Lectins modulate the microbiota of social amoebae

See allHide authors and affiliations

Science  27 Jul 2018:
Vol. 361, Issue 6400, pp. 402-406
DOI: 10.1126/science.aat2058

Sticky bacteria tolerated as future food

Dictyostelium discoideum amoebae consume bacteria until the supply is exhausted. Then the amoeba cells clump together into a “slug” and initiate a complex multicellular reproductive phase. Specialized cells within aggregates rid the slug of any extracellular bacteria. However, some strains of amoeba tolerate live, intracellular bacteria. Dinh et al. discovered that these carrier strains bear surface lectins that bind Klebsiella bacteria, promote cell entry, and prevent the bacteria from being immediately digested. These bacteria then provide a future food source. Moreover, the internalized bacteria transfer DNA into the amoeba nucleus, resulting in transient genetic transformation.

Science, this issue p. 402


The social amoeba Dictyostelium discoideum maintains a microbiome during multicellular development; bacteria are carried in migrating slugs and as endosymbionts within amoebae and spores. Bacterial carriage and endosymbiosis are induced by the secreted lectin discoidin I that binds bacteria, protects them from extracellular killing, and alters their retention within amoebae. This altered handling of bacteria also occurs with bacteria coated by plant lectins and leads to DNA transfer from bacteria to amoebae. Thus, lectins alter the cellular response of D. discoideum to bacteria to establish the amoebae’s microbiome. Mammalian cells can also maintain intracellular bacteria when presented with bacteria coated with lectins, so heterologous lectins may induce endosymbiosis in animals. Our results suggest that endogenous or environmental lectins may influence microbiome homeostasis across eukaryotic phylogeny.

Dictyostelium discoideum are amoebae that live in the soil and feed on bacteria; as food becomes scarce, they aggregate into a mound and form a migrating slug, which eventually develops into a fruiting body containing a spore-filled sorus on a cellular stalk (1). It had been thought that development occurred free of bacteria and that the final fruiting body was sterile (2). D. discoideum has an innate immune system composed of “sentinel” cells that are able to rid the slug of interstitial bacteria (3, 4); however, it was recently reported that about one-third of wild isolates maintain symbiotic interactions with the bacteria in their soil environment and carry bacteria stably through cycles of growth and development (5). In a form of microbial farming, carried bacteria can seed a new food supply for germinating spores as the spores establish a new colony of amoebae (5). Additional interactions between D. discoideum and their bacterial associates have been documented, and there is evidence that they are controlled by specific signals (68). Our finding that amoebal lectins mediate bacterial carriage suggests that D. discoideum actively controls its microbiome and that amoeba-bacterium interactions may serve as a useful model for microbiome homeostasis.

Amoebae that carry bacteria are easily distinguished from noncarriers because migrating carrier slugs leave behind bacteria that form colonies visible to the naked eye (fig. S1A). To assay for carriage, we allowed amoebae to initiate development in the presence of food bacteria such as Klebsiella pneumoniae (fig. S1B) and scored individual fruiting bodies for the presence of viable bacteria (fig. S1C). By this measure, >90% of a carrier strain’s sori contained bacteria (table S1) and bacterial endosymbionts in their spores (fig. S1D), whereas <10% of sori from noncarrier strains contained viable bacteria (table S1). It is unclear how bacterial carriage occurs, given the existence of an innate immune system, although significantly reduced sentinel cell populations in carrier strains have been reported (9). All of our carrier strains had sentinel cells (table S1), so we tested whether they cleared bacteria during extended slug migration. We found that carriers maintained their bacteria through >6 cm of slug migration, whereas no viable bacteria were recovered from noncarriers after slug migration, suggesting that sentinel cells do not influence bacterial carriage (table S1).

