The genetics of pathogen avoidance in Caenorhabditis elegans


  • Hinrich Schulenburg,

    Corresponding author
    1. Department of Animal Evolutionary Ecology, University of Tübingen, Auf der Morgenstelle 28, 72076 Tübingen, Germany.
      *E-mail; Tel. (+49) 7071 2975342; Fax (+49) 7071 295634.
    Search for more papers by this author
  • Jonathan J. Ewbank

    1. Centre d'Immunologie de Marseille-Luminy, Université de la Méditerranée, Case 906, 13288 Marseille cedex 9, France.
    2. INSERM, U631, 13288 Marseille, France.
    3. CNRS, UMR6102, 13288 Marseille, France.
    Search for more papers by this author

*E-mail; Tel. (+49) 7071 2975342; Fax (+49) 7071 295634.


Much attention is rightly focused on how microbes cause disease, but they can also affect other aspects of host physiology, including behaviour. Indeed, pathogen avoidance behaviours are seen across animal taxa and are probably of major importance in nature. Here, we review what is known about the molecular genetics underlying pathogen avoidance in the nematode Caenorhabditis elegans. In its natural environment, the soil, this animal feeds on microbes and is continuously exposed to a diverse mix of microorganisms. Nematodes that develop efficient behavioural responses that enhance their attraction to sources of food and avoidance of pathogens will have an evolutionary advantage. C. elegans can specifically detect natural products of bacteria, including surfactants (such as serrawettin) and acylated homoserine lactone autoinducers, and it can learn to avoid pathogenic species. To date, several distinct mechanisms have been shown to be involved in pathogen avoidance. They are based on G protein-like, insulin-like and neuronal serotonin signalling. We discuss recent findings on the mechanisms of pathogen recognition in C. elegans, the relationship between alternative behavioural defences and also between these and other life-history traits. We propose that the selective pressure associated with avoidance behaviours influence both pathogen and host evolution.

The importance of pathogen avoidance behaviours

When an animal is challenged with an infectious microbe it faces two main options. It can try to fight the threat by immediately activating physiological and cellular defences. These often include the expression of antimicrobial peptides or the mobilization of phagocytosing cells. Although generally extremely efficacious, such innate immune mechanisms are costly in terms of energy, and can cause self-damage (autoimmunity or immunopathology) (Schmid-Hempel, 2003). The alternative option is to avoid the infectious agent, for example, by moving away from pathogen-rich compounds (Moore, 2002). Although this may come with its own direct or indirect costs (e.g. energy needed for enhanced movement or reduced access to resources respectively), the second option has two major advantages: it minimizes the infection risk and economizes the animal's resources. These avoidance behaviours are likely to be of major importance in the natural ecological niches of many animal species, affecting the evolution of foraging strategies, habitat choice, social interactions and also mate choice (Moore, 2002). An increasing body of data supports a similar role in humans. Examples include innate disgust of microbe-infested compounds, hygienic behaviour and possibly xenophobia (Curtis and Biran, 2001; Faulkner et al., 2004).

The molecular genetics of pathogen avoidance behaviours have been explored in only a few organisms. Studies in mice, for example, have focused on the avoidance of parasitized con-specifics. This social aversive behaviour was found to depend on oxytocin and oestrogen signalling, which appear to interact in the hypothalamus and the amygdala within the brain (Kavaliers et al., 2005). Currently, however, the most detailed data are available for the nematode Caenorhabditis elegans. This free-living worm feeds on microorganisms and is particularly common in microbe-rich environments, such as compost or rotting fruits (Barriere and Felix, 2007; Hodgkin and Doniach, 1997). These environments may also contain pathogens, which usually infect worms upon oral uptake during feeding, although alternative infection routes are known (reviewed in Schulenburg et al., 2004). When C. elegans is infected, a complex innate immune response is triggered, involving interconnected signalling cascades such as the p38 MAPK and TGF-β pathways (see review by Ewbank, 2006). In addition, C. elegans appears to have evolved behavioural defences against harmful microbes, and thus it should provide a powerful model for the genetic dissection of pathogen avoidance.

