In Vivo and In Vitro Antimalarial Properties of Azithromycin-Chloroquine Combinations That Include the Resistance Reversal Agent Amlodipine

Abstract

Evidence of emerging Plasmodium falciparum resistance to artemisinin-based combination therapies, documented in western Cambodia, underscores the continuing need to identify new antimalarial combinations. Given recent reports of the resurgence of chloroquine-sensitive P. falciparum parasites in Malawi, after the enforced and prolonged withdrawal of this drug, and indications of a possible synergistic interaction with the macrolide azithromycin, we sought to further characterize chloroquine-azithromycin combinations for their in vitro and in vivo antimalarial properties. In vitro 96-h susceptibility testing of chloroquine-azithromycin combinations showed mostly additive interactions against freshly cultured P. falciparum field isolates obtained from Mali. Some evidence of synergy, however, was apparent at the fractional 90% inhibitory concentration level. Additional in vitro testing highlighted the resistance reversal properties of amlodipine for both chloroquine and quinine. In vivo experiments, using the Peters 4-day suppressive test in a P. yoelii mouse model, revealed up to 99.9% suppression of parasitemia following treatment with chloroquine-azithromycin plus the R enantiomer of amlodipine. This enantiomer was chosen because it does not manifest the cardiac toxicities observed with the racemic mixture. Pharmacokinetic/pharmacodynamic analyses in this rodent model and subsequent extrapolation to a 65-kg adult led to the estimation that 1.8 g daily of R-amlodipine would be required to achieve similar efficacy in humans, for whom this is likely an unsafe dose. While these data discount amlodipine as an additional partner for chloroquine-based combination therapy, our studies continue to support azithromycin as a safe and effective addition to antimalarial combination therapies.

INTRODUCTION

In the last decade, artemisinin-based combination therapies (ACTs) have become widely adopted as first-line treatment in almost all countries where malaria is endemic (90). These drug combinations display excellent clinical efficacy against Plasmodium falciparum infection, yet recent studies from western Cambodia report decreases in parasite clearance rates following artesunate monotherapy or artesunate-mefloquine combination therapy (18, 59, 60, 76, 92). The possible emergence of resistance underscores a clear need to find alternative regimens. While several promising agents are in the pipeline, the development of antimalarial drugs that are effective, well tolerated, and safe remains a very challenging task (6, 27, 77, 89). In addition, the new paradigm of antimalarial therapies based on the combination of drugs that have additive or preferably synergistic properties raises the threshold for drug discovery even further.

Chloroquine (CQ) was the most important drug for the treatment of malaria for many decades, until widespread resistance led to its replacement, most recently by ACTs (21, 93). Recent data from Malawi and Kenya have shown that in those regions, the removal of CQ from local use has led to the resurgence of CQ-sensitive (CQS) strains that presumably are not subject to the negative fitness cost imparted by mutant P. falciparum CRT (PfCRT)-mediated CQ resistance (37, 50, 53). It is thought that in Malawi, the attrition of CQ-resistant (CQR) P. falciparum parasites has been faster and more pronounced than in Kenya due to Malawi's more effective removal of CQ from widespread use. These findings suggest that CQ might once again become a useful antimalarial drug, if used in combination, in areas where CQ resistance has waned.

Similar to CQ, the macrolide azithromycin (AZ) has a long track record of tolerability and safety in both children and pregnant women, the populations most affected by malaria (19, 20, 34). As a widely used antibiotic, AZ has a very broad spectrum of activity against many important bacteria, including Streptococcus, Staphylococcus, Chlamydia, and Neisseria, and is a first-line indication for the treatment of common infections, including community-acquired pneumonia, exacerbation of chronic bronchitis, pharyngitis, sinusitis, and several sexually transmitted diseases. Furthermore, AZ has excellent pharmacological properties, including good bioavailability (up to 40% with or without food), wide distribution in the body, high intracellular concentration with subsequent slow release into the bloodstream, a long half-life (t_1/2, ∼68 h), and simple dosing regimens that do not need to be adjusted in cases of hepatic or renal dysfunction (32, 74).

In vitro studies have shown that AZ has reasonably good activity against P. falciparum asexual blood-stage parasites, including CQR strains (7, 30, 58, 63). This was confirmed in animal studies against CQS and CQR strains of rodent Plasmodium species (1, 31, 73). In humans, evidence of AZ antimalarial activity has been shown in P. falciparum prophylactic studies (2, 46) as well as in a large randomized trial of AZ monotherapy in patients infected with P. vivax in India (20). In vitro studies have revealed that AZ antimalarial activity requires parasite exposure to drug for at least two cycles of intraerythrocytic development (one cycle of invasion, development, and egress lasts 42 to 48 h) for the drug to exert its maximum antiparasitic killing effect. This is similar to the activities of other antibiotic drugs active against Plasmodium, such as tetracycline, doxycycline, and clindamycin (15, 33, 94). This “delayed death” effect indicates that AZ will likely need to be combined with a faster-acting drug for maximum therapeutic effect (57).

In this light, the combination of CQ with AZ is an interesting possibility. Early studies with field isolates of P. falciparum reported AZ-CQ effects that were additive to synergistic (54, 63). However, more recent studies with long-term culture-adapted parasite lines observed only additive interactions between these two drugs (81). These disparate results raised the possibility that the findings of studies with long-term lab-adapted P. falciparum lines may not accurately reflect possible synergy between CQ and AZ. Interestingly, human trials using AZ and CQ in combination against P. falciparum in India revealed cure rates of 97%, a value that was far higher than that achieved with either agent used as monotherapy and that exceeded expectations based on the local prevalence of CQR parasites (19). In subsequent randomized clinical trials conducted in India, Latin America, and sub-Saharan Africa, the coadministration of AZ and CQ demonstrated over 95% efficacy in the treatment of adults with symptomatic, uncomplicated falciparum malaria (unpublished data presented at the American Society of Tropical Medicine and Hygiene Annual Meetings in 2007 and 2008).

