OUP user menu

A molecular map of chloroquine resistance in Mali

Abdoulaye A. Djimde, Breanna Barger, Aminatou Kone, Abdoul H. Beavogui, Mamadou Tekete, Bakary Fofana, Antoine Dara, Hamma Maiga, Demba Dembele, Sekou Toure, Souleymane Dama, Dinkorma Ouologuem, Cheick Papa Oumar Sangare, Amagana Dolo, Nofomo Sogoba, Karamoko Nimaga, Yacouba Kone, Ogobara K. Doumbo
DOI: http://dx.doi.org/10.1111/j.1574-695X.2009.00641.x 113-118 First published online: 1 February 2010


Plasmodium falciparum chloroquine resistance (CQR) transporter point mutation (PfCRT 76T) is known to be the key determinant of CQR. Molecular detection of PfCRT 76T in field samples may be used for the surveillance of CQR in malaria-endemic countries. The genotype-resistance index (GRI), which is obtained as the ratio of the prevalence of PfCRT 76T to the incidence of CQR in a clinical trial, was proposed as a simple and practical molecular-based addition to the tools currently available for monitoring CQR in the field. In order to validate the GRI model across populations, time, and resistance patterns, we compiled data from the literature and generated new data from 12 sites across Mali. We found a mean PfCRT 76T mutation prevalence of 84.5% (range 60.9–95.1%) across all sites. CQR rates predicted from the GRI model were extrapolated onto a map of Mali to show the patterns of resistance throughout the participating regions. We present a comprehensive map of CQR in Mali, which strongly supports recent changes in drug policy away from chloroquine.

  • malaria
  • drug resistance
  • surveillance
  • molecular methods


Twelve years after the introduction of chloroquine in the 1940s, Plasmodium falciparum resistance to the drug developed and spread steadily throughout malaria-endemic regions. From Asia and South America, then into much of Africa, the once-efficacious medication has grown increasingly ineffective (Wellems & Plowe, 2001; Wellems, 2002). Chloroquine is no longer recommended as the first-line treatment in Africa, but the readily available and inexpensive drug persists throughout the continent (Hopkins et al., 2007; Etuk et al., 2008). Therefore, tracking parasite susceptibility to chloroquine may still have important public policy implications as well as research and clinical utility.

The standards for measuring chloroquine resistance (CQR) are the World Health Organization's (WHO's) in vivo drug efficacy tests (WHO, 1973, 1996, 2003). However, these tests require 2–4 weeks to perform and are difficult to interpret in endemic areas where transmission is high. In addition, they require multisite, representative, and repeated studies with sufficient sample size to detect changes. Alternative tests have included a variety of in vitro tests (Rieckmann et al., 1978; Desjardins et al., 1979; Druilhe et al., 2001). They all involve venous blood to be drawn and some level of parasite cultivation that present high rates of assay failure even in highly equipped laboratories and in expert hands (Olliaro & Bloland, 2001). Therefore, neither the in vivo tests nor the currently available in vitro tests are suitable for large-scale epidemiological surveys. Point mutations in the P. falciparum CQR transporter gene, pfcrt, have been shown to be the primary molecular basis of CQR (Fidock et al., 2000). Molecular methods were developed to detect these mutations in field samples. Previous studies showed that the prevalence of the PfCRT K76T mutation was consistently higher than the prevalence of in vivo CQR. However, prospective chloroquine efficacy studies at different sites in Mali found that age-adjusted genotype-resistance indices (GRIs), which are obtained as the ratio of the prevalence of PfCRT 76T to the incidence of CQR in a clinical study, were remarkably stable (mean GRI=2.1). Thus, in different settings in Mali, molecular surveys for PfCRT K76T could potentially be used to predict rates of in vivo CQR (Djimde et al., 2001a). If validated in new settings of Mali and in other countries, this GRI model represents a potentially powerful and novel tool for the detailed mapping of CQR at the country or regional level. To this end, we set out to confirm the GRI model through molecular and in vivo data from a variety of villages in Mali and in the West African subregion through a review of the literature.

Materials and methods

Model verification

In order to confirm the GRI model across population, time, and resistance patterns, we compiled data from the literature, and new data from our team, and from other Malian scientists. New blood samples blotted onto filter papers were collected from Kollé and Bandiagara, two of the original sites used in the development of the GRI model. Additional samples were collected in the village of Faladiè, a new field site for our team. Faladiè (3000 inhabitants) is a rural village located at 80 km north-west from Bamako. Malaria transmission is seasonal, with a peak in October. The highest rates of in vivo CQR in Mali (60%) were documented in that site in 2003 and rates of PfCRT 76T were above 80% (Sangho et al., 2004). These samples were taken from patients presenting with symptoms consistent with uncomplicated falciparum malaria in 2002 and 2003, i.e. 5–7 years after the initial report of the GRI model.

