All metazoans have evolved means to protect themselves from threats present in the environment: injuries, viruses, fungi, bacteria and other parasites. Insect protection includes innate physical barriers and both cellular and humoral responses. The insect innate immune response, best characterized in Drosophila melanogaster, is a rapid broad response, triggered by pathogen-associated molecular patterns (PAMPs) recognition, which produces a limited range of effectors that does not alter upon continued pathogen exposure and lacks immunological memory. The Drosophila response, particularly its humoral response, has been investigated by both low and high-throughput methods. Three signalling pathways conserved between insects and mammals have been implicated in this response: Toll (equivalent to mammalian TLR), Imd (equivalent to TNFα) and Hop (equivalent to JAK/STAT). This review provides an entry point to the insect immune system literature outlining the main themes in D. melanogaster bacterial pathogen detection and humoral and cellular immune responses. The Drosophila immune response is compared with other insects and the mammalian immune system.
All metazoans have evolved means to protect themselves from threats present in the environment: injuries, virus, fungi, bacteria and parasites. Devoid of an adaptive immune system, insects rely on innate defences comprising physical barriers (chitinous peritrophic membranes lining the gut, chitin exoskeletons and barrier epithelia) plus cell-based (cellular) and cell-mediated extracellular (humoral) response strategies. An example of the former is the encapsulation of parasitic wasp eggs deposited in the haemocoel of Drosophila larvae (reviewed in Meister and Lagueux, 2003), while the latter is exemplified by the synthesis and secretion into the haemolymph mostly by the fly fat body (equivalent in function to the mammalian liver) of a limited repertoire of small antimicrobial peptides (AMPs) against body cavity-injected Escherichia coli and Micrococcus luteus (Lemaitre et al., 1997). These insect innate immune response mechanisms are best characterized in Drosophila melanogaster (see below).
Antibody-mediated adaptive immunity generates a complex repertoire of immune receptors by somatic gene rearrangement after antigen exposure, followed by clonal expansion of cells expressing antigen-specific receptors. It is only present among jawed vertebrates (Gnathostomes) where it has evolved to supplement the more ancient innate immunity.
Two protein families in insects, that recognize unique peptidoglycan components of the two main types of bacterial cell walls (Gram-positive and Gram-negative), are central to the detection and recognition of bacterial pathogens via the Toll- and Imd- pathways. These are the PGRP (peptidoglycan recognition proteins) and GNBP (Gram-negative-binding proteins).
Peptidoglycan recognition proteins (reviewed in Steiner, 2004) were first identified and characterized in Bombyx mori (Yoshida et al., 1996) as proteins involved in the triggering of the phenoloxidase-mediated melanization cascade upon binding to M. luteus peptidoglycan. Genome analysis identified 13 PGRP homologues (plus a number of potential alternative splicing isoforms) and proposed a division based on transcript length and 5′ untranslated regions: S (short) and L (long) (Werner et al., 2000). All Drosophila PGRP-S have short untranslated 5′ regions, code for short peptides (∼20 kDa) and have an N-terminus secretion signal peptide, suggesting they are extracellular proteins. On the other hand, PGRP-L have long untranslated 5′ regions and code for longer peptides (30–90 kDa), some of which have predicted transmembrane domains, suggesting they are intracellular or membrane bound.
Peptidoglycan recognition protein domains can be identified both from primary sequence and tertiary structure (Werner et al., 2000; Liepinsh et al., 2003) and are homologous to T7 bacteriophage lysozyme, a zinc-dependent N-acetylmuramyl-L-alanine amidase known to hydrolyse bacterial peptidogylycan. Some members of the PGRP family have been shown to have amidase activity, e.g. PGRP-SC1 (Mellroth et al., 2003). It has been proposed that these PGRP with enzymatic activity act as scavengers clearing bacterial cell wall debris from the insect that would otherwise elicit an immune response (Mellroth et al., 2003). PGRP without enzymatic activity are still capable of binding peptidoglycan and have been shown to elicit an immune response, e.g. PGRP-SA and PGRP-LC (Michel et al., 2001; Gottar et al., 2002).
Drosophila PGRP-SA (semmelweis mutation), the first fly homologue to be characterized, was identified as a component of the Toll Gram-positive response pathway by screening chemically mutagenized flies for mutations that inactivated genes involved in the control of AMP inducibility (Michel et al., 2001).