The first step of behavioural defence consists of the recognition of pathogens, which could involve chemosensation and/or mechanosensation. C. elegans can discern differences in the physical properties of what it eats (Sawin et al., 2000), and thereby appears to discriminate between different bacterial species (Shtonda and Avery, 2006). For the time being, there is no evidence that this capacity is used to distinguish between pathogenic and non-pathogenic bacteria, and for the remainder of the review, we will concentrate on chemosensory discrimination of bacteria. C. elegans has a simple and well-characterized nervous system consisting of 302 neurons, including 32 presumed chemosensory neurons with sensory cilia exposed to the environment (reviewed in Bargmann, 2006). Most are found in the two amphid organs in the head; in addition to 11 chemosensory neurons they each contain one thermosensory neuron (Fig. 1). Additional chemosensory neurons are found in the six inner labial organs in the head (Fig. 1) and the two phasmid organs in the tail. Perception of chemical compounds involves G protein-coupled chemoreceptors and a G protein-signalling network. The chemosensory system mediates specific behaviours to a broad range of compounds (Bargmann, 2006).

Figure 1.

The head olfactory system of C. elegans.
A. Scanning electron micrograph of the head of an adult C. elegans hermaphrodite, showing the opening of the buccal cavity (towards the top). The filled arrowhead indicates one of the two amphid openings, the open arrowhead one of the six inner labial organ endings in the apex of each lip. Image very kindly provided by Jürgen Berger, MPI for Developmental Biology, Tübingen.
B. Schematic overview of the chemosensory organs in the head. Mouth opening is towards the top. Lateral view of only one-half of the symmetrically organized amphid organs (dark yellow; total number: two) and the inner labial organs (red; total number: six). The neuronal ganglia are given as ovals, including ADF, ADL, AFD, ASE, ASG, ASH, ASI, ASJ, ASK, AWA, AWB, AWC for the amphid organ, and IL1, IL1D, IL1V, IL2, IL2D, IL2V for the inner labial organs. The dendrites of one amphid or one inner labial organ are, respectively, embedded in a tube-like sheath cell and one socket cell. The amphid organs are open to the outside at the base of the lip, whereas the inner labial organs end in the apex of each lip. The pharynx with its two bulbi is given in grey. More details can be found at

Distinct behavioural responses have long been noted towards different microbes such as Acinetobacter calcoaceticus, Escherichia coli, Serratia liquefaciens or different Bacillus, Enterobacter and Pseudomonas species (Andrew and Nicholas, 1976; Grewal and Wright, 1992). Most bacteria attract worms from a distance, suggesting the existence of volatile cues. C. elegans is exquisitely sensitive to oxygen concentrations; on an agar plate they can even sense local dips provoked by growing bacteria (Gray et al., 2004). Thus part of the attraction may be linked to the sensing of metabolically active bacteria that are likely to be a good food source. However, oxygen is certainly not the only signal, and some species, including A. calcoaceticus, are still somewhat attractive, even when dead (Grewal and Wright, 1992). In fact, many additional diffusible compounds, such as cAMP, also play a role in attracting C. elegans. Not surprisingly, many of the identified attractants are natural products of bacterial metabolism (Bargmann, 2006).

On the other hand, microbes can also repel C. elegans. Avoidance may be due to low food quality of the bacterium, as shown for Bacillus megaterium (Andrew and Nicholas, 1976; Shtonda and Avery, 2006). Importantly, however, the aversive response can be specifically directed against the pathogenic strains of different bacterial species, including Bacillus thuringiensis (Schulenburg and Müller, 2004; Hasshoff et al., 2007), Microbacterium nematophilum (Yook and Hodgkin, 2007), Photorhabdus luminescens (Sicard et al., 2007), Pseudomonas aeruginosa (Zhang et al., 2005; Beale et al., 2006; Laws et al., 2006; Yook and Hodgkin, 2007) and Serratia marcescens (Pujol et al., 2001; Zhang et al., 2005; Pradel et al., 2007). In this review, we will discuss pathogen avoidance in terms of: (i) immediate physical evasion of pathogens, (ii) learning of pathogen evasion and (iii) physical evasion and reduced feeding as part of a general stress response. We will explore the relationship between the underlying mechanisms, the specificity of pathogen recognition in C. elegans, the role of pathogen avoidance in nematode life history and its potential impact on host and pathogen evolution.

Physical evasion of pathogens

Caenorhabditis elegans is attracted from a distance by the Gram-negative bacterium S. marcescens, but when it comes into contact with certain strains, it exhibits an avoidance behaviour and crawls away (Pujol et al., 2001). A screen of a transposon-induced mutant library of the repellent S. marcescens strain Db10 indicated that avoidance depended on the expression of the cyclic lipodepsipentapeptide serrawettin W2 – a secreted surfactant, which is produced in large quantities by some S. marcescens strains. Serrawettins are required for swarming motility, an essential attribute of these bacteria as it allows movement from soil particle to soil particle. As described below, serrawettin W2 is directly sensed as an aversive stimulus by C. elegans (Pradel et al., 2007).

