8 G. Dimopoulos
 2003 Blackwell Publishing Ltd, Cellular Microbiology, 5, 3­14
appear to either have been lost or forced to diversify.
Interestingly, the largest number of orthologues and small-
est diversification between the two species is found in
gene families that are involved in the highly conserved
signal transduction pathways of the innate immune sys-
tem and developmental processes. Mosquito orthologues
have for instance been identified for all Toll pathway com-
ponents. In contrast, the largest family expansions and
diversifications have occurred within gene families impli-
cated in pathogen recognition, serine protease cascades
and effector systems. Diversification of the protein reper-
toire for some gene families is also increased through
alternative splicing of transcripts, as in the case of the
PGRP genes (discussed above). The TEP families (dis-
cussed above) are represented by six and 15 members
in D. melanogaster and A. gambiae, respectively. Only
one 1:1 orthologous pair has been identified and most of
the other members have resulted from species-specific
expansions which may reflect adaptations to different
microbial exposure (Christophides et al., 2002). One of
the most remarkable examples of a massive mosquito-
specific gene family expansion is represented by the
FREP family (described in earlier sections). Whereas only
13 FREP homologues have been identified in the D. mel-
anogaster genome, the A. gambiae has as many as 58
different members (Zdobnov et al., 2002). The reason for
this massive expansion of the A. gambiae family is prob-
ably linked to haematophagy and exposure to Plasmo-
dium.The bacteria and blood cell binding nature of FREPs
may be important in controlling the midgut bacteria flora,
which is highly boosted upon blood feeding, or prevent
blood coagulation through competitive inhibition (Demaio
et al., 1996; Gokudan et al., 1999). The malaria infection
responsive nature of some mosquito members suggest
them being a part of the antimalarial defence system
(Christophides et al., 2002; Dimopoulos et al., 2000;
2002; unpublished data). The A. gambiae antimicrobial
peptide, Gambicin, has also been identified in the Aedes
aegypti mosquito but does not exist in the D. melano-
gaster genome (Dimopoulos et al., 2000; Vizioli et al.,
2001; Christophides et al., 2002). Gambicin may have
evolved specifically to cope with the mosquitos' microbial
flora or the malaria parasite.
The mosquitos' innate immune responses and
Plasmodium killing
The lifecycle of Plasmodium in the mosquito is complex
and involve several developmental transformations and
spatial translocations through and between epithelial
tissues (Fig. 1). The parasite suffers large losses during
its sporogonic development. A small proportion of
ingested gametocytes will develop into ookinetes and of
these only a fraction will reach the oocysts stage. At the
later stages of infection, more than 80% of the haemocoel
sporozoites will fail to relocate into the salivary glands and
are instead rapidly cleared from the haemolymph through
unknown mechanisms. The magnitude of these losses
can differ greatly between infections with different parasite
and mosquito species combinations and their molecular
basis is unknown (Beier, 1998; Ghosh et al., 2000;
Dimopolous et al., 2002b). Parasite elimination in mosqui-
toes have been linked to the innate immune system and
may be crucial for the successful transmission of malaria
(discussed below; Luckhart et al., 1998; Lowenberger
et al., 1999). High infection levels may seriously affect the
mosquitos' fitness whereas a too low infection level may
prevent transmission of the parasite. The lowest possible
infection level that can permit transmission appears to be
most favourable for both the parasite and the vector. In
fact, several studies have indicated increased mosquito
mortality upon high malaria infection levels that are more
likely to occur in infections between unnatural mosquito­
parasite species as opposed to low infection levels that
are characteristic for natural mosquito­parasite combina-
tions (Ferguson and Read, 2002). A finely tuned equilib-
rium between the mosquitos' immune system and the
parasites immune evasive capability is likely to contribute
towards the establishment of a low but transmissive infec-
tion level. Incompatibility of the biochemical environment
in the mosquito, the receptor­ligand interactions involved
in epithelial invasion and the temporal kinetics of parasite
development and mosquito processes, such as blood
digestion, are also likely to constitute important determi-
nants of mosquito permissiveness to malaria infection and
thus important regulators of transmission.