Drosophila PGRP-SD, identified as the other member of the PGRP-S subfamily without predicted amidase activity, has also been implicated in the eukaryotic detection of Gram-positive bacteria through the Toll pathway and has been shown to be at least partially redundant to PGRP-SA (Bischoff et al., 2004). PGRP-SD null mutants showed increased susceptibility to Gram-positive bacterial infections. In vivo transcription levels of the AMP Drosomycin gene following Gram-positive bacterial infection (which generally is not significantly affected by single null mutations in PGRP-SA, PGRP-SD and GNBP1) was greatly reduced in the PGRP-SA and PGRP-SD double mutant for a range of bacterial pathogens (Bischoff et al., 2004). Lack of Drosomycin in vivo activation on PGRP-SD overexpression may suggest that PGRP-SD requires an accessory protein, such as GNBP1 identified for PGRP-SA, for downstream signalling activation (Bischoff et al., 2004). The partial redundancy observed between PGRP-SA and -SD may be evidence that evolutionary divergence in the PGRP-S family has increased its recognition potential. On the other hand, it also allows for the possibility that the downstream antimicrobial response may be differently modulated by the two PGRP-S proteins, thus creating the possibility for eukaryotic response specificity to emerge.
PGRP-LC has been shown to have three alternative splicing isoforms (PGRP-LCa, -LCx and -LCy) two of which (PGRP-LCa and -LCx) have been implicated in Gram-negative bacterial infection immune response through Imd (Werner et al., 2003; Kaneko et al., 2004; Kaneko and Silverman, 2005). PGRP-LCx homodimers and PGRP-LCx and -LCa heterodimers recognize different peptidoglycan elicitors, polymeric and monomeric meso-diaminopimelic acid moieties, respectively, increasing the range of pathogen-associated molecular patterns (PAMPs) detected (Kaneko and Silverman, 2005). As is discussed below, alternative splicing of Drosophila Dscam (Down syndrome cell adhesion molecule) (Watson et al., 2005) has been proposed as a means of providing the insect with an increased defensive repertoire that, if regulated, could enable the insect's immune system to display both immunological memory and an adaptive response.
Despite their name, GNBP in Drosophila have so far been linked to Gram-positive (GNBP1) and fungal (GNBP3) infections (Gobert et al., 2003; Leclerc and Reichhart, 2004). GNBP1 was originally shown to bind the insoluble lipopolysaccharide and β-1,3-glucans that are components of the walls of Gram-negative bacteria and fungi respectively (Kim et al., 2000). Consequently, it was proposed that GNBP1 could be a Gram-negative or fungal receptor. However, mutation (osiris) and RNA interference experiments on whole organisms have implicated GNBP1 in the Gram-positive immune response pathway, possibly forming a complex with PGRP-SA (Leclerc and Reichhart, 2004; Pili-Floury et al., 2004).
All the GNBP proteins characterized to date have a secretion signal peptide and an inactive β-glucanase domain prompting the suggestion that they are secreted proteins that, like some PGRP, can bind but not hydrolyse the conserved domain of their targets (Royet et al., 2005).
Gram-negative bacteria and Bacillus ssp. (Bacillus ssp. are Gram-positive bacteria with a meso-diaminopimelic acid peptidoglycan structure) have been shown to activate the Imd pathway through PGRP-LC and PGRP-LE (Choe et al., 2002; Takehana et al., 2002). PGRP-LC has a single predicted transmembrane domain and as such is expected to bind the extracellular bacterial peptidoglycan and interact with intracellular Imd (Kaneko and Silverman, 2005). Immunoprecipitation of tagged PGRP-LC in a Drosophila S2 cell line has confirmed PGRP-LC interaction with Imd (Choe et al., 2005). Although PGRP-LE lacks a transmembrane domain and a secretion signal peptide it has been shown to activate the Imd pathway systemically when overexpressed in the fat body and consequently has been suggested to be an extracellular protein (Takehana et al., 2004). PGRP-LE has been shown to bind peptidoglycans and to activate Imd at least in part through PGRP-LC (Takehana et al., 2004).
Activation of Imd leads to the activation of an intracellular signalling complex involving dFADD (Drosophila Fas-associated death domain protein), an adaptor protein, and DREDD (death-related ced-3/Nedd-2-like protein), a caspase, which have both been implicated in apoptosis (Hu and Yang, 2000).
Immunoprecipitation has shown that dFADD interacts with DREDD and yeast two-hybrid analysis has shown an interaction between Imd and dFADD (Hu and Yang, 2000; Naitza et al., 2002). Genetic analysis suggests Imd is upstream of dFADD which is in turn upstream of DREDD (Naitza et al., 2002), but the most likely hypothesis is that the three proteins form a single signalling complex (Leclerc and Reichhart, 2004).
Relish, a dual domain protein harbouring both the transcription factor NF-κB-like domain and the inhibitory IκB domain, is associated with DREDD in vivo, as shown by co-immunoprecipitation (Stoven et al., 2003). DREDD is thought to be involved in the cleavage of the phosphorylated Ankyrin repeats, part of the IκB domain, of Relish (Stoven et al., 2003).
