Natural polymorphisms in C. elegans HECW-1 E3 ligase affect pathogen avoidance behaviour.

Heritable variation in behavioural traits generally has a complex genetic basis, and thus naturally occurring polymorphisms that influence behaviour have been defined only in rare instances. The isolation of wild strains of Caenorhabditis elegans has facilitated the study of natural genetic variation in this species and provided insights into its diverse microbial ecology. C. elegans responds to bacterial infection with conserved innate immune responses and, although lacking the immunological memory of vertebrate adaptive immunity, shows an aversive learning response to pathogenic bacteria. Here, we report the molecular characterization of naturally occurring coding polymorphisms in a C. elegans gene encoding a conserved HECT domain-containing E3 ubiquitin ligase, HECW-1. We show that two distinct polymorphisms in neighbouring residues of HECW-1 each affect C. elegans behavioural avoidance of a lawn of Pseudomonas aeruginosa. Neuron-specific rescue and ablation experiments and genetic interaction analysis indicate that HECW-1 functions in a pair of sensory neurons to inhibit P. aeruginosa lawn avoidance behaviour through inhibition of the neuropeptide receptor NPR-1 (ref. 10), which we have previously shown promotes P. aeruginosa lawn avoidance behaviour. Our data establish a molecular basis for natural variation in a C. elegans behaviour that may undergo adaptive changes in response to microbial pathogens.

C. elegans is initially attracted to pathogenic P. aeruginosa PA14, but within a few hours develops an avoidance response to P. aeruginosa (Supplementary Fig. 1). P. aeruginosa lawn avoidance behaviour confers increased survival in the presence of a lawn of pathogenic bacteria[11-14]. We followed the kinetics of P. aeruginosa lawn avoidance behaviour of the laboratory wild type strain N2 (Bristol, England) after transfer from E. coli OP50 (Fig. 1a). We observed that over 50% of the N2 population vacated the lawn of P. aeruginosa by t = 8 h, and by t = 24 h, over 90% of the animals were found outside of the lawn of P. aeruginosa (Fig. 1b and Supplementary Fig. 1).

Figure 1

Natural variation in C. elegans hecw-1 modulates behavioural avoidance of P. aeruginosa

(a) Schematic of the assay of pathogen avoidance behaviour in which C. elegans were transferred to plates containing a lawn of P. aeruginosa PA14, and lawn occupancy was scored over time. (b) P. aeruginosa lawn occupancy of C. elegans strains assayed at t = 24 h. Full time course in Supplementary Fig. 1. (c) Schematic of the 8 kb rescuing genomic fragment encompassing hecw-1 (Supplementary Fig. 2c). The two polymorphisms that vary between CB4856 and DA650 within this region are shown, as is the region deleted in the hecw-1(ok1347) mutant. (d) The coding polymorphisms in hecw-1 result in a Q325P substitution in CB4856 and a Y322C substitution in the AB2 and CB3198 wild strains. (e) P. aeruginosa lawn avoidance behaviour of strains carrying hecw-1 322Y 325Q, 322Y 325P, and 322C 325Q alleles in the CB4856 background and constructed as indicated. (f) P. aeruginosa lawn avoidance behaviour of animals each carrying a single integrated copy of hecw-1 genomic sequence with the naturally occurring 322Y 325Q, 322C 325Q, and 322Y 325P alleles of hecw-1 in the N2-derived hecw-1(ok1347) background. In (b) and (e) * P < 0.001 was determined by the ANOVA multiple comparisons test. In (f) * P < 0.001 and ** P < 0.02 was determined by the Student's t test based on the results of four independent experiments. Error bars indicate s.e.m.

Characterization of the npr-1 gene has shown that the 215V allele of npr-1 found in the N2 strain, which has likely been derived in the laboratory[15,16], has increased activity compared with the ancestral 215F allele of npr-1, modifying multiple behavioural phenotypes including aggregation, aerotaxis, and locomotion specifically in the presence of bacterial food[10,17-19]. In addition, we have shown previously that the 215V allele of npr-1 confers N2 with enhanced behavioural avoidance of a lawn of P. aeruginosa compared with wild isolates such as CB4856 (Hawaii, USA) that carry the 215F allele of npr-1[11]. Consistent with these prior observations, we observed that the wild isolates CB4856 and RC301 (Freiburg, Germany) that carry the 215F allele of npr-1 exhibited markedly delayed avoidance times from a lawn of P. aeruginosa compared with the P. aeruginosa avoidance times observed for N2 (Fig. 1b and Supplementary Fig. 1).

