Bacterial culture

Pa^ATCC43949 was routinely grown in Lysogeny Broth (LB) with shaking (250 rpm) or on LB agar plates for 48 h at 28°C, unless otherwise stated. Where relevant, microaerobic conditions were produced by incubating agar plates in a 2.5L anaerobic jar with an AnaeroGen Sachet (Oxoid AN0025) at the indicated temperatures for 3–4 days, which lowers O₂ levels to less than 1%. Full anaerobic conditions were achieved using a Modular Atmospheric Controlled System anaerobic cabinet (DW Scientific). Haemolysis was determined on Tryptic Soy Agar base supplemented with 5% Sheep blood v/v (Oxoid). Carbon utilisation was determined using a modified version of Oxidation/Fermentation media (0.2% w/v Casein, 85 mM NaCl, 1.7 mM K₂HPO₄, 0.008% w/v Bromothymol blue, 0.8% w/v agar).

Thermal tolerance assays

Cultures were initially grown overnight at 28°C in LB media with aeration. Overnight cultures were then diluted to 10⁵ cfu/ml into fresh LB media and added to the wells of a sterile 96 well PCR plate. Using the gradient function on a thermocycler cultures were exposed to temperatures of 30–45°C for either 4 or 18 h. To assess viability from the large number of samples following heat exposure a kinetics-based method utilizing a bacterial growth curve was used. To do this heat-exposed cultures were diluted 1:100 into fresh LB within a sterile flat bottom 96-well plate (Costar, Corning). Cultures were then incubated at 28°C with orbital shaking for 24 h with optical density (OD₆₀₀) readings taken every 10 minutes using a BMG labtech microtitre plate reader.

Bioluminescence

Photorhabdus cultures were grown in LB broth at 28°C, 250 rpm for 16–18 h, diluted 1:100 in LB and 200 μl aliquoted in a 96-well “white” microtitre plate. Plates were incubated at the specified temperature in a Fluostar plate reader (BMG) with luminescence optic installed and regular measurements taken with continuous orbital shaking in between measurements for 24 h.

RNA purification

Pa^ATCC43949 was grown overnight in 10 ml LB at 28°C, 250 rpm, and sub-cultured (1:100 dilution) in 50 ml media at either 28°C or 37°C in 250 ml flasks for 4 h to mid-log phase (OD₆₀₀ ~0.6). To extract total RNA, a 10 ml aliquot of each culture was added to 25 ml RNAlater (Ambion) and centrifuged at 10,000 rpm, 4°C in a JA-25.50 rotor centrifuge (Beckman). RNA was isolated from bacterial pellets using the miRNeasy kit (Qiagen), with an on-column DNase treatment (Qiagen), and RNA eluted in 70 μl RNAse-free H₂0. To ensure complete removal of DNA, a subsequent DNase treatment was performed using the Turbo^™ DNase-free kit (Ambion). The concentration and integrity of RNA samples was determined with an Experion RNA StdSens analysis kit (Bio-Rad Laboratories).

RNA-seq library preparation

Ribosomal RNA was depleted from RNA samples using the Ribozero kit for Gram-negative bacteria (Epicentre). rRNA depletion was verified and samples quantified using a 2100 Bioanalyzer with a RNA 6000 Pico kit (Agilent). Strand-specific RNA-seq libraries were constructed using the Illumina compatible ScriptSeq mRNA-seq library preparation kit (Epicentre). cDNA libraries were quantified using a 2100 Bioanalyzer with a DNA 1000 kit (Agilent). To provide experimental robustness we constructed three RNA-seq libraries for each condition using RNA extracted from independent biological replicates and sequenced each using a different sequencing strategy. The first replicate of each library was sequenced in two separate lanes on the Genome Analyzer IIx (Illumina), with 36 bp paired-end reads at the University of Exeter Sequencing Service (Accession numbers SRR1555173 and SRR1555174). The second replicate of each library was sequenced on the HiSeq 2000 (Illumina), in two separate lanes with 70 bp paired-end reads at the University of Exeter Sequencing Service (Accession numbers SRR1555126 and SRR1555127). The third replicate of each library was sequenced on the HiSeq 2000 (Illumina), multiplexed with 10 libraries/lane with 100 bp paired-end reads (Accession numbers SRR1555150 and SRR1555151) by Source Bioscience (UK).

RNA-seq analysis

The analysis of the RNA-seq data was performed on linux servers running debian OS, with 96GB RAM. A comparative analysis was performed between treatments. For each treatment, the raw data in fastq format was converted to bfq format. The MAQ alignment software (version 0.7.1) [21] was used to align the Illumina data to the Photorhabdus asymbiotica ATCC43949 genome and plasmid gene models obtained from Genbank. Custom PERL scripts were then used to count the number of reads aligned to each gene and convert the data into a format suitable for statistical analysis. The data was normalised by reads per kilobase of exon per million mapped reads (RPKM). The RPKM measure was chosen as read density reflects the molar concentration of a transcript in the starting sample by normalizing for RNA length and for the total number of reads in the measurement. RPKM normalization enables comparison of transcript levels both within and between samples [22]. Treatments were then compared using DESeq, an R package that estimates variance-mean dependence in count data derived from RNA-seq experiments and tests for differential expression based on a model using the negative binomial distribution, to identify differentially expressed genes from different samples [23]. Using DESeq it was possible to generate text files containing the expression values for the samples, and a P-value for each gene to denote its expression difference between libraries. In addition the RNA-seq data was visualised using the methods described by Croucher et al [24]. Reads were aligned to the genome and plasmid of P. asymbiotica ATCC43949 using SSAHA2 (version1.0.9) [25]. The cigar2Coverage PERL script was used to convert the SSAHA2 output into a format compatible with the Artemis genome browser [26]. This allowed the mapped transcriptome data to be viewed, in a strand-specific manner, as a graph relative to the genome annotation.