The Imd/dFADD/DREDD complex is also associated with activating the serine/threonine kinase dTAK1 (Drosophila transforming-growth-factor-β activating kinase) and has been shown to be dependent on Ubiquitin ligases (bendless and dUev1A) and dTAB2 (a protein of unknown function that contains a K63-ubiquitin-binding domain) (Kaneko and Silverman, 2005). dTAK1 is required to activate the Relish (involved in AMP expression), Kay (kayak) and Jra (Jun-related antigen) transcription factors (Silverman et al., 2003). Kay and Jra are involved in the regulation of genes involved in wound repair and stress responses and are part of a signalling cascade homologous to the mammalian JNK (c-Jun N-terminal kinase) MAPK (Mitogen-activated protein kinase) signalling pathway.
The human homologue of dTAK1 (TAK) has been shown to directly phosphorylate the human homologues of Kenny and Ird5 (IKK complex) (Wang et al., 2001). Once activated, the Drosophila IKK complex is believed to phosphorylate Relish thus triggering activated DREDD endoproteolytic caspase activity (Stoven et al., 2003). Once cleaved, Relish translocates to the nucleus and activates antimicrobial gene expression.
Nitric oxide has also been implicated in the Drosophila immune response (Foley and O’Farrell, 2003). Oral Pectobacterium carotovorum subsp. carotovorum infection leads to nitric oxide synthase (NOS) upregulation in the larval gut and haemocytes. Chemical inhibition of NOS increased larval sensitivity to Gram-negative infections and inhibited the induction of the AMP Diptericin. Nitric oxide was sufficient to upregulate a diptericin-GFP reporter construct through Imd (Foley and O’Farrell, 2003).
The Toll-signalling pathway, shown in Fig. 1A, is activated by fungal and Gram-positive infections through serine proteases that have so far not been fully characterized. It is known that the serine proteases involved in dorsal-ventral embryonic patterning (Gastrulation defective, Snake and Easter) are not involved in either pathway (Lemaitre et al., 1996). Persephone has been shown to be a serine protease, active in the fungal response (Ligoxygakis et al., 2002a) and inhibited by the serpin (serine protease inhibitor) SPN43Ac (Necrotic) (Robertson et al., 2003). Serine protease activity leads to cleavage and activation of Spätzle for both Gram-positive and fungal infections. Spätzle binds to Toll, a transmembrane receptor, leading to receptor dimerization and activation of the intracellular signalling domain: TIR (Toll/interleukin-1 receptor) domain (Weber et al., 2003). Recently, more detail of the possible events between recognition of an invading microorganism by PGRP, proteolysis of the Necrotic serpin and Toll activation have been determined (Pelte et al., 2006).
The activated receptor complex recruits the adaptor proteins dMyD88 (Drosophila myeloid-differentiation factor 88) and Tube, and the Pelle serine kinase (Horng and Medzhitov, 2001; Sun et al., 2002; Tauszig-Delamasure et al., 2002). The latter starts a kinase cascade that leads to the phosphorylation of an IκB homologue, Cactus and its subsequent degradation (Nicolas et al., 1998). Dissociation from Cactus, allows NF-κB-like transcription factors Dorsal and DIF (Dorsal-related immune factor) to translocate to the nucleus and activate the antimicrobial response, which includes induction of WntD expression, a Wnt-family (Nusse et al., 1991) factor that inhibits Dorsal translocation (Gordon et al., 2005). Upregulation of transcripts coding for inhibitory proteins creates a negative-feedback loop that helps ensure response termination.
The Hop-signalling pathway, shown in Fig. 1B, was originally characterized by its involvement with embryonic segmentation (Binari and Perrimon, 1994). Its role in the insect immune response was first identified in Anopheles gambiae: the transcription factor Ag-STAT, a member of the STAT (signal transducer and activator of transcription) protein family and homologous to Drosophila STAT92E, was shown to accumulate in the cell nuclei upon immune challenge of the mosquito (Barillas-Mury et al., 1999). A similar response has been reported for D. melanogaster STAT92E, and shown to be dependent on the cytokine receptor Domeless (Dome) and the Janus kinase (JAK) homologue Hopscotch (Hop) (Agaisse et al., 2003).
Hop pathway activation in the fat body (equivalent to mammalian liver and the main site of expression of immune humoral factors) is dependent on the cytokine Upd3 (unpaired3), a Dome ligand, produced in blood cells upon immune challenge (Agaisse et al., 2003) thus providing a potential link between humoral and cellular responses.
