Innate immune‐related gene expression in the major disease vector mosquito Anopheles gambiae has been analyzed following infection by the malaria parasite, Plasmodium berghei. Substantially increased levels of mRNAs encoding the antibacterial peptide defensin and a putative Gram‐negative bacteria‐binding protein (GNBP) are observed 20–30 h after ingestion of an infected blood‐meal, at a time which indicates that this induction is a response to parasite invasion of the midgut epithelium. The induction is dependent upon the ingestion of infective, sexual‐stage parasites, and is not due to opportunistic co‐penetration of resident gut micro‐organisms into the hemocoel. The response is activated following infection both locally (in the midgut) and systemically (in remaining tissues, presumably fat body and/or hemocytes). The observation that Plasmodium can trigger a molecularly defined immune response in the vector constitutes an important advance in our understanding of parasite–vector interactions that are potentially involved in malaria transmission, and extends knowledge of the innate immune system of insects to encompass responses to protozoan parasites.
The mosquito Anopheles gambiae is the principal vector of human malaria in Africa, where the disease is estimated to cause more than 1 million deaths per year (Collins and Paskewitz, 1995). The absence of effective malaria vaccines, and the spread of drug‐resistant Plasmodium parasites as well as insecticide‐resistance in vector populations, render the need for novel malaria control strategies increasingly acute. Historically, successful malaria control has been effected through management or control of vector mosquito populations, thereby breaking the cycle of human disease transmission (Collins and Besansky, 1994; Collins and Paskewitz, 1995). Advances in the molecular genetic manipulation of certain insect species have led to speculation that malaria could be controlled through genetic alteration of Anopheline vector mosquitoes rendered refractory to Plasmodium growth and differentiation (Collins, 1994; Curtis, 1994). Research efforts directed towards such a goal will greatly benefit from enhanced understanding at the molecular level of the interactions between vector insect and protozoan parasite. These interactions are also of broad biological interest, as they relate to the central paradox of metazoan parasitism: successful growth and differentiation of a parasite within a host endowed with elaborate defense mechanisms.
Plasmodium enters the female mosquito during blood‐feeding upon an infected vertebrate host. Within the insect midgut, parasite gametocytes are rapidly activated to produce gametes. Fertilization follows, leading to formation of a motile ookinete which penetrates the epithelial cell monolayer surrounding the gut lumen. The ookinete ceases its migration upon reaching the basal lamina separating the midgut and hemocoel compartments. Here the ookinete undergoes differentiation into the oocyst form, which grows over a period of days and produces sporozoites; ultimately these are released into the hemolymph to colonize the salivary gland, where they become the only insect stage of the parasite that is again infective to vertebrates upon subsequent blood‐feeding by the mosquito (Touray et al., 1992). The great majority of Plasmodium–mosquito combinations are incompatible, and it is thought that multiple factors contribute to the inherent capacity of a particular Anopheline mosquito species to support growth and differentiation of a particular Plasmodium parasite (Warburg and Miller, 1991). At present these factors are not well understood at the molecular level, although laboratory studies have documented that several distinct types of incompatibility in some mosquito–Plasmodium combinations have a relatively simple genetic basis (Kilama and Craig, 1969; Vernick et al., 1989, 1995; Zheng et al., 1997). At the descriptive level, incompatibility might be ascribed to physiological or structural barriers, or to active innate defense (immune) responses of the insect, which result in killing specific stages of the parasite (Warburg and Miller, 1991).
