Feeding behaviour of Caenorhabditis elegans is an indicator of Pseudomonas aeruginosa PAO1 virulence

Strains and growth conditions

The Caenorhabditis elegans Bristol strain N2 was used for feeding and bacterial killing experiments (provided by Dr Kunyan Zhang). The tph-1 mutant worm is deficient in tryptophan hydroxylase required for the biosynthesis of serotonin (provided by Dr Jim McGhee). Escherichia coli OP50 was used as a non-pathogenic food source for cultivating C. elegans. The panel of type III secretion mutant strains and the wild type PAO1F (PAO1 from Alain Filloux) are described in Table 1 (provided by Dr Arne Rietsch). Wild type P. aeruginosa PAO1 and all mini-Tn5-lux transposon mutants (see Table 2) were previously described (Lewenza et al., 2005). Fifty µl of overnight OP50 cultures was spread and grown on NGM, which was used as a food source to cultivate nematodes. The NGM medium is composed of double distilled water, 0.25% (w/v) Bacto-Peptone (BD), 0.3% (w/v) NaCl, and 2% (w/v) Bacto-Agar (BD), 5 µg/ml cholesterol, 1 mM MgSO₄, 25 mM KH₂PO₄ (pH 6) and 1 mM CaCl₂. The SK medium is similar to NGM but contains 0.35% (w/v) Bacto-Peptone (BD).

C. elegans feeding behaviour assays

Ten µl of normalized bacterial food sources (1 × 10⁷ cfu/ml) were spotted on SK plates and grown overnight. Twenty L4 stage hermaphrodite nematodes were then transferred from E. coli OP50 NGM plates to the 6 cm assay SK plates containing the pre-grown P. aeruginosa food sources to be tested. SK plates with two, three or four P. aeruginosa food sources were monitored throughout a 72-h period. The number of worms in each bacterial colony was counted as an indicator of food preference. Within this time period, the original L4 worms could be easily differentiated from all new offspring on the basis of size. Feeding assays were performed at 25 °C. The unpaired t-test was used to compare the number of worms that selected mutant or wild type bacterial food sources at each time point. Images of colonies and worms were captured with a Motic DM143 digital microscope. When using tph-1 worms, the choice index was calculated using a previously described formula (choice index = test worms minus OP50 worms/total worms) (Zhang, Lu & Bargmann, 2005).

Fluorescence microscopy of C. elegans

L4 nematodes were given the choice of P. aeruginosa producing cherry fluorescent protein (ChFP) or green fluorescent protein (Gfp). The plasmid pCHAP6656 was introduced into wild type PAO1 as the Chfp-tagged strain, and exoS::GL3 ΔexsE was used as the Gfp-tagged strain. The pCHAP6656 plasmid expresses Chfp as outer membrane-localized lipoprotein (Lewenza, Mhlanga & Pugsley, 2008) and the plasmid did not affect the C. elegans response to PAO1. Nematodes were removed from either the Chfp or Gfp-tagged bacterial colonies and mounted on microscope slides for visualization. Fluorescent bacteria in the nematode gut were observed using a Leica DMI4000 B inverted microscope equipped with an ORCA R2 digital camera. The following excitation and emission filters were used to monitor red and green fluorescence, respectively (Ex 555/25; Em 605/52; Ex 490/20; Em 525/36). The Quorum Angstrom Optigrid (MetaMorph) acquisition software was used for image acquisition with a 63 × 1.4 objective and deconvolution was performed with Huygens Essential (Scientific Volume Imaging B.V.).

