Neuronal Goα and CAPS Regulate Behavioral and Immune Responses to Bacterial Pore-Forming Toxins

Pore-forming toxins (PFTs) are abundant bacterial virulence factors that attack host cell plasma membranes. Host defense mechanisms against PFTs described to date all function in the host tissue that is directly attacked by the PFT. Here we characterize a rapid and fully penetrant cessation of feeding of Caenorhabditis elegans in response to PFT attack. We demonstrate via analyses of C. elegans mutants that inhibition of feeding by PFT requires the neuronal G protein Goα subunit goa-1, and that maintenance of this response requires neuronally expressed calcium activator for protein secretion (CAPS) homolog unc-31. Independently from their role in feeding cessation, we find that goa-1 and unc-31 are additionally required for immune protection against PFTs. We thus demonstrate that the behavioral and immune responses to bacterial PFT attack involve the cross-talk between the nervous system and the cells directly under attack.


Introduction
Bacterial infectious diseases rank among the top leading causes of death worldwide. Mycobacterium tuberculosis, Streptococcus pneumonia and Staphylococcus aureus, among others, are especially problematic because of antibiotic resistance [1]. These, and many other bacterial pathogens, produce pore-forming toxins (PFTs) that contribute significantly to their virulence [2,3,4,5].
Cells possess PFT-defense mechanisms involved in membrane resealing, as well as various molecular defenses [6,7,8,9,10,11,12,13,14]. Many genes involved in protection against PFTs were discovered in C. elegans using the PFT Cry5B and, where tested, were found to have conserved roles in mammalian cells [7,11,14]. All PFT-defense studies published to date involve the cells directly under attack by the PFT. However, as we have previously noted [13,15], the PFT Cry5B causes an inhibition of feeding behavior in C. elegans, suggestive of a neuronal component to PFT responses.
The nervous system can function to fight off pathogens by altering host behavior in response to infection, ranging from classic sickness behavior and emotional responses in humans [16], to pathogen avoidance in C. elegans [17,18,19]. The neuronal and immune systems are furthermore intimately connected through molecular pathways, likewise observed in humans (e.g., neuronal control of cytokine production [20]), as well as in C. elegans (e.g., neuronal control of antimicrobial peptide expression and noncanonical unfolded protein response genes [21,22]). Neuronal pathways are involved in C. elegans' defense against Pseudomonas aeruginosa and fungal infections, controlling behavioral responses [17,18,19] as well as downstream molecular defense pathways in the affected tissues [21,22,23,24,25].
Based on the ease of studying cellular PFT-defenses, innate immunity and the nervous system, we used C. elegans to determine whether a neuronal component exists in the defensive responses to a specific class of bacterial virulence factors, PFTs. Here, we report two neuronal pathways that regulate independent behavioral and defensive responses of C. elegans to PFTs.

PFTs rapidly and reversibly induce feeding cessation
To measure the kinetics of feeding inhibition by PFT, we determined fractions of animals feeding after various exposure times to Cry5B expressed from Escherichia coli, C. elegans' normal lab food source. We found that in wild-type animals feeding has ceased in a significant fraction of the population five minutes after transfer to Cry5B, and in the entire population after eight minutes (Fig. 1A). After 30 minutes or two hours animals are still not feeding, but after 24 hours about 40% of the same population has resumed feeding (Fig. 1B). When we exposed animals to Vibrio cholerae expressing or lacking their native PFT, V. cholerae cytolysin (VCC; a small-pore PFT like Cry5B [26]), we found that it also triggers feeding cessation (Fig. 1C). The same is true for Cry21A, a crystal toxin that belongs to the same family as Cry5B but shares only ,40% amino acid identity [15] (Fig. S1A). Inhibition of feeding induced by Cry5B, VCC and Cry21A follow very similar kinetics (Fig. 1A, C, S1A), and it thus appears that rapid inhibition of feeding is part of a generalized response of C. elegans to smallpore bacterial PFTs.