We gathered additional filter papers from the villages of Mantéourou and Bancoumana, two sites used by other Malian investigators. In addition to the collection of blood filter papers from cases of fever, field clinicians kept detailed accounts of the level of CQR seen in vivo in the locality using the WHO in vivo protocols (WHO, 2003).

DNA was extracted from each sample and the PfCRT K76T genotype was determined by nested PCR as described previously (Djimde et al., 2001b). PCR banding was grouped as no band, wild type, mutant, and mixed. Mixed infections were considered to have the mutant phenotype. Samples yielding no band were excluded from further analysis.

To further test the validity and applicability of the model, the published literature was searched for studies that had compared the prevalence of the PfCRT 76 marker with in vivo resistance patterns. Inclusion criteria for the literature reviews were that the data be published in a peer-reviewed journal and have both data on in vivo CQR and molecular data on the PfCRT 76T mutation from the same geographic location. The reported PfCRT 76T prevalence had to be <100%.

Data analysis

From the new Mali molecular and in vivo data, we calculated the ‘predicted in vivo CQR’ using the published GRI formula as follows: PfCRT 76T prevalence/2.1=predicted in vivo CQR. This predicted in vivo resistance was then compared with the actual observed day 14 in vivo CQR for that same site and year. This process was repeated for the data extracted from the literature.

Using the model to predict in vivo CQR

Following the verification of the GRI model, the next step was to field test the model and develop a map of CQR in Mali. Wanting to test the method in the most practical and real-world settings, villages that had yet to be included in research studies were targeted. Importantly, these villages had no prior association with our team. The association of Malian ‘bush doctors,’Médecins de Campagne, who provide medical care in rural areas throughout Mali, was invited to participate in the study. These medical doctors practice in rural villages where neither electricity nor running water is available. They perform their rounds in a radius of 5–15 km around their headquarters on foot, bicycles, or motorcycles. Diagnosis of malaria and other illnesses is solely based on clinical symptoms as no laboratory is available. Eighteen physicians from 18 villages widely dispersed throughout the Sahel region of Mali to Bamako, the capital city of Mali, were invited to participate in this pilot implementation of the GRI model. These physicians were asked to participate in the collection of blood spot filter papers for use in molecular testing. They received a day-long training session, which included information on the molecular basis of CQR, the importance of accurate drug resistance monitoring, and a detailed description of their role within the study. They were instructed on proper informed consent and sampling techniques. At the end of the training session, each physician was given a ‘GRI kit’ containing adequate gloves, alcohol, filter paper, sterile lancets, pencils, and tape to allow for the collection of 100 samples. They were instructed to offer participation and collect a filter paper for each individual presenting with a suspected case of uncomplicated malaria based on their clinical evaluation. No other data were collected from the study participants.

Filter papers were then sent from these villages to the Malaria Research and Training Center (MRTC) laboratory for molecular analysis. Once at the laboratory in Bamako, Mali, the same technique as that described in the model verification above was used to detect the PfCRT 76T mutation.

The study protocol received ethical clearance from the Ethics Committee of the Faculty of Medicine, Pharmacy and Odonto-Stomatology, University of Bamako.

Map development

Once all of the season's samples were collected and analyzed, the GRI model was used to calculate the predicted level of CQR by village. The data were then extrapolated onto a map of Mali to show the patterns of resistance throughout the participating regions. In addition to the participating villages, all villages that were used in the confirmation of the model were mapped. Data published in 2004 from the region of Kidal where the GRI model was first field tested in Mali and used to inform policy in the management of an epidemic of malaria were also mapped (Djimde et al., 2004).


From the five Malian villages, a total of 951 blood filter paper samples were collected. The rate of PfCRT 76T ranged from a low of 32.4% in Mantéourou to a high of 91.5% in Bancoumana. The predicted in vivo resistance and the actual observed in vivo resistance are shown in Fig. 1. The difference between the observed and the predicted in vivo resistance ranged from a low of 3.9 to a high of 14.3 percentage points.


Predicted and observed in vivo resistance based on the prevalence of the PfCRT 76T mutation using the GRI model (GRI=2.1).