Quantitative Real-Time PCR

Quantitative real time-PCR (qRT-PCR) was performed using the OneStep RT-PCR kit (Qiagen) as recommended by the manufacturer, with 100 ng of the appropriate RNA sample. PCR reactions were performed using the StepOnePlus Real Time PCR System (Agilent Biosystems) with reaction conditions as follows: 50°C x 30 min, 95°C x 15 min and 30 cycles of 95°C x 30 s, 58°C x 30 s, 72°C x 45 s. Cycle threshold (C_t) values were calculated using StepOne software v2.2 (Agilent Biosystems) and comparative analysis (ΔΔC_t) performed using the Relative Expression Software Tool (REST-2009, Pfaffl 2002) with rpoA and csrB as reference genes. qRT-PCR primers are listed in S13 Table.

2-Dimensional—Difference Gel Electrophoresis (2D-DIGE)

A 10 ml aliquot of each exponential phase culture used for the RNA purification for the RNA-seq experiment (see above) was centrifuged at 4500 rpm for 10 min. The cell pellets were sent on dry ice to Cambridge Centre for Proteomics. Cells were lysed in CHAPS/Thiourea buffer (6 M Urea, 2M Thiourea, 4% Chaps, 5mM Mg Acetate, 10mM Tris pH 8.5) by standard sonication. Samples were quantified using Quick Start^™ Bradford (ref²) and 50 μg of each sample was labeled with 250pmol Cy3 dye (GE Healthcare) for 30 min at room temperature, protected from light. A 50μg pool of all samples was labeled with 250 pmol Cy5 in order to normalize results from replicate gels. 10 mM lysine was added to quench the reaction and incubated for 10 min at room temperature. Equivalent concentrations of Cy3-labeled sample and Cy5-labeled pool were mixed and diluted in 2x Sample buffer (8M Urea, 4% Chaps, 2% DTT, 2% IPG buffer 3-10NL (GE Healthcare). Samples were incubated in the dark for 15 min and then diluted in De-streak rehydration solution (GE Healthcare). 2D-DIGE was performed as previously described [27].

Phenotype Microarrays

Pa^ATCC43949 and Pl^TT01 were grown from frozen glycerol stock on LB agar plates for 48 h at 28°C, a swab of culture was re-suspended in the appropriate media for each plate and adjusted to 85% transmittance measured using a Biolog turbidimeter. These cultures were diluted 1:100 in the appropriate media and PM plates were inoculated with 100 μl per well (three repeats per plate). We found that the recommended inoculating fluid (IF0a) supplied by Biolog interferes with the growth and respiration of Photorhabdus spp. Therefore, we used an adapted version of M9 media (12.5 mM Na₂HPO₄, 22 mM KH2PO4, 8.5 mM NaCl, 2 mM MgSO4, 100 uM CaCl2, pH 6.8) supplemented with 1x Dye A (Biolog). In addition, plates PM1 and PM2 were also supplemented with 0.5% (w/v) casein and plates PM3B, PM6 and PM7 were supplemented with 20 mM mannose. Plates PM1, PM2, PM3B, PM6, PM7 and PM8 were supplemented with 1x RPMI vitamin mix when incubated at 37°C. Plates PM9 and PM10 utilised the manufacturer’s medium (2.0 g of tryptone, 1.0 g of yeast extract, and 1.0 g of NaCl per liter). Plates were incubated at the indicated temperature in an OmniLog incubator, formazan formation was monitored every 15 min for 48 h and kinetic data was analysed using OmniLog-PM software (Biolog).

Secondary metabolite analysis

Pa^ATCC43949 was grown at 28°C and 37°C in LB medium upon addition of 10% (w/v) amberlite XAD-16 adsorber resin. Supernatant and XAD-16 were separated by decantation. After washing the adsorber resin the bound compounds were eluted upon incubating twice with 2 ml methanol at room temperature for 20 min and the solvent was removed under reduced pressure. Then the samples were re-dissolved in 0.5 ml methanol, centrifuged (10 min, 13500 rpm, room temperature), diluted 1:10 with methanol and analysed with a Dionex Ultimate 3000 system (Thermo Scientific, Dreieich, Germany) using an Acquity UPLC BEH C18 1.7μm RP column (Waters GmbH, Eschborn, Germany) coupled to an AmaZon X mass spectrometer (Bruker Daltonik GmbH, Bremen, Germany). Metabolites were gradually eluted from the column in 22 min (gradient from 5% to 95% acetonitrile in H₂O with 0.1% formic acid; flowrate of 0.6 mL/min).

Insect virulence assays

Caterpillar virulence assays were conducted essentially as previously reported [28]. Briefly, first day fifth instar Manduca sexta larvae raised on an antibiotic free artificial wheat-germ diet were injected into an ethanol (70% v/v) swabbed region (using a 27-gauge insulin needle (VWR) just above the first proleg with a 50 μl of a dilution (in 1xPBS) of mid-exponentially grown Pa^ATCC43949 bacteria (taken at OD₆₀₀ = 0.4 with aeration at 28°C). Caterpillars were placed on ice for 10 minutes prior to injection to reduce mobility. Two cohorts of 30 insects were injected with estimated 1000 cells each. One cohort was then placed at 28°C and the other at 37°C. The larvae were monitored over 7 days using physical stimulus to assess their status.

