Research ArticleMICROBIOLOGY

Social amoebae establish a protective interface with their bacterial associates by lectin agglutination

See allHide authors and affiliations

Science Advances  24 Jul 2019:
Vol. 5, no. 7, eaav4367
DOI: 10.1126/sciadv.aav4367


Both animals and amoebae use phagocytosis and DNA-based extracellular traps as anti-bacterial defense mechanisms. Whether, like animals, amoebae also use tissue-level barriers to reduce direct contact with bacteria has remained unclear. We have explored this question in the social amoeba Dictyostelium discoideum, which forms plaques on lawns of bacteria that expand as amoebae divide and bacteria are consumed. We show that CadA, a cell adhesion protein that functions in D. discoideum development, is also a bacterial agglutinin that forms a protective interface at the plaque edge that limits exposure of vegetative amoebae to bacteria. This interface is important for amoebal survival when bacteria-to-amoebae ratios are high, optimizing amoebal feeding behavior, and protecting amoebae from oxidative stress. Lectins also control bacterial access to the gut epithelium of mammals to limit inflammatory processes; thus, this strategy of antibacterial defense is shared across a broad spectrum of eukaryotic taxa.


Characterizing the effector proteins that host organisms use to manage their microbiota is important for understanding how animals and their resident microbiota coevolve as a “superorganism” or holobiont (13). Host animals manipulate bacterial populations in their gut through the retention and clearance of specific species, while the bacteria produce metabolites that affect host processes from neurological function to immune signaling (4, 5). Over evolutionary time, these forces would be expected to shape host behaviors and microbiota composition to increase host fitness. A number of recent studies support the idea that the immune system and stable gut microbes work in concert to maintain intestinal homeostasis by limiting colonization by pathogenic bacteria. Homeostasis is achieved through the creation of the gut barrier by Paneth, goblet, and innate immune cells using mucus and other secreted proteins (6, 7). Host proteins necessary for establishing this barrier provide direct antimicrobial activity, a physical barrier, or agglutination, while symbiotic bacteria have evolved mechanisms to survive under the harsh conditions that these functions impose by developing resistance or inducing immune tolerance (8, 9). Mucins, a class of highly cross-linked proteins, create a physical mucus barrier that limits the exposure of epithelial cells to microbes. Mucus provides the scaffolding necessary for bacterial attachment and biofilm formation that is sometimes pathogenic (9, 10). Proteins and peptides secreted into the mucin layer, such as secretory immunoglobulin A (sIgA) and α-defensin 5, allow for biofilm formation and colonization by agglutination of symbionts, while lectins prevent bacteria from gaining direct access to the epithelial cells by holding them within the mucin layer (7, 11).

Microbes have evolved for billions of years within the context of predator-prey relationships, as exemplified by the relationship between amoebae and bacteria. Amoebae are effective predators that track their bacterial prey by chemotaxis toward the chemicals that bacteria release and efficiently phagocytizing them, while their bacterial prey evolved defense mechanisms that prevent amoebae from feeding on them (12, 13). Dictyostelium discoideum social amoebae feed on soil bacteria until the food supply is exhausted they then aggregate and develop through distinct stages, culminating in a fruiting body of dead stalk cells that hold up a ball of spores (14). The soil presents amoebae with a heterogeneous group of bacterial species, and D. discoideum amoebae recognize specific species and alter their metabolism to optimize their survival (15, 16). Developing D. discoideum amoebae carry a microbiome, commensal bacteria that can provide a future food supply, toxin resistance, or defense from nonkin amoebal strains (1720). We recently showed that the lectins discoidin IA and IC establish the D. discoideum microbiome (21). Wild microbiome-carrying strains secrete discoidin IA and IC that bind to the bacteria, protecting them from killing by D. discoideum antimicrobials and inducing the microbiome. This work showed that the selective retention of bacteria by amoebae during development likely begins at the plaque edge. Here, we describe a new role for CadA (also known as DdCAD-1 and GP24) at the plaque edge as a bacterial agglutinating protein important for amoebal survival. Our results suggest that CadA imposes structure on the bacterial population near growing amoebae, forming a beneficial interface that dampens bacterial pathogenicity and promotes amoebal feeding.


CadA is important for viability when amoeba grow on high-density bacteria

When individual D. discoideum amoebae grow on lawns of food bacteria, they form circular plaques as the amoebal colony expands through rounds of cell division (Fig. 1A). When working with the cadA mutant amoebae described previously, we noticed that their plaques were variable in size, suggesting that some plaques had a growth defect (22). To explore this possibility, we deleted the coding sequence of cadA in our laboratory strain AX4 to test the effect of this deletion on growth on bacteria. Unexpectedly, we found that cadA mutant amoebae had poor viability when plated on Klebsiella pneumoniae bacteria; <20% formed plaques under standard laboratory growth conditions (Fig. 1, A and B). However, the viability of the cadA mutant as judged by this assay was indistinguishable from the parental strain when cocultured with Micrococcus luteus, Bacillus subtilis, or Staphylococcus aureus (fig. S1A). The low viability of cadA amoebae when cultured with K. pneumoniae and their normal viability when cultured with Gram-positive bacteria suggested that secreted CadA is critical for plaque formation on at least some species of Gram-negative bacteria. A role for CadA during D. discoideum’s growth stage was unexpected since it has a well-characterized role as a cell adhesion protein during multicellular development (2325).

Fig. 1 CadA promotes amoebal plaque formation on dense bacterial lawns.

