Abstract
Professional phagocytes detect, pursue, engulf, and kill bacteria. While all professional phagocytes use chemotaxis to locate bacteria, little is known about whether they can discriminate among them, responding preferentially to some bacteria over others. Here we examine the chemotaxis of the soil amoeba and professional phagocyte Dictyostelium discoideum in assays where amoebae were presented with a paired choice of different bacteria. We observed variation in the extent to which they pursue different types of bacteria and preferential migration towards Gram-negative over Gram-positive bacteria. Response profiles were similar for amoebae isolated from different geographic locations, suggesting that chemotaxis preferences are not strongly influenced by any local variation in the bacterial community. Because cyclic adenosine monophosphate (cAMP) is a known chemoattractant for D. discoideum, we tested whether it mediated the preference for Gram-negative bacteria. Chemotaxis was diminished in response to a cAMP-deficient strain of Escherichia coli and enhanced in response to an E. coli strain that overproduces cAMP. We conclude that D. discoideum discriminates at a distance among bacteria and that discrimination is mediated in part by sensing of cAMP. Preferential sensing and response to different bacteria may help to explain why some amoeba-bacterial associations are more prevalent in nature than others.
Background
There is a growing appreciation that nearly all multicellular organisms are colonized by large numbers of bacteria that approach or even outnumber an organism’s own cells [1–3]. These bacteria can influence everything from a host’s physiology to its development and behavior [reviewed in 4–6]. Despite the near universality of colonization by bacteria, however, little is known about the extent to which hosts can selectively maintain or remove different bacteria— which in turn requires the ability to sense and respond preferentially to some bacteria over others.
Professional phagocytes include motile cells of the mammalian immune system, such as neutrophils or macrophages, that locate and engulf bacterial invaders [7–9]. However, phagocytosis is also used by free-living amoebae to hunt and kill bacteria for food [10,11]. Phagocytosis involves invagination of the cell membrane, engulfment, and internalization of a bacterial cell or object. The resulting vacuole, called a phagosome, subsequently fuses with a lysosome that contains a variety of hydrolytic enzymes and usually results in death and digestion of the bacteria. Because the molecular basis of phagocytosis is conserved between amoebae and mammals, many bacteria that can evade intracellular killing and replicate inside immune phagocytes also infect and kill free-living amoebae [12,13]. Indeed, amoebae are thought to be the primary reservoir for these bacterial pathogens in nature and potentially a natural predator that selected for their pathogenic features [14–20].
Phagocytosis is a specific process, requiring host receptors that recognize and bind ligands on the bacterial cell surface [21–25]. This specificity alone leads to some degree of discrimination on the part of phagocytes in their ability to attach to, engulf, and kill bacteria they encounter [26]. In the soil amoeba Dictyostelium discoideum, for example, distinct sets of genes are required for killing and digestion of Gram-negative versus Gram-positive bacteria, and different genes respond transcriptionally to phagocytosis of these two classes of bacteria [27].
In addition to binding and engulfing bacteria once they have been encountered, professional phagocytes show a variety of chemotactic or “chase” behaviors [28] that enable them to detect and respond to bacteria at a distance (e.g., neutrophil chase [29]). In some cases, the bacteria- or host-produced chemoattractants have been identified. For example, N-formylated peptides, produced by bacteria, are chemoattractive to neutrophils [30,31]. Macrophage and neutrophils are also attracted to components of the complement system that are released by other cells upon contact with bacteria [32–34].
Despite work on immune phagocyte discrimination upon physical contact, there have been few attempts to demonstrate that phagocytes can discriminate among different types of bacteria at a distance, allowing them to selectively pursue certain bacteria, despite the importance of any such discrimination for outcomes such as clearance, infection, or colonization. In one notable exception, previous work showed that neutrophils will extend more pseudopods towards E. coli and Salmonella enterica serovar Typhimurium than they do towards S. enterica serovar Typhi. It was also shown that expression of the Vi capsular polysaccharide by the latter impedes neutrophil chemotaxis and leads to the discriminatory response [35].