Complement activity assay

The Wielisa Total Complement System Screen (Wieslab), described by Seelen and co-workers [29], was used to detect inhibition of the classical (CP), mannose-binding lectin (MBLP) and alternative (AP) complement pathways by rhabduscin-containing supernatants from strains Pl^TT01, Pa^ATCC43949 and Pa^Kc. After 24 hours, the induced cultures were centrifuged at 3900 g for 20min and the supernatants were used in Wielisa complement activity assay. 6μl of rhabduscin-containing supernatant was added to 14 μl of concentrated human serum (positive control serum, supplied with the kit) and assayed for complement activity after pre-incubation at 30°C or 37°C for 30 min. The assay was completed in duplicate, according to the manufacturers instructions, and included a blank, a positive control (human serum from healthy individuals) and a negative control (heat inactivated serum). Complement activity inhibition was quantified from the absorbance at 405 _nm using the equation: (sample-negative control)/(positive control-negative control) x 100%. Rhabduscin analog was produced from heterologous expression of isnAB from P. luminescens^TTO1 in E. coli as described previously [30] and its structure was confirmed by detailed NMR analysis.

Serum resistance

Human serum (male) from type AB plasma was obtained from Sigma-Aldrich (Catalog number H4522). Overnight cultures of Pl^TT01, Pa^ATCC43949 and E. coli ^DH5α were diluted in LB or LB supplemented with 20% (v/v) human serum type AB, Heat Inactivated human serum type AB and 0.9% saline at a starting optical density of 0.05 at 600 nm. A total of 150 μl of each diluted cell suspension was added in triplicate into wells of a 96-well plate (Corning^® Costar). The assay was conducted at 28°C, shaking, and growth was monitored using an automated plate reader FluoStar Omega (BMG Labtech) with measurements taken every 30 min for 24 h.

Murine macrophage challenge

J774.2 (ECACC reference 85011428) murine macrophage-like cells (2x10⁶ cells per well on poly L-lysine-coated plastic coverslips) were washed twice with pre-warmed medium, and triplicate wells were infected with GFP labeled Pa^ATCC43949 (using the gfp-expressing plasmid pHC60) at a multiplicity of infection of ten bacteria per cell in a 5% CO₂ atmosphere. Bacteria and mammalian cells were centrifuged for 5 min at 800g to enhance bacteria—host interactions. After incubation at 37°C for 1 h, cells were washed five times with pre-warmed medium without serum and prepared for microscopic observation as follows. Cells were fixed with 4% (w/v) PBS-paraformaldehyde at room temperature for 1 h, washed three times with PBS and incubated with 50 mM ammonium chloride to minimize auto-fluorescence. Coverslips were further washed with PBS and the cells were permeabilized with 0.5% (v/v) Triton X-100 in PBS for 10 min. Non-specific binding sites were blocked with 0.5% (w/v) bovine serum albumin (BSA) in PBS. F-actin was stained with TRITC-phalloidin (Sigma). Stained cells were washed extensively with PBS, mounted overnight at 4°C with Mowiol and analyzed with a Zeiss confocal microscopy.

Manduca hemocyte in vivo challenge

Manduca sexta were individually reared as described [31]. Briefly, larvae were maintained individually at 25°C under a photoperiod of 17 hours light: 7 hours dark and fed on an artificial diet based on wheat germ. Manduca larvae (taken 1 day after ecdysis to the 5^th instar) were injected with GFP labeled Pa^ATCC43949 (using the gfp-expressing plasmid pHC60). After 4 hours the insects were bled (~100 μl) from the cut dorsal horn. Monolayers were formed on a coverslip from a suspension of ~5 x 10⁶ hemocytes ml^-1 in Grace’s insect medium (GIM) (Sigma), and observed in a Zeiss laser-scanning confocal microscope.

Biofilm visualisation

Visualisation of surface-associated biofilms was done by inoculating a chambered microscopy slide (μ-slide 8 well, Ibidi) with 300μl of mid log Pa^ATCC43949 containing the gfp-expressing plasmid pHC60, grown with aeration at 28°C in LB. Static cultures were then incubated for up to 24h in either 28°C in normal air, at 37°C in 5% C0₂ or at 37°C in normal air. Fluorescent images from washed slides were obtained using an inverted confocal microscope (Zeiss LSM510, 40x oil immersion objective).

Accession Numbers

NCBI (SRP045656) accession contains the experimental metadata and the individual fastq files for the triplicated data from the RNA-seq experiments.

Temperature tolerance of P. asymbiotica strains

We examined the ability of various Photorhabdus strains to survive an 18h exposure to various temperatures and their subsequent recovery and growth under static growth conditions at 28°C in LB medium (Fig 2). We also compared their growth dynamics after a shorter pre-exposure period of 4h (S1 Fig). We note that increasing the challenge temperature increased the lag times for recovery. The P. asymbiotica strains showed relatively short recovery times after exposure to 38.8°C compared to the temperature “intolerant” strains. Nevertheless there was great variation in the tolerance of strains. For example while Pa^ATCC43948 was not seen to recover at all after exposure to 38.8°C (within the time frame tested), the Pa^Goldcoast strain was not affected even by 45°C exposure for 4h. The reason for this temperature induced lag is the beyond the scope of this paper and the subject of a subsequent publication. Nevertheless it should be noted that while all clinical P. asymbiotica strains grow readily in rich LB medium during continued exposure to 37°C (from a low inoculum), none of the other strains tested here could sustain growth in this temperature regime.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Clinical Photorhabdus isolates are able to survive exposure to higher temperatures than most non-clinical isolates.