PFT-induced cessation of feeding can be inhibited by cotreatment with exogenous serotonin, a neurotransmitter known to induce feeding in C. elegans [27] (Fig. 1D). Combined with the fact that 40% of animals have resumed feeding after 24 hours (Fig. 1B), this serotonin result indicates that cessation of feeding is reversible and not likely caused by physical damage to the pharynx and suggests that the neural circuitry that controls feeding remains functional in the presence of PFT.
Goa pathway components are required for cessation of feeding in response to PFTs Neuronal G-protein signaling pathways control many behaviors in C. elegans. Through screening of numerous Goa, Gqa, and Gsa pathway mutants we found a single mutant that shows a penetrant, dramatic loss of feeding inhibition on Cry5B after 30 minutes and two hours, goa-1(sa734) (Table S1). sa734 is a null allele of goa-1, the single C. elegans homolog of the most abundant G-protein in the mammalian brain, Goa [28,29]. Goa mediates the signals of many neurotransmitters [29,30], and in C. elegans GOA-1 controls various behaviors such as locomotion, egg-laying, feeding and olfactory adaptation [31,32,33].
Further analysis of goa-1(sa734) feeding behavior shows that after 10, 30 and 120 minutes exposure to Cry5B it has significantly higher fractions of animals feeding than wild type ( Fig. 2A). The goa-1(sa734) mutation also causes impaired cessation of feeding in response to VCC and Cry21A after 10, 30 and 120 minutes ( Fig. 2B, S2A). A different null allele of goa-1, ep275, similarly shows significantly increased fractions of animals feeding on Cry5B at all three time points (Fig. S2B), indicating that the phenotype is caused by a loss of goa-1 function.
Next, we measured pharyngeal pumping rates of several of the Goa and Gqa pathway mutants in absence or presence of Cry5B (The pumping rate is the frequency of muscle contractions in the nematode's posterior pharynx, a measure for its feeding rate [34]). In all strains tested, Cry5B causes a significant reduction of pumping rates after 30 minutes exposure (Fig. 2C, Table S2). However, goa-1(sa734) shows significantly higher pumping rates than wild type on Cry5B (Fig. 2C). Interestingly, where complete loss of goa-1 function results in constitutive pumping on Cry5B, a weak reduction-of-function mutation of goa-1, n1134, still allows for normal inhibition of feeding by Cry5B (Fig. 2C). goa-1(n1134) mutants have a N-terminal truncation of first 4 amino acids causing it to lack the consensus site for myristoylation [33]. They show increased rates of egg laying, an abnormal olfactory response that they share with goa-1(sa734), and a general resistance to phenotypes induced by administration of exogenous serotonin [31,33]. This may indicate that the reduction of function is too weak to show a phenotype, or that the N-terminal part or proper membrane localization is not required for GOA-1 to inhibit  (Fig. S1B).) (B) 0.5 and 2 hr after transfer to E. coli-Cry5B, animals are not feeding, but after 24 hr almost half of the population has resumed. (This is not due to reduced activity of Cry5B (Fig. S1C).) (C) V. cholerae expressing VCC induces cessation of feeding, following similar kinetics as Cry5B, whereas V. cholerae lacking VCC does not (blue line). (D) Exogenous serotonin causes constitutive feeding on Cry5B. In this and subsequent figures, graphs show mean 6 standard error of three experiments unless otherwise described, and statistics indicated are: ns not significant, * p,0.05, ** p,0.01, *** p,0.001. Lack of any symbol indicates no significant difference. Here, statistics indicate significance of difference between PFT and control at the same time point. In all subsequent figures, statistics indicate the difference between mutant and wild type on the same treatment, and where applicable additional statistics are provided in Table S2. doi:10.1371/journal.pone.0054528.g001 feeding. Thus, Goa, a major pan-neuronal gene in C. elegans [32,33], plays a critical role in regulating PFT-induced inhibition of feeding.
unc-31(e928), a null allele [40], was also notably different from wild type. These animals initiate the normal feeding-cessation response at 30 minutes PFT exposure, but it is not maintained since by two hours a significant fraction of the animals have resumed feeding (Table S1). Maintenance of feeding cessation is restored when unc-31 is selectively expressed in neurons (Table S1), indicating it is required in the nervous system to maintain, but not initiate, PFT-induced cessation of feeding. unc-31 is further discussed below.

