Microarray studies have shown recently that microbial infection leads to extensive changes in the Drosophila gene expression programme. However, little is known about the control of most of the fly immune‐responsive genes, except for the antimicrobial peptide (AMP)‐encoding genes, which are regulated by the Toll and Imd pathways. Here, we used oligonucleotide microarrays to monitor the effect of mutations affecting the Toll and Imd pathways on the expression programme induced by septic injury in Drosophila adults. We found that the Toll and Imd cascades control the majority of the genes regulated by microbial infection in addition to AMP genes and are involved in nearly all known Drosophila innate immune reactions. However, we identified some genes controlled by septic injury that are not affected in double mutant flies where both Toll and Imd pathways are defective, suggesting that other unidentified signalling cascades are activated by infection. Interestingly, we observed that some Drosophila immune‐responsive genes are located in gene clusters, which often are transcriptionally co‐regulated.
Innate immunity plays a very important role in combating microbial infection in all animals. The innate immune response is activated by receptors that recognize surface determinants conserved among microbes but absent in the host, such as lipopolysaccharides, peptidoglycans and mannans (Medzhitov and Janeway, 1997). Upon recognition, these receptors activate multiple and complex signalling cascades that ultimately regulate the transcription of target genes encoding effector molecules. Importantly, different pathogens elicit specific transcription programmes that can now be investigated by using microarray technology (De Gregorio et al., 2001; Huang et al., 2001; Irving et al., 2001).
Drosophila is devoid of an adaptive immune system and relies only on innate immune reactions for its defence. Genetic and molecular approaches have shown that Drosophila is a powerful model system to study innate immunity, which seems to be remarkably conserved from flies to mammals (Hoffmann and Reichhart, 2002; Tzou et al., 2002a). To combat microbial infection, Drosophila activates multiple cellular and humoral responses including, for example, proteolytic cascades that lead to blood coagulation and melanization, the production of several effector molecules such as antimicrobial peptides (AMPs) and the uptake of microorganisms by blood cells (Tzou et al., 2002a). AMPs are made in the fat body, a functional equivalent of mammalian liver, and secreted in the haemolymph, where they directly kill invading microorganisms (Hoffmann and Reichhart, 2002). Genetic analyses have shown that AMP genes are regulated by the Toll and Imd pathways (Hoffmann and Reichhart, 2002; Tzou et al., 2002a). The Toll pathway is activated mainly by Gram‐positive bacteria and fungi and controls in large part the expression of AMPs active against fungi, while the Imd pathway responds mainly to Gram‐negative bacteria infection and controls antibacterial peptide gene expression (Hoffmann and Reichhart, 2002; Tzou et al., 2002a). However, most of the AMP genes can be regulated by either pathway, depending on the type of infection, and the selective activation of Toll or Imd by different classes of pathogens leads to specific AMP gene expression programmes adapted to the aggressors. Thus, the control of AMP genes by the Toll and Imd pathways provides a good model to study how recognition of distinct microbes generates adequate responses to infection.
The Imd and Toll pathways do not appear to share any intermediate components and mediate differential expression of AMP‐encoding genes via distinct NF‐κB‐like transcription factors (Hoffmann and Reichhart, 2002; Tzou et al., 2002a). Upon infection, the Toll pathway is activated in the haemolymph by an uncharacterized serine protease cascade that involves the serpin Necrotic and leads to the processing of Spaetzle, the putative Toll ligand. Binding of Spaetzle to Toll activates an intracellular signalling cascade, involving the adaptor proteins dMyD88 and Tube, and the kinase Pelle, that leads to degradation of the Iκ‐B‐like protein Cactus and the nuclear translocation of the NF‐κB‐like transcription factors Dif and Dorsal (Hoffmann and Reichhart, 2002; Tzou et al., 2002a). An extracellular recognition factor, peptidoglycan recognition protein (PGRP)‐SA, belonging to a large family of proteins that bind to peptidoglycan has been implicated in the activation of the Toll pathway in response to Gram‐positive bacteria but not fungi (Michel et al., 2001). These data support the idea that the Toll pathway is activated by soluble recognition molecules that trigger distinct proteolytic cascades converging to Spaetzle.
