The reservoir of Plasmodium infection in humans has traditionally been defined by blood slide positivity. This study was designed to characterize the local reservoir of infection in relation to the diverse var genes that encode the major surface antigen of Plasmodium falciparum blood stages and underlie the parasite's ability to establish chronic infection and transmit from human to mosquito.
We investigated the molecular epidemiology of the var multigene family at local sites in Gabon, Senegal and Kenya which differ in parasite prevalence and transmission intensity. 1839 distinct var gene types were defined by sequencing DBLα domains in the three sites. Only 76 (4.1%) var types were found in more than one population indicating spatial heterogeneity in var types across the African continent. The majority of var types appeared only once in the population sample. Non-parametric statistical estimators predict in each population at minimum five to seven thousand distinct var types. Similar diversity of var types was seen in sites with different parasite prevalences.
Var population genomics provides new insights into the epidemiology of P. falciparum in Africa where malaria has never been conquered. In particular, we have described the extensive reservoir of infection in local African sites and discovered a unique var population structure that can facilitate superinfection through minimal overlap in var repertoires among parasite genomes. Our findings show that var typing as a molecular surveillance system defines the extent of genetic complexity in the reservoir of infection to complement measures of malaria prevalence. The observed small scale spatial diversity of var genes suggests that var genetics could greatly inform current malaria mapping approaches and predict complex malaria population dynamics due to the import of var types to areas where no widespread pre-existing immunity in the population exists.
Citation: Chen DS, Barry AE, Leliwa-Sytek A, Smith T-A, Peterson I, Brown SM, et al. (2011) A Molecular Epidemiological Study of var Gene Diversity to Characterize the Reservoir of Plasmodium falciparum in Humans in Africa. PLoS ONE 6(2): e16629. https://doi.org/10.1371/journal.pone.0016629
Editor: Anne Charlotte Gruner, Agency for Science, Technology and Research (A*STAR), Singapore
Received: August 31, 2010; Accepted: January 6, 2011; Published: February 9, 2011
Copyright: © 2011 Chen et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: The authors would like to acknowledge NIH/Fogarty International TW001503: "Training in Malaria Pathogenesis and Diversity in Africa", Bill and Melinda Gates Foundation/Broad Institute: "P. falciparum Genetic Diversity" and ExxonMobil-HSPH Program on Malaria for support of the Senegal field sample collection. Isolates were collected in Gabon thanks to the European Union INCO Programme (contract no. IC18-CT98-0359). This work was supported by Wellcome Trust Programme grant No. 041354 awarded to KPD, NIH/National Institute of General Medical Sciences grant GM061351-07, and NIH/National Institute of Allergy and Infectious Disease grant R01 AI084156. DSC was supported by a NIH Ruth L. Kirschstein National Research Service Award (F32AI071765). AEB was supported by an Innovation Fellowship from the Victorian Endowment for Science Knowledge and Innovation and a Howard Florey Centenary Fellowship from the National Health and Medical Research Council, Australia. The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Hundreds of millions of Plasmodium falciparum infections persist for many months in the human population to sustain malaria transmission where anopheline mosquito vectors are only seasonally available. These chronic, largely undetected, infections constitute the reservoir of infection and serve to fuel continued malaria transmission.
During the past era of malaria eradication the size and persistence of the reservoir of infection were considered major obstacles to the eradication of P. falciparum in Africa . The duration of a single infection was measured to be 150 to 200 days using models and epidemiological observations of blood smear data . Today, with the advantage of molecular tools, we understand that individuals in endemic areas harbor chronic P. falciparum infections composed of several genetically distinct haploid parasite genomes introduced by single and/or multiple (superinfection) mosquito bites , , . The duration of infection of individual genotypes is influenced by age, prior exposure, coinfection and superinfection , , . Density-dependent regulation of parasitemia has also been shown to occur among Plasmodium spp and genotypes in coinfected semi-immune children . Considering these infection dynamics, the parasite rate (prevalence) as defined by blood slide positivity does not accurately describe the complexity of the reservoir of infection and yet remains the mainstay for monitoring and evaluating malaria control . Consequently, we considered how best to characterize the reservoir of infection in humans using parasite genetics/population genomics.
Diversity measures based on polymorphic antigen encoding genes, barcoding, microsatellite loci and genome-wide single nucleotide polymorphisms , ,  define the number of distinct alleles and/or genomes per infected human, the latter termed multiplicity of infection (MOI). They also define the estimated number of distinct genomes in a population. This molecular epidemiological information, while useful to evaluate changes in transmission intensity, fails to capture any specific information about the reservoir. As the ability of the parasite to persist by clonal antigenic variation (i.e. achieve a chronic infection) plays a key role in transmission from human to mosquito, we propose that the diversity of the genes underlying this phenotype provides a genetic fitness estimate of the reservoir, to complement existing molecular surveillance for drug resistance mutations , genes encoding vaccine antigens , , and microsatellite loci .
P. falciparum Erythrocyte Membrane Protein-1 (PfEMP1), encoded by the var multigene family, is considered the major polymorphic surface antigen of the infected erythrocyte involved in clonal antigenic variation (reviewed in , ). PfEMP1, expressed during maturation by trophozoite (blood stage) and early gametocyte (transmission stage) parasites , mediates binding to endothelial cell receptors in the deep vasculature of host tissues , . To achieve chronic infection by immune evasion, parasites periodically switch PfEMP1 isoforms displayed on the erythrocyte membrane by differential expression of up to 60 members of the var multigene family found throughout each parasite genome , , . This process of clonal antigenic variation is believed to be responsible for parasite persistence and chronic infection seen in naturally infected individuals, as depicted in Figure 1 where a chronically-infected individual displays successive waves of parasitemia mediated by antigenic switching.
Asexual P. falciparum parasitemia followed over time in a naturally infected Puerto Rican child. The parasitemia follows a pattern of recurrent peaks that decline in amplitude with time. The parasitemia in this child, believed to be a clone, lasted nearly 800 days. These successive peaks of parasitemia are consistent with antigenically distinct waves of parasitemia in P. falciparum infection believed to be mediated by PfEMP1 that allow for parasite persistence . The intra-host dynamics of parasitemia observed in semi-immune children  and induced human infections  are best explained by variant-specific immunity to PfEMP1 variants encoded by the var multigene family  rather than by immunity to single-copy antigen genes. Figure composed using data from .