The viability of D. discoideum amoebae was assessed by plating them clonally on growing lawns of K. pneumoniae (K. p.). (A) Plaques resulting from equal numbers of AX4 (top) and cadA mutant (bottom) amoebae plated on K. pneumoniae on SM agar plates (left) or SM agar plates with the nutrient components diluted fivefold (SM/5) (right). (B) Amoebal viability as assessed by plating efficiency of AX4 and cadA (y axis) (left) on K. pneumoniae lawns growing on SM agar plates in which the nutrient components were diluted 1.5- to 5-fold, as indicted. Number of K. pneumoniae cells in representative samples of these platings when grown without amoeba (red line; y axis) (right). Lower and upper bounds of the box plots represent the first and third quartile, respectively, while the whiskers represent maximum and minimum values; the centerline is the median, and the “x” represents mean of triplicate samples for three biological replicates [Kruskal-Wallis one-way analysis of variance (ANOVA) and ad hoc pairwise Wilcoxon rank sum test]. ns, not significant; CFU, colony-forming units. (C) Plating efficiency of AX4 and cadA amoebae on SM agar in the presence of exogenous CadA protein. Statistical significance was observed between cadA and cadA amoebae plated with exogenous CadA protein produced in E. coli [recombinant CadA protein (rCadA)], where ***P < 0.001. Box plots represent data from triplicate samples of three biological replicates where statistical significance was estimated by the Wilcoxon rank sum test.

Mutants of D. discoideum that are defective in their responses to Gram-negative bacteria display reduced viability on dense lawns of K. pneumoniae bacteria but grow normally on dead K. pneumoniae (16, 26). This indicates that growth on “preferred” food bacteria such as K. pneumoniae can be toxic for amoebae. The low viability of cadA mutants might also be due to K. pneumoniae toxicity, so we compared their viability during coculture with K. pneumoniae on SM agar plates containing successively lower nutrient levels, which produced bacterial lawns of successively lower density (Fig. 1B). The cadA amoebae were about twice as efficient in forming plaques on SM/1.8 agar (1.8-fold diluted nutrients; Fig. 1B) compared to SM. On SM/2.5 and SM/5, the plating efficiency of the cadA amoebae was indistinguishable from the parental AX4 (Fig. 1, A and B). Decreasing the initial seeding density of K. pneumoniae eightfold before coculture with cadA amoebae on undiluted SM agar also yielded plating efficiencies comparable to AX4 (fig. S1B). These data suggest that cadA is necessary for the survival of amoebae when cocultured with dense bacteria, but it is also possible that K. pneumoniae bacteria are less toxic to amoebae when they are grown under suboptimal nutrient conditions (e.g., SM/2.5 agar) or when bacterial lawn formation requires that the seeded bacteria undergo additional doublings on full-strength SM agar. To address this, we tested whether exogenous CadA protein could rescue the viability of the cadA amoebae with bacteria growing under standard nutrient conditions. We applied recombinant CadA, purified from CadA-expressing Escherichia coli to individual cadA mutant amoebae, and plated the mixture on SM agar plates seeded with K. pneumoniae. cadA mutant amoebae produced greater than threefold more plaques when exogenous CadA was present, up to ~70% of the viability of the AX4 parent (Fig. 1C). Together, these results suggest that extracellular CadA is important for the survival of D. discoideum amoebae during the initial stages of plaque formation.

What about the cadA mutant plaques that do form on K. pneumoniae lawns? We measured their rate of expansion from the time of their delayed appearance on the K. pneumoniae bacterial lawn and found that the rate was indistinguishable, on average, from the wild-type strain AX4 (fig. S1C). This suggests that, once a cadA mutant plaque achieves a particular size or maturity, CadA is no longer essential for viability on K. pneumoniae lawns.

The compromised viability of cadA mutants suggests a role for the CadA protein in protecting amoebae from bacterial toxicity. To test this, we examined the production of reactive oxygen species (ROS) in amoebae as one indication of toxic stress. We sampled amoebae feeding on bacteria at the plaque edge and measured ROS using a ROS-activated dye that becomes fluorescent orange when exposed to ROS. By this measure, the cadA mutant amoebae had much higher levels of ROS compared to wild-type cells (fig. S2A). After culturing them with K. pneumoniae for 24 hours on non-nutrient agar, roughly 15-fold more cadA mutant amoebae displayed high levels of cytoplasmic ROS compared to wild-type cells, but comparable numbers of cadA mutant and wild-type amoebae displayed lower levels of ROS when exposed to B. subtilis (fig. S2B). These observations suggest that CadA suppresses the toxicity of K. pneumoniae to D. discoideum amoebae that are actively feeding at the plaque edge.

CadA prevents mixing of bacteria and amoebae across plaque borders

CadA is secreted unconventionally through the contractile vacuole during development (27, 28), and we recently found that it is secreted during the vegetative growth of some D. discoideum isolates collected from the wild (21). In addition, the CadA gene is expressed during growth on E. coli and K. pneumoniae but not during growth on S. aureus or B. subtilis, suggesting that CadA is produced during growth on specific species of bacteria (16). If CadA functions to protect amoebae from bacterial toxicity, then it should be localized where the amoebae meet bacteria at the plaque edge. Attempting to demonstrate this, we used a Western protein blotting procedure on the agar, upon which wild-type and cadA mutant plaques had been growing on K. pneumoniae lawns and probed the blots with monoclonal antibodies directed against CadA protein. We observed CadA protein signal corresponding to the places on the agar, where the edges of wild-type plaques were growing and no signal from where cadA mutant plaques were growing (Fig. 2A). Since all of the amoebae and bacteria were removed from the agar before blotting, these results suggest that CadA is secreted at the growing edge of plaques. To examine CadA secretion directly, we harvested amoebae growing on bacteria, disaggregated them into single-cell suspensions in buffer, and separated amoebae, bacteria, and the buffer supernatant by differential centrifugation (see Materials and Methods). CadA protein was detected in all three fractions, but cell-associated actin was only found in the cell pellet fractions (Fig. 2B), suggesting that CadA is produced and secreted by vegetative amoebae growing on K. pneumoniae bacteria. We carried out similar experiments on amoebae growing, axenically, in liquid media and found that CadA was not produced at all under these conditions (Fig. 2B).