Recently, several studies have led to the genetic and molecular identification of seven components of the Imd pathway (Hoffmann and Reichhart, 2002; Tzou et al., 2002a). The ultimate target of the Imd pathway is Relish, a rel/NF‐κB transactivator related to mammalian P105. Current models suggest that this protein needs to be processed in order to translocate to the nucleus. Its cleavage is dependent on both the caspase Dredd and the fly Iκ‐B–kinase (IKK) complex. Epistatic experiments suggest that dTAK1, a MAPKKK, functions upstream of the IKK complex and downstream of Imd, a protein with a death domain similar to that of mammalian receptor‐interacting protein (Hoffmann and Reichhart, 2002; Tzou et al., 2002a). Recently, three independent studies have shown that a putative transmembrane protein, PGRP‐LC acts upstream of Imd and probably functions in sensing microbial infection (Choe et al., 2002; Gottar et al., 2002; Rämet et al., 2002b). The Drosophila Toll and Imd pathways share many features with the mammalian TLR/IL‐1 and TNF‐R signalling pathways that regulate NF‐κB, pointing to an evolutionary link between the regulation of AMP gene expression in flies and the mammalian innate immune response (Hoffmann and Reichhart, 2002; Tzou et al., 2002a)
We identified 400 Drosophila immune‐regulated genes (DIRGs) through a microarray analysis of the transcription programmes induced by septic injury and by natural fungal infection (De Gregorio et al., 2001). Many of these genes were assigned to functions related to the immune response including, in addition to the AMP response, microbial recognition, phagocytosis, melanization, coagulation, reactive oxygen species (ROS) production, wound healing and iron sequestration. Although the regulation of AMP genes by Imd and Toll pathways has been studied extensively, little is known about the role of these two pathways in the control of other genes regulated by infection in Drosophila. In this study, we have characterized further the role of the Toll and Imd pathways in the Drosophila host defence. To study the contribution of each pathway in the resistance to infection, we first compared the susceptibility of flies carrying mutations affecting the Toll, Imd or both signalling cascades with several types of bacterial and fungal infection by a survival test. Secondly, we analysed, using northern blots, the expression of AMP genes after different types of infection in the same mutants. Finally, we monitored by microarray analysis the effect of mutations affecting the Toll and Imd cascades on the transcriptional reprogramming induced by septic injury. Our study demonstrates that the Toll and Imd pathways are the major regulators of the immune response in Drosophila adults.
Contribution of the Toll and Imd pathways to resist microbial infection
It has been shown that mutants of the Imd pathway are more susceptible than wild‐type flies to Gram‐negative bacterial infection, while mutants of the Toll pathway are more susceptible to fungal and Gram‐positive bacterial infection (Lemaitre et al., 1996; Rutschmann et al., 2002). Three Drosophila lines carrying mutations affecting both the Toll and Imd pathways (imd;spz, imd;Tl and dif,kenny), have been reported to be sensitive to both bacterial and fungal infections (Lemaitre et al., 1996; Leulier et al., 2000; Rutschmann et al., 2000, 2002). However, these double mutant lines probably retain limited Toll or Imd activity because the imd allele is a hypomorph and dif mutants retain Dorsal activity (Georgel et al., 2001). We have generated double mutant Drosophila lines by recombining two strong alleles of the Toll pathway (spz: spzrm7 and Tl: Tl1‐RXA/Tlr632) with a null allele of relish (rel: relE20). The comparison of the susceptibility to microbial infection of flies deficient for either the Toll (spz or Tl) or Imd (rel) pathway and flies mutated for both rel,spz or rel,Tl allows us to analyse in detail the contribution of each pathway to host defence.
Wild‐type OregonR (wt), single (rel, Tl and spz) and double mutant (rel,spz and rel,Tl) adult flies were injected with Gram‐negative bacteria (Escherichia coli), Gram‐positive bacteria (Micrococcus luteus and Enterococcus faecalis) and fungi (Aspergillus fumigatus) or were naturally infected with the spores of the entomopathogenic fungus Beauvaria bassiana (Figure 1). As previously observed, Tl and spz mutants are resistant to E.coli injection while rel flies are highly susceptible, dying within 3 days (Lemaitre et al., 1996; Hedengren et al., 1999; Leulier et al., 2000). Surprisingly, both double mutants (rel,spz and rel,Tl) are more susceptible than rel to E.coli infection, suggesting that the Toll pathway triggers a significant response against Gram‐negative bacteria (Figure 1A). This is in agreement with a previous study showing that dif,kenny double mutants die earlier than kenny flies after infection by E.coli (Rutschmann et al., 2002). To study the contribution of Toll and Imd pathways to resist Gram‐positive bacteria infection, we injected two bacterial strains: M.luteus, which does not kill flies deficient in the Toll or Imd pathway (Leulier et al., 2000); and E.faecalis, which kills spz flies very rapidly (Rutschmann et al., 2002). Interestingly, we noticed that double mutant lines are very sensitive to infection by M.luteus (Figure 1B) and that rel,spz double mutants are slightly more susceptible than spz flies to E.faecalis infection (Figure 1C). These data confirm that the Toll pathway is the most important pathway in fighting Gram‐positive bacterial infection but indicate that the Imd pathway can also play a significant role. Finally, we observed that rel,spz and rel,Tl are almost equally as susceptible as single mutants in the Toll pathway (Tl or spz) to injection of A.fumigatus and to natural infection by B.bassiana, suggesting that the Imd pathway is not essential for the antifungal response (Figure 1D and E).
