physiological pH. This was in fact the case. As shown in Fig. 4a, the amount of AP4b or AM-1 trapped in the normal RBC membrane is higher than that of CQ (about 18 to 19% of AP4b or AM-1 versus 10% of CQ). The quinolizidinyl compounds displayed a more strongly lipophilic character than CQ even at pH 5.2 (which is supposed to reflect the pH of the digestive vacuole) (18) (Table 1). This means that whereas CQ concentrates predominantly in the aqueous regions, AM-1 and, even more, AP4b should concentrate in the lipid regions or at the interface between lipids and water. This may facilitate their interaction with heme molecules during the process of crystallization to HZ, which is reported to occur in vitro at the lipid-water interface or in vivo within the neutral lipid nanosphere (13, 32). In agreement, LAR values were significantly higher for the quinolizidinyl compounds than for CQ. This finding may indicate the ability of the quinolizidinyl compounds to block drug efflux by hydrophobic interaction with PfCRT and explain the lower resistance index and the higher efficacy toward CQ-R strains of P. falciparum (37).

FIG. 3.

(a) Concentration-dependent binding of CQ, AM-1, and AP4b to human RBC (500 μl) at 15% Htc. (b) Uptake of the compounds (50 μM) as a function of the number of RBC. Incubation was performed at 37°C for 30 min. Data represent the means of five experiments run in triplicate. (c) Uptake of the compounds to normal (nRBC) or parasitized RBC (D10, CQ-S; W2, CQ-R). The results are from a representative experiment out of three conducted under the same conditions. Compounds (50 μM) were incubated for 30 min at 37°C with RBC suspensions at 15% Htc (500 μl, final volume). Parasitemia of pRBC was 4,0%. *, P < 0.01 versus results for normal RBC; #, P < 0.01 versus results for CQ with the same strain.

TABLE 1.

Physicochemical characteristics and antimalarial activities of AM1, AP4b, and CQ^a

Compound	pKa1	pKa2	Log P	Log D7.4		Log D5.2 (Calc.)b	VARpH 5.2c	LAR		IC50 (nM) for:		IC90 (nM) for:		RId
				Obs.	Calc.b					CQ-S strain	CQ-R strain	CQ-S strain	CQ-R strain
AM1	9.11	8.33	4.94	2.26	2.26	−2.09	22,645	182e	180f	26.4 ± 15g	41.8 ± 19g	34.3 ± 12	54.9 ± 22	1.59
AP4b	8.87	7.56	4.57	2.69	2.69	−1.48	14,893	490e	485f	24.2 ± 14	21.3 ± 7	35.1 ± 2	33.1 ± 6	0.88
CQ	10.20, 10.18h	8.41, 8.38h	4.73, 4.72h	0.90, 0.96h	0.87	−3.49	22,856	7.9e	7.4f	24.7 ± 15	276.5 ± 149	44.6 ± 3	441.2 ± 41	11.2

^aPK_a, log P, and log D values were calculated as described in Materials and Methods. The experiments were conducted at 25°C. Obs, observed; Calc, calculated; IC₉₀, 90% inhibitory concentration; log D_7.4 and log D_5.2, log D values at pHs 7.4 and 5.2, respectively.

^bCalculated from the following equation: log D = log P − log[1 + 10^{(pKa1 − pH)} + 10^{(pKa1 + pKa2 − 2pH)}].

^cVAR, antilog (calculated log D_7.4 − calculated log D_5.2).

^dRI, resistance index: IC₅₀ for CQ-R strain/IC₅₀ for CQ-S strain.

^eAntilog for observed log D_7.4.

^fAntilog for calculated log D_7.4.

^gData and methods as in reference 33.

^hFrom reference 38.

FIG. 4.

Binding of compounds to RBC (total uptake and amount bound to the membrane). RBC were treated (30 min, 37°C) with 50 μM compounds in the absence (a) or presence (b) of 10 μM FP. *, P < 0.01 versus results for CQ; #, P < 0.05 versus results for FP-CQ. The data are the means of a representative experiment run in triplicate.

Calculated VAR values, as well as their activities against CQ-S strains of P. falciparum, are comparable for the three compounds, in agreement with the known finding that a high VAR in addition to the ability of inhibiting in vitro the formation of BH is important for activity against CQ-S parasites (37).

When drugs and RBC were coincubated in the presence of FP (10 μM), the amount of compound bound to RBC was more than twice that measured in the absence of FP and the quinoline entrapped in the membrane increased for all compounds to about 50% of the total (Fig. 4b). The enhanced binding of the drugs in the presence of FP is in agreement with that reported by Bray et al. (4), confirming that FP-drug interaction could be the main driving force for intracellular quinoline uptake. In addition, although AP4b and AM-1 were trapped in the membrane in larger amounts than CQ, they were more also concentrated in the cytosolic fraction and therefore were potentially available to accumulate within the parasite and exert their antimalarial activities.