Despite these promising results with AZ-CQ combinations, widespread resistance to CQ and the potential to leave AZ as the sole effective antimalarial agent represent threats to effective malarial control. One strategy to improve efficacy would be to add an agent capable of reversing the CQ resistance phenotype (43, 51). Reversal agents are postulated to act by preventing mutant PfCRT, the primary mediator of CQ resistance, from effluxing CQ from its site of action in the digestive vacuole (48). Calcium channel blockers such as verapamil (VP) and amlodipine (AMLO) are among the best-studied CQ resistance reversal agents (4, 17, 43, 51). Given the large amounts of drug required for reversal, however, concerns with toxicity have prevented these agents from being pursued clinically in combination with CQ (22). Nevertheless, in the case of AMLO, its chiral nature points to a possible solution for reducing host toxicity. Clinically, AMLO is used to lower blood pressure, by virtue of the calcium channel-blocking activity that is considerably more pronounced in the S than in the R enantiomer (49, 67, 95), yet both enantiomers are equally potent CQ resistance reversal agents both in vitro and in rodent malaria models (4, 17).

Here, we have sought to investigate whether the addition of R-AMLO might increase the antimalarial activity of AZ-CQ. In vitro drug assays were used to measure the interactions of CQ and AZ in fresh field isolates from Mali, as well as to measure the potentiation of CQ by R-AMLO or other CQ resistance reversal agents, when they were assayed against a panel of well-characterized, culture-adapted lines. Interactions of CQ, AZ, and R-AMLO were also evaluated in a P. yoelii rodent malaria model, including pharmacokinetic/pharmacodynamic analyses. R-AMLO drug levels necessary for a curative response in humans were then extrapolated on the basis of previously published data. The results support a synergistic interaction between CQ and AZ and indicate that R-AMLO provides a therapeutic benefit to AZ-CQ combinations, though at concentrations that are likely to be too high for clinical use.

MATERIALS AND METHODS

Drugs.

Chlorpheniramine, CQ, cyproheptadine, desipramine, diltiazem, and VP were purchased from Sigma or Winthrop-Breon. AZ, [³H]AZ, AMLO, and R-AMLO were kindly supplied by Pfizer, Inc. (New York, NY).

Parasite lines.

The long-term-culture-adapted P. falciparum 3D7, Dd2, and 7G8 parasite lines and the P. yoelii nigeriensis N67 line were obtained from the Malaria Research and Reference Reagent Resource Center (MR4; ATCC, Manassas, VA). P. falciparum field isolates QK025, QK028, QK030, QK031, and QK040 were obtained from infected patients in Mali and were culture adapted at Columbia University for at most 5 weeks. P. falciparum parasites were propagated in leukocyte-free human red blood cells (RBCs) at 4% hematocrit in malaria culture medium (RPMI 1640 with l-glutamine [Invitrogen], 50 mg/liter hypoxanthine, 10 mg/liter gentamicin, 25 mM HEPES, 0.225% NaHCO₃, and either 0.5% lipid-rich bovine serum albumin [Albumax I; for 3D7, Dd2, and 7G8] or 10% human serum [for Malian isolates]). Parasites were propagated at 37°C in tissue culture plates kept in sealable modular incubator chambers (Billups-Rothernberg Inc.) that were gassed with 5% CO₂-5% O₂-90% N₂.

Female CD-1 mice were purchased from Charles River Laboratories (Wilmington, MA). These were fed ad libitum with water and irradiated standard pellet diet. The Columbia University Institutional Animal Care and Use Committee approved all protocols for animal experimentation.

In vitro chloroquine-azithromycin drug interactions.

CQ and AZ drug interactions were assessed using 96-h [³H]hypoxanthine incorporation assays, as described previously (81). In these assays, each drug was tested alone or in combination at fixed ratios of its previously determined 50% inhibitory concentration (IC₅₀) (CQ:AZ ratios of 0:1, 1:3, 1:5, 1:1, 3:1, 5:1, and 1:0) in 96-well plates with an initial parasitemia of 0.2% synchronized ring-stage parasites. Medium was replaced after 48 and 72 h, and [³H]hypoxanthine (0.5 μCi/well) was added at the 72-h time point. Medium replacements contained appropriate amounts of reagents to ensure that drug concentrations remained constant in each well throughout the assay. Fractional IC₅₀ (FIC₅₀) and FIC₉₀ values were derived by curve fitting the data from at least two 96-h assays and were used to plot isobolograms (5).

Azithromycin uptake assays.

RBCs infected with trophozoite-stage P. falciparum Dd2 parasites were enriched to >90% purity following retention on MACS columns placed in a SuperMACS separator, which purifies parasitized RBCs on the basis of the presence of paramagnetic parasite-produced hemozoin (87). Infected RBCs were incubated for 30 min in malaria parasite culture medium in the presence of 20 mM glucose and 1 μM [³H]AZ. Cells were separated from the supernatant containing the free radiolabel and lysed overnight in an ethanol-tissue solubilizer solution to release the incorporated drug, as described previously (79). Uninfected RBC preparations were included as a control. Total counts from the pellet and supernatant fractions were calculated as a function of volume, in order to calculate the ratio of intracellular to extracellular [³H]AZ.

In vitro investigations of chloroquine resistance reversal agents.

