Report

Genetic Loci Affecting Resistance to Human Malaria Parasites in a West African Mosquito Vector Population

See allHide authors and affiliations

Science  04 Oct 2002:
Vol. 298, Issue 5591, pp. 213-216
DOI: 10.1126/science.1073420

Abstract

Successful propagation of the malaria parasitePlasmodium falciparum within a susceptible mosquito vector is a prerequisite for the transmission of malaria. A field-based genetic analysis of the major human malaria vector, Anopheles gambiae, has revealed natural factors that reduce the transmission of P. falciparum. Differences in P. falciparumoocyst numbers between mosquito isofemale families fed on the same infected blood indicated a large genetic component affecting resistance to the parasite, and genome-wide scanning in pedigrees of wild mosquitoes detected segregating resistance alleles. The apparently high natural frequency of resistance alleles suggests that malaria parasites (or a similar pathogen) exert a significant selective pressure on vector populations.

Mosquitoes infected with malaria incur quantifiable costs in reproductive fitness for measures such as longevity, fecundity, and flight distance (1–3). In addition, mosquito immune genes respond transcriptionally to malaria parasite infection (4–6), indicating that malaria parasites are detected by mosquito immune surveillance. Inbred lines selected from mosquito colonies can inhibit parasite development through at least two distinct mechanisms: melanotic encapsulation and intracellular lysis (7, 8). Therefore, selective pressure by parasite infection could drive the natural selection of resistance mechanisms in mosquitoes. However, previous studies of resistance used laboratory strains of mosquitoes and malaria parasites, which are known to display biological aberrations in comparison to natural populations (9–11), and thus the relevance of these studies to natural malaria transmission may be questioned. Malaria resistance mechanisms and their significance for the biology of malaria transmission have not been examined previously in natural field populations of vectors.

For study of the vector-parasite interaction using field populations, standard genetic techniques that require the creation of inbred or isogenic lines were not applicable. However, the large number of progeny produced by single wild females permitted a novel study design based on isofemale pedigrees, which is quite distinct from designs previously used in human and animal genetics (fig. S1). In nature,A. gambiae females mate only once (12), and the stored sperm is used to fertilize successive egg batches that mature when the female feeds on blood. We captured pre-mated blood-fed females resting in village dwellings in Bancoumana, Mali (West Africa); allowed them to oviposit in an environmentally controlled chamber; and then raised their respective progeny in the environmental chamber (13). We used the isofemale families at either the F1 or the F2 stage (the latter produced by mass mating of F1 mosquitoes). All females of a family were challenged with malaria parasites by being fed on blood from the sameP. falciparum gametocyte carrier (14). Unfed mosquitoes were removed (15). At 8 days after infection, the progeny were dissected, and the surviving oocyst-stage parasites were counted in each midgut. The number of oocysts per mosquito constituted the quantitative infection phenotype.

We investigated the genetic dependence of mosquito resistance to malaria infection in two stages. First, in a familiality study, we compared parasite infection levels between mosquito families that were each generated by a different founding pair of wild mosquitoes. The goal was to measure interfamily differences in infection distributions, and therefore only phenotype measurements were used. Groups of three F1 families were reared together in an environmental chamber and were fed simultaneously on blood from a single P. falciparum gametocyte carrier. The three families constituted a comparison group; within such a group, environmental differences could be discounted. We collected six such three-family comparison groups. If there were no segregating mosquito genes that influence the infection phenotype, then parasite numbers in the different families of the same group should appear to be drawn from the same distribution. Conversely, the action of genes that affect resistance should cause family-specific changes in the distribution of parasite numbers. We performed pairwise comparisons (a:b, a:c, and b:c) of infection phenotype for the three families of each comparison group. Because the phenotypic distribution of oocyst infection in mosquitoes tends to be non-normal (16–18), we used the nonparametric Kolmogorov-Smirnov (KS) test to evaluate significance [see methods (13)]. Of 18 pairwise comparisons, 9 were significantly different at P < 0.05, including 5 that were significant atP < 0.01 and 3 that were significant at P < 0.001 (table S1). There was at least one significant pairwise difference in all but one of the comparison groups.