Mosquito refractoriness to Plasmodium infection
Several Anopheles­Plasmodium combinations are incom-
patible because of highly efficient parasite killing by mos-
quito refractory mechanisms. The best characterized
refractory mosquitoes belong to the genetically selected
L3-5 A gambiae strain which melanotically encapsulates
late ookinetes and early oocysts as they reach the basal
side of the midgut epithelium between 20 and 40 h after
infection (Fig. 1). This humoral encapsulation mechanism
has a certain degree of specificity and key components of
the melanization reaction appear to originate from the
haemolymph (Collins et al., 1986; Gorman et al., 1998;
Paskewitz et al., 1998; Dimopoulos et al., 2001).
Melanized Plasmodia are rarely seen in field collected
A. gambiae mosquitoes (Schwartz and Koella, 2002).
Three quantitative trait loci (QTL) are controlling the
refractory trait (Zheng et al., 1997) and a 528 kb section
of the major QTL, Pen1, chromosomal region has been
sequenced and revealed 48 genes that are currently being
analysed for potential implication in the refractory mech-
Mosquito immunity to malaria 9
 2003 Blackwell Publishing Ltd, Cellular Microbiology, 5, 3­14
anism (Thomasova et al., 2002). Factors such as pattern
recognition receptor molecules or components of serine
protease cascades represent promising candidates for the
regulation of melanotic encapsulation. Microarray gene
expression studies on the refractory L 3­5 mosquitoes
have indicated major differences from the susceptible
mosquitoes in expression signatures involving immunity
and oxidoreductive genes. The potential implication of the
innate immune system and/or the mosquitos' oxidative
status in the determinants of the refractory phenotype is
currently investigated (C. Barillas-Mury, R. Cantera, G.
Christophides, G. Dimopoulos, F. C. Kafatos, unpublished
data).
Other refractory mosquito species and strains that kill
Plasmodia with different mechanisms exist. A genetically
selected strain of the oriental vector Anopheles dirus is
almost totally refractory to Plasmodium yoelii nigeriensis
infection whereas it will permit normal development of P.
falciparum and Plasmodium vivax. In this strain, Plasmo-
dium yoelii oocyst development is arrested within 12 h and
then followed by melanization of dead early oocysts. The
melanization reaction does not seem mediate the killing
itself but occurs as a subsequent event (Somboon et al.,
1999). In a genetically selected A. gambiae strain, all
Plasmodium gallinaceum ookinetes die in the midgut epi-
thelium through a lytic mechanism. The dying ookinetes
are first vacuolated and will then appear as degraded and
broken within the midgut cell cytoplasm. Genetic crossing
experiments suggest that this lytic killing mechanism is
controlled by a single dominant locus (Vernick et al.,
1995). The Mexican malaria vector Anopheles albimanus
is highly refractory to the P. vivax CS VK247 variant. The
development of this parasite is compromised at three dif-
ferent stages in the midgut: the first portion of parasites
is believed to be destroyed by mosquito digestive
enzymes in the ectoperitrophic space close to the internal
midgut surface; a second portion disintegrates within the
midgut epithelium and; a third portion is arrested during
early oocyst development on the basal side of the midgut
epithelium. (Gonzalez-Ceron et al., 2001). Although the
implication of the mosquitos' immune system in these
refractory mechanisms is strongly suggested, their molec-
ular basis remains unknown. The same mechanisms that
are responsible for total refractoriness are also most likely
operating at a lower level in susceptible mosquitoes where
they reduce parasites at the crucial transition stages
within and between epithelia. For instance, melanized
Plasmodia have been documented in natural mosquito
populations (Schwartz and Koella, 2002). Other killing
mechanisms, such as lysis, are more difficult to document
because of the lack of a visible phenotype. Hence, genet-
ically selected refractory strains are important for the
study of mechanisms mediating Plasmodium killing in nat-
ural susceptible mosquitoes.