Dicytostelium discoideum is soil amoeba and major model system for phagocytosis, chemotaxis, and host-pathogen interactions. As a single-celled amoeba, it phagocytoses bacteria for food during the unicellular stage of its life cycle. In response to starvation, it undergoes aggregative multicellular development, mediated by cyclic AMP signaling, first forming a migratory slug, and later a fruiting body that consists of a ball of spores atop a stalk of dead cells. Many bacteria that infect and kill macrophage or neutrophils also infect and kill Dictyostelium [12,36–39], suggesting that bacteria can function as both predator and prey to D. discoideum. Early studies by Konijn and colleagues [22] demonstrated that D. discoideum will respond chemotactically to E. coli and possibly other species of bacteria, but whether the amoebae discriminate among different bacteria and respond selectively is not known. Moreover, while it was long thought that the multicellular stages of the life cycle (the slug and fruiting body) were sterile, a recent study showed that some D. discoideum strains are colonized by bacteria [40]. Additional work demonstrated variable impacts of these colonizing bacteria on the amoebae that range from beneficial to commensal [41–43]. These diverse impacts of bacteria on amoebae, in addition to its reliance on bacteria for food, makes D. discoideum a useful model system to test for the possibility of discrimination at a distance.
To determine whether D. discoideum can discriminate among different bacteria, we performed behavioral choice assays, where amoebae were placed equidistant between different species of bacteria, and we quantified the movement by the amoebae to each one. Our experiments consisted of pairwise combinations of four Gram-negative and four Gram-positive species. We chose Gram-negatives Klebsiella pneuomoniae and E. coli because these bacteria are high quality prey for D. discoideum, which can devour them rapidly and develops well on lawns of both species. Gram-negatives Pseudomonas fluorescens and Burkholderia spp. have both been repeatedly isolated from D. discoideum fruiting bodies [40], and we hypothesized that the amoebae might show greater sensitivity to, or even preference for, these bacterial species. Indeed, the particular isolate of Burkholderia we used was isolated from a fruiting body of a D. discoideum natural isolate in the lab. Among the Gram-positives, Bacillus subtilis is a common soil bacterium that D. discoideum might encounter. Microccocus luteus is likewise common in the environment, but also found on skin. Enterococcus facaelis is a Gram-positive enteric bacteria that Dictyostelium could potentially encounter in nature, since naturally occurring D. discoideum fruiting bodies have only been found on dung [e.g., 45]. Finally, Staphylococcus aureus is a Gram-positive bacterium commonly found on the skin and in the respiratory tract. Overall, we chose a range of bacteria that likely vary with respect to the extent of D. discoideum’s prior experience with them during its evolutionary history.
Finally, we tested three different D. discoideum genotypes from distinct geographic regions (Massachusetts, Texas, and Tennessee) that comprise a large portion of the species’ natural range, which is primarily the eastern half of the United States, as well as parts of central America and east Asia [45]. The degree to which chemotaxis behaviors are similar or different among geographically and genetically diverged D. discoideum can help to determine whether bacteria-seeking behaviors are evolutionarily malleable and responsive to natural variation in the bacterial community across environments.
Methods
Dictyostelium Strains and Culture Conditions
D. discoideum strains QS32 (Houston, Texas), QS39 (Indian Gap, Tennessee), and QS40 (Mt. Mitchell, Massachusetts) were spotted from frozen spore stocks onto lawns of Klebsiella pneumoniae bacteria on SM agar plates (per liter: 10 g glucose, 10 g Bacto Peptone (Oxoid), 1 g yeast extract (Oxoid), 1g MgSO4, 1.9 g KH2PO4, 0.6 g K2HPO4, 20 g agar). Following germination and growth, amoebae were harvested from a leading edge and inoculated into petri dishes containing 10 ml of HL5 medium (ForMedium Ltd) supplemented with 10% fetal bovine serum and 2% PSV (10 mg/ml penicillin, 50 mg/ml streptomycin sulfate, 20 μg/ml folate, and 60 μg/ml vitamin B12). Cells were maintained as axenic (bacteria free) adherent cultures for approximately one week prior to the start of the experiment.
Bacterial Strains and Culture Conditions
Eight bacterial species were streaked onto lysogeny broth (LB) agar plates from frozen cultures: Gram-negative: K. pneumonaie (Kp), E. coli (Ec), P. fluorescens (Pf), Burkholderia spp. (Bu); Gram-positive: E. facaelis (Ef), B. subtillus (Bs), S. aureus (Sa), or M.(Ml). Strain information is provided in Table S1; most were standard lab strains, except for Burkholderia spp., which we isolated from the sorus of a D. discoideum fruiting body. Following inoculation from frozen stocks, colonies were inoculated into 1 ml of LB medium and grown overnight at their respective optimal temperatures: 37C (Ec, Bs, Kp, Ml, and Sa), 31C (Bu, Ef) or room temperature (Pf).
