In life science research, model organisms greatly facilitate the identification of fundamental mechanisms that can be tested later in less tractable systems. For example, the key developmental mechanisms that were discovered in Drosophila melanogaster were later shown to be pertinent, even if not invariant, to most metazoa. The mouse model has been instrumental in biomedical research, even though its distinct physiology requires that discoveries in the mouse be validated in primates or humans before potential application to human medicine. Similarly, for complex diseases such as malaria in which three organisms—the human host, the Plasmodium parasite and the Anopheles vector—are implicated, model organisms greatly facilitate discovery. In recent years, the development of genomic and functional genomic approaches to malarial research has favoured the use of the principal human vector A. gambiae, although in combination with tractable non-human hosts and parasites. Such studies have explained important mechanisms of parasite killing in vectors. Testing their relevance to the human malarial system is now the primary goal.

Plasmodium parasites need to overcome several bottlenecks in the vector that result in most wild mosquitoes bearing few, if any, oocysts—that is, showing low infection prevalence and low parasite load (Vaughan et al, 1992; Gouagna et al, 1998; Sinden, 1999). Studies on the immune system of A. gambiae have shown that a fine balance exists between mosquito factors that affect ookinete survival to the oocyst stage, positively (agonists) or negatively (antagonists; Blandin et al, 2004; Osta et al, 2004). Such factors, and the molecular mechanisms they serve, might prove to be favourable targets for novel interventions towards blocking malaria transmission.

Putative components of the A. gambiae immune system were identified by using comparative genomic analysis (Christophides et al, 2002; Zdobnov et al, 2002) combined with transcriptomic studies of the A. gambiae response to Plasmodium berghei infections (Oduol et al, 2000; Christophides et al, 2002; Dimopoulos et al, 2002; Vlachou et al, 2005). Functional analysis by RNA interference-mediated gene silencing showed that certain genes, which are upregulated in the mosquito during ookinete invasion of the midgut, participate in humoral mechanisms of P. berghei killing in the vector. Thus, silencing of the antagonist LRIM1 (leucine-rich repeat immune protein 1) gene markedly increases the number of developing oocysts (Osta et al, 2004), as does the silencing of the complement-like gene thioester-containing protein (gene) 1 (TEP1; Blandin et al, 2004). Both LRIM1 and TEP1 encode circulating haemolymph (blood) proteins that favour the killing of ookinetes by lysis and melanization after midgut invasion. By contrast, two circulating C-type lectins—C-type lectin 4 (CTL4) and CTL mannose binding 2 (CTLMA2)—inhibit the melanization response and protect the development of ookinetes to the oocyst stage (Osta et al, 2004). More recent studies have established that antagonistic immune reactions are controlled, at least in part, by immune signalling pathways such as the Relish/Imd pathway (Meister et al, 2005). The number of P. berghei oocysts in A. gambiae is also regulated by local midgut epithelial responses that engage diverse—both negative and positive—factors (Vlachou & Kafatos, 2005; Vlachou et al, 2005).

The mean number of oocysts per midgut in all control group mosquitoes (even when lacking oocysts) was 3.02 (s.e.=0.37); the oocyst load (mean oocyst number among mosquitoes showing at least one oocyst) was 4.98 (s.e.=0.47). These values are consistent with previous data on natural infections with P. falciparum in human blood (Tahar et al, 2002; Gouagna et al, 2004; Boudin et al, 2005; Lambrechts et al, 2005). By contrast, infections of the same vector with P. berghei frequently show mean oocyst numbers higher than 50 (Blandin et al, 2004; Osta et al, 2004; Abraham et al, 2005). This difference, between the laboratory model and the human malaria system in the field, might be inherent to the parasite itself or the host–parasite interactions.