It has long been recognized that insects possess highly effective innate defense mechanisms of both cellular and humoral nature (Boman and Hultmark, 1987; Lackie, 1988; Hoffmann et al., 1996). The latter has received considerable recent attention and much is now known about the structure and activity of secreted antimicrobial peptides, as well as their regulated synthesis in model dipteran and lepidopteran species upon bacterial or fungal infection (Cociancich et al., 1994; Hultmark, 1994; Hoffmann et al., 1996). Studies of humoral immunity in medically important vector species are more limited and recent (Kaaya et al., 1987; Chalk et al., 1994; Ham et al., 1994; Lowenberger et al., 1995; Richman and Kafatos, 1996). In A.gambiae they include the cloning of sequences encoding the antimicrobial peptide defensin (Richman et al., 1996), lysozyme protein (Kang et al., 1996), a putative immune‐related transcription factor (Barillas‐Mury et al., 1996) and several mRNAs upregulated after bacterial infection (Dimopoulos et al., 1996, 1997; and unpublished observations). The latter group includes the A.gambiae homolog (Dimopoulos et al., 1997) of a Gram‐negative bacteria‐binding protein (GNBP) identified originally in the silk moth Bombyx mori (Lee et al., 1996). Exogenously administered antimicrobial peptides have been shown to be active against various insect stages of Plasmodium (Gwadz et al., 1989; Rodriguez et al., 1995). The availability of molecular markers, in particular defensin and GNBP, has permitted us to investigate the possibility that immune activation may be involved in insect–protozoan interactions.
Defensin and GNBP RNA levels are induced locally and systemically during midgut tissue invasion
Figure 1A demonstrates elevated RNA levels for two immune markers in mosquitoes fed on Plasmodium berghei‐infected as compared to naive mice. In this and subsequent experiments reverse transcription PCR (RT‐PCR) was used as a convenient, sensitive, and semi‐quantitative method to analyze changes in gene expression (Dimarcq et al., 1994; Richman et al., 1996); levels of immune marker RNAs were normalized relative to a control RNA encoding the ribosomal protein S7 (Salazar et al., 1993). Figure 1B confirms this induction using quantitative RNA blot analysis of defensin RNA (the low abundance of GNBP mRNA renders blot analysis of this sequence impractical; data not shown). The expression of both defensin and GNBP is reproducibly induced 20 to 30 h after ingestion of infected blood. For defensin, induced levels are comparable to those observed in larval and adult stage mosquitoes following experimental infection with bacteria (Richman et al., 1996). Defensin induction is not yet observed at 14 h after P.berghei infection, but is prominent at 24 h (Figure 1C). Thus, induction corresponds to the period when motile ookinete stage parasites are invading the epithelium of the posterior midgut (Meis et al., 1989 and data not shown). Between 40 and 48 h post‐feeding defensin induction is variable, possibly reflecting variable levels of late midgut invasion by a minority of ookinetes (data not shown). Induction is not due to ingestion of a normal blood‐meal alone, as it is evident in comparison to mosquitoes fed upon uninfected mice.
While Figure 1 shows induction in intact mosquitoes, Figure 2 demonstrates that P.berghei infection induces defensin and GNBP RNA levels both locally in the midgut, and systemically in the remaining carcass. Carcass expression is presumably occurring primarily in cells of the fat body and also in some hemocyte cell types, as has been observed in model insect species challenged with bacteria (Hoffmann et al., 1996).
Immune gene induction results from ingestion of infectious Plasmodium parasites, and is not elicited by vertebrate host factors or endogenous microorganisms
Rodent malaria elicits significant alterations in blood physiology of the vertebrate host, particularly the loss of red blood cells (Garnham, 1966). Levels of cytokines and acute phase response proteins may also be altered in the vertebrate, and opportunistic infectious microorganisms may be present. It is therefore conceivable that blood ingested from infected as opposed to naive mice may contain host factors or biological agents other than Plasmodium, which are themselves capable of eliciting mosquito immune induction. To control for this possibility, the mosquito responses to two different strains of P.berghei (Paton et al., 1993) were compared: a strain (Anka 2.33) which undergoes normal asexual development in the mouse but does not produce any gametocyte stage parasites (and is thus non‐infective for the mosquito), and a ‘wild type’ strain (Anka 2.34) which produces high numbers of gametocytes. Figure 3 compares the levels of gene activation of defensin and GNBP in A.gambiae challenged with these non‐gametocyte producing (G–) and wild type (WT) strains of P.berghei, assayed at approximately equal levels of parasitemia. Both RNAs are substantially enhanced in mosquitoes infected with the wild type as compared with the non‐infectious parasite strain. The difference in infectivity of the strains was confirmed by the ultimate appearance of oocysts in the midgut of mosquitoes fed on the WT but not the G– strain (Figure 3). In the WT strain, the vertebrate asexual stage parasites are preponderant relative to gametocytes (Janse and Waters, 1995). Taken together, the results comparing WT and G– strains, and the time course studies, indicate that the observed immune response is elicited by infectious ookinetes rather than vertebrate host factors, erythrocyte stage parasites, or compromised blood‐meal quality (such as anemia).