Chemotaxis Assay
Following overnight growth, bacterial cultures were concentrated to an OD600 of 5 by centrifugation. Cells of D. discoideum were harvested from the petri dishes by pipetting, washed with KK2 buffer (14.0 mM K2HPO4 and 3.4 mM K2HPO4, pH = 6.4), and re-suspended at a density of 1 × 107 cells/ml. Two microliters of each D. discoideum suspension (corresponding to 2 × 104 cells) were spotted onto a Nunc Omni tray containing 10 ml of KK2 with 2% agar noble. A grid was placed beneath the plate to ensure equal spacing of the amoeba spots, which were placed at distances of approximately 1.3 cm. The D. discoideum spots were first allowed to dry (~15 min), and then 2 μl spots of bacteria were placed in between each D. discoideum spot, again using a grid beneath the plate for placement, such that the distance between the center of each D. discoideum spot and those of its bacterial choices was approximately 0.65 cm (see Figure 1). Each D. discoideum genotype was placed in a different row of the plate, and each bacterial species in a different column. Across replicate experiments, we changed the order of column and rows so that the bacteria and D. discoideum cells were in different locations. Experiments were carried out in three temporally independent blocks, each commenced by re-inoculating from frozen stocks.
Time Lapse Videos of Chemotaxis
We recorded the movement of the D. discoideum cells toward the bacteria by time-lapse microscopy, which commenced two hours after the spots were deposited. Images were taken on a Leica DMI6000B microscope with an automated stage at 50X magnification using the LAS software suite. The left and right edges of each amoeba spot (representing the area the amoebae would enter once they commenced chemotaxis) were imaged every five minutes. From these image stacks, we extracted the image corresponding to four hours from the start of the experiment, which was two hours after the start of the time lapse. We chose this time-point based on preliminary experiments: by this time, the amoebae had usually exited the boundaries of the spots, but not yet filled the imaged area completely.
Image Analysis
The number of cells that moved towards each bacterial species was counted in ImageJ. Images were first cropped (at a tangent to the spot) so that only cells that had migrated outside the boundaries of the spot would be counted. The cropped image, consisting of a rectangular selection of dimensions 10 × 12.5, was converted to an 8-bit image, thresholded to achieve a binary image, and the cells were counted using the module “Analyze Particles”. Some manual adjustment was necessary to ensure clumps were split into single cells and any holes in the cells were filled. See Fig. S1 for an example of a raw image and the same image following thresholding and counting.
Adenlyate Cyclase E. coli Mutants
To assess the impact of cAMP on D. discoideum chemotaxis, we obtained a cAMP-deficient E. coli strain, where the only adenylate cyclase gene (cyaA), which is necessary for cAMP production, had been deleted (Keio:JW3778; ∆cyaA::kanR allele) and its wild-type progenitor (BW25113) from GE Dharmacon [46]. We also obtained an E. coli strain that carries an additional copy of the cyaA gene on an inducible plasmid (cyaAOE; ASKA E. coli K12 W3110; gift of Tim Cooper, University of Houston) [47]. The presence and absence of the cyaA gene in the ∆cyaA and wild-type E. coli, respectively, was confirmed by PCR using cyaA specific primers, whereas the ASKA strain was confirmed by sequencing the plasmid insert.
Prior to the assays, the ∆cyaA, wild-type, and cyaOE strains were streaked onto LB plates, grown overnight at 37 C, and single colonies were inoculated into test tubes containing 1 ml LB. The ∆cyaA culture was supplemented with 30 μg/ml kanamycin to ensure retention of the deletion allele, and the cyaAOE culture was supplemented with 10 μg/ml chloramphenicol and 1 mM IPTG to promote plasmid maintenance and induce expression of the cyaA gene, respectively. All cultures were incubated overnight at 37 C and concentrated to an OD600 of 5 prior to the start of the experiment. Chemotaxis assays were carried out as described above. We used all possible pairwise combinations of the three bacterial genotypes, with replicate spots on the plate for each D. discoideum - E. coli genotype pair. The entire assay was then repeated in a temporally independent replicate.
