Two distinct sensing pathways allow recognition of Klebsiella pneumoniae by Dictyostelium amoebae

Authors

  • Wanessa C. Lima,

    Corresponding author
    1. Department for Cell Physiology and Metabolism, Centre Medical Universitaire, University of Geneva, Geneva 4, Switzerland
    Search for more papers by this author
  • Damien Balestrino,

    1. Laboratoire Microorganismes: Génome et Environnement (LMGE) UMR CNRS 6023, Faculté de Pharmacie, Clermont Université, Clermont-Ferrand, France
    Search for more papers by this author
  • Christiane Forestier,

    1. Laboratoire Microorganismes: Génome et Environnement (LMGE) UMR CNRS 6023, Faculté de Pharmacie, Clermont Université, Clermont-Ferrand, France
    Search for more papers by this author
  • Pierre Cosson

    1. Department for Cell Physiology and Metabolism, Centre Medical Universitaire, University of Geneva, Geneva 4, Switzerland
    Search for more papers by this author

  • The copyright line for this article was changed on 8 April 2015 after original online publication.

Summary

Recognition of bacteria by metazoans is mediated by receptors that recognize different types of microorganisms and elicit specific cellular responses. The soil amoebae Dictyostelium discoideum feeds upon a variable mixture of environmental bacteria, and it is expected to recognize and adapt to various food sources. To date, however, no bacteria-sensing mechanisms have been described. In this study, we isolated a Dictyostelium mutant (fspA KO) unable to grow in the presence of non-capsulated Klebsiella pneumoniae bacteria, but growing as efficiently as wild-type cells in the presence of other bacteria, such as Bacillus subtilis. fspA KO cells were also unable to respond to K. pneumoniae and more specifically to bacterially secreted folate in a chemokinetic assay, while they responded readily to B. subtilis. Remarkably, both WT and fspA KO cells were able to grow in the presence of capsulated LM21 K. pneumoniae, and responded to purified capsule, indicating that capsule recognition may represent an alternative, FspA-independent mechanism for K. pneumoniae sensing. When LM21 capsule synthesis genes were deleted, growth and chemokinetic response were lost for fspA KO cells, but not for WT cells. Altogether, these results indicate that Dictyostelium amoebae use specific recognition mechanisms to respond to different K. pneumoniae elements.

Introduction

In multicellular organisms, recognition of bacterial pathogens is essential to stimulate specific antibacterial responses and to allow the defence of the organism against infections. Mammalian and Drosophila cells make use of a collection of receptors that directly detect molecules exposed at the surface of bacteria (e.g. cell-wall components such as lipopolysaccharides or peptidoglycans) or secreted by bacteria (e.g. exotoxins or heat shock proteins) (Flannagan et al., 2009). In Drosophila melanogaster, the search for mechanisms involved in recognition of microorganisms has notably led to the discovery and characterization of Toll receptors that were later shown to also play a key role in bacterial recognition in mammalian organisms (Leulier and Lemaitre, 2008). Toll receptors are part of a signalling pathway ultimately controlling the transcriptional activity of NF-κB like transcription factors and the production of antimicrobial peptides. Recognition of bacteria by a eukaryotic cell and subsequent cell activation can be linked to initiation of phagocytosis (as in the case of lipopolysaccharides (LPS)-recognition by CD14 receptor, which triggers bacterial uptake and induces inflammatory response; Melendez and Tay, 2008), but in many instances different receptors are involved in recognition/activation and in phagocytosis of bacteria. For example, in mammalian cells, Toll-like receptors (TLRs) recognize bacterial LPS, but phagocytosis is achieved by the Fcγ and scavenger receptors upregulated by the same TLRs (Underhill and Gantner, 2004).

While specialized phagocytic cells are essential for elimination of invading microorganisms and for clearance of dying cells in multicellular organisms, in unicellular organisms such as the amoeba Dictyostelium discoideum phagocytosis is the main mechanism for acquiring food (Bozzaro et al., 2008). Dictyostelium amoebae exist as single cells in the soil, where they feed phagocytically upon bacteria. The molecular mechanisms that amoebae use to bind to and to ingest microorganisms are analogous to those found in multicellular organisms (Cosson and Soldati, 2008). They notably exhibit a cellular adhesion system composed of a surface receptor presenting features of integrin beta chains and binding to talin (Cornillon et al., 2006). Subsequent to recognition and binding, the microorganism is internalized, and ingested bacteria are killed by exposure to the harsh environment of the mature phagosomes, characterized by an acidic pH, the presence of proteolytic enzymes, and production of reactive oxygen species (Cosson and Soldati, 2008).

Previous studies have shown that Dictyostelium can be attacked by a variety of bacterial pathogens (Hilbi et al., 2007; Bozzaro and Eichinger, 2011; Lima et al., 2011; Steinert, 2011). On the other hand, when confronted with non-pathogenic bacteria such as non-virulent Klebsiella pneumoniae or Bacillus subtilis, Dictyostelium cells use specific molecular mechanisms to kill different bacteria (Benghezal et al., 2006; Lelong et al., 2011). In addition, large-scale transcriptional studies point to differential metabolic and signalling pathways being activated when Dictyostelium cells are grown in different sources of bacteria (Sillo et al., 2008; Nasser et al., 2013). It is thus likely that Dictyostelium recognizes various kinds of bacteria and adapts its physiology. To date, it is, however, not clear whether Dictyostelium amoebae simply adapt their metabolism to the nutrients that they can obtain from each bacteria, or if they specifically recognize and respond to each type of bacteria by modulating their physiology and gene expression. Several key elements allowing recognition of microorganisms in metazoans such as Toll receptors or NF-κB transcription factors are absent in Dictyostelium. On the other hand, some proteins like CD36 that have been shown to participate in microorganism recognition in metazoans have clear orthologues in Dictyostelium (Cosson and Soldati, 2008), although it remains to be seen if they play equivalent roles.