Next, we analysed the effect of mutations affecting Imd and Toll pathways on the expression of AMP genes after injection of E.coli, M.luteus or A.fumigatus. Figure 2 shows a northern blot analysis of two antibacterial peptide genes (attacin and diptericin) and two antifungal peptide genes (drosomycin and metchnikowin). The double mutants rel,spz and rel,Tl failed to show induction of AMP genes. In fact, the only AMP transcript detectable in these flies is the antifungal drosomycin, which is present at a level similar to that in unchallenged flies. Diptericin is regulated by the Imd pathway, while metchnikowin, attacin and drosomycin are regulated by both pathways. Interestingly, the contribution of each pathway to the expression of each AMP gene depends on the type of infection. For example, in agreement with previous studies (Leulier et al., 2000), drosomycin expression is affected similarly by the rel and Toll pathway mutants after E.coli infection, but is regulated predominantly by the Toll pathway during M.luteus or A.fumigatus infections (Figure 2).
The results obtained by northern blot analysis correlate with the data from survival experiments. The contribution of the two pathways to the control of the antibacterial peptides (Figure 2) is consistent with the augmented sensitivity to bacterial infection of double mutant flies versus single mutants (Figure 1A–C). The level of the antifungal peptide Drosomycin transcript after fungal infection is very similar in the double mutant flies (rel,spz and rel,Tl) compared with Tl and spz single mutants (Figure 2C), consistent with a similar resistance to A.fumigatus and B.bassiana displayed by these four lines (Figure 1C and D). Importantly, the Tl and spz alleles alone, or in combination with rel, display the same behaviour in all survival experiments performed (Figure 1) and have a similar pattern of AMP gene expression (Figure 2), suggesting that Spaetzle is the sole extracellular activator of the Toll pathway in response to microbial infection. However, we noticed that attacin and diptericin expression after A.fumigatus infection is reduced in Tl but not in spz flies (Figure 2C). We extended the analysis of A.fumigatus infection to pelle, tube and dif mutants (data not shown), which display the same AMP expression profile as spz, suggesting that the effect observed in Tl flies is due to the genetic background of the strain used. The complete survival and northern analysis presented here was extended to a strong allele of pelle alone or in combination with rel, which gave similar results to spz and Tl alleles (data not shown).
The Toll and Imd pathways control the majority of Drosophila immune‐regulated genes
To identify which of the 400 previously identified DIRGs are controlled by the Imd and/or Toll pathways, mRNA samples from spz, rel and rel,spz adult males, collected after septic injury with a mixture of E.coli and M.luteus, were hybridized to Affymetrix DrosGenome1 GeneChips capable of measuring mRNA levels for nearly every gene in the Drosophila genome. The gene expression profiles obtained for the mutants flies were compared with our previous analysis of wild‐type flies. Since double mutants start to die within 1 day after bacterial infection (Figure 1A and B), we limited our analysis to the first 6 h of the immune response (time points: 0, 1.5, 3 and 6 h), ensuring that the changes in expression profiles are not an indirect consequence of the sickness of the flies. In addition to loss‐of‐function mutants, we also observed the genome‐wide changes in gene expression of uninfected Tl10b/+ flies carrying a gain‐of‐function allele that constitutively activates the Toll pathway. Each time series was observed in duplicate, while the Tl10b allele was assayed three times. Complete results can be found at http://www.fruitfly.org/expression/immunity/.