We observed that P. aeruginosa lawn avoidance behaviour by the CB4856 strain was also diminished relative to RC301, with approximately 75% of CB4856 animals remaining inside the lawn of P. aeruginosa after t = 24 h, compared with 25% lawn occupancy by RC301 animals after t = 24 h (Fig. 1b and Supplementary Fig. 1). We observed that the DA650 strain[10], which was generated by backcrossing the 215F npr-1(g320) allele of the RC301 strain into the N2 genetic background, exhibited kinetics of P. aeruginosa lawn avoidance comparable to RC301 (Fig. 1b and Supplementary Fig. 1). These observations confirmed that the npr-1 polymorphism is a major determinant of behavioural avoidance of the P. aeruginosa lawn, but also revealed that additional genetic differences among wild isolates of C. elegans influence the behavioural avoidance of pathogenic bacteria.

We sought to identify these additional genetic loci that modulate the behavioural avoidance of pathogenic bacteria. We found a substantial contribution from a single locus in determining the different pathogen avoidance behaviours observed in CB4856 and DA650 (Supplementary Fig. 2). We performed positional mapping and transgenic rescue to narrow the genomic interval containing causative genetic differences to an 8 kb fragment (genetic analysis and positional mapping detailed in Supplementary Fig. 2). This 8 kb genomic interval included a single gene, F45H7.6 (Fig. 1c), which encodes an E3 ubiquitin ligase with two WW domains and a HECT domain (Fig. 1d), with strong homology to the human HECT, C2, and WW domain containing E3 ubiquitin ligase 1 (HECW1/NEDL1)[20]. Thus, we named the C. elegans F45H7.6 gene hecw-1. We identified two polymorphisms in hecw-1 (Fig. 1c), including one coding polymorphism in exon 8 of hecw-1 that results in a Q325P change in HECW-1 in CB4856 (Fig. 1d). Consistent with our mapping data implicating hecw-1 as a determinant of P. aeruginosa lawn avoidance behaviour, we found that a deletion allele of hecw-1, ok1347 (Fig. 1c), exhibited enhanced P. aeruginosa lawn avoidance behaviour (Supplementary Fig. 3).

We sequenced the region of the hecw-1 gene region encompassing amino acid 325 from a total of 162 C. elegans strains (Supplementary Table 1), and we found that only CB4856 carries the hecw-1 325P polymorphism. Unexpectedly, we identified a second polymorphism, which results in a Y322C substitution in HECW-1 (Fig. 1d) in 11 wild strains, including AB2 (Adelaide, Australia) and CB3198 (Pasadena, California, USA). Comparison of the C. elegans amino acid sequence flanking and including the polymorphisms with the sequences from other Caenorhabditis species and mammalian hecw-1 orthologues suggested that these polymorphisms lie in a conserved, but relatively rapidly evolving region (Supplementary Fig. 4a).

Whereas the phenotypic effect of the hecw-1 325P polymorphism in CB4856 was suggested by mapping and rescue experiments, the hecw-1 322C polymorphism was detected by sequencing. To determine the effect of Y322C substitution on pathogen avoidance behaviour, we crossed the hecw-1 322C polymorphism in AB2 and CB3198 strains into the CB4856 background (thus substituting the hecw-1 322C 325Q alleles for the hecw-1 322Y 325P alleles of CB4856). We also crossed the hecw-1 322Y 325Q allele of N2 into the CB4856 background. After six backcrosses to the CB4856 strain, we observed that each of the two strains carrying the hecw-1 322C 325Q alleles exhibited P. aeruginosa lawn avoidance behaviour that was equivalent to that observed for the CB4856 strain (Fig. 1e). These data indicated that the hecw-1 322C (in AB2 and CB3198) and hecw-1 325P (in CB4856) polymorphisms confer comparable effects on P. aeruginosa lawn avoidance behaviour, with delayed avoidance relative to the hecw-1 322Y 325Q allele crossed into the CB4856 background (Fig. 1e).