In this study we investigated if and how Dictyostelium cells specifically recognize bacteria and respond to their presence. Our results indicate that at least two distinct mechanisms allow recognition of K. pneumoniae bacteria by Dictyostelium, one sensing the folate produced by bacteria, the other one responding to bacterial capsule.

Results

FspA, a protein essential for Dictyostelium growth in the presence of Klebsiella pneumoniae

The fspA KO cell line described in this work was isolated from a population of random insertion Dictyostelium mutants based on their inability to grow on a lawn of K. pneumoniae bacteria. We used for this initial screen a non-pathogenic, non-capsulated laboratory strain of K. pneumoniae (unless otherwise specified, the abbreviation Kp is used to designate this laboratory strain). Wild-type (WT) Dictyostelium cells were able to ingest bacteria and to form phagocytic plaques in a lawn of K. pneumoniae (Fig. 1A and B). On the contrary, fspA KO cells grew very poorly in the presence of Kp (Fig. 1B), although they grew as well as WT cells in rich liquid medium (data not shown). The mutagenic plasmid inserted in the genome of fspA KO cells was recovered with its flanking regions, and the site of insertion sequenced, revealing that the fspA KO was caused by the insertion of the mutagenic vector 98 nucleotides downstream from the start codon of the DDB_G0277237 gene (Fig. S1).

Figure 1.

fspA KO cells are defective for growth on K. pneumoniae.

A. Schematic representation of growth of Dictyostelium cells on a lawn of bacteria. Cells growing on bacteria form a phagocytic plaque.

B. Growth of WT and fspA KO cells was tested on different strains of Gram-positive and Gram-negative bacteria (described in Table S2). WT cells were able to grow on all bacteria tested. fspA KO cells presented a specific growth defect on some Gram-negative bacteria (E. coli B/r and K. pneumoniae), but were able to grow on all the other strains tested.

C. Complementation of fspA KO cells with a vector harbouring FspA cDNA restored growth on K. pneumoniae. NT: not transfected cells.

To confirm that the phenotype of the KO cells was indeed due to this insertion in the fspA gene, we disrupted the gene by homologous recombination in WT cells (Fig. S1) and observed the same growth defect in three independent fspA KO clones, which were used for all further analysis. To further characterize fspA KO cells, we assessed their growth on a large array of bacteria, including representatives of both Gram-positive and Gram-negative bacteria (Fig. 1B). KO cells were able to grow on all Gram-positive bacteria tested (Bacillus subtilis, Micrococcus luteus and Staphylococcus aureus). They also grew on several non-pathogenic Gram-negative bacteria (Pseudomonas aeruginosa, Aeromonas salmonicida, and some Escherichia coli strains), but not on the laboratory strain of K. pneumoniae or on a mucoid strain of E.coli (strain B/r) (Fig. 1B). Growth of fspA KO cells on K. pneumoniae bacteria was restored by introducing in the KO cells a plasmid expressing FspA tagged with a Flag epitope (Fig. 1C).

The predicted FspA protein is 333 aminoacids long, and it exhibits one putative signal peptide and potentially nine transmembrane domains (Fig. 2A). The C-terminal segment of the protein (329–332) contains a putative di-lysine ER retention motif. FspA is conserved inside the Amoebozoa group, but no clear homologues can be readily detected outside this group by primary sequence similarity searches (using BlastP at the NCBI server). However, similarity searches using AmiGO server (release GO_20130330; Carbon et al., 2009) revealed a weak similarity (39%) with the SRE-19 protein of C. elegans, a chemosensory G protein-coupled receptor (GPCR) from the nematode serpentine family (Troemel et al., 1995). Phylogenetic analysis was conducted using orthologues from the C. elegans SRE family and orthologues from rhodopsin-like olfactory receptors from diverse metazoans (identified previously as having weak similarity to the nematode genes; Sengupta et al., 1996). According to this analysis, amoebozoan FspA orthologues lie closer to the nematode SRE family (Fig. 2B). In addition, a search for protein domains at PROSITE webserver (Sigrist et al., 2010) returned a low confidence hit to GPCRs family 1 (entry PS50262, from amino acid 1 to 190). The low levels of similarity among members of the GPCR superfamily are a known feature, and classification of proteins into the group often relies on traits other than the primary sequence. Prediction of protein function using specific dedicated webservers for GPCR signature profiling (as GPCR-pred, PCA-GPCR and GPCR-CA) confirmed FspA protein as a potential putative GPCR. Collectively, these observations suggest that FspA may indeed play a role in intracellular signalling, although given the low degrees of similarity observed, their functional significance remains unclear.

Figure 2.

FspA structure and evolutionary history.