The in vitro CQ resistance reversal properties of VP, diltiazem, and AMLO (either the racemic mixture or the R enantiomer), reported to have activity against calcium channel blockers in mammalian cells, were studied against the standard culture-adapted P. falciparum lines 3D7, Dd2, and 7G8. In parallel, we evaluated the non-calcium channel blocker reversal agents chlorpheniramine, cyproheptadine, and desipramine (8, 22, 47, 55, 69, 83). Drug responses were calculated by adding fixed amounts of these agents to CQ in 72-h [³H]hypoxanthine incorporation assays, as described previously (26). Concentrations were selected to be lower than the IC₅₀s of these reversal agents when they were tested alone (4, 8, 17, 47, 51, 69). The IC₅₀ of CQ in the absence or presence of each reversal agent was calculated via nonlinear (curve-fitting) regression analysis of dose-response curves.

To further characterize the different reversal properties of each agent, we also calculated the response modification index (RMI) using the formula IC₅₀(A + B)/IC₅₀(A), where A is the antimalarial drug and B is the resistance-reversing drug. For purposes of interpretation, when RMI is ≈1, there is no change in antimalarial activity, whereas an RMI of ≪1 represents potentiation of antimalarial activity (i.e., evidence of synergy) and an RMI of ≫1 represents antagonism (62).

In vivo drug interactions of chloroquine, azithromycin, and R-amlodipine.

CQ, AZ, and R-AMLO, tested alone and in various combinations, were evaluated for in vivo efficacy against P. yoelii nigeriensis N67-infected CD-1 mice, using the Peters 4-day suppressive test (68, 83). Briefly, 6- to 8-week-old, female CD-1 mice were inoculated on day 0 via intraperitoneal injections with 10⁷ P. yoelii-infected RBCs. Beginning 2 h after inoculation, once-daily treatments with drug were administered for four consecutive days. Drug concentrations were calculated in mg/kg of body weight and adjusted so that aliquots of 0.2 ml would contain the desired dose. CQ and AZ were given subcutaneously, while R-AMLO was administered via oral gavage. We also included a control group with drug vehicle alone. Day 4 parasitemias in all mice were measured via Giemsa-stained thin smears, and the percent inhibition for each treatment group was calculated according to the formula [(A − B)/A] × 100, where A is the day 4 average parasitemia for the control group and B is the day 4 average parasitemia for the treatment group. All data were aggregated from groups of three to five mice in 1 to 3 separate trials.

Pharmacologic analysis of R-AMLO in a mouse model.

Pharmacokinetic and pharmacodynamic properties of R-AMLO were investigated using the P. yoelii model of infection, as described above. In this experiment, five groups of three CD-1 mice were treated with 40 mg/kg R-AMLO, 1.5 mg/kg CQ, and 10 mg/kg AZ. The effect of drug treatments on parasitemias was measured as described above. Mouse blood was serially collected at 0, 3, 6, 12, and 20 h posttreatment over 4 days, using a sampling regimen that ensured that less than 200 μl total blood was collected per mouse. Whole blood was separated into plasma and RBCs via centrifugation and stored at −80°C. For pharmacokinetic analyses, samples were thawed and transferred (in 20-μl aliquots) to a 96-well Marsh tube box, precipitated with 100 μl of internal standard (200 ng/ml; a dimethyl analog of AMLO), resuspended for 30 s, and centrifuged for 10 min. Supernatants (80-μl aliquots) were transferred to a 96-well deep-well block containing 80 μl of acidified water (0.1% formic acid) and resuspended.

Concentrations of R-AMLO were determined using liquid chromatography-tandem mass spectrometry (LC-MS/MS). For this, 10-μl samples were injected onto a Phenomenex monolithic C₁₈ column (5 μm, 25 by 4.6 mm). The column was equilibrated with mobile phase (0.1% formic acid in water [solution A] and 0.1% formic acid in acetonitrile [solution B]) at a flow rate of 1.70 ml/min. The gradient was started at 2% B, increased to 98% B from 0.5 to 1 min, and then returned to starting conditions by 1.5 min and held for an additional 0.7 min, for a total run time of 2.2 min. Under these conditions the retention times for R-AMLO and internal standard were 1.15 and 1.14 min, respectively. The effluent was analyzed by a mass spectrometer detector (API-4000; AB Sciex) fitted with a turbo ion-spray interface and operated in positive-ion mode with a declustering potential of 51 V, a temperature of 550°C, and a collision energy of 15 V. R-AMLO and internal standard were monitored by multiple reaction monitoring at transitions of 409.3/238.0 and 395.4/237.9, respectively (29). The dynamic range of the assay was 0.001 to 10 μg/ml for both plasma and RBCs. Pharmacokinetic parameters of R-AMLO were calculated using WinNonlin (version 5.2) software (Pharsight).

R-AMLO protein binding.

R-AMLO protein binding was determined in both mouse and human plasma utilizing a 96-well Teflon equilibrium dialysis unit. Plasma and buffer, loaded with R-AMLO at two different concentrations (19 and 190 μM; n = 6 replicates per concentration), were added to opposing chambers of each well and incubated for 4 h at 37°C. Protein binding was calculated as the concentration in the buffer side divided by the total concentration in the plasma side, subtracted from 100.

Statistical analysis.

For all experiments described above, statistical significance was determined using the nonparametric Mann-Whitney U test, calculated with GraphPad Prism (version 4) software.

RESULTS

In vitro synergy testing of azithromycin-chloroquine combinations in freshly cultured P. falciparum field isolates from Mali.