The review of the literature yielded three relevant studies from three West African countries: Mauritania (Jelinek et al., 2002), Ghana (Mockenhaupt et al., 2005), and Burkina Faso (Tinto et al., 2003), with a total representation of six villages. We compared the prevalence of the PfCRT 76T marker and levels of in vivo resistance. Of the six representative villages, the difference between the predicted and the observed CQR values ranged from 0.2 to 19 percentage points (Fig. 2). The Burkina Faso village of Bama was noted to have a ratio of PfCRT 76T to in vivo resistance that approached 1. These studies reported their own calculated GRI as follows: the Mauritanian study reported a calculated GRI of 1.7, the study in Ghana used the genotype-failure index (GFI) model reporting a ratio of 2.8 for clinical failure, and the Burkina Faso data from four villages yielded a range of GRI from 1.1 to 3.0. The Ghana authors used the GFI, which is calculated using the same formula as the GRI, but uses rates of clinical failures (WHO, 1996) instead of rates of parasitological resistance (WHO, 1973). The simple nonweighted average for all sites in our review of the literature review was 2.0.


Prevalence of the PfCRT 76T mutation and the predicted and observed rate of in vivo resistance across three West African countries based on the Mali GRI of 2.1. *Villages from Burkina Faso.

Of the 18 invited rural doctors, 13 made the trip to Bamako. Of the 13 who attended the training session, 12 (92.3%) sent filter paper samples. From these 12 participating Médecins de Campagne, we collected a total of 1260 filter paper samples from 12 villages throughout the Sahel region of Mali. Filter papers from one village failed to yield a PCR product and was removed from further analysis. The post-PCR results are detailed in Table 1. There was a total ‘no band’ (no amplification despite a nested PCR) rate of 41.1% (range 1.6–73.9%). It should be noted that the typical rate of PCR failure in our laboratory is <6% (data not shown). From the remaining 645 samples, we found a prevalence rate of 84.6% (range 60.9–95.1%) of the PfCRT 76T mutation across all sites (see Table 1). We combined these data with our previous results from other Malian sites used in the model verification to construct a new map of CQR village by village across Mali (Fig. 3).

View this table:

Prevalance of PfCRT 76 wild-type and mutant alleles and the subsequent predicted in vivo CQR in 11 selected Malian villages

SitesWild type PfCRT K76 (%)Mutant type PfCRT 76T (%)Predicted CQR (%)

Map of GRI predicted in vivo resistance to chloroquine in Mali.


We present the most comprehensive map of CQR in Mali to date. This map was made in 2005, when the Malian National Malaria Control Program was reviewing the antimalarial treatment policy. It demonstrated the extent of CQR in Mali and contributed to the decision to move the first-line antimalarial drug from chloroquine- to artemisinin-based combination therapies. These data support other studies that showed that in areas with low to moderate transmission patterns, the rate of PfCRT 76T in the population is roughly two to three times the observed in vivo resistance (Jelinek et al., 2002; Tinto et al., 2003; Djimde et al., 2004; Mockenhaupt et al., 2005). We did not have data to test the limits of the GRI as no village had PfCRT 76T prevalence >96%. Other sites with a higher PfCRT 76T prevalence, such as the Congo, have shown that as PfCRT 76T reaches 100%, the predictive value of the GRI diminishes (Mayengue et al., 2005). However, it is important to note that the WHO advocates for a change in drug policy once in vivo resistant rates to a drug reach 25%. Importantly, the GRI is most accurate in areas where it stands to be of greatest clinical and political utility. Day 14 in vivo CQR data were provided here because the GRI model was developed when the norm was to use day 14 drug efficacy in the field.

We found that the prevalence of the PfCRT 76T mutation varied widely throughout the country from just over 30% to well over 90%. These data strongly support changes in drug policy away from chloroquine. Indeed, since 2006, the Malian government's official policy is to use artemisinin-based combination therapy (ACT) as the first-line treatment for uncomplicated malaria. To this end, the country currently provides free ACTs to children under 5 presenting with uncomplicated malaria.

After a nested PCR amplification of DNA extracted from filter papers obtained from presumptive cases of malaria, there was a mean rate of absence of PCR product of 41.1% (range 1.6–73.9%). Although PCR amplification could fail for a number of reasons, the great majority of these samples would contain no malaria parasites. This highlights the problem of overdiagnosis of malaria in rural settings where the treatment decision is solely based on clinical symptoms. These data were used to convince the National Malaria Control Program of Mali to require malaria diagnosis with microscopy or rapid diagnostic tests (RDTs) before the prescription of ACTs to children <5 years of age. The use of RDTs is now part of the new guidelines for the management of malaria in Mali.