A. Hydrophobicity plot (Kyte-Doolittle, window size of 13 residues) and predicted structure of the FspA protein. Based on primary sequence analysis, FspA exhibits a signal peptide (SP, red box) and 9 transmembrane domains (TM, grey boxes); a di-leucine ER-retention signal is present at the C-terminus (blue box).

B. Unrooted maximum likelihood tree of FspA with orthologues from nematode SRE family and metazoan rhodopsin-like olfactory receptors (refer to Table S3 for a complete list of protein accession numbers). Branch lengths are proportional to the number of amino acid substitutions per site. Numbers at the nodes represent the percentage of bootstrap support (only values > 50% are shown).

fspA KO cells ingest and kill bacteria efficiently

The same defective growth observed for fspA KO cells, i.e. a specific inability to grow on the laboratory K. pneumoniae strain and on the mucoid E. coli B/r strain, has been previously reported for two other Dictyostelium KO cells unable to kill ingested K. pneumoniae bacteria (phg1a and kil2 KO cells; Benghezal et al., 2006; Lelong et al., 2011). This led us to investigate if fspA KO cells can efficiently internalize and process K. pneumoniae. We first tested the ability of fspA KO cells to engulf different types of particles. Ingestion of latex beads, dead FITC-labelled or live GFP-expressing K. pneumoniae was not affected in fspA KO cells compared with WT cells (Fig. 3A), indicating that the KO cells retained their competence to ingest bacteria. fspA KO cells were also able to kill ingested K. pneumoniae bacteria as efficiently as WT cells (Fig. 3B), demonstrating that the defective growth phenotype was not linked to impaired killing.

Figure 3.

fspA KO cells internalize and kill bacteria efficiently.

A. Internalization of 1 μm latex beads, live Kp (GFP-expressing) or dead Kp (FITC-labelled) was measured by flow cytometry. Values were normalized to 100% of WT (grey line). Phagocytosis of beads or bacteria was as efficient in fspA KO cells as in WT cells.

B. Killing of bacteria was assessed by mixing Dictyostelium (WT, fspA KO and the killing-defective kil2 KO) cells and K. pneumoniae, and measuring the number of remaining live bacteria after different incubation times. WT cells eliminate K. pneumoniae rapidly, with only 25% of bacteria remaining after 2 h and less than 5% after 5 h. A similar curve was seen for the fspA KO cells. For comparison, kil2 KO cells were unable to eliminate bacteria efficiently: 80% of bacteria were still alive after 2 h, and almost 40% after 5 h. **P < 0.01, to WT values at each given time point.

Further characterization of the endocytic and phagocytic pathways did not reveal any other phenotypic anomaly. Notably, endosomal acidification (Fig. S2A), ingestion and recycling of fluid-phase (Fig. S2B), proteolysis of ingested substrates (Fig. S2C), and intracellular targeting of lysosomal enzymes (Fig. S2D) were not significantly different between WT and fspA KO cells. Together these results suggested that the inability of fspA KO cells to grow in the presence of K. pneumoniae was not caused by a defect in the function of the phagocytic pathway or in the ingestion and intracellular processing of phagocytosed bacteria.

Sensing of K. pneumoniae and of folate is defective in fspA KO cells

The impaired growth of fspA KO cells in the presence of K. pneumoniae could be linked to a defective recognition of the bacteria by the mutant cells. To test this, we developed a sensing assay based on the observation that WT Dictyostelium cells respond to the presence of different types of bacteria by significantly increasing their random migration speed (Fig. 4A). A similar method has been used previously to assess the activation state of Dictyostelium cells upon exposure to different compounds (Sroka et al., 2002; Rifkin and Goldberg, 2006). WT cells exposed to K. pneumoniae or to Bacillus subtilis bacteria migrated 2 to 3 times faster than unstimulated cells (Fig. 4B). Strikingly, in the same assay fspA KO cells failed to respond to the presence of K. pneumoniae (Fig. 4A and B), while they still reacted to B. subtilis (Fig. 4B).

Figure 4.

Sensing of K. pneumoniae and folate is defective in fspA KO cells.

A. Trajectories of WT and fspA KO cells migrating for 30 min in the absence or presence of K. pneumoniae (the origins are set to 0). Axes represent distance in μm.

B. Average speed of WT and fspA KO cells in the presence of bacteria (B. subtilis and K. pneumoniae), folate and its analogue methotrexate (MTX). WT cells responded to the presence of both bacteria, while fspA KO cells lost the ability to sense K. pneumoniae. fspA KO cells were also impaired in the sensing of folate and MTX. **P < 0.01, to control condition (in phosphate buffer).

C. Complementation of fspA KO cells with FspA restores the ability of cells to sense folate. **P < 0.01, to fspA KO values.

D. K. pneumoniae bacteria in exponential growth phase were treated with the antibiotic sulfathiazole (ST), an inhibitor of folate biosynthesis. K. pneumoniae growing in the same conditions, but in the absence of antibiotics, were used as control. WT cells did not respond to folate-depleted K. pneumoniae (the grey line indicates speed of WT cells in phosphate buffer). *P < 0.05, to non-treated Kp.

E. Induction of transcription upon exposure to folate is defective in fspA KO cells. Genes upregulated more than fourfold in WT cells upon exposure to folate (Table S1) were also analysed in fspA KO cells: in these cells, upregulation was not as effective as in WT cells, showing a 35–65% reduced induction. *P < 0.05 and **P < 0.01, to WT fold change values.

