A high rate of treatment failure for commonly anti-malarial drugs used in Plasmodium falciparum infections has been reported in numerous endemic areas. The recommended treatment policy is now to use drug combinations 1. The atovaquone-proguanil (AP) drug combination, distributed under the trade name of Malarone^®, is one of the treatment and prophylaxis options. AP has a high cure rate, limited mild side effects 2 and proved efficacious against multi-drug resistant P. falciparum malaria 34. Atovaquone (a hydroxy-naphthoquinone) is a potent inhibitor of the cytochrome bc1 (cytbc1) complex 5678, a key respiratory enzyme from the mitochondrial membrane, while proguanil (an isopropylbiguanide) inhibits the plasmodial dihydrofolate reductase 4910. The association synergizes to collapse the mitochondrial membrane 811.

Plasmodium falciparum in vitro resistance to atovaquone has been associated with specific point mutations in the cytochrome b gene (Pfcytb) in the region spanning codons 271–284 712. A high frequency of recrudescence was observed in patients receiving atovaquone as a single drug therapy against P. falciparum 1314. A Y268S point mutation in the Pfcytb gene, distinct from the mutations observed in the lines selected in vitro for atovaquone resistance, was detected in the recrudescing parasites 7. Codon 268 polymorphism was used as marker for a molecular surveillance of atovaquone-proguanil resistance 1516171819. AP treatment failures were increasingly reported a few years after its introduction, with recrudescing parasites presenting a markedly increased IC₅₀ for atovaquone 152021. In most cases, recrudescence was associated with a mutant 268 codon, either a Y268S 312202223242526, a Y268N 15 or a Y268C mutation 21. However, the presence of a mutant 268 codon was not observed in all cases of AP failure 2127.

Along with the key issue of emergence and spreading of polymorphisms conferring atovaquone resistance, analysis of Pfcytb field diversity presents an interest in population genetics 28293031. Indeed, the cytb gene is encoded by the mitochondrial DNA and as a consequence, is of uniparental inheritance and under quite different evolution constraints compared to nuclear genes 3233. In particular, interallelic recombination is not possible and polymorphisms such as base substitutions or insertions may accumulate over time. In mammals, the cytb locus displays an approximately 10-fold higher mutation rate than nuclear genes 34. A rapid mutation rate (one mutation in 10⁵ parasites) was reported for Pfcytb 35, but 100–1,000 lower rates were described subsequently 7.

Sequence polymorphism of the near to full gene sequence has been explored in laboratory 736 and in vitro resistant isolates 712. It has also been looked for in cases of treatment failures from several countries, mostly African countries 712152122242526. Systematic analysis of field polymorphism in African settings 3738 and in isolates from the Thai Myanmar border 39 essentially focused on the region coding for the atovaquone-binding site. Full gene sequence analysis of field samples has been restricted to isolates from few patients in India 36 and from patients returning to France from West Africa, Central Africa or the Indian Ocean 212540. This provides an interesting picture of the overall polymorphism of the gene, but little clues on possible population signatures related to this gene. Sequence analysis of mitochondrial DNA from 100 independent isolates collected worlwide provided evidence for geographical clustering 2930.

To further document Pfcytb population polymorphism, 671 isolates from six different areas: Africa (Senegal, Ivory Coast, Central African Republic, and Madagascar), Asia (Cambodia) and South America (French Guiana) were AP had not been in regular use were analysed. This identified numerous novel, low frequency polymorphisms in Africa and Cambodia, most of which were country-specific, together with a high frequency signature that was specific for the South American isolates. None of the polymorphisms previously associated with atovaquone resistance in vitro or AP treatment failure was observed in this panel of isolates.

Among the 671 isolates, a full length Pfcytb sequence was successfully established for 576 isolates and partial sequence was obtained for 95 isolates (Figure 1). Overall, 11 polymorphic sites were observed (Figure 1), of which eight were novel mutations compared with published data (Figure 2).