Laboratory colonies of Anopheline mosquitoes have been reported to harbor endogenous bacteria (Seitz et al., 1987). Conceivably, immune gene induction might be due to opportunistic bacterial co‐penetration of the midgut during parasite invasion. Mosquitoes reared in the EMBL insectary were examined for the presence of microorganisms. Figure 4C shows ookinetes in the midgut lumen of fed mosquitoes reared under standard conditions: ookinetes are observed in association with rod‐shaped microorganisms, presumably bacteria. Fungal infection was not observed (data not shown). Antibiotic treatment of mosquitoes prior to blood‐feeding visibly eliminated gut bacteria (Figure 4A), without affecting development of ookinete stage parasites (Figure 4B). In the absence of bacteria, P.berghei infection strongly induced defensin RNA levels 26 h after infection (Figure 4, bottom panels), confirming that the response is due to the parasite. The relative induction of defensin by Plasmodium in antibiotic‐treated mosquitoes was unusually high: ∼12‐fold as compared with the ∼3‐fold induction shown in Figure 1. This difference may be due in part to the effect of resident gut bacteria upon the background levels of defensin RNA expression (comparing the RNA blots in Figures 1 and 4, defensin RNA levels in antibiotic‐treated control mosquitoes are ∼4.5‐fold lower than observed in untreated control animals). The relevance of precise quantitative comparison of induced defensin levels is further limited by possible differences in infectivity of the mice used in the experiments, reflected in differing parasitemias and average numbers of exflagellation centers (see legends to Figures 1 and 4). The important conclusion is that the response to Plasmodium is robust, as shown by the data presented here from five independent experiments.
Mosquito–parasite interactions as related to innate immunity
The insect innate defense system represents a potentially formidable obstacle to the survival and growth of infecting microorganisms and eukaryotic parasites, in particular those which, like Plasmodium, undergo profound developmental changes associated with the invasion of multiple host tissues. Plasmodium transmission requires that the vector midgut provides an environment permissive for sexual reproduction and subsequent coordinated differentiation, translocation, and multiplication events leading to sporozoites: it also requires the survival of the sporozoites as they traverse the hemocoel and specifically ‘colonize’ cells of the salivary gland, where they acquire vertebrate infectivity (Touray et al., 1992). In principle, successful parasites may circumvent vector defense mechanisms during their developmental progression in a number of ways: by evading recognition, by suppressing the response, or through insensitivity to effector molecules (Loker, 1994). The inherent ease of triggering the defenses may also differ in different mosquitoes. The importance of potential insect defenses is highlighted by what happens to the parasite in non‐permissive (refractory) mosquitoes. Two distinct types of refractoriness have been documented in A.gambiae: lysis of migrating ookinetes within midgut epithelial cells (Vernick et al., 1995), and melanotic encapsulation of Plasmodium at a later stage as they initiate oocyst development (Collins et al., 1986). These responses are genetically as well as morphologically distinct. Three genetic loci have been mapped recently, which cooperate to impart to the mosquito the capacity for semi‐dominant melanotic encapsulation of Plasmodium; none of these loci are involved in prior killing of invading ookinetes (Zheng et al., 1997). Studies in other insects have documented a variety of additional immune responses, both cellular and humoral: phagocytosis, nodule formation, cellular encapsulation and secretion of antimicrobial peptides (Boman and Hultmark, 1987; Lackie, 1988; Hoffmann and Hoffmann, 1990). Recent studies in Drosophila melanogaster have shown that the humoral response is itself regulated through multiple genetic pathways (Lemaitre et al., 1995; Hoffmann et al., 1996). In contrast to the dramatic refractory phenotypes, the nature of the specific molecular pathways that underlie them remains to be defined; similarly, the molecular mechanisms whereby parasites circumvent the immune mechanisms of susceptible mosquitoes are unknown.
