Multinormal In Vitro Distribution Model Suitable for the Distribution of ...

ANTIMICROBIAL AGENTS AND CHEMOTHERAPY, Feb. 2009, p. 688–695 0066-4804/09/$08.00⫹0 doi:10.1128/AAC.00546-08 Copyright © 2009, American Society for Microbiology. All Rights Reserved.

Vol. 53, No. 2

Multinormal In Vitro Distribution Model Suitable for the Distribution of Plasmodium falciparum Chemosusceptibility to Doxycycline䌤 Se´bastien Briolant,1,2 Meili Baragatti,1,2 Philippe Parola,1,2,3 Fabrice Simon,2,4 Adama Tall,5 Cheikh Sokhna,6 Philippe Hovette,7 Modeste Mabika Mamfoumbi,8 Jean-Louis Koeck,9 Jean Delmont,2,3,10 Andre´ Spiegel,5 Jacky Castello,7 Jean Pierre Gardair,11 Jean Francois Trape,2,6 Maryvonne Kombila,8 Philippe Minodier,12 Thierry Fusai,1,2 Christophe Rogier,1,2 and Bruno Pradines1,2,13* Unite´ de Recherche en Biologie et Epide´miologie Parasitaires, Institut de Me´decine Tropicale du Service de Sante´ des Armeés, Le Pharo, 13998 Marseille, France1; Unite´ de Recherche sur les Maladies Infectieuses et Tropicales Emergentes, Unite´ mixte de recherche 6236, Marseille, France2; Service des Maladies Infectieuses et Tropicales, Ho ˆpital Nord, Assistance Publique des Ho ˆpitaux de Marseille, Marseille, France3; Service de Pathologies Infectieuses et Tropicales, Ho ˆ pital d’Instruction des Armeés Laveran, Marseille, France4; Service d’Epide´miologie, Institut Pasteur, Dakar, Seńe´gal5; Unite´ de Recherche 077 de Paludologie Afrotropicale, Institut pour la Recherche et le De´veloppement, Dakar, Seńe´gal6; Centre Me´dical de Secours, Total Exploration et Production Congo, Pointe-Noire, Re´publique du Congo7; De´partement de Parasitologie-Mycologie, Faculte´ de Me´decine, Universite´ des Sciences de la Sante´, Libreville, Gabon8; Service de Biologie Me´dicale Centre Hospitalier des Armeés Bouffard, Djibouti9; Centre de Formation et Recherche en Me´decine et Sante´ Tropicales, Faculte´ de Me´decine Secteur Nord10; De´partement Me´dical International, Total, 92078 Paris la De´fense, France11; Service de Pe´diatrie, Ho ˆpital Nord, Assistance Publique des Ho ˆpitaux de Marseille, Marseille, France12; and Centre National de Re´fe´rence du Paludisme, France13 Received 28 April 2008/Returned for modification 20 July 2008/Accepted 14 November 2008

The distribution and range of 50% inhibitory concentrations (IC50s) of doxycycline were determined for 747 isolates obtained between 1997 and 2006 from patients living in Senegal, Republic of the Congo, and Gabon and patients hospitalized in France for imported malaria. The statistical analysis was designed to answer the specific question of whether Plasmodium falciparum has different phenotypes of susceptibility to doxycycline. A triple normal distribution was fitted to the data using a Bayesian mixture modeling approach. The IC50 geometric mean ranged from 6.2 ␮M to 11.1 ␮M according to the geographical origin, with a mean of 9.3 ␮M for all 747 parasites. The values for all 747 isolates were classified into three components: component A, with an IC50 mean of 4.9 ␮M (ⴞ2.1 ␮M [standard deviation]); component B, with an IC50 mean of 7.7 ␮M (ⴞ1.2 ␮M); and component C, with an IC50 mean of 17.9 ␮M (ⴞ1.4 ␮M). According to the origin of the P. falciparum isolates, the triple normal distribution was found in each subgroup. However, the proportion of isolates predicted to belong to component B was most important in isolates from Gabon and Congo and in isolates imported from Africa (from 46 to 56%). In Senegal, 55% of the P. falciparum isolates were predicted to be classified as component C. The cutoff of reduced susceptibility to doxycycline in vitro was estimated to be 35 ␮M. Over the past 20 years, strains of Plasmodium falciparum have become resistant to chloroquine (CQ) and other antimalarial drugs (48, 51). This has led to a search for effective alternative antimalarial drugs with minimal side effects. The emergence and spread of parasite resistance to antimalarial drugs necessitates the discovery and development of novel compounds via the identification of novel chemotherapeutic targets. Thirty years ago, tetracyclines were found to have antimalarial activity (37). Experimental observations obtained in vitro (13, 22) and in clinical studies (52) demonstrated the antimalarial activity of tetracycline and its derivatives. Daily doxycy-

cline (DOX) has been shown to be an effective causal chemoprophylactic in Thailand (41), Indonesia (27), and Kenya (1). DOX is currently one of the recommended chemoprophylactic regimens for travelers or soldiers visiting areas of endemicity in Southeast Asia (15), Africa, and South America. DOX is now recommended by the French Consensus conference for chemoprophylaxis in countries with a high prevalence of resistance to CQ or multiresistance (group 3 countries) (3). Since September 2002, French troops have participated in a peacekeeping operation, Operation Licorne, on the Ivory Coast. Soldiers were prescribed DOX (100 mg) daily for prophylaxis. Many cases of malaria have been reported, but most of them are believed to have resulted from poor compliance (23, 24). From 2002 to 2006, 1,787 malaria cases were found among French soldiers expected to be taking DOX. A surge in the number of malaria cases within 3 weeks after discontinuing DOX prophylaxis is often observed (29, 43). DOX failures soon after termination of drug therapy suggest that DOX serves primarily as a suppressive agent. It is recommended that