To characterize the reservoir of infection, we investigated the molecular epidemiology of the var multigene family in three local African parasite populations representing a range of epidemiologic characteristics (Table 1) with an expectation that we would see more var diversity in high transmission sites. Bakoumba, Gabon is a rural village of ∼3000 people in the South East in an area of meso- to hyperendemic P. falciparum. Pikine, Senegal is a suburb east of Dakar city with ∼770,000 inhabitants living in an area of hypoendemic malaria. Kilifi, Kenya is a rural district of ∼600,000 people on the East Coast, where malaria is hypo- to mesoendemic. Var gene diversity data were collected using a previously published population genomic framework . We set out to determine i) the extent of var gene diversity in three African sites, ii) the level of var gene differentiation across the African continent, and iii) the population structure of var genes in Africa.
Var genes are composed of multiple Duffy-binding-like (DBL) and cysteine-interdomain-rich (CIDR) domains that adhere to host tissues to enhance parasite survival and contribute to pathogenesis . DBLα, as the only domain found in nearly all var genes , is the most informative molecular marker of var gene diversity. We applied the sampling framework to survey the var DBLα domain in 29–30 natural P. falciparum isolates from each of three African populations. Using a degenerate PCR strategy, we randomly sampled var genes by amplifying, cloning, and sequencing approximately 100 copies of the DBLα domain from each isolate in the population samples . Within each isolate, we eliminated redundancy (repeated sampling of the same var gene) by defining distinct var genes as those with less than 96% nucleotide identity to any other DBLα sequence. At the population level, we designated distinct ‘var’ types by clustering the var genes, again using the 96% identity rule . Previous studies have shown that DBLα sequences share on average ∼45–66% amino acid identity, and the vast majority of sequences share <75% amino acid identity (e.g. , ). We compared the diversity in African var DBLα sequence data with published DBLα data from two local, non-African populations: a cluster of Amele villages in Papua New Guinea (PNG) , and the municipality of Porto Velho (and suburbs), in the Western Amazonian state of Rondônia, Brazil .
Materials and Methods
Informed consent was obtained from all participants in the study. All data were analyzed anonymously. For the study in Gabon, ethical permission was obtained from the Ethics Committee for Research in Human Medicine, International Center for Medical Research, Gabon, and NYU School of Medicine Institutional Review Board. The informed consent procedure for the study consisted of a presentation of the aims of the study to the community. Individuals (or their parents) were asked to volunteer. At the time of collection, the purpose and design of the study were explained to each individual and verbal informed consent was collected by a minimum of two people. The verbal consent process was consistent with the ethical expectation at the time of enrollment. The ethics committees specifically approved this procedure. For the study in Senegal, ethical permission was obtained from the Ethics Committee of Cheikh Anta Diop University in Dakar and NYU School of Medicine Institutional Review Board. A verbal form of consent was chosen by the local scientific team as they considered it to be the most appropriate method of consent for this type of study. Verbal informed consent was obtained by the clinician reading the consent form to each participant. After verbal informed consent was obtained, the clinician signed and documented consent on the written consent form. The ethics committees specifically approved this procedure. For the study in Kenya, ethical approval was given by the Kenya Medical Research Institute (KEMRI) National Ethical Review Committee and NYU School of Medicine Institutional Review Board. Written consent was obtained from each participant.
Study sites and populations
We compared three epidemiologically and geographically distinct P. falciparum populations of Central Africa (Bakoumba, Gabon), West Africa (Pikine, Senegal) and East Africa (Kilifi, Kenya). Bakoumba, Gabon isolates were collected from volunteers from the rural village of Bakoumba in the Haut-Ogooue Province . This village consists of around 3000 people within a 10 km radius. Bakoumba is meso- to hyperendemic for P. falciparum with highly seasonal transmission. The entomological inoculation rate (EIR) during a peak transmission time has been measured as 0.83 infective bites/person/night , with an estimated mean EIR of 0.28 infective bites/person/night over a full year . Isolates were collected from 604 children aged between 6 months to 10 years, presenting with non-clinical (asymptomatic) malaria in May 2000 . We sampled var genes from 29 of these isolates.
Pikine, Senegal isolates were collected from individuals attending an outpatient clinic in Pikine, a suburb to the East of Dakar city. The population of Pikine near the time of the study was around 770,000 people living within a 10 km radius. In this area, malaria is hypoendemic with seasonal transmission between August and December, and an estimated annual EIR range of 0.014 to 0.86 infective bites/person/year (averages to a daily EIR of 0.00004 to 0.002 infective bites/person/night) . Isolates were collected as part of malaria treatment studies from adults and children 5 years of age or greater presenting with symptoms consistent with mild malaria including fever, chills, and headache, and a positive malaria blood smear . Samples were collected between 2000 and 2005. We sampled var genes from 29 of these isolates.
Kilifi, Kenya isolates were collected from individuals from the rural District of Kilifi, located in the Coast Province of Kenya. Kilifi District, at the time of the study, had a population of around 600,000 people living in rural villages within 75 km of each other. The estimated population from which the study participants came was around 350,000 people at the time of the study. Malaria is endemic in the area, with two annual peaks of transmission and subsequent disease in May to July, and in November . Mean daily EIR varies by specific location, ranging from 0.000–0.145 infective bites/person/night . Blood samples were taken from children (aged 2-104 months) admitted to the pediatric ward of the Kilifi District Hospital with a final diagnosis of non-severe malaria (no prostration, no respiratory distress, no prolonged seizures or coma). Samples were collected in 2002. We sampled var genes from 30 of these isolates.
We selected for inclusion in this study isolates identified as single-genome infections by genotyping of Merozoite Surface Protein 2 (msp2) . This biased the investigation towards younger children in the study populations. In summary, Kilifi isolates were from hospitalized, non-severe clinical cases from a large rural catchment with transmission ranging from low to high; Pikine isolates were from outpatient, non-severe clinical cases from an urban population with low transmission; and Bakoumba isolates were from asymptomatic, non-clinical cases from a smaller, rural catchment with high transmission. Together, these populations represent in a wide range of malaria epidemiologies in Africa.
DNA extraction and genotyping
Genomic DNA for each isolate was obtained by extraction of dried blood spots or whole blood with the QIAamp DNA Blood Mini Kit (Qiagen) according to the manufacturer's instructions. Isolates used for var gene sampling were identified as single genotype infections by genotyping of msp2 using PCR amplification and restriction enzyme digestion . In addition, we genotyped each isolate at 12 microsatellite repeat loci amplified by PCR with fluorescent-labeled primers . Microsatellite alleles were differentiated by DNA fragment length as determined on an ABI 3700 sequencer and scored using GeneMapper software (Applied Biosystems). Some field isolates used in the study contained two genotypes when both methods were used. These methods confirmed that no two African isolates in the study were genotypically identical.