Localized and systemic response of the mosquito to Plasmodium
We have documented that infection of the mosquito by the rodent malaria parasite P.berghei results in robust induction of RNAs that encode the known immune effector molecule defensin and the immune marker GNBP. In parallel experiments, this response has been detected in both encapsulation‐refractory and susceptible mosquitoes (data not shown); although these parallel studies have not been quantitative, and none have been conducted at the protein as opposed to the RNA level, it appears likely that the encapsulation response is regulated independently of the particular immune pathway(s) that we have tested thus far. Be that as it may, our observations are important because they constitute the first molecular evidence that the Plasmodium parasite does not evade detection by the (poorly understood) immune surveillance mechanisms of the mosquito.
We have shown that the observed immune response is due to parasite infection, rather than blood‐feeding per se, prior effects of the parasite on the vertebrate host, the presence of vertebrate stage parasites in the blood‐meal, or opportunistic bacterial infections during Plasmodium invasion of the midgut. The full physiological significance and scope of the observed response remain to be determined in future studies. It will be of interest to examine the interaction between A.gambiae and P.falciparum (the major human pathogen) relative to the results reported here using a model rodent malaria system. Moreover, it is not known whether the induction of gene expression results from receptor‐mediated immune recognition of P.berghei by the mosquito, or is a consequence of midgut injury associated with parasite invasion. Mechanical injury is associated with activation of humoral immunity in insects (Boman and Hultmark, 1987).
A second important conclusion of the present work is that mosquito midgut cells can themselves mount a robust immune response to parasite infection. While antimicrobial peptides have been reported in the alimentary systems of some insects (Kylsten et al., 1992; Tryselius et al., 1992; Dunn et al., 1994) they have not been identified previously in the mosquito midgut. In vitro studies on midgut explants exposed to bacteria have provided further evidence that the A.gambiae midgut is an autonomous, immune‐competent organ (data not shown). The alimentary tract is the first organ system that ingested parasites (indeed ingested microorganisms) encounter, and thus it is not surprising that it can also function as a first line of immune defense. This function clearly bears close analysis. It may well be related to the prominence of midgut‐localized parasite refractoriness mechanisms whose molecular nature remains unknown.
Induction of defensin RNA, and indeed the presence of active defensin peptide (Richman et al., 1996) in whole extracts of parasite‐susceptible mosquitoes, may seem paradoxical in view of evidence that exogenously administered insect defensin can kill Plasmodium in the mosquito; interestingly, the pertinent studies show that sensitivity of the parasite is stage‐specific and that only mature oocysts and sporozoites are killed (M.Shahabbudin, personal communication). It is worth recalling that the difference between susceptible and refractory mosquitoes is quantitative: although refractory mosquitoes can completely prevent parasite transmission, massive parasite losses occur at every stage of its development even within a susceptible mosquito (Vaughan et al., 1994). It appears that in any mosquito, whether or not the ingested parasites can lead to malaria transmission depends on a balance between multiplication and differentiation of the parasite, and its elimination by physical barriers and active innate immune responses. Quantitative studies of parasite‐induced endogenous defensin peptide levels in different mosquito strains, particularly in the midgut, will be needed to clarify the role of this factor in defense against Plasmodium. It will also be interesting to compare the repertoire of antimicrobial peptides in the midgut immune responses of different mosquitoes as related to the sensitivity of their co‐adapted parasite species. It appears that defensin is the major antimicrobial peptide in adult mosquitoes (Lowenberger et al., 1995; A.Richman, unpublished observations) and it is tempting to speculate that Plasmodium may have evolved mechanisms conferring specific insensitivity to defensin during the period of ookinete‐induced expression in the midgut.