Given the conflicting results of previous in vitro drug interaction studies of AZ-CQ (54, 63, 81), we investigated the hypothesis that CQ and AZ might display synergy against field isolates that could not be observed in long-term, culture-adapted lines. To achieve this, we obtained fresh field isolates from Mali, where CQ resistance was widespread at the time of collection. These isolates (QK025, QK028, QK030, QK031, and QK040) were propagated for a maximum of 5 weeks before synergy testing was completed. Table 1 shows drug susceptibilities for both CQ and AZ in 96-h assays, where AZ exerts its optimal antimalarial activity via the delayed death mechanism (81). All five Malian lines were found to be CQR in vitro, with IC₅₀s in the range of 121 to 173 nM, exceeding the typical in vitro CQ resistance cutoff of 80 to 100 nM (23). By comparison, the CQS 3D7 line produced a mean CQ IC₅₀ of 24 nM. In these 96-h assays, AZ IC₅₀s were in the range of 136 to 406 nM. These values are similar to our previous measurements obtained upon exposure of P. falciparum culture-adapted lines to AZ for 96 h (range, 103 to 561 nM; in that study, AZ IC₅₀s after 48 h, i.e., a single generation, were in the range of 3.5 to 22.6 μM, reflecting 30- to 230-fold gains in potency after two generations [81]).

Table 1.

Susceptibility of freshly adapted P. falciparum isolates to chloroquine and azithromycin after 96 h of incubation^a

Isolate	CQ		AZ
	IC50	IC90	IC50	IC90
QK025	148 ± 25	266 ± 38	211 ± 62	2,173 ± 157
QK028	129 ± 17	241 ± 23	314 ± 55	2,820 ± 108
QK030	147 ± 22	256 ± 28	406 ± 52	2,363 ± 264
QK031	167 ± 37	276 ± 43	156 ± 60	2,036 ± 213
QK040	121 ± 21	216 ± 27	163 ± 57	2,184 ± 176
Dd2	173 ± 13	256 ± 15	136 ± 25	562 ± 153
3D7	24 ± 4	31 ± 6	170 ± 41	2,744 ± 313

^aValues were calculated from five separate assays performed in duplicate and are expressed in nM as arithmetic means ± SEMs.

Of note, in our present assays, IC₉₀s for AZ were high, in the range of 2.0 to 2.8 μM for the Malian isolates and the CQS line 3D7, with a lower value of 0.6 μM observed for the reference line Dd2 (Table 1). These data reveal that AZ alone is considerably less effective than CQ in fully suppressing parasite growth, even at high concentrations, in some parasite backgrounds. Other studies in our laboratory support this observation that some lines display relatively low IC₉₀/IC₅₀ ratios (<5 for strains Dd2, K1, and FCR3), while others show high IC₉₀/IC₅₀ ratios (in the range of 10 to 40 for 3D7, GC03, and 7G8). The mechanistic basis for this difference in susceptibility to high concentrations of AZ has not been elucidated.

Figure 1 shows the FIC₅₀ and FIC₉₀ isobologram profiles of the 96-h AZ-CQ combination assays for these five isolates. With the FIC₅₀ data, the 1:1 ratio combination repeatedly provided evidence of synergistic interactions, with sum FIC₅₀ values in the range of 0.44 to 0.53 (see Table S1 in the supplemental material). Meanwhile, sum FIC₅₀ values for AZ-CQ at 1:5, 1:3, 3:1, and 5:1 ratios were more variable, ranging from 0.67 to 1.21, indicating an essentially additive response.

Fig. 1.

FIC₅₀ and FIC₉₀ isobolograms of AZ-CQ against five freshly cultured CQR P. falciparum isolates from Mali tested in a 96-h assay. Assays were performed using 5 CQ-AZ combinations on the basis of relative ratios of their IC₅₀s: 5:1, 3:1, 1:1, 1:3, and 1:5. As controls, CQ and AZ were tested alone to define their respective IC₅₀ and IC₉₀ values and derive FIC₅₀ and FIC₉₀ values in the combinations. The line of additivity is shown in light gray. Combinations were tested on two separate occasions, and FIC₅₀ (A) or FIC₉₀ (B) values for each occasion and each ratio are indicated with a line connecting the two separate values. Results show an essentially additive set of interactions at the FIC₅₀ level, except for the 1:1 combinations that displayed evidence of synergy. At the IC₉₀ level, all drug combinations provided evidence of some synergy. For each panel, sum FIC values are listed as means ± SDs (values are provided in Table S1 in the supplemental material).

With the FIC₉₀ values, their mean sums for each clinical isolate with all drug combinations were always less than 1. At the 1:1 ratio, the combination of AZ-CQ produced sum FIC₉₀ values in the range of 0.33 to 0.42 for these parasite lines, consistent with synergy (see Table S1 in the supplemental material).

When the sum FIC values were calculated across all drug dilutions, the means ± standard deviations (SDs) for the field isolates were in the ranges of 0.83 ± 0.22 to 0.92 ± 0.29 for the sum FIC₅₀ values and 0.64 ± 0.24 to 0.75 ± 0.26 for the sum FIC₉₀ values (Fig. 1; see Table S1 in the supplemental material). These values were consistent with a mild degree of synergy manifesting at the FIC₉₀ level.

In vitro azithromycin uptake in P. falciparum asexual blood-stage parasites.

Prior studies have documented that CQ concentrates up to millimolar levels in infected RBCs, by virtue of the weak base properties of the drug and its binding to heme products, both of which drive its accumulation in the acidic digestive vacuole (10, 78). Similar studies on whether AZ also concentrates in P. falciparum-infected RBCs have, to our knowledge, not been previously reported. To investigate this important parameter of antimalarial drug action, we measured the accumulation of tritium-labeled AZ in both uninfected and infected RBCs. This was measured with the P. falciparum Dd2 strain and used cells exposed to AZ for 30 min. While [³H]AZ was observed to be equally distributed between uninfected RBCs and extracellular medium, the accumulation of [³H]AZ in parasitized RBCs was 3.9 ± 0.8 (mean ± standard error of the mean [SEM]) times higher than that in extracellular medium, as calculated from three independent experiments conducted in duplicate (data not shown).

In vitro testing of chloroquine combined with chloroquine resistance reversal agents against P. falciparum.