High throughput var DBLα sequencing of African isolates
Var gene sampling based on the DBLα domain was performed as described elsewhere . Briefly, DBLα domains were amplified from individual isolate genomic DNA samples using degenerate primers to blocks B, D, and/or H . For eight of the 29 Bakoumba isolates, DBLα domains were amplified using primers to blocks D and H (Fwd-AGRAGYTTYGCNGAYATHGG Rev-AACCAYCTYAARTAYTGNGG) with PCR conditions as published . For the remaining 21 Bakoumba isolates, DBLα domains were amplified using primers to blocks B and H (Fwd-GCMTGYGCDCCRTWYMGAMG Rev-TCKGCCCATTCYTCRAACCA; primer sequences provided by J. Smith). PCR conditions for these primers were as follows: 2 µl isolate genomic DNA, 1x Ampli Taq Gold reaction buffer (Applied Biosystems), 2.5 mM MgCl2, 0.15 mM dNTPs, 20 pmoles of each primer, 2.5 Units Ampli Taq Gold (Applied Biosystems). PCR cycling was carried out on a Perkin Elmer 9600 thermal cycler and involved 1 cycle of 95°C for 10 mins; 35 cycles of 95°C for 5 secs, 50°C for 20 secs and 60°C for 45 secs; 1 cycle of 60°C for 2 mins. For the Kilifi and Pikine isolates, DBLα domains were amplified using the primers to blocks B and H noted above. PCR conditions were 1 µl genomic DNA, 0.15 mM dNTPs, 2.5 mM MgCl2, 20 pmoles of each primer, 2.5 units Hotmaster Taq DNA Polymerase (Eppendorf). Reactions were performed in an Eppendorf EP Gradient Mastercycler thermal cycler at 94°C for 2 min; 35 cycles of 94°C for 5 sec, 50°C for 20 sec, and 60°C for 45 sec; and a final cycle at 60°C for 2 min. PCR bands of between 450–700 bp were purified using agarose gel electrophoresis and extraction with the Qiaquick Gel Extraction kit (Qiagen) according to manufacturer's instructions. PCR products were cloned and at least 96 clones sequenced with Universal forward and reverse primers as described previously . Sequences from Gabon have been deposited in GenBank under the following accession numbers: DQ134044-6; DQ134086-92; DQ134281-2; DQ134514-5181; DQ135183-232; DQ135243-8; DQ135261-8; DQ135349-51; DQ135354; DQ135382; DQ135400; DQ135406-7; DQ135409-22; DQ135426-41; DQ135444-48; DQ135460-597. Sequences from Kilifi and Pikine have been deposited under the accession numbers HQ732288-HQ733853.
Var sequence data from Papua New Guinea and Brazil
Var sequence data from all 30 isolates comprising the ‘local population’ from Amele, Papua New Guinea  were downloaded from GenBank. Accession numbers for downloaded Amele sequences are listed in Table S1. Var sequence data from 42 isolates from Porto Velho (and suburbs), Brazil ,  were downloaded from the supplementary online file Supplemental Table 4 (applic6.txt) in . Of the 43 Porto Velho isolates described , we excluded isolate S20 collected in 1985 and included the remaining 42 isolates collected from 2001–2006. A list of the 42 isolates included in the analysis can be found in Table S2. These sequence data encompass a region spanning the semi-conserved blocks D and H of the DBLα domain.
Var sequence analysis
We have previously published our framework for analyzing var sequence data . For comparability, block B-H var sequences were trimmed at the 5′ end before analysis so that only the block D-H region was compared. Quality controlled var gene sequences were aligned at 96% sequence identity using Sequencher 4.6 software (Genecodes) to remove redundancy among the reads from each isolate. We classified var sequences into distinct types through alignment using Sequencher 4.6 with a 96% sequence identity cutoff to identify matching sequences among isolates.
Twelve microsatellite markers were genotyped in P. falciparum isolates from Kilifi, Kenya (n = 47) Pikine, Senegal (n = 30); Bakoumba, Gabon (n = 24) were genotyped as previously described . The isolates used for microsatellite analysis included 30 Kilifi isolates, 29 Pikine isolates, and 24 Bakoumba isolates used in the var gene analysis. Microsatellite haplotypes were defined for each parasite isolate and used for population genetic analysis. Microsatellite alleles were scored using GeneMapper software (Applied Biosystems). For the population genetic analyses of microsatellites we used the software program FSTAT .
Cumulative diversity curves
Smoothed cumulative diversity curves, analogous to species accumulation curves, were plotted on Microsoft Excel with averaged data generated from EstimateS 8.0 software  set to 50 runs.
Non-parametric statistical estimates of richness, Chao1 and Abundance-based Coverage Estimator (ACE), and 95% confidence intervals were calculated using EstimateS 8.0 . The Chao1 statistic  estimates the total number of types in a population using frequency data on types seen once only and types seen twice only. While the Chao1 estimator relies on singleton and doubleton data to estimate unseen species, the ACE estimator  utilizes all frequency data, divided into abundant types and rare types. We applied a cutoff value for rare groups of 10 as suggested . Chao1 estimates were derived as a lower-bound estimate of richness, whereas ACE better approximates a point estimate of richness . In the setting of equal probability of distribution and sampling of types, the Chao1 estimator yield a point estimate of richness. These estimators cannot predict a probable maximum richness. Curves depicting the stability of richness estimates with sample size were calculated using EstimateS .
Frequency plots and rank abundance curves
Frequency plots and rank abundance curves  were tabulated and plotted using Microsoft Excel.
Estimates of the number of genomes in survey populations
An approximation of the number of circulating genomes in a survey population was calculated as follows:In this case, the survey population refers to the age-specific population from which P. falciparum isolates were taken, and the population size reflects the estimated population of that age group. This survey population size does differentiate between symptomatic and asymptomatic individuals, but rather includes all individuals of the specified age range. Population size estimates for each geographic unit were obtained from multiple sources as indicated in the Table 1. These population sizes were then adjusted to reflect the age-group specific population using published data on the age structure of each population. Sources for estimates of parasite prevalence and of MOI are referenced in Table 1. Data on parasite prevalence and MOI were selected to match to specific age groups and geographic locations surveyed where possible.
Estimates of further sampling efforts
Using the ACE estimate of var type richness and a statistical predictor of the number of new species or types encountered in further sampling of a population , we estimated the proportion of var types in a population that will be identified given a range of sample sizes in each population. The model uses frequency information from ‘rare’ types, and we have used the suggested frequency cutoff of 10 for rare types . The projections account for sampling of multiple infections in the population. Calculations were performed using the ‘Multinomial Model’ implemented in the software package SPADE , and plotted on Microsoft Excel.