The avoidance response to serrawettin W2 depends primarily on a single pair of chemosensory neurons called AWB, which are part of the amphid organs. Although a specific receptor required for serrawettin detection has not, as yet, been identified, it is likely to be one of the more than 1000 G protein-coupled receptors encoded in the C. elegans genome. Thus, avoidance is compromised by mutations in the Gi-like protein ODR-3, the G protein receptor kinase GRK-2 and the TAX-4/TAX-2 cGMP-gated channel (Figs 1 and 2A). Animals in which the AWB neurons are genetically ablated show a decreased avoidance of S. marcescens, rather than no response at all, and do retain some capacity to distinguish between Db10 and a serrawettin W2-deficient (swrA mutant) strain (Pradel et al., 2007). Interestingly, the biochemically distinct serrawettin molecules W1 and W3 from other S. marcescens strains induce an avoidance response that clearly differs from that against serrawettin W2 (Pradel et al., 2007).

Figure 2.

A schematic overview of pathogen avoidance in C. elegans.
A. Physical pathogen evasion is controlled by unknown G protein-coupled receptors (GPCR), the Gi-like protein ODR-3, the cGMP-gated TAX-2/TAX-4 channel, the G protein receptor kinase GRK-2 and the Toll-like receptor TOL-1; it involves signalling in AWB and possibly other neurons.
B. Aversive learning is regulated by serotonin (5-HT) signalling in ADF neurons and thus it depends on genes involved in its biosynthesis (CAT-1, CAT-4, TPH-1); signal transduction involves the serotonin-gated chloride channel MOD-1, possibly in interneurons AIB, AIY or AIZ.
C. Reduced insulin-like signalling (DAF-2 and AGE-1) and thus activation of the DAF-16 transcription factor mediate pathogen-induced feeding inhibition and escape as part of a stress response.

Physical evasion of S. marcescens and its surfactant serrawettin also involves a second, possibly non-independent mechanism, which is based on the only Toll-like receptor gene (tol-1) in C. elegans. Although tol-1 mutants are strongly repelled by purified serrawettin W2, they have a reduced capacity to discriminate between wild-type and swrA mutant strains. Thus, tol-1 may have a neuronal function in the integration of attractive and repulsive stimuli from S. marcescens (Pujol et al., 2001; Pradel et al., 2007). This contrasts with the role of Toll-like receptors in controlling antimicrobial gene expression via activation of Rel-family transcription factors in a wide range of vertebrate and invertebrate species.

Chemosensation is also involved in the avoidance of other pathogens by C. elegans. In particular, the repulsion seen with pathogenic M. nematophilum and two Pseudomonas strains was suppressed by mutations in the genes tax-4 and tax-2 (Yook and Hodgkin, 2007). As discussed further below, recognition of P. aeruginosa may involve detection of acylated homoserine lactone autoinducers.

Learning of pathogen avoidance

Nematodes appear to be assiduous learners and exposure to chemical cues in early life is remembered into adulthood (Remy and Hobert, 2005). They are also able to adjust their foraging behaviour based on previous experience (Shtonda and Avery, 2006). A similar learning response towards pathogens should be evolutionary advantageous, because it would enhance rapid avoidance of the repertoire of pathogen species to which worms are exposed. This possibility was explored using pathogenic P. aeruginosa PA14 and S. marcescens ATCC 13880 (Zhang et al., 2005). Avoidance is significantly induced if animals are raised in the presence of the respective pathogenic strains but not avirulent bacteria. Aversive learning is rapid, can be induced within a few hours and is highly specific to the pathogen species experienced during development (Zhang et al., 2005). A functional genetic analysis showed that mutations in genes required for serotonin synthesis (cat-1, cat-4 and particularly tph-1) and also the gene mod-1, which encodes a serotonin-gated chloride channel, impaired the learning response. This response could be restored in the serotonin-deficient tph-1 mutant by ectopic expression of tph-1 in the serotonergic ADF neurons, but not the other pair of serotonergic head neurons, NSM (Zhang et al., 2005). Furthermore, exposure to pathogens increased tph-1 expression and consequently serotonin levels in the ADF neurons while exogenous serotonin added to bacterial lawns enhanced the learning response (Zhang et al., 2005). Thus, this comprehensive data set convincingly demonstrates that aversive olfactory learning directed against pathogens is mediated by an enhanced serotonin signalling in the ADF sensory neurons (Figs 1 and 2B).