Although the strains carrying the 322C and 325Q alleles of hecw-1 in the CB4856 background were generated from two independent strains, AB2 and CB3198, we considered that these strains may nevertheless carry linked loci that could modulate pathogen avoidance behaviour. Thus, we sought to determine the relative activities of the naturally occurring alleles of hecw-1 in an otherwise isogenic strain background. We generated single-copy insertions each carrying an 8 kb hecw-1 genomic sequence fragment in the N2-derived hecw-1(ok1347) background, with the transgenes differing only at HECW-1 amino acid positions 322 (Y/C) and/or 325 (Q/P). Consistent with our observations of strains carrying the naturally-occurring hecw-1 polymorphisms in the CB4856 background, we observed that animals carrying either the hecw-1 322C or the hecw-1 325P substitutions rescued the hecw-1(ok1347) P. aeruginosa lawn avoidance phenotype with a delayed avoidance of the P. aeruginosa lawn compared with animals carrying the hecw-1 322Y 325Q allele (Fig. 1f). These data show that both the hecw-1 322C 325Q and 322Y 325P polymorphisms result in increased HECW-1 activity relative to the hecw-1 322Y 325Q allele. Structural modeling of HECW-1, based on the crystal structure of the corresponding domain of human HECW1, suggested that the hecw-1 325P and hecw-1 322C polymorphisms may alter the same protein-protein interaction interface on the surface of HECW-1 without a radical disruption in overall structure (Supplementary Fig. 4b).

To determine the cells in which HECW-1 is expressed, we generated a transcriptional reporter consisting of Green Fluorescent Protein (GFP) under the control of the hecw-1 promoter, which contains 0.9 kb sequence upstream of the first exon of hecw-1 that was sufficient to rescue the hecw-1(ok1347) mutant. GFP fluorescence was observed in the nervous system throughout the body, but in the anterior ganglion was conspicuously limited to two neurons located posterior to the anterior bulb of the pharynx, each of which have projections extending anteriorly (Fig. 2a). We confirmed that these two neurons are the OLLL and OLLR neurons by co-localization experiments using the OLL-specific reporter ser-2dp::gfp[21] and the hecw-1p::mCherry transgene (Figs. 2b-2d).

Figure 2

HECW-1 functions in the OLL sensory neuron pair to negatively regulatepathogen avoidance behaviour

(a) Expression of GFP under the control of a hecw-1p showing fluorescence throughout the nervous system of the body, but limited expression in the anterior ganglion (arrow points to single pair of head neurons showing expression). Scale bar indicates 50 μm. (b-d) HECW-1 is expressed in the OLL sensory neurons. (b) Anterior ganglion expression of hecw-1p::mCherry. (c) Expression of the OLL sensory neuron marker ser-2dp::gfp. (d) Merge of Nomarski, (b) and (c). Scale bar indicates 5 μm. (e) Expression of a ser-2dp::hecw-1 cDNA::gfp translational fusion protein in the OLL neuron pair (indicated by arrow). Scale bar indicates 10 μm. (f) P. aeruginosa lawn avoidance behaviour of N2, hecw-1(ok1347), and three independent transgenic lines of the hecw-1(ok1347) mutant expressing ser-2p::hecw-1:gfp. * P < 0.001 was determined by the ANOVA multiple comparisons test. Error bars indicate s.e.m.

To determine whether expression of HECW-1 in the OLL neuron pair is sufficient to regulate pathogen avoidance behaviour, we performed rescue experiments utilizing transgenes comprised of hecw-1 cDNA fused to GFP under the control of neuron-specific promoters. Expression of a hecw-1 cDNA::gfp transgene under the control of either its endogenous hecw-1 promoter (Supplementary Fig. 5) or the OLL specific ser-2d promoter[21] (Fig. 2e) was sufficient to rescue the hecw-1(ok1347) P. aeruginosa lawn avoidance phenotype (Fig. 2f). Although the ser-2d promoter directs additional expression in PVD neurons[21], heterologous expression of hecw-1::gfp in PVD did not rescue the P. aeruginosa lawn avoidance phenotype of hecw-1 (Supplementary Fig. 6a). In addition, an alternative promoter directing hecw-1::gfp expression in neurons including the OLL neuron pair, but not PVD neurons, rescues the hecw-1 mutant phenotype (Supplementary Fig. 6b). These data suggest that HECW-1 activity in the OLL neuron pair is sufficient to rescue the P. aeruginosa lawn avoidance phenotype of hecw-1(ok1347).

