To determine the degree of specificity in the response of C. elegans to bacterial infection, we carried out a whole-genome, comparative analysis of worms infected with one Gram-positive and three Gram-negative bacterial pathogens using long-oligo microarrays. We first looked at the response to S. marcescens and found less than a 2% overlap between the genes identified as being up-regulated by S. marcescens in this study (supplementary Table 1a in Additional data file 3) and a previous investigation, which employed a different microarray platform based on nylon cDNA filters with partial genome coverage 8. This underlines the difficulty in making direct comparisons between studies employing different experimental designs.

Studies with C. elegans generally use worms cultured on the standard nematode growth medium (NGM) agar. On the other hand, the Gram-positive bacterium Enterococcus faecalis is most pathogenic when cultured on a rich medium (brain heart infusion (BHI) agar) 15. To eliminate possible effects of the medium on nematode physiology, we wished to carry out all infections on worms grown on NGM agar. We determined that E. faecalis was still pathogenic to C. elegans when grown on NGM agar, if pre-cultured in liquid BHI medium (supplementary Figure 1 in Additional data file 1), and adopted this protocol for our analyses.

Comparing the levels of expression for genes that were up- or down-regulated at a single time point by each individual bacterial pathogen (S. marcescens; E. faecalis; Erwinia carotovora; Photorhabdus luminescens), we observed expression profiles that were characteristically unique, or 'pathogen-specific signatures'. For example, the majority of genes with expression levels altered in one direction following infection by P. luminescens were either unchanged or responded differently in infections with other pathogens (Figure 1a,b; supplementary Table 1a,b in Additional data file 3). Thus, 24 h post-infection, C. elegans is clearly capable of mounting a response that is principally different for each of the pathogens used in this study. From non-redundant groups of 2,171 genes up-regulated and 2,025 genes down-regulated after infection with at least one pathogen, only 254 and 266 genes, respectively, were identified to be commonly regulated by more than one pathogen (supplementary Table 1c in Additional data file 3). These comparatively small numbers reinforce the notion of pathogen-specific responses, while at the same time suggesting that host responses to different pathogens have common facets. To examine this further, we performed clustering analyses with both the commonly up- and down-regulated genes. In both cases, groupings composed of genes responding similarly to different pathogens were observed (Figure 1c). Surprisingly, the response to the Gram-positive bacterium, E. faecalis, overlapped to a greater extent with those provoked by the Gram-negative bacteria P. luminescens and E. carotovora than did the response provoked by a third Gram-negative bacterium, S. marcescens. Thus, for example, one grouping was identified for genes with altered expression following infection with the first three bacteria, to the exclusion of S. marcescens (Figure 1d). Overall, highest similarity existed between the genes whose expression was altered following infection with E. carotovora and P. luminescens.

The large numbers of genes identified as being transcriptionally regulated upon infection represents a challenge for meaningful interpretation. In our study this problem was further compounded by the inclusion of multiple pathogens, which as a consequence, required the analysis of diverse datasets. The use of Gene Class Testing 22 to identify functional associations can, however, help in the identification of biologically relevant themes. We therefore used the freely available Expression Analysis Systematic Explorer (EASE) 23 to identify gene classes significantly over-represented among genes regulated as a consequence of infection. In our analyses, we looked at gene classes derived using Gene Ontology, euKaryotic Orthologous Groups and functional information from published experiments using C. elegans (see Materials and methods). Biological themes were formed via the grouping of gene classes in an ad hoc fashion, with all members of a group having similar biological functions. For example, the 'infection-related response' class includes genes described in published studies as being up- or down-regulated by infection, together with any structurally homologous genes.

With EASE we identified two major groupings of gene classes. The first, termed 'pathogen-shared', is composed of gene classes identified across infections with different pathogens (Figure 2a; supplementary Table 2a in Additional data file 3). These include classes shared by genes with similar expression profiles in E. faecalis, E. carotovora and P. luminescens infections and that can be further associated with proteolysis, cell death, insulin signaling and stress responses. Other gene classes shared by E. faecalis and P. luminescens include lysozymes, genes expressed in the intestine and genes implicated in the response to infection with Microbacterium nematophilum, a Gram-positive nematode-specific pathogen 9. There was similarly an over-representation of genes up-regulated following infections with E. carotovora and P. luminescens that are associated with infection by another Gram-negative pathogen, P. aeruginosa 11. A second grouping defined the 'pathogen-specific' responses (Figure 2b; supplementary Table 2b in Additional data file 3). For example, only E. faecalis infection was associated with a significant down-regulation of hormone receptors, while P. luminescens infection involved a significant elevation of the proportion of genes described to be under the control of p38 MAPK and TGF-β signaling pathways 1024. Biological themes associated with host response to adverse conditions, including infection, can be found within both the pathogen-specific and pathogen-shared groupings (Figure 2). Thus, as further discussed below, clustering analysis of gene expression and gene class testing are both consistent with the notion that the response of C. elegans to infection can be defined by two biologically relevant signatures, one being pathogen-shared and the other, pathogen-specific.