The Goa G-protein pathway is required for PFT defense
The mutant strains screened for constitutive pumping in the presence of Cry5B were also examined for qualitatively altered immunity to PFT after 24 and 48 hours exposure to three doses of Cry5B PFT. Mutation of several G-protein signaling pathway genes leads to moderate hypersensitivity to the pore-forming toxin Cry5B (''Hpo'' phenotype [11]), but only loss of goa-1 or eat-16 leads to severe Hpo phenotypes at both time points (Table S1, Fig. 3A). The Hpo phenotype is due to loss of goa-1, as it is also apparent with the ep275 null allele (Fig. S3).
For a quantitative analysis of the sensitivities of several Goa and Gqa pathway mutants, we performed dose response assays with purified Cry5B. Consistent with the qualitative assays, goa-1(sa734) and eat-16(ce71) mutants show significantly decreased survival compared to wild type on the three Cry5B doses tested (Fig. 3B). The goa-1(n1134) mild reduction-of-function mutant also shows significant hypersensitivity at two Cry5B doses, although not to the extent of goa-1(sa734) and eat-16(ce71). This innate immunity role appears uncoupled from cessation of feeding as eat-16(ce71) mutants show only a modest defect of feeding inhibition and goa-1(n1134) mutants display a normal feeding-cessation response.
Next, we exposed mutants lacking Goa components to V. cholerae expressing VCC and scored 24 hours later for survival. goa-1(sa734) and eat-16(ce71) null mutant animals show significantly decreased survival after 24 hours on various doses of V. cholerae expressing VCC compared to a V. cholerae strain lacking the PFT, and for eat-16(ce71) a dose-response to VCC is evident (Fig. 3C). V. cholerae also induces significantly increased lethality in the absence of VCC in both these mutants compared to wild type (Fig. 3C) (further enhanced by the presence of the PFT), which is consistent with the fact that goa-1(sa734) animals are hypersensitive to infection by the bacterial pathogen Pseudomonas aeruginosa [42]. goa-1(n1134) mild reduction-of-function animals show significantly increased lethality on the highest VCC dose only (Fig. 3C). Animals overexpressing egl-10 show no increased lethality on V. cholerae, regardless of the presence or absence of VCC (Fig. 3C). (egl-30 and egl-10 reduction-of-function or null mutants could not be tested on V. cholerae due to internal hatching of progeny.) Hypersensitivity of these mutants is not likely due to general illness, as in a previous study goa-1(sa734) animals showed no hypersensitivity to the heavy metal cadmium, or high osmolarity [42].
Calcium activator for protein secretion (CAPS) is required in neurons for PFT defense goa-1 mutants are hypersensitive to P. aeruginosa infection, which has been linked to hypersecretion in neurons [42]. Consistent with this, calcium activator for protein secretion (CAPS)/unc-31, egl-3 (protein convertase type 2) and egl-21 (carboxypeptidase E) mutants, which have reduced neurosecretion because of reduced dense-core vesicle (DCV) exocytosis or defects in DCV contents, are resistant to P. aeruginosa infection [42,43,44,45]. To determine if a similar correlation exists for PFT defense, we included unc-31, egl-3 and egl-21 mutants in our screen (Table S1). Opposite to the P. aeruginosa resistance phenotype, loss of these genes causes hypersensitivity to Cry5B (Table S1, Fig. 4A). In quantitative analyses unc-31 and egl-21 mutants show significantly reduced survival on purified Cry5B relative to wild-type controls (Fig. 4B). The unc-31(e928) Hpo phenotype is not likely due to general sickness, as loss of unc-31 increases life span and confers resistance to P. aeruginosa [42,46]. unc-31 is expressed in all neurons and other secretory cells, but not in the intestine [45]. unc-31(e928) animals in which unc-31 expression was selectively restored to the neurons via the rab-3 promoter [47] show wild-type sensitivity to Cry5B (Fig. 4B, Table S1). Thus, UNC-31 functions in the neurons to protect C. elegans from an attack by PFT on its intestinal cells.