For each African isolate, the sampling yielded high quality DNA sequence data for a median of 83 (range 12–181) var clones. This sampling provided approximately 50% (median 24, range 3–47 unique var DBLα D–H tags) of the estimated 50–60 var genes per genome (Table 2). We measured depth of var gene sampling in each population with a cumulative diversity curve  depicting the rate at which new var types were identified with the collection of unique sequences from each isolate. Despite identifying ∼700 unique var sequences per African population, many var types remained undiscovered in each population as indicated by failure of the curves to plateau (Figure 2A). In contrast, sampling of the non-African populations of Amele and Porto Velho was more complete (Figure 2A).
A) Cumulative diversity curves for each of the five populations. These averaged curves plot the cumulative number of var types observed with successive sampling of var sequences. A well-sampled population will show a curve that levels off and approaches an asymptote, that would approximate the total number of types in the population. The curves from the three African populations (Bakoumba, Pikine, Kilifi) did not show evidence of leveling off, in contrast to the curves from Amele, Papua New Guinea  or Porto Velho, Brazil . Sampling of the African populations has not yet begun to approach the limits of diversity. B) Sharing of var types among the three African population samples. Between 26 and 41 var types were shared among any two population samples; only 10 var types were found in all three populations. C) The majority of var types in each continent were not found in the other continents. Only 5 var types were found in all three continents. Samples from Bakoumba, Pikine, and Kilifi represent Africa, samples from Amele represent Asia-Pacific, and samples from Porto Velho represent the Americas.
Proportion of genomes surveyed
To determine the proportion of P. falciparum genomes surveyed in each population, we used parameter estimates from the literature to calculate the number of circulating P. falciparum genomes in each age-specific population:
While population size and parasite prevalence varied greatly among populations, overall we sampled a minority of the circulating P. falciparum genomes in each population (Table 1). We estimated that in Bakoumba we sampled 29 of 877 circulating genomes (3.3%); in Pikine, 29 of 32487 genomes (0.09%); and in Kilifi, 30 of 47070 genomes (0.06%). Correspondingly, in Amele , 30 of an estimated 1446 parasite genomes (2.0%) were sampled, and in Porto Velho , 42 of an estimated 3136 genomes (1.3%) were sampled (Table 1).
Estimates of var richness
Due to this shallow sampling, we applied statistical estimators designed for incompletely sampled data sets to estimate var richness (the total number of var types in the population). Var gene diversity was found to be high in all three African parasite populations using the Chao1 non-parametric statistical estimator, suitable for data sets with numerous rare types . The richness estimates considered the age and clinical symptoms of each human population. We estimated 4540 var types in the Bakoumba population, 5116 types in Pikine, and 7565 types in Kilifi. In contrast, a log-order fewer var types were projected in the non-African populations: 369 in Amele (Papua New Guinea) and 232 in Porto Velho (Brazil). Projections using another non-parametric estimator (Abundance Coverage Estimator) yielded similar results (Table 2).
Both Bakoumba and Amele represent surveys of asymptomatic children in small (<10 km diameter), high transmission villages. A comparison of data from these two sites revealed considerably greater local var diversity specific to Africa. In Amele nearly half of the ∼400 projected var types were identified after sampling 30 genomes, whereas a similar sampling effort in Bakoumba yielded only 15% of the ∼4500 projected var types (Table 2). It follows that nearly a log-order more genomes must be sampled in Bakoumba compared to Amele to fully catalogue local var diversity (Figure S1). The references cited in Figure S1 and in other Supporting Information figures and tables are listed in Text S1.
Geographic differentiation of var types and microsatellite markers
Minimal sharing of var types was observed among the three African populations (Figure 2B) suggesting spatial diversity. Of the 1839 African var types sampled, 66 (3.6%) were shared between any two sites and only 10 (0.5%) were found in all three sites. Genotyping of microsatellite repeat markers  in these samples demonstrated evidence of geographic differentiation at the genome-wide level between West Africa and East/Central Africa. Each parasite isolate had a unique microsatellite haplotype, but there was overlap among haplotypes (ie: matching alleles at one or more loci). The microsatellite diversity statistics are shown in Table S3. Analysis of the 3 African populations revealed no significant differentiation among the Kilifi and Bakoumba populations as measured by Wrights FST (FST = 0.019) (Table S4). However, Pikine and Kilifi populations were significantly differentiated (FST = 0.039; P<0.01) and Pikine and Bakoumba showed low levels of differentiation (FST = 0.019) which did not reach statistical significance. When grouped together, Kilifi and Bakoumba populations remained significantly differentiated from Pikine (FST = 0.033; P<0.01). While these data together suggest geographic differentiation of var genes within Africa, we have interpreted this observation with care as the paucity of shared var types may simply reflect shallow sampling in any site (Figure S2). Comparisons across continents revealed few shared var types between African sites and the well-sampled Amele or Porto Velho sites (Figure 2C) demonstrating larger scale geographic variation in var genes. Two of the shared var types have high sequence homology with the previously described unusual semi-conserved var gene var1CSA found in most parasite isolates  (Table S5).
Var gene population structure
The population structure of var genes in Africa, as defined by organization of var gene repertoires within and among isolates, differed greatly from that observed in PNG or Brazil (Figures 3 and 4). Within a local African population, the majority of var types were rare, i.e. identified in only one of the 29–30 isolates sampled. In contrast, there were fewer rare types and considerable overlap in var repertoires in local Amele and Porto Velho parasite populations. It is noteworthy that the three African populations showed similar patterns of abundance (predominantly rare types) (Figure S3) despite marked differences in transmission and epidemiology.
In parenthesis next to each population is the number of isolates (n) sampled from the population. On the horizontal axis is the frequency class (in number of isolates) for each var type. The vertical axis depicts the number of var types found in each frequency class. For example, in the Kilifi dataset 622 var types were found in one isolate; 28 var types were each found in two isolates, etc. In the African populations, the overwhelming majority of var types were found in only one isolate. Differences in frequency distribution of var sequences were statistically significant by χ2 analysis (p<0.0001) (Table S6).
Within each labeled population, columns represent individual parasite isolates, and the black boxes represent var genes found in that isolate. Black boxes at the top of Figure 4 represent rare var types (found only in one isolate). Black boxes in the lower portion of Figure 4 depict var types that were found in more than one isolate within a population. For var types found in more than one isolate, each row represents a distinct type within the population. The key at the bottom-left of Figure 4 depicts frequency (in number of isolates) with which a particular var type was found in the population sample. Var types found more frequently were placed towards the bottom of the figure. White space represents an unknown number of var types that were not sequenced. Note that the amount of whitespace is not associated with numbers of genes missing, but was necessary to demonstrate sharing among repertoires. There was greater sharing of var types among isolates in the non-African populations compared to the African populations. Among the African populations, there appears to be more var type sharing in the Bakoumba population.
