Identification of three single nucleotide polymorphisms in Anopheles gambiae immune signaling genes that are associated with natural Plasmodium falciparum infection

The causative agents of malaria are protozoan parasites of the genus Plasmodium, which are transmitted to humans by anopheline mosquitoes. The highest number of cases occurs in sub-Saharan Africa, where the most deadly parasite Plasmodium falciparum is transmitted primarily by Anopheles gambiae sensu stricto. With over 500 million new cases a year [1] and over half of the world's population at risk for malaria, increased understanding of the complex factors governing transmission is a pressing need.

Extensive genetic structuring among natural populations of A. gambiae is likely to have an impact on parasite transmission. In particular, frequencies of paracentric inversions on the right arm of the second chromosome (2R) have revealed that many A. gambiae populations deviate strongly from Hardy-Weinberg Equilibrium. On the basis of these findings, Mopti, Savanna, and Bamako chromosomal forms were described [2]. Genetic differentiation among these chromosomal forms revealed the presence of molecular forms characterized by fixed nucleotide differences in the intergenic spacer of the X-linked ribosomal DNA [3]. In Mali, the M molecular form corresponds to the Mopti chromosomal form and the S molecular form corresponds to the Savanna and Bamako forms. The distribution of the M molecular form appears mostly limited to West and Central Africa, whereas the S molecular form is found throughout the range of A. gambiae [4]. Gene flow between the S and Mopti-M forms is severely restricted [5, 6].

In addition to chromosomal complexity, the genome of A. gambiae is characterized by high level nucleotide polymorphism. The published genome of A. gambiae PEST strain was reported to have a frequency of single nucleotide polymorphisms (SNPs) of 1.6 × 10^-3 or approximately 1 SNP per 625 bp [7]. Morlais et al [8] found similarly high levels of polymorphism in other laboratory strains of A. gambiae. The most recent genome assembly of A. gambiae in the Ensembl database [9] reports a SNP density of 1 per 247 bp.

In 35 genes putatively involved in mosquito-pathogen interactions and behaviour, Morlais et al [8] identified 460 SNPs, 140 of which encode nonsynonymous substitutions. The authors also examined SNP frequencies in genes encoding fibronectin, thioester-containing protein 3 (TEP3), and peptidoglycan recognition protein (PGRP) in field-caught A. gambiae from sites in Senegal, Burkina Faso and Cameroon where the M and S molecular forms are sympatric [8]. Although the genes selected by Morlais et al [8] have been shown to function in anti-pathogen defense in A. gambiae [10], no data on SNP associations with parasite infection and molecular form - a key variable based on aforementioned predictions of gene flow [5, 6]- were presented.

In addition to studies that have reported SNPs in genes implicated in anti-parasite immunity, a number of studies have taken a genome-wide approach to identify factors that regulate parasite transmission. In particular, Riehle et al [11] determined that colocalized quantitative trait loci (QTL) within a region of chromosome 2L in A. gambiae collected from a single site in Mali accounted for a significant amount of the variation in P. falciparum infection. The authors designated these QTL as a Plasmodium-resistance island or PRI, but also acknowledged that there were numerous other loci throughout the genome that likely contributed to variation in infection [11]. Riehle et al [11] also hypothesized that resistance is the ancestral phenotype, whereas susceptibility to infection with P. falciparum results from mutations in the A. gambiae genome that result in failure of anti-parasite immunity.

Based on these general concepts - that A. gambiae populations exhibit significant genetic structuring, that SNPs are present in anti-pathogen genes in natural populations of A. gambiae, that susceptibility to infection may result from a failure of anti-parasite immunity- and the use of association mapping to link SNPs in functional genes with disease states [12], it was hypothesized that SNPs predicted to alter immune signaling could be linked to natural P. falciparum infection in A. gambiae. Furthermore, the strength of these associations may be a reflection of the different genetic backgrounds of A. gambiae, a phenomenon that will dictate the success of genetic interventions (e.g., creation of refractory mosquito strains) to block parasite transmission in natural populations [13].

To begin testing these hypotheses, SNPs in a subset of immune signaling genes were identified using direct sequencing of conserved domains, and the association of these SNPs with P. falciparum infection in A. gambiae collected in Mali were analysed. This association study revealed significant molecular form-dependent associations of P. falciparum infection with three SNPs in genes encoding Toll5B and an insulin-like peptide 3 precursor. In addition, these analyses revealed an epistatically associated SNP in a gene encoding MKK4 - a mitogen-activated protein kinase that has been predicted to function in the Toll signaling pathway [14‐16]. Hence, these data support the proposed hypotheses and provide new insights on regulatory factors that are associated with control of P. falciparum infection in A. gambiae.