Statistical analysis
The experiment results were analyzed as a mixed-model ANOVA using PROC MIXED in SAS (v. 9). The number of cells migrating after 4 hours was modeled as a function of the D. discoideum (Dd) genotype (QS32, QS39, or QS40), the focal bacterial species, the alternative bacterial choice, and the block (i.e., experimental replicate). Focal bacteria and alternative bacteria were nested within Gram-status (positive or negative). We included interaction terms in the model, Dd genotype × bacteria, Dd genotype × alternative bacteria, and Dd genotype × block. The significance of random effects factors was assessed using a likelihood ratio test that compared the difference in the -2 residual likelihoods of nested models that did or did not include the factor of interest. This difference was compared to the Chi-square distribution, with the degrees of freedom equal to the difference in the number of parameters between the full and reduced models. Gram status was modeled as a fixed effects factor, and the significance of this term was determined by a standard F-test.
Results
In 21 of 24 pairwise choices between a Gram-positive and a Gram-negative bacterium, D. discoideum preferentially migrated towards the Gram-negative bacterial species (Fig. 2; sign-test: two-tailed P<0.0003). There were three exceptions to this pattern: QS39 showed a weak preference for Gram-positive B. subtilis over Gram-negatives K. pneumonaie and E. coli, and QS40 weakly favored B. subtilis in a pairwise contest against K. pneumonaie (see Fig. 2).
We modeled the chemotactic response, quantified as the number of cells migrating towards a bacterial spot after 4 hours, as a function of the focal bacteria, the alternative choice of bacteria (see Fig. 1), the Gram status, and the D. discoideum genotype. All three interaction terms (Dd genotype × focal bacteria, Dd genotype × alternative bacteria, and alternative bacteria × focal bacteria) were highly non-significant and dropped from the model (Table 1). The identity of the alternative bacteria in each pairwise choice was also highly non-significant and likewise dropped from the model. The lack of significance for this term indicates that the chemotaxis of the amoebae to any given bacteria was similar regardless of its alternative choice bacteria.
Overall, we observed significant variation in the response to different bacteria (focal bacteria; χ2=29, df=1, P<0.0001). The strongest response was to P. fluorescens, a bacterial species that has been found in the fruiting bodies of D. discoideum [40,42]. The weakest chemotactic responses were to the Gram-positive bacteria M. luteus, S. aureus, and E. facaelis (Fig. 3).
Consistent with our conclusions based on the pairwise analysis, our full statistical model showed a significant effect of Gram-status on chemotaxis (Fig. 4; F1,5.87=9.23, P=0.02), which reflected an overall preference for Gram-negative bacteria. Despite some differences among D. discoideum genotypes in the level of response to B. subtilis, the overall response profile to different bacteria was strikingly similar across the geographically widespread D. discoideum strains (Table 1, non-significant effect of Dd genotype of chemotaxis: χ2 =0, df=1, P=1).
Why does D. discoideum exhibit a chemotactic preference for Gram-negative bacteria? We hypothesized that this effect might be mediated by cAMP, a known chemoattractant for D. discoideum that is also differentially produced by Gram-negative and Gram-positive bacteria. In Gram-negative bacteria, cAMP functions in catabolite repression and is produced in large quantities intracellularly and released into the medium when glucose is absent. In contrast to the central role of cAMP in Gram-negative bacteria, Gram-positive bacteria do not use cAMP to mediate catabolite repression, there are few reports of its detection in Gram-positive bacteria except under limited conditions (e.g., O2 limitation [48]), and there are no established signaling roles for cyclic mononucleotides in Gram-positive bacteria [49].
To test the hypothesis that amoebae are specifically sensing and responding to cAMP, we compared the amoeba response to an E. coli mutant with a gene deletion of cyaA, which encodes adenylate cyclase, the enzyme that catalyzes the production of cAMP from adenosine triphosphate (ATP). We compared the amoebae responsiveness to the deletion mutant to that of its wild-type progenitor and to an E. coli strain that overexpresses cyaA—and thus these strains should vary in their production of cAMP. Consistent with our hypothesis, chemotaxis was strongest to cyaAOE and weakest to ΔcyaA (see Fig. 5 for example images for strain QS39), and the pattern was consistent across all three D. discoideum genotypes (Fig. 6). Taken together, this indicates that the positive response to E. coli is mediated, at least in part, by sensing of bacterially produced cAMP.