The optical density achieved by representative strains after overnight growth in static conditions (at 28°C in LB medium) after prior 18 h exposure to a range of temperatures. A range of clinical (N. American and Australian) and non-clinical (European) strains of P. asymbiotica (Pa) were tested, and the well-studied P. luminescens strain (Pl^TT01) was included for comparison. Green stars and red diamonds indicate thermal tolerance and intolerance respectively. Pa strain designations are indicated as superscripts.

https://doi.org/10.1371/journal.pone.0144937.g002

Temperature dependent differences in RNA abundance

If P. asymbiotica is specifically adapted to infect a mammalian host it is likely that the endothermic host temperature is a signal for the production of mammalian relevant pathogenicity determinants. To investigate this we performed strand-specific RNA-seq using triplicate biological samples of Pa^ATCC43949 RNA isolated from exponential phase cultures grown in LB medium with aeration. To facilitate mapping of the RNA-seq reads we chose to focus our studies on Pa^ATCC43949, as the genome has been fully sequenced and closed [16]. The annotated genome contains 4401 predicted ORFs on the chromosome and 29 annotated ORFs on the pPAU1 plasmid. Differential gene expression was analysed using DESeq and significant changes determined using cut-off scores of +/- 1.95 log₂-fold change, P-value ≤ 0.1 and minimum base mean read depth of 10 in the reference condition (S1 Data). This analysis found only 100 transcripts (with 45 of unknown function) significantly up-regulated at 37°C compared to 28°C (S1 Table), and 83 down regulated (with 31 of unknown function) (S2 Table).

Of the 37°C up-regulated chromosomal genes, the most strongly induced encode products involved in the acquisition and metabolism of amino acids and peptides. These include genes for a secreted metalloprotease prtA (a putative virulence factor, see discussion) and its exporter complex prtBCD (Fig 3), four solute-binding proteins of OppA-like oligopeptide ABC transporters, a predicted amino acid ABC transporter (PAU_03993) and several proteases/peptidases (D-aminopeptidase, enhancin and a putative thermostable carboxypeptidase (PAU_00753). We also note the up regulation of several genes involved in linking amino acids to central metabolism (asnA, AdoMet-synthetase and aroG) and vitamin biosynthesis (cbi genes). Several genes encoding virulence factors are significantly up regulated at 37°C including the potent mcf1 toxin [28], several genes of PVC-unit1 [32] and its putative effectors (PAU_02805/6), the insect toxin pirB [33], a putative “invasin” containing operon (PAU_02531–37) and a sepC toxin-like gene (PAU_0214). Secondary metabolism genes showing significantly higher expression at 37°C include; an operon containing genes similar to those used for the production of the antibiotic Fortimicin (PAU_01180–4), the luxCD substrate recycling genes, two NRPS-related genes (PAU_02219–20) and the isnAB-like genes (PAU_01720–21) used for the synthesis of the aglycon precursor of rhabducsin [30]. Several genes encoding chaperone and heat shock proteins (clpB, htpG/hsp90) were also significantly up-regulated at 37°C. Whilst DESeq did not indicate a significant increase in transcription (given the cut off criteria) of the chaperonin genes groL and groS, these genes do show some up regulation, although they are already amongst the most highly expressed transcripts at both temperatures suggesting their importance (S2 Fig). Interestingly the most abundant transcripts overall are from the tmRNA ssrA-PAU_01268 nucleoprotein complex. While these do not vary significantly with temperature the very high levels of expression suggest a high demand for re-cycling stalled ribosomes. We also note increased abundance of PAU_01563 mRNA encoding a predicted stress-response protein.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. The secreted metalloprotease PrtA is one of the most highly up regulated genes at 37°C.

An Artemis view of mapped RNA-seq data showing higher transcription of the prtA gene at 37°C compared to 28°C. A slight increase is also seen in the associated ABC transporter genes, prtBCD, and the predicted inhibitor gene inh.

https://doi.org/10.1371/journal.pone.0144937.g003

The majority of chromosomal genes down-regulated at 37°C encode products involved in the uptake and metabolism of carbohydrates (e.g. tctABC, lamB, mal-genes, pfkF) and nucleotides (e.g. purD, purH purL, purE, purK, guaA, guaB and xanthine/uracil permease PAU_00190). Several genes predicted to encode regulators of carbohydrate metabolism were also strongly down at 37°C, including the carbon starvation protein (cstA), a putative ner-like sugar fermentation stimulation protein regulator (PAU_03919) and the tricarboxylate transport two-component system (tctDE). Only a few putative virulence genes are down regulated at 37°C; those that are include those whose products show some similarity to type VI secretion system proteins (PAU_02313–02321), the xaxA insect toxin gene and a papH fimbrial regulator homologue. Finally an uncharacterised operon (PAU_03768–03772) that contains a homologue of the Streptomyces afs gene (involved in Streptomyces quorum sensing) is also strongly down regulated.

All P. asymbiotica strains so far analysed carry a plasmid with homology to pPAU1 [16], which shows limited homology to the pMT1 plasmid of Y. pestis essential for flea vectoring. This plasmid has not yet been seen in any strains of P. luminescens and P. temperata, and it is therefore tempting to speculate that it encodes determinants for human pathogenicity. A previous proteomic analysis comparing the secretome of Pa^ATCC43949 at 28°C and 37°C revealed a significant decrease in the abundance of the plasmid-encoded protein pPAU_0028 at 37°C. This protein has an unknown function however it does contain a bacterial Ig-like domain also found in invasins [34]. Nevertheless we could detect no significant difference in the transcription of any of the plasmid genes between the two temperatures using DESeq (S1 Data). Decreased secretion of the pPAU_0028-encoded protein is therefore a consequence of post-transcriptional changes.