General statistics and hierarchical cluster analysis
Out of the 400 DIRGs previously identified, the majority (283) display a significant change in the expression pattern in the first 6 h (for each gene P <0.0025) (Figure 3A). Using an automated approach, we determined for each gene whether the gene expression profile is significantly different in a mutant background compared with wild‐type (see Materials and methods). We observed that half of the 162 up‐regulated genes examined (86) are not induced in rel,spz double mutants, 32 are partially affected and only 44 are still fully induced in this background (Figure 3A, upper table). Similar data were obtained from the analysis of 121 down‐regulated genes, the majority of which are dependent (46) or partially dependent (27) on Relish and Spaetzle for their regulation, while 48 genes show no significant difference in rel,spz compared with wild‐type flies (Figure 3A, lower table). Within the group of DIRGs affected in rel,spz flies, we could distinguish four categories based on their differential response to spz and rel mutations (see Venn diagrams in Figure 3A). Genes affected in rel but not in spz flies are probably controlled by the Imd pathway. In contrast, genes affected in spz flies but not in rel are probably controlled by the Toll pathway. We also found genes that are affected in both single mutants, which are probably regulated by both Imd and Toll pathways, and genes that are affected only in double mutant flies, suggesting that the two pathways play redundant roles in their regulation (Figure 3A). The tables in Figure 3A also show that 34 induced and 12 repressed DIRGs are regulated in Tl10b flies in the absence of infection. Interestingly, 34 of them are significantly affected in the rel,spz background and 23 in spz flies. It does appear that the Imd pathway may be less important in repressing DIRGs, as very few repressed DIRGs are only dependent on Relish (Figure 3A).
To analyse the gene expression profile in more detail, we hierarchically clustered all 400 DIRGs including previous data obtained after B.bassiana infection (time points: 0, 12, 24 and 24 h) and the complete kinetics of wild‐type flies after septic injury (time points: 0, 1.5, 3, 6, 12, 24 and 48 h) (Figure 3B). We observe that most of the genes induced after fungal infection display a late or sustained response after septic injury and are not induced in spz and rel,spz flies, while they are fully induced in rel mutants and are up‐regulated in unchallenged Tl10b flies (Figure 3B, cluster UP6). These data strongly suggest that this group of genes is controlled by the Toll pathway. In addition to the UP6 cluster, Spaetzle can regulate acute phase genes (UP5). In contrast to Spaetzle, Relish controls predominantly early and sustained phase genes, which are not induced by fungal infection or in Tl10b flies (UP4). Up‐regulated genes independent from Relish and Spaetzle (clusters UP1 and UP3) are generally weakly induced by fungal infection and not affected by Tl10b.
The analysis of repressed DIRGs shows that Spaetzle can regulate both early (D5 and D9) and late/sustained (D1 and D3) phase genes, while Relish partially controls a small number of early phase genes (D8). A large group of late phase genes repressed after fungal infection are regulated by both Relish and Spaetzle (D10). Interestingly, a second group of genes strongly repressed after fungal infection (D2) are not affected by rel and spz mutations but affected in Tl10b flies.
Our analysis shows that the Toll and Imd pathways regulate the majority of the immune‐responsive genes. However, the presence of genes not affected, or only partially affected, in the rel,spz background suggests that other pathways regulate the Drosophila immune response. Consistent with the survival experiments (Figure 1), we found that most of the genes induced by fungal infection are regulated by the Toll pathway after septic injury, without contribution of the Imd pathway, and that the two pathways contribute to the control of many genes induced only by bacterial infection.
Target genes of Toll and Imd pathways
To address which immune reactions are controlled by Relish, Spaetzle, both Relish and Spaetzle or by a still unknown mechanism, we examined the effect of the rel and spz mutations on the expression of selected DIRGs (Table I). Unlike our previous automated analysis, we used a less stringent approach. In Table I, we considered each gene affected by a mutation when we detected at least a 2‐fold change in one time point in the mutant line compared with the corresponding time point in wild‐type flies. As previously shown by northern blot analysis (Lemaitre et al., 1996; Hedengren et al., 2000), we found that most AMP genes are regulated by both Relish and Spaetzle, with the exceptions of attacin D and diptericin A, which are controlled only by Relish. Most of our results correlate with previous analyses. However, in contrast to northern blot analysis, we failed to detect an effect of the single spz mutation on the induction of drosomycin, and of the rel mutation on attacin A activation, suggesting that the cRNA probes from these genes can saturate the oligonucleotide microarray.
Among the genes regulated only by the Imd pathway (affected in rel flies, but not in spz), we found several PGRPs encoding genes (PGRP‐LB, SB1, SD and one PRGP‐like), which play a role in the detection of bacteria (Werner et al., 2000). We also identified three genes encoding enzymes involved in the melanization process (Pale, Punch and Dhpr) and one encoding a putative prophenoloxidase‐activating enzyme (proPO‐AE). Finally, we identified genes coding for an uncharacterized serpin (Sp4), one induced and three repressed serine proteases (see also Table I, part 2), one factor involved in iron metabolism (Zip3), one stress response peptide (TotM) and Imd.