Fig. 2 CadA prevents mixing of bacteria and amoebae across plaque borders.

(A) Growing edge localization of CadA protein. AX4 and cadA plaques growing on K. pneumoniae on SM/5 agar were extracted and electroblotted onto nitrocellulose membranes. CadA protein was visualized using anti-CadA monoclonal antibody MLJ11. Scale bars, 1 cm. The inset shows the result of 2 μg of purified CadA spotted on top of a cadA plaque before blotting. (B) AX4 and cadA amoebae were plated with K. pneumoniae onto SM agar plates and incubated overnight. Plates were harvested and amoebal (AP) and bacterial (BP) pellets were separated from the supernatant (S) by successive centrifugation and probed for the presence of CadA by Western blot, with pure CadA as a control. AX4 cells were also recovered from growth media (HL5) after overnight incubation and processed such as the plated cells into amoebal pellet and supernatant fractions. Anti-actin monoclonal antibodies (mAb) were used to detect an amoebae-associated protein in the same samples. (C) Representative images of AX4 (top) and cadA (bottom) plaques grown on K. pneumoniae on SM/2.5 agar. AX4 plaques form roughly circular plaques with defined edges (top left), while cadA plaques appear erose with satellite plaques (bottom left). No satellite colonies are observed when cadA amoebae are grown with exogenously added recombinant CadA protein. Representative images of H2B-mCherry–expressing AX4 amoebae, showing a defined plaque edge with few cells outside of the plaque, and H2B-mCherry–expressing cadA amoebae migrating away from the plaque edge into the bacterial lawn. The table shows the numbers of cells observed outside of plaques. (D) Representative images of the edge of AX4 (top) and cadA (bottom) plaques on K. pneumoniae on SM/2.5 agar, stained with LIVE/DEAD BacLight stain and visualized by differential interference contrast (DIC) or fluorescence microscopy. SYTO9 (green) labels live bacteria and propidium iodide (PI; red) labels dead bacteria. In the composite images of the plaque edges (far right), live bacteria are seen within the growing edge of the cadA plaque, while the AX4 has predominantly dead bacteria within the border of the plaque. Scale bars, 100 μm.

When cadA amoebae produced plaques on low-density K. pneumoniae lawns, they were irregular in shape with many small satellite plaques outside of the borders of the main plaques; they lacked the symmetry and distinct borders of wild-type plaques (Fig. 2C). Satellite plaques are occasionally formed by the wild type when motile amoebae migrate into the bacterial lawn beyond the edge of the main plaque. To visualize the dispersal of amoebae beyond the plaque border, we labeled their nuclei with a protein fusion of histone H2b and a red fluorescent protein, H2b-mCherry (see Materials and Methods). We examined plaques on lawns of K. pneumoniae by fluorescence microscopy and observed many cadA amoebae beyond the growing edge of every mutant plaque and very few AX4 amoebae outside of any wild-type plaque (Fig. 2C). We also found hundreds of cadA amoebae outside of mutant plaques on thick lawns of K. pneumoniae (SM/1.9 and higher; fig. S1D). Very few satellite plaques ever formed from these cells, so we infer that they have low viability. In addition, wild-type cells and cadA plaques appeared indistinguishable when grown on B. subtilis with very few amoebae found dispersed outside of the plaque borders (fig. S1E). To test whether this phenotype resulted from a lack of extracellular CadA, we mixed CadA with cadA amoebae and spotted the mixture on SM/2.5 agar plates seeded with K. pneumoniae. The cadA mutant plaques formed in the presence of CadA protein were symmetrical, with very few cells migrating beyond the plaque edge, and they resembled wild-type plaques (Fig. 2C).

Most bacteria found within D. discoideum plaques are dead, presumably killed by antibacterial proteins secreted by the amoebae (21). Since cadA amoebae do not remain within the borders of plaques, we examined whether live bacteria were restricted to the bacterial lawn surrounding cadA plaques. We visualized the bacteria at the growing edge of mutant and wild-type amoebal plaques growing on K. pneumoniae lawns on SM/2.5. While the growing edges within AX4 plaques contained mainly dead bacteria, the edges of cadA plaques contained patches of live bacteria (Fig. 2D). Note that D. discoideum amoebae appear free of any fluorescence, presumably due to robust transporters pumping out the dyes (29, 30). These results suggest that CadA is necessary for the proper organization of bacteria and amoebae at the growing edge of D. discoideum plaques, establishing a border between them.

CadA is a bacterial agglutinin that restricts expansion of bacterial colonies

We next tested whether CadA protects K. pneumoniae bacteria from D. discoideum antibacterial proteins, as we recently demonstrated for discoidin I (21). We isolated Dictyostelium antibacterial proteins (Dabs) that are naturally secreted by AX4 during feeding on K. pneumoniae bacteria, as described previously (21). We mixed CadA with K. pneumoniae bacteria suspended in growth media containing partially purified Dabs. Without added CadA, we observed complete killing of the bacteria, as expected, but at higher CadA concentrations, we observed robust bacterial growth, suggesting that CadA had protected the bacteria from killing (fig. S3, A and B). To ensure that CadA had not inactivated or sequestered the Dabs, we incubated them together before the assay, and after removing CadA, the Dabs retained their ability to kill bacteria, suggesting that CadA protection works through direct interaction with bacteria (fig. S3A).