It has been known for several years that bacteria secrete folate, which acts as a chemoattractant for Dictyostelium (Burkholder and McVeigh, 1942; Pan et al., 1972). Folate and its analogue methotrexate also act as general activators of Dictyostelium and can induce a variety of cellular responses, notably an increase in random cell migration in WT cells (Rifkin and Goldberg, 2006). Unlike WT cells, fspA KO cells did not respond to folate or to methotrexate (Fig. 4B), and this phenotype was restored in KO cells expressing FspA-Flag (Fig. 4C). To verify the specificity of this phenotype, we also assessed the ability of kil2 KO cells to respond to K. pneumoniae or folate (Fig. S2E). As expected, kil2 KO cells are able to sense both K. pneumoniae and folate, indicating that recognition and killing of K. pneumoniae are two processes involving different cellular pathways.

In order to test whether folate was a major K. pneumoniae factor stimulating Dictyostelium cells, we attempted to produce folate-depleted Kp by incubating them in the presence of sulfathiazole, an antibiotic that inhibits bacterial folate biosynthesis by preventing the formation of the precursor dihydropteroate (Herrington, 1994). WT Dictyostelium cells did not respond to folate-depleted K. pneumoniae (Fig. 4D), suggesting that in these conditions, Dictyostelium cells may respond to the presence of K. pneumoniae bacteria mostly through folate sensing. We named the FspA protein on the basis of its essential role in folate sensing (Folate Sensing Protein A).

To further characterize the cell response to folate in WT and fspA KO cells, we analysed the transcriptional response of several genes after 1 h of incubation in phosphate buffer in the presence or absence of folate. The 50 genes analysed here (Table S1) were chosen partly based on microarray studies of Dictyostelium cells growing in the presence of bacteria (Farbrother et al., 2006; Carilla-Latorre et al., 2008; Sillo et al., 2008), but also included genes potentially involved in intracellular signalling, such as GPCRs. In WT cells, only 3 genes (all coding for GPCRs) were induced more than fourfold upon exposure to folate: grlC, grlD, and grlM (Fig. 4E and Table S1). In fspA KO cells exposed to folate, the expression levels of these three genes were 35–65% lower than in WT cells (Fig. 4E). This result also points to an inefficient response of fspA KO cells to folate.

Dictyostelium cells sense K. pneumoniae capsule in an FspA-independent manner

The laboratory strain of K. pneumoniae used in this study is characterized by its lack of pathogenicity, as well as by the lack of a visible capsule (Fig. 5A and Benghezal et al., 2006). Measurement of capsule content also revealed no detectable capsule production (undistinguished from background values). Most clinical or environmental isolates of K. pneumoniae exhibit a capsule, which represents one of the key virulence factors of K. pneumoniae (Podschun and Ullmann, 1998). In order to check if the presence of a capsule affects interactions between K. pneumoniae and Dictyostelium, we tested the ability of WT and fspA KO cells to grow in the presence of LM21, a well-characterized capsulated strain of K. pneumoniae (Favre-Bonte et al., 1999). Surprisingly, both WT as well as fspA KO Dictyostelium cells were able to grow in the presence of LM21 (Fig. 5B).

Figure 5.

An FspA-independent pathway allows sensing of K. pneumoniae capsule.

A. Electron micrographs of K. pneumoniae bacteria (laboratory Kp strain, LM21 and the LM21 capsule-defective Δwzx mutant). A capsule is visible at the surface of LM21, but not of other strains. Scale bar: 0.5 μm.

B. Growth of WT and fspA KO cells was tested in the presence of LM21 WT or of LM21 mutants devoid of LPS (ΔwecA) or of capsule (Δwza and Δwzx). WT Dictyostelium cells were able to grow on all bacterial strains tested. While fspA KO cells grew on LM21 and on an LPS mutant, they were unable to grow on mutants devoid of capsule.

C. Average speed of WT and fspA KO cells in the presence of LM21 WT and capsule mutants (Δwza and Δwzx). Dictyostelium WT cells moved faster in the presence of the three bacteria tested, but fspA KO cells did not respond to the capsule mutants. **P < 0.01, to control condition (phosphate buffer).

D. Average speed of WT and fspA KO cells in the presence of purified capsule from LM21 ΔwecA. Both WT and KO cells responded to capsule, as indicated by a threefold increase in speed when compared with control values. *P < 0.05, to control condition (phosphate buffer).

To determine the nature of this FspA-independent growth in the presence of LM21 K. pneumoniae, we tested several LM21 mutant strains, impaired in the production of different exopolysaccharides, capsule and LPS. LM21 Δwza and Δwzx, both devoid of capsule (Fig. 5A, and Balestrino et al., 2008), did not support growth of fspA KO cells (Fig. 5B), indicating that the presence of a capsule is necessary and sufficient to allow growth of fspA KO cells in the presence of K. pneumoniae. Accordingly, both WT as fspA KO cells presented increased motility in the presence of LM21 K. pneumoniae, but only WT cells could respond to the capsule mutants (Fig. 5C). To further characterize the effect of capsule, we assessed the ability of cells to respond to purified capsule extracted from an LPS-deficient strain (LM21 ΔwecA): both WT and fspA KO cells responded to the presence of an LPS-free capsule extract (Fig. 5D).