A number of studies have raised the possibility that CQ could be used in combination with a CQ resistance reversal agent as a means to more effectively treat CQR parasites (12, 22, 41). To further investigate this, we searched for compounds that could potentially be incorporated into CQ-AZ combinations. For this study, we combined CQ with fixed amounts of several reversal agents against both CQS (3D7) and CQR (Dd2, 7G8) lab-adapted lines. The fixed amounts of CQ resistance reversal agents used in these experiments were 0.8 μM verapamil, 2 μM diltiazem, 0.6 μM chlorpheniramine, 0.9 μM cyproheptadine, and 0.6 μM desipramine. For the AMLO assays, we used 2.5 μM AMLO (containing equal amounts of the R and S enantiomers) as well as 1.25 and 2.5 μM R-AMLO. These concentrations are below the IC₅₀ of AMLO, which has been observed to be in the range of 5 to 12 μM (4, 17).

Figure 2 shows the RMI (defined in Materials and Methods) for the above compounds tested in combination with CQ (values are presented in Table S2 in the supplemental material). This index measures the extent to which the reversal agent modifies the CQ IC₅₀. As expected, the RMI in the CQS line 3D7 for all reversal agents was close to or above 1. In contrast, in the CQR lines Dd2 and 7G8, CQ was potentiated to various degrees by the calcium channel blockers (VP, AMLO, diltiazem). VP, a model CQ resistance reversal agent (43, 51), strongly potentiated CQ against Dd2 (RMI, 0.20 ± 0.02) but only weakly so against 7G8 (RMI, 0.75 ± 0.07). This was comparable to previously observed values and has been attributed to sequence differences in the CQ resistance transporter PfCRT (52, 82, 88). On the other hand, diltiazem displayed an RMI of 0.53 ± 0.10 against Dd2 and showed no potentiation against 7G8 (RMI, 0.97 ± 0.23). Against Dd2, both racemic AMLO and the enantiomer R-AMLO (at 2.5 μM and 1.25 μM, respectively) strongly potentiated CQ (0.16 ± 0.02 and 0.30 ± 0.17, respectively). Use of 2.5 μM R-AMLO did not significantly improve the RMI (0.27 ± 0.11) against Dd2. Similarly, both AMLO and 1.25 μM R-AMLO strongly reversed the CQ resistance of 7G8 (0.34 ± 0.07 and 0.38 ± 0.16, respectively). Meanwhile, the higher concentration of R-AMLO had an RMI of 0.26 ± 0.08 against 7G8. Thus, of the calcium channel blockers studied, both AMLO and R-AMLO had the most pronounced effect, especially against 7G8.

Fig. 2.

Response modification index of various CQ resistance reversal agents against standard culture-adapted P. falciparum lines. An index of 1 implies no reversal, whereas low values imply substantial reversal. All reversal agents were tested in the presence of CQ. Values were determined from three independent experiments performed in duplicate and are presented as means ± SEMs (values are presented in Table S2 in the supplemental material). VP, verapamil; AMLO, amlodipine; R-AMLO, R enantiomer of AMLO; DILT, diltiazem; DES, desipramine; CHLOR, chlorpheniramine; CYP, cyproheptadine.

In parallel, we compared the CQ resistance reversal properties of the non-calcium channel blockers desipramine, chlorpheniramine, and cyproheptadine against the same strains. Overall, all three agents displayed good potentiation of CQ against both Dd2 and 7G8 (see Table S2 in the supplemental material). Based on these results, we chose to further investigate R-AMLO because of the combination of effective CQ resistance reversal and the potential for safe dosing with the R enantiomer (14).

In vivo chloroquine, azithromycin, and R-amlodipine interactions.

To study the in vivo interactions of CQ, AZ, and R-AMLO, we used a model of Plasmodium yoelii infection and measured drug efficacy using the Peters 4-day suppressive test (68). For these studies, CD-1 mice were infected on day 0 and treated once daily for 4 days. Parasitemias were then measured on day 4 and compared to those of placebo-treated controls. Efficacy studies were performed on one to three separate occasions (corresponding to 5 to 15 mice per treatment group; see Table S3 in the supplemental material). As seen in Fig. 3, R-AMLO alone suppressed day 4 parasitemias by 36.6% ± 9.6% (mean ± SEM) at doses up to 40 mg/kg. This was similar to the suppressive activity of CQ (40.6% ± 4.9%) when it was administered at 1.5 mg/kg. Together with 20 mg/kg R-AMLO, CQ efficacy increased to 71.5% ± 5.9%, comparable to CQ activity in combination with 2.5 mg/kg AZ (74.8% ± 4.8%). When AZ was combined with 20 mg/kg R-AMLO, the suppressive activity of AZ increased from 49.3% ± 5.6% to 84.2% ± 3.6%. The addition of 20 mg/kg R-AMLO to AZ-CQ (at subcurative doses of 2.5 mg/kg and 1.5 mg/kg, respectively) improved the suppression to 95.0% ± 1.1%, while 40 mg/kg R-AMLO added to AZ-CQ (2.5 mg/kg and 10 mg/kg, respectively) produced 99.9% ± 0.1% suppression (see Table S3 in the supplemental material). These results point to an additive effect of AMLO combined with AZ or CQ in this in vivo rodent malarial model. There were no apparent toxicities to the mice undergoing any of these treatments, including with the R-AMLO high-dose regimen.

Fig. 3.

Day 4 suppressive activity of CQ, AZ, and R-AMLO in CD-1 mice infected with P. yoelii nigeriensis N67. Drug concentrations are in mg/kg, abbreviated MKG. Data were obtained from 5 to 15 mice per group and are presented as mean + SEM suppression (values are listed in Table S3 in the supplemental material).

Pharmacokinetic/pharmacodynamic measurements of R-amlodipine in a mouse model.

