Specificity and signaling in the <em>Drosophila</em> immune response

Overview of Drosophila Immunity

Insects, such as Drosophila, thrive in microbe-rich environments. Not surprisingly, they have evolved complex mechanisms to combat microbial infection. These defenses include structural barriers to infection, such as the cuticle and peritrophic membrane. Insects also rely on inducible responses such as phagocytosis, the production of antimicrobial compounds, and homeostatic mechanisms that help repair the damage caused by infection (Lemaitre and Hoffmann, 2007). Together, these defense mechanisms allow insects to be broadly resistant to a large range of pathogens without an acquired immune response.

The inducible humoral insect immune response has been most widely studied in the favorite model system Drosophila melanogaster, where microbial challenge leads to the rapid and robust production of a battery of antimicrobial peptides (AMPs). Several families of AMPs have been described in Drosophila, with antifungal and anti-bacterial (both anti-Gram-negative or anti-Gram-positive) activities. Some of these AMPs appear to be unique to insects, e.g., Diptericin, while others have homologs in mammals, e.g., Defensins, Cecropins and Drosomycin (Lee et al., 1989; Simon et al., 2008).

As best we know, production of AMPs is regulated at the transcriptional level. To date, nearly all AMP genes have been found to be controlled by NF-κB family transcription factors. Drosophila have two distinct pathways which activate NF-κB factors and drive transcription of AMP genes following infection. The Toll pathway responds to several different types of microbes, including fungi and many Gram-positive bacteria, and leads to the activation of the NF-κB family members DIF and Dorsal. On the other hand, the IMD pathway is activated by Gram-negative and certain types of Gram-positive bacteria and leads to the activation of the NF-κB precursor Relish. The details of the how different microbes are detected and discriminated by these two pathways is the main focus of this review. First, the basic outline of these two signaling pathways will be summarized.

Toll and IMD Signaling

As mentioned above, the Toll pathway is able to activate two NF-κB homologs, DIF and Dorsal. Both of these proteins are similar to mammalian p65, and are held in the cytoplasm of unstimulated cells by the Drosophila IκB homolog Cactus. Like mammalian IκBs, Cactus is phosphorylated and degraded upon stimulation (Fernandez et al., 2001). One outstanding question within this pathway is the identity of the Cactus kinase. To date, only the kinase Pelle, homologous to the mammalian IRAK family of kinases, has been shown to function in the Toll pathway. In mammals, IRAK kinases are indirectly involved in the phosphorylation of IκBs, being required for the initiation of a kinase cascade that culminates in the activation of the IκB kinase (IKK) complex, which directly phosphorylates IκBα (Skaug et al., 2009). In flies, a similar kinase cascade may be involved, although the components are not yet identified. Alternatively, it remains possible that Pelle directly phosphorylates Cactus on the residues necessary for its degradation. This issue remains unresolved.

It is clear that Pelle is a component of a trimeric complex that associates with the active form of the transmembrane receptor Toll. Like all the mammalian Toll-like Receptors and the mammalian IL-1 Receptor, the cytoplasmic domain of Drosophila Toll contains a TIR domain. This TIR domain interacts with the Drosophila MyD88 homolog via a homotypic TIR:TIR interaction (Tauszig-Delamasure et al., 2002; Charatsi et al., 2003; Kambris et al., 2003). MyD88 also contains a Death domain (DD), which interacts with the protein Tube. Through its other face, the Tube DD also interacts with Pelle (Towb et al., 1998; Sun et al., 2002; Sun et al., 2004). This trimeric MyD88/Tube/Pelle complex is thought to interact transiently with the cytosolic TIR domain of Toll. This association likely leads to the auto-phosphorylation and activation of Pelle, which in turn leads, directly or indirectly, to the phosphorylation of Cactus.

Once Cactus is degraded, the NF-κB factors DIF and Dorsal translocate to the nucleus where they control the transcription of target genes. In the adult fly, DIF is critical for the activation of AMP gene transcription (Meng et al., 1999; Rutschmann et al., 2000a; De Gregorio et al., 2001, 2002; Irving et al., 2001). In larvae, redundancy is observed between DIF and dorsal, and only a double mutant fails to induce AMP genes. In addition to controlling AMP gene expression, DIF and dorsal together also seem to regulate the survival of hemocytes in larvae (Qiu et al., 1998; Matova and Anderson, 2006). Dorsal (and much of the rest of the Toll pathway) also plays a critical role in early embryonic development, in patterning the dorso-ventral axis. During development, phosphorylaztion of Dorsal is linked to enhanced nuclear localization and transcriptional activation (Drier et al., 1999, 2000). DIF is not required in development but can partially substitute for Dorsal in this process when expressed in the embryo (Stein et al., 1998). The role of phosphorylation of DIF, in the context of the immune response, has not been examined (for an overview of the Toll signaling pathway, see Fig. 1).