* Corresponding author. Mailing address: Unite´ de Recherche en Biologie et Epide´miologie Parasitaires, Institut de Me´decine Tropicale du Service de Sante´ des Armeés, Boulevard Charles Livon, Parc le Pharo, 13998 Marseille Armeés, France. Phone: 33 4 91 15 01 10. Fax: 33 4 91 15 01 64. E-mail: [email protected]. 䌤 Published ahead of print on 1 December 2008. 688

VOL. 53, 2009

DISTRIBUTION OF 747 DOXYCYCLINE IC50s

DOX be taken for 4 weeks after returning from an area of endemicity. In vitro analysis of the susceptibility of P. falciparum strains to antimalarial drugs has three important attributes. First, this approach allows one to assay the response of clinical isolates to individual drugs that are unmodified by important host factors that influence drug efficacy in vivo. Second, the progressive decline in drug susceptibility of isolates from the same site may be utilized for the development of a sensitive method for identifying incipient resistance in the parasite population. Finally, strains with reduced antimalarial susceptibilities can then be established in continuous culture to provide the material needed to investigate novel molecular mechanisms of resistance as well as for tests of susceptibility to other antimalarial agents. Longitudinal studies and comparison of in vitro parasite responses within a defined human population can reveal trends in susceptibility to a particular drug and thus represent the earliest possible warning of developing resistance in local parasite populations. Early detection of resistance to DOX requires that baseline parasite chemosusceptibility of current isolates from regions of endemicity be established. The aim of the present work was to determine the distribution and range of 50% inhibitory concentrations (IC50s) of DOX for 747 isolates obtained between 1997 and 2006 from patients living in Senegal, Republic of Congo, Gabon, and Djibouti and patients hospitalized in Marseille (France) for imported malaria.

MATERIALS AND METHODS P. falciparum isolates. Between January 1999 and December 2006, 309 fresh P. falciparum isolates were obtained in Marseille from patients hospitalized with malaria after having returned to France from the following locations: Comoros (n ⫽ 146), the Ivory Coast (n ⫽ 51), Cameroon (n ⫽ 18), Senegal (n ⫽ 15), Burkina Faso (n ⫽ 11), Madagascar (n ⫽ 10), Gabon (n ⫽ 7), Guinea (n ⫽ 7), Democratic Republic of the Congo (n ⫽ 6), Mali (n ⫽ 5), Benin (n ⫽ 5), and other areas in Africa (n ⫽ 28). Between October 1997 and February 1998, September and November 1998, and September and December 1999, 214 P. falciparum isolates were obtained in Dielmo and Ndiop (280 km southeast of Dakar) in the Fatick region of Senegal. Patients from Dielmo and Ndiop were recruited at home by daily active case detection during a longitudinal study of the mechanisms of protective immunity to malaria (25, 39). Between March 2005 and January 2006, 114 children aged 16 months to 17 years were enrolled from among those who went to the Medical Service of Total-Elf, Pointe-Noire, Republic of Congo, due to febrile illness and uncomplicated malaria. Between June and December 1999, 72 fresh P. falciparum isolates were obtained from children with uncomplicated malaria from Libreville, Gabon. Twenty blood samples were collected in April 1999 from patients with P. falciparum who had not traveled outside Djibouti and 18 from patients in different parts of Africa. Informed consent was obtained individually from all adults and from the parents of children before blood collection, and the local institutional ethical guidelines were followed. Venous blood samples were collected in Vacutainer ACD tubes (Becton Dickinson, Rutherford, NJ) before treatment and transported at 4°C to our laboratory in Marseille within less than 48 h of collection. Thin blood smears were stained using a RAL kit (Reáctifs RAL, Paris, France) and examined to determine the P. falciparum density. Parasitized erythrocytes were washed three times in RPMI 1640 medium (Invitrogen, Paisley, United Kingdom). If parasitemia exceeded 0.8%, infected erythrocytes were diluted to 0.5 to 0.8% with uninfected erythrocytes and resuspended in culture medium to a hematocrit of 1.5%. Susceptibility to DOX, CQ, quinine (QN), mefloquine (MQ), or atovaquone (ATV) was determined after suspension in RPMI 1640 medium, which was supplemented with 10% human serum and buffered with 25 mM HEPES and 25 mM NaHCO3. Susceptibility to cycloguanil (CY) was determined after suspension in RPMI 1640 SP823 with reduced p-aminobenzoic acid (0.5 ␮g/liter) and low folates (10 ␮g/liter) (Invitrogen).