Together these results suggest that an FspA-independent mechanism allows Dictyostelium to sense K. pneumoniae capsule, and that this activation is sufficient to support growth of fspA KO cells in the presence of K. pneumoniae.

Intracellular localization of FspA

Since Flag-tagged FspA restored a wild-type phenotype in fspA KO cells (Figs 1C and 4C), at least part of this tagged protein was presumably correctly localized within the cell. When assessed by immunofluorescence, the bulk of FspA-Flag appeared colocalized with PDI (Fig. 6A), a marker of the endoplasmic reticulum. No significant colocalization was seen with a marker of endosomal compartments (p80 protein; Ravanel et al., 2001), contractile vacuole (Rhesus protein; Benghezal et al., 2001), or plasma membrane (H36) (Fig. 6A).

Figure 6.

FspA localizes mostly in the endoplasmic reticulum.

A. Colocalization of FspA-Flag with markers for endosomal compartments (p80), plasma membrane (H36) and endoplasmic reticulum (PDI). FspA colocalizes extensively with ER. Scale bar: 5 μm.

B. To assess the presence of FspA-Flag at the cell surface, plasma membrane proteins were biotinylated, purified on neutravidin beads, and the presence of FspA was revealed with an anti-Flag antibody in the total cell lysate and in the biotinylated protein fraction. The Western blot of a representative experiment is shown; all lanes are from the same gel and have the same exposure time. 20% of the total biotinylated fraction volume was loaded, and dilutions of the lysate (from 0.025% to 0.005% of the total lysate volume) were used to estimate the amount of biotinylated proteins. Only a very small fraction of the FspA protein was detected at the cell surface (< 1%). However, as a measure of background, a 5- to 10-fold lower fraction was detected at the cell surface for two negative controls: a cytoplasmic protein (actin) or an ER-resident protein (PDI) (Fig. S3).

However, we could not exclude based on immunofluorescence that a small fraction of FspA might be present at the cell surface. In order to investigate this point, the surface of intact cells was biotinylated and the presence of FspA-Flag was assessed among the purified biotinylated proteins. A small fraction of the FspA protein was biotinylated, corresponding to less than 1% of the total protein (Fig. 6B). This was however higher (5- to 10-fold) than detected for a cytosolic protein (actin) or a resident ER protein (PDI) (Fig. S3). This result suggests that, while a major part of FspA localizes in the ER, a fraction of it may be present at the cell surface. This surface localization of FspA is however tentative, since accidental rupture of a few cells during biotinylation can result in artefactual biotinylation of a small percentage of intracellular proteins.

Discussion

Dictyostelium sensing of bacterial-secreted folate

In metazoan organisms, recognition of pathogens is mostly mediated by cellular receptors that detect bacterial components and trigger specific signal transduction pathways. Environmental amoebae, such as Dictyostelium discoideum, rely on bacteria present on the soil as food source. To achieve this aim they must forage, internalize, kill and digest bacteria. Dictyostelium is able to grow on many different types of bacteria, and specific transcriptional responses have been observed in different conditions (Farbrother et al., 2006; Carilla-Latorre et al., 2008; Sillo et al., 2008; Nasser et al., 2013). However, most of these studies analysed steady-state situations, and to date it is still not clear if Dictyostelium specifically recognizes different types of bacteria and triggers defined adaptation pathways, or if the specific response occurs at a metabolic level, as an a posteriori adaptation to the different intracellular levels of available metabolites. In this study, we show that at least two distinct signalling pathways allow recognition of K. pneumoniae by Dictyostelium cells: one involves sensing of bacterial capsule, while the other relies on the detection of folate secreted by the bacteria (Fig. 7).

Figure 7.

Distinct pathways allow Dictyostelium amoebae to recognize bacteria. Our results suggest that Dictyostelium cells are able to recognize K. pneumoniae bacteria through at least two distinct pathways, one sensing bacterial capsule, and the other sensing folate. The FspA protein is an essential component of the folate-sensing pathway, but it does not play a role in capsule sensing. Recognition of bacterial components leads to different cellular responses: transcriptional regulation, motility and growth. At least one additional mechanism must be at play for the recognition of other bacteria like the Gram-positive Bacillus subtilis.

The isolation and characterization of fspA KO cells allowed us to show that, in order to grow in the presence of K. pneumoniae, Dictyostelium must specifically recognize these bacteria. When Dictyostelium is exposed to a laboratory, non-capsulated strain of K. pneumoniae, the sole activation pathway at play involves sensing of folate secreted by bacteria, and the FspA protein is an essential element for the activation of this pathway. The FspA-dependent folate-sensing pathway regulates the immediate response of Dictyostelium to folate (chemokinetic responses), a rapid transcriptional response (within 1 h of exposure), and it is essential for sustained growth in the presence of bacteria. Note that, when analysing transcriptional responses, we focused on relatively short-term responses (1 h), and it thus seems likely that we are measuring the primary response of Dictyostelium to folate, rather than a long-term adaptation process. This is in stark difference to most published studies that analysed cell response to bacteria after longer exposure times (from 2 h to 14 h or more) (Carilla-Latorre et al., 2008; Sillo et al., 2008; Nasser et al., 2013), and may account for the fact that none of the genes identified in these previous studies was upregulated upon short exposure to folate.