We recently showed that the D. discoideum lectin discoidin I binds to K. pneumoniae and modulates the interactions of bacteria and amoebae (21). Since our results suggested that CadA also interacts directly with bacteria, we examined CadA’s ability to bind K. pneumoniae. Using recombinant CadA protein and a bacterial binding assay (fig. S3C), we observed saturable binding of CadA to K. pneumoniae and could estimate an apparent Ka of 1.0 × 107 M−1 and 7.5 × 105 binding sites per bacterium (Fig. 3A). To begin to characterize the CadA binding target on the bacteria, we attempted to inhibit CadA binding with various monosaccharides. We observed significantly reduced CadA binding to K. pneumoniae in the presence of 100 mM galactose and 300 mM glucosamine (Fig. 3B). This suggests that CadA is a lectin that binds to carbohydrates on the surface of K. pneumoniae, perhaps to their lipopolysaccharide, which is rich in galactose moieties (31).

Fig. 3 CadA binds and agglutinates bacteria.

(A) Increasing concentrations of recombinant CadA (1.25 to 74 μg) were incubated with 109 K. pneumoniae in 600 μl for 60 min. After bacteria were removed by centrifugation, unbound CadA was used to determine the amount of CadA bound to bacteria. Saturation binding of CadA suggests an association constant (Ka) of 1.0 × 107 M−1 and 7.5 × 105 binding sites per bacterium. Line shown is best fit for specific binding with Hill coefficient with the Bmax and Hill coefficient calculated to be 15.58 and 1.16, respectively. (B) K. pneumoniae (109) were incubated with recombinant CadA (8 μg) for 60 min, with increasing concentrations (10 to 300 mM) of d-glucose (white), d-glucosamine (black), and d-galactose (gray). Unbound CadA was quantified and used to calculate the amount of CadA bound to the bacteria. Data represent the means ± SD values of three technical replicates of three independent biological replicates, and the presence of glucosamine and galactose shows significantly reduced CadA binding relative to glucose (Wilcoxon rank sum test, ***P < 0.001). (C) K. pneumoniae (109) bacteria were incubated with buffer (top) or recombinant CadA protein (0.66 mg/ml) (bottom), with glucose (middle) or galactose (right) for 1 hour and visualized by fluorescence microscopy. Scale bars, 100 μm. GFP, green fluorescent protein.

By microscopic examination of the Dabs protection assay, we observed nonuniform bacterial growth at CadA concentrations that were protective to the bacteria. Using K. pneumoniae–expressing green fluorescent protein, we observed a few concentrated foci of surviving bacteria adhering to the surface of the plate (fig. S3A), suggesting that CadA may protect the bacteria by sequestering them within clumps, possibly reducing their exposure to the Dabs. We directly tested CadA for agglutinating activity by incubating K. pneumoniae suspensions with increasing concentrations of CadA protein and followed bacterial clump formation by light scattering and by direct microscopic examination. We observed bacterial clumps at CadA (0.16 mg/ml), and larger clumps were produced as we increased the concentration of CadA (Fig. 3C and fig. S3D). Biofilms provide important niches required for gut homeostasis; some Lactobacillus species require biofilm formation for colonization in the gut, and sIgA induces biofilm formation in some commensal bacteria, establishing a barrier to pathogenic bacteria (32, 33). Visualization of CadA agglutinated K. pneumoniae with Live/Dead staining shows similarities to biofilms with an outer layer of live bacteria around a core of dead bacteria (fig. S3E) (34, 35).

Structural studies of CadA revealed bound calcium ions in the N-terminal domain that are necessary for amoebal cell-cell adhesion (24). To determine whether calcium is necessary for bacterial agglutination, we incubated CadA with the calcium chelator EGTA for 1 hour and then removed the EGTA. The agglutinating activity of EGTA-treated CadA was greatly diminished compared to untreated protein but could be restored by adding calcium chloride to the assay, indicating that calcium is necessary for agglutination (fig. S3E). Bacterial agglutination by CadA (Fig. 3C) and its protection of bacteria from Dabs (fig. S3B) were completely abolished by 300 mM galactose, suggesting that the lectin function of CadA is necessary for bacterial agglutination and for the protection from antibacterial proteins that we observed.

The agglutinating activity of CadA provides a potential mechanism for its role in amoebal survival on dense bacterial lawns. One possibility is that CadA forms a lattice out of the proximal bacteria at the plaque edge, forming a barrier that modulates the exposure of amoebae to bacteria. To explore this idea, we attempted to produce a bacterial barrier with pure CadA protein without any amoebae present. We deposited recombinant CadA protein on the surface of nutrient agar plates and spotted bacterial suspensions adjacent to the protein deposits. As the K. pneumoniae, or E. coli B/r, bacteria grew and expanded beyond the initial deposition of cells, and they were unable to occupy areas containing CadA protein, but they did expand into deposits of mock-purified protein or bovine serum albumin (BSA; Fig. 4A). Bacterial spots of the other species that we tested expanded symmetrically into the areas of CadA and BSA protein deposition (Fig. 4A). CadA was able to agglutinate E. coli B/r, as described for K. pneumoniae above, but CadA did not agglutinate S. aureus, B. subtilis, Pseudomonas aeruginosa, or M. luteus (Fig. 4B). Thus, CadA’s ability to restrict colony expansion of a given bacterial species correlated with its ability to agglutinate those bacteria (compare Fig. 4, A and B). These results, together with the plaque phenotype of cadA mutants, suggest that Dictyostelium amoebae secrete CadA to form a dynamic interface at the plaque edge to control their interaction with K. pneumoniae bacteria.