All isolates from Senegal, Central African Republic and Madagascar had the same sequence, which was identical to the Camp/3D7 reference sequence. One synonymous point mutation and two coding mutations were observed in the set of isolates from Ivory Coast (Figure 1). None of these had been published previously and each was observed in a single isolate.

In Cambodia, two synonymous mutations (T42T and L265L) were observed (Figure 1). Each was detected in one isolate, the remaining 177 isolates harboured the Camp/3D7 reference Pfcytb gene sequence.

The French Guiana samples presented the largest polymorphism, with seven mutated sites affecting seven codons. Five alleles were observed (Figure 1). Unlike the other settings where alleles had a single mutated position, three out of the five alleles from French Guiana were multiple mutants with two (allele FG2) or three (alleles FG3 and FG4) mutated codons. The Camp/3D7 reference allele was present in 22 out of 160 isolates (14%). Allele FG1, which carried a silent L230L mutation, was highly predominant, accounting for 135 of 160 (84%) isolates (Figures 1 and 3). The same mutation was associated with additional SNPs in allele FG4, which therefore probably derive from FG1. Allele FG4 which carried additional mutations N7I and L114S, was observed only once (frequency 0.6%). The N7I polymorphism was detected in allele FG3 as well, but in this case it was associated with N2D and I342L (Figure 1).

In the panel of isolates studied here, 11 polymorphic sites were observed, resulting in 11 codons displaying a single point mutation. Ten distinct alleles could be identified, as shown in Figure 1. Apart from the a126t and c688t silent mutations at codon T42T and L230L, respectively, and a581g non synonymous mutation at codon H194R, all mutations observed here were novel (Figures 1 and 2). Five synonymous and six non-synonymous mutations were detected. The observed ratio of synonymous to non synonymous mutations is in line with four synonymous and five non synonymous mutations detected in 270 full gene sequences of isolates from West Africa and the Indian Ocean 25 and four synonymous and six non synonymous mutations in 14 isolates from India 36, but lower than the three to one ratio observed in Gabon 37 and the nine to three ratio reported for 135 isolates from West and Central Africa 40. The compilation of all Pfcytb field polymorphisms identified so far (Figure 2) highlights 50 polymorphic sites, affecting 45 codons, with 30 mutant amino acid residues.

There was a clear geographical heterogeneity in the level and type of polymorphism (Figure 3). The Camp/3D7 type was the single allelic form detected in three African settings studied here (Madagascar, Senegal and Central African Republic). Similar observations were made in Ethiopia 37. In Ivory Coast, the same allele was observed in 93% of the patients, a frequency similar to the 90% prevalence observed in Gabon 37, or in travellers returning to France from West and Central Africa 40. As observed in other African endemic settings 37, the mutant alleles were single mutants and each mutant had a low frequency, being observed in one or two isolates. A similar low frequency of single mutant alleles has been observed in the panel of isolates from patients returning from travel to West and Central Africa and the Indian Ocean 2540.

In Cambodia, the Camp/3D7 wild type allele was also highly predominant, consistent with recent sequence data of the atovaquone binding site in patients from the Thai/Myanmar border 39. The two mutations observed were both synonymous and detected at low frequency.

French Guiana showed a quite different profile, and was the most polymorphic of the six countries explored. Seven of 11 mutant sites were observed in the set of isolates from French Guiana. Furthermore, this area was the only one where multiple mutant alleles were detected. Importantly, the Camp/3D7 wild type allele, which was dominant in the other settings was observed with a 14% frequency only, while the dominant allele (84% frequency) was a single, silent L230L mutant. This was in accordance with data from other localities from South America 2930. Triple mutants, most probably originating from the L230L parent for FG4 isolate, were observed along with a double mutant possibly derived from the Camp/3D7 type. Thus, the French Guiana parasite population had a Pfcytb pattern dissimilar from the African and Cambodian settings. This was consistent with data from others 2930. A specific, high frequency geographical signature seems to prevail in India as well, where 13 or 14 isolates carried a N2N F3Y Y4Y haplotype 36. However, polymorphism in French Guiana was larger than in India, where three allelic forms have been reported. Thus, of all geographical areas studied so far, French Guiana presents the largest Pfcytb gene polymorphism. This is interpreted as a consequence of its population structure, with hypoendemic, isolated foci that are propitious to genetic drift 47.