The aversive learning response was subsequently confirmed in two independent studies. In one study, avoidance of pathogenic P. aeruginosa PA14 was enhanced upon prior exposure to the pathogen, but not when UV-killed or chloramphenicol growth-inhibited pathogens were used. This suggests that aversive learning requires live bacteria (Laws et al., 2006). The other study tested the role of the P. aeruginosa acylated homoserine lactone autoinducers, which mediate quorum sensing in diverse bacteria and thereby regulate production of virulence factors (Beale et al., 2006). Generally, different autoinducers attracted worms. Nematodes were, however, repelled by one of these autoinducers (3O-C8 homoserine lactone) if they were pre-exposed to the pathogenic P. aeruginosa PAO1 strain that produces exactly the same autoinducer. Repulsion was not observed upon pre-exposure to a PAO1-derived strain deficient in autoinducer synthesis. Thus, autoinducers can serve as a specific signal to condition pathogen avoidance behaviour in C. elegans. This response depended on the serotonin-gated chloride channel MOD-1, again highlighting the role of serotonin signalling in aversive learning (Beale et al., 2006).

Pathogen avoidance behaviour as part of a general stress response

Evolutionary theory suggests that different costly defence mechanisms are traded off against each other; i.e. lineages with high physiological immunity would show little ability to avoid pathogens and vice versa. This prediction has been tested by analysing natural variation in C. elegans defences against B. thuringiensis. In contrast to expectations, however, the study revealed a positive correlation between resistance and behavioural defence (Schulenburg and Müller, 2004). One possible explanation for this pattern is that both traits are linked genetically as part of a general stress response. In C. elegans, the insulin-like signalling pathway is involved in the response to diverse environmental stressors. It consists of a series of highly conserved proteins, including the insulin-like receptor DAF-2, which is a transmembrane tyrosine kinase, and the FOXO/forkhead family transcription factor DAF-16. Activated DAF-2 signals through a kinase cascade, which also involves the phosphatidylinositol-3-OH kinase, AGE-1. The ultimate result is phosphorylation and thus cytoplasmic retention of DAF-16. Nuclear translocation of DAF-16 is possible when the pathway is inactive or downregulated. In this case, nuclear DAF-16 regulates the expression of a large diversity of genes, including those involved in detoxification, the heat shock response and also those with antimicrobial properties (Murphy et al., 2003; McElwee et al., 2004). Indeed, under standard laboratory conditions, daf-2 mutants have an increased resistance to many pathogens (Garsin et al., 2003; Troemel et al., 2006).

The analysis of different insulin-signalling mutants demonstrated that this pathway also mediates pathogen resistance in environments where at least some of the behavioural defences like physical evasion are less pertinent (e.g. in liquid medium or wormballs) (Hasshoff et al., 2007). At the same time, insulin-like signalling affects two distinct behavioural defences. A reduced-function mutation of the daf-2 insulin-like receptor led to significantly increased physical evasion and a significantly reduced oral uptake of pathogenic strains of the Gram-positive bacterium B. thuringiensis (Hasshoff et al., 2007) (Fig. 2C). The same mutation also enhanced longevity as well as pathogen resistance. Both behavioural responses were particularly pronounced shortly after pathogen exposure, indicating that they constitute a fast defence reaction (Hasshoff et al., 2007). Interestingly, wild-type worms did not display similar behavioural responses. This may suggest that insulin-like signalling only contributes to pathogen avoidance after its activity is modulated by some other factor, e.g. heat stress or starvation. Consistent with this suggestion, the nuclear translocation of the transcription factor DAF-16 was not found in previous studies to be influenced by pathogenic P. aeruginosa (Shapira et al., 2006; Troemel et al., 2006). Similar tests need to be performed for B. thuringiensis. Conversely, it still remains to be determined whether daf-2-mediated avoidance behaviours extend to pathogens other than B. thuringiensis.

Taken together, insulin-like signalling appears to provide a combined defence response against certain pathogens without a trade-off between individual mechanisms (e.g. physiological versus behavioural defences). This combined strategy seems to be part of a general stress response that serves to protect the nematode in an unfavourable environment.