Figure 1

Toll signaling pathway. The Toll signaling pathway and its multiple modes of activation in the Drosophila immune response. Three distinct mechanisms for microbial recognition, leading to the cleavage of Spätzle and activation of Toll, are illustrated. The mechanisms include peptidoglycan detection, by PGRP-SA, PGRP-SD and GNBP1, β-glucan detection through GNBP3, and protease activity sensing via the serine protease Persephone. All these detection modalities lead to the activation of the Spätzle processing enzyme (SPE) which converts this cytokine into its active form, for binding and activating Toll. The intracellular signal transduction downstream of Toll is very similar to the MyD88-dependent pathway, which functions downstream of most mammalian TLRs. The key features of this pathway include a trimeric receptor associated complex, containing MyD88, Tube and Pelle, which ultimately lead to the phosphorylation and degradation of the IκB homology Cactus and the nuclear translocation of NF-κB homologs Dif and Dorsal.

The IMD pathway culminates in the activation of a third NF-κB homolog, known as Relish. Like the mammalian NF-κB precursor proteins p100 or p105, Relish contains an N-terminal Rel Homology Domain and C-terminal IκB-like ankyrin repeats. In unstimulated cells, the C-terminal IκB-like domain is believed to hold full length Relish in the cytoplasm (Stöven et al., 2000). After immune stimulation Relish is endoproteolytically cleaved, and the N-terminal transcription factor module translocates into the nucleus while the C-terminal IκB-like domain remains in the cytoplasm. In addition to this proteolytic cleavage, full activation of Relish also requires phosphorylation on two residues, serines 528 and 529, (Erturk-Hasdemir et al., 2009). This phosphorylation is not required for Relish cleavage, but instead seems to be crucial for the transcriptional activation of some Relish target genes. Another protein, known as Akirin, also functions in the nucleus for the induction of AMP gene expression, but how it interacts with Relish, if at all, is unclear (Goto et al., 2008).

Relish cleavage and phosphorylation are controlled by two interconnected branches of the IMD signaling pathway. Serines 528 and 529 are directly phosphorylated by the Drosophila IKK complex. This kinase complex includes Drosophila IKKβ and IKKγ homologs, also known as IRD5 and Kenny, respectively (Rutschmann et al., 2000b; Silverman et al., 2000; Lu et al., 2001). IKK activation, in turn, requires the MAP3K TAK1 and its binding partner TAB2. Interestingly, both TAB2 and IKKγ include conserved K63-polyubiquitin binding domains. These non-degratory ubiquitin chains are thought to function in the IMD pathway, however the exact molecular mechanisms involved are not yet clear (Zhou et al., 2005). The putative E3 ubiquitin ligase DIAP2 is also required in the IMD pathway and may promote K63-chain formation (Kleino et al., 2005; Leulier et al., 2006; Huh et al., 2007). Further upstream, kinase activation also requires the imd protein, the Drosophila FADD homolog, and the caspase-8 like DREDD (Leulier et al., 2000; Georgel et al., 2001; Leulier et al., 2002; Naitza et al., 2002). These three proteins may form a trimeric complex as FADD can interact with both IMD and DREDD. How this complex signals the activation of TAK1 is still under investigation. In addition to a poorly defined role in the activation of TAK1, DREDD is also required for the cleavage of Relish. Interestingly, the IKK complex is also required for Relish cleavage, but its kinase activity is not involved in this function (Stöven et al., 2003; Erturk-Hasdemir et al., 2009). Instead, the IKK complex may function as a scaffold facilitating the cleavage of Relish by DREDD (for an overview of the IMD signaling pathway, see Fig. 2).