Visualisation of the mapped RNA-seq reads in Artemis (S2 and S3 Data) revealed one of the most highly expressed transcripts is a putative sRNA (nt 721652–721861) which is in the same genomic location and has 85% nucleotide identity to csrB in P. luminescens. We also note expression of several other novel sRNA with homology to known sRNA in the rfam database. A transcript that shows homology to the spot42 sRNA, which regulates carbohydrate metabolism and uptake in E. coli is expressed at both temperatures in LB during exponential growth from the forward strand (nt 336528–336582) upstream of PAU_00301. A putative gcvB sRNA is also expressed at both temperatures from the reverse strand (nt c683067-683243) upstream of gcvA. The sRNA gcvB negatively regulates the translation of mRNA encoding solute-binding proteins of amino acid and peptide ABC transport systems in enteric bacteria [35]. Putative sRNA with homology to glmY (nt c1505935-1506092) and glmZ (nt c4794954-4795177) are also expressed at both temperatures. In E. coli, these sRNA act in a regulatory cascade to control the level of intracellular glucosamine-6-phosphate [36]. These novel transcripts were not included in the genome annotation used to perform the DESeq analyses, but visualisation of the plots in Artemis suggests no obvious deviation in any putative sRNA expression levels at the two temperatures.

We obtained further validation of the RNA-seq dataset using qRT-PCR to confirm the changes of several distinct transcripts as exemplars, using RNA extracted from additional independent biological replicates to add confidence to our interpretation of the data (S3 Fig).

Temperature dependent differences in protein abundance

Changes in relative mRNA abundance do not always correlate with protein abundance due to post-transcriptional regulation and differential protein stability/turnover. Therefore we also conducted triplicate 2D-DIGE proteome analyses [37] of protein extracted from cell lysates derived from the same exponential phase cultures used for RNA extractions. Six 2D-DIGE gels (three biological replicates for each temperature against pooled standard) were run and protein spots were identified and quantified using DeCyder. Graphical images of the individual 2D-proteomics gels, the DIGE-comparison gel and the DeCyder statistical analysis for each relevant protein spot are included in supplementary data (S5 Data). This analysis found 30 protein spots significantly more abundant at 37°C and 25 spots significantly less abundant at 37°C compared to 28°C (> = -1.8-fold difference with a T-test P value <0.01). Using LC-MS/MS we were able to unambiguously identify 15 proteins more abundant at 37°C than 28°C and 15 less abundant at 37°C than at 28°C (S3 Table). Proteins more abundant at 37°C were seen to be involved in chaperone activities (GroEL, GroES, DnaK ClpB), oxidative stress resistance (Bcp, Gst), autoinducer AI-2 catabolism (LsrF), lipopolysaccharide (LPS) biosynthesis (GmhA), iron acquisition (PAU_03286) and a secreted Asparaginase (AnsB). The majority of proteins less abundant at 37°C are involved in metabolic processes. These include proteins for maltose transport (MalE); purine metabolism (PurH, PurL, GuaAB); gluconeogenesis/anapluerosis (PckA, AspC, SfcA) and glycine detoxification/one carbon pool metabolism (GcvT). A protein potentially involved in immune evasion, a homologue of the lipid-A modification/ AMP resistance gene ArnA, also showed lower abundance at 37°C.

Temperature dependent differences in metabolism and stress tolerance

Our results from the RNA-seq and proteomics experiments suggest that P. asymbiotica alters its metabolism in response to host temperature. Whilst P. asymbiotica may be cultured easily at 28°C and 37°C in LB with aeration, our initial attempts to cultivate this strain in minimal media were problematic at 37°C. We therefore tested the ability of Pa^ATCC43949 to use different carbon and nitrogen sources and to tolerate a range of pH and osmolytes at the two host temperatures using the Biolog Phenotype Microarray system. We included the insect host restricted Pl^TT01 as a comparator. In order to use the Phenotype Microarray system it was necessary to make certain adaptations to the standard protocols. We replaced the standard IF0A media (Biolog) with a modified M9 salts medium (see methods) in plates PM01, PM02, PM3B, PM06, PM07 and PM08 as we found IF0A to be inhibitory to Photorhabdus. Furthermore it was necessary to supplement the carbon plates (PM01 and PM02) with a low concentration of Casamino Acids (0.05% w/v), as a mixed nitrogen source, as in these studies Photorhabdus could not utilise inorganic nitrogen. In the nitrogen (PM3B) and peptide plates (PM06, PM07, PM08) we supplemented 20 mM D-mannose as a carbon source. In addition, it was necessary to supplement the carbon, nitrogen and peptide plates with a 1x RPMI vitamin mix to support respiration at 37°C. Individual and combined (mean) respiration data replicates (per well of each plate) were separately visualised in a series of plots using the lattice package in R. From these, it was possible to obtain an overall, visual assessment of general trends of compounds that support respiration. For simplicity we have summarised the data as a qualitative analysis in S4 to S9 Tables, highlighting substrates that supported respiration above the level of the negative control wells.

This analysis showed that at 28°C both Pl^TT01 and Pa^ATCC43949 could utilise a diverse range of compounds as a sole carbon source for respiration at 28°C. However at 37°C Pa^ATCC43949 was seen to only utilise D-mannose, uridine, adenosine, inosine, L-serine or glycyl-L-proline as a sole carbon source. At 28°C, Pa^ATCC43949 could not utilise the majority of amino acids as a sole nitrogen source. Exceptions include L-glutamate, L-glutamine, L-tyrosine, L-serine and L-tryptophan. However when Pa^ATCC43949 was incubated at 37°C this list became further restricted to L-glutamate, L-glutamine and L-tyrosine. Interestingly we noted that Pa^ATCC43949 could use L-aspartate and L-asparagine as a sole carbon source at 28°C but not at 37°C while conversely it could use these amino acids as a sole nitrogen source at 37°C but not at 28°C. Other organic nitrogen sources that could be used as a sole nitrogen source by Pa^ATCC43949 at 28°C included N-acetyl-D-glucosamine, adenosine, cytidine and xanthine, with only xanthine being used at 37°C. The Phenotype Microarray analyses suggested Pa^ATCC43949 is only able to utilise a restricted subset of di- and tri-peptides as a sole nitrogen source at either temperature.