To investigate the pharmacological properties of R-AMLO in mice, we measured drug concentrations in RBCs and plasma. These were collected from CD-1 mice infected with P. yoelii and treated with 4 days of 1.5 mg/kg CQ, 10 mg/kg AZ, and 40 mg/kg R-AMLO, which had shown 99.9% suppression in parasitemias (Fig. 3; see Table S3 in the supplemental material). There were no complications to the mice from this procedure.

Pharmacokinetic parameters of R-AMLO are shown in Table 2. In mouse plasma, the concentration of R-AMLO steadily increased during the experiment, reaching a 3-fold accumulation on day 4 in relation to day 1 (area under the concentration-time curve [AUC] from 3 to 24 h [AUC_3-24] = 6.7 μg·h/ml versus AUC_72-96 = 21.7 μg·h/ml). Plasma steady-state levels were achieved between days 3 and 4 due to the minimal increase in exposure (20.4 versus 21.7 μg·h/ml) during that time and the ∼3 to 4 half-lives that had lapsed (day 4 plasma t_1/2 ≈ 9.8 h). Overall, R-AMLO showed a 3-fold greater partitioning into RBCs than plasma (205.2 versus 62.6 μg·h/ml, respectively) at the end of the 4-day period. A concentration-versus-time plot can be seen in Fig. 4.

Table 2.

Pharmacokinetic parameters of R-amlodipine

Compartment	Area under the curve (μg·h/ml)					Cmaxa (μg/ml)
	Total	3–24 h	24–48 h	48–72 h	72–96 h
Plasma	62.6	6.7	13.5	20.4	21.7	1.4
Red blood cells	205.2	15.7	42.1	46.0	101.5	5.0

^aC_max, maximum concentration of drug.

Fig. 4.

Curves of plasma and RBC concentrations as a function of time, obtained from mice administered R-AMLO orally at time points 0, 24, 48, and 72 h. Values are presented as means ± SDs from the groups of three mice sampled at each time point.

Extrapolation of mouse pharmacokinetic/pharmacodynamic parameters to humans.

To extrapolate the above parameters into a human dose prediction, we set out to quantify R-AMLO protein binding in mouse and human plasma. Percent recovery from the buffer controls was at least 89%. Protein binding of R-AMLO was high in both human and mouse plasma and appeared to be concentration dependent over the 10-fold range. The average of the two different concentrations yielded values of 96.4% and 98.0% (mouse and human, respectively) for the measured protein binding of R-AMLO. These results are in alignment with previously reported S-AMLO protein binding values as well as human data on AMLO ([84]; Norvasc package insert).

Given that R-AMLO is 96.4% protein bound in mice, the day 4 free AUC (fAUC_72–96) can be calculated to be 0.78 μg·h/ml. This can subsequently be used as the concentration of free drug needed to achieve a greater than 99% cure rate in mice after 4 days of treatment. Thus, assuming dose-proportional pharmacokinetics, an estimate of the R-AMLO dose needed to achieve comparable free drug concentrations in humans on day 4 can be calculated. On the basis of previously published human pharmacokinetic data on racemic AMLO (repeated oral administration of 15 mg once daily for 14 days) (25) and with adjustment for the contribution of the R enantiomer (50%), the day 14 human free AUC was reported to be 3.26 × 10⁻³ μg·h/ml. Comparing the ratio of mouse free AUC to that reported for a human, one would thus require a predicted 1.8 g of R-AMLO daily for 4 days to achieve similar free drug exposure. Given the site of parasitic infection and comparable percent hematocrit level in mouse and human (mouse, 36 to 48%; human, 39 to 43%), the data provide a fair degree of confidence in translating the concentration-response observed in mice to that in humans.

DISCUSSION

Drug resistance continues to be a major problem in the treatment of malaria throughout the world, leading to the adoption of combination therapy as the new standard of care. As seen in HIV, tuberculosis, and other infections, combination therapy in malaria can increase efficacy, decrease toxicity, and extend the life of individual drugs by delaying or preventing the development of resistance (21). In the present study, we have further explored the efficacy of AZ-CQ combinations and evaluated the possibility of adding the CQ resistance reversal agent amlodipine, using in vitro and in vivo systems.

Given the discrepant results regarding in vitro interactions between CQ and AZ (54, 63, 81), we chose freshly obtained field isolates out of concern that long-term-culture-adapted parasite lines may display inconsistent susceptibilities, especially in terms of synergy testing. We also extended the duration of the assay to 96 h in order to maximize the delayed death phenotype observed in P. falciparum parasites exposed in vitro to AZ, which manifests as potent inhibition of the progeny of drug-treated parasites (63, 81). This design differed from that of an earlier study (63) that measured CQ-AZ in vitro interactions after 68 h of drug incubation, at which time AZ showed IC₅₀s that were in the 0.6 to 10.2 μM range presumably because drug killing was not yet maximal. Performing 96-h assays can be predicted to affect in vitro drug interactions by maximizing the efficacy of AZ action. Our studies provided evidence of some synergy between CQ and AZ at the FIC₉₀ level with most of the ratios, while the effect was more often additive at the FIC₅₀ level. Synergy at the FIC₉₀ level was typically greater as the ratio increased in favor of AZ, suggesting that the high concentrations of AZ caused these CQR isolates to become more susceptible to CQ. Of note, the greatest evidence for synergy, at both FIC₅₀ and FIC₉₀ levels, was consistently observed at the 1:1 ratio. This suggests that dosing each component of the AZ-CQ combination at a concentration that affords equivalent potency is ideal in achieving the greatest degree of antimalarial activity.