689

Drugs. DOX hydrochloride was obtained from Sigma (St. Louis, MO), and the stock solution was prepared in methanol. Twofold serial dilutions were prepared in sterile water to obtain final concentration ranges from 0.1 to 502 ␮M and distributed in triplicate into Falcon 96-well flat-bottomed plates (Becton Dickinson, Franklin Lakes, NJ). CQ diphosphate and QN hydrochloride were purchased from Sigma. MQ was obtained from Hoffman-LaRoche (Bale, Switzerland), CY from Zeneca Pharma (Reims, France), and ATV from GlaxoSmithKline (Research Triangle Park, NC). CQ and CY were resuspended in water at concentrations ranging from 5 to 3,200 nM and 10 to 20,000 nM, respectively. QN, MQ and ATV were first dissolved in methanol and then diluted in water to obtain final concentration ranges from 5 to 3,200 nM for QN, 3.2 to 400 nM for MQ, and 0.3 to 100 nM for ATV. The CQ-susceptible P. falciparum clone 3D7 (Africa) and the CQ-resistant clone W2 (Indochina) were used as references to test each batch of plates. Reference clones were maintained in continuous culture and synchronized twice with sorbitol (19). Batches were validated when IC50s for 3D7 were 8 to 14 ␮M for DOX, 13 to 35 nM for CQ, 60 to 180 nM for QN, 35 to 69 nM for MQ, 1.5 to 5.5 for ATV, and ⬍10 nM for CY and those for W2 were 10 to 16 ␮M for DOX, 345 to 625 nM for CQ, 520 to 840 nM for QN, 23 to 43 nM for MQ, 2.0 to 4.6 for ATV, and 900 to 2,900 nM for CY. Drug assay. For in vitro isotopic microtests, 200 ␮l/well of the suspension of parasitized erythrocytes was distributed in 96-well plates predosed with drugs. Parasite growth was assessed by adding 1 ␮Ci of [3H]hypoxanthine with a specific activity of 14.1 Ci/mmol (NEN Products, Dreiech, Germany) to each well. Plates were incubated for 42 h at 37°C in an atmosphere of 10% O2, 5% CO2, 85% N2, and 95% humidity. Immediately after incubation, the plates were frozen and then thawed to lyse erythrocytes. The contents of each well were collected on standard filter microplates (Unifilter GF/B, Perkin Elmer, Meriden, CT) and washed using a cell harvester (FilterMate; Packard). Filter microplates were dried, and 25 ␮l of scintillation cocktail (Microscint O; Perkin Elmer) was placed in each well. The radioactivity incorporated by the parasites was measured using a scintillation counter (Top Count; Perkin Elmer). The IC50, i.e., the drug concentration corresponding to 50% of the uptake of [3H]hypoxanthine by parasites in drug-free control wells, was determined by nonlinear regression analysis of the log dose-response curves (Riasmart; Packard, Meriden, CT). Statistical analysis. The statistical analysis was designed to answer the specific question of whether P. falciparum has different phenotypes of susceptibility to DOX. Parasite susceptibility is expressed as the IC50. As a result of the prima facie appearance of the pooled distribution of log10(IC50) looking like a threecomponent mixture distribution, a triple normal distribution was fitted to the data using a Bayesian mixture modeling approach (12, 38) with Markov chain Monte Carlo posterior simulation techniques available in WinBUGS version 1.4.3 (Imperial College and MRC, United Kingdom; free download available at http://www.mrc-bsu.cam.ac.uk/bugs). It was assumed that the proportions (pi) of the observations arose from three distinct normal distributions (i.e., components) with means (␭i) and standard deviations (SDs) (␴i), where p1 ⫹ p2⫹ p3 ⫽ 1 and ␭1 ⬍ ␭2 ⬍ ␭3. Indicator variables allowed the classification of observations according to the component of the mixture from which they were obtained. The parameters of the posterior distributions were estimated by simulation using the Gibbs sampler (46). Convergence of the model was assessed using the GelmanRubin statistic, as modified by Brooks and Gelman (8). To check the validity of our model, we used pivotal quantities, as explained by Johnson (16). Johnson proved that for a pivotal quantity, its sampling distribution evaluated at a posterior draw of the model parameters is the same as the sampling distribution evaluated at the true values of the model parameters that give rise to the observed data. The pivotal quantity used can be a function of both the data and parameters. We applied the results of Johnson to the pivotal 747 ˜T⬃i Yi ⫺ ␭ ˜ i being the posterior , with Yi being the ith observation, T quantities ⬃ ␴ ˜ T i i⫽1 ˜T⬃i being ˜ estimation of the component of the ith observation (Ti ⫽ 1,2,3), ␭ ˜T⬃i ⫽ ␭ ˜1,␭ ˜2,␭ ˜3), the posterior estimation of the means of the components (␭ and ␴ ˜ T⬃i being the posterior estimations of the SDs of the components ˜ 1,␴ ˜ 2,␴ ˜ 3). According to Johnson, these quantities have a chi-square (␴ ˜ T⬃i ⫽ ␴ distribution with 747 degrees of freedom under our proposed model. A double normal distribution model and a quadruple normal distribution model were also fitted to the data to look at the validity of considering a three-component mixture. The proportions of observations from subgroups (i.e., subgroups defined by geographical origin of the isolates or date of sampling) that were predicted as belonging to each of the components of the mixture distribution were estimated based on the hypothesis of identical parameters (i.e., mean and SD) of compo-

冘冉冊

690

BRIOLANT ET AL.

ANTIMICROB. AGENTS CHEMOTHER.

TABLE 1. In vitro susceptibility of P. falciparum isolates to antimalarial drugs

TABLE 3. Correlation of in vitro responses of P. falciparum isolates to DOX, CQ, QN, MQ, CY, and ATV

Antimalarial

No. of isolates

Geometric mean IC50

95% Confidence interval

Drug

DOX CQ QN MQ CY ATV

747 742 730 537 662 628

9.3 ␮M 105.6 nM 208.1 nM 9.9 nM 156.7 nM 3.6 nM

8.8–9.6 96.6–115.3 195.1–221.9 9.2–10.6 132.8–185.0 3.4–3.9

Chloroquine Quinine Mefloquine Cycloguanil Atovaquone

nent-specific normal distributions over all subgroups (i.e., hypothesis of stable phenotypes of susceptibility of P. falciparum to DOX in all subgroups). Assessment of standard antimalarial drug cross-resistance with DOX was estimated by the coefficient of the correlation coefficient (r) and the coefficient of determination (r2).

RESULTS Average parameter estimates for CQ, QN, MQ, CY, ATV, and DOX against the P. falciparum isolates are given in Table 1. The geometric mean IC50 of DOX for 747 was 9.3 ␮M (95% confidence interval, 8.8 to 9.6 ␮M). The IC50s for DOX were analyzed by the geographical origin of the isolates (Table 2). The IC50 geometric mean ranged from 6.2 ␮M to 11.1 ␮M according to the geographical origin. Significant differences between IC50s were observed according to the geographical origin (Kruskal-Wallis test, P ⬍ 0.001). Correlations of in vitro responses of isolates to DOX and standard antimalarial drugs are shown in Table 3. A positive correlation between the responses to two antimalarial drugs suggests in vitro cross-resistance but does not necessarily imply cross-resistance in vivo. For DOX, all r2 values were ⬍10%. The triple normal distribution model is represented in Fig. 1. The parameter estimates for the three-component mixture model, i.e., proportions of isolates in each normal distribution, the mean of log10(IC50) values, and the SD for each distribution, are summarized in Table 4. A double normal distribution model and a quadruple normal distribution model were also fitted to the data to assess the validity of considering a threecomponent mixture (Fig. 2). Using Johnson’s method, we obtained Bayesian P values for lack of fit of our models. A P value close to 0 or 1 indicates a model lack of fit. In contrast, the closer the P value is to 0.5, the better the model fits the data. After 1,000 simulations to obtain the Bayesian P value, the result showed that the triple normal distribution model fits the data well (P ⫽ 0.609). Then we ran 1,000 simulations for each of the two other models and showed that the double normal