Among the 11 SNPs analysed, the frequency of exact matches between direct sequencing and Luminex for individual genotypes (n = 48) was greater than 98%, indicating that Luminex was highly accurate for genotyping.

Based on SNP genotypes at the 11 loci selected, at least three genetically distinct populations in Mali (Figure 1) were identified. In agreement with population genetic studies based on microsatellites [39, 40], significant genetic differentiation between M and S molecular forms was observed. Further, Pimperena S (S2) form mosquitoes were differentiated from S forms collected in other villages (S1). This outcome may be due to temporal variation in the relative abundance of the Bamako and Savanna chromosomal forms, both of which are the S molecular form and therefore indistinguishable using the Fanello et al [3] molecular form diagnostic.

The association between SNP genotypes at each locus and infection status within each of the three populations indicated in Figure 1 was evaluated using Chi-square and Fisher's exact tests. Infection rates of the genotyped A. gambiae ranged from 3.27% in Selinkenyi to a high of 15.09% in Doneguebougou (Table 1). Genotype frequencies at the Ins35 locus were significantly different between the Pimperena (S2) and S1 groups, while genotype frequencies at the Toll5B1 locus were significantly different between S2 and all other groups (M and S1 in Table 2). All SNPs identified to be associated with infection status had an effect size (h) greater than 1 (Table 3) and power greater than 0.85. Those SNPs that were determined to be associated with population groups had power greater than 0.83 (Table 3). SNPs with smaller h, such as Ins32 and Ins35 (0.75 < h < 1), could be associated with infection if a larger sample size was scored. For example, if 30 infected specimens and 62 uninfected specimens for Ins32 were genotyped and a similar genotype frequency distribution was observed, Ins32 could be identified as significantly associated with infection status. The sample sizes of this study were not sufficient to confirm involvement of SNPs with smaller h - such as Ins32 or Ins35 - in the regulation of P. falciparum development in A. gambiae.

Three SNPs - Ins34 (3L), Toll5B1 (3L), and Toll5B6 (3L) (Table 2) - were found to be associated with P. falciparum infection status. Ins34 is a synonymous SNP resulting in a change from GGC to GGT at nucleotide position 462 in the Insulin-like peptide 3 precursor gene. Toll5B1 introduces a synonymous SNP at nucleotide position 129, changing the codon from ATC, a common codon (28.3%), to ATT, a rarer codon (12.9%). Toll5B6 results in a non-synonymous codon change from AGC (Ser, S) to AAC (Asn, N) at amino acid position 454 in the predicted translation of Toll5B. These SNPs were not identified in either the NCBI or Anobase SNP databases generated by reads from the PEST strain (10× coverage) or from the MOPTI strain (1.2× coverage) aligned to Celera A. gambiae contigs, although Ins34 and Toll5B6 were both present in the M and S form scans available from VectorBase [41]. However, these loci were not identified to be polymorphic in either scan [41]. The CC genotype at the Ins34 locus in M form mosquitoes was more common in samples that were not infected with P. falciparum. In both M and S1 populations, Toll5B1 C alleles were significantly associated with infection status. While the Toll5B1 TT genotype was rare, TC heterozygotes were more common in uninfected A. gambiae than in P. falciparum-infected A. gambiae. Within the S1 population, individuals with the AG genotype at the Toll5B6 locus were less likely to be infected with P. falciparum.

Significant differences in median P. falciparum sporozoite infection intensities were observed in the collected samples (Kruskal-Wallis rank sum test, p = 1.88x10^-5, Figure 2, Table 4). However, no significant association between P. falciparum infection intensity and any SNP locus genotype was observed. The sporozoite infection intensities were highly variable in M and S1 populations as indicated by the standard deviations being more than double their associated means shown in Table 4. Among infected mosquitoes, samples from Pimperena (S2) had higher median sporozoite intensities than those from other villages. VectorBase A. gambiae population data indicated that samples from Pimperena are mostly Savanna chromosomal form while other sites like Doneguebougou and Selinkenyi are composed of a mixture of Bamako and Savanna forms. Thus, the possibility of chromosomal form affecting the density of P. falciparum in the S1 group cannot be ruled out.