One interpretation of these data could be that high concentrations of the weak base AZ might cause this drug to accumulate in the acidic digestive vacuole of the intraerythrocytic parasite, akin to known accumulation of AZ in acidic lysosomes of phagocytic cells (74). Intravacuolar accumulation could occur to such an extent that it indirectly increases the potency of CQ, perhaps through an effect on pH and/or the kinetics of hemoglobin degradation. CQ, also a weak base, is known to accumulate in the digestive vacuole as a diprotonated species, as predicted by the Henderson-Hasselbalch equation, and is thought to exert its antimalarial activity by binding heme species and interfering with their parasite-mediated detoxification (44, 66, 85). Further studies on a possible effect of high AZ concentrations on vacuolar pH could benefit from the recent development of methods to measure compartmental pH in parasites transformed with the ratiometric fluorescent protein pHluorin (45).

It is, nevertheless, likely that the primary mode of antimalarial action of AZ is to bind to the cyanobacterium-like large subunit rRNA in the parasite's apicoplast and inhibit protein translation in this organelle (81). This inhibition is thought to result in delayed death by preventing the synthesis of proteins that are essential for the normal development of the progeny of drug-treated parasites. This proposed primary mode of antimalarial action is supported by the recent finding of a mutation in the 50S ribosomal large subunit sequence in AZ-pressured P. falciparum parasites that had acquired resistance and is consistent with how this macrolide works on the 50S ribosomal subunit in bacteria (72, 81).

While AZ-CQ combinations have proven quite effective clinically in treating P. falciparum or P. vivax malaria (13), relatively high doses of AZ are required and preexisting resistance to CQ poses a problem that potentially could be minimized by the addition of a CQ resistance reversal agent. Our results indicate that the reversal agent AMLO has greater potency in reversing the CQ resistance phenotype than either VP or diltiazem and AMLO is comparable to other agents such as chlorpheniramine (Fig. 2; see Table S2 in the supplemental material). This was true not only for the highly CQR line Dd2 (harboring a PfCRT haplotype that is associated with high-level VP reversibility) but also for 7G8, a less resistant line that harbors a mutant PfCRT haplotype associated with reduced VP reversibility (52, 82). Importantly, R-AMLO was equally effective as a reversal agent against both CQR lines. We note that the 2.5 μM concentration tested herein displayed some intrinsic activity against Dd2 but not against 3D7 (mean ± SEM inhibition, 35% ± 10% and 12% ± 5%, respectively; n = 4), consistent with earlier estimates of AMLO IC₅₀s of 5 to 7 μM for CQR lines and 12 μM for a CQS line (4, 17). In parallel to the above studies, we also investigated the resistance reversal properties of these same agents in combination with quinine (QN). Of all agents tested, AMLO displayed the greatest potentiation of QN against both Dd2 and 7G8, which manifest low-level QN resistance. Notably, AMLO was also the only agent that potentiated QN in the CQS line 3D7 (see Table S4 in the supplemental material).

When it was tested in a 4-day suppressive treatment model of mice infected with P. yoelii nigeriensis N67, the combination of CQ and AZ produced an additive drug-drug interaction. Introducing R-AMLO to the AZ-CQ combination further increased the potency of the regimen, leading to 99.9% suppression on day 4 when it was tested at the highest doses. Of note, although P. yoelii nigeriensis has been referred to as a CQR line (83), we observed that a regimen of 3 mg/kg CQ achieved 89.6% ± 7.5% suppression on day 4 (see Table S3 in the supplemental material). This is comparable to reported cutoff values of 3 mg/kg CQ in similar tests against CQS P. berghei strains (36; also confirmed in our laboratory). These results raise the question of whether P. yoelii N67 actually displays innate CQ resistance (70). Based on our data, we would argue that this strain should instead be classified as CQS and caution against assumptions that P. yoelii N67 and P. berghei differ significantly in their CQ susceptibility when citing the literature.

While R-AMLO increased CQ efficacy, it also improved the activity of AZ at a similar rate (∼70% relative or ∼30% absolute increase in day 4 parasite suppression; see Table S3 in the supplemental material). This result is consistent with a lack of a clear CQ resistance phenotype in P. yoelii that could be potentiated by AMLO. Alternatively, these results could indicate a previously undetected interaction between AZ and R-AMLO. Studies on the interaction between AZ and multidrug resistance (MDR; P-glycoprotein) transporters in murine macrophage and human intestinal cells have provided evidence that VP can inhibit efflux of AZ via MDR transporters (65, 80). Thus, it is possible that R-AMLO, also a calcium channel blocker, might inhibit MDR or other transporters in Plasmodium and thus increase the intracellular concentration of AZ. Our finding of a nearly 4-fold preferential AZ accumulation into parasitized RBCs, although far less than the accumulation observed with CQ (64, 75), serves as a baseline for future experiments aimed at understanding the combined effects of AZ, CQ, and R-AMLO.

Our data also prompt a discussion about whether AMLO is truly a CQ resistance reversal agent or is merely acting as a compound with weak antimalarial activity. Evidence for the latter comes from our studies with P. yoelii, which shows activity with AMLO alone. Nonetheless, our data revealed that AMLO potentiated CQ action only when they were tested in combination against CQR P. falciparum lines and not CQS lines, in agreement with the findings of former studies (4, 17), thus arguing for CQ resistance reversal activity of AMLO, in addition to some weak intrinsic activity.

Regarding the pharmacokinetic/pharmacodynamic studies, our results predict that in order to achieve cure with AZ-CQ plus R-AMLO combination therapy, a 65-kg adult would need a daily dose of 1.8 g of R-AMLO. There are two caveats to predicting this concentration. First, studies have shown that in rats, first-pass metabolism of AMLO is saturated at doses higher than 8 mg/kg. Given that human pharmacokinetic data are available only with low doses (∼0.14 mg/kg in 70-kg adults) (84), this point of first-pass saturation might be achieved in humans at lower doses than in mice, leading to higher exposure from doses lower than 1.8 g per day. Second, the blood/plasma ratio in mice was roughly 3-fold. If the human blood/plasma ratio were drastically higher, then increased efficacy might be achieved at a lower dose. Nevertheless, a daily dose of 1.8 g of R-AMLO is likely unsafe for humans, especially when it is combined with CQ and AZ, which can also cause arrhythmias as a result of prolonged QT (Zithromax package insert).