TABLE 2. In vitro susceptibility of P. falciparum isolates from different areas in Africa to doxycycline Source

No. of isolates

Geometric mean IC50 (␮M)

95% Confidence interval

Imported from Africa Senegal Congo Gabon Other African sites Africa (all areas)

309 214 114 72 38 747

8.5 11.1 8.2 11.0 6.2 9.3

7.9–9.1 10.0–12.3 7.4–9.1 9.7–12.4 5.3–7.3 8.8–9.6

Value for comparison to DOX

No. of isolates

r

r2

P

742 730 537 662 628

⫺0.0005 ⫹0.1967 ⫹0.1775 ⫹0.0835 ⫹0.3095

⬍0.0001 0.0387 0.0315 0.0070 0.0958

0.9895 ⬍0.0001 ⬍0.0001 0.0316 ⬍0.0001

distribution model (P ⫽ 0.13) and the quadruple normal distribution model (P ⫽ 1) fit the data worse than the triple normal distribution model. Separate analyses of the first and second halves of the data set (1997 to 2001 and 2002 to 2006, respectively) gave similar estimates of the parameters of the distributions of IC50s in the three components (data not shown). The distribution of the IC50s of the P. falciparum isolates in the three-component mixture model according to geographical origin is shown in Fig. 3. The three-component model appears to fit the data from imported malaria or from Congo well. A separate analysis of Senegalese data showed point estimates of the geometric means in the three components (4.2, 8.2, and 19.5 ␮M in components A, B, and C, respectively) that were similar to those obtained with the whole data set (Table 4) and 95% prediction intervals of these estimates (2.1 to 6.9, 6.8 to 9.7, and 17.7 to 21.8 for components A, B, and C, respectively) that included the point estimates obtained with the whole data set (Table 4). Separate analysis of the data from Gabon was not possible because of the small sample size. The proportions of the observations predicted to belong to each of the components of the mixture distribution according to geographical origin are summarized in Table 5. About 50% of the IC50s for DOX were distributed in the B component for imported malaria isolates from Africa as well as isolates from Congo and Gabon. The B component represented only 13% of the isolates from Senegal. The cutoff for reduced susceptibility to DOX in vitro was estimated by the geometric mean plus 2 SDs of the IC50s of the P. falciparum isolates associated with component C, that is, 34.2 ␮M. Isolates with an IC50 greater than 35 ␮M are considered isolates with reduced susceptibility to DOX in vitro. DISCUSSION Most prophylactic failures of DOX against P. falciparum are associated with the use of standard doses resulting in lowerthan-expected serum drug levels (50), inadequate low doses (30), or poor compliance (42, 49). Moreover, DOX pharmacokinetic parameters could explain some of these cases. DOX has a short elimination half-life (16 h) compared to proguanil (24 h), ATV (31 to 73 h), CQ (2 to 3 days), and MQ (6 to 41 days) and a short mean residence time (63% of the administered dose is eliminated in 27 h) (43). In addition, its slow action in vitro has a delayed effect upon growth and requires prolonged incubation of parasites (34, 35). Early detection of reduced susceptibility or resistance to DOX will require that baseline parasite chemosusceptibility of

VOL. 53, 2009


691

FIG. 1. Distribution of the log10(IC50) values of the 747 P. falciparum isolates from Africa (1997 to 2006) in the three-component mixture model (Bayesian mixture modeling approach). The dotted lines represent the three fitted mixtures, and the continuous line represents the overall mixture fit.

current isolates from regions of endemicity be established. Maximizing the efficacy and longevity of antimalarial drugs as a tool to control malaria will critically depend on pursuing intensive research into identifying in vitro markers as well as implementing in vitro and in vivo surveillance programs such as those championed by the World Antimalarial Resistance Network (44, 45). In this context, there is a need to identify in vitro and molecular markers that predict QN resistance which can provide an active surveillance method to monitor temporal trends in parasite susceptibility (4, 32). We opted for a Bayesian mixture modeling approach. A Bayesian approach has already been proposed for antimalarial in vitro susceptibilities, though from a different angle: a Bayesian approach was used to estimate IC50 in the outlying most resistant isolates to reduce overestimation of IC50 (47). All 747 values were classified into three components: component A (IC50 geometric mean of 4.9 ␮M [SD ⫽ 2.1 ␮M]), component B (IC50 geometric mean of 7.7 ␮M [SD ⫽ 1.2 ␮M]), and component C (IC50 geometric mean of 17.9 ␮M [SD ⫽ 1.4 ␮M]). The double normal and quadruple normal distribution models were less suitable for the distribution of the 747 DOX IC50s. However, the hypothesis of a quadruple normal distri-

TABLE 4. Parameter estimates for the three-component mixture model for the 747 P. falciparum isolates from Africaa IC50 (␮M)

Proportion (%) Component

A B C

Geometric mean

SD

Mean

95% PI

Mean

95% PI

SD

95% PI

24 44 32

15–34 37–52 25–39

4.9 7.7 17.9

3.5–6.6 7.5–8.0 16.4–19.3

2.1 1.2 1.4

1.9–2.5 1.17–1.25 1.3–1.5

a The number of iterations was 60,000, with a burn-in period of 10,000 iterations. Prediction intervals (PI) are the cutoff points for the 2.5% highest and lowest realizations of the estimate for each parameter.