Discussion
We present data showing that bacterial PFTs rapidly and reversibly inhibit feeding in C. elegans, and that the Goa pathway components goa-1 and downstream eat-16 are required for this feeding-cessation response. goa-1 and eat-16 are furthermore required for defense against PFTs. egl-21 and neuronal unc-31 are also required for PFT defense, and neuronal unc-31, but not egl-21, is required for maintenance of PFT-induced feeding cessation.
As mentioned above, the rapid inhibition of feeding by PFT is not likely caused by damage to the cells or neural circuitry of the pharynx, as PFT-induced feeding inhibition is reversible (Fig. 1B,  1D). Additionally, the pharynx is not directly affected by Cry5Bremoving the Cry5B receptor from the intestine but not the pharynx leads to Cry5B resistance and expression of the Cry5B receptor in the intestine but not the pharynx results in Cry5B sensitivity [48,49]. Cry5B furthermore does not inhibit feeding in intestine-receptor-negative animals (10, 30, 60 and 90 minutes after transfer to Cry5B, 94.9%, 98.2%, 84.7% and 82.0% respectively of bre-3(ye28) animals showed normal feeding rates [average of 2 experiments, 23-30 animals each]), and exposure of C. elegans to Cry5B results in permeabilization of the intestinal cells but not the pharyngeal cells [13]. These data suggest that rather than the PFT itself, the effects of pore formation in the intestine are sensed and cause a signal to be sent to the pharynx to inhibit pumping. Such a signal could be relayed via synaptic connections or nonsynaptic communication (e.g., neuropeptides) between the somatic and pharyngeal nervous systems, or the pharyngeal nervous system could autonomously sense the effects of Cry5B (e.g., ion fluxes). Consistent with this model (Fig. 4C), (1) unc-31 is required in the neurons for a normal feeding-cessation response (Table S1), (2) goa-1 and eat-16 are required for a normal feedingcessation response ( Fig. 2A, B, C, S2A, B) and are predominantly (although not exclusively) expressed in neurons [32,33], and (3) exogenous administration of the neurotransmitter serotonin can suppress the feeding-cessation response (Fig. 1D).
Since loss of goa-1 results in both continued feeding on Cry5B and increased sensitivity to Cry5B, the two phenotypes could be related-i.e., increased consumption of the PFT could cause increased sensitivity because of and increased PFT dose. However, several lines of evidence suggest against such a relationship. Loss of goa-1 results in hypersensitivity to V. cholerae lacking VCC (Fig. 3C). However, wild type animals also continuously feed on V. cholerae lacking VCC (Fig. 1C). Furthermore, goa-1(n1134) reduction-offunction animals show normal inhibition of feeding (Fig. 2C) but are hypersensitive to Cry5B (Fig. 3B), and a lack of correlation between the expressivities of the feeding-cessation response and PFT immunity was similarly found for the eat-16(ce71) mutant (Fig. 2C, 3B). So, although we do not exclude that increased intake of PFT in part contributes to mutant hypersensitivity, we hypothesize that goa-1 affects feeding inhibition and innate defenses via independent downstream signaling pathways. Such independent roles in behavioral and molecular defensive responses for a single pathway have been shown before in C. elegans for the insulin/IGF-I pathway, in the context of a challenge with a pathogenic Bacillus thuringiensis strain (likely expressing nematicidal crystal toxins) [50].