Many factors contribute to mosquito infection and successful transmission of malaria parasites, including innate immunity. As such, considerable efforts have been focused on understanding the mosquito immune system and natural variability in these defenses. In humans, numerous reports have demonstrated that natural variability to infection can be associated with single nucleotide polymorphisms (SNPs) in genes that regulate host immunity [42‐44]. Although SNP associations have been described for a variety of human infections and diseases, these have not been identified for P. falciparum infection in genetically defined natural populations of A. gambiae until this study.

The Toll and Imd signaling pathways are important regulators of innate immunity in A. gambiae. In particular, Garver et al [45] reported that, under laboratory conditions, the Toll pathway controlled P. falciparum infection intensity in A. gambiae, while the Imd pathway appeared to regulate resistance to infection. Two Toll5B SNPs - Toll5B1 and Toll5B6 - were significantly associated with P. falciparum infection status among field-collected A. gambiae. Anopheles gambiae Toll5B is orthologous to Drosophila melanogaster Toll-5, also known as Tehao [10], and Toll5 in Aedes aegypti [46]. In Ae. aegypti, Toll5B was inducibly expressed in the mosquito fat body following fungal infection [47]. In A. gambiae, Pinto et al [48] reported that Toll5B expression is significantly upregulated in hemocytes of adult female mosquitoes infected with Plasmodium berghei at 24-28 hours post-infection, a period associated with active parasite invasion of the midgut epithelium. Based on these observations and these data, Toll5B is likely to be responsive to P. falciparum infection in A. gambiae under natural conditions.

Both Toll5B SNPs showed divergence from Hardy-Weinberg equilibrium. In particular, Toll5B1 showed significant divergence from Hardy-Weinberg expectation in both the M (p = 0.00016) and S1 (p = 0.00077) forms, while Toll5B6 showed significant divergence from Hardy-Weinberg expectation in the M form (p = 0.00035; Table 5), suggesting that selection pressure(s) acting on these loci may skew the genotype frequencies of this gene. For Toll5B1, the observed heterozygosity was greater than expected, both in M and S1 groups (Table 5). For this SNP, CT heterozygotes were more common (96%) in uninfected samples while 72% of infected samples were CC homozygotes. TT homozygotes were very rare and only one sample of this genotype was identified in the entire collection. These circumstances indicate the possibility of heterosis or balancing selection on this locus. For Toll5B6, the GG homozygote was the most common form in all groups and the proportion of GG was highest in the M uninfected group. The effect size of this SNP was 0.514, so a larger sample size (83 infected and 183 uninfected) would be needed (Table 3) to confirm involvement of this SNP with respect to infection status in the M form population.

Toll5B1 occurs in a highly polymorphic region, containing 13 polymorphic loci within a stretch of 166 bp. Six of these SNPs occur in clusters of rare codons or result in the change from a common codon to a rare one. Clusters of rare codons have been shown to alter protein production in a synergistic manner [49‐51]. Genotype frequencies of Toll5B1 and infection associations of this SNP were significantly different between Pimperena (S2) and A. gambiae M and S1 groups (Table 2). Taken together with observations on codon frequency and protein function, these data suggest that within the M and S1 molecular forms, this SNP may have a functional effect on P. falciparum infection, but this association may be population-specific, as indicated by the lack of any Toll5B1-P. falciparum infection association in the Pimperena (S2) population. Toll5B6 is a non-synonymous SNP that introduces the amino acid change S454N. The replacement of serine by the more bulky asparagine could introduce subtle changes in the 3-dimensional structure of the Toll leucine-rich repeat (LRR) in which this mutation occurs.

Additional analyses revealed that MKK43, a SNP in the MAPK kinase MKK4 gene, located on chromosome 2, was in linkage disequilibrium with infection-associated Toll5B1, a SNP on chromosome 3, in the S1 population (p = 0.0001). In mammals, MKK4 can be activated by TLR3 [16] and TLR2 signaling [14, 15], indicating that MKK4 activation is functionally linked to Toll signaling. Downstream of this activation, MKK4 functions with MKK7 as the primary activator of c-Jun N-terminal kinase or JNK, one of three immunity-associated MAPKs. In D. melanogaster, Toll activation of MKK4/7 and JNK during septic injury regulates cytoskeletal genes typically associated with a wound healing response to infection [52]. A potential Toll/MKK4/JNK signaling module - if biologically functional in A. gambiae - could be linked to the profound cytoskeletal changes in the midgut epithelium that have been described during P. berghei and P. falciparum infection of laboratory and field specimens of A. gambiae, respectively [53‐55]. This possibility is currently being investigated.