Fig. 4 Species-specific restriction of bacterial colony expansion by CadA correlates with CadA bacterial agglutination.

(A) A fraction from a mock protein purification from E. coli, recombinant CadA (10 μg), or BSA (10 μg) was spotted onto SM agar and allowed to dry. Overnight cultures of K. pneumoniae (top left) and other species of bacteria, as indicated, were spotted adjacent but not touching the CadA spot. Dashed circles represent the original deposition of the protein solutions (or mock fraction), and the bars indicate the diameter of the original bacterial spot. (B) The bacterial species in (A) were incubated with recombinant CadA protein (0.66 mg/ml) for 1 hour and visualized by fluorescence microscopy after labeling the bacteria with the cell-permeable dye SYTO9. Scale bars, 100 μm.

Altered amoebal predation of CadA-agglutinated bacteria

The agglutination of bacteria at the edge of Dictyostelium plaques would be expected to affect amoebal predation and feeding. We examined the feeding behavior of amoebae with a model of the plaque edge that has the inverse geometry of a plaque, amoebae surrounding CadA-agglutinated clumps of bacteria. We observed amoebae presented with a mixture of CadA-agglutinated K. pneumoniae bacteria and planktonic bacteria using an under-agar assay for optimal visualization. As a control, we presented amoebae with clumps of bacteria that were generated mechanically from packed pellets formed by centrifugation of bacterial suspensions. Amoebae actively moved toward and surrounded the CadA-induced bacterial clumps, and as they divided, the daughter cells remained feeding at the edge of the same clump (Fig. 5, A and B, and movie S1). Amoebae approached mechanically generated clumps chaotically, some penetrated to the interior of clumps, and some left the clumps altogether (Fig. 5, A and B, and movie S2). Both wild-type and cadA mutant amoebae moved faster and more persistently toward the CadA-induced clumps as compared to the mechanically generated clumps (fig. S4, A and B), and they chemotaxed more efficiently to CadA-induced clumps (Fig. 5C and fig. S4C). Directed movement of amoebae toward CadA-induced clumps appears to represent folate chemotaxis since the addition of 100 μM folate (Fig. 5C) or the use of dead bacteria (fig. S4D) abolishes the effect. The addition of 100 μM folate only partially reduced the speed of amoebae approaching CadA-induced clumps, suggesting that other factors emanate from the bacterial clumps that stimulate cell movement (fig. S4A). These results suggest that CadA-induced clumping provides agency to social predation of bacteria. These data suggest that CadA agglutinates bacteria in a way that signals amoebae to prey on bacteria cooperatively and in a way that appears to limit their exposure to the toxicity of live bacteria.

Fig. 5 Altered amoebal predation on CadA-agglutinated bacteria.

H2B-mCherry–expressing AX4 amoebae (A) or cadA mutant amoebae (B) were mixed with K. pneumoniae clumped by CadA (A and B, top) or by mechanically by centrifugation (A and B, bottom) under agar on glass bottom plates. White arrows highlight amoebae traveling through a clump, while orange arrows highlight amoebae traveling away from a clump. Individual cells were imaged and tracked for 3 hours using Nikon Eclipse Ti. t, minutes. Scale bars, 100 μm. (C) Violin plots of the chemotactic index of individual AX4 or cadA amoebae for clumps of K. pneumoniae that were mechanically produced, CadA-induced, or CadA-induced with 100 μM folic acid added to the buffer before plating.


Our results indicate that the well-characterized cell adhesion protein CadA is also a lectin and bacterial agglutinin. We hypothesize that Dictyostelium amoebae respond to the bacteria at the edge of a plaque much like they do to the surface of a clump of bacteria that have been agglutinated by CadA. The immunoblots of growing Dictyostelium plaques support this possibility since the highest concentration of CadA appears to be at the plaque edge. We also showed that Dictyostelium amoebae display positive chemotaxis to CadA-agglutinated bacteria and feed on those clumps at the edges, only moving to the center of the clumps by consuming the outer surface bacteria. Given the geometry of the natural plaque, CadA presumably would have the effect of attracting the amoebae to feed on the bacteria at the plaque edge without penetrating the plaque border to move into the bacterial lawn. This is what is observed in wild-type plaques and precisely what is defective in cadA mutant plaques. The fact that the plaque formation defect of cadA mutants can be rescued by adding exogenous CadA protein supports these ideas and suggests that the porous plaque borders of cadA mutant plaques are converted to the less porous borders seen in the wild type. The viability defect of cadA mutants suggests that the functional requirement for CadA is particularly stringent when the bacteria-to-amoebae ratio is high such as when a single cell is attempting to produce a plaque. We do not know when the amoebae die when plated on a lawn of bacteria in the viability assay, but we do see single cells that venture outside of mature cadA mutant plaques, and they rarely divide even once (unpublished observations). This suggests that the function of CadA is directed at the bacteria rather than toward, say, cell-cell adhesion between sister cells as a nascent colony expands. When cadA mutants do form plaques, they grow at wild-type rates, suggesting that additional mechanisms come into play to protect amoebae from K. pneumoniae bacteria as the plaques mature. All of these ideas suggest that CadA is part of the system of cooperative feeding where a minimum number of amoebae secrete CadA to “wall off” the bacteria, by agglutination, long enough to establish a plaque.