The Drosophila Hop pathway has been shown to control expression of TEP1, part of a four-member family of thioester-containing proteins with significant similarities to the complement C3/α2-macroglobulin superfamily (Lagueux et al., 2000). TEP1 is thought to function as an opsonin, promoting phagocytosis by analogy to aTEP, an A. gambiae TEP1 homologue, which has been shown to be induced by immune challenge in vivo and to be involved in promoting phagocytosis in cultured mosquito cells (Levashina et al., 2001).
Turandot (Tot) proteins, produced by the larval fat body have been shown to accumulate in the haemolymph in response to various stress conditions (Ekengren et al., 2001). TotA, in particular, has been shown to be upregulated due to septic injury and regulated by Hop-signalling (Agaisse et al., 2003). Other proteins that are still uncharacterized, e.g. CG11501 (Boutros et al., 2002) have also been shown to be regulated by the Hop pathway. So far, none of these proteins that has been identified has shown direct antibacterial activity (Agaisse and Perrimon, 2004).
Genomic-wise RNAi screens monitoring the effect on the expression of a STAT92E-dependent reporter construct have implicated a further 117 genes in the Hopscotch intracellular signalling pathway, including both positive and negative regulators (Baeg et al., 2005).
The best characterized response of immune challenged flies is the production, mostly by the fly fat body and haemocytes of short extracellular AMP with different activity spectra such as Diptericin (active mostly against Gram-negative bacteria), Drosomycin (anti-fungal) and Defensin (active against Gram-positive bacteria). Eight classes of AMP have been characterized (Hultmark, 2003) but transcriptional analysis of immune challenged flies suggests there may be more (De Gregorio et al., 2001). In addition to the fat body, it has been shown that barrier epithelia in Drosophila can express at least one AMP in an inducible tissue-specific manner (Tzou et al., 2000). It has been proposed that this surface epithelial immunity system is the ancestral immune defence mechanism that has been elaborated during evolution into concentrations of more specialized cell types such as in the fat bodies of higher insects and the white blood cells of mammals (Tzou et al., 2000). In accord with this proposal is the recent discovery that the Drosophila kidney equivalent (Malpighian tubule) functions as an autonomous immune response system that detects and responds to bacterial infection independently of the fat body (McGettigan et al., 2005; Dow and Davies, 2006). Echoing the earlier observations with fat bodies and haemocytes is the finding that the Malpighian tubule also utilizes the nitric oxide-signalling pathway (McGettigan et al., 2005).
Although the Drosophila antimicrobial response is modular (Boutros et al., 2002), local differences in pathways, pathway crosstalk and overlapping pathways have been reported. For example, drosomycin transcription (under the control of Toll in the systemic response) has been shown to be dependent on Imd in the tracheae and Malpighian tubules (Tzou et al., 2000); totA transcription (under the control of Hop) is dependent on Relish activity (Agaisse et al., 2003), and cecropin, attacin and defensin (all coding for AMPs) are regulated by both Toll and Imd pathways (Hergannan and Rechhart, 1997).
However, AMPs are only a small fraction of the fly's response to invading pathogens. Transcriptional profiling has not only been used to identify immune inducible systemic changes of transcription by comparing naïve and immune challenged wild-type flies (De Gregorio et al., 2001; Irving et al., 2001), but also to identify pathway-specific responses by comparing the immune responses of different signalling pathway mutants (Boutros et al., 2002; De Gregorio et al., 2002a).
Over 543 genes were identified whose transcription was altered due to immune challenge by septic injury (De Gregorio et al., 2001; Irving et al., 2001). These included all characterized AMPs, 25 unknown peptides (between 43 and 134 amino acids long) and proteins involved in pathogen recognition and phagocytosis (e.g. PGRP, GNBP and TEP). Trypsin-like proteases and serpins have also been identified as well as genes probably involved in melanization and coagulation. Other components of the immune-signalling pathways that were detected included dorsal, cactus, relish, genes involved in iron metabolism, in reactive oxygen species production and detoxification as well as genes that showed no homology to genes in databases (De Gregorio et al., 2001). Some of the reported transcriptional changes have been confirmed by haemolymph proteome analysis of cultured mbn-2 cells, larvae and adult flies (Levy et al., 2004; Vierstraete et al., 2004; de Morais Guedes et al., 2005).
Comparing signalling pathway mutants confirmed that Imd and Toll pathways are involved in most of the observed antimicrobial responses (De Gregorio et al., 2002a).