In the group of genes controlled only by the Toll pathway (affected in spz mutants and not in relish), we found most of the up‐regulated genes of the Toll pathway itself (necrotic, spaetzle, Toll, pelle, cactus and Dif). These findings extend previous data showing that Cactus regulates its own expression (Nicolas et al., 1998). In this group, we also found genes encoding two short proteins with significant similarities to the N‐terminal domain of Gram‐negative‐binding proteins (GNBPs), that have been isolated as proteins that bind to β‐1‐3 glucan, a component found on the cell wall of fungi (Kim et al., 2000). Therefore, these two new GNBPs are potential candidates for the recognition protein that activates the Toll pathway in response to fungal infection. The Toll pathway also controls genes encoding three uncharacterized serine proteases, two serpins, one kunitz‐type serine protease inhibitor, three putative proPO‐AEs that may play a role in the melanization reaction and one peroxidase gene that could mediate the production of ROS. Finally, we identified several genes coding for unknown small peptides, including the IM2 family, that are induced by B.bassiana infection and could function as new antifungal effector molecules controlled by the Toll cascade. Spaetzle also regulates several repressed DIRGs (Table I, part 2) including genes encoding serine proteases, serine protease inhibitors and one lysozyme. It has been reported that infection inhibits the expression of cytochrome P450 detoxification genes in vertebrates (Renton, 2000). Interestingly, we found that in Drosophila, the Toll pathway mediates the repression of seven cytochrome P450 genes and of other detoxification enzymes (sodh‐1, CG3699 and CG7322) in response to bacterial infection. The rel mutation used in this study (relE20) abolishes the transcription of relish mRNA (Hedengren et al., 1999); therefore, we limited the analysis of relish expression to the wt, spz and Tl10b lines. Interestingly, we found a partial effect of the spz mutant on relish expression at the 6 h time point.
The genes regulated by the Imd and Toll pathways can be divided into three groups as shown in Table I: group A genes are affected only in rel,spz flies; group B genes are affected in both rel and spz flies; and group C is composed of AMP genes weakly affected in rel but strongly affected in rel,spz flies. Imd and Toll pathways are redundant in the regulation of genes in group A. This group includes genes probably involved in melanization (putative proPO‐AE, yellow f; Cp19), one component of the Toll pathway (dorsal), two components of the JNK pathway (d‐Jun and puc) involved in wound healing (Rämet et al., 2002a) and one putative chitin‐binding lectin (idgf3) that could recognize endogenous chitin at the injury site. Both Imd and Toll pathways affect the genes in groups B and C. This groups include three PGRP genes (PGRP‐SA, SC2 and LC); two complement‐like genes (Tep2 and Tep4) and one complement‐binding receptor gene that could be involved in phagocytosis; one fibrinogen‐like gene potentially involved in coagulation; one gene involved in melanization (Ddc); and one transferrin gene mediating iron sequestration. In addition, we found in this category genes encoding five up‐regulated and two down‐regulated serine proteases, two serine protease inhibitors and several unknown small peptides highly induced by infection (Table I, parts 1 and 2).
Among the DIRGs independent of Imd and Toll pathways (not affected in the double mutant rel,spz), we identified genes encoding a putative binding lectin (Idgf1), a putative coagulation factor (annexin IX), one enzyme potentially involved in melanization (laccase‐like), two homologous small peptides and several serine proteases (one up‐regulated and seven repressed).
Genes responding to microbial infection can be located in co‐regulated genomic clusters
The identification of a large number of DIRGs, coupled with the analysis of the mutations affecting the Imd and Toll pathways, allowed us to examine on a large scale the chromosomal localization of co‐regulated genes. An automated statistical analysis helped us to identify 36 DIRGs significantly clustered in the genome. A few examples of genomic clusters identified through this method are given in Figure 4A. In addition, we found other associations of DIRGs not identified by the automated analysis, which are shown in Figure 4B. Finally, Figure 4C shows an example of DIRGs that, although not associated, are encoded in the same cytological region (spaetzle, Toll, pelle and CG5909 in 97A4‐F4). Some of the clusters include copies of homologous genes (three IM‐2‐like genes; attacins A and B1; three cecropin genes; and Dif and dorsal), whose association can be explained by duplication events. However, we identified several gene clusters whose members do not share sequence similarities.