Are the behavioural defence mechanisms related?

To date, it is unknown whether the characterized mechanisms are independent or interact with each other to mediate behavioural defence. For instance, it is highly likely that pathogens induce a general stress response. Hence, the mechanism described above could perhaps contribute to the other patterns of behavioural defence. Similarly, a certain degree of learning may have influenced avoidance of pathogenic B. thuringiensis in the above example. Interestingly, the evasion responses appear to converge on the chemosensation machinery. Avoidance of S. marcescens and aversive learning were shown to rely on chemosensory neurons – in the first case primarily involving AWB and in the second case ADF (Zhang et al., 2005; Pradel et al., 2007). Furthermore, in adults, as judged by reporter gene expression, the tol-1 Toll-like receptor gene is expressed in the nervous system: the four URY neurons, which resemble sensory neurons, and two neurons in the retrovesicular ganglion (Pujol et al., 2001; Pradel et al., 2007). Similarly, although insulin-like signalling can act in a cell autonomous manner in the intestine, there is also a clear contribution from the nervous system, including chemosensory neurons (reviewed in Baumeister et al., 2006).

Intriguingly, two recent studies provide evidence for a twofold link between serotonin and insulin-like signalling in chemosensory neurons. In the one case, the stress-dependent nuclear translocation and thus activation of DAF-16 was enhanced by serotonin release in NSM neurons and suppressed by serotonin release in ADF neurons (Liang et al., 2006). As a consequence, serotonin in ADF neurons may have opposing effects on behavioural responses, as it appears to decrease DAF-16-mediated defences while boosting aversive learning (Zhang et al., 2005). In the second study, serotonin synthesis in ADF neurons was shown to be influenced by DAF-16 activity (Estevez et al., 2006). Consistent with these findings, serotonin regulates not only nematode locomotion (Zhang et al., 2005; Dernovici et al., 2007) but also pharyngeal muscle activity (Niacaris and Avery, 2003). Therefore, it may mediate the reduced feeding response observed in the insulin-signalling pathway mutants. Furthermore, the sensory Gi-like protein ODR-3 was shown to act via the transcription factor DAF-16 to extend life span (Lans and Jansen, 2007), thus providing another direct link between chemosensation and the insulin-like pathway. Nevertheless, the neuronal integration of the signals that underlie pathogen avoidance warrants further research. Full understanding of the relationship of the different mechanisms also requires a more detailed dissection of the exact time-course of sensory and locomotory processes, e.g. using the methodology developed by Shtonda and Avery (2006).

How are pathogens recognized and how specific is recognition?

While C. elegans does employ non-chemosensory mechanisms to avoid some bacteria because they represent bad food (Shtonda and Avery, 2006), worms appear to be able to discern specifically pathogenic from non-pathogenic bacteria (Pujol et al., 2001; Zhang et al., 2005; Beale et al., 2006; Laws et al., 2006; Hasshoff et al., 2007; Pradel et al., 2007; Sicard et al., 2007). Two alternative non-exclusive mechanisms may explain this discriminatory capacity. On the one hand, C. elegans may deduce the presence of harmful microbes from the cellular damage or stress that they provoke. However, in the case of S. marcescens, mutants that are strongly attenuated, such as 20C2 (Kurz et al., 2003), still repel worms and conversely swrA mutants that are not so repellent are as pathogenic as Db10 (Pradel et al., 2007). Thus, C. elegans appears to detect pathogens directly, presumably by binding pathogen-associated molecules. To date, the precise underlying mechanism is unknown. The results on S. marcescens avoidance implicate G protein-coupled chemoreceptors (> 1000 genes). Other candidate pathogen recognition receptors include proteins with leucine-rich repeat domains, the numerous (c. 250) C-type lectin domain-containing proteins (Schulenburg et al., 2004), or the very large family (> 500 genes) of F-box domain proteins that encode ubiquitin-dependent proteasome adapters (Thomas, 2006).

The high level of specificity seen in the capacity of C. elegans to distinguish between different serrawettin molecules (Pradel et al., 2007) came as something of a surprise, because invertebrate responses towards pathogens are generally thought to be rather unspecific. Recent work in two crustacean species revealed, however, a similarly high level of immune specificity that allows the discrimination of different strains of the same parasite species (Kurtz and Franz, 2003; Little et al., 2003). New molecular genetic data additionally point to potential mechanisms for specificity in invertebrates, which may involve the diversification of the Dscam immunoglobulin receptor via alternative splicing in arthropods and also, more generally, synergistic interactions among different components of invertebrate immune systems (Schulenburg et al., 2007).