Pfcytb nucleotide polymorphism was larger than the deduced protein sequence. All three alleles from Cambodia coded for the same, wild type protein sequence. In French Guiana the wild type deduced protein sequence accounted for 158 of 160 alleles. Thus in all geographic regions, the wild type protein sequence was the highly dominant if not the only predicted isoform. Altogether, these data point to an elevated mutation rate of the locus, with albeit a large dominance of the wild type protein sequence, probably indicating its optimal fitness. At the nucleotide level, interesting geographic clustering of molecular signatures were observed (Figure 4), suggesting different ancestral types in French Guiana (and in South America) as well as in India 36. Specificities of the parasite population in India is also suggested by polymorphism of the Pfcrt locus 48. The data reported here further support the evidence that the present parasite population from South America differs from the parasites from Africa, regarding surface antigens 49, loci such as Pfcg2 31, Pfcrt 505152, Pfdhfr and Pfdhps 51 and numerous genome-wide scattered SNPs 53. The large Pfcytb polymorphism together with the observation that the var gene repertoire in Brazilian isolates is small and highly redundant unlike in Africa and Southeast Asia 54, raises the question of the structuring of the South American P. falciparum population.

None of the field polymorphism observed here or in other settings concerned the residues that are mutated in lines selected in vitro for atovaquone resistance 5712. Furthermore, no mutant Y268S, Y268C and Y268N Pfcytb, selected under atovaquone or AP pressure 3712152122242526 was detected in any of the settings explored here. The 268N mutant, which has been observed with a 4.5% frequency in Nigeria in the absence of AP pressure 38 was not detected. This mutation has been reported so far essentially in Nigeria 1538 and, as discussed by Happi et al 38 may have arisen under pressure by related drugs. The Y268S polymorphism was not detected in the isolates from French Guiana studied here, which were collected before implementation of AP as prophylaxis and as second line treatment in 2002. It was however observed one year after implementation in a second line AP-treatment failure 20, further substantiating the interpretation of its selection during AP treatment.

Overall, the available data indicate an elevated mutation frequency of the Pfcytb locus, with multiple polymorphic sites, and in some cases more than one polymorphism per site, consistent with the elevated mutation frequency of mitochondrial genes 3435. The geographical clustering observed here adds further support to the evidence that the parasite population from South America differs from the parasites from Africa. Importantly, the absence of a mutated codon 268 in all settings investigated here indicates that AP remains an interesting treatment option for these areas. In view of the high mutation rate and of the rapid selection of mutants under AP pressure, careful surveillance of emerging Pfcytb mutants and resistance to atovaquone used as prophylaxis in travellers or as treatment is warranted.

MTE did the sequencing of isolates except for Ivory Coast, was responsible for data collection, entry in the database and drafted the manuscript.

RJ (Senegal), MR (Madagascar), EL (French Guiana), DM (Republic Central Africa), SBA, MCH, CR (Ivory Coast) were responsible for sample collection and coordination of laboratory in the field.

NK, EL, MR, RJ, DM and CR performed all PCR amplifications and actively participated in sample collection.

TF established suitable protocols for Pfcytb amplification, conducted the phyologenetic study, participated in drafting the manuscript and was responsible for coordination of laboratory work in Cambodia.

CB was responsible for the sequencing procedure.

OMP conceived the study, helped for sequence analysis, drafted and revised the manuscript.

All authors read and gave the final approval of the version to be published.