The RNA-seq and proteomics analyses show an increase in transcription and/or abundance of chaperones and stress response proteins at 37°C. To determine if the higher temperature affects the stress tolerance of P. asymbiotica we performed triplicate biological replicates of the Biolog osmolyte and pH plates (PM09 and PM10). The standard Biolog media recommended for these plates (IF10) is a similar composition to LB and therefore we could use this without additional supplements. Pa^ATCC43949 became more sensitive to osmotic stress and tolerated a narrower range of pH at 37°C (S10 and S11 Tables).

Temperature dependent differences in oxygen requirement and bioluminescence

Our RNA-seq analysis showed a significant increase in the expression of luxCD, which encode the enzymes that recycle the aldehyde substrate for luciferase. We tested the light emission from Pa^ATCC43949, Pa^Kingscliff and Pl^TT01 at 28°C and 37°C in LB medium with aeration. Light emission was shown to increase at 37°C for both clinical strains, which correlates with the increase in luxCD mRNA abundance. A surprising observation is that the temperature intolerant Pl^TT01 also emitted high levels of light up to around 10h when cultured at 37°C before apparently succumbing to temperature stress (S4 Fig). While the biological significance of bioluminescence in Photorhabdus remains unclear, conservation of the lux operon in all strains so far isolated argues for a selective advantage in nature. It is possible that bioluminescence protects Photorhabdus against ROS because the luciferase reaction consumes high levels of molecular oxygen. We therefore tested the ability of P. asymbiotica to grow in different oxygen environments at the two temperatures on various types of solid media. This showed that P. asymbiotica is unable to grow under full anaerobic conditions at either temperature on any media tested. Growth under oxygen-limited conditions (microaerobic) is poor compared to aerobic (normal air) at 28°C, and only full aerobic conditions support growth at 37°C (S5 Fig). We suggest that the increased consumption of oxygen by the luciferase reduces the availability of oxygen at 37°C thereby preventing microaerobic growth.

Temperature dependent differences in secondary metabolite production

All Photorhabdus genomes so far examined encode a large number of genes for enzymes dedicated to the production of secondary metabolites, suggesting natural products (NPs) are important to the EPN life cycle [38]. For example Pl ^TT01 encodes over 24 clusters comprising approximately 8% of the genome. Secreted NPs produced by Pa^ATCC43949 were compared from the supernatants of stationary phase cultures grown in LB with aeration for 48 h at 28°C and 37°C, using LC-MS/MS. This revealed very few differences in the overall profile of NPs produced at the two temperatures. Of the known NPs that have been characterised to date, we detected a 2.6-fold increase in the production of GameXPeptide [39] and a 1.5-fold increase in iso-propyl-stilbene (IPS) [40, 41] at 37°C compared to 28°C. While the DESeq analysis did confirm up regulation of transcription of the GameXPeptide synthesis gene (gxpS) at 37°C (S6 Fig and S1 Table) the slight increase in the transcription of the IPS synthetic genes fell below the 1.95 log2fold cut off (S7 Fig). The natural product rhabduscin, produced by several Photorhabdus and Xenorhabdus strains, was recently shown to inhibit the insect Phenol-Oxidase (PO) cascade [30]. Furthermore homologues of key rhabduscin synthetic genes, isnAB, can also be identified in many diverse bacteria (S4 Data). While RNA-seq showed a significant increase in the expression of the isnAB-like rhabduscin synthesis genes at 37°C (Fig 4A), we did not detect increased secretion of rhabduscin itself. Crawford et al [30] demonstrated that the majority of rhabduscin localised to the cell surface, which may explain why we did not detect it directly in the clarified supernatants tested. We hypothesised that the decoration of the cell surface with this PO inhibitor might also afford protection against components of the human innate immune system analogous to the insect PO cascade. We therefore investigated whether heterologously produced and purified aglycon precursor of rhabduscin could offer protection against human complement, which like PO also constitutes a serine protease cascade. Our results confirmed that the aglycon precursor alone (synthesised by IsnAB) does act as a potent inhibitor of the mammalian alternative complement pathway (Fig 4B). Furthermore cell-free supernatants from all Photorhabdus strains we tested were also able to specifically inhibit the alternative pathway of complement (Fig 4C) suggesting rhabduscin is released in sufficient quantities from the cell surface.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. The expression and function of the Photorhabdus natural product rhabduscin.

(A) Artemis views of the RNA-seq reads of the three replicates mapped onto the Pa^ATCC43949 operons responsible for rhabduscin synthesis. The isnAB genes are responsible for synthesis of the aglycon precursor shown above the left panel. The PAU_02755–7 genes encode glycosidase enzymes that add the sugar groups to produce the final rhabduscin molecule. Note PAU_02756 is unique to the P. asymbiotica (replaced by a transposase in Pl^TT01) and so the final Pa^ATCC43949 rhabduscin structure from Pa^ATCC43949 may not be the same as that shown from Pl^TT01 (above the right panel). (B) The purified aglycon precursor of rhabduscin (shown above the key) is able to completely inhibit the human alternative complement pathway. (C) Cell free supernatants from Pa^ATCC43949 (PaATCC43949), Pa^Kingscliff (Pa Kc) and Pl^TT01 (Pl TT01) can all inhibit the human alternative complement pathway (AP). Note the classical (CP) is only partially inhibited, while LB alone also inhibits the Maltose binding lectin (MBLP) pathway to some extent.

https://doi.org/10.1371/journal.pone.0144937.g004

As complement is not the only means of killing bacteria by human serum we decided to also test the ability of Photorhabdus to resist killing by commercially available serum. This is an important prerequisite for the survival of P. asymbiotica during human infection. Unlike the E. coli control both Pa^ATCC43949 and Pl ^TT01 could resist killing by 20% (v/v) commercially available human serum (S8 Fig). In addition we also tested the ability of Pa^ATCC43949 and Pl ^TT01 to resist killing by pig and rabbit serum (S9 Fig). Interestingly while Pa^ATCC43949 was resistant to both, the growth of Pl ^TT01 was strongly inhibited by pig serum suggesting Pa^ATCC43949 is tolerant to serum from a greater range of mammals.