Discussion
Professional phagocytes locate bacteria in part based on signals secreted by bacteria. The ability to find bacteria at a distance is likely crucial for amoebae that seek bacteria for food and phagocytic immune cells that must arrive at a site of infection and mount an immune response or else allow a beneficial symbiont to remain and colonize. In the social amoeba D. discoideum, environmental bacteria represent not only potential prey, but some bacteria can establish intracellular infections that range from pathogenic to mutualistic [11,40–42,50,51]. Discrimination may be critical to explain why some bacteria are readily localized and removed by phagocytes whereas others evade detection.
In paired choice assays, we observed variation in the responsiveness of amoebae to different bacteria, and the overall pattern was stronger in favor of Gram-negative compared to Gram-positive bacteria. While it is known that D. discoideum can respond to folate derivatives released by bacteria [52,53] and it has been suggested previously that cAMP might mediate bacterial sensing [54,55], no prior work has quantified or demonstrated differential chemotaxis that effectively discriminates among different types of bacteria.
We initially hypothesized that, because chemotaxis towards bacteria is likely an adaptive response by amoebae to locate potential food sources, we might observe different chemotactic profiles of D. discoideum isolates from different geographic locations where the local bacterial community may vary. While we did observe some differences in response to the Gram-positive B. subtilis, overall we found no significant difference among the three geographically divergent D. discoideum genotypes in their bacterial chemotaxis profiles. This result suggests that discrimination ability may relate to fundamental characteristics of the bacteria, in that it is not highly malleable to fine-scale differences in the composition of the bacterial community across sites.
In light of cAMP as a strong chemoattractant for D. discoideum during its developmental transition to multicellularity and also its differential use by Gram-negative and Gram-positive bacteria, we examined the potential role of this molecule in mediating the chemotaxis behavior of D. discoideum. Consistent with earlier work, all three amoeba genotypes responded as expected under our hypothesis: their chemotaxis response was strongest to the cyaAOE (which should produce the highest levels of cAMP), weaker towards the wild-type, and the lowest to cyaA null mutants that lack cAMP. Taken together, this result suggests that D. discoideum’s preference for Gram-negative bacteria is mediated in part by differences in production of cAMP among bacteria. We note, however, that we observed stronger migration towards the cyaA- strain than a control that lacked any bacteria, indicating that cAMP is not the sole molecule that induces chemotaxis towards E. coli. The likely explanation for the migration towards the cyaA- strain is that the cells are responding to folate, which has been previously associated with localization of bacteria by Dictyostelium [52,56], or some other unidentified chemoattractant.
The specific bacteria that induced the strongest amoeba response are also interesting to consider individually. P. fluorescens, which elicited the strongest overall response, has been repeatedly isolated from the fruiting bodies of D. discoideum and shown to provide some benefit to the amoeba, as either food or through the production of antifungal compounds [42]. Two other bacteria that also elicited strong responses were E. coli and K. pneumoniae – these are both species that are routinely used to cultivate D. discoideum in the laboratory and constitute the best known prey bacteria for this species.
In prior work, Burkholderia, Enterobacter sakazakii, and P. fluorescens were commonly isolated from D. discoideum sori [40], and our characterization of the bacterial soil community at sites in Virginia where D. discoideum is unusually prevalent also indicates that these species are common in the soil (unpublished results). Taken together, these findings could suggest that D. discoideum may be most capable of detecting the bacteria with which it associates or encounters frequently in nature. However, the directionality of this finding is not obvious. In other words, does D. discoideum become preferentially associated with these bacteria because it readily detects and finds them in nature? Or has its detection method been fine-tuned to preferentially sense and find the bacteria with which it best associates? While we cannot decisively answer this question, the lack of any geographic pattern in the response of D. discoideum genotypes to different bacterial species, as well as the potential role of cAMP as the molecule that D. discoideum senses, lends greater credence to the first hypothesis – that D. discoideum is biased towards finding and associating with certain bacteria because of the conserved mechanism by which it senses them. More generally, our results suggest that long-term colonization, particularly of professional phagocytes, might be mediated in part by the mechanisms by which they detect and respond to bacteria and suggests the possibility that hosts might actively pursue and determine colonization outcomes.
Author Contributions
GR carried out experiments and analyzed data. EAO carried out experiments, performed statistical analyses, and wrote the paper. All authors gave final approval for publication.
Funding Sources
GR received a research scholarship for this work from the University of Houston Stutzenberg Undergraduate Research Endowment. EAO was funded by the National Science Foundation DEB-1557023. The authors thank Adam Kuspa, in whose lab we conducted preliminary experiments for this manuscript.