To determine whether the OLL neuron pair is necessary for the negative regulation of pathogen avoidance behaviour, we carried out laser ablation of the OLL neuron pair using transgenic animals carrying the ser-2dp::gfp reporter to mark the OLL neurons for ablation. We ablated and mock ablated the OLL neuron pair and allowed the animals to recover overnight on plates seeded with E. coli OP50. To confirm that the OLL neurons were successfully ablated, we checked the OLL GFP signals of the ablated animals before we transferred them to P. aeruginosa plates. We found that animals with the OLL neuron pair ablated exhibited markedly enhanced P. aeruginosa lawn avoidance behaviour relative to mock-ablated animals as determined by diminished occupancy of the P. aeruginosa lawn after 5 h (Fig. 3a). These data suggest a role for the OLL neuron pair in the negative regulation of pathogen avoidance behaviour in C. elegans.

Figure 3

The OLL sensory neuron pair negatively regulates pathogen avoidance behaviour

(a) P. aeruginosa lawn avoidance behaviour of N2 animals carrying the ser-2dp::gfp transgene and subjected to laser ablation of OLL. Mock animals were treated in parallel without laser treatment. Two independent experiments were performed, and 15 and 30 animals were ablated in each experiment. (b) Fluorescence microscopy of animals expressing the ser-2dp::gfp transgene in OLL, and (c) carrying the ser-2dp::gfp transgene along with the csp-1b caspase under the control of ser-2d promoter that results in genetic ablation of OLL. Pharyngeal red fluorescence signals arise from co-injection marker myo-2p:mcherry. Scale bar indicates 20 μm. Genetically ablated animals were then transferred to plates seeded with (d) P. aeruginosa PA14 and (e) E. coli OP50 respectively, and lawn occupancy was recorded at the indicated times. * P < 0.001 was determined by the ANOVA multiple comparisons test. **P = 0.23 was determined by the ANOVA multiple comparisons test. Error bars indicate s.e.m.

We also generated transgenic animals in the N2 background expressing the caspase csp-1b cDNA (Denning, D. and Horvitz, H.R. unpublished results), which induces cell death, under the control of the OLL-specific promoter, ser-2dp. Using two transgenic lines with more than 90% of the animals lacking the OLL neuron pair (Figs. 3b and 3c), we found that animals lacking the OLL neuron pair exhibited an enhanced pathogen avoidance behaviour phenotype as was observed in OLL laser-ablated animals (Fig. 3d). The enhanced P. aeruginosa lawn avoidance behaviour of the hecw-1(ok1347) mutant and OLL-ablated strains led us to ask whether the lawn avoidance behaviour might be observed even in the presence of nonpathogenic E. coli OP50. However, in contrast to the marked avoidance behaviour observed in the presence of a lawn of pathogenic P. aeruginosa, we found that neither the hecw-1(ok1347) deletion, nor genetic ablation of OLL mediated by ser-2dp::csp-1b conferred a discernible lawn avoidance behavioural phenotype in the presence of E. coli OP50 (Fig. 3e).

Ultrastructural studies suggest a mechanosensory function for the OLL neuron pair[22]. Consistent with these data, we found that OLL-ablated and hecw-1(ok1347) mutant animals exhibited a defective withdrawal response of C. elegans to a light touch on the tip of its nose, the Nose Touch phenotype[23] (Supplementary Fig. 7a and 7b). The involvement of the OLL neuron pair in P. aeruginosa lawn avoidance behaviour suggests that the mechanosensory detection of bacteria may contribute to HECW-1-regulated pathogen avoidance behaviour. A role for mechanosensation in C. elegans-bacteria interactions has been previously implicated in studies of dopamine-dependent mechanosensory signaling in the C. elegans basal slowing response when encountering a lawn of bacteria[24].