The third important observation in the present work is the marked ‘systemic’ immune response in the carcass, at a time when the parasite is confined within the midgut compartment by the basal lamina. The possibility of an immune‐related signalling process between tissues, whether involving diffusible parasite products or specific factors of insect origin, is worth considering. Cytokine‐like activities have been described in some invertebrate species (Raftos et al., 1991; Granath et al., 1994), and it has been proposed that in Drosophila, expression of the gene encoding the antifungal factor drosomycin is mediated through soluble ligand binding and activation of a cell surface receptor (Lemaitre et al., 1996).
The observed activation of two immune‐responsive genes by a parasite entering via the natural route of infection for hematophagous insects clearly establishes a link between the better characterized invertebrate response to experimental bacterial infection, and the mosquito response to challenge by Plasmodium. In follow‐up studies, isolation and utilization of additional markers representing various independent immune response pathways should prove revealing. In addition, high‐resolution spatial analysis of gene expression patterns will be needed to identify the immune responding cells, both in the midgut and in the carcass, thus characterizing in molecular and cellular terms the multifaceted interactions between mosquito and parasite.
Materials and methods
Mosquito rearing and parasite infection were performed essentially as described (Sinden, 1997). The rodent malaria P.berghei provides a versatile, readily manipulated and safe model system for the study of A.gambiae–Plasmodium interactions (Janse and Waters, 1995). P.berghei strains Anka 2.33 and 2.34 were passaged in Balb/c mice. Experiments were performed with Anka 2.34 unless otherwise specified. All care and handling of experimental animals was in accordance with the guidelines of the European Molecular Biology Laboratory. The parasitemias of infected mice were determined from Giemsa‐stained blood films prepared from tail blood. The average number of exflagellation centers per field was used as an index of mouse infectivity for mosquitoes. Infectivity of a given feed was further monitored by determining the incidence and density of oocysts in a sample of dissected midguts 8–10 days following infection. The melanotically encapsulating A.gambiae strain L3‐5 was used throughout with one exception: strain G3 was used to demonstrate differential oocyst formation (see Figure 3, lower panel).
Antibiotic pretreatment and blood‐meal analysis
Five‐ to eight‐day‐old female mosquitoes were maintained for 4 days on 10% sucrose plus penicillin (100 U/ml)/streptomycin (100 μg/ml) in phosphate buffered saline (PBS). 18 h after blood‐feeding midguts were dissected out, the contents suspended in a small volume of PBS and ‘smeared’ on a glass slide, fixed in methanol, and Giemsa stained.
Twenty‐four hours after blood‐feeding, mosquitoes were anaesthetized on ice and dissected in ice cold PBS. Anterior and posterior midgut tissues were carefully separated from hindgut and malphigian tubules, rinsed in clean PBS, and immediately frozen in liquid nitrogen. All remaining tissues, designated ‘carcass’, were similarly frozen.
RT–PCR procedures and sequences of defensin and S7‐specific primers were as described (Richman et al., 1996). Sequences of GNBP‐specific primers were; 5′ GCAACGAGAATCTGTACC 3′ and 5′ TAACCACCAGCAACGAGG 3′ (Dimopoulos et al., 1997). Total RNA was extracted from whole or dissected mosquitoes at the indicated time points, using the RNaid PLUS kit (Bio 101). For RNA blot analysis, the electrophoresis, transfer to Hybond‐N+ membrane (Amersham), probe preparation, and hybridization were performed as described (Sambrook et al., 1989).
We thank R.Sinden, G.Butcher and J.Thompson for materials and helpful suggestions, Claudia Blass for expert technical assistance, colleagues in the Kafatos laboratory for discussions and comments, and B.Seraphin for comments. Supported by the John D. and Catherine T. MacArthur Foundation, the Human Frontier Science Program, the UNDP/World Bank/WHO Special Programme for Research and Training in Tropical Diseases (TDR), the European Union Training and Mobility of Researchers (TMR) Network on Insect–Parasite Interactions, and an EU TMR postdoctoral fellowship to G.D.
- Copyright © 1997 European Molecular Biology Organization