Clean injury (sterile needle penetration) has also been reported to induce a largely overlapping set of genes, but to a lower extent than septic injury (Boutros et al., 2002). Importantly, transcriptional analysis showed that clean injury is sufficient to activate the cellular immune response in flies (Markus et al., 2005). Fungal induction of AMPs was dependent on the route of infection, with a natural infection eliciting a strong persistent anti-fungal response without activation of anti-bacterial genes (Lemaitre et al., 1997). Similarly, D. melanogaster response to natural bacterial pathogens after oral infection differs from the previously characterized septic injury response. Such differences reflect the route of infection and pathogen-specific responses (Vodovar et al., 2005).
When Drosophila cell lines (S2 and mbn-2) were challenged with crude Gram-negative lipopolysaccharides (containing meso-diaminopimelic-acid-type peptidoglycans), largely similar transcriptional changes were observed which led to a humoral-like response upregulating expression of AMPs (Lindmark et al., 2001; Boutros et al., 2002). Challenging mbn-2, a tumorous Drosophila blood cell line with live Gram-negative bacteria at low multiplicity of infection triggers a response that mimics lipopolysaccharide challenge (Lindmark et al., 2001; Johansson et al., 2005). At high multiplicities, the response was compatible with a cellular response, i.e. the transcription of prophenoloxidase, unpaired cytokine genes and genes involved in cell differentiation were upregulated (Johansson et al., 2005).
Innate immune system activation has been shown to be age dependent in Drosophila, i.e. it undergoes a senescence during which basal transcription levels of AMP genes increase even in the absence of immune stimulation (Seroude et al., 2002). This transcription activation weakens and lengthens upon immune challenge (Zerofsky et al., 2005). As a result, older flies are more susceptible to infections due to the delayed response and are less fertile due to the prolonged expression of AMPs.
Drosophila can be infected by several viruses and the advantages of its genetics base have prompted both basic studies on viral infection mechanisms per se (Cherry and Perrimon, 2004; Cherry et al., 2005) and applied studies on the specific responses of the Drosophila innate immune system.
Infection of adults with Drosophila X virus, a member of the dsRNA birnaviridae, followed by anoxia challenge has been shown to lead to sustained expression of AMPs to levels comparable to bacterial challenges (Zambon et al., 2005). Expression of AMPs did not influence viral titres and comparison between signalling pathway mutants, implicated only the Toll pathway in the response, which has been proposed to be mostly cellular (Zambon et al., 2005). A different response was obtained in a different infection system, where Drosophila adults were solely challenged with Drosophila C virus (DCV), a member of the +ssRNA dicistroviridae (Dostert et al., 2005). Their results indicated no induction of AMPs and no involvement of the Toll and Imd pathways, but provided evidence that virus induced transcription products were generated in part through the action of the JAK/STAT (Hopscotch) pathway. However, not all DCV-induced genes are JAK/STAT dependent, suggesting that the full antiviral response involves the participation/crosstalk with a second pathway (Dostert et al., 2005).
Cellular and local responses
The cellular response in Drosophila includes three processes attributed to different lineages of haemocytes (reviewed in Lavine and Strand, 2002; Meister, 2004; Irving et al., 2005): phagocytosis (plasmatocytes), encapsulation (lamellocytes) and melanization (crystal cells).
However, the three lineages are only present during larval stages where the dedicated phagocytic cells or plasmatocytes comprise the large majority of haemocytes and crystal cells represent a small population (5%); lamellocytes are only present upon parasitization (Lanot et al., 2001). The haemopoietic organs, the lymph glands, are lost during metamorphosis and only plasmatocytes (both of embryonic and larval origins) are present in the adult population (Holz et al., 2003).
Phagocytosis relies on recognition of non-self or other specific signals, e.g. phosphatidylserine-containing phospholipids on the external surface of an apoptotic cell's plasma membrane (reviewed in Fadok et al., 1998). In Drosophila, the transmembrane scavenging receptor class C type I (dSR-CI) (Pearson et al., 1995; Ramet et al., 2001), PGRP-LC (Ramet et al., 2002) and the transmembrane receptor containing EGF-repeats Eater (Kocks et al., 2005) have been implicated in promoting phagocytosis of invading organisms.
Drosophila dSR-CI, expressed only in haemocytes (Pearson et al., 1995), has been shown to increase bacterial binding to transfected CHO (Chinese hamster ovary) cells and to decrease bacterial binding to RNAi-treated Drosophila S2 cells in culture; RNAi treatment reduced phagocytosis of heat-killed bacteria by as much as 20% (Ramet et al., 2001).
RNAi depletion of PGRP-LC in S2 cells, a cell line established from Drosophila OregonR embryos (Schneider, 1972), suggests PGRP-LC plays a role in phagocytosis, however, it is not clear whether it functions as a receptor recognizing Gram-negative bacteria or whether Imd-signalling is involved (Ramet et al., 2002).