Anopheles gambiae immune responses
Upon malaria infection, the mosquito is mounting robust
local and systemic immune responses that correlate
temporally and spatially with the parasite's development.
During midgut invasion by the ookinete, several immune
genes have been shown to be upregulated in both the
midgut epithelium and in the fat body. The activation of
immune responses in the fat body at this stage, when
the parasites are still located within the midgut epithe-
lium, strongly suggest the existence of immune signal-
ling cascades between the different tissues. At the later
stages of infection, when sporozoites translocate from
the oocysts to the salivary glands, immune responses
have been documented in the salivary gland and in the
fat body. Genes implicated in these immune responses
encode putative pattern recognition receptors, serine
proteases, signalling pathway components, antimicrobial
peptides and nitricoxide synthase (Dimopoulos et al.,
1997; 1998; 2002; Richman et al., 1997; Luckhart et al.,
1998; Oduol et al., 2000). A significant degree of over-
lap between responses to bacteria and malaria chal-
lenge has been demonstrated by gene expression
analysis of 2300 genes on a microarray. The spectrum
of genes that are regulated by malaria infection is signif-
icantly smaller than that of the bacteria infection induced
genes and do not overlap with sterile injury induced
genes. In this assay, the malaria infection responsive
immune genes included a GNBP, a PGRP, a FREP, a
TEP, a serine protease, a phagocytic component and a
leucin rich repeat protein gene that share homology with
Toll receptors (Dimopoulos et al., 2002). The docu-
mented immune responses to malaria infection may
partly result from the injury that is caused by parasite
invasion of epithelial tissues as well as microbial compo-
nents of the midgut that may be present at the invasion
site. However, the lack of overlap between sterile injury
responsive and malaria infection responsive gene
expression signatures, and the strong activation of
immune genes in antibiotic-treated malaria-infected
mosquitoes strongly suggest the existence of a Plasmo-
dium recognition-specific mechanism of immune induc-
tion (Richman et al., 1997; Dimopoulos et al., 2002). A
comprehensive gene expression study using a whole
genome microarray, representing the entire A. gambiae
transcriptome, will provide a much more detailed view
on the regulation of malaria infection responses and
the implicated components. The significance of the
documented immune responses in the elimination of
parasites is strongly supported by the lower prevalence
of malaria infection in mosquitoes that have been
preimmune challenged with bacteria, the implication
of the immune responsive nitricoxide synthase in
Plasmodium killing and the correlation of immune
10 G. Dimopoulos
 2003 Blackwell Publishing Ltd, Cellular Microbiology, 5, 3­14
responses and parasite losses (Dimopoulos et al., 1997;
1998, 2000; 2002; Richman et al., 1997; Luckhart et al.,
1998; Lowenberger et al., 1999). The immune respon-
sive mosquito-specific antimicrobial peptide Gambicin
has been shown to possess lethal activity against
ookinete stage Plasmodia (Vizioli et al., 2001). Anti-
plasmodial activity against the oocyst and sporozoite
stages of Plasmodium has also been shown in vitro for
defensins from other insects (Shahabuddin et al., 1998).
Other insect­parasite models
Whereas most of our knowledge on the insect innate
immune system and its interactions with parasites has
derived from studies in D. melanogaster and A. gambiae,
respectively, studies of other insect­pathogen models are
revealing novel and complementary aspects on insects'
immune system and resistance to human parasites.