Kolmogorov–Smirnov tests showed that at day 7 after infection, the distributions of developing P. falciparum oocysts in control and CTL4, CTLMA2 or LRIM1 knockdown mosquitoes are statistically similar (Table 1). Additionally, melanized ookinetes were not detected in the CTL knockdowns. These results are in contrast with our previous results from gene-silencing laboratory experiments on P. berghei infections (Osta et al, 2004). These previous experiments were carried out in the G3 strain, which originated from a mix of various populations of M and S molecular forms of A. gambiae and is polymorphic for chromosomal inversions. The G3 strain differs considerably from the Yaoundé strain used here, which is the offspring of a pure, standard chromosomal and molecular M form population from Cameroon. As an appropriate control, we used P. berghei to infect Yaoundé mosquitoes and confirmed both their high susceptibility to P. berghei and pronounced effects of gene silencing (Table 2); the results were comparable with those reported previously for infections of the G3 strain (Osta et al, 2004). Therefore, the effects of gene silencing on P. berghei and P. falciparum infections differ because of the parasite species and not the mosquito strain.

We compared the expression levels of CTL4, CTLMA2 and LRIM1 24 h after feeding Yaoundé mosquitoes with human blood, which was either non-infected or infected with P. falciparum. At this crucial time, parasites emerge at the basal surface of the midgut, come into contact with haemolymph components and either begin the transformation into oocysts or die. The data from two independent experiments showed an average of 2.4-, 1.7- and 1.5-fold upregulation of these genes in the carcasses of infected mosquitoes (Fig 1). A comparable level of upregulation in carcasses—where these genes are predominantly expressed—was detected in previous infections with P. berghei (Osta et al, 2004). Carcasses are enriched in haemocytes (mosquito blood cells) that predominantly expresses these genes (M.A.O., unpublished data). The midguts showed low expression levels of these genes and no significant upregulation on infection. It should be noted that, at least with respect to P. berghei, the upregulated genes include both a parasite antagonist (LRIM1) and two agonists (CTL4, CTLMA2). Thus, transcriptional upregulation does not correlate directly with the outcome of infections, although further studies are warranted to test this conclusion more rigorously.

In conclusion, we have observed an important difference in the impact of silencing immunity genes on infection levels between the laboratory model (rodent) parasite and the sympatric natural (human) parasite isolates. Three A. gambiae genes encoding systemically circulating proteins, which strongly influence P. berghei development (positively or negatively), did not show comparable effects on P. falciparum infections in the same mosquito species and strain. It could be argued that the latter result might be due to the low infection levels of that parasite in A. gambiae. Future experiments to test this hypothesis definitively, by manipulating infection levels for both parasites, are conceivable but difficult—especially in the field—and are required to address this hypothesis. At least in terms of transcription, P. falciparum indeed affects the expression of some immune genes (Fig 1). Furthermore, an extensive series of studies on CTL4 failed to show a correlation between the degree of P. berghei ookinete melanization and infection levels (M.A.O., unpublished data). At present, we favour the hypothesis that the observed contrasting phenotypes might reflect co-evolution of the human parasite and its natural vector. Indeed, P. falciparum is naturally transmitted, but P. berghei might encounter A. gambiae only accidentally. This rodent parasite is naturally transmitted by A. dureni in Central Africa (reviewed by Gillies & De Meillon, 1968), and is unlikely to encounter the highly anthropophilic A. gambiae. The opposite is true for the human parasite P. falciparum, which shares an evolutionary history with the strongly anthropophilic A. gambiae, potentially resulting in unique co-adaptations. This supposition is consistent with the tantalizing observation that the A. gambiae refractory strain L3-5 is able to melanize numerous species of Plasmodium, including non-African P. falciparum isolates, but not—or only poorly—African P. falciparum isolates that are sympatric with A. gambiae (Collins et al, 1986). Co-adaptation would have favoured low infection levels if, as proposed by Boëte (2005), both parasite and host benefit by limiting the parasite numbers; infection is costly for the vector (Anderson et al, 2000; Ferguson & Read, 2002) and reducing the burden might favour transmission to the vertebrate. The bottleneck that natural parasites suffer at the ookinete-to-oocyst transition (Gouagna et al, 1998) might at first be interpreted as evidence for selection of host responses that reduce the parasite load. However, such selection is unlikely to be unconstrained: an excessive immune response could be detrimental not only for the parasites but also for the host itself, as immune system overactivation might have costly side effects (Moret & Schmid-Hempel, 2000; Schmid-Hempel, 2003). Thus, co-adaptation might have benefited both parties by favouring a robust equilibrium on the basis of limited, but sufficient to keep the parasite load low, responses of several A. gambiae genes (not just the ones we tested) to P. falciparum. Our data suggest that, despite the lower infection levels compared with P. berghei, P. falciparum induces an immune response of A. gambiae genes including, but probably not limited to, upregulation of LRIM1, CTL4 and CTLMA2 24 h after blood feeding. In fact, a recent genome-wide comparison of A. gambiae expression profiles after infection with P. berghei and P. falciparum parasites (Dong et al, 2006) showed that the former induces a wide and profound impact on the transcriptome, whereas the latter elicits a more specific immune response, possibly suggesting co-adaptation of the immune system to P. falciparum infection.