The same RNAi screen using S2 cells also implicated the GATA transcription factor Serpent, involved in Drosophila haematopoiesis (reviewed in Evans et al., 2003), in phagocytosis (Ramet et al., 2002). Comparison of transcription profiles between Serpent RNAi-depleted and control S2 cells identified Eater (CG6124), a predicted cell-surface receptor harbouring 32 EGF-like repeats, as a candidate gene involved in bacterial phagocytosis (Kocks et al., 2005). RNAi depletion of Eater significantly decreased bacterial (Staphylococcus aureus and E. coli) binding to and uptake by S2 cells (Kocks et al., 2005). A truncated form of Eater, encoding only the four first EGF-like domains (which are the most sequence variable repeats and are also glycosylated), expressed in and purified from S2 cells (expression construct contained an affinity tag) has been shown to bind both bacterial (S. aureus and E. coli ) and fungal cells (Candida silvativa) in vitro (Kocks et al., 2005).
TEP proteins, which have been implicated in opsonization in A. gambiae (Levashina et al., 2001) may also be involved in phagocytosis in Drosophila: TEP genes are induced in Drosophila by septic injury (De Gregorio et al., 2001) and two family members are mainly expressed in haemocytes (Meister, 2004). In mbn-2 cells, none of the tep genes were reported to be induced by lipopolysaccharide but tep4 was upregulated by E. coli infection (Johansson et al., 2005). Haemolymph proteome analysis has shown that TEP2 expression levels increase in larvae fed bacterial membrane preparations (de Morais Guedes et al., 2005). TEP4 expression is upregulated during fungal infection of adult flies (Levy et al., 2004).
Recognition of non-self targets too large for phagocytic clearance is believed to trigger the encapsulation response in Drosophila larvae; a particularly important response against parasitization in the wild (reviewed in Meister, 2004 and Lanot et al., 2001). Parasitic wasp eggs deposited in the larval haemocoel of resistant insects are quickly recognized by plasmatocytes in the haemocoel that have been shown to attach to the wasp's egg chorion (Russo et al., 1996). Activation of the encapsulation response then triggers prohaemocyte differentiation in the lymph glands to lamellocytes that are subsequently released into the larval haemocoel. Lamellocytes attach to the parasitic wasp egg and to other lamellocytes (the latter through septate junctions) forming a multilayered capsule surrounding the egg, which later blackens probably due to melanization. Transcription profiling of haemocytes isolated from naïve mutant larvae with altered haemocyte population composition has implicated the Hopscotch pathway as well as integrins (particularly βPS and αPS4) in lamellocyte function, and have identified DoxA3 as a potentially lamellocyte-specific prophenoloxidase (Irving et al., 2005). The encapsulation response against Leptopilina boulardi wasp eggs was significantly affected in thermosensitive myospheroid (gene coding for βPS) mutants under restrictive conditions (Irving et al., 2005). Comparisons of transcription profiling of Drosophila larvae during the course of a parasitoid infection (the wasp Asobara tabida) with control larvae have identified over 100 genes involved in the Drosophila parasitoid response, including members of the Toll and Hopscotch pathways (Wertheim et al., 2005).
As mentioned previously, crystal cells are expected to play a role in melanization and in encapsulation (which also involves melanization) during the larval stages. Crystal cells have been shown to express two of the three genes encoding phenoloxidases and store the precursors in intracellular structures that give the lineage its name (Meister, 2004). Phenoloxidase catalyses oxidation of phenolic compounds to quinones that can polymerize to form melanin. Melanin and intermediates of the melanization cascade are toxic to microorganisms (reviewed in Nappi and Vass, 1993). In the adult fly, melanization may be a humoral response with all key players already present in the haemocoel.
As noted earlier, local responses at barrier epithelia have been proposed to be the true ancestors of the metazoan immune system (Tzou et al., 2000). Tissue-specific expression has been shown to be dependent on Imd- rather than on Toll-signalling as Toll loss- and gain-of-function mutations did not alter the tissue-specific expression patterns, while expression levels of drosomycin (controlled by Toll in the fat body during systemic response) were not upregulated upon infection in imd– null mutant flies (Tzou et al., 2000). Local responses are also believed to be dependent on as yet uncharacterized tissue-specific mechanisms because bacteria injected directly into the insect body cavity induce antimicrobial response in the fat body but not in epithelia (Tzou et al., 2000).
Immune systems in other insects
Comparison of the immune-related genes from D. melanogaster and A. gambiae suggests that signal transduction is the most conserved aspect of these two systems, with abundant 1:1 orthologues. While other aspects of the immune response (recognition, modulation and effectors) have been less conserved over the estimated 250 million years since the two species diverged, it is apparent that the defence mechanisms have been conserved (Christophides et al., 2002; Zdobnov et al., 2002).