bution suggested by a wider prediction interval of the IC50 SD of component A, a larger IC50 SD of component A, and the distribution of the lowest IC50s obtained for Senegalese isolates cannot be strictly ruled out. The observations used to fit the models were supposed to be conditionally independent given the value of the model parameters. However, the 747 isolates were not randomly sampled in either space or time. Could the time and space clustering of the data explain the appearance of the IC50s’ distribution in three components? Separate analyses of the first and second halves of the data set gave similar estimates of the parameters of the distributions of IC50s in the three components. Therefore, the different components identified in the present study did not reflect temporal trends. The three-component model fit the data from imported malaria and from Congo well. Moreover, a separate analysis suggested that the distribution of IC50s in Senegal was compatible with the assumption of a mixture of three components that had similar parameters in all places, even if the three components model did not seem to fit the data from Senegal well. According to the origin of the P. falciparum isolates, the triple normal distribution has been found in each subgroup. However, the proportion of isolates predicted to belong to component B was most important for isolates from Gabon and Congo and for cases imported from Africa (from 46 to 63%). In Gabon, the isolates were mainly distributed in two components, B (46%) and C (49%). In Senegal, 55% of the P. falciparum isolates were predicted to be classified as component C. We demonstrated the existence of at least three phenotypes for DOX in Africa. Are these three phenotypes associated with three different genotypes? Genotyping analysis is currently required in order to identify the molecular basis of the susceptibility differences and correlate genetic profiles to the phenotypes.

692

BRIOLANT ET AL.


FIG. 2. Double normal distribution model (A) and quadruple normal distribution model (B) of the log10(IC50) values of the P. falciparum isolates. The dotted lines represent the two or four fitted mixtures, and the continuous line represents the overall mixture fit.

A cutoff for reduced susceptibility to DOX in vitro was estimated by the geometric mean plus 2 SDs of the IC50s of the P. falciparum isolates in the C component, and the value obtained was 34.2 ␮M. Isolates with an IC50 greater than 35 ␮M were considered isolates with reduced susceptibility to DOX in vitro. Of the 747 P. falciparum isolates, 9 (1.2%) were considered to have decreased susceptibility to DOX in vitro. These data are consistent with clinical observations of limited prophylactic failures when DOX is used according to recommendations (6, 7). This 1.2% prevalence is very different from the 67% of isolates with decreased susceptibility observed in 2005 in French Guiana (20). The estimate of 67% prevalence was based on a susceptibility threshold of 9.6 ␮M (40). It seems that the 9.6 ␮M threshold was too low to define reduced susceptibility to DOX. The mean IC50 for DOX increased from 9.6 ␮M in 1996 to 13.1 ␮M in 2005 in French Guiana. However, these values are consistent with our observations in Africa. Using their cutoff set at 9.6 ␮M, the prevalence of isolates with an IC50 above this cutoff would be 41%. In our triple normal distribution model, all isolates predicted to belong to component C, 32%, would be resistant to DOX. These

data, with a threshold set at 9.6 ␮M, would not be consistent with the limited clinical observations of DOX prophylactic failures. DOX did not show significant cross-resistance with CQ, QN, MQ, CY, or ATV. A positive correlation between the IC50s of two antimalarial drugs may suggest in vitro cross-resistance or common mechanisms of action, but the relationship between in vitro and in vivo resistance depends on the level of resistance and the coefficients of correlation (r) and determination (r2). To involve the same mechanism of action for two compounds, which could induce cross-resistance, the coefficient of correlation must be high (r2 ⬎ 0.7), such as that for MQ and halofantrine (r2 ⫽ 0.777) (33). All of the coefficients of determination calculated between the responses to DOX and those of all the drugs tested are inferior to 0.09, suggesting that less than 9% of the variations in responses to DOX are explained by variations in responses to the other drugs. The lack of correlation between DOX and other antimalarial drugs is likely due to differences in drug targets or drug resistance mechanisms. Cyclines are a family of antibiotics that have long been known to inhibit protein synthesis in bacteria. Their mechanism of

VOL. 53, 2009


693

FIG. 3. Distribution of the log10(IC50) values of the P. falciparum isolates in the three-component mixture model (Bayesian mixture modeling approach) according to geographical origin. The dotted lines represent the three fitted mixtures, and the continuous line represents the overall mixture fit. TABLE 5. Proportions of observations predicted to belong to each of the components of the mixture distribution according to geographical origin Proportion (%) Origin of isolates

Component Mean

95% PI

Imported

A B C

22 56 22

15–29 49–63 17–28

Senegal

A B C

32 13 55

24–41 6–21 47–62

Congo

A B C

18 63 19

10–29 52–73 11–27

Gabon

A B C

5 46 49

1–14 33–59 36–62

Total

A B C

24 44 32

15–34 37–52 25–39

The number of iterations was 60,000, with a burn-in period of 10,000 iterations. Prediction intervals (PI) are the cutoff points for the 2.5% highest and lowest realizations of the estimate for each parameter.

action was further elucidated at the molecular level when it was determined that they were bound to proteins S4, S7, and S9 of the 30S small ribosomal subunit and various ribonucleic acids of the 16S rRNA (2, 26, 28, 31), preventing the binding of aminoacyl tRNA to site A of the ribosome and thus blocking the elongation step of translation. However, the mechanisms of action are much less clearly identified for Plasmodium. Several studies showed an action of the cyclines on the plasmodial mitochondrion at concentrations of 1 ␮M to 10 ␮M, which are similar to clinical concentrations. One study highlighted an in vitro synergistic activity between exposure time of the parasite to tetracyclines and the increase in oxygen content of the Plasmodium environment (13), suggesting an action of the tetracyclines on mitochondria. The latter is implicated in the control of oxidative stress and the energy production of anaerobic plasmodia (18). According to three previous studies, cyclines can directly inhibit mitochondrial protein synthesis (5, 9, 17) and also decrease the activity of an enzyme, the dihydroorotate dehydrogenase involved in the de novo synthesis of pyrimidines (36). DOX would inhibit the synthesis of nucleotides and deoxynucleotides in P. falciparum (53) at concentrations much higher than those observed in vivo (200 ␮M). The in vitro exposure of P. falciparum to minocycline would also