Since the RBC uptake of AP4b or AM-1 was higher in the presence of FP, we investigated in vitro the extent of interaction of the new compounds with FP. In mixed aqueous solution (40% DMSO [vol/vol]), FP is expected to be monomeric, as shown by the sharp Soret band at 400 nm (21). It is known that the UV-visible spectrum of FP changes upon addition of CQ, with a decrease in the absorption of the Soret band which is indicative of the formation of CQ-FP complexes (9). The interaction of the aminoquinolines with FP was investigated by monitoring the quenching of the Soret band in 40% DMSO-10 mM sodium phosphate buffer, pH 6.0. As shown in Fig. 5a, the addition of AP4b or AM-1 to FP resulted, at each drug concentration (1 to 200 μM), in a reduction in the Soret absorption of FP. This is indicative of an association between FP and the new quinolines. However, under these experimental conditions, the effect of AM-1 or AP4b on the Soret band is smaller than that of CQ, demonstrating a weaker interaction with FP. About 1.5- to 2.0-fold higher concentrations of these compounds than of CQ were required to produce comparable changes in FP absorption (data not shown). This feature suggests an effect of the quinolizidine ring of AM-1 or AP4b on the 7-chloro-4-aminoquinoline heterocyclic ring, which is thought to be responsible of the formation of π-π interaction with FP (14, 36). However, the stoichiometry of the interaction of AP4b or AM-1 with FP is the same as that for CQ (about 1:1.7), as shown in Fig. 5b. Job's graphs correlate the changes in the absorbance intensity of each compound to the molar fraction drug/FP. A linear correlation is observed between the strength of complexes with FP and the hemolytic activity of the compounds when they are coincubated with FP (Fig. 2), suggesting that the damage of the RBC membrane might be a function of the stability of the FP/compound adducts once they have been inserted into the phospholipid bilayer.

FIG. 5.

(a) Decrease of the Soret band of 15 μM FP in the presence of increasing concentrations of CQ, AM1, and AP4b (1 to 200 μM). (b) Job's plot of compound binding to FP. The total concentration (FP plus compound) was 10 μM in 10 mM sodium phosphate (pH 6.0) containing 40% DMSO. This is a representative experiment out of three made in triplicate with similar results.

Recently it has been proposed that an excess of denatured globin, caused by binding of CQ to FP, could contribute to the antimalarial activity of CQ by inhibiting endosomal maturation and ultimately FP crystallization (17). According to this hypothesis, we measured the capability of AP4b and AM-1 to enhance denaturation of the Hb after incubation with RBC under acidic conditions (similar to those present in maturing endosomes of parasites). The increase in denatured Hb was identical to that induced by CQ (about 40 nmol denatured Hb/h/ml of packed RBC) (data not shown), confirming the capability of AP4b and AM-1 of interacting with FP.

Last, both AM-1 and AP4b were able to inhibit the formation of BH in a dose-dependent manner when tested with the spectrophotometric BHIA developed by our group (31). The calculated IC₅₀s for AM-1 and AP4b were 1.52 and 1.12 drug molar equivalents to hemin, respectively, and were comparable to that of CQ (1.65) (Fig. 6). These data confirm that the ability to inhibit BH formation is not directly related to the strength of association of the drugs with FP, as clearly demonstrated by Egan et al. (14, 22). Both the ability to form a complex with hematin and that to inhibit BH formation are prerequisites for strong antiplasmodial activity. However, a direct correlation cannot be established unless the data are corrected for lipophilicity and pK_a of the tested compounds (12).

FIG. 6.

Inhibitory effects of CQ, AM1, and AP4b on BH formation in vitro (BHIA). This is a representative experiment out of four made in triplicate with similar results.

The fact that AM-1 and AP4b can overcome CQ resistance may be related to their higher level of accumulation inside RBC infected by the resistant strains and/or a reduced affinity for PfCRT.

In conclusion, we have shown that the two novel 4-aminoquinolines, AM-1 and AP4a, which show in vitro strong activity against both CQ-S and CQ-R parasites (33), interact with hematin and inhibit BH formation in vitro similarly to CQ. Therefore, their antimalarial activities are likely to be due to inhibition of HZ formation and buildup of toxic heme molecules, as hypothesized for the other members of the 4-aminoquinoline family.

Both compounds bind quite well to normal RBC and pRBC and are easily internalized as a function of their pK_a and lipophilicity, but they are not hemolytic at therapeutic doses. They are less hemolytic than CQ even when hemolysis is accelerated (induced) by coincubation with FP.

Therefore, AM-1 and AP4b seem to have an interesting safety profile for normal cells, which combined with their in vitro potency against P. falciparum parasites and in vivo oral bioavailability makes them good lead candidates for new antimalarials.