How do behavioural defences integrate into nematode life history?

Full appreciation of the molecular genetics of avoidance behaviours is only possible by taking into account the associations with other life-history demands, because these are likely to determine the underlying signalling hierarchies and the connection with other physiological processes that might at first sight appear unrelated. Indeed, avoidance behaviours are expected to be modulated dynamically depending on both external and also internal stimuli in order to economize on the worm's energetic resources and increase fitness (i.e. reproductive output). Copper homeostasis in the fruit fly Drosophila melanogaster may serve as an example for avoidance behaviour being determined by the complex integration of external and internal cues. Copper is an essential trace element but is toxic at high concentrations. If flies are well fed and have adequate copper stocks, then they avoid it, but if their reserves are depleted, this response is dampened (Balamurugan et al., 2007).

In C. elegans, the insulin-like pathway may provide a link between alternative defences as part of a general stress response (Baumeister et al., 2006; Hasshoff et al., 2007). The stress response is expressed when the activity of the pathway is downregulated. Activation of the pathway on the other hand leads to high reproductive rates via changes in the worm's metabolism (Gems et al., 1998). Hence, insulin-like signalling may integrate information on food availability, presence of pathogens and noxious substances, unfavourable temperatures, and also the organism's energy state to mediate between alternative life-history strategies – one devoted to development and reproduction and the other to defence against all kinds of insults.

The chemosensation machinery is obviously at the heart of integrating biotic and abiotic environmental signals. For example, serotonin levels in ADF neurons also act as a convergence point for the regulation of hyperoxia avoidance (Chang et al., 2006) and the chemosensory neuron AWB required for the response to serrawettin W2 is known to mediate aversive behaviours against chemical stimuli such as octanol (Bargmann, 2006). These responses are likely regulated by G protein-like signalling networks. Thus, chemosensory neurons as well as G protein signalling may directly link pathogen avoidance to other behavioural defences as well as other life-history traits. In this context, it is interesting to note that an intact TAX-4/TAX-2 cGMP-gated channel, required in the chemosensory neurons for pathogen as well as hyperoxia avoidance (Chang et al., 2006), decreases resistance to M. nematophilum infection. It is thought that TAX-4/TAX-2 has a role in co-ordinating the secretion of cuticle components that are required by the bacteria to establish an infection in the nematode's anal region (Yook and Hodgkin, 2007). This example clearly illustrates how different aspects of organismal physiology that have an impact on host–pathogen interactions are interlinked, placing particular constraints on the way that innate immune defences can evolve.

Evolutionary perspective

Rich soils can harbour large populations of nematodes that therefore exert a potentially powerful selective force on microorganisms. As nematodes have sophisticated pathogen avoidance responses, one can expect a continuously evolving modulation of both nematode and microbe physiology. On the one hand, pathogens that gain a fitness advantage from infecting worms are selected to disguise their presence, e.g. by expressing secreted and surface molecules that cannot be detected or that may even attract worms. On the other hand, soil microbes hunted by nematodes as food may be selected to repel worms. In the latter case, one possibility is to produce substances that mimic pathogen molecules that induce the avoidance response. For example, a non-pathogenic species that produces a serrawettin W2 mimetic would be expected to have a selective advantage. There would then be a selective pressure on nematodes to uncover the microbial deceit, potentially resulting in a co-evolutionary arms race that consists of repeated cycles of microbial adaptations followed by host counter-adaptations. These dynamics are expected to shape the evolution of microbial molecules or metabolic pathways that are accessible to detection by nematode chemosensation. Consequently, taking into account the ecological, behavioural and evolutionary links between microbes and their natural predators, such as C. elegans, should prove extremely fruitful for our understanding not only of nematode life-history and defence mechanisms but also of microbial biology and pathogenesis.


We would like to thank Jürgen Berger and Kwang-Zin Lee for SEM, Anne Hart, Léo Kurz, Volkhard Kempf, Andreas Peschel, Nathalie Pujol and Yun Zhang for critical comments on the manuscript. H.S. acknowledges support from the Wissenschaftskolleg zu Berlin and the German Science Foundation (Grant SCHU1415/3-2); work in the Ewbank lab is supported by institutional grants from INSERM and the CNRS, the French Ministry of Research, the ANR and FRM.