694

BRIOLANT ET AL.

decrease the transcription of mitochondrial genes (subunit I of the cytochrome c oxidase and apocytochrome b) and plastid genes (subunit RpoB/C of the RNA polymerase), suggesting an activity that affects these two organelles (21). One study showed that DOX acts specifically on the apicoplast of P. falciparum (11) and, to a lesser extent, on the mitochondrion, the division of which is inhibited at the end of the parasite cycle. Those authors concluded that this phenomenon is due to the apicoplastic attack (the two organelles have common metabolic pathways). A parasite exposed to 1 ␮M DOX for 20 h presents an inhibition of apicoplastic replication during the following cycle. This can be visualized by fluorescence confocal microscopy, electron microscopy, and analysis of the parasitic transcriptome. Two recent studies confirmed the specific action of the cyclines on the apicoplast of P. falciparum (10, 14). DOX is currently the only antimalarial drug for which no P. falciparum resistance has been described. The in vitro susceptibility responses to DOX were distributed among three components. Genotyping analysis is required in order to identify the molecular basis of the susceptibility differences. Only 1.2% of the 747 tested isolates had reduced susceptibility to DOX in vitro, consistent with clinical observations. It will be critical to identify early signs of resistance before these strains become prevalent and compromise the clinical utility of DOX. Maximizing the efficacy and longevity of DOX as a tool to control malaria will critically depend on pursuing intensive research into identifying in vitro markers as well as implementing in vitro and in vivo surveillance programs, such as those championed by the World Antimalarial Resistance Network.


5. 6. 7. 8. 9. 10. 11.

12. 13.

14. 15.

16. 17. 18.

ACKNOWLEDGMENTS

19.

We thank all the participants in the different countries and the staff of the Service des Maladies Infectieuses et Tropicales de l’Ho ˆpital Nord in Marseille, the Service de Pathologies Infectieuses et Tropicales de l’Ho ˆpital Laveran in Marseille, the Institute Pasteur in Dakar, the Unite´ de Recherche 077 de Paludologie Tropicale in Dakar, the Centre Me´dical de Secours of Total Exploration et Production Congo in Pointe-Noire, the De´partement Me´dical International of Total in Paris, the De´partement de Parasitologie-Mycologie in Libreville, and the Service de Biologie Me´dicale de l’Ho ˆpital Bouffard in Djibouti. We thank R. Amalvict, E. Baret, W. Daries, M. Fortunee, R. Ges, J. Mosnier, D. Ragot, D. Ramarlah, and Y. Trullemans from the Institut de Me´decine Tropicale du Service de Sante´ des Armeés in Marseilles for their technical support since 1997. We acknowledge the financial support of the De´le´gation Geńe´rale pour l’Armement, the Direction Centrale du Service de Sante´ des Armeés (grant no. 06CO009), and the Programme Hospitalier de Recherche Clinique Re´gional APHM 2003. We have no conflicts of interest concerning the work reported in this paper. None of the authors owns stocks or shares in a company that might be financially affected by the conclusions of this article. The conclusions of this article were not financially affected.

20.

21. 22. 23.

24.

25.

26.

REFERENCES 1. Andersen, S. L., A. J. Oloo, D. M. Gordon, O. B. Ragama, G. M. Aleman, J. D. Berman, D. B. Tang, M. W. Dunne, and G. D. Shanks. 1998. Successful double-blinded, randomized, placebo-controlled field trial of azithromycin and doxycycline as prophylaxis for malaria in Western Kenya. Clin. Infect. Dis. 26:146–150. 2. Anokhina, M. M., A. Barta, K. H. Nierhaus, V. A. Spiridonova, and A. M. Kopylov. 2004. Mapping of the second tetracycline binding site on the ribosomal small subunit of E. coli. Nucleic Acids Res. 32:2594–2597. 3. Anonymous. 2008. Recommendations for clinical practice. Management and prevention of imported Plasmodium falciparum malaria: revision 2007 of the 1999 consensus conference. Med. Malaria Infect. 38:54–67. 4. Bacon, D. J., R. Jambou, T. Fandeur, J. Le Bras, C. Wongsrichanalai, M. M. Fukuda, P. Ringwald, C. H. Sibley, and D. E. Kyle. 2007. World Antimalaria

27.

28. 29. 30. 31.