Figure 2

IMD signaling pathway. This pathway is preferentially triggered by DAP-type peptidoglycan, common to Gram-negative bacteria and certain Gram-positives, especially Bacillus spp. DAP-type peptidoglycan can be recognized by different receptors, depending on its location and size. Large insoluble peptidoglycan is recognized on the cell surface by PGRP-LCx, while smaller peptidoglycan fragments, like TCT, can be recognized at the cell surface by a ligand-induced dimer of PGRP-LCx and PGRP-LCa. In addition, DAP-type peptidoglycan that reaches the cytosol can trigger another receptor, PGRP-LE. Both PGRP-LC and PGRP-LE trigger a similar intracellular signal transduction pathway, as outlined here, that culminates in the activation of the NF-κB precursor Relish. In addition, recognition of intracellular DAP-type peptidoglycan by PGRP-LE also triggers an autophagic response, which is important in the protection against intracellular bacteria.

In addition to playing a key role in activation of the IKK complex and Relish, TAK1 also activates Drosophila JNK signaling through Hemipterous (MKK7) and Basket (JNK) (Sluss et al., 1996; Holland et al., 1997; Chen et al., 2002). Thus, TAK1 plays a crucial role at the nexus of JNK and NF-κB signaling in this innate immune signaling pathway. NF-κB (Relish) plays a critical role in the induction of AMP genes; without Relish, no AMP gene expression is detected. However, the role of the JNK pathway in AMP regulation remains controversial. Several reports have argued that JNK signaling actually down-modulates AMP gene expression (Kim et al., 2005, 2007), while another report has argued that JNK signaling is required for AMP induction (Delaney et al., 2006). More studies are required to resolve these conflicting conclusions. In addition to the possible inhibitory activity of JNK on NF-κB-responsive AMP expression, the Relish branch of the pathway also seems to generate an inhibitor of JNK signaling (Park et al., 2004). The mechanism by which Relish-induced gene products interfere with JNK signaling has been suggested to involve the ubiquitin E3 ligase POSH. POSH is thought to target TAK1 for degradation (Tsuda et al., 2005), however it is not clear how this would preferentially inhibit JNK but not Relish dependent responses. In addition to activating JNK and NF-κB/Relish signaling, the IMD pathway also induces the activation of the Drosophila p38 pathway, which requires imd protein but not TAK1 or any of the downstream components (Zhuang et al., 2006). Little is known about p38 signaling in the Drosophila immune response, however it does appear to play a critical role in regulating the ROS generating enzyme DOUX in the gut, thereby controlling the intestinal microflora (Ha et al., 2009).

Microbial recognition

Both the Toll and IMD pathways are stimulated by bacterial peptidoglycans (Leulier et al., 2003; Kaneko et al., 2004). In addition, the Toll pathway can be triggered by β-glucans, from fungal cell wall, or by proteases directly released from pathogens (Gottar et al., 2006; El Chamy et al., 2008). Peptidoglycan (PGN) is the major structural component of the bacterial cell wall. PGN structures display a great deal of diversity, varying widely among different classes of bacteria. However, all PGNs include a carbohydrate backbone, usually consisting of alternating N-acetyl-Glucosamine and N-acetyl-Muramic acid residues and short stem-peptides containing both L and D amino acids. These stem-peptides are often cross-linked to each other to stiffen the cell wall; the precise structure of these cross-linking structures is highly variable. The stem-peptides also display a great deal of variation in their amino acid constituents. The carbohydrate backbone is more constant but also can be modified by various chemical substitutions, such as acetylation (Schleifer and Kandler, 1972; Mengin-Lecreulx and Lemaitre, 2005) (see Fig. 3 for a diagram of PGN structures).

Figure 3

Peptidoglycan structure. (A) As shown, peptidoglycan has a common core structure with a great deal of inherent variation. Most notably, the constituent of the third position of the stem-peptide can vary, and are most commonly L-lysine or meso-DAP. In addition, the amount and exact chemical nature of the crosslinking bridges can vary, with some examples noted in the box below. TCT is a monomeric unit of the DAP-type peptidoglycan chain, and is indicated in the dashed box. (B) Structures of lysine and DAP.