High throughput assay to identify preferred bacterial food sources

We previously constructed a mini-Tn5-lux transposon mutant library and mapped the transposon insertion site in 2,370 individual mutants (Lewenza et al., 2005). As a primary screen, this collection of transposon mutants was grown overnight in LB medium in 25 × 96 -well microplates. A 48-pin stamp was used to stamp transfer ∼5 µl of liquid LB culture onto a 6 × 8 grid of colonies on 15 cm SK agar plates and grown overnight at 37 °C. Ten L4 nematodes were deposited on each side of the grid (20 worms in total) and were allowed to eat and reproduce over the course of 7 days at 25 °C. Each plate with 48 different food sources was monitored daily for the first disappearance of specific colonies, which we identified as preferred food sources. In general, the first colonies were consumed by day 4. After identification of the preferred food source colonies, the nematodes would continue to reproduce and eat all the remaining bacterial colonies to completion. Mutants identified in the primary screen of 2,370 strains were retested in secondary screens to confirm the preferred food source phenotype. In method A, 6 mutants were arrayed in 6 specific positions within a 6 × 8 grid and surrounded by wild type PAO1. Alternatively for method B, individual mutants were positioned in 3 consistent well positions in the middle of a 6 × 8 grid of wild type PAO1 (see Fig. 3). Ten L4 worms were added to the both sides of the plates (20 worms in total), which were monitored daily to identify the first colonies eaten to completion.

Growth defect analysis of transposon mutants

All mini-Tn5-lux mutants that served as preferred food sources for C. elegans in the primary screen were tested for growth defects in LB and SK liquid media. Briefly, each strain was grown overnight in LB cultures and diluted 1/500 into 100 µl of fresh LB or SK liquid media in 96-well microplates. Cultures were grown and monitored for the OD₆₀₀ values every 20 min throughout 18 h growth at 37 °C using the Wallac V ictor³ luminescence plate reader. Mutations that caused growth defects in SK growth medium were excluded from further analysis to rule out the possibility that weakly grown colonies might be preferred by the nematodes due to the low cell density.

Nematode chemotaxis assays

Bacterial cultures were grown overnight and 1 ml of supernatant was collected by centrifugation and filter sterilized. Twenty µl of culture supernatant was spotted at either end of 6 cm SK plates and 15 L4 nematodes were transferred to the center of SK plates. The plates were monitored every 15 min throughout a 2 h period to count the number of nematodes that moved into the dried spot of bacterial supernatant. Each supernatant was tested three times.

Slow killing assays

As previously described (Powell & Ausubel, 2008), 10 µl (1 × 10⁷ cfu/ml) of bacterial cultures were transferred to 6 cm SK plates and spread to form a bacterial lawn. Thirty L4 stage nematodes were transferred from NGM-OP50 to SK plates containing the pre-grown bacterial lawn of P. aeruginosa and incubated at 25 °C. SK plates were prepared with 25 µg/ml of 5-fluoro-2′-deoxyuridine (FUdR), a eukaryote DNA synthesis inhibitor that prevents the growth of egg offspring during the experiment. E. coli OP50 was fed nematodes as a negative control strain with limited killing of C. elegans after 10 days. Nematodes were monitored daily under a dissection microscope to detect unresponsive and dead worms. Kaplan–Meier survival curves were plotted for triplicate experiments (n = 30) with a total of 90 nematodes. Significant differences in C. elegans survival were determined using the log-rank test (Graph Pad Prism) as previously described (Hasshoff et al., 2007; Feinbaum et al., 2012; Korgaonkar et al., 2013).

Ingestion and food preference in binary feeding assays withC. elegans

The type III secretion system is an important virulence factor in P. aeruginosa but was shown previously to not have a role in C. elegans virulence (Wareham, Papakonstantinopoulou & Curtis, 2005). Regulation of the type III secretion system involves a negative regulator called ExsE, and an ΔexsE mutant overexpresses and showed increased secretion of the type III effectors ExoT, ExoY and ExoS into the culture supernatant after growth in inducing, low calcium conditions (Rietsch et al., 2005). The ΔexsE mutant also demonstrated increased cytotoxicity to host cells, presumably due to the increased gene expression of all type III secreted effectors that were injected into host cells upon contact (Rietsch et al., 2005). We were interested to determine if the ΔexsE mutant that overexpresses the type III effectors had an effect on C. elegans feeding behaviour. Using a binary feeding assay where nematodes are given the choice of two bacterial food sources, L4 stage nematodes preferred wild type PAO1 to the ΔexsE strain (Fig. 1A). This result suggested that overexpression of type III effectors was detected and avoided by the nematode.

To determine if the nematodes ingested both possible food sources, C. elegans was fed a Chfp-tagged wild type PAO1 (pCHAP6656) or Gfp-tagged, ΔexsE exoS-gfp strain and nematode intestinal tracts were monitored with microscopy. The ΔexsE exoS-gfp strain overexpresses exoY and exoT and was also avoided by nematodes, comparable to the ΔexsE strain that overexpressed exoY, exoT and exoS (data not shown). Although the worm body was autofluorescent in both green and red channels, the early stages of bacterial accumulation were observed (Fig. 1). At early time points, worms that were removed from wild type PAO1 (pCHAP6656) colonies were shown to have exclusively red fluorescent bacteria in the gastrointestinal tract (Fig. 1B). Similarly, worms taken at early time points from the ΔexsE exoS-gfp strain had exclusively green fluorescent bacteria in the GI tract (Fig. 1C). At later time points, worms were shown to have both green and red fluorescent bacteria in the GI tract (Fig. 1D). This indicates that worms ingested both food sources and did not simply avoid feeding on the ΔexsE strain. Some nematodes chose neither of the two food sources but as the exposure time increased, more worms made a food choice (data not shown). However, there were still worms that avoided both food sources and did not have any bacteria in the GI tract (1E). We concluded that the nematodes ingested and compared both possible food sources, but ultimately preferred the wild type strain.

Type III secretion mutants are preferred food sources and impaired for virulence

We reasoned that if type III overexpression strains were avoided, than secretion defective strains might be preferred. Indeed, C. elegans preferred strains that were defective in the type III secretion machinery (ΔexsEΔpscD) and were defective for all type III secretion effectors (ΔexoSY T) (Fig. 2A). After longer periods of feeding (∼4 days), the ΔexsEΔpscD food source was the most preferred as it was the first colony consumed (Fig. 2C). To compare the potential toxicity of type III effectors, nematodes were exposed to strains that overexpress individual effectors. In these feeding assays, nematodes consistently preferred the ExoY overexpressing strain suggesting ExoY is the least toxic effector (Fig. 2B). ExoY is an adenylate cyclase and our findings are comparable with a report showing that ExoY was not required for in vitro cytotoxicity and had no impact on dissemination during infection (Lee et al., 2005). To determine if the type III secretion system of PAO1 was required for C. elegans virulence, we performed slow killing assays with individual P. aeruginosa food sources. The ΔexsEΔpscD showed a significant reduction in slow killing of C. elegans, confirming that the most preferred food sources also show reduced virulence (Fig. 2D).

High throughput feeding behaviour assays

Given the relationship between feeding preferences and virulence observed with strains defective for type III secretion, we wanted to screen a large collection of mutants to find additional preferred food sources. We previously constructed a large library of mini-Tn5-lux mutants with mapped insertion sites, which were arrayed into a library of 2,370 mutants in 96-well microplates (Lewenza et al., 2005). Using standard petri dishes with SK agar, we stamped liquid LB cultures onto a grid of 48 colonies (6 × 8), and introduced 20 L4 stage nematodes. The plates were incubated at 25 °C and observed daily to identify colonies that were preferentially eaten to completion. In the primary screen of 2,370 mutants, we identified 191 strains that were preferred food sources. Growth curves were performed on all candidates from the primary screen and those mutants with growth defects in SK medium were excluded from further study. We developed secondary screens to confirm this phenotype. In method A, we arranged 6 unique preferred food sources within a grid of 48 colonies, with PAO1 in all other positions (Fig. 3A). This method did confirm the preferred feeding source phenotype of some mutants. Since the worm was given a choice of multiple preferred food sources, there may have been competition for the most preferred. To reduce the competition observed in method A, we used an alternative secondary screen where individual candidates were situated in triplicate positions within a 48-grid of wild type PAO1 (Fig. 3B). Using these two secondary screens, we confirmed the preferred food source phenotype of 37 mutants (Table 2). After sufficient incubation time, the nematodes would reproduce to high numbers and eat all the bacterial colonies to completion.

Preferred food sources for C. elegans included mutants in many known virulence factors

Table 2 summarizes the transposon insertion sites of genes that led to the preferred food source status for C. elegans. The genes identified in this screen can be grouped into the following categories: known virulence factors, virulence regulatory systems, nutrient utilization and metabolism, and hypothetical proteins. The majority of the genes identified in PAO1 as having a preferred food source phenotype (Table 2) are present in PA14 but were not identified in the genome-wide screen for slow killing determinants in PA14 (Feinbaum et al., 2012). This indicates that the two assays are unique and identify different bacterial phenotypes. However, mutations in the pqs biosynthesis and type IV pili genes resulted in slow killing defects (Feinbaum et al., 2012) and led to preferred food source status (Table 2), indicating a small overlap between the two screens.

Mutations in genes encoding the PQS biosynthesis genes, type IV pili and the TypA GTPase had a preferred food source phenotype, and all were previously recognized as virulence factors in P. aeruginosa (Zaborin et al., 2009; Feinbaum et al., 2012; Neidig et al., 2013). Among the known regulators of virulence, we identified the PprB, PhoQ, PqsR and CifR regulatory systems (Cao et al., 2001; MacEachran, Stanton & O’Toole, 2008; Gooderham et al., 2009; de Bentzmann et al., 2012). The C. elegans virulence screens frequently identify global regulators and two-component system regulators, probably due to the pleiotropic effects of these mutations, given the large number of virulence genes controlled by these systems (Sifri, Begun & Ausubel, 2005). Both PqsR and PprB are known regulators of the pqsABCDE biosynthesis genes (Cao et al., 2001; de Bentzmann et al., 2012), which were previously shown to be involved in the C. elegans red death and slow killing phenotypes (Zaborin et al., 2009; Feinbaum et al., 2012). PhoQ is a two-component sensor that responds to limiting Mg²⁺ and controls numerous genes including antimicrobial peptide resistance modification to LPS (Macfarlane et al., 1999), and is required for virulence in plant and chronic rat lung infections (Gooderham et al., 2009). The CifR repressor controls the Cif secreted toxin that reduces the apical expression of CFTR and chloride secretion in epithelial cells (MacEachran et al., 2007; MacEachran, Stanton & O’Toole, 2008). PA14, but not PAO1, expresses Cif activity (Swiatecka-Urban et al., 2006), suggesting that the CifR mutant phenotype in this assay is independent of Cif activity.

Consistent with the identification of regulatory proteins in C. elegans virulence screens, we also identified mutations in several transcriptional regulators and two-component sensors and response regulators (PA0056, PA0929–PA0930, PA4983) that led a preferred food source phenotype (Table 2). Intergenic insertions between PA0120 and PA0121 likely disrupted the downstream PA0121, an uncharacterized transcriptional regulator. It is unclear what gene is affected by an intergenic insertion between PA4353–PA4354, due to their divergent orientation. The functions of PA0056, PA4983 and PA0929–PA0930 are currently unknown, and the latter two-component system is adjacent to gacS, a global virulence regulator required for C. elegans killing (Sifri, Begun & Ausubel, 2005).

Preferred food sources for C. elegans included mutants in nutrient acquisition pathways

This screen also identified mutants with insertions into pathways required for the uptake and utilization of various nutrient sources. The DctD two-component response regulator controls expression of the ABC transporters that take up C-4 dicarboxylates (Valentini, Storelli & Lapouge, 2011), and a mutation in this gene led to a preferred food source for C. elegans (Table 2). There were also mutations in multiple ABC transporters that led to preferred food source status, including transporters involved in the uptake of dipeptides, quaternary ammonium compounds (QAC), zinc, as well as genes involved in amino acid metabolism (Table 2). The AgtS (PA0600) two-component sensor regulates an ABC transporter that is required for the uptake of δ-aminovalerate and γ-aminobutyrate (Chou, Li & Lu, 2014). The cognate response regulator AgtR is also required to sense N-acetylglucosamine shed from gram-positive bacterial cell wall, leading to increased pyocyanin production and enhanced fruit fly killing during mixed infection (Korgaonkar et al., 2013). Mutations in this two-component system may affect both nutrient acquisition and pyocyanin production, which is directly toxic to C. elegans (Cezairliyan et al., 2013). These results are consistent with a previous screen that also identified various metabolic and nutrient acquisition genes required for slow killing of C. elegans by PA14 (Feinbaum et al., 2012).

Preferred food sources do not cause universal chemoattraction of C. elegans

A simple explanation for the preferred feeding phenotype is that bacterial mutants may have had increased production of an attractant or lacked production of a repellant that caused increased nematode accumulation in a colony. To test this hypothesis, we compared the nematode attraction to filtered culture supernatants from wild type PAO1 and preferred food sources in a binary chemoattraction assay. Culture supernatants were collected and spotted on SK plates at equal distances from where 15 L4 worms were transferred (Fig. 4A). Within 30 min, worms were more attracted to the culture supernatants of E. coli OP50 than PAO1 (Fig. 4B), which was also a preferred food source (Fig. 3). Of the supernatants tested, nematodes were more chemoattracted to supernatants from the PA4497 and PA1291 mutants (Fig. 4). In contrast, nematodes were equally attracted to the parent and mutant supernatant in the other combinations tested (Fig. 4). In one instance, C. elegans was repelled by one culture supernatant after 30 min (Fig. 4E). While it appeared that some supernatants might have contained a chemoattractant for C. elegans, no chemoattraction was observed in most cases.

tph-1 worms demonstrated preferred feeding behaviour

The tph-1 nematode strain is defective for the enzyme tryptophan hydroxylase, which is the rate-limiting step in the biosynthesis of serotonin. This neurotransmitter was previously shown to be important for aversive learning and avoidance of P. aeruginosa (Zhang, Lu & Bargmann, 2005). Naive worms have a preference for P. aeruginosa over E. coli OP50, but learn to avoid virulent P. aeruginosa strains in binary choice assays if they were pre-exposed to P. aeruginosa for 4 h prior to the food choice experiment (Zhang, Lu & Bargmann, 2005). The avoidance behaviour is specific to the strains that were pre-exposed to the worms, and requires the serotonin neurotransmitter pathway (Zhang, Lu & Bargmann, 2005). Here we tested tph-1 worms for their preference of either the type III overexpression ΔexsE strain of P. aeruginosa or E. coli OP50. At an early time point (5 h), wild type N2 worms preferred P. aeruginosa ΔexsE, but preferred OP50 between 24 and 48 h exposure (Fig. 5A). The tph-1 mutant strain did not have any preference at 5 h, but also demonstrated a similar preference for OP50 between 24 and 48 h (Fig. 5A).

Given the rapid, aversive learning behaviour of C. elegans after only 4 h of pathogen pre-exposure (Zhang, Lu & Bargmann, 2005), we wanted to determine if the tph-1 worms were still capable of detecting preferred P. aeruginosa food sources over the time course of a few days in the feeding assay with a grid of 48 colonies. We tested a subset of the N2 preferred food sources and showed that tph-1 worms were still capable of detecting and eating those preferred sources to completion before moving on to eat the wild type PAO1 (Fig. 5B). This indicated that the serotonin chemosensory pathway was not required for C. elegans to differentiate the preferred food sources from the wild type PAO1 colonies that occupied the majority of positions in the food grid.

Preferred food sources were defective for full virulence inC. elegans slow killing assays

It is known that preferred bacterial food sources are associated with health benefits such as increased C. elegans life span and reproductive fitness (Abada et al., 2009; Freyth et al., 2010). We wanted to further establish a relationship between the preferred food status and virulence in C. elegans. We tested a panel of 11 mutants identified as preferred food sources (Table 2) for defects in slow killing of C. elegans. Interestingly, 45% of the preferred food sources tested showed significant defects in slow killing relative to the wild type PAO1 (Fig. 6). A pqsD mutant in the PQS biosynthesis pathway showed the strongest virulence defect, comparable to the non-pathogenic E. coli OP50 food source (Fig. 6). The pqs operon is required for the synthesis of the Pseudomonas quinolone signal, a quorum sensing signal molecule that contributes to slow killing and the red death phenotype in C. elegans (Zaborin et al., 2009; Feinbaum et al., 2012). PA4296 (PprB) is a two-component response regulator that regulates several adhesins, pili and fimbriae, the pqs operon, and is attenuated in the Drosophila acute and chronic infection models (de Bentzmann et al., 2012). PA0929–PA0930 encode a two-component system that has no known function, but is required for full virulence in C. elegans (Fig. 6). Lastly, PA4983 also encodes a two-component response regulator with no known function. These regulatory systems are excellent candidates as novel important regulators of virulence of C. elegans.