Resistance Network (WARN) II: in vitro antimalarial drug susceptibility. Malaria J. 6:120. Blum, J. J., A. Yayon, S. Friedman, and H. Ginsburg. 1984. Effects of mitochondrial protein synthesis inhibitors on the incorporation of isoleucine into Plasmodium falciparum in vitro. J. Protozool. 31:475–479. Briolant, S., L. Almeras, T. Fusai, C. Rogier, and B. Pradines. 2007. Cyclines and malaria. Med. Trop. 67:86–96. Briolant, S., T. Fusai, C. Rogier, and B. Pradines. 2008. Tetracyclines antibiotics in Malaria. Open Trop. Med. J. 1:31–46. Brooks, S. P., and A. Gelman. 1998. Alternative methods for monitoring convergence of iterative simulations. J. Comput. Graph. Stat. 7:434–455. Budimulja, A. S., Syafruddin, P. Tapchaisri, P. Wilairat, and S. Marzuki. 1997. The sensitivity of Plasmodium protein synthesis to prokaryotic ribosomal inhibitors. Mol. Biochem. Parasitol. 84:137–141. Dahl, E. L., and P. J. Rosenthal. 2007. Multiple antibiotics exert delayed effects against the Plasmodium falciparum apicoplast. Antimicrob. Agents Chemother. 51:3485–3490. Dahl, E. L., J. L. Shock, B. R. Shenai, J. Gut, J. L. DeRisi, and P. J. Rosenthal. 2006. Tetracyclines specifically target the apicoplast of the malaria parasite Plasmodium falciparum. Antimicrob. Agents Chemother. 50: 3124–3131. Diebolt, J., and C. P. Robert. 1994. Estimation of finite mixture distributions through Bayesian sampling. J. R. Stat. Soc. B 56:363–375. Divo, A. A., T. G. Geary, and J. B. Jensen. 1985. Oxygen- and timedependent effects of antibiotics and selected mitochondrial inhibitors on Plasmodium falciparum in culture. Antimicrob. Agents Chemother. 27: 21–27. Goodman, C. D., V. Su, and G. I. McFadden. 2007. The effects of antibacterials on the malaria parasite Plasmodium falciparum. Mol. Biochem. Parasitol. 152:181–191. Gras, C., R. Laroche, L. Guelain, G. Martet, M. Merlin, G. Pottier, M. Guisset, and J. E. Touze. 1993. Place actuelle de la doxycycline dans la chimioprophylaxie du paludisme à Plasmodium falciparum. Bull. Soc. Pathol. Exot. 86:52–55. Johnson, V. 2004. A Bayesian chi-square test for goodness-of-fit. Ann. Stat. 32:2361–2384. Kiatfuengfoo, R., T. Suthiphongchai, P. Prapunwattana, and Y. Yuthavong. 1989. Mitochondria as the site of action of tetracycline on Plasmodium falciparum. Mol. Biochem. Parasitol. 34:109–115. Krungkrai, J. 2004. The multiple roles of the mitochondrion of the malarial parasite. Parasitology 129:511–524. Lambros, C., and J. P. Vanderberg. 1979. Synchronization of Plasmodium falciparum erythrocytic stages in culture. J. Parasitol. 65:418–420. Legrand, E., B. Volney, J. B. Meynard, O. Mercereau-Puijalon, and P. Esterre. 2008. In vitro monitoring of Plasmodium falciparum drug resistance in French Guiana: a synopsis of continuous assessment from 1994 to 2005. Antimicrob. Agents Chemother. 52:288–298. Lin, Q., K. Katakura, and M. Suzuki. 2002. Inhibition of mitochondrial and plastid activity of Plasmodium falciparum by minocycline. FEBS Lett. 515: 71–74. McColm, A. A., and N. McHardy. 1984. Evaluation of a range of antimicrobial agents against the parasitic protozoa, Plasmodium falciparum, Babesia rhohaini and Theileria parva in vitro. Ann. Trop. Med. Parasitol. 78:345–354. Migliani, R., R. Josse, P. Hovette, A. Keundjian, F. Pages, J. B. Meynard, L. Ollivier, K. Sbai Idrissi, K. Tifratene, E. Orlandi, C. Rogier, and J. P. Boutin. 2003. Le paludisme vu des trancheés: le cas de la Co ˆte d’Ivoire en 2002–2003. Med. Trop. 63:282–286. Migliani, R., L. Ollivier, O. Romand, C. Verret, R. Haus-Cheymol, A. Todesco, F. Pages, B. Pradines, B. Queyriaux, G. Texier, R. Michel, A. Spiegel, and J. P. Boutin. 2008. Paludisme chez les militaires français en Co ˆte d’Ivoire de 1998 à 2006. Bull. Epidemiol. Hebdom. 23–24:209–212. Norate, N., R. Durand, A. Tall, L. Marrama, C. Sohna, B. Pradines, S. Cojean, M. Guillotte, E. Bischoff, M. T. Ekala, C. Bouchier, T. Fandeur, F. Ariey, J. Patarapotikul, J. Le Bras, J. F. Trape, C. Rogier, and O. Mercereau-Puijalon. 2007. Rapid dissemination of Plasmodium falciparum drug resistance despite strictly controlled antimalarial use. PLoS One 1:139. Oehler, R., N. Polacek, G. Steiner, and A. Barta. 1997. Interaction of tetracycline with RNA: photoincorporation into ribosomal RNA of Escherichia coli. Nucleic Acids Res. 25:1219–1224. Ohrt, C., T. L. Richie, H. Widjaja, G. D. Shanhs, J. Fitriadi, D. J. Fryauff, J. Handschin, D. Tang, B. Sandjaja, E. Tjitra, L. Hadiarso, G. Watt, and S. Wignall. 1997. Mefloquine compared with doxycycline for the prophylaxis of malaria in Indonesian soldiers. Ann. Intern. Med. 126:963–972. Olson, M. W., A. Ruzin, E. Feyfant, T. S. Rush, J. O’Connell, and P. A. Bradford. 2006. Functional, biophysical, and structural bases for antibacterial activity of tigecycline. Antimicrob. Agents Chemother. 50:2156–2166. Pang, L. W., N. Limsomwong, E. F. Boudreau, and P. Singharaj. 1987. Doxycycline prophylaxis for falciparum malaria. Lancet i:1161–1164. Pang, L., N. Limsomwong, and P. Singhara. 1988. Prophylactic treatment of vivax and falciparum malaria with low-dose doxycycline. J. Infect. Dis. 158: 1124–1127. Pioletti, M., F. Schlunzen, J. Harms, R. Zarivach, M. Gluhmann, H. Avila,

VOL. 53, 2009

32.

33.

34. 35.

36. 37.

38.

39. 40. 41. 42.