Interestingly, most of the genes inside each cluster (Figure 4A and B) or in the same genomic region (Figure 4C) share a similar type of regulation. For example, the genes Ady43A and CG11086 are both induced upon infection by the Toll pathway, suggesting that they are targets of the same transcription factors (either Dif or Dorsal) (Figure 4C). In support of this finding, a cluster of three Dorsal optimal binding sites has been identified recently between Ady43A and CG11086 (Markstein et al., 2002).
To identify the target genes of the Toll and Imd pathways in response to microbial infection, we have compared the gene expression programmes induced by septic injury in wild‐type and mutant adult male flies using oligonucleotide microarrays. In parallel, we have monitored the survival rate and the expression level of various AMP genes after infection by various microorganisms. For the Toll pathway, we selected a strong homozygous viable allele of spz (rm7). We observed that the spz, Tl and pll mutations alone or in combination with rel (Figures 1 and 2; data not shown) have similar effects on both the survival rate and pattern of AMP gene expression after microbial infection. These findings suggest that the effects of spz mutation on the transcription programme induced by infection reflect the role of the entire Toll pathway in the immune response. For the Imd pathway, we selected a null viable allele of relish (E20). Similarly to the Toll pathway, previous comparative studies did not reveal any striking difference between mutations in relish and null mutations in the genes encoding the other members of the Imd pathway such as kenny, ird5 and dredd, with the sole exception of mutations in dTAK1, which have a slightly weaker phenotype (Leulier et al., 2000; Rutschmann et al., 2000; Lu et al., 2001; Vidal et al., 2001). Again, these data suggest that the effects of rel mutation on the immune response reflect the role of the whole Imd pathway. However, we cannot exclude other pathways, including Toll, from having a minor role in Relish activation.
Based on known data on AMP gene expression, we predicted redundant functions for the Imd and Toll pathways in the control of some of their target genes; thus, we generated a double mutant rel,spz strain devoid of all Toll and Imd activity. Finally, we extended the microarray analysis to a gain‐of‐function allele of Toll (Tl10b) that is constitutively active even in the absence of infection. The Drosophila lines used in this study are not isogenic, thus some of the changes in the gene expression programmes might arise from the genetic background. In addition, developmental or physiological defects induced by the mutations could also affect the adult expression profile. Tl10b flies, for example, show a melanotic tumour phenotype (Lemaitre et al., 1995). However, spz and rel adults do not show any detectable defect (Hedengren et al., 1999); therefore, we believe that most of the changes in the expression profiles in these mutants reflect the direct or indirect effects of the Toll and Imd pathways on transcriptional reprogramming during the immune response.
The septic injury experiments were performed using a mixture of Gram‐positive and Gram‐negative bacteria. This type of infection activates a wide immune response and allows the simultaneous analysis of several categories of immune‐responsive genes (De Gregorio et al., 2001). However, it has been shown that Toll and Imd pathways are activated selectively by different classes of microorganisms; thus, the use of a bacterial mixture might increase the redundancy of the two pathways in the control of common target genes.
In our previous microarray analysis, we observed a very high correlation with published data: 34 out of 35 genes induced by infection identified by northern blot were also detected as up‐regulated with the microarray approach (see http://www.cnrs‐gif.fr/cgm/immunity/). Here, we found that the effects of mutations in Toll and Imd pathways on most of the AMP‐expressing genes and on several genes expressing regulatory factors (necrotic, cactus and relish) corroborate previous studies using northern blots (Nicolas et al., 1998; Levashina et al., 1999; unpublished data). However, we failed to detect the partial effects of single mutations spz or rel on the induction of a subset of AMP genes, suggesting that some genes expressed at high levels (like AMP genes) have saturated binding to the arrays, preventing accurate measurements.
Toll and Imd control the majority of the Drosophila immune response
The microarray analysis demonstrates that the functions of Toll and Imd pathways in Drosophila immunity can be extended beyond the regulation of AMP genes. The majority of the DIRGs are affected by the mutations in the Toll or Imd pathways (Figure 3). Many of these genes are unknown (see http://www.fruitfly.org/expression/immunity/ for a complete list); others can be assigned to several immune functions (Table I). The susceptibility of the Imd and Toll pathway mutants to different types of microbial infection suggested a control of the antifungal response by the Toll pathway: a major role for the Toll pathway for the response to Gram‐positive bacteria with a minor contribution of Imd, and a predominant role of Imd with a minor contribution of Toll to the resistance against Gram‐negative bacteria (Figure 1). In agreement, microarray analysis shows that the Toll pathway controls most of the late genes induced by fungal infection and cooperates with the Imd pathway for the control of genes implicated in several immune reactions such as coagulation, AMP production, opsonization, iron sequestration and wound healing. Interestingly, defensin, which encodes the most effective antimicrobial peptide directed against Gram‐positive bacteria (Tzou et al., 2002c), is co‐regulated by both the Imd and Toll pathways. Our hierarchical cluster analysis of the expression profiles combining the effect of the mutations after septic injury with the response to fungal infection provides a wealth of information that may help to elucidate the function of some of the uncharacterized DIRGs. Until now, the increased susceptibility to infection of Imd‐ or Toll‐deficient flies has been attributed to the lack of expression of AMP genes, and it has been shown recently that the constitutive expression of single AMP genes in imd;spz double mutant flies can increase the survival rate of some types of bacterial infection (Tzou et al., 2002c). Our finding that the Toll and Imd pathways are the major regulators of the Drosophila immune response now suggests that other immune defence mechanisms might contribute to the increased susceptibility to infection displayed by mutant flies.