To date, our results do not allow us to define the exact role of FspA in folate sensing. Based on the phenotype of fspA KO cells, FspA may act as a folate receptor acting either at the cell surface or in the endoplasmic reticulum, where we detected most of the Flag-tagged FspA protein. This would be compatible with the observation that FspA is a polytopic membrane protein homologous to C. elegans GPCR chemosensory receptors (Fig. 2). Alternatively FspA may represent another essential element in the folate-sensing pathway, acting either as a regulator of another, unidentified, receptor, or in the regulation of the intracellular signal relay. It is known that the response of Dictyostelium cells to folate induces increases in cGMP content, indicating that the response is modulated by heterotrimeric G proteins, notably via Gα4 (Kesbeke et al., 1990; Hadwiger et al., 1994). Transcriptional response to folate includes a strong induction of a few GPCRs (Table S1), and FspA is partially involved in this response. FspA might be a first-line signal transducer, further activating other GPCR proteins also capable of sensing folate.

It is also worth noting the striking phenotypic similarity between fspA KO cells and two other KO cells previously isolated in our laboratory, the killing-deficient mutants kil2 and phg1a KO cells. The three cell lines have a very specific defective growth on E. coli B/r and the laboratory K. pneumoniae strain, but as discussed above the molecular mechanisms mediating growth impairment are different. This suggests the existence of a Klebsiella-adaptation module, in which FspA, Kil2 and Phg1A play different roles in sensing and killing.

Dictyostelium sensing of bacterial capsule

Our results imply that folate sensing is critical for sustained growth of Dictyostelium in the presence of a laboratory, non-capsulated strain of K. pneumoniae. However, fspA KO cells grew efficiently in the presence of LM21, a capsulated strain of K. pneumoniae. Analysis of LM21 mutants indicates that the presence of a capsule makes LM21 permissive for growth of fspA KO cells. The fact that fspA KO cells are unable to grow on, and also unable to respond to, capsule-deficient bacteria (as the Kp laboratory strain and LM21 Δwza and Δwzx mutants) points to a dual recognition mechanism of K. pneumoniae by Dictyostelium: in the absence of capsule, Dictyostelium relies solely on folate sensing for detecting bacteria, and this recognition is mediated via FspA. On the other hand, recognition of capsule is achieved by an alternative FspA-independent pathway (Fig. 7).

In K. pneumoniae, the capsule is one of the key virulence determinants, and it is often related to reduced ability of mammalian host cells to phagocyte bacteria (Podschun and Ullmann, 1998; Cortes et al., 2002; Evrard et al., 2010). In Dictyostelium amoebae, capsule-defective mutants of the capsulated Kp52145 were also killed more efficiently than the corresponding WT bacteria (March et al., 2013). When comparing the ability of Dictyostelium cells to grow on the capsulated LM21 strain or on its isogenic capsule mutants, WT Dictyostelium cells grew faster on the latter: it took 5–6 days for WT Dictyostelium to form a phagocytosis plaque on WT K. pneumoniae, but only 2–3 days on capsule-defective bacteria. This observation indicates that bacterial capsule plays a dual role in the interaction between K. pneumoniae and Dictyostelium: on the one hand it increases resistance of bacteria to intracellular phagocytosis and killing; on the other hand Dictyostelium has evolved a recognition system that allows it to sense the capsule and to modulate its response when facing these bacteria. Further genetic screens will be necessary to identify FspA-independent mechanisms allowing sensing of K. pneumoniae capsule or of Gram-positive bacteria.

Altogether, these results indicate that amoeba cells use a variety of mechanisms to sense different bacterial components and to grow in the presence of bacteria (Fig. 7). This is in accordance with the ecological niche that amoebae occupy: given the diversity of available food in the soil, cells should be able to recognize different bacterial components and to respond accordingly. Although previous studies, focused on transcriptional changes, already highlighted how the metabolism of amoebae adapts to different food sources (Carilla-Latorre et al., 2008; Sillo et al., 2008; Nasser et al., 2013), this study goes one step further with the identification of the molecular pathways involved on bacterial recognition.

Experimental procedures

Screening for growth-deficient Dictyostelium mutants

Mutants unable to grow on a laboratory strain of Klebsiella pneumoniae (Kp; Benghezal et al., 2006) bacteria were selected by restriction enzyme-mediated integration (REMI) mutagenesis (Kuspa, 2006) as described previously (Cornillon et al., 2000; Lelong et al., 2011). Mutants in which the plasmid insertion was in a coding region were selected, re-transfected with the recovered knockout plasmid (Fig. S1) and analysed again for defective growth. One mutant (herein denominated fspA) proved to be defective after this second screening round and was selected for further characterization.

Cell culture, plasmids and bacterial strains

All Dictyostelium strains used in this study were derived from the subclone DH1–10 (Cornillon et al., 2000) of the DH1 strain, referred to as wild-type for simplicity. Cells were grown at 21°C in HL5 medium and subcultured twice a week to maintain a maximal density of 106 cells ml−1.