In mammals, the insulin/insulin-like growth factor signaling cascade (IIS) has been shown to regulate innate immunity through Toll- and NF-κB-dependent pathways [56, 57]. In particular, IIS activation can induce or inhibit NF-κB-dependent signaling and is, thereby, capable of exerting both pro- and anti-inflammatory effects on the host immune response. In Anopheles stephensi, control of malaria parasite development is regulated by signaling proteins associated with the IIS cascade [[22], Corby-Harris et al. unpublished]. The IIS is highly conserved [23] and critical components of the cascade, as well as a variety of insulin-like peptides, including insulin-like peptide 3 precursor, are expressed in the midgut of A. gambiae [[58]; Luckhart, unpublished].

Four SNPs in the insulin-like peptide 3 precursor gene - Ins32, Ins33, Ins34 and Ins35 - were analysed and significant infection (Ins34) and molecular form (Ins35) associations were found for two of these (Table 2). Ins34 introduces a synonymous mutation, with no remarkable change in codon frequency. Although Ins34 is predicted to have little to no effect on protein function, this infection-associated SNP was in linkage disequilibrium in the S1 population with Toll5B2 (p = 0.03307, Additional file 2), a SNP that introduces the non-synonymous mutation D56A into Toll5B. This SNP encodes a mutation close to the N-terminus and outside of the predicted LRRs, in an ectodomain region identified in other Toll proteins as the cysteine-rich capping structure [59]. The N-terminal capping structure may participate in protein-protein interactions that are critical for Toll receptor function [59].

Within the S1 population, Ins32, Ins33, and Ins35 showed levels of linkage disequilibrium with Toll5B2 similar to that observed for Ins34 (p = 0.01238, 0.02931, 0.00851, respectively, Additional file 2), which likely reflects the fact that these loci are closely linked physically. Although Ins35 was not significantly associated with infection status (Table 2), this SNP was also in linkage disequilibrium with Toll5B2 in the S2 population (p-value = 0.00406, Additional file 2). However, in the S2 population, the neighboring SNPs - Ins32, Ins33, Ins34 - did not show similar levels of linkage disequilibrium with Toll5B2 (p = 0.09584, 0.14396, 0.13238, respectively, Additional file 2), suggesting that the associations of Ins34 and Ins35 with Toll5B2 may, in fact, represent novel biological functionality that is population-specific in A. gambiae.

The positional effects of SNPs in the architecture of signaling cascades have been investigated in a series of relevant studies. Riley et al [60] examined SNPs in the D. melanogaster Ras-mediated signal transduction pathway. The least polymorphic signaling protein genes (e.g., Ras, Dr, and Polehole) were those that were proximal to the origin of signaling at the cell surface, while the most polymorphic genes were located farther downstream in the signaling cascade (e.g., Dsor1, Csw and Ksr; [60]). Computational simulations of MAPK signaling predicted similar constraints in that if signal amplification was crucial, the upstream signaling proteins were more constrained than were the downstream components [61]. Together these findings suggest that a SNP in an upstream component of a signaling cascade, such as Toll or an ILP, could have pleiotropic effects on the downstream components of the signaling cascade and an increased potential for a deleterious outcome [62]. In studies of the Tor signaling pathway, however, selection constraints were found to be greater in the downstream components rather than the upstream components of the cascade [62]. Regardless of the pattern of selection constraints, the polarity of a signal transduction pathway is an integral part of determining the effect of a SNP. As such, the molecular cell biology of the signaling pathway protein networks highlighted herein can complement studies of phenotype-associated SNPs in natural A. gambiae populations.

In summary, the infection-associated SNPs that have been identified here could be used as genetic markers for the susceptibility of an anopheline population to malaria parasite infection. More importantly, however, these SNPs confirm previous laboratory associations of the target genes with the regulation of parasite infection in A. gambiae, extending possible functional linkages among these target genes and confirming the influence of population structuring on biological associations of the SNPs under study. In particular, these data together with associated laboratory data that implicate these gene products in anti-parasite immunity suggest that Toll5B, MKK4 and insulin-like peptide 3 alone or perhaps in some combination may be involved in the regulation of P. falciparum development in A. gambiae under natural conditions. The alternative hypothesis - that the true functional genes are linked to these SNPs - will be evaluated as haplotype maps are developed. In addition, these data have revealed that SNPs are not equally distributed among A. gambiae molecular forms and that selection pressures may be driving deviations from Hardy-Weinberg equilibrium. These insights suggest that gene flow may impede the spread of transgenes under field conditions and that selection may alter the distribution of transgenes, factors that must be accommodated in the development of any field strategies.