Thus, the familiality study provided strong evidence that alleles segregating in a natural population of A. gambiae have a significant effect on susceptibility to P. falciparum, and that such alleles are collectively frequent. This information has not been available for natural populations of malaria vectors and was further validated in the second stage of this study. In the second stage, we performed a genome-wide scan for loci that influence resistance or susceptibility. Two independent wild isofemale F2 families were infected on the blood of different gametocyte carriers, and then we determined phenotypes and genotypes of each mosquito at a broadly distributed set of 24 microsatellite loci (fig. S2).

Because the genetic size of the A. gambiae genome is 215 centimorgans (cM) (19), the resolution of the scan was approximately 9 cM. At this marker density, an unknown locus of interest is, on average, 4.5% recombination distance (∼6 Mb) away from a marker locus. For the genome scan, we compared parasite count distributions for mosquitoes that shared a single allele or allelic combination (genotype) versus all other mosquitoes in the same family. We used F2 families because they offered a substantial increase in the number of mosquitoes per family, as compared to F1s. It should be noted that although either generation (F1 or F2) has a maximum of four alleles sampled from the natural population, the maximum number of genotypes per locus is four for the F1 and 10 for the F2 generation.

Table 1 presents the results from the two F2 families. At every marker, we split the members of the F2 generation into two samples, using a genotype or allele observed at that marker as the selection criterion. For example, if six genotypes were observed in the F2 sample, we performed six genotype tests for the marker: We compared the phenotype distribution for each genotype against the pooled phenotype distribution of all other genotypes. For every test, we computed the probability of encountering the observed difference under the null hypothesis of no linkage between this marker and malaria resistance (20). As was the case in the familiality study, we used a nonparametric test to compare phenotype distributions. For each family, we performed approximately 60 allele tests and 100 genotype tests. To adjust for multiple testing, we used an empirical permutation test (21,22). Statistical test results were interpreted according to accepted criteria (23).

Table 1

Statistical tests for genetic loci controlling midgut oocyst number. F2 families of A. gambiae from wild single-pair crosses were fed on natural P. falciparum–infected blood, and individual microsatellite genotypes were determined at a resolution of ∼9 cM. Nominal (P n) and genome-wide (corrected for multiple testing) P values are shown for the KS test. Genome-wideP values were generated by randomizing the correlation between marker genotype and phenotype distributions. We repeated the genome scan in 105 (or 107 for 98BF214 allele 1) randomized data sets.

View this table:

Both of the F2 families displayed significant linkage between parasite count and DNA markers (Table 1). In family 98BF214 (n = 83 mosquitoes), one of the two observed marker alleles at the chromosome 2L marker H603 had a strong correlation with infection phenotype (genome-wide Pvalue = 4 × 10−7; phenotype distributions are shown in Fig. 1A). The effect was seen in both genotypes that included allele 1, which suggests that this marker is located close to a locus where the resistance allele is semidominant. The putative resistance allele linked to marker allele 1 can explain almost all of the parasite-free mosquitoes (24 out of 27) in the family. The linked resistance locus is provisionally namedPfin1, for P. falciparum infection intensity 1. In family 97F2B4A5 (n = 82), a different marker (H290 on chromosome 2R) showed significant linkage between genotype and parasite count (Fig. 1B). Here, one of the eight genotypes at H290 (genotype 2–4) showed significant resistance (genome-wide P value = 0.025), suggesting recessive inheritance of resistance (24). This locus is provisionally named Pfin2.

Figure 1

Infection phenotype of wild A. gambiae pedigrees with natural P. falciparum. Histograms show the infection phenotype of F2 isofemale families by microsatellite marker genotype. Green bars show phenotypic distributions for the genotype indicated in the panel. Open bars show the overall phenotypic distribution for all phenotypes of the family combined. (A) Family 98BF214. Green bars show the phenotype for the indicated genotype at marker H603. Mean oocyst number per mosquito for genotype 1-1 (n = 6) is 0.17 oocyst; for genotype 1-2 (n = 44), 9.9 oocysts; for genotype 2-2 (n = 33), 50.6 oocysts. (B) Family 97F2B4A5. Green bars show the phenotype for genotype 2-4 (n = 8) at marker H290. This family has seven other genotypes (Table 1). Mean oocyst number per mosquito for genotype 2-4 (n = 8) is 8.0 oocysts; for all other genotypes (n = 74), it is 28.6 oocysts. In all panels, the width of the bars represents a range of eight counts per bin. The first bar is centered on 0 and has an effective width of 0 to 3 oocysts. The second bar includes the bin from 4 to 11 oocysts, and subsequent bins follow the same pattern.