Comparison of var diversity with malariometric indices
Var diversity data was compared with standard malariometric indices such as prevalence of a positive blood smear and entomological inoculation rate (EIR), i.e. the number of bites by infectious mosquitoes received per person per year (Figure 5). We observed high var type richness throughout a range of parasite prevalence and transmission intensities in the African populations. Across continents, despite similar transmission levels and parasite prevalence in Amele and Bakoumba, large differences in var type richness were seen. Conversely, despite large differences in transmission and parasite prevalence, Amele and Porto Velho displayed similar var richness that was an order of magnitude lower than in the African populations. Strikingly, no correlation existed between var richness and prevalence, or var richness and EIR among the sites. The finding of high var type richness in low prevalence/low transmission settings such as Pikine demonstrates that var genetics can quantitatively discern complexity in the parasite reservoir.
For each population sampled, var richness estimated using the Chao1 equation is plotted against A) parasite prevalence and B) transmission as represented by entomological inoculation rate (EIR). We have used published parasite prevalence and EIR figures which are listed in Table 1. Where EIR has been reported as a range, we have plotted the midpoint of the range. The bars above and below each point represent the 95% confidence interval of the Chao1 estimate. Despite differences in transmission intensity and parasite prevalence, the local African populations all exhibited high estimates of var richness, roughly a log-order greater than the non-African populations.
The goal of any malaria elimination campaign is to reduce the reservoir of malaria infection in humans to zero locally. Success towards elimination would be monitored by measuring reductions in malaria transmission intensity using both EIR  and antigen-specific serological methods . Reductions in the size of the reservoir of infection in humans would be monitored by blood smear positivity . This measure is well known to be insensitive to changes in transmission intensity when parasite prevalence or rates reach 1 to 5% . It also does not identify the extent of diversity of the reservoir nor the existence of multiple genomes per host. Similarly, more sensitive species-specific diagnostic methods currently in development will not capture this information. Current efforts in malaria genomics have neglected to analyze the reservoir of infection.
Here we present the first study to explore the reservoir of infection in local parasite populations, through the lens of the diversity of the major surface antigen of the blood stage of P. falciparum. We have focused on the var genes as a measure of parasite fitness due to their role in both establishing chronic P. falciparum infections and promoting transmission to the mosquito. Our choice of this antigen fits the well-established microbiological paradigm that global surveillance of any antigenically diverse pathogen (e.g Influenza A, HIV-1) focuses on documenting diversity of the major surface antigen involved in immune evasion.
A detailed analysis of var genetics within local African sites compared to previously published data from local sites in PNG and Brazil was completed. Analysis of 2158 unique var sequences from 88 African isolates did not reveal the limit to var richness. We estimate var richness in each local African population to be five to seven thousand types. These are likely to be underestimates for several reasons including limited sampling (Figure S4); Chao1 estimates representing a lower bound of richness ; greater diversity when entire var genes are analyzed; and observed fine spatial heterogeneity such as that seen in plant communities  resulting in non-representative sampling. The richness may be even higher within the general population of each site if the var types seen within an age-restricted subset differ from those in the population at large. Data presented in Figure 5 show that the depth of diversity affords greater sensitivity to define the reservoir of infection and measure incremental reductions in the size of the reservoir with control, particularly in the African sites, compared to parasite prevalence. Clearly, this extensive variation presents an opportunity for parasite surveillance, although a major challenge to vaccine design.
The genomic var diversity data presented complement those previously reported as our study has focused on local parasite populations. We showed in the past that high worldwide var diversity existed by comparing Amele (a local PNG population) with a global dataset . Herein we describe high var diversity in local African populations, greatly exceeding that in local PNG or South American populations by at least an order of magnitude. Supporting our observations, others have described extensive local African var diversity in studies of severe malaria, although these studies mostly analyzed transcribed var genes rather than genomic var genes that constitute the reservoir (e.g. , ). The extent of var diversity in local African parasite populations compared to PNG and South America is consistent to that seen for genome-wide SNPs, microsatellites and mitochondrial genome diversity data, albeit with greater diversity in the var system , , . This is best explained by the long population history and documented high recombination rates of P. falciparum in Africa .
The observed spatial diversity in var types warrants follow up as such geographic variation could be extremely important in the maintenance of diversity, as is the case for plant species . How var gene variation exists in space and time needs to be elucidated by appropriate sampling and next-generation sequencing. In addition, the observed small scale spatial diversity of var genes suggests that var genetics could greatly inform current malaria mapping approaches  as well as predict complex malaria population dynamics due to the import of var types to areas where there is no widespread pre-existing immunity in the population.
Presently, limited var gene data are available from the P. falciparum genome sequencing projects, and planned 3X coverage for the majority of genomes is insufficient for assembly of the recombinogenic telomeric regions where most var genes reside . Despite next-generation sequencing methods, var genes have largely been ignored due to these assembly issues. This seems extraordinary given their importance for parasite survival. Our African data suggest that we would need to sequence the var genes of hundreds of genomes to observe the full extent of var DBLa diversity in one local population (Figure S1). These data also establish the scale of the next generation sequencing effort and the requirement for high density spatial sampling in a few endemic areas.
Figure 3 and 4 present the first comparative analysis of the population structure of the var multigene family in multiple local populations. Despite different epidemiologies in the African populations, var type frequency distributions (Figure 3) and levels of repertoire overlap (Figure 4) appeared essentially the same but significantly different from local PNG and South American populations. Indeed, negligible var repertoire overlap was seen within any of three local African settings indicating that reinfected individuals would be more likely to encounter parasites with repertoires of var types to which they lack immunity. We propose that this lack of var repertoire overlap in parasite genomes within each African site favors superinfection and persistence in a semi-immune host. This observation can explain why semi-immune African children harbor multiple parasite genomes (peak mean MOI of 5 in children 3–7 years old) and the number of genomes reduces as children become more immune .
Immense var gene diversity and the observed lack of repertoire overlap in Africa pose a great challenge to malaria control and elimination in that immunity to infection only develops after many exposures which may be due to the numerous (antigenically distinct) var types in each population. Furthermore, the extent of overlap among the var repertoires will determine cross-immunity in the host population to distinct parasite genomes. Consequently, levels of cross-immunity would be higher in PNG and South American sites compared to African sites. Finally, the low repertoire overlap in Africa demonstrates the enormous potential for outcrossing in these parasite populations ,  to quickly generate further genomic diversity of var repertoires unless elimination strategies are implemented and sustained.
We observed high var diversity in both low and high prevalence sites in Africa. This striking lack of correlation between prevalence and var diversity shows that var typing provides a complementary measure of the reservoir of infection to parasite prevalence. We propose that this molecular epidemiological surveillance system will prove particularly useful in monitoring incremental changes in the population structure and size of the reservoir of infection at low parasite prevalence in the new era of malaria eradication.