For example, ookinete stage Plasmodia that have been
injected into the D. melanogaster hemocoel will develop
oocysts and produce infectious sporozoites that are rap-
idly cleared from the haemolymph. Parasite development
was not compromised in mutants with constitutively
active Toll receptors, nor was transcription of antimicro-
bial peptide genes induced in infected flies. Killing of
sporozoites in the haemolymph appeared to involve
haemocytes and other unknown components (Schneider
and Shahabuddin, 2000). Drosophila immune responses
and killing mechanisms of Plasmodia are likely to differ
significantly from those taking place in the co-adapted
mosquito vector. However, the fly with its powerful genetic
and transgenic tools may provide a useful model to iden-
tify, study and engineer anti-Plasmodial proteins.
Studies in tsetse (Glossina spp.) vectors of African
sleeping sickness and nagana have shown upregulation
of several antimicrobial peptides upon infection with Try-
panosoma. Interestingly, the tsetses' immune surveil-
lance system appears to be capable of discriminating
between parasites and bacteria, and between different
parasite life stages. The implication of tsetse immune
responses in parasite killing is suggested by the signifi-
cantly lower Trypanosoma infection levels in flies that
have been preimmune challenged with bacteria (Hao
et al., 2001; Boulanger et al., 2002). A recently imple-
mented gene discovery project have identified over 60
tsetse fly genes with potential implication in its immune
system and will permit a more detailed dissection of
tsetse responses to Trypanosoma infection (M. J.
Lehane, personal communication).
Conclusions and discussion
The past decade has experienced a revolution in our
knowledge on the insect innate immune system and
novel aspects are continuously added. For instance, the
apparent connections between immune cascades and
the apoptotic machinery may prove to play a key role in
the mosquitos' interactions with the malaria parasite
(Christophides et al., 2002; Hoffmann and Reichhart,
2002). Plasmodium infection has been suggested to
induce apoptosis of both invaded mosquito midgut cells
and follicular epithelial cells (Han et al., 2000; Hopwood
et al., 2001).
The available A. gambiae genome sequence in combi-
nation with high throughput gene expression analysis
and transgenic technologies will allow a more compre-
hensive dissection of the mosquitos' responses to infec-
tion. These studies can be expanded and include
different refractory mosquito strains, that are easily
selected in the laboratory (H. Hurd, personal communica-
tion), and mosquitoes from the field that may have
diverged significantly from the currently studied lab
strains.
The accumulated knowledge on vector­parasite inter-
actions will ultimately allow the development of disease
control strategies based on transgenic refractory insects
and other transmission blocking approaches where host
antibodies can block the parasites' lifecycle in the mos-
quito (Collins, 1994; Stowers and Carter, 2001; Ito et al.,
2002). The mosquitos' innate immune system could be
utilized in various ways for the development of malaria
control strategies based on genetically engineered mos-
quitoes or the spread of resistance genes in mosquito
populations with mobile elements. In one scenario the
mosquito could have a boosted anti-Plasmodial immune-
surveillance system that would recognize the parasite as
non-self more efficiently and/or stronger activate Plasmo-
dium killing mechanisms. In another scenario, the engi-
neered mosquito could express a blood meal inducible
immune protein that would kill ookinetes in the midgut
epithelium. The implementation of such control strategies
will, in addition to a comprehensive dissection of the
mosquitos' immune system, also require consideration of
the transgenic mosquitos' fitness, the possible selection
of resistant parasites to the killing mechanism and a
detailed knowledge of the field mosquito populations that
frequently are composed of several sympatric, but repro-
ductively isolated, sibling species (Coetzee et al., 2000;
Alphey et al., 2002; Enserink, 2002).
Acknowledgements
I acknowledge contributions of former and current members of
Professor F. C. Kafatos' laboratory and fruitful interactions with
the laboratories of Professors J. A. Hoffmann, R. E. Sinden, M.
J. Lehane, H. Hurd, F. Collin, A. Ezekowitz and C. Janeway. I
thank Dr Nadiem Mohammed for critical reading of the manu-
script. Most of the presented work on A. gambiae immunity has
been done in Professor F. C. Kafatos laboratory and supported