In fact, the key defence mechanisms (haemolymph coagulation, phenoloxidase-mediated melanization, lectin, complement, agglutinin response, reactive oxygen and nitrogen species production, phagocytosis, Toll- and Imd-linked AMP production) have remained conserved and are present to different extents in all invertebrates so far investigated (reviewed in Iwanaga and Lee, 2005).
Regardless of how efficient an innate immune system can become, it is evolutionary unstable as pathogens would be expected to adapt and eventually acquire resistance mechanisms particularly in longer living invertebrates.
Innate immunity per se cannot account for a number of longer lasting immunological responses in which insects are observed to become more responsive when subjected to a second homologous challenge (Schmid-Hempel, 2005). An example of this is the line-specific immunological response of the copepod Macrocyclops albidus against the tapeworm Schistocephalus solidus (Kurtz and Franz, 2003). Success and intensity of reinfection of the copepods by the natural parasite S. solidus were lower if the tapeworms used on reinfection were related to the first group. Different tapeworm sibling groups did not lead to a significant reduction of reinfection success or intensity (Kurtz and Franz, 2003). Cockroaches (Periplaneta americana) first challenged with either bee or snake toxin and then challenged again 7 days later survived longer when challenged with the homologous toxin than with the heterologous toxin. This protective effect could further be extended to a period of 7 weeks (Rheins and Karp, 1985). Cockroaches (P. americana) have also been reported to reject allografts faster than they reject isografts from the same donor if previously exposed to a tissue graft (Hartman and Karp, 1989).
While somatic recombination which could provide an alternative defence mechanism (Loker et al., 2004), is as yet undetected in insects, in plants, exposure to fungi or viruses increased the rate of homologous recombination (Lucht et al., 2002; Kovalchuk et al., 2003). If alternative splicing, such as reported for Drosophila Dscam (Watson et al., 2005), could be regulated within an immune response context it could provide the insect with an increased defensive repertoire and enable the insect immune system to display both immunological memory and an adaptive response. Changes in splicing patterns due to continued exposure to PAMPs would lead to an immune response that is adaptable. If the splicing changes are regulated and effective splice variants could be selected for, the response would be a true adaptive response. Moreover, if the effective splice variant could be maintained in the absence of the PAMPs that triggered its selection, the immune system would display immunological memory.
Commensal microbiota may also play a role in insect immunity as has been shown for Pantoea agglomerans (main component of the locust microbiota) and Metarhizium anisopliae (pathogenic fungus) in Schistocera gregaria (desert locust). In this case, germination of fungal spores is inhibited by phenolic compounds produced by the bacteria (Dillon and Charnley, 2002).
Insect versus mammalian immune systems
Apart from phenoloxidase-mediated melanization and encapsulation all other major defence mechanisms are present in the mammalian innate immune system. Evolution of closed circulatory systems may have limited the use of melanization and encapsulation as immune mechanisms, due to potentially lethal thrombosis, being retained only in invertebrates with open circulatory systems (Theopold et al., 2002). The most significant differences are in pathogen detection and in the subsequent signalling.
As detailed above, pathogen recognition in Drosophila is carried out by extracellular (e.g. PGRP-SA) or membrane-associated receptors (e.g. PGRP-LC) that directly bind to microbial cell wall components (bacterial peptidoglycans and fungal cell wall glucans) and trigger signalling either through a serine protease cascade or directly. Two signalling pathways are responsible for most of the antimicrobial response observed (Toll and Imd) while a third (Hop) has been implicated in modulating and in linking humoral to cellular responses.
In mammals, all three pathways have homologues that have been implicated in the immune system: Toll and Toll-like receptors (TLRs)/IL-1R (reviewed in Janeway and Medzhitov, 2002; Takeda, 2005), Imd and Tumour necrosis factor α (TNFα) (reviewed in Aggarwal, 2003) and Hop and JAK/STAT (reviewed in Mitchell and John, 2005). However, the interplay between these pathways is very different in mammals from that in insects.
Peptidoglycan recognition proteins are present in humans and have been shown to bind bacterial peptidoglycan (Liu et al., 2001) but their contribution to pathogen recognition has not yet been demonstrated (Liu et al., 2001; Xu et al., 2004).
To date, only TLRs have been implicated in pathogen detection and recognition in mammals. So far, 11 TLR genes have been discovered of which eight have been implicated in the immune response. All TLRs have an extracellular leucine-rich domain that binds PAMPs (e.g. peptidoglycan, flagellin, unmethylated CpG DNA or dsRNA) directly or through adaptor proteins as in human TLR4. The latter requires a 136-residue secreted protein, MD-2, and a GPI-anchored cell-surface protein, CD14, for full activation upon bacterial lipopolysaccharide binding (Shimazu et al., 1999; Janeway and Medzhitov, 2002).