Because carrier strains maintain bacteria during development, we assessed their ability to kill bacteria as they transition from growth to development. The noncarrier laboratory strain AX4 killed extracellular bacteria within the plaque, whereas just outside the plaque, the bacteria were alive (Fig. 1A). Plaques of noncarrier strains were similar to those of AX4, whereas plaques of carriers had a mottled appearance, with live and dead bacteria segregated into distinct areas near the edges (Fig. 1B). Thus, extracellular bacterial killing within carrier plaques is uneven, and this might explain the presence of live bacteria as the amoebae begin to develop.

Fig. 1 Extracellular killing of bacteria by D. discoideum and secreted antibacterial proteins.

(A) Plaques of D. discoideum strain AX4 growing on lawns of K. pneumoniae were stained with fluorescent dyes that distinguish live and dead bacteria, as visualized by fluorescence microscopy (upper left panel) and differential interference contrast (DIC) microscopy (lower left panel). Numbers (1 and 2) indicate the areas that were sampled to reveal dead (red) and live (green) bacteria (right panels). (B) Plaques of wild carrier (car+) and noncarrier (car) strains, stained as in (A). (C) Carrier (circles) and noncarrier (squares and diamonds) amoebae were mixed with K. pneumoniae in buffer, and the number of viable bacteria were determined over time and normalized to 100% at the start of the assay. Values are means ± SEM for three independent experiments. (D) Secretomes from QS23 (noncarrier) and QS37 (carrier) were resolved by ion-exchange chromatography (DEAE-Sepharose) with a step-elution of increasing salt (NaCl, dashed line), and the fractions (1 ml) were assayed for antibacterial activity against K. pneumoniae. Trace activity was sometimes observed in carrier secretomes and elutes near the peak observed in noncarriers (asterisk). When discoidin I is removed from carrier secretomes, antibacterial activity becomes apparent (open circles).

We tested extracellular killing of K. pneumoniae by amoebae in suspension and found that noncarrier strains killed bacteria efficiently, but carrier strains did not (Fig. 1C). We tested whether this difference in bacterial killing might be explained by the proteins that they secrete during growth by examining the secretomes harvested from amoebae consuming K. pneumoniae. We fractionated the secretomes by ion-exchange chromatography and followed antibacterial activity by using an end-point dilution assay that measures the killing of K. pneumoniae (fig. S2A). Antibacterial activity in noncarrier secretomes eluted from the column in a broad peak and was heat-labile (10 min, 65°C), but carriers produced very little, or undetectable, activity (Fig. 1D; fig. S2, B to D; and table S2). This prompted us to identify regulators of bacterial carriage by screening for proteins that are differentially secreted by carrier versus noncarrier amoebae. The most prominent differences between carriers and noncarriers were two proteins in carriers with apparent molecular weights of ~24 and ~30 kDa (fig. S3A). We used mass spectrometry to determine that the 30-kDa band contained a mixture of the well-studied discoidin I lectins discoidin A (DscA) and discoidin C (DscC) (table S3) and that the 24-kDa band contained a mixture of the calcium-dependent adhesion protein CadA and the related proteins Cad2 and Cad3 (table S4) (1013).