A. Bashan, H. Bartels, T. Auerbach, C. Jacobi, T. Hartsch, A. Yonath, and F. Franceschi. 2001. Crystal structures of complexes of the small ribosomal subunit with tetracycline, edeine and IF3. EMBO J. 20:1829–1839. Plowe, C. V., C. Rooper, J. W. Barnwell, C. T. Happi, H. H. Joshi, W. Mbacham, S. R. Meshnick, K. Mugittu, I. Naidoo, R. N. Price, R. W. Shafer, C. H. Sibley, C. J. Sutherland, P. A. Zimmerman, and P. J. Rosenthal. 2007. World Antimalarial Resistance Network (WARN) III: molecular markers for drug resistant malaria. Malaria J. 6:121. Pradines, B., P. Hovette, T. Fusai, H. L. Atanda, E. Baret, P. Cheval, J. Mosnier, A. Callec, J. Cren, R. Amalvict, J. P. Gardair, and C. Rogier. 2006. Prevalence of in vitro resistance to eleven standard or new antimalarial drugs among Plasmodium falciparum isolates from Pointe-Noire, Republic of the Congo. J. Clin. Microbiol. 44:2404–2408. Pradines, B., C. Rogier, T. Fusai, J. Mosnier, W. Daries, E. Baret, and D. Parzy. 2001. In vitro activities of antibiotics against Plasmodium falciparum are inhibited by iron. Antimicrob. Agent Chemother. 45:1746–1750. Pradines, B., A. Spiegel, C. Rogier, A. Tall, J. Mosnier, T. Fusai, J. F. Trape, and D. Parzy. 2000. Antibiotics for prophylaxis of Plasmodium falciparum infections: in vitro activity of doxycycline against Senegalese isolates. Am. J. Trop. Med. Hyg. 62:82–85. Prapunwattana, P., W. J. O’Sullivan, and Y. Yuthavong. 1988. Depression of Plasmodium falciparum dihydroorotate dehydrogenase activity in in vitro culture by tetracycline. Mol. Biochem. Parasitol. 27:119–124. Rieckmann, K. H., R. D. Powell, J. V. McNamara, D. Willerson, L. Kass, H. Frischer, and P. E. Carson. 1971. Effects of tetracycline against chloroquineresistant and chloroquine-sensitive Plasmodium falciparum. Am. J. Trop. Med. Hyg. 20:811–815. Robert, C. P. 1996. Mixtures of distributions: inference and estimation, p. 441–464. In W. R. Gilks, S. Richardson and D. J. Spiegelhalter (ed.), Markov chain Monte Carlo in practice. Chapman and Hall, London, United Kingdom. Rogier, C., and J. F. Trape. 1996. Etude de l’acquisition de la pre´munition en zones d’holo et me´soende´mie palustre à Dielmo et à Ndiop (Seńe´gal). Re´sultats pre´liminaires, 1990–1994. Med. Trop. 55:71–76. Sarthou, J. L., and J. M. Reynes. 1996. Malaria in French Guiana: chemoresistance of Plasmodium falciparum isolates. Rev. Epidemiol. Sante´ Publique 44:42. Shanks, G. D., M. D. Edstein, V. Suriyamongkol, S. Timsaad, and H. K. Webster. 1992. Malaria chemoprophylaxis using proguanil/dapsone combinations on Thai-Cambodian border. Am. J. Trop. Med. Hyg. 46:643–648. Shanks, G. D., P. Roessler, M. D. Edstein, and K. H. Rieckmann. 1995.


43.

44. 45.

46.

47.

48.

49.

50.

51. 52.

53.

695

Doxycycline for malaria prophylaxis in Australian soldiers deployed to United Nations missions in Somalia and Cambodia. Mil. Med. 160:443–444. Shmuklarsky, M. J., E. F. Boudreau, L. W. Pang, J. I. Smith, I. Schneider, L. Fleckenstein, M. M. Abdelrahim, C. J. Canfield, and B. Schuster. 1994. Failure of doxycycline as a causal prophylactic agent against Plasmodium falciparum malaria in healthy nonimmune volunteers. Ann. Intern. Med. 120:294–299. Sibley, C. H., K. I. Barnes, and C. V. Plowe. 2007. The rationale and plan for creating a world antimalarial resistance network (WARN). Malaria J. 6:118. Sibley, C. H., K. I. Barnes, W. M. Watkins, and C. V. Plowe. 2008. A network to monitor antimalarial drug resistance: a plan for moving forward. Trends Parasitol. 24:43–48. Smith, A. F. M., and G. O. Roberts. 1993. Bayesian computation via Gibbs sampler and related Markov chain Monte Carlo methods (with discussion). J. R. Stat. Soc. B 55:3–23. Stepniewska, K., K. Chotivanich, A. Brockman, N. P. J. Day, and N. J. White. 2007. Overestimating resistance in field testing of malaria parasites: simple methods for estimating high EC50 values using a Bayesian approach. Malaria J. 6:4. Trape, J. F., G. Pison, M. P. Preziosi, C. Enel, A. Desgreés du Lou ˆ, V. Delauney, B. Samb, E. Lagarde, J. F. Molez, and F. Simondon. 1998. Impact of chloroquine resistance on malaria mortality. C. R. Acad. Sci. Paris 321: 689–697. Wallace, M. R., T. W. Sharp, B. Smoak, B., C. Iriye, P. Rozmajzl, S. A. Thornton, R. Batchelor, A. J. Magill, H. O. Lobel, C. F. Longer, and J. P. Burans. 1996. Malaria among United States troops in Somalia. Am. J. Med. 100:49–55. Weiss, W. R., A. J. Oloo, A. Johnson, D. Koech, and S. L. Hoffman. 1995. Daily primaquine is effective for prophylaxis against falciparum malaria in Kenya: comparison with mefloquine, doxycycline, and chloroquine plus proguanil. J. Infect. Dis. 171:1569–1575. Wernsdorfer, W. H. 1991. The development and spread of drug-resistant malaria. Parasitol. Today 7:297–303. Willerson, D., K. H. Rieckmann, P. E. Carson, and H. Frischer. 1972. Effects of minocycline against chloroquine-resistant falciparum malaria. Am. J. Trop. Med. Hyg. 21:857–862. Yeo, A. E., K. H. Rieckmann, and R. I. Christopherson. 1998. Indirect inhibition by antibiotics of nucleotide and deoxynucleotide biosynthesis in Plasmodium falciparum. Southeast Asian J. Trop. Med. Public Health 29: 24–26.