While fold change measurements are conceptually useful when performing exploratory analyses, they lack known and controllable long-range error rates 22. We therefore performed complementary analyses in which exploratory findings using fold change-derived data were combined with results obtained using two established statistical tools, MAANOVA and BRB-ArrayTools (see Materials and methods). With the two exploratory analyses, a grouping of host-responses observed following infection with E. carotovora, E. faecalis and P. luminescens was the most consistent (Figures 1c and 2a). We therefore used MAANOVA and BRB-ArrayTools on microarray data obtained with these three pathogens to investigate further the nature of this apparent pathogen-shared host-response. We identified a total of 22 high-confidence genes with significant differences in expression between control animals and animals infected with the three pathogens (Table 1; supplementary Table 3a in Additional data file 3). Prominent among these 'common response genes' is lys-1, which was one of the first infection-inducible genes to be identified in C. elegans 8. Following the demonstration that it was up-regulated by S. marcescens infection, lys-1 has also been shown to be part of the response of the worm to P. aeruginosa 11. The list also includes a gene that encodes a lipase, a class of protein important in the response to S. marcescens 8 and M. nematophilum 9, as well as a saposin-encoding gene. All the corresponding proteins are expected to have antimicrobial activity and, therefore, to contribute directly to defense 2526. Other genes correspond to a C-type lectin (clec-63), a putative LPS-binding protein (F44A2.3), and proteins containing Complement Uegf Bmp1 (CUB) and von Willebrand Factor (vWF) domains and vWF, epidermal growth factor (EGF) and lectin domains, respectively; all of these could be involved in pathogen recognition 2526. Members of the largest class of genes, however, encode aspartyl proteases not previously associated with the response to infection in C. elegans.

Neither up- nor down-regulated genes exhibited any substantial genomic clustering of the type described for genes involved in the response to M. nematophilum infection 9. With regards to down-regulated genes within the pathogen-shared response identified in this study, they are all seemingly metabolism-related; a similar phenomenon has been previously described in worms infected with M. nematophilum 9.

To validate these results, we examined in more detail the regulation of three asp genes encoding aspartyl proteases, as well as a C-lectin, encoded by clec-63, using quantitative real time-PCR (qRT-PCR). Since only a small number of common response genes was identified during statistical testing, we also looked at the expression of two other clec genes, one being clec-65, the genomic neighbor of clec-63, and the other clec-67, reported to be induced by M. nematophilum 9. At 24 h, all six genes showed a marked up-regulation following infection by E. faecalis, E. carotovora and P. luminescens, whereas they did not show a substantial change in expression following S. marcescens infection (Figure 3a). We hypothesized that this result could be a consequence of the different pathogenicities of the bacteria. To investigate this, we carried out a time course study over a period of five days, using qRT-PCR to follow relative expression levels of asp-3, asp-6 and clec-63 in worms infected by S. marcescens. The expression levels of these three genes indeed increased over this period (Figure 3b), suggesting that their induction is linked to pathogenesis more than to pathogen recognition.

In contrast to the bacterial pathogens used in this study that infect C. elegans via the intestine, the fungus Drechmeria coniospora infects nematodes via the cuticle 27. A comparison of the common response genes with those having an altered expression following infection with D. coniospora, determined under similar experimental conditions to those used in this study (Pujol et al., submitted), showed absolutely no overlap (results not shown). This clear distinction between bacterial and fungal infection was unexpected since we had previously reported, based on our results using cDNA microarrays, that the antimicrobial peptide gene nlp-29 was induced upon infection both by S. marcescens and D. coniospora 27. This gene appeared not to be up-regulated, however, by any of the bacterial pathogens used in this study, including S. marcescens. When we assayed the level of nlp-29 expression in worms infected by the different pathogens using qRT-PCR, we found that only D. coniospora induced a substantial increase (Figure 3c). We recently found that nlp-29 is induced under conditions of high osmolarity (Pujol et al., submitted), including when plates used for culturing worms become drier after a few days storage. The age of plates was not a variable that was previously controlled, and we now believe this to be the most likely reason for having erroneously identified nlp-29 as a gene induced by S. marcescens infection. These results underline the fact that C. elegans mounts distinct responses to bacterial and fungal infection.