Interactions between the Imd and Toll pathways
The interactions between the Toll and Imd pathways are more complex than merely regulating the same target genes. In agreement with northern blot analysis (unpublished data), we show that the transcriptional control of relish in response to infection receives a modest input from the Toll pathway, revealing an additional level of interaction between the two cascades. The activation of Toll may increase the level of Relish to allow a more efficient response to bacterial infection. This finding is in agreement with previous observations showing that in mutants where the Toll pathway is constitutively active (Tl10b), all the antibacterial peptides genes, including diptericin, are induced with more rapid kinetics than in wild‐type flies (Lemaitre et al., 1996). Furthermore, the higher susceptibility to E.coli infection of the rel,spz double mutant compared with the rel single mutants flies indicates that Toll also has a direct, Relish‐independent effect on the resistance to infection by Gram‐negative bacteria (Figure 1A). Northern blot analysis shows that relish induction in response to infection is significantly reduced in dTAK1 and dredd mutants, indicating that the Imd pathway undergoes autoregulation (unpublished results). Interestingly, the Imd pathway can influence the Toll pathway through the control of PGRP‐SA, which encodes a recognition protein essential for the activation of the Toll pathway by Gram‐positive bacteria (Michel et al., 2001). Again, it is interesting to notice that this interaction between the Toll and Imd pathways correlates with the contribution of both pathways to fight infection with Gram‐positive bacteria (Figure 1B and C). Interestingly, all the genes encoding components of the Toll pathway required for both antibacterial and antifungal responses (necrotic, spaetzle, Toll, pelle, cactus and Dif) are not controlled by the Imd pathway and are subjected to autoregulation.
Other pathways controlling the Drosophila immune response
The Rel/NF‐κB proteins Dif, Dorsal and Relish, which are the transactivators induced by the Toll and Imd pathways, bind to the κB sites present in the promoters of target genes, such as AMP genes, regulating their expression. Therefore, the analysis of the promoters of the DIRGs controlled by Toll or Imd pathways could help to identify all the direct NF‐κB targets during infection. However, some of the effects of mutations affecting the Toll or Imd pathways that we monitored by microarray analysis might be mediated by the regulation of other transcription factors or signalling cascades. It has been shown recently in larvae that the Tep1 gene is regulated by the JAK–STAT pathway and can be activated by the Toll pathway, suggesting that Toll can control, at least partially, the JAK–STAT cascade (Lagueux et al., 2000). Here we report that two genes encoding components of the JNK pathway (puc and d‐Jun) are partially regulated by Toll and Imd in response to septic injury.
The presence of DIRGs independent of or only partially dependent on both the Imd and Toll pathways suggests the presence of other signalling cascades activated after septic injury. Potential candidates are MAPK and JAK–STAT pathways. Beside their developmental functions, the MAPK pathways have been implicated in wound healing (JNK) and the stress response (MEKK) (Sluss et al., 1996; Inoue et al., 2001; Rämet et al., 2002a). The JAK–STAT pathway, as we mentioned above, controls the Drosophila complement‐like gene TepI (Lagueux et al., 2000). The stimuli that trigger these cascade are not known and it is not clear if these cascades are activated by exogenous or host factors. Interestingly, in vertebrates, the JAK–STAT pathway is activated by cytokines during the immune response. The microarray analysis of mutants in these pathways might help to reveal their exact contribution to the Drosophila immune response. Our observation that Toll and Imd pathways control most of the DIRGs raises the question of whether these two pathways are the sole signalling cascades directly activated by microbial elictors, while the other signalling pathways are triggered by other stimuli associated with infection such as wound, stress, cytokine‐like factors and Toll and Imd activities.