The inverse correlation that was found with regard to P. aeruginosa resistance for unc-31 and goa-1 loss-of-function mutants [42] does not exist for Cry5B PFT defenses (Fig. 3A, B, C, 4A, B). An analogous observation was made for the hypoxia pathway, which when activated protects C. elegans against V. cholerae expressing VCC but increases its sensitivity to V. cholerae lacking VCC [6]. Likewise, UNC-31 activity protects against PFTs but may compromise the resistance to other virulence factors. This finding is likely in accordance with the evolution of the nematode in its natural environment, where different pathogens have evolved different attack strategies that require more complex defensive responses from the nematode.
No direct genetic link was shown between unc-31 and goa-1 in P. aeruginosa defense [42]. GOA-1 is furthermore thought to be mainly involved in the secretion of synaptic vesicles that contain neurotransmitters such as acetylcholine, whereas UNC-31 is thought to control secretion of dense-core vesicles that contain hormones, serotonin, and neuropeptides such as insulin [47,51]. unc-31(e928) stops feeding after 30 minutes on Cry5B but prematurely resumes feeding at two hours (Table S1), whereas goa-1(sa734) is feeding at all time points tested, and to a larger extent than the unc-31 mutant. In addition, the effects of goa-1 and unc-31 mutation on pumping on Cry5B appear to be cumulative, i.e., the goa-1(sa734); unc-31(e928) double mutant shows higher fractions pumping on Cry5B than either single mutant (Table S1). Therefore, we propose that the PFT-defense functions of goa-1 and unc-31 are not coupled to levels of neuronal secretion per se, as is hypothesized to be the case for P. aeruginosa defense [42], but rather depend on the contents of the different secreted vesicles.
In summary, we show that the C. elegans host responses to PFTs include two defense pathways that function outside the tissue under attack, involving the conserved genes goa-1 and unc-31, and that the nematode's nervous system can modulate behavior and immune defenses in response to a specific and highly abundant class of bacterial toxins -PFTs.

C. elegans and bacterial strains
Worm strains used in this study are outlined in Table S3, and were maintained at 20uC on E. coli strain OP50, as described [52]. Mutations were confirmed by phenotype where possible, otherwise by PCR band size comparison or DNA sequencing.

Feeding assays
To determine fractions of animals in a population that are feeding, E. coli-Cry5B, E. coli-Cry21A, V. cholerae-VCC, and no-PFT control plates were prepared as described [6,53,54]. 10-15 L4 animals were transferred to each plate, and the assay was incubated at room temperature and observed after the indicated times. Using a dissecting microscope, any single animal was observed for a maximum of 5-10 seconds, and scored as ''feeding'' if rhythmical backward movements of the grinder were observed [34]. To determine feeding on 24-hr old plates, plates were prepared as normal, but incubated at 20uC for 24 hr before use. Three independent repeats were performed for each assay. To quantitatively assess feeding rates of individual animals, we determined pumping rates. Pumping entails the backward movements of the posterior grinder of the pharynx, which can be observed through a dissecting microscope [34]. E. coli-Cry5B plates were prepared as above. 10-13 L4 animals were transferred to each plate, and pumping rates were measured for 30 seconds, before transfer, or after 30 minutes incubation at room temperature. Each assay was repeated independently three times. It is of note that goa-1(n1134) mutants were reported to have slightly decreased pumping rates [33], which we also observed here for n1134 and sa734 (Fig. 2C).

PFT toxicity assays
For qualitative toxicity assays, including the screening (Table S1), 0%, 10%, 25% and 100% Cry5B plates were used, which are prepared by diluting E. coli-Cry5B (100%) bacteria with empty vector control (0%) bacteria at the indicated ratios, as described [53]. To prevent behavioral artifacts due to worms wandering off the toxin [19,50] bacterial lawns were spread to cover the entire agar plate. 15 L4 animals were transferred to each plate. Assays were incubated at 20uC and observed after 24 and 48 hr. Representative images were taken after 48 hr, using an Olympus SZ60 dissecting microscope linked to a Canon Powershot A620 digital camera, and using Canon Remote Capture software. Three independent repeats were performed for each mutant strain. The same plates prepared for screening Cry5B sensitivity (Table S1) were also used to screen feeding behavior (see below).