Thus, we conclude that var population genomics provides new insights into the epidemiology of P. falciparum in Africa where malaria has never been conquered. In particular, we have been able to describe the nature and extensive size of the reservoir of infection in local African sites as well as discover a unique var population structure that can facilitate superinfection with minimal overlap among var repertoires.
The yield of further var sampling efforts. The predicted proportion of total types observed was plotted as a function of sample size, using a non-parametric estimator  modeled on data from Bakoumba, Gabon and Amele, Papua New Guinea. Applying our present framework for sampling, we would need to sample 320 isolates to achieve 80% coverage of var types in the surveyed age group in Bakoumba, Gabon. In contrast, in Amele, PNG a sample of 40 isolates would suffice to achieve similar coverage. It is clear that to achieve similar coverage of var types in Africa and PNG requires vastly greater sampling effort in African populations. These projections were based on sampling of all parasitemic individuals, and not limited only to those with single clone infections. Application of next generation sequencing technologies may result in a higher yield of var sequences per isolate, and decrease the total number of isolates needed; however sampling in Africa would still require much greater number of isolates as compared to PNG.
Potential effect of sampling depth on appearance of differentiation. Hypothetical sampling of three identical populations, each containing the same highly diverse set of 1000 distinct var types. A) A random sample of 10% (100 types) is taken from each of the three populations. B) Assuming equal probability of sampling each var type, the resulting 10% samples from each population would demonstrate few var types shared between populations, even though the composition of each population is identical. C) A random sample of 80% (800 types) is taken from each of the three populations. D) If 80% of each population is sampled, the samples would display greater sharing of var types and would more closely resemble the true relationship among the populations (D). This exercise depicts one scenario where shallow sampling of populations may give the appearance of differentiation. Outcomes will vary with differences in total diversity, distribution of diversity, and sharing in the populations compared.
Var rank abundance of African and non-African populations. For each population, the relative abundance of each var type was plotted against the abundance rank of that var type. Relative abundance refers to the proportion of the total var sequences in the population sample. The slope of the curve reflects the evenness of the relative abundances of the var types sampled. A steep curve, as seen in Amele, denotes a more heterogeneous distribution of types.
Stability of richness estimators with sample size. In this figure, averaged Chao1 and ACE richness estimates were plotted against the number of var genes sampled for each population. Estimates of richness for the African populations (Kilifi, Pikine, Bakoumba) were a log-order of magnitude greater than those of the S. American (Porto Velho) or PNG (Amele) populations. Richness estimates demonstrated stable increases (slope of curve) with greater sampling .
List of accession numbers for Amele sequences used in the analysis. GenBank Accession numbers for the 460 Amele DBLα sequences used in this analysis .
List of Porto Velho isolates used in the analysis. DBLα sequences from 42 Porto Velho isolates ,  used in this analysis. Sequence data was downloaded from the supplementary online file Supplemental Table 4 (applic6.txt) in reference .
Allelic diversity of African local populations based on analysis of 12 microsatellite loci. Kilifi was the most diverse parasite population, followed by Bakoumba and Pikine. n = number of samples, SE = standard error.
Geographic differentiation of African local populations based on analysis of 12 microsatellite loci. Wrights FST values for population comparisons by microsatellite alleles calculated using FSTAT .
African var types shared across populations. From the datasets analyzed, this table lists the African var types also found in Amele and/or Porto Velho, as well as var types found in all three African populations. Under ‘Accession Number,’ a representative African sequence is listed. The table indicates the populations in which each type was found. Two shared var types, DQ135232 and DQ135587, demonstrated high homology to var1CSA, an unusual semi-conserved var gene . No common sequences matched the conserved var gene var2csa, however the primers were not designed to amplify this unique gene.
Chi-square test on the distribution var sequence frequencies among the five populations. Differences in frequency distribution among the five populations are statistically significant by χ2 analysis (χ2 = 1152; p<0.0001).
The authors wish to thank Dyann Wirth for providing malaria samples. We would like to thank the volunteers and communities involved in the study; Brett Lowe for logistical support; Mercedes Pascual for advice on the rank abundance curve; and Martine Zilversmit for critical evaluation of the manuscript.
Conceived and designed the experiments: DSC AEB KPD. Performed the experiments: DSC AEB AL-S T-AS. Analyzed the data: DSC AEB T-AS IP SMB KPD. Contributed reagents/materials/analysis tools: FM-N PD MMK KM JPD DN OS SM KPD. Wrote the paper: DSC AEB KPD.
- 1. Macdonald G (1956) Epidemiological basis of malaria control. Bulletin of the World Health Organization 15: 613–626.
- 2. Macdonald G, Goeckel GW (1964) The Malaria Parasite Rate and Interruption of Transmission. Bull World Health Organ 31: 365–377.
- 3. Bruce MC, Donnelly CA, Alpers MP, Galinski MR, Barnwell JW, et al. (2000) Cross-Species Interactions Between Malaria Parasites in Humans. Science 287: 845–848.
- 4. Nassir E, Abdel-Muhsin A-MA, Suliaman S, Kenyon F, Kheir A, et al. (2005) Impact of genetic complexity on longevity and gametocytogenesis of Plasmodium falciparum during the dry and transmission-free season of eastern Sudan. International Journal for Parasitology 35: 49–55.
- 5. Smith T, Felger I, Fraser-Hurt N, Beck HP (1999) 10. Effect of insecticide-treated bed nets on the dynamics of multiple Plasmodium falciparum infections. Transactions of the Royal Society of Tropical Medicine and Hygiene 93: 53–57.
- 6. Bruce MC, Donnelly CA, Packer M, Lagog M, Gibson N, et al. (2000) Age- and species-specific duration of infection in asymptomatic malaria infections in Papua New Guinea. Parasitology 121: 247–256.
- 7. Hay SI, Smith DL, Snow RW (2008) Measuring malaria endemicity from intense to interrupted transmission. The Lancet Infectious Diseases 8: 369–378.
- 8. Anderson TJC, Haubold B, Williams JT, Estrada-Franco JG, Richardson L, et al. (2000) Microsatellite Markers Reveal a Spectrum of Population Structures in the Malaria Parasite Plasmodium falciparum. Mol Biol Evol 17: 1467–1482.
- 9. Daniels R, Volkman SK, Milner DA, Mahesh N, Neafsey DE, et al. (2008) A general SNP-based molecular barcode for Plasmodium falciparum identification and tracking. Malar J 7: 223.
- 10. Guerin PJ, Bates SJ, Sibley CH (2009) Global resistance surveillance: ensuring antimalarial efficacy in the future. Current Opinion in Infectious Diseases 22: 593–600.