Both the Toll and IMD pathways rely on peptidoglycan recognition proteins (PGRPs) for sensing PGNs (Table 1). This family of proteins is structurally similar to type 2 amidases (N-acetylmuramyl-L-alanine amidases), a class of enzymes that hydrolyze the bond between the lactyl group in acetylmuramic acid and L-alanine in the stem-peptide of PGN. In fact, some PGRPs are type 2 amidases, while others lack the catalytic cysteine and instead function by binding PGN (Mellroth et al., 2003). Drosophila encode for 13 PGRP genes, making about 17 distinct proteins through alternative splicing (Werner et al., 2000). Six of the Drosophila PGRPs (PGRP-SB1, -SB2, SC1a/b, SC2, -LB) are known or predicted type 2 amidases, that are involved in degrading PGN and dampening immune activation (Mellroth et al., 2003; Bischoff et al., 2006; Zaidman-Remy et al., 2006). The other seven lack type 2 amidase activity but function through binding PGN. In particular, 4 PGRPs (PGRP-SA, -SD, -LC, and LE) function as receptors in the IMD or Toll pathways, as detailed below. PGRP-LF seems to function as a decoy receptor, binding PGN but not activating immune signaling (Persson et al., 2007; Maillet et al., 2008), while the functions of PGRP-LA and -LD remain elusive (Royet and Dziarski, 2007).

Initial studies suggested that the IMD pathway was activated preferentially by Gram-negative bacteria, which lead many to assume that LPS, the most potent activator of the mammalian innate immune response, would be the main agonist of this pathway (Samakovlis et al., 1990; Werner et al., 2003). However, a careful analysis of published results suggested otherwise. In addition to Gram-negatives, certain Gram-positive bacteria, e.g., Bacillus spp, were also IMD pathway activators (Lemaitre et al., 1997). Subsequently, Lemaitre's group showed that diaminopimelic acid (DAP)-type PGN, from Escherichia coli or Bacillus thurengensis, activated the IMD pathway, while the Silverman group demonstrated that purified LPS samples were unable to trigger the IMD pathway, and IMD agonistic activity could be traced to DAP-type PGN (Leulier et al., 2003; Kaneko et al., 2004). Lemaitre's group also showed that the Toll pathway was activated by PGN, but in this case lysine-type PGN from Gram-positives like M. luteus and E. fecalis was more potent.

Following these discoveries, a great deal has been learned about the molecular mechanisms involved in detecting various types of PGN. PGRP-SA and PGRP-SD are required for the recognition of PGN and the activation of Toll signaling (see below for more detail on the Toll pathway) (Michel et al., 2001; Bischoff et al., 2004), while in the IMD pathway either of two receptors, PGRP-LC or PGRP-LE, are capable of recognizing DAP-type PGN (Gottar et al., 2002; Ramet et al., 2002; Kaneko et al., 2004, 2006; Takehana et al., 2004; Choe et al., 2005). PGN binding to either PGRP-LC or -LE triggers the IMD signaling pathway, described above, via a short conserved domain found in the N-terminus of both receptors (Kaneko et al., 2006). This conserved signaling domain has some similarity to the RHIM domain found in the mammalian proteins RIP1, RIP3 and TRIF (Meylan et al., 2004). However the molecular mechanisms involved in signaling by the PGRP-LC/LE RHIM-like domain still remain to be determined. Regardless of the mechanisms involved, activation of these receptors leads to activation of both Relish cleavage and Relish phosphorylation.

One key discovery, which enabled detailed molecular and biophysical analyses of PGRP-LC and PGRP-LE, was that a monomeric fragment of PGN from Gram-negative bacteria, known as Trachael Cytotoxin (TCT), potently activates the IMD pathway (Kaneko et al., 2004; Stenbak et al., 2004). TCT is a disaccharide-tetrapeptide, featuring DAP at the third position of its stem-peptide, isolated from culture supernatants of B. pertussis (Fig. 3A). Synthetic lactyl-tetrapeptides (substructures of TCT) are able to serve as weak agonists of the IMD pathway only if they contain DAP at this third position, providing further demonstration that the DAP residue is key to triggering the IMD pathway (Kaneko et al., 2004). PGRP-LC and PGRP-LE preferentially bind DAP containing PGN (Takehana et al., 2002; Swaminathan et al., 2006). The crystal structures of both PGRP-LC and PGRP-LE bound to TCT have been solved and show that a key arginine, Arg254 in PGRP-LE, provides the critical DAP-specific interaction (Chang et al., 2006; Lim et al., 2006). Compared to lysine, DAP contains an additional carboxylate group (Figure 3B), and the guanidinium group of Arg254 forms a bidentate salt bridge with this moiety . In fact, this key arginine residue is common to all known DAP-type recognizing PGRPs and is found in the base of a deep cleft in which the PGN stem-peptide binds. In addition, DAP binding PGRPs also contain nearby glycine and tryptophan residues that are involved in other DAP-specific interactions (Swaminathan et al., 2006).