Quantitative survival assays with purified Cry5B were performed as described [53]. Cry5B was purified as described [11]. Three independent repeats were performed. Note that toxicity of purified Cry5B on a mg/mL basis sometimes differs from batch to batch, which is why different Cry5B concentrations give similar survival rates for wild-type animals in different experiments (e.g., Fig. 3B versus Fig. 4B). The same batch of Cry5B is used within each complete set of experiments.
VCC survival assays were performed essentially as described [6], except assays were incubated at 20uC, and scored after 24 hr. VCC dilutions were prepared the same way as E. coli-Cry5B dilutions. Animals that showed internal hatching of progeny (''bagging'') or the intestine protruding through the vulva (''exploding'') were censored. Pilot assays involving the use of 5fluoro-29-deoxy-uridine (FUdR) to prevent the development (and internal hatching) of progeny revealed that FUdR causes V. cholerae to lose its pathogenicity. Therefore, egl-30-like strains could not be tested here, as without inhibiting the development of their progeny the terminal phenotype of most of these animals is bagging due to their low egg-laying rates. An E. coli OP50 control was included with each assay, which showed no significant lethality for any of the strains after 24 hr (data not shown). Three independent repeats were performed with 30-50 animals per treatment. After 24 hours, wild-type animals are qualitatively affected by the presence of VCC but show no lethality on any of the VCC doses (Fig. 3C). VCC-induced lethality is noticeable in wild-type animals at later time points [6].

Statistical analyses
For Fig. 1A-C, and S1A-C t-tests were used to compare each pair of measurements. For Fig. 1D, 2A, B, 3B, C, 4B and S2A, B one-way ANOVA with Dunnett's comparison of means was performed for each treatment. For Fig. 2C and S2C a Wilcoxon test was performed. For Table S1 fractions feeding on Cry5B were normalized to fractions feeding on control plates at each time point and one-way ANOVA of these normalized values was performed. Means of the mutants were compared to mean of wild type using a Dunnett's test.
Statistics were calculated using JMP 9.0 software (SAS Institute). Prism 5.0 Software (GraphPad) was used to draw graphs. p values, means or medians and confidence intervals or interquartile ranges of all experiments performed for this study are provided in Table S2. Figure S1 Cry21A inhibits feeding, transfer does not alter feeding, and Cry5B plates retain potency over 24 hr. (A) Animals transferred to E. coli expressing Cry21A rapidly stop feeding, whereas animals transferred to control plates do not. (B) Fractions of wild-type animals feeding at various time points after transfer are the same as before transfer. (C) Wild-type animals transferred to 24-hr old E. coli-Cry5B plates are not feeding 10 or 30 min after transfer. Statistics indicate difference between 24-hr old control and 24-hr old Cry5B plate. Here and in all subsequent supplemental figures graphs show mean 6 standard error of 3 experiments, and statistics indicated are: ns not significant, * p,0.05, ** p,0.01, *** p,0.001. Additional statistics are provided in Table S2. (TIF) Figure S2 goa-1 null mutants constitutively feed on Cry21A and Cry5B, and transferring worms does not affect their pumping rates. (A) E. coli-expressed Cry21A inhibits feeding in wild-type animals after the indicated exposure times, but does not inhibit feeding in goa-1(sa734) mutants. (B) goa-1(ep275) constitutively feeds on E. coli-expressed Cry5B. (C) 30 minutes after transfer to plates with E. coli not expressing PFT, pumping rates are the same as before transfer. Bars show mean 6 standard error of 3 experiments, and dots are individual measurements of all three experiments. Additional statistics are provided in Table S2. (TIF) Figure S3 Goa is required for PFT defense. After 48 hr, goa-1(ep275) mutants are qualitatively hypersensitive to E. coliexpressed Cry5B. Scale bar = 500 mm.