- 11. Barry AE, Schultz L, Buckee CO, Reeder JC (2009) Contrasting Population Structures of the Genes Encoding Ten Leading Vaccine-Candidate Antigens of the Human Malaria Parasite, Plasmodium falciparum. PLoS ONE 4: e8497.
- 12. Takala SL, Coulibaly D, Thera MA, Batchelor AH, Cummings MP, et al. (2009) Extreme Polymorphism in a Vaccine Antigen and Risk of Clinical Malaria: Implications for Vaccine Development. Science Translational Medicine 1: 2ra5.
- 13. Kyes SA, Kraemer SM, Smith JD (2007) Antigenic Variation in Plasmodium falciparum: Gene Organization and Regulation of the var Multigene Family. Eukaryotic Cell 6: 1511–1520.
- 14. Miller LH, Baruch DI, Marsh K, Doumbo OK (2002) The pathogenic basis of malaria. Nature 415: 673.
- 15. Hayward RE, Tiwari B, Piper KP, Baruch DI, Day KP (1999) Virulence and transmission success of the malarial parasite Plasmodium falciparum. Proc Natl Acad Sci U S A 96: 4563–4568.
- 16. Staalsoe T, Hamad AA, Hviid L, Elhassan IM, Arnot DE, et al. (2002) In vivo switching between variant surface antigens in human Plasmodium falciparum infection. J Infect Dis 186: 719–722.
- 17. Barry AE, Leliwa-Sytek A, Tavul L, Imrie H, Migot-Nabias F, et al. (2007) Population Genomics of the Immune Evasion (var) Genes of Plasmodium falciparum. PLoS Pathog 3: e34.
- 18. Fowler EV, Peters JM, Gatton ML, Chen N, Cheng Q (2002) Genetic diversity of the DBLalpha region in Plasmodium falciparum var genes among Asia-Pacific isolates. Mol Biochem Parasitol 120: 117–126.
- 19. Albrecht L, Castiñeiras C, Carvalho BO, Ladeia-Andrade S, Santos da Silva N, et al. (2010) The South American Plasmodium falciparum var gene repertoire is limited, highly shared and possibly lacks several antigenic types. Gene 453: 37–44.
- 20. Ntoumi F, Ekala M-T, Makuwa M, Lekoulou F, Mercereau-Puijalon O, et al. (2002) Sickle cell trait carriage: imbalanced distribution of IgG subclass antibodies reactive to Plasmodium falciparum family-specific MSP2 peptides in serum samples from Gabonese children. Immunology Letters 84: 9–16.
- 21. Aubouy A, Migot-Nabias F, Deloron P (2003) Polymorphism in two merozoite surface proteins of Plasmodium falciparum isolates from Gabon. Malaria Journal 2: 12.
- 22. Elissa N, Migot-Nabias F, Luty A, Renaut A, Toure F, et al. (2003) Relationship between entomological inoculation rate, Plasmodium falciparum prevalence rate, and incidence of malaria attack in rural Gabon. Acta Tropica 85: 355–361.
- 23. Cabrera G, Cot M, Migot-Nabias F, Kremsner PG, Deloron P, et al. (2005) The sickle cell trait is associated with enhanced immunoglobulin G antibody responses to Plasmodium falciparum variant surface antigens. J Infect Dis 191: 1631–1638.
- 24. Trape JF, Lefebvre-Zante E, Legros F, Ndiaye G, Bouganali H, et al. (1992) Vector density gradients and the epidemiology of urban malaria in Dakar, Senegal. Am J Trop Med Hyg 47: 181–189.
- 25. Ndiaye D, Daily JP, Sarr O, Ndir O, Gaye O, et al. (2005) Mutations in Plasmodium falciparum dihydrofolate reductase and dihydropteroate synthase genes in Senegal. Tropical Medicine & International Health 10: 1176–1179.
- 26. Mwangi TW, Ross A, Snow RW, Marsh K (2005) Case definitions of clinical malaria under different transmission conditions in Kilifi District, Kenya. J Infect Dis 191: 1932–1939.
- 27. Mbogo CM, Mwangangi JM, Nzovu J, Gu W, Yan G, et al. (2003) Spatial and temporal heterogeneity of Anopheles mosquitoes and Plasmodium falciparum transmission along the Kenyan coast. Am J Trop Med Hyg 68: 734–742.
- 28. Felger I, Tavul L, Beck HP (1993) Plasmodium falciparum: A Rapid Technique For Genotyping The Merozoite Surface Protein-2. Experimental Parasitology 77: 372–375.
- 29. Anderson TJC, Su XZ, Bockarie M, Lagog M, Day KP (1999) Twelve microsatellite markers for characterization of Plasmodium falciparum from finger-prick blood samples. Parasitology 119: 113–125.
- 30. Smith JD, Subramanian G, Gamain B, Baruch DI, Miller LH (2000) Classification of adhesive domains in the Plasmodium falciparum Erythrocyte Membrane Protein 1 family. Molecular and Biochemical Parasitology 110: 293.
- 31. Albrecht L, Merino EF, Hoffmann EHE, Ferreira MU, de Mattos Ferreira RG, et al. (2006) Extense variant gene family repertoire overlap in Western Amazon Plasmodium falciparum isolates. Molecular and Biochemical Parasitology 150: 157–165.
- 32. Goudet J (1995) FSTAT (Version 1.2): A Computer Program to Calculate F-Statistics. J Hered 86: 485–486.
- 33. Colwell RK (2006) EstimateS: Statistical estimation of species richness and shared species from samples. Version 8.0. Available: http://purl.oclc.org/estimates.
- 34. Chao A (1984) Non-parametric estimation of the number of classes in a population. Scandinavian Journal of Statistics 11: 265–270.
- 35. Chao A, Lee S-M (1992) Estimating the Number of Classes via Sample Coverage. Journal of the American Statistical Association 87: 210–217.
- 36. Chao A, Shen T-J (2009) Program SPADE (Species Prediction And Diversity Estimation). Version 2009/Feb. Available: http://chao.stat.nthu.edu.tw.
- 37. Whittaker RH (1965) Dominance and Diversity in Land Plant Communities: Numerical relations of species express the importance of competition in community function and evolution. Science 147: 250–260.
- 38. Shen T-J, Chao A, Lin C-F (2003) Predicting the number of new species in further taxonomic sampling. Ecology 84: 798–804.
- 39. Drakeley CJ, Corran PH, Coleman PG, Tongren JE, McDonald SLR, et al. (2005) Estimating medium- and long-term trends in malaria transmission by using serological markers of malaria exposure. Proceedings of the National Academy of Sciences of the United States of America 102: 5108–5113.