PGRP-LC and PGRP-LE are found in different subcellular compartments and detect microbes in these distinct environments. PGRP-LC is a single pass, type 2 transmembrane receptor that is found primarily on the cell surface (Kaneko et al., 2004; Mellroth et al., 2005). Through alternative splicing, PGRP-LC encodes for 3 different receptors (PGRP-LCa, -LCx, and -LCy) each with an identical cytosolic domain but distinct extracellular ligand binding PGRP domains. On the other hand, PGRP-LE encodes for only one protein isoform, which lacks a transmembrane domain and functions as a cytosolic receptor (Werner et al., 2000). While PGRP-LC is critical for recognizing extracellular bacteria (and extracellular PGN), PGRP-LE surveils the cytosol for intracellular bacteria and/or small fragments of PGN that enter cells (Kaneko et al., 2006; Yano et al., 2008). RNAi-based studies showed that the different splice isoforms of PGRP-LC are involved in recognizing different types of PGN. In particular, the recognition of polymeric PGN, as isolated from E. coli, requires only PGRP-LCx. Moreover, PGRP-LCx mutants do not induce AMP genes in response to E. coli infection. However, the response to extracellular monomeric PGN (TCT) requires both PGRP-LCx and PGR-LCa, while intracellular TCT triggers PGRP-LE.

The finding that polymeric and monomeric DAP-type PGNs trigger distinct PGRP-LC receptors was explained, in part, by the crystal structure of PGRP-LCa/x bound to TCT. As mentioned above, most PGRPs contain a deep cleft in which peptidoglycan fragments (sometimes referred to as muropeptides) bind. PGRP-LCa is the exception; it contains two unique dipeptide sequences that disrupt the PGN binding cleft and occlude PGN binding. Conversely, PGRP-LCx has a typical DAP-type PGN binding cleft and can avidly bind TCT or polymeric PGN. Upon binding TCT, PGRP-LCx and PGRP-LCa heterodimerize (Chang et al., 2005; Mellroth et al., 2005). The crystal structure of this ligand bound dimeric complex shows that the carbohydrate portion of TCT makes key contributions to the dimerization interface, providing a clear explanation for the TCT-induced dimerization (Chang et al., 2006). It is reasonable to hypothesize that the ligand-induced heterodimerization of PGRP-LCa and PGRP-LCx is critical for activation of downstream signaling, although this has not yet been demonstrated. This model of dimerization-induced signaling does not explain how PGRP-LCx alone is sufficient for IMD signaling triggered by polymeric PGN. The crystal structures suggest that PGRP-LCx is unlikely to form homo-multimers upon binding polymeric PGN, because of a steric clash at the putative dimerization interface. Therefore, a distinct model must be proposed for signaling by polymeric PGN and PGRP-LCx. In this case, the ligand is polyvalent and likely binds to multiple individual PGRP-LCx receptors, perhaps creating a ‘cluster’ of receptors, thereby generating a density of PGRP-LCx cytosolic domains. This clustering, per se, may be sufficient to activate signal transduction, or perhaps the intracellular domains actually form higher order protein-protein interactions while the extracellular domains remain clustered on one large fragment of PGN, but not in direct contact with each other. Future studies are required to examine these possibilities.

Like the PGRP-LCa/x heterodimers, PGRP-LE also multimerizes upon binding TCT. However, the PGRP-LE-TCT complex forms very high order multimers in a ‘head to tail’ fashion. Like the PGRP-LC structures, the PGRP-LE structure was solved with the isolated PGRP domain, and we cannot be certain what quaternary structure the holo-receptor forms upon TCT binding. However, these biophysical studies clearly demonstrate that TCT causes PGRP-LE to multimerize into large complexes (Lim et al., 2006). As mentioned above, PGRP-LE detects DAP-type PGN that enters the cytosol. PGN may enter the cytosol from infection with intracellular bacteria, like Listeria monocytogenes, or small PGN fragments, such as TCT, appear to directly enter into cells. Upon binding these PGNs, PGRP-LE likely forms higher order multimers and triggers IMD signaling. In addition, PGRP-LE activation can also induce an autophagic response that helps protect against intracellular microbes (Yano et al., 2008). The role of ligand-induced receptor multimerization in the activation of IMD signaling pathway, via PGRP-LCs and PGRP-LE, requires further study. As mentioned above, these receptors signal through a RHIM-like domain in their N-terminal domains. The mechanism by which the RHIM-like domain functions in the context of dimerized, multimerized or clustered PGRP receptors is unknown.