Molecular epidemiology of the PfCRT K76T has also been used to guide research efforts. Laufer and colleagues used the diminishing prevalence of the PfCRT 76T marker in Malawi as a basis for a clinical trial that showed that 13 years after the withdrawal of chloroquine from the Malawian market, the drug's clinical efficacy returned. The authors showed that in the absence of the drug's selection pressure, the PfCRT 76T mutation could disappear in a population, inferring that the resistance trait did not incur a fitness advantage on the parasite. The authors were careful not to advocate a return to chloroquine to the Malawian market, but rather cautioned that chloroquine may eventually see a return of efficacy on the continent should the drug be wisely removed from the continent's markets (Laufer & Plowe, 2004; Laufer et al., 2006). Figure 1 shows that the prevalence of PfCRT 76T in Bandiagara increased from 39% (n=47) in 2002 to 68.9% (n=75) in 2003 (P<0.001) while in Kollé the prevalence of PfCRT 76T decreased from 85% (n=127) in 2002 to 64.5% (n=172) in 2003 (P<0.001). The only drug used in uncomplicated malaria in Bandiagara in 2002–2003 was chloroquine, which may have resulted in the increased prevalence of PfCRT 76T in that site. In contrast, in Kollé, a small rural village, there was a randomized trial that compared chloroquine, amodiaquine, and sulfadoxine–pyrimethamine (Tekete et al., 2009). Because both amodiaquine and sulfadoxine–pyrimethamine were highly efficacious on P. falciparum in Kollé, the PfCRT 76T mutant had no fitness advantage, which may explain the observed decrease in the prevalence of that allele in this village. Taken together, our data and the available literature demonstrate the power of drugs in shaping malaria parasite populations.

Today, despite chloroquine's diminished clinical efficacy, the drug continues to be used as an inexpensive malaria treatment in many parts of the world, notably in sub-Saharan Africa, as often treatment alternatives are not available or are beyond the means of the population (Djimde et al., 2004; Mayengue et al., 2005). Indeed, given the enormous price difference between chloroquine and ACTs, changes in prescribing patterns and pharmacy stocks are likely to take time. Nevertheless, most African countries with high levels of P. falciparum resistance to chloroquine are currently changing policy away from the use of chloroquine (Laufer et al., 2007). With the change in prescribing patterns from chloroquine to other antimalarials, tracking resistance patterns to these drugs will become increasingly important. While the GRI model may not fit with all antimalarial drugs or all malarious areas, the molecular approach used in our model may serve as a foundation on which other markers and resistance patterns can be developed and tracked. It remains to be seen whether this shift in policy will eventually lead to a decline in the prevalence of mutant PfCRT and in turn in vivo CQR. Certainly, data describing Malawi's transition away from chloroquine support this theory (Kublin et al., 2003; Laufer & Plowe, 2004; Laufer et al., 2006). Indeed, the GRI may be used to help track how successful the country's efforts have been in shifting the drug policy away from chloroquine.

Perhaps, most importantly, this study demonstrated that with very limited resources, an efficient, cost-conscious, and scientifically sound drug resistance monitoring project is feasible.

From the outset there were numerous obstacles to address: education and training of physicians not familiar with molecular techniques, limited financial resources, project initiation in villages not familiar with medical research, and most importantly the transportation of filter papers back to the laboratory in Bamako. Most of the villages used in the project were far from principal roads. The samples arrived by whatever means possible, including by foot, motorcycle courier, bus, plane, and public transport, often by way of a veritable relay of transport from villages along the Sahel belt of southern Mali to as far away as to the North. Despite these obstacles, we show that drug resistance tracking using a molecular biology tool is possible even in the world's poorest countries. Indeed, by maximizing the resources that were already in place, notably the Médecins de Campagne and the laboratory of MRTC in Bamako, the foundation for a country-wide drug resistance monitoring program was laid.


We thank the study populations for their participation in this study. We thank ‘Association des Médecins de Campagne’ for their contribution to this study. Financial support was provided by grants from MIM-UNICEF-UNDP-World Bank-WHO Special Programme for Research and Training in Tropical Diseases (TDR) Grant # A20238, NIAID/NIH Supplement Award 5 RO1 AI44824-03, and International Atomic Energy Agency (IAEA) RAF/6025. A.A.D. is supported by European and Developing Countries Clinical Trail Partnership Senior Fellowship (Grant # 2004.2.C.f1) and Howard Hughes Medical Institution International Scholarship (Grant # 55005502). B.B. was supported by Fogarty International Center grant # TW007988 to Vanderbilt University. The Bancoumana site is supported by the MVDB/NIAID intramural grant for malaria vaccine testing site and the Mantéourou site is supported by the BioMalPar grant. PCR controls were provided by MR4.


  • Editor: Monique Capron


View Abstract