Co‐regulated genomic clusters
In vertebrates, many genes involved in the immune response are grouped in large chromosomal complexes. The recent completion of the Drosophila genome did not reveal any striking chromosomal organization beside clustering of genes belonging to the same family, probably reflecting recent duplication events (Khush and Lemaitre, 2000). In this study, we observed that some of the genes responding to microbial infection are located in the same cytological region or are associated in transcriptionally co‐regulated genomic clusters. Interestingly, microarray analysis of circadian gene expression in Drosophila has led to the identification of similar clusters of genes (McDonald and Rosbash, 2001). Other microarray analyses might reveal the importance of the genome organization in the definition of adequate transcription programmes in response to a variety of stimuli.
Materials and methods
OregonR flies were used as a wild‐type standard. Exact genotypes of the flies analysed in this study are: spaetzlerm7/spaetzlerm7 (spz); Tl1–RXA, e/Tlr632 (Tl); relishE20, e/relishE20, e (rel); pll7/pll78 (pll); spaetzlerm7, relishE20/spaetzlerm, relishE20 (rel,spz); Tl1‐RXA, relish E20 e/Tlr632, relishE20 (rel,Tl); and Tl10b, e/+ (Tl10b). spaetzlerm7, Tl1‐RXA and relishE20 are strong or null alleles of spz; Tl, and rel; Tlr632 is a thermosensitive allele of Tl with a strong phenotype at 29°C (Lemaitre et al., 1996; Hedengren et al., 1999). rel,spz flies were obtained by recombining spaetzlerm7 and relishE20 on the third chromosome. The alleles Tl1‐RXA and Tlr632 were recombined with relishE20, and the resulting double mutant lines were crossed to generate the line rel,Tl. The Tl10b allele is a gain‐of‐function allele of Toll. Tl10b/TM3 males were crossed to wild‐type female flies, and Tl10b/+ males were subjected to microarray analysis.
For septic injury and natural infection experiments, we used Drosophila adults, aged 3–4 days, at 25°C. Septic injury was produced by pricking the thorax of the flies with a needle previously dipped into a concentrated culture of E.coli, M.luteus and E.faecalis or in a suspension of A.fumigatus spores (Tzou et al., 2002b). Natural infection was initiated by shaking anaesthetized flies in a Petri dish containing a sporulating culture of the entomopathogenic fungus B.bassiana. For survival experiments, 60 flies were infected in the morning and incubated at 29°C (except for E.faecalis infection that was performed at 25°C). For northern blotting and microarray analysis, flies were incubated at 25°C and collected at specific times after infection.
Analysis of mRNA expression using oligonucleotide arrays
Microrray analysis was performed with Affymetrix Drosophila GeneChips using poly(A) RNA from adult males as previously described (De Gregorio et al., 2001). To identify genes that show changes between conditions, t‐tests were performed. Due to the limited number of arrays used, we agglomerated all infected time points and treated them equivalently. We did restrict our analysis to equally represented time points, those from 90 min to 6 h after septic infection. It should be noted that this method of analysis prevents us from observing real differences between genotypes for genes that are particularly dynamic. Five sets of tests were performed for the 400 DIRGs with the following comparisons: wild‐type uninfected samples (n = 5) with wild‐type bacterially infected (n = 12); wild‐type infected (n = 12) with spz infected (n = 6); wild‐type infected (n = 12) with rel infected (n = 6); wild‐type infected (n = 12) with rel,spz infected (n = 6); and wild‐type uninfected (n = 5) with Tl10b uninfected (n = 3). To mitigate false positives, the P‐value for each t‐test was considered significant if it was <0.0025 so that we expect approximately one false positive for each of the five sets of tests.
DIRG gene clusters were identified in the genome by first finding all DIRG genes that were adjacent in the genome. Forty‐three such pairs of DIRGs exist. Using the binomial distribution, it was calculated that a total of four genes (including the pair) within a 16 gene window (seven genes on either side of each pair) was significant at P < 0.05. A total of six gene clusters comprising a total of 36 genes met these criteria.
supplementary data for this paper are available at The EMBO Journal Online.
We thank François Leulier for critical reading of the manuscript. E.D.G. was supported by a Human Frontier Science Program fellowship. The laboratory of B.L. was funded by the Association pour la Recherche contre le Cancer (ARC), the Fondation pour la Recherche Médicale (FRM) and Programme Microbiologie (PRMMIP00). P.S. was supported by an NSF Biocomputing post‐doctoral fellowship.
- Copyright © 2002 European Molecular Biology Organization