The observed P. aeruginosa lawn avoidance phenotype likely results from the integration of multiple attractive and repulsive behaviours induced by nutritional, metabolic, and pathogenic aspects of the bacterial lawn. We have previously shown that P. aeruginosa lawn avoidance behaviour of the npr-1 mutant contributes to enhanced survival in a pathogen killing assay with P. aeruginosa[11,14]. Consistent with these prior studies, we observed that the hecw-1(ok1347) mutant exhibited enhanced survival compared with N2 (Fig. 4a). We proceeded to ask whether HECW-1 might act through NPR-1 to modulate behavioural avoidance of pathogenic bacteria. Whereas the hecw-1(ok1347) mutation conferred an enhanced P. aeruginosa lawn avoidance phenotype in the presence of the npr-1 215V allele and npr-1(g320) 215F alleles, we found that the putative null alleles, npr-1(ad609) and npr-1(ky13), suppressed the enhanced P. aeruginosa avoidance phenotype conferred by the hecw-1(ok1347) mutation (Fig. 4b and Supplementary Fig. 8a). Consistent with the observed genetic interaction between hecw-1 and npr-1 with regard to P. aeruginosa lawn avoidance behaviour, we observed that both the pathogenesis survival and Nose Touch phenotypes conferred by the hecw-1 loss-of-function in the presence of P. aeruginosa were also suppressed by null mutations in npr-1 (Fig. 4a and Supplementary Fig. 7b). The OLL neurons may connect through the IL2 and CEP neurons to the RMG neurons, in which NPR-1 activity exerts its diverse influence over C. elegans behaviour[25]. Partial rescue of the sensitivity of the npr-1(ky13) mutant to the effect of the hecw-1(ok1347) mutation on pathogen avoidance was observed using an npr-1 cDNA transgene under the control of the flp-5 promoter, which directs expression in RMG and ASE neurons (Supplementary Fig. 8b). These data support a model in which HECW-1 in the OLL neuron pair functions to inhibit the activity of NPR-1 in the RMG inter/motor neuron (Fig. 4c).

Figure 4

Regulation of pathogen avoidance behaviour and survival by HECW-1 is dependent on NPR-1

(a) Survival assay of indicated mutants on P. aeruginosa PA14 at 22.5°C. (b) P. aeruginosa lawn avoidance behaviour of double mutants in the N2 background carrying the hecw-1(ok1347) mutation in combination with three different npr-1 alleles— the N2 wild type npr-1 215V allele, the 215F allele npr-1(g320), and the putative null npr-1(ad609). *P < 0.001 was determined by the ANOVA multiple comparisons test. Error bars indicate s.e.m. (c) Model for the function of HECW-1 in the regulation of pathogen avoidance behaviour.

The ecological and evolutionary impact that microbes may have on host organisms has been increasingly appreciated to extend beyond the immune system, encompassing diverse aspects of host physiology[26]. Our work provides a molecular basis for how natural variation can lead to changes in behaviour that may facilitate adaptation of C. elegans to microbes.

Methods Summary

C. elegans were cultivated and strains constructed using standard methods[27]. Transgenic strains were generated using indicated constructs and standard microinjection methods[28]. MosSCI single-copy insertions were generated as described[29]. Assays measuring C. elegans pathogen avoidance behaviour utilized P. aeruginosa PA14 plates prepared as follows: a 100 mL solution of LB was inoculated with a single colony of P. aeruginosa PA14 and grown overnight without shaking at 37°C (OD=0.2-0.3). 30 μL of this culture was used to seed the center of each 100-mm NGM plate, and the seeded plates were incubated for 24 h at room temperature (22.5°C). Approximately 30 Larval Stage 4 (L4) animals were transferred onto plates containing the P. aeruginosa PA14 lawn at 22.5°C, and occupancy was determined at the indicated times. Survival assays were carried out on 35 mm Slow-Killing Assay plates[30] supplemented with 5-fluorodeoxyuridine (0.05 mg/mL) and seeded with P. aeruginosa PA14 prepared as above and maintained at 22.5°C. Laser ablations were performed on L3-stage larvae as described in text. Microscopy and image analysis was carried out on an AxioImager Z1 fluorescence microscope fitted with CCD camera (AxioCam) and processed with Axioplan image processor software (Zeiss). The statistical analyses were performed using GraphPad Prism software.