In discussing our findings, we note two limitations to our study. First, our P. falciparum in vitro data with CQ and AZ tested alone and in combination included field isolates that had not been cloned. Genotyping of these isolates with primer pairs that amplified polymorphic regions of the msp1 and msp2 genes (42) and the microsatellite marker TA81 (3) revealed these to be polyclonal infections. It is well established that mixed infections can produce biphasic dose-response profiles with antimalarial drugs, in cases where these harbor a mixture of drug-sensitive and -resistant clones, and that over time certain clones can become dominant in the culture (91). With the Malian isolates, genotyping of pfcrt exon 2 and CQ responses provided evidence of monophasic responses, consistent with all parasites being CQR. Other fast-acting drugs were not tested to check for admixtures of drug-resistant and -sensitive parasites. Second, our in vivo data were obtained with a rodent parasite that appears to not be CQR, despite literature reports indicating that this P. yoelii N67 should have been resistant to CQ (17, 70). We are unaware of any other P. yoelii strain that is truly CQR, and the only highly CQR P. berghei line that we have been able to test is the RC strain; however, we find that the RC strain cannot propagate to sufficiently high parasitemias (>2%) in mice to enable us to perform any drug efficacy studies. Thus, our mouse studies do not enable us to test whether AMLO is acting as a reversal agent in vivo.

Despite our findings on AMLO, the potential of AZ as an antimalarial, whether or not in combination with CQ, continues to be of interest. Besides having significant antimalarial activity, an empirical antimalarial regimen that includes AZ would have the advantage of treating a wide array of common bacterial infections. This is important because, as multiple studies have shown, a significant clinical overlap can exist between malaria and other infections, particularly those cased by bacterial pathogens, among children in sub-Saharan Africa (24, 39, 40, 61). In cases of pneumonia, for example, this overlap can lead to misdiagnosis and delays adequate therapy, potentially increasing morbidity and mortality among individuals in that group (24, 39, 61). In addition, other studies have found that coinfections with nontyphoidal Salmonella, which is typically sensitive to AZ, can account for more than half of all bloodstream infections in patients with severe malaria (11). Thus, given the distinct challenges in promptly diagnosing malaria and its overlap with common bacterial infections, there have been calls for the creation of an empirical antimalarial combination treatment that would include a well-tolerated antibacterial agent with a broad spectrum of activity, such as AZ (56). AZ-CQ combinations are also being considered for potential use in intermittent preventive treatment of malaria in pregnancy (13). With this application in mind, we note that a recent study of intermittent preventive treatment in children found no long-term interference with the antibody response to P. falciparum antigens, thereby reducing a concern that this form of intervention might compromise patient immunity to subsequent infections (9).

A recent paper by Friesen et al. points to another potential benefit of having AZ as part of an empirical antimalarial combination regimen (28). In a mouse model of causal prophylaxis using P. berghei, AZ not only prevented the parasite from advancing beyond the liver stage but it also induced potent immune protection against subsequent sporozoite infections. Thus, periodic concomitant exposure to both AZ and P. falciparum could potentially lead to a vaccine-like protective effect. If such a strategy were successful, the benefits in terms of reduced rates of mortality and morbidity could be substantial.

In light of preexisting CQ resistance in P. falciparum, AZ has also been studied in combination with other antimalarials, yielding promising results. These include in vitro synergism reported with dihydroartemisinin (58) and fast clinical and parasitological cure of uncomplicated malaria in patients treated with artesunate and AZ (57, 86). AZ combined with sulfadoxine-pyrimethamine was also recently reported to be safe, well tolerated, and efficacious for the treatment of malaria during pregnancy (38). Another interesting candidate would be piperaquine, a bisquinoline that is generally highly active against CQR parasites and that is now being used in combination with dihydroartemisinin for the treatment of uncomplicated malaria in Southeast Asia (16, 21, 89). Our analysis of piperaquine-AZ combinations in vitro shows that they display additive interactions at the FIC₉₀ levels against the same Malian strains tested with AZ-CQ (see Fig. S1 in the supplemental material). These data provide a useful comparison to our AZ-CQ studies that showed marked synergy at the FIC₉₀ level (Fig. 1).

Prior to any large-scale deployment of AZ as a component in antimalarial combination therapy, an important issue to address would be the possible emergence of AZ resistance among bacteria and/or Plasmodium parasites. Given its long half-life, lingering subtherapeutic levels of AZ would create selective pressure for the emergence of resistance. Examples of resistance in a setting of mass use already exist. Studies in France have shown the steady increase of AZ resistance in Mycoplasma pneumoniae, a common cause of community-acquired pneumonia worldwide (71). As recently documented by Haug et al. (35), after 3 years of biannual mass distributions of AZ to treat trachoma, almost 80% of Streptococcus pneumoniae nasopharyngeal isolates recovered from treated patients were resistant to this macrolide. This resistance declined toward earlier levels after mass deployment was terminated. Additionally, low-level AZ resistance has been selected in vitro in P. falciparum (81).

In sum, AZ presents a potentially attractive component of future antimalarial combination therapies because of its safety, efficacy, and broad-spectrum activity. Clinical evaluation of AZ-containing combinations to treat uncomplicated malaria and investigations of its potential use in intermittent preventive treatment of malaria in pregnancy are ongoing and will be highly informative in deciding the role that AZ can play in malaria treatment programs.