It remains unclear whether natural parasite–mosquito interactions are sufficiently distinct to preclude the possibility of extrapolating results obtained in P. berghei–A. gambiae to the natural transmission system. Future work will need to test whether other potent circulating immune-active proteins of the mosquito, such as TEP1 (Blandin et al, 2004) or RFABG (Vlachou et al, 2005), also show disparate effects on the rodent model compared with field isolates of the human parasite. In addition, proteins that are synthesized and locally active within the invaded epithelium, such as cytoskeleton-modifying components (Vlachou et al, 2005), also need to be tested. Genes in quantitative trait loci, shown by investigations in field conditions, such as Anopheles Plasmodium-responsive leucine-rich repeat 1 (APL1; Riehle et al, 2006) are also excellent candidates for affecting P. falciparum in natural transmission systems.

Gene silencing was achieved by injecting 207 ng of dsRNA into the thorax of 1-day-old females, as described previously (Blandin et al, 2002). In each of five replicates carried out, 60–100 females were injected with dsRNA for each of the genes tested. DsRNA for GFP was used as reference. The efficiency of gene silencing was determined as described by Osta et al (2004).

Plasmodium falciparum gametocyte carriers-recruitment and experimental infection. P. falciparum gametocyte carriers were recruited among 5- to 12-year-old children at schools in Mfou, a town located 30 km from Yaoundé city in an area endemic for malaria. Blood samples were collected by finger-prick from each volunteer and thick blood smears were stained with Giemsa and examined by microscopy for the presence of P. falciparum. Sexual and asexual stages were counted in observed fields that cumulatively contained at least 500 leukocytes; an estimate of parasite density was obtained by assuming a standard number of 8,000 leukocytes/μl of blood. Children with asexual parasitaemia exceeding 1,000 parasites/μl were immediately treated with the amodiaquine and artesunate drug combination according to national guidelines. Asymptomatic gametocyte-positive children were enrolled as volunteers. The recruitment procedures were approved by the Cameroonian and WHO ethical review committees.

For each experimental replicate, mosquitoes were injected with dsRNA and 4 days later were fed on blood from different P. falciparum gametocyte carriers. Five millilitres of venous blood was collected in heparinized vacutainers. To limit the effect of human factors, such as transmission-blocking immunity (Boudin et al, 2005), the blood was centrifuged for 3 min at 2,000 r.p.m. and the serum was replaced with the same volume of AB serum from a French donor with no exposure to malaria. Mosquitoes were starved for 16 h and allowed to feed for 15 min on this mixture using standard membrane-feeding (Tchuinkam et al, 1993). Unfed and partially fed mosquitoes were discarded. After 7 days, mosquitoes were dissected and midguts were stained with 0.4% mercurochrome in distilled water to count oocyst numbers by light microscopy. Experiments were considered successful when the infection prevalence of control mosquitoes (GFP dsRNA) was at least 30%. Mosquito infections with the P. berghei GFP-CON strain (Franke-Fayard et al, 2004) were carried out as described by Osta et al (2004).