The response of C. elegans to infection by S. marcescens and P. aeruginosa involves predominantly genes expressed in the intestine 811. Information regarding the expression patterns for 19 of the 22 common response genes differentially regulated after infections with E. faecalis, E. carotovora and P. luminescens is available (supplementary Table 3a in Additional data file 3). Of these, 16 are expressed in the intestine of the adult animal. Examination of their proximal promoter regions using BioProspector 28 revealed GATA motifs in 43% of these genes (supplementary Table 3a in Additional data file 3), consistent with similar findings from a recent study 11. Two other genes, npp-13 and K06G5.1, are known to be expressed in the gonad. By in situ hybridization, the remaining gene, F44A2.3, is reported to show weak but specific expression at the vulva and in the head. This gene also attracted our attention as it encodes a protein containing a lipopolysaccharide-binding protein (LBP)/bactericidal permeability-increasing protein (BPI)/cholesteryl ester transfer protein carboxy-terminal domain (Pfam accession number PF02886), associated with bacterial recognition or killing in many species 2930. We determined its expression pattern by generating transgenic strains carrying green fluorescence protein (GFP) under the control of the F44A2.3 promoter. We observed high levels of constitutive GFP expression in the pre-anal, vulval, hypodermal, glial amphid socket and excretory duct cells of the adult animal (Figure 4a-i). Upon infection of worms carrying the reporter gene with E. carotovora or P. luminescens, there was no perceptible change in the level of GFP expression at 24, 48 or 72 h post-infection (results not shown). Similarly, these two pathogens caused no discernable induction of GFP expression at any time up till 72 h post-infection in strains carrying 5 other transcriptional reporter genes (asp-5 and -6, clec-63, -65 and -67; results not shown). Thus, based on the genes tested, we were unable to identify robust in vivo reporters for the response to bacterial infection. The cells that expressed pF44A2.3::GFP are in privileged sites, in contact with the external environment, hinting at a potential front-line role for F44A2.3 in pathogen recognition. We addressed any potential role in resistance to infection by inactivating its expression by RNAi, but did not see any significant effect on survival (supplementary Figure 2 in Additional data file 1).

In contrast to the reporter genes listed above, we observed a clear and reproducible induction of expression of the asp-3 and -4 reporter genes. In the absence of infection, virtually no GFP was detectable, while after exposure to E. carotovora or P. luminescens there was an accumulation of GFP within large vacuoles formed in the intestine (Figure 4j-k). We observed a qualitatively similar induction of reporter gene expression following infection with E. faecalis but of a lower magnitude (results not shown).

When the asp-4::GFP reporter was transferred by mating into pmk-1(km25) or dbl-1(nk3) mutant backgrounds, we observed an induction of GFP expression following infection with E. carotovora that was similar to that seen in wild-type worms (results not shown). The two mutants respectively affect the p38 MAPK and TGF-β pathways, important for resistance to bacterial infection. Thus, these results suggest that infection-induced expression of ASP-4 is independent of the two pathways.

Both asp-3 and -4 have been specifically associated with the execution of necrotic cell death in C. elegans 17. Indeed, inspection of worms during infection revealed the frequent incidence of necrotic cell death in the intestine, which is manifested by the vacuole-like appearance of cells (Figure 4j), not seen within the intestine of healthy animals. These dramatically swollen cells have distorted nuclei restricted in the periphery, a most prominent characteristic of necrotic cell death. Preliminary observations suggested that infection under different culture temperatures (25°C and 20°C) progresses similarly in terms of symptoms and asp::GFP reporter gene expression, except that at 25°C everything was more rapid. In subsequent experiments, we therefore conducted infections at 20°C to increase the temporal resolution. The appearance of necrosis follows the spatiotemporal progression of infection. The first tissue affected is the intestine, where vacuolated cells were observed around 24 h post-infection. After the second day of infection, the epidermis and the gonad become severely distorted and displayed similar necrotic vacuoles. This pattern of necrotic death, observed following infection with different pathogens, could be part of an inducible defense mechanism contributing to host survival, or a deleterious consequence of infection. To differentiate between these two possibilities, we assayed the resistance to infection of two necrosis-deficient C. elegans mutants, vha-12(n2915) and unc-32(e189), that both affect V-ATPase activity 3132. The two mutants showed enhanced survival, relative to wild-type N2 worms in infections with E. carotovora (Figure 5a) and P. luminescens (Figure 5b). Given that these mutants display abnormal pharyngeal pumping, we were concerned that resistance might be the consequence of a reduced bacterial load. We therefore directly assayed the number of viable bacteria within worm intestines at 24 h post-infection. With E. carotovora, there was no difference between infected wild-type and mutant animals (Figure 5c), while for P. luminescens, unc-32 animals had a higher bacterial load (Figure 5d). Therefore, differences in bacterial accumulation are not correlated with resistance of the two mutants to infection. Certain mutants of the insulin/insulin growth factor signaling pathway, such as daf-2, exhibit increased pathogen resistance and longevity 33. To examine whether vha-12 and unc-32 are more infection-resistant due to general effects in survival and ageing, we measured the longevity of these mutants on non-pathogenic E. coli and found that they had similar lifespans to wild-type animals (Figure 5e), consistent with previous findings 34. We also observed that the induction of asp-4::GFP by E. carotovora and P. luminescens was unchanged in a vha-12 mutant background (supplementary Figure 3 in Additional data file 1). Thus, mutants that have a defect in intracellular organelle acidification are necrosis-deficient and exhibit a specific increase in their resistance to infection that appears to be independent of asp-4 activity.