A major genetic locus controlling natural Plasmodium falciparum infection is shared by East and West African Anopheles gambiae

The mosquito Anopheles gambiae is the major African vector of human malaria caused by Plasmodium falciparum. More than 300 million cases of acute malaria occur globally each year resulting in over one million deaths with greater than 90% of these deaths occurring in children from Sub-Saharan Africa [1]. In malaria endemic countries where the poorest sections of society are most affected, inadequate public health initiatives and the selection of insecticide resistance hinder the efforts to control malaria transmission. Recent proposals to curb the malaria burden involve more technological approaches including the generation and introduction of transgenic mosquitoes [2]. However, all approaches have their own logistical, political and ethical limitations.

Here the genetics of natural vector resistance to P. falciparum is dissected, with the ultimate goal to utilize successful existing natural Plasmodium resistance mechanisms for control of human malaria transmission. Knowledge of the prevalence, strength and genomic location of loci that control mosquito resistance to malaria parasites in vector populations could aid future efforts to control parasite transmission. Genetically resistant and susceptible mosquitoes exist in nature, and phenotypic variation can segregate as a simple Mendelian trait [3]. A recent comprehensive study reported the prevalence, strength and genomic location of natural P. falciparum resistance loci in A. gambiae in Mali, West Africa [4]. The loci with the strongest phenotypic effect on parasite numbers clustered to form a Plasmodium-resistance island (PRI) on chromosome 2L. It would be valuable to know whether the genetic mechanism of the PRI identified in Mali is specific to that population, suggesting that there might be many geographically local A. gambiae resistance mechanisms across Africa, or conversely whether the mechanism identified by the PRI exists in A. gambiae throughout Africa. A major barrier to gene flow in A. gambiae occurs between West and East African populations [5], and thus comparing these populations should offer the most stringent test of the geographic generality of genetic loci controlling resistance mechanisms.

The source of the pedigrees was previously reported [6]. Briefly, isofemale pedigrees were initiated from a laboratory colony (established in Mbita, Kenya in 2001, approximately three years prior to use in this study) and raised to the F1 generation at which point they were challenged on gametocyte containing blood. The same pedigrees were previously used to investigate the role of genotype by genotype interactions in controlling the parasite load of vector mosquitoes [6]. Thus the parasite challenge was heterogeneous: each F1 pedigree was split into 3–5 groups that were each fed on independent natural gametocytemic blood samples. Unfed and partially fed females were excluded from further analysis. Seven to eight days after the infective blood meal, mosquito midguts were dissected and stained with 2% mercurochrome, and the numbers of oocysts were counted by light microscopy. Mosquito carcasses were saved for later DNA extraction.

DNA was isolated from mosquitoes using Qiagen DNeasy (Qiagen, CA, USA). Microsatellite genotyping was performed multiplexed as described previously [4]. Fragments were separated on an ABI3100 Genetic Analyzer and sized using GeneMapper Version 3.5 (both from Applied Biosystems, CA, USA). All automated genotype calls were visually verified. Mapping was carried out in those pedigrees deemed potentially informative using previously described guidelines; ≥20 females with ≥30% infection prevalence [4]. Six pedigrees, named 57, 92, 141, 142, 146 and 1102 met the typing criteria. Twenty-eight microsatellite loci were screened in these informative pedigrees and as many as 15 additional loci were typed in some, but not all pedigrees (Figure 1). Many of these additional markers were added to increase the signal resolution around marker H325. All primer sequences are available elsewhere [4, 7] except 2L.24D left 5' TGCCAATCAATCAGTGTGCT 3' and right 5' CACTAGCAACGGCACTACCC 3' and 2L.26D left 5' GAAGTGGAAGAACACGCTCA and right 5' CACACGATTGCAGATGAGTT 3'. Physical positions given are based on Ensembl A. gambiae genome sequence version 39, converted to the inverted 2La/a form by use of inversion breakpoint coordinates from [8].

Linkage analysis methods are described in detail elsewhere [3]. Despite the fact that each pedigree was divided into groups that were fed on different gametocyte carriers, in the current study all individuals within a pedigree were first analysed together. In one pedigree yielding a linkage signal, the infections from different parasite isolates were also analysed individually. In brief, at every marker, the pedigree members were split into two samples, using an allele or genotype observed at that marker as the selection criterion. At each locus, the distribution of oocyst numbers for each allele/genotype was compared against the pooled oocyst distribution for all other alleles/genotypes. Permutations were used to establish empirical genome-wide p-values. Wilcoxon-Mann-Whitney (WMW), a nonparametric statistical test, was used to determine statistical significance.

A novel and important aspect of this mapping methodolgy is that it examines each locus in isolation from all other loci and does not require knowledge of relative marker position. Thus, the result is robust to the presence of segregating chromosomal inversions, which are common in A. gambiae, and can alter the physical relationship between markers in different individuals within a pedigree. For example, another recent mapping study of P. falciparum resistance in other Kenyan pedigrees [9] used a mapping approach in which valid linkage results are dependent on correct knowledge of marker order, but marker order was not empirically determined. That study assumed a marker order [7] that predated release of the A. gambiae genome sequence. In fact, marker order in the genome sequence differs markedly from that in [7, 9] (e.g., chromosome 2L ref [7, 9] order: AG2H46-AG2H197-AG2H786-AG2H135-AG2H603-AG2H787-AG2H325; genome order and current study: AG2H46-AG2H197-AG2H135-AG2H786-AG2H325-AG2H787-AG2H603).