Toll activation by PGN and beyond

Unlike the IMD pathway, which is activated in a fairly specific manner by DAP-type PGN, Toll activation occurs indirectly by a wider array of immune stimuli. Toll functions more like a cytokine receptor, binding a processed form of the cytokine Spätzle, a member of the cysteine knot family of growth factors and cytokines (Weber et al., 2003; Hu et al., 2004; Hoffmann et al., 2008). Spätzle is made as a pro-protein that is found circulating in the hemolymph. Upon immune activation (or developmental cues), serine protease cascades are triggered that culminate in the cleavage of Spätzle. Once processed, mature Spätzle binds to and dimerizes the transmembrane receptor Toll, initiating the intracellular signaling pathway described above. Four different serine protease cascades appear to converge on the cleavage of Spätzle. In early development, the protease Easter is responsible for cleaving Spätzle. During the immune response, bacterial PGN, fungal β-glucans, and microbial proteases are sensed by three distinct mechanisms, but converge upon activation of one serine protease, known as the Spätzle processing enzyme (SPE), which in turn cleaves and activates Spätzle (Fig. 1).

The serine protease Persephone appears to function as a sensor for proteases secreted by both fungal and bacterial pathogens (Gottar et al., 2006; El Chamy et al., 2008). Persephone is likely activated by cleavage after a histidine residue, unlike the other proteases involved in the Toll signaling pathways, and maybe a good target for subtilisin-like proteases produced by microbial pathogens (El Chamy et al., 2008). The activation of the Persephone-Toll pathway by microbial proteases occurs independently of recognition of microbial cell wall material, which can also stimulate the Toll pathway through more classical receptor-mediated recognition.

For example, β-glucans from the cell wall of yeast are recognized by Gram-negative binding protein 3 (GNBP3) (Gottar et al., 2006), while two secreted PGRP receptors and GNBP1 are involved together in PGN recognition (Gobert et al., 2003; Bischoff et al., 2004; Wang et al., 2006, 2008). [Despite their name, none of the GNBPs have been linked to the response to Gram-negative bacteria, but GNBP1 and GNPB3 are involved in the recognition of fungal or Gram-positive bacterial cell walls]. The N-terminus of GNBP3 binds to long β-1,3 glucans, common to the cell walls of many types of fungi, especially yeast (Mishima et al., 2009). GNBP1, on the other hand, is involved in PGN recognition, although its role is controversial.

Lysine-type PGNs, common to many Gram-positive bacteria, are potent agonist of the Toll pathway. As in the IMD pathway, PGRP receptors are critical for the recognition of lysine-type PGN. In particular, two secreted PGRPs, PGRP-SA and -SD, are involved in the Toll pathway. Genetic studies have shown that some Gram-positive bacteria and the PGN from these same species, such as M. luteus, are sensed through PGRP-SA (Michel et al., 2001). In fact, the structure of PGRP-SA bound to a lysine-containing muropeptide has been solved. PGRP-SA can also bind DAP-type PGN, albeit to a lesser degree. However, PGRP-SA appears to be able to specifically cleave DAP-type muropeptides, removing the final amino acid in the stem-peptide. It has been postulated that this carboxypeptidase activity prevents DAP-type PGN from stimulating the Toll pathway via PGRP-SA (Chang et al., 2004).

Interestingly, not all lysine-type PGN requires PGRP-SA to trigger the Toll pathway. In particular, Staphylococcus aureus, Enterococcus faecalis, Streptococcus pyogenes, and Staphylococcus saprophyticus infections still produce strong AMP gene responses in PGRP-SA mutant (seml) flies. Response to these bacteria or their PGNs instead requires either PGRP-SA or PGRP-SD (Bischoff et al., 2004). The mechanism of PGRP-SD-mediated recognition of some, but not all, lysine-type PGN producing bacteria remains unclear. One possibility is a structural difference common to those PGNs sensed by PGRP-SD, which prevents detection by PGRP-SA, or vice versa. However, biochemical studies of PGRP-SD do not support the notion that it is involved in recognizing lysine-type PGN. Crystallographic studies show that PGRP-SD has a deep PGN binding cleft, typical of all PGRPs, and this binding cleft includes an arginine (Arg90) in the key position typical of DAP-PGN specific recognition. In fact, binding studies confirm that PGRP-SD binds DAP-type PGN, from B. subtilis, but not Lys-type from S. aureus (Leone et al., 2008). In addition, the moderate induction of drosomycin observed following either B. subtilis or E. coli infection, which is Toll dependent (Leulier et al., 2003), required both PGRP-SA and PGRP-SD. So, somehow PGRP-SA and PGRP-SD function together in the recognition of DAP-type PGN, for moderate Toll induction, but function in a more redundant manner for the recognition of certain Lys-type PGN, for robust Toll induction. However, the recognition of M. luteus PGN appears to more simply require only PGRP-SA. The molecular mechanisms involved in the recognition of various PGNs by PGRP-SA and/or PGRP-SD remain to be determined.