In mammals activation of TLR leads to receptor dimerization and downstream signalling generally culminating in NF-κB activation through the MyD88 and IKK complex. Alternative pathways are also present, e.g. MyD88-independent and IKK-independent TLR3 activation of the interferon-β gene (Yamamoto et al., 2002; Meylan et al., 2004).
Activation of the TLR pathway leads to expression of antimicrobial gene, e.g. β-defensin (Vora et al., 2004), as well as signalling molecules (cytokines) that lead to activation of other pathways and modulate the adaptive response, e.g. TNF-α and IFN-γ which activate the TNFR and JAK/STAT pathways respectively.
Despite the major recent advances in our understanding of the innate immune system in insects many questions remain. For example, are there other pattern recognition receptors (PRRs) or strategies which function to increase the diversity of the system? For example, how important are galectins in insect immunity as PRRs, signal modulators or effectors? These phylogenetically ancient β-galactoside binding lectins are known to be involved in mammalian innate immunity and to date examples have been found in Diptera (Rabinovich et al., 2002; Pace and Baum, 2004; Rabinovich and Gruppi, 2005). Galectins can contain more than one carbohydrate recognition domain (CRD) sometimes fused to short amino repeat sequences and can therefore be bivalent or multivalent when binding to complex carbohydrates, similar to the cross-linking of antibodies and multivalent antigens. Differing carbohydrate specificity in a galectin containing two CRDs can enable it to bind two distinct bacteria (Rabinovich and Gruppi, 2005). Glimpses of a galectin role in insect innate immunity come from the discovery that a putative galectin homologue was upregulated in Anopheles after challenge with bacteria (Dimopoulos et al., 1998), and that the PRRs in the midgut of the sand fly, which recognizes the lipophosphoglycan that enables the Leishmania parasite to remain adherent to the midgut of its vector, is a galectin (Kamhawi et al., 2004). The observation that alternative splicing of a single mammalian lectin can generate as many as six different isoforms (Rabinovich et al., 2002) adds to the potential immune diversity inherent in the galectin family. Regulated alternative splicing of other immune-related factors (as suggested by Watson et al., 2005) including PGRP (see above) constitute further reservoirs for diversity in insects. Two recent reviews of the innate system in higher organisms (Hargreaves and Medzhitov, 2005; Martin and Frevert, 2005) reveal again the close parallels with insects, but in so doing draw attention to diversity opportunities that are as yet relatively unexplored. Some of these have been mentioned above, but they include the possible functional diversities that derive from combinatorial engagement of PRRs, cooperativity among receptors, cell type-specific expression and modulation by accessory proteins. Insect immune potential is enriched by the fact that the immune response system often shares the same components with insect processes such as apoptosis and development and more dual or multifunctional ‘recruits’ will surely be found. Galectins may expand the insect PRR repertoire, but even in the higher organisms there are examples of pattern recognition that cannot be ascribed to any known PRRs (Hargreaves and Medzhitov, 2005). This case for the existence of novel PRRs is also true of insects. If all the known and potential mechanisms for diversity of insect pattern recognition and response are considered therefore, it may not be unreasonable to conclude that sufficient mechanisms exist to account for the observed insect immune responses.
A key question for the future is whether the invertebrate response to viral infections is, like the invertebrate response to bacterial pathogens, comparable to the mammalian responses.
The similarity in defence mechanisms against bacterial infections in both invertebrates and mammals has allowed invertebrates to be effectively used as infection models for mammalian pathogens such as Staphylococcus aureus (Needham et al., 2004; Garcia-Lara et al., 2005). Insect models that can be reared at mammalian temperatures, such as Manduca sexta (Tobacco hornworm) or Galleria mellonella (Greater Wax moth) (reviewed in Kanost et al., 2004; Kavanagh and Reeves, 2004), may not only be appropriate invertebrate models to investigate mammalian bacterial pathogens, but also may enable a reduced use of mammalian models.
On the other hand, the similar defence mechanisms in invertebrates and mammals should alert us to the possibility that pathogenicity mechanisms acquired to infect invertebrates may be adapted to mammalian pathogenicity (Harb et al., 2000; Waterfield et al., 2005). Insects and other invertebrates are likely sources of mammalian pathogens and a selection ground for virulence factors that may lead to novel human afflictions. This should be a further incentive to improve our understanding of the insect and mammalian immune systems.
We thank the reviewers for their perceptive and helpful suggestions. V.P. thanks the Gates Cambridge Trust and Universities UK for supporting this work.