The statistical analysis was run using both the program OUTBRED_LINES as previously described [3] and a new implementation of the same algorithm in the statistical analysis environment R [10], called OUTBRED_LINES-R. The motivation was to avoid use of approximations employed in OUTBRED_LINES and instead use the built-in exact WMW tests. However, very similar results were obtained from OUTBRED_LINES and OUTBRED_LINES-R. Exact p-values from the R implementation OUTBRED_LINES-R are presented throughout the manuscript.

Six Kenyan pedigrees met the minimum informativeness criteria for genotyping (≥20 mosquitoes surviving to dissection, with infection prevalence ≥ 30%), and these pedigrees were completely genotyped (Figure 2). One of the pedigrees (n = 25, pedigree #57 in [6]) yielded a significant locus for P. falciparum resistance (nominal p = 0.002, genome-wide p = 0.040). The locus displays significant linkage to 5 nearby markers (genome wide p <0.05; from left to right; 2L.H13808912, H796, 2L.17686896, 2L.25C, 2L.25A; Figure 1, Table 1). The Kenyan resistance locus is here named EA_Pfin1 (E ast A frica_ P lasmodium f alciparum Infection In tensity), to distinguish it from the previously named Pfin1 locus, which it overlaps, identified in the chromosome 2L PRI of West African A. gambiae [3, 4]. This mapping with greater marker density shows solid statistical support for the EA_Pfin1 resistance locus from multiple genetic markers (Figure 3). The detection of maximum linkage over a large physical region (~16 Mb) is due to the limited number of recombinants expected in such a small pedigree (n = 25). Despite small pedigree size, the EA_Pfin1 locus had high explanatory power for infection phenotype: the marker allele linked to resistance explained 75% of completely uninfected mosquitoes in the pedigree, and the susceptible allele explained 89% of oocyst-containing mosquitoes (Figure 4). The other five informative pedigrees gave no significant linkage signal (discussed below).

Careful examination of the EA_Pfin1 pedigree indicates that one of the three parasite isolates used failed to infect any mosquitoes in the pedigree, and infected only five mosquitoes of 76 total assayed across all pedigrees [6]. Due to the lack of infectivity of the parasites in this blood sample for the EA_Pfin1 pedigree, the data from mosquitoes fed on this isolate were removed and the pedigree was reanalysed. Without these six mosquitoes, the remaining 19 mosquitoes still give significant linkage signal at the same microsatellite loci despite the small sample size (Table 2). This result indicates that the linkage signal did not depend solely or disproportionately upon defeat by the mosquitoes of one of the three parasite isolates fed to them. Indeed, even with loss of 24% of the pedigree sample size by removal of the six mosquitoes, the linkage signal was still comparable to that from the entire pedigree. The surviving linkage signal in the 19 mosquitoes was due to mosquitoes in the pedigree whose parasite challenge was divided between two different gametocyte carriers. Analysis of the remaining two groups fed on either gametocyte carrier alone did not yield a linkage signal (not shown).

Thus, the detected linkage signal in the entire pedigree or any subset did not result from a strong protective effect by mosquitoes against any single parasite isolate. Instead, detection of linkage required the larger sample size obtained by combining groups of mosquitoes that fed on at least two different parasite isolates, and consequently the signal must have originated in the pedigree's consistent genetic response to at least two different parasite isolates. Taken together, these results indicate that Plasmodium resistance controlled by the A. gambiae EA_Pfin1 locus can be independent of and transcend parasite genotype.

A P. falciparum resistance locus, EA_Pfin1, is reported in East African A. gambiae that resides within the chromosomal Plasmodium-resistance Island (PRI) detected in West African A. gambiae. The phenotype is also the same: parasite elimination prior to oocyst development. The vast majority (>75%) of mosquitoes in the EA_Pfin1 pedigree carrying the resistance allele completely blocked parasite development, and thus were parasite-free. This is comparable to the phenotypic effect size of resistance loci in the 2L PRI in West Africa [4]. Due to their chromosomal co-localization and similarity of phenotype, EA_Pfin1 and the 2L PRI in West Africa likely control the same underlying Plasmodium resistance mechanism.

Here it is established that the genomic region of A. gambiae chromosome 2L, containing strong Plasmodium-resistance loci originally identified in pedigrees from a West Africa population of vector mosquitoes, is not a geographically local phenomenon. At the level of resolution of the genetic mapping, the same genetic factor also exists in East African A. gambiae. The intriguing prospect of shared P. falciparum resistance loci across Africa makes it a promising approach to consider taking advantage of widespread natural Plasmodium resistance mechanisms for the control of malaria transmission [18, 19].