Innate immune effector proteins are necessary for barrier formation in the gut as perturbations in the microbiota result in diseases such as Crohn’s disease and ulcerative colitis (36, 37). We have shown that CadA functions as a lectin during the vegetative growth of D. discoideum and establishes a barrier between the amoebae and specific species of bacteria upon which they feed. CadA agglutinates the two species of Enterobacteriaceae, but not the Gram-positive bacteria that we tested, and forms a barrier to Enterobacteriaceae colony expansion on agar surfaces and clumps of bacteria in suspension. The absence of the CadA-bacteria barrier leads to bacterial penetration of the growing plaque and amoebal penetration into the surrounding bacterial lawn, and amoebae display increased production of ROS and have reduced viability. Perturbation of barrier forming genes in animals yields similar results. Decreased expression of the agglutinating peptide α-defensin, or homozygous null mutations in the lectin ZG16, causes a breakdown in the gut epithelium barrier, leading to increased bacterial invasion (7, 11, 36, 37). In addition, sIgA agglutinates bacteria in a process known as immune exclusion, preventing the attachment to the gut epithelium and allowing for clearance (9, 33, 38). Additional D. discoideum lectins that we have identified by mass spectrometry of secreted proteins (Cad2 and Cad3) (21), or genetically (iliE) (16), may modulate amoebal interactions with different bacteria. The iliE gene is predicted to encode a concanavalin A lectin homolog and mutations in the gene render amoebae unable to form plaques on Gram-positive bacteria such as B. subtilis (16). The phosphorylation status of CadA may also play a role in altering immune function of the protein (39). Thus, the growing edge D. discoideum plaques may provide a model for barrier immunity where the separation between “host” amoebae and their bacterial associates is mediated by a number of secreted proteins. Our results suggest that common selective pressures have shaped bacteria-eukaryote interactions from amoebae to animals, since both agglutinate bacteria with lectins to protect against potential toxic effects of their bacterial associates.


Strains, growth, and plasmids

K. pneumoniae was grown in SM media or on SM agar (21). D. discoideum strains were derived from the axenic laboratory strain AX4 and maintained in HL5 media with penicillin (50 U/ml) and streptomycin (50 μg/ml) or in coculture with K. pneumoniae on SM agar plates [2% Bacto agar (BD Difco) with SM media] in standard 10-cm plastic petri dishes as described previously (21). Blasticidin-resistant D. discoideum strains were grown in HL5 media supplemented with blasticidin (4 μg/ml; Thermo Fisher Scientific). Neomycin-resistant strains were grown in HL5 media supplemented with geneticin (20 μg/ml; Thermo Fisher Scientific).

DNA constructs

pDM304 mCherry-H2b was transformed into AX4 and cadA mutant strains to fluorescently tag their nuclei (21). Construction of the cadA knockout vector pLPBLP_cadA was accomplished using the following primers and homologous recombination: 5′ homologous arm was amplified with aaaggtaccgaagacTTCCTGATGGTGATGATGGTTATGA (cadA_HA5_fwd KpnI/BbsI) and tttgtcgacATTCAAATGATTCACCAGTGCAGTT (cadA_HA5_rev SalI), and 3′ homologous arm was amplified with aaaggatccAGACATTCCCAAAGAATATGACTGT (cadA_HA3_fwd BamHI) and tttactagtgaagacAAAAAAAATTTCCCGCTTTGAAGGG (cadA_HA3_rev SpeI/BbsI).

Homologous arms were amplified from AX4 genomic DNA. 5′ homology arm and pLPBLP vector were separately digested with KpnI and SalI and ligated together. The resulting vector and the 3′ homology arm were digested separately with BamHI and SpeI and ligated together. pLPBLP_cadA was digested with BbsI, resulting in linear fragments containing only homologous DNA and the blasticidin resistance cassette. The linear DNA was electroporated into amoebae, and mutants were selected with blasticidin (10 μg/ml) (40).

Plating efficiency and plaque growth

Plating efficiency assay was performed by plating 100 amoebae with bacteria onto nutrient agar plates. After 4 days, the number of observed plaques was counted. Plating efficiency was calculated by dividing the number of observed plaques by the number of plated amoebae. To measure plaque growth, 50 amoebae were added to 400 μl of overnight K. pneumoniae culture and plated on SM agar. Plates were observed until “pin prick” plaques became visible. The diameter of each plaque was then tracked for 4 days, and the area was calculated [area = π*(d/2)2].

To test the effect of CadA protein on plating efficiency, K. pneumoniae was plated on SM in a OneWell plate (Greiner Bio-One ). CadA protein was then mixed with AX4 or cadA amoebae in buffer and spotted with a multichannel pipette in a 96-spot format. Each 2-μl spot contained 20 μg of CadA and an average of three amoebal cells. Spots were scored on the basis of plaque growth.

Measurement of ROS

We examined cellular stress through the visualization of ROS formation in amoebae upon exposure to bacteria in two ways: at the growing edge of a plaque and after acute exposure on washed agars plates for 24 hours. AX4 and cadA amoebae were collected at mid-log phase (2 × 106) in HL5, washed, and resuspended in KK2 buffer at 1 × 107 cells/ml. Amoebae were mixed with 400 μl of overnight K. pneumoniae [optical density at 600 nm (OD600) = 2] in SM and plated on to SM agar. After 24 hours, the plates were scraped and washed three times with KK2 buffer. CellROX Deep Red (Thermo Fisher Scientific) was added to a final concentration of 5 μM, incubated for 30 min, harvested, and washed once with KK2 buffer before fluorescence imaging. To quantitate and compare the relative proportion of CellROX-positive cells, a fluorescence signal threshold that produced 1% positive AX4 cells was applied to all of the images. This threshold was chosen because high ROS levels are not typically observed in wild-type amoebae.