A knockout vector was obtained during the mutagenesis, consisting of the REMI vector flanked by genomic regions comprising fspA gene (Fig. S1). The ClaI-digested plasmid was introduced into WT cells by electroporation, transfected cells were selected in the presence of 10 μg ml−1 blasticidin, and individual clones were screened by PCR (Fig. S1E). Three independent KO clones were generated and analysed in this study, and presented identical phenotypes.

An expression vector carrying a tagged version of the fspA gene was constructed by introducing a C-terminal Flag epitope (DYKDDDDK) into the FspA cDNA coding sequence and transfected in WT and fspA KO cells by electroporation; transfected cells were selected in the presence of 10 μg ml−1 G418.

The full list of bacterial strains used in this study is presented in Table S2. In all experiments, bacteria were grown in SM medium (10 g l−1 bacteriological peptone, 1 g l−1 yeast extract, 16.2 mM KH2PO4, 5.7 mM K2HPO4 and 4 mM MgSO4) supplemented with 1% glucose. For the study of bacteria morphology by electron microscopy, bacteria were fixed in 2% glutaraldehyde (1 h at room temperature) and post-fixed in 2% osmium tetroxide (1 h at 4°C), dehydrated and embedded in Epon resin and processed for conventional electron microscopy as previously described (Marchetti et al., 2004).

For capsule quantification, capsular polysaccharides were extracted from an overnight bacterial culture growing in DW medium (a defined medium designed for optimal capsule production) and quantified by measuring uronic acid concentration (compared with a standard curve of d-glucuronic acid) (Domenico et al., 1989; Balestrino et al., 2008). To produce folate-depleted K. pneumoniae, cells growing in exponential phase (OD 0.3–0.4) were washed twice in phosphate buffer and incubated for 2 h with 250 μg ml−1 sulfathiazole (Sigma-Aldrich).

Growth of Dictyostelium on bacteria and intracellular killing

The procedure to test growth of Dictyostelium cells on a bacterial lawn has been described previously (Froquet et al., 2009). Briefly, 103 Dictyostelium cells were deposited on a bacterial lawn (50 μl of an overnight culture) and allowed to grow until the WT cells form a phagocytic plaque. The number of days for such depends on the bacteria tested: Bs (4 days), Ml (4), Sa (4), As (2), PT531 (4), DH5α (2), EcBr (3), Kp (3), LM21 (5), LM21 Δwza (2), LM21 Δwzx (3), and LM21 ΔwecA (5).

Killing of bacteria was also assessed as described previously (Lelong et al., 2011). Briefly, 2 × 106 Dictyostelium cells were mixed with 2 × 104 Kp bacteria, and incubated at 21°C for up to 5 h. After 0, 1, 2, 3, 4 or 5 h of incubation, a 10 μl aliquot of the suspension was collected and diluted in 100 μl of 0.1% Triton X-100, plated in LB agar and the number of colonies counted to assess the number of live bacteria remaining.

Chemokinetic assay

For chemokinetic measurements, 5 × 105 Dictyostelium cells were allowed to attach on glass bottom 35 mm fluorodishes (WPI, World Precision Instruments, Sarasota, FL) for 20 min in phosphate buffer (2 mM Na2HPO4, 14.7 mM KH2PO4, pH 6.0) with or without addition of folate or methotrexate (2 mM), purified capsule (1 μg ml−1) or bacteria (1:100 v/v., from an overnight culture washed twice in phosphate buffer). Cells were then imaged for 30 min (pictures every 15 s) in a wide-field inverted Zeiss Axiovert 100 M, with a Plan-Neofluar 10× objective. The time series images were acquired with a Hamamatsu CCD cooled camera and assembled into a movie using Metamorph (Molecular Devices, Sunnyvale, CA). Particle tracking application for Metamorph was used to track the individual trajectories and total distance of at least 15 cells per experiment.

Endocytic pathway assessment

The internalization of phagocytic particles was measured as follows: 105 cells were incubated at 21°C in HL5 medium containing 1 μl ml−1 of 1 μm diameter FITC-fluorescent latex beads (Polysciences, Warrington, PA) or 107 fluorescent Kp (either live-GFP-labelled or dead-FITC-labelled), harvested after 30 min, washed with ice-cold HL5 containing 0.1% sodium azide, and analysed by flow cytometry (FACS Calibur, Beckton Dickinson, San Jose, CA).

Endosomal pH was measured by flow cytometry as described previously (Marchetti et al., 2009). The activity of lysosomal enzymes was measured using a colorimetric assay (Froquet et al., 2008) and the kinetics of phagosomal proteolytic activity using the un-quenching of DQ Green BSA-tagged beads (Sattler et al., 2013).

Fluorescence microscopy

To perform immunofluorescence, 106 cells were incubated on a glass coverslip for 3 h at 21°C, fixed in 4% paraformaldehyde, permeabilized with 0.07% Triton X-100 and labelled as already described (Charette et al., 2006). Cells were visualized with an LSM700 confocal microscope (Carl Zeiss). Mouse monoclonal antibodies against the endosomal marker p80 (H161) and the plasma membrane protein H36 and a rabbit antiserum against the contractile vacuole marker Rh were described previously (Benghezal et al., 2001; Ravanel et al., 2001; Mercanti et al., 2006). Endoplasmic reticulum marker PDI (protein disulfide isomerase) was detected using 221-64-1, a mouse monoclonal antibody (Monnat et al., 1997). Mouse monoclonal anti-Flag antibody (clone M2) was from Sigma-Aldrich, and fluorescent secondary goat anti-mouse or anti-rabbit IgG from Molecular Probes.