Genetic control of the previously described encapsulation-mediated resistance mechanism in A. gambiae (7) was composed of three loci, where the major locus, Pen1, was closely linked to marker H175 and near markerH290 (19, 25). Despite this genetic proximity, it is unlikely that Pfin2 is the same asPen1 for two reasons. First, the phenotypes are distinct. The influence of Pen1 was expressed as melanotic encapsulation of malaria ookinetes, with no effect on infection intensity (25), whereas the influence of Pfin2was exclusively expressed as variation in infection intensity, without melanotic encapsulation. Second, the encapsulation trait largely controlled by Pen1 was strongest against allopatric (non-African) strains of P. falciparum but had a much reduced effect on African parasite strains (7), whereas Pfin2 was identified specifically because of its action against sympatric parasites. It is likely that Pfin2and Pen1 are distinct loci, perhaps in a resistance gene cluster, although the possibility cannot currently be excluded that they might represent disparate alleles at the same locus.

The Pfin1 marker H603 lies within the large chromosome inversion 2La (19). Karyotype analysis of polytene ovarian nurse cell chromosomes showed that family 98BF214 was fixed for the homokaryotypic inverted 2La/a chromosome arrangement (26). Thus, the linkage between H603 and resistance was due to linkage between two freely segregating loci and not due to disequilibrium by suppression of recombination within a polymorphic inversion.

Outbred animal populations have been used previously for fine mapping of traits (27, 28), but rarely for initial gene discovery by genome scan in field populations (29). The novelty of the current study design lies in the mass mating step used to produce the F2 generation. We chose the F2study design because it offers a large sample size. Although mass mating among F1 mosquitoes leads to loss of some inheritance information (such as paternal relationships), phenotypic effect and marker informativeness were sufficiently large to help us identify genomic regions with a strong influence on vector-parasite interaction. We demonstrate that field populations of A. gambiae exhibit significant variation in permissiveness for parasite development, and we point to two genomic regions that affect this trait.

It is likely that there are other such resistance loci, and the strategy we present can be used to more extensively screen natural mosquito populations. By implementation of a modified protocol to maintain pedigrees beyond the initial genome scan of the F2generation, segregating resistance alleles could be mapped at high resolution and positional candidates could be identified. Extant pedigrees would also permit characterization of the mechanisms of resistance. In the current work, we detected resistance as a reduction in oocyst number 8 days after a blood meal. This is an aggregate phenotype that summarizes all preceding events in parasite development, and it is likely that different resistance alleles cause developmental blockades at different critical points. It is also possible that some resistance mechanisms may interact or synergize with human host factors such as transmission-blocking antibodies or cytokines in the infecting blood meal.

It would be of interest to identify the molecular nature of these resistance loci, which are probably rapidly evolving components of the genome that define the points of greatest adaptive friction between parasite virulence factors and the mosquito host. In plants, most genetically identified resistance genes are pattern-recognition receptors for pathogen virulence factors (30,31), but comparable information is lacking in mosquitoes. Continued parasite transmission by a vector with a high frequency of segregating resistance factors suggests that the parasite has made adaptive responses to mosquito resistance, perhaps by evolving multigenic and/or polymorphic virulence factors for the insect stages of the life cycle analogous to those found in asexual-stage parasites (32). Virulence in the mosquito host is under parasite genetic control (33), but specific virulence factors are not known.

Malaria-infected mosquitoes in nature typically carry fewer than 10 oocysts (18, 34). Far higher parasite intensities can be achieved in laboratory infections of mosquitoes, particularly when genetically selected susceptible lines are used (7, 8). Natural resistance alleles that limit parasite development in the vector, such as those described here, may prove to be an important factor underlying the small numbers of oocysts observed in wild infected mosquitoes. Genetic studies in the field as well as in the laboratory will be needed to elucidate the mechanisms that can limit the propagation of P. falciparum in A. gambiae and their respective importance for natural transmission of the disease.

Supporting Online Material

www.sciencemag.org/cgi/content/full/298/5591/213/DC1

Materials and Methods

Figs. S1 and S2

Table S1

Data Sets

  • * These authors contributed equally to this work.

  • To whom correspondence should be addressed. E-mail: kenneth.vernick{at}nyu.edu

REFERENCES AND NOTES

View Abstract

Navigate This Article