As mentioned above, GNBP1 also functions in PGN recognition and Toll signaling. In fact, PGRP-SA, -SD, and GNBP1 form a trimeric complex together in the presence of PGN fragments (Wang et al., 2008). Some groups have reported that GNBP1 provides a critical PGN processing activity to this complex, required to generate small PGN fragments which are bound by PGRP-SA and/or PGRP-SD for Toll activation (Filipe et al., 2005; Wang et al., 2006). However, another group has reported that they do not observe a similar PGN digesting activity associated with GNBP1, in Drosophila or Tenebrio molitor (Buchon et al., 2009). Instead, this group proposes that GNBP1 serves to link the PGRPs with the downstream serine protease cascade, described below. Thus, it appears that GNBP1 functions in a complex with the PGN sensing receptors PGRP-SA and PGRP-SD, but the biochemical mechanism by which it contributes to immune recognition or Toll signaling are not yet clear.

Both the GNBP3-mediated recognition of β-glucans and the PGRP-SA/SD/GNBP1-mediated recognition of bacterial PGNs trigger Toll signaling through the same serine protease cascade. This cascade involves the modular serine protease (ModSP), which is probably directly activated by these microbial sensing receptor complexes, and at least two downstream CLIP-domain serine proteases - Grass and SPE. As mentioned above, SPE cleaves and thereby activates Spätzle, the ligand for Toll. Another protease, known as Spirit may function between Grass and SPE, and other non-catalytic serine-protease homologs, Sphinx 1/2 and Spheroide, were also implicated this pathway by RNAi based studies (Kambris et al., 2006). However, the assignment of these factors to this pathway requires further genetic and biochemical characterization. The pathways presented in Fig. 1 suggest a protease cascade that is consistent with the genetic analysis of mutants in Drosophila and the biochemical analysis of the cascade from the hemolymph of Tenebrio. However, biochemistry of the Drosophila serine protease cascade still requires further study, as several issues remain unresolved, including the role of Spirit. In addition, the predicted specificity of the Drosophila ModSP does not match the predicted cleavage site of the downstream serine protease Grass, and ModSP does not cleave Grass in vitro (Buchon et al., 2009). Thus, it remains possible that other factors may be involved. Currently, it is not clear how (or even if) PGN binding to PGRP-SA/SD/GNBP1 leads to the activation of ModSP, and, as mentioned above the exact biochemical role of GNBP1 remains controversial.

Concluding remarks

The goal of this review is to summarize recent work on the Toll and IMD pathways, two important innate immune signaling pathways in Drosophila, with an emphasis on the molecular mechanisms involved in microbial recognition in this model system. This review is not meant to be a comprehensive analysis of all aspect of the insect immune response. Notably, many very exciting studies have been published recently on phagocytosis, anti-viral immunity, melanization and clotting - all topics not covered here. Instead, significant detail is presented on the biochemistry and genetics of bacterial recognition by the PGRP family of innate immune receptors. Over the past 5 years, much progress been made in this area, and the basic underpinnings of how bacteria are recognized by PGRGP receptors, which bind the bacterial cell-wall derived compound peptidoglycan, has been resolved. In addition, the preferential binding of DAP-type PGN by some of these receptors, notably PGRP-LC and PGRP-LE, provides a firm explanation on the specific activation of the IMD pathway by Gram-negative and certain Gram-positive bacteria. However, the specificity, or lack thereof, in activating the Toll pathway by different types of peptidoglycans still requires further investigation. Moreover, in both IMD and Toll signaling, it remains unclear how peptidoglycan binding by these PGRP receptors leads to the activation of downstream signaling events. These questions will be the focus of future study.