Surface biotinylation

To detect biotinylated FspA-Flag at the cell surface, 3 × 107 cells were washed twice in ice-cold phosphate buffer containing 120 mM sorbitol (pH 6.0), resuspended and incubated for 10 min in ice-cold phosphate buffer with 120 mM sorbitol (pH 8.0) containing 0.5 mg ml−1 sulfo-NHS-SS-biotin (Thermo Scientific, Rockford, IL), then washed once in phosphate buffer containing 100 mM glycine and 4 times in phosphate buffer containing 120 mM sorbitol (pH 6.0). Cells were lysed 15 min in lysis buffer (50 mM Tris pH 7.4, 150 mM NaCl, 1% Triton X-100, 1% sodium deoxycholate) containing protease inhibitors (2 mg ml−1 iodoacetamide, 2 μg ml−1 phenylmethylsulfonylfluoride, 1 μg ml−1 aprotinin, and 2 μg ml−1 leupeptin), and the supernatant was incubated overnight with Neutravidin Ultralink beads (Thermo Scientific, Rockford, IL). Beads were incubated 15 min in 6 M urea, washed 3 times in lysis buffer, heated at 60°C, and the eluate resuspended in 2× sample buffer (0.6 M sucrose, 0.1 M Tris pH 6.8, 10 mM EDTA pH 8.0, 1 mg ml−1 bromophenol blue, 4% SDS). Biotinylated proteins were separated on a 12% SDS-PAGE gel, transferred to nitrocellulose membranes, detected with a primary mouse anti-Flag antibody (Sigma-Aldrich) and a secondary HRP-coupled goat anti-mouse IgG (Bio-Rad, Hercules, CA) and revealed using a chemiluminescence imager (PXi Syngene, Cambridge, UK). As a positive control, a cell surface protein (SibA; Cornillon et al., 2006) was analysed. Cytoplasmic actin and ER-resident PDI (protein disulfide isomerase, Monnat et al., 1997) were used as negative controls.

RT-PCR

5 × 106 Dictyostelium cells were incubated for 1 h in phosphate buffer with or without addition of 2 mM folate. Cells were then harvested and RNAs purified with a Qiagen RNeasy kit. cDNA was synthesized from 1 μg of total RNA using random hexamers and Superscript II reverse transcriptase (Invitrogen). Amplicons (50 to 100 bp) were designed using the program PrimerQuest (Integrated DNA Technologies, Coralville, IO) with parameters adjusted for GC content (40%) and maximum TM difference (1°C). Oligonucleotide sequences were aligned against the Dictyostelium coding sequence database by blast to ensure that they were specific for the gene tested. Oligonucleotides were obtained from Invitrogen, and Table S1 shows a list of all genes analysed. PCR reactions (14 μl) contained SYBR Green Master Mix (Applied Biosystems), diluted cDNA (1 ng) and 500 nM of forward and reverse primers, and were analysed in a StepOnePlus cycler (Invitrogen) with the following parameters: 95°C for 20 s, 40 cycles of 95°C/3 s and 60°C/30 s. Raw CT values were imported into Excel (Microsoft Corporation), and fold changes were calculated as math formula, where ΔCT = CT (target) − CT (control genes: act and rnl) and Δ(ΔCT) = ΔCT (stimulated: 1 h in the presence of folate) − ΔCT (control condition: 1 h without folate).

Sequence and phylogenetic analysis

Sequence similarity analyses were performed using BlastP program against the protein databases deposited at NCBI, AmiGO and DictyBase servers (Johnson et al., 2008; Carbon et al., 2009; Fey et al., 2009). Signal peptide and transmembrane domains were detected using, respectively, SignalP and TMHMM (Krogh et al., 2001; Petersen et al., 2011). Protein domains and motifs were scanned using InterPro and PROSITE (Hunter et al., 2012; Sigrist et al., 2013), and prediction of primary sequences as GPCR proteins done via GPCR-pred (Bhasin and Raghava, 2004), PCA-GPCR (Peng et al., 2010) and GPCR-CA (Xiao et al., 2009) webservers. For phylogenetic analyses, protein sequences were aligned with CLUSTALX 2.0 program (Larkin et al., 2007). Maximum likelihood trees were computed by using MEGA 5.0 (Tamura et al., 2011), with the WAG+F model, and parameters for invariable sites and gamma-distributed rate heterogeneity (5 categories). One hundred bootstrap replicates were executed and bootstrap values drawn up on the consensus ML-tree.

Statistical analysis

Unless otherwise specified, for quantified data, the values represent the arithmetical mean and s.e.m. of at least three independent experiments. Statistical comparisons were done with Student's t-tests (two-tailed, unpaired).

Acknowledgements

We thank the personal from the Pôle Facultaire de Microscopie Ultrastructurale (PFMU) of the University of Geneva for the preparation of the electron micrograph samples. We also thank Thierry Soldati laboratory for providing the DQ Green BSA-tagged beads. PC laboratory is funded by a grant from the Swiss National Science Foundation (number 31003A-135789), and WCL was partially supported by a Téléthon Action Suisse grant.

Conflict of interest

The authors declare that they have no conflict of interest.

Ancillary