Temperature dependent differences in insect virulence

Our results show that growth of Pa^ATCC43949 at 37°C restricts its ability to utilise a range of nutrients and increases sensitivity to pH and osmolyte stress. It also leads to an increase in the production of specific virulence factors and secondary metabolites. To determine how these changes affect the ability of P. asymbiotica to infect insects, we performed standard infection assays of Manduca sexta larvae at 28°C and 37°C [28]. As expected, 100% of the cohort of 5^th instar larvae (n = 30) were killed within 48 h by an injected low dose of approximately 1000 cells at 28°C. Importantly however, when the infected insects were incubated at 37°C the bacteria became avirulent and the whole cohort survived for up to 7 days.

Temperature tolerance

The ability of the clinical strains to tolerate growth at 37°C or above is unusual in the genus as, with the exception of certain P. luminescens strains, the majority of Photorhabdus cannot replicate above 34°C. This is clearly a prerequisite for bacteraemia and the establishment of a sustained human infection. Previous MLST and a recent whole genome phylogeny analysis of eight strains of P. asymbiotica and three strains of P. luminescens (using predicted proteomes) suggest a monophyletic origin of the clinical strains and their pPAU1-like plasmids (data not shown). Nevertheless it cannot be ascertained from this study whether 37°C tolerance was inherent in the last common ancestor of all the clinical strains, or whether it evolved independently on the different continents. The identification of clinical isolates from USA, Australia and Nepal (Pa^Nep) and a clinical isolate-like strain from Thailand (Pa^Thai), in addition to the two closely related temperature intolerant strains from northern Europe (Pa^Hit and Pa^Jun), does suggest the P. asymbiotica phylogenomic group are globally widespread. The higher abundance of Pa^ATCC423949 heat shock chaperone at 37°C is not surprising, although at this point it is unclear if these levels indicate a stress-response per se or an appropriate temperature specific induction. However the decreased tolerance to both salt stress and a wider pH range does suggest a lower integrity of the cell membrane and/or an impairment of osmotic homeostasis at the higher temperature. Importantly the observation that Pa^ATCC423949 becomes avirulent to M. sexta at 37°C suggests that it is using temperature as a cue to orchestrate changes in transcription and translation appropriate for human infection but inappropriate for insect virulence. The alternative hypothesis that the higher temperature causes dysfunction of key proteins or pathways is unlikely given their ability to cause human infection.

Immune evasion

It is generally accepted that insect immune systems share a great deal in common, both mechanistically and genetically, with the innate immune system of mammals. Upon entry into the human host P. asymbiotica must be able to resist immediate killing by both humoral and cellular arms of the innate immune system.

Other secondary metabolites

Pa^ATCC43949 shares only 15 of its 25 known operons for natural product synthesis with Pl^TT01 suggesting niche or host specific roles for many of these. In addition to IPS and rhabduscin we also see significant increase in transcription of the non-ribosomal-peptide-synthesis gene gxpS (PAU_03067), which produces the natural product GameXPeptide. While the exact function of this is unknown preliminary work suggests a role in which immune suppression (unpublished).

Changes in metabolism as a nutritional virulence strategy

Previous studies have indicated that Photorhabdus employs a metabolic switch to regulate its transition from insect pathogen to nematode mutualist [45]. The significant reduction in the range of metabolic capabilities of Pa^ATCC43949 at 37°C could be the result of an adapted and regulated response, a consequence of unintended pathway failures, or a combination of both. Nevertheless, when taken together the RNA-seq, proteomics and phenotype microarray data is suggestive of a regulated response.

One possible reason for the observed changes is that by reducing the levels of enzymes involved in the metabolism of pyruvate, PEP and acetyl-CoA, the cells could deliberately slow down metabolite flux feeding into the TCA cycle at 37°C. When taken together with the reduction in the one-carbon pool NADPH production implied by the data, it may indicate that the bacteria are attempting to counter unavoidable and undesirably high metabolism arising from increased enzyme rates at the higher temperature. Such an explanation would imply that Pa^ATCC43949 is not well adapted to the mammalian temperature regime. Nevertheless, we argue that a more likely interpretation of the data is that the bacteria are using 37°C as a cue to deploy a “nutritional virulence” strategy, as seen in several other human pathogens. This is a strategy by which pathogenic bacteria selectively use certain metabolites in order to deprive the host of molecules necessary for resisting infection [57, 58]. We will expand on this hypothesis below. Fig 5 illustrates some of the relevant metabolic pathway changes we discuss.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. A schematic summarising some key differences in metabolism at 37°C compared to 28°C, centred on glutamate/asparagine metabolism and the TCA cycle.