Expression of His6-tagged CadA

His6-tagged CadA was expressed from the pETM-CadA vector from C.-H. Siu, as described previously (24). Overnight E. coli cultures were diluted 1:100 into LB with carbenicillin (50 ug/ml) at 37°C. Cultures were grown to an OD600 of 0.8. Protein production was induced with 0.4 mM isopropyl-β-d-thiogalactopyranoside for 3 hours. Cells were spun down at 14,000g for 20 min at 4°C. Pellets were resuspended in sonication buffer containing lysozyme (1 mg/ml). The bacteria were sonicated on ice to lyse. Samples were centrifuged at 20,000g for 20 min at 4°C. The supernatant was collected, and CadA was purified using HisPur Ni-NTA resin (Thermo Fisher Scientific) and the corresponding protocol. Samples were then concentrated and buffer-exchanged with Sorenson’s (Sor) buffer (2.0 g KH2PO4 and 0.29 g Na2HPO4/liter) using a 3000-kDa molecular weight cutoff spin column (Centricon Ultracel-3K). For mock-purified samples, E. coli containing empty pETM vector were put through the same procedure, and a volume equal to the CadA sample was used in the experiment in Fig. 4A.

Biochemical methods

SDS–polyacrylamide gel electrophoresis (SDS-PAGE) was performed on 12% gels with 30% acrylamide/bis-acrylamide (29:1) solution (Bio-Rad). Western blot transfer from SDS-PAGE gels to nitrocellulose membranes were performed with methanol/tris transfer buffer [25 mM tris (pH 7.6) and 20% (v/v) methanol] at 100 V for 1 hour. Membranes were blocked with tris-buffered saline with Tween 20 (TBST; pH 7.6, 0.5% Tween 20) and 5% nonfat milk for 1 hour at room temperature. Monoclonal anti-CadA antibodies (MLJ11, obtained from D. Fuller; 1:10,000 dilution) were incubated for 1 hour at room temperature. Filters were washed with TBST three times for 10 min each. Secondary horseradish peroxidase–conjugated goat anti-mouse antibodies (Thermo Fisher Scientific) were incubated for 1 hour at 1:10,000 before repeating the wash. Blots were then visualized with a chemiluminescent substrate kit (SuperSignal WestPico, Thermo Fisher Scientific).

To determine CadA localization during growth on bacterial lawns, we constructed layered 10-cm agar plates by laying the agar of one SM agar plate (10-ml volume) on top of a 25-ml SM agar plate. An overnight culture of K. pneumoniae was mixed with ~100 amoebae and plated onto the layered agar plates. After the amoebal plaques appeared, the top 10-ml agar layer was separated from the petri dish and gently washed to remove bacteria and amoebae. The 10-ml agar disk was then placed on a nitrocellulose membrane with the side previously containing bacteria and amoeba facing away from the membrane. The contents of the agar were then electroblotted onto the membrane with a standard Western blot apparatus and buffer system, as described above, and CadA protein localization was determined as described above. As a control, 2 μg of recombinant CadA protein was spotted on top of a cadA mutant plaque to demonstrate that CadA could migrate through the SM agar.

CadA secretion during vegetative growth was determined by plating AX4 or cadA amoebae with K. pneumoniae on SM agar and incubating them overnight at room temperature. Before any clearing of the bacteria by the amoebae had occurred, the bacteria and amoebae were collected and suspended in Sor buffer. Differential low-speed centrifugation was used to separate amoebae (400g for 5 min), bacteria (3000g for 10 min), and supernatant. The presence of CadA was probed by Western blot as described above, and anti-actin monoclonal antibodies were used to detect cell-associated actin (224-236-1, Developmental Studies Hybridoma Bank, University of Iowa).

We determined the apparent binding affinity of CadA for K. pneumoniae by measuring the amount of protein remaining unbound. K. pneumoniae was grown to an OD600 of 1.0 in SM. One milliliter of bacteria was spun down and washed with Sor buffer. The bacteria were then resuspended in 60 μl of Sor buffer with 0 to 70 μg of CadA and incubated for 1 hour. The bacteria were pelleted, 30 μl of supernatant was collected, and unbound protein was measured by Bio-Rad protein assay. By subtracting the unbound protein from the known input concentration, we calculated the amount of CadA bound to the bacteria. Estimated binding affinity of CadA for K. pneumoniae was calculated using the method of Steck and Wallach (41).

To determine the possible lectin-binding activity of CadA, we repeated the above experiment in the presence of 0 to 300 mM d-galactose, d-trehalose, or d-glucosamine (EMD Millipore). K. pneumoniae was grown as previously described. Next, bacteria, saccharide, and 10 μg of CadA were mixed in 60 μl of Sor buffer for 1 hour. Supernatants were collected, and unbound protein was measured by the Bio-Rad protein assay.

Purification of secreted antibacterials

Secreted antibacterials were purified as described previously (21). Briefly, 10 SM plates with AX4 growing on K. pneumoniae were scraped, and amoebae were separated by low-speed centrifugation. The amoebae were washed once with Sor buffer and brought up to 40 ml of Sor buffer. Overnight culture (10 ml) of K. pneumoniae grown in SM was pelleted at 4000 rpm and added to the amoebae. The amoebae and bacteria were gently mixed on an orbital shaker for 2 hours. The sample was then spun at 1500 rpm for 5 min in a Triton X-100 swinging bucket rotor (Sorvall Legend RT) to pellet the amoebae. Next, the supernatant was centrifuged at 4000 rpm for 5 min to pellet the bacteria. Last, the supernatant was centrifuged at 20,000 rpm (Sorvall 4B SS-34 rotor) for 10 min at 4°C. The supernatant was then filtered [0.22-μm-pore polyethersulfone filter (Corning)] and loaded onto a 10-ml DEAE Sepharose CL-6B (GE Healthcare) column. The column was washed with three column volumes of 10 mM tris-HCl (pH 8) buffer. Elution fractions (1 ml) were then collected from 100 mM steps of NaCl in 10 mM tris-HCl (pH 8).

Antibacterial activity assay

Overnight K. pneumoniae cultures were diluted 100,000-fold in SM media, and 50 μl of this dilution were put into wells of a 96-well plate. Samples (antibacterials, CadA, or antibacterials previously incubated with CadA) were then deposited in the first row of wells in a twofold dilution to bring the total well volume to 100 μl. Each sample was then diluted across the plate twofold. The plates were then placed at 37°C overnight. Growth was observed as cloudy wells and checked by dilution onto SM agar plates that were placed at 37°C overnight.

Spatial inhibition of bacterial growth

Bacteria were grown overnight at 37°C. Each bacterial suspension (2 μl) was spotted onto the respective bacterial nutrient agar and allowed to dry. Once dried, 10 μg of CadA, an equivalent volume of mock-purified CadA (from E. coli with and empty expression plasmid), or 10 μg of BSA was spotted adjacent to but not touching the bacterial colonies and allowed to dry. After drying completely, the plates were placed in a 37°C incubator and imaged every day for 4 days.


Bacteria were prepared as before in overnight cultures. One milliliter of bacteria (OD600 = 1.0) was centrifuged and washed with 200 μl of Sor buffer. The bacteria were then resuspended with CadA to 60 μl and incubated for 45 min to 2 hours. Bacteria were centrifuged at 7000 rpm for 4 min and resuspended in 60 μl of Sor buffer, and 5 μl of each bacteria were placed on slides and imaged. Removal of divalent cations was accomplished by incubating CadA with 10 mM EGTA in Sor buffer for 1 hour. EGTA was removed by buffer exchange in a 3000-kDa molecular weight cutoff spin column (Centricon Ultracel-3K) with Sor buffer. Mechanical agglutination was achieved by centrifugation of bacterial suspensions at 12,000 rpm for 5 min. Bacteria were then gently resuspended in 60 μl to maintain large aggregates.

Predation videos

Amoebae were grown to mid-log phase (1 × 106 to 2 × 106 cells/ml) in HL5 with penicillin (50 U/ml) and streptomycin (50 μg/ml). K. pneumoniae was grown and agglutinated as above. A total of 5 × 105 AX4 or cadA amoebae were mixed with 5 μl of aggregated bacteria in 300 μl and spread onto 60-mm 1.5% KK2 [2.2 g KH2PO4, 0.7 g K2HPO4/liter (pH 6.4)] buffered agar plates. The cells were allowed to dry before 50 mm by 50 mm squares were cut and placed upside down onto glass-bottomed 6- or 12-well dishes (MatTek 12-well glass bottom culture plate). To test the effect of folate on cell movement, folic acid was added to the suspension of amoeba and bacteria at 100 μM before deposition on the agar plate. Images were taken on Nikon Eclipse Ti using NIS-Elements imaging software version 4.51.00 (build 1143). NIS-Elements imaging software was used to track individual amoebae over the course of the experiment. Chemotactic index was calculated as the distance moved toward the center of a bacterial aggregate divided by the total distance (42). Cells beginning the experiment in contact with an aggregate were excluded. Speed (fig. S3A) and path length (fig. S3B) were calculated by NIS-Elements tracking.

Statistics and reproducibility

Non-normal distribution of plating efficiency data was determined by Shapiro-Wilk test. Statistical significance for plating efficiency was calculated by Kruskal-Wallis one-way analysis of variance (ANOVA), followed by an ad hoc pairwise Wilcoxon rank sum test (Fig. 1B) from three biological replicates of three technical replicates. Statistical significance of exogenous CadA rescue (Fig. 1C) was calculated by Wilcoxon rank sum test. Non-normal distribution of saccharide inhibition data was determined by Shapiro-Wilk test. Statistical significance of saccharide inhibition of CadA binding (Fig. 3B) was determined by Wilcoxon rank sum test. Chemotactic index data represents n > 40 cells in each condition and was non-normally distributed by Shapiro-Wilk test (Fig. 5C). Significance for chemotactic indices was determined by Wilcoxon rank sum test (Fig. 5C). Non-normal distribution of speed measurements was determined by visual inspection of quantile-quantile plots and histograms and statistical significance was determined by Wilcoxon rank sum test (fig. S4A). CellROX-positive statistics were calculated using a two-tailed Student t test (Fig. 2B). Path length was determined to be normally distributed by Shapiro-Wilk test and statistical significance was determined by Welch’s t test for unequal variance (fig. S4B).


Supplementary material for this article is available at

Fig. S1. Growth of cadA mutant amoebae on dense lawns of bacteria.

Fig. S2. Elevated ROS in cadA mutant amoebae feeding on bacteria.

Fig. S3. CadA agglutinates K. pneumoniae and protects them from killing.

Fig. S4. Differential amoebal movement toward CadA clumped bacteria.

Movie S1. Amoebal predation on CadA-agglutinated bacteria.

Movie S2. Amoebal predation on mechanically generated bacterial clumps.

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.


Acknowledgments: We thank O. Zhuchenko for preliminary observations of the cadA mutants and W. F. Loomis and G. Shaulsky for providing insights during the course of this work. Funding: This work was supported by the Dictyostelium Functional Genomics Program project grant from the NIH (PO1 HD39691). Author contributions: T.F. and A.K. conceived and designed experiments, T.F. directed and performed all of the experiments with assistance from C.D. T.F. and A.K. wrote the manuscript with scientific and editorial input from C.D. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors.

Stay Connected to Science Advances

Navigate This Article