- 40. Burdon JJ, Thrall PH (1999) Spatial and Temporal Patterns in Coevolving Plant and Pathogen Associations. The American Naturalist 153: S15–S33.
- 41. Kyriacou HM, Stone GN, Challis RJ, Raza A, Lyke KE, et al. (2006) Differential var gene transcription in Plasmodium falciparum isolates from patients with cerebral malaria compared to hyperparasitaemia. Molecular and Biochemical Parasitology 150: 211–218.
- 42. Warimwe GM, Keane TM, Fegan G, Musyoki JN, Newton CRJC, et al. (2009) Plasmodium falciparum var gene expression is modified by host immunity. Proceedings of the National Academy of Sciences 106: 21801–21806.
- 43. Joy DA, Feng X, Mu J, Furuya T, Chotivanich K, et al. (2003) Early Origin and Recent Expansion of Plasmodium falciparum. Science 300: 318–321.
- 44. Mu J, Awadalla P, Duan J, McGee KM, Joy DA, et al. (2005) Recombination Hotspots and Population Structure in Plasmodium falciparum. PLoS Biology 3:
- 45. Hay SI, Guerra CA, Gething PW, Patil AP, Tatem AJ, et al. (2009) A world malaria map: Plasmodium falciparum endemicity in 2007. PLoS Med 6: e1000048.
- 46. Plasmodium Writing Group (2008) Comprehensive Sequencing Proposal for Plasmodium. 29: Available: www.genome.gov/Pages/Research/DER/PathogensandVectors/PlasmodiumWhitePaperV8.pdf. Accessed 2008 Sep.
- 47. Babiker HA, Ranford-Cartwright LC, Currie D, Charlwood JD, Billingsley P, et al. (1994) Random mating in a natural population of the malaria parasite Plasmodium falciparum. Parasitology 109(Pt 4): 413–421.
- 48. Paul RE, Packer MJ, Walmsley M, Lagog M, Ranford-Cartwright LC, et al. (1995) Mating patterns in malaria parasite populations of Papua New Guinea. Science 269: 1709–1711.
- 49. Eyles D, Young M (1951) The duration of untreated or inadequately treated Plasmodium falciparum infections in the human host. Journal National Malaria Society 10: 327.
- 50. Molineaux L, Diebner HH, Eichner M, Collins WE, Jeffery GM, et al. (2001) Plasmodium falciparum parasitaemia described by a new mathematical model. Parasitology 122: 379–391.
- 51. Earle WC (1962) The course of naturally acquired malaria. Geneva: World Health Organization. Document WHO/Mal/333. 14: Available: http://whqlibdoc.who.int/malaria/WHO_Mal_333.pdf. Accessed 2009 Apr.
- 52. Burkot TR, Graves PM, Paru R, Wirtz RA, Heywood PF (1988) Human malaria transmission studies in the Anopheles punctulatus complex in Papua New Guinea: sporozoite rates, inoculation rates, and sporozoite densities. Am J Trop Med Hyg 39: 135–144.
- 53. Gil LH, Alves FP, Zieler H, Salcedo JM, Durlacher RR, et al. (2003) Seasonal malaria transmission and variation of anopheline density in two distinct endemic areas in Brazilian Amazonia. J Med Entomol 40: 636–641.
- 54. Aubouy A, Bakary M, Keundjian A, Mbomat B, Makita JR, et al. (2003) Combination of Drug Level Measurement and Parasite Genotyping Data for Improved Assessment of Amodiaquine and Sulfadoxine-Pyrimethamine Efficacies in Treating Plasmodium falciparum Malaria in Gabonese Children. Antimicrob Agents Chemother 47: 231–237.
- 55. Direction de la Prévision et de la Statistique (2004) Projections de population du Sénégal issues du recensement de 2002. 31. Available: http://www.ansd.org/Rslt_prov_RGPHIII.pdf. Accessed 2008 Aug.
- 56. O'Meara WP, Bejon P, Mwangi TW, Okiro EA, Peshu N, et al. (2008) Effect of a fall in malaria transmission on morbidity and mortality in Kilifi, Kenya. The Lancet 372: 1555–1562.
- 57. Lewis MP (2009) Amele: A language of Papua New Guinea. In: Ethnologue: Languages of the World. Dallas: SIL International. 26: Available: http://www.ethnologue.com/show_language.asp?code=aey. Accessed 2009 May.
- 58. Instituto Brasileiro de Geografia e Estatística (2004) Census of the Munincipality of Porto Velho 2004. 26: Available: ftp://ftp.ibge.gov.br/Estimativas_Projecoes_Populacao/Estimativas_2004/UF_Municipio.zip. Accessed 2009 May.
- 59. Mapping Malaria Risk in Africa: MARA (2004) Populations at Risk of Malaria. 26: Available: http://www.mara.org.za/popatrisk.htm. Accessed 2009 May.
- 60. Cattani JA, Tulloch JL, Vrbova H, Jolley D, Gibson FD, et al. (1986) The Epidemiology of Malaria in a Population Surrounding Madang, Papua New Guinea. Am J Trop Med Hyg 35: 3–15.
- 61. Fowkes FJI, Imrie H, Migot-Nabias F, Michon P, Justice A, et al. (2006) Association of haptoglobin levels with age, parasite density, and haptoglobin genotype in a malaria-endemic area of Gabon. Am J Trop Med Hyg 74: 26–30.
- 62. Imrie H, Fowkes FJ, Michon P, Tavul L, Hume JC, et al. (2006) Haptoglobin levels are associated with haptoglobin genotype and alpha+ -Thalassemia in a malaria-endemic area. Am J Trop Med Hyg 74: 965–971.
- 63. Camargo LM, dal Colletto GM, Ferreira MU, Gurgel Sde M, Escobar AL, et al. (1996) Hypoendemic malaria in Rondonia (Brazil, western Amazon region): seasonal variation and risk groups in an urban locality. Am J Trop Med Hyg 55: 32–38.
- 64. Ndiaye D, Daily JP, Sarr O, Ndir O, Gaye O, et al. (2006) Defining the origin of Plasmodium falciparum resistant dhfr isolates in Senegal. Acta Tropica 99: 106–111.
- 65. Färnert A, Williams Thomas N, Mwangi Tabitha W, Ehlin A, Fegan G, et al. (2009) Transmission-Dependent Tolerance to Multiclonal Plasmodium falciparum Infection. The Journal of Infectious Diseases 200: 1166–1175.
- 66. Sallenave-Sales S, Daubersies P, Mercereau-Puijalon O, Rahimalala L, Contamin H, et al. (2000) Plasmodium falciparum: a comparative analysis of the genetic diversity in malaria-mesoendemic areas of Brazil and Madagascar. Parasitology Research 86: 692–698.