Discoidin I forms dumbbell-shaped trimers with the subunits oriented in parallel, positioning the C-terminal H-type lectin domains at one end (14). Discoidin I binds to polysaccharides containing N-acetylgalactosamine (GalNAc) and galactosamine in β1–β3 linkages that are found in D. discoideum cell surface glycoproteins and bacterial carbohydrates (15, 16). In laboratory strains, discoidin I is found inside amoebae, and it is secreted when they aggregate after the start of development (17, 18). We surveyed the secretion of discoidin I in wild strains, using antibodies against DscA and DscC, and found that carriers appeared to secret higher levels of these proteins at the start of development than noncarriers (Fig. 2A). One carrier that we examined in detail secreted >80 times more discoidin I than noncarriers (fig. S3B). We examined the timing of discoidin I secretion from the end of vegetative growth through developmental aggregation and found that carriers secreted higher amounts of discoidin I at earlier times compared with noncarriers (e.g., Fig. 2B). These results suggest that discoidin I secretion before development is a physiological feature of carriers that may determine bacterial carriage. To explore this, we first measured bacterial binding and found that DscA binds quantitatively to K. pneumoniae bacteria, likely through the GalNAc moieties present in the lipopolysaccharide of many K. pneumoniae strains (Fig. 3A) (14, 19). DscA displayed saturation binding to K. pneumoniae and Escherichia coli (Fig. 3B), and we estimated that DscA binds to K. pneumoniae with micromolar affinity (association constant Ka = 1.2 × 106 M−1) through ~1.4 × 106 binding sites on each bacterium (supplementary materials) (20). About 2.75 × 105 DscA proteins would be required to form a monolayer covering the surface of one K. pneumoniae bacterium, assuming that the lectin ends of the roughly cylindrical trimers were opposing an idealized smooth surface with the shape and volume of the bacterium (supplementary materials). Thus, additional binding sites within the glycocalyx of K. pneumoniae likely account for the fivefold higher DscA binding that we observed.

Fig. 2 Discoidin I secretion by wild carrier strains.

Wild strains were harvested during late-stage growth on lawns of K. pneumoniae bacteria (when most of the bacteria had been consumed), shaken in Sor buffer for 1 hour before being separated into supernatant (S) and pellet (P) fractions, and resolved on SDS–polyacrylamide gel electrophoresis (SDS-PAGE) protein gels. (A) Western blots stained with antibodies against discoidin I, showing soluble and cell-associated discoidin I in carrier and noncarrier strains. The DscA/C control (left lane) was purified from AX4 (supplementary materials). (B) A representative time course of discoidin I production, from before bacteria are consumed to the start of the growth-to-development transition, for a noncarrier (QS14) and a carrier (QS68) strain. The 3-hour point is roughly equivalent to the time of harvest in (A).

Fig. 3 Discoidin I binds and protects bacteria.

(A) Increasing numbers of K. pneumoniae (K.p.) bacteria (1X = 2.3 × 107 bacteria) were incubated with purified DscA (300 μg/ml) in 100 μl buffer for 60 min at room temperature. Bacteria were separated from soluble DscA by centrifugation, and the supernatant (S) and pellet (P) fractions were resolved by SDS-PAGE and visualized by Coomassie blue staining. (B) DscA (12.5 to 62.5 μg) was incubated with 1.3 × 108 K. pneumoniae or E. coli in 100 μl for 60 min. Bacteria were removed by centrifugation, and the amount of unbound DscA was determined and used to calculate the amount of DscA bound to the bacteria (shown as means ± SEM). The asterisk indicates that no unbound DscA was detected. The dashed line shows the saturation point estimate. (C) K. pneumoniae bacteria, pretreated with DscA or bovine serum albumin (BSA), were suspended in buffer and challenged with Dabs. Bacterial viability was quantitated by counting live bacteria at various times, using a live-dead staining reagent and fluorescence microscopy, normalized to the zero-time samples. Statistically significant differences (**P < 0.01, ***P < 0.001) were observed between DscA-treated and BSA-treated bacteria at all time points (one-way pairwise analysis of variance using Tukey contrasts for multiple comparisons of means). Values are means ± SEM.

We tested whether discoidin I binding alters K. pneumoniae’s interaction with amoebae by examining bacterial killing, carriage, and endosymbiosis. First, we tested whether discoidin I protects bacteria from killing by D. discoideum antibacterial proteins (Dabs). DscA-bound K. pneumoniae were protected from killing by Dabs in an overnight outgrowth assay (fig. S2F) and in a short-term viability assay (Fig. 3C). Notably, the protection in the outgrowth assay diminished when the amount of discoidin added was below that needed to achieve saturation binding to the bacteria (fig. S2F). However, DscA-coated K. pneumoniae were not resistant to killing by heat (65°C, 10 min) or antibiotic treatment (kanamycin at 50 μg/ml). If protection by discoidin I is a mechanism by which carrier strains spare bacteria from killing, carriers would have to secrete sufficient discoidin to protect some bacteria as the amoebae enter development. Thus, we reexamined carrier secretomes for cryptic Dabs that might have been masked by discoidin I. Removing discoidin I from carrier secretomes by passing them through polygalactose affinity columns revealed latent antibacterial activity (Fig. 1D, fig. S2E, and table S2). Our results indicate that discoidin I can counteract the Dabs secreted by carriers, at least in vitro, and this suggests that discoidin I is secreted by carrier strains at levels sufficient to spare some bacteria for carriage as amoebae enter development.