This model is predicted by integrating data from the RNA-seq, proteomics and phenotype microarray studies. Intermediates (boxes) and pathways (arrows) predicted to be down regulated at 37°C are in red while those up regulated are in green. Data suggests TCA cycle intermediates (back boxes) would be relatively isolated from glutamate/asparagine metabolism and could be maintained via the conversion of L-serine into citrate via pyruvate. Black arrows indicate certain potential enzyme pathways that are present and predicted to be unchanged at 37°C. The data suggests a central role for imported peptides and amino acids in metabolism at 37°C. Opp/Dpp represent oligo- and di-peptide importers, TCT represents tricarboxylic acid and PEP is Phosphoenolpyruvate.

https://doi.org/10.1371/journal.pone.0144937.g005

Quorum sensing?

It has recently been discovered that P. asymbiotica can use dialkylresorcinol molecules for QS [77]. These molecules are synthesised by the products of the darABC operon (PAU_02400–02402) [78], which are also involved in synthesis of the IPS molecule (S7 Fig). In QS, the dialkylresorcinol molecules are sensed the PauR receptor (PAU_4062), which is a LuxR like protein, that once activated induces transcription of the pcfABCDEF operon (PAU_04063–68). Transcription of this operon induces a cell-clumping phenotype which has been shown to contribute to high pathogenicity of P. asymbiotica toward insect larvae [77]. Interestingly in Pl^TT01 a close homologue of PauR, named PluR, senses α-pyrones as QS signals which again induce the same cell clumping response from the Pl^TT01 pcfABCDEF operon [79]. It has been suggested that the use of dialkylresorcinol molecules instead of α-pyrones, which are not produced by P. asymbiotica strains, as quorum sensing molecules might be an important evolutionary step from invertebrate to human pathogenicity [77]. Nevertheless neither the Pa^ATCC43949 pcfABCDEF nor pluR genes show any significant difference in expression between the two temperatures. However we do see changes in the expression of other genes that might be suggestive of alternations in QS activity. For example, the AI-2 pathway genes lsrFG are induced at 37°C which would act to degrade the AI-2 signal if present. Interestingly, in Pl^TT01, knocking out the AI-2 synthesis gene luxS increases the oxidative stress response [80], which may indicate a role for this QS system in Pa^ATCC43949 as it attempts to adapt to a stringent mammalian host environment. More speculatively, an operon encoding genes homologous to those for the synthesis of a known Streptomyces QS is also strongly repressed (PAU_03767–03778) again suggesting that QS systems relevant to insect infection may be inappropriate to the human host environment. There is an interesting parallel here to the necessary loss of the plcR QS system of Bacillus anthracis in its adaptation to the mammalian host from the invertebrate associated B. cereus ancestor [81].

Temperature sensing and regulators

While we have described in detail the changes in gene expression at 37°C, we cannot yet attribute this to any specific regulators. It is interesting that many of the down-regulated regulators are luxR gene homologues. However, without further experimentation we cannot determine if changes in the expression of these plays any role in human virulence. Nevertheless this remains a formal possibility [82]. For example any one could act as a repressor, the down-regulation of which could lead to the activation of mammalian specific genes at 37°C. Alternatively down regulation or dysfunction of certain response regulators could “blind” the cell to the presence of otherwise abundant nutrients. This would have the effect of forcing them into becoming nutritional specialists and therefore facilitating a nutritional virulence strategy.

Many diverse mechanisms for bacterial temperature sensing have been described. These range from RNA-thermometers [83] to temperature dependent conformation changes in specific regulator proteins. A good example of the latter is the intrinsic thermal sensing by the Yersinia virulence regulator RovA [84], which becomes proteolytically cleaved upon thermally induced conformation change. We actually see a slight increase in the expression of the Pa^ATCC43949 slyA gene (a rovA homologue) at 37°C although this is not significant by our criteria. A recent study in Y. pseudotuberculosis revealed that, like Pa^ATCC43949, they also reprogram a large set of catabolic/energy production genes in response to temperature increase. In the Yersinia the effects are mediated through a massive remodelling of the cAMP Regulatory Protein CRP-controlled network [85]. How this is accomplished has not yet been fully determined, however a previous study revealed many of virulence-associated functions are co-regulated with carbohydrate metabolism. This is mediated through the carbon storage regulator system Csr, involving the csrB and csrC regulatory sRNAs and CRP [86]. In the Yersinia the Csr system consists of an RNA binding protein CsrA that binds the Shine-Dalgarno regions of target mRNAs to repress and destabilises them. CsrA itself can be bound and sequestered by either csrB or csrC. Normally CsrA inhibits the production of the RovA virulence regulator, so increasing the expression of csrB leads to the de-repression of RovA and activation of several virulence factors. It is a CRP-mediated repression of the response regulator UvrY that leads to the activation of csrB transcription. Interestingly UvrY has been shown to be essential for the adaptation of Photorhabdus to the insect host [87]. Knock out the uvrY gene in Pl^TT01, resulted in changes in gene expression and phenotype that showed much in common with Pa^ATCC43949, grown at 37°C. For example the UvrY deficient Pl^TT01 strain showed increased transcription of groESL and luxCD, together with an increase in motility and iron transport and a decrease in carbohydrate transport. Conversely, the uvrY mutant also showed expression changes not consistent with Pa^ATCC43949 growth at 37°C. For example Pl^TT01 uvrY induces transcription of components of the AI-2 QS system, which would be theoretically repressed in Pa^ATCC43949 due to the increase in the AI-2 degrading enzymes, LsrFG. Pl^TT01 uvrY also increases csrB and represses tldD protease expression, which together would lead to the repression of CsrA activity. In our experiments we see no change in either csrB or tldD transcription. Nevertheless Pl^TT01 uvrY also becomes avirulent to insects consistent with Pa^ATCC43949 at 37°C.

Our challenge now is to determine the key regulators that allow Pa^ATCC43949 to exhibit such large differences in temperature dependent phenotype despite relatively limited changes in gene expression.