We tested whether discoidin I influences the bacterial carriage and endosymbiosis that define the carrier phenotype. We mixed D. discoideum amoebae with K. pneumoniae or with DscA-coated K. pneumoniae at the start of development and tested the resulting fruiting-body sori for live bacteria. A single treatment of K. pneumoniae with DscA resulted in bacterial carriage by the noncarrier strain QS4 and the noncarrier laboratory strain AX4 (table S1). We also observed live green fluorescent protein (GFP)–expressing K. pneumoniae within the D. discoideum spores, but only when the bacteria were pretreated with DscA (Fig. 4A). These results suggest that extracellular discoidin I is sufficient to account for bacterial carriage in D. discoideum by binding bacteria, protecting them from secreted antibacterial proteins, and inducing carriage and endosymbiosis.

Fig. 4 Lectin-induced modified bacterial internalization (LIMBI) results in bacterial endosymbiosis, increased persistence of bacteria within amoebae, and genetic transformation.

(A) D. discoideum amoebae (QS4) were mixed with GFP-expressing K. pneumoniae that had been pretreated with buffer (mock) or DscA and allowed to develop into fruiting bodies. Fluorescence microscopy of the resulting spores shows intact bacteria in spores after LIMBI with DscA. (B) An overlay image of DIC and fluorescence microscopy, showing vegetative amoebae and GFP-expressing and DscA-coated E. coli under agar. The persistence of the bacteria within amoebae was quantified by determining the number of minutes that each bacterium was observed to remain intact. (C) Box plots of the persistence time of discoidin-coated bacteria within amoebae after LIMBI. The edges of the boxes represent the 75th and 25th percentiles, thick lines represent medians, the whiskers are the minimums and maximums, and dots are outliers. The difference between LIMBI and the no-lectin control was statistically significant (P < 0.001, Mann-Whitney-Wilcoxon rank-sum test). (D) LIMBI-mediated gene transfer from E. coli into D. discoideum. Discoidin-coated E. coli harboring a plasmid engineered to express a red fluorescent protein fusion to D. discoideum histone H2b (mCherry-H2b) were mixed with AX4 amoebae, followed by 10 days of drug selection for the plasmid. Shown are a DIC microscopy image, fluorescence microscopy images of nuclear DNA stained with DAPI (4′,6-diamidino-2-phenylindole) or mCherry, and an overlay showing nuclear expression of mCherry-H2b.

Some of the internalized discoidin-coated bacteria escape intracellular killing and digestion by the phagolysosomal pathway of the amoebae and end up in the spores. Thus, discoidin I alters the handling of bacteria by amoebae so that live bacteria persist, a phenomenon that we have termed lectin-induced modified bacterial internalization, or LIMBI. We demonstrated the specificity of LIMBI for lectin-coated bacteria with mixing experiments using discoidin-coated bacteria and uncoated bacteria. We presented DscA-bound red fluorescent protein (RFP)–expressing E. coli and GFP-expressing E. coli together to D. discoideum and followed their uptake by the amoebae over several hours by means of time-lapsed fluorescence imaging. Amoebae took up and digested uncoated bacteria, whereas some discoidin-coated bacteria remained intact and were scattered throughout the cytoplasm. The reciprocal experiment with discoidin-coated GFP-expressing E. coli resulted in retention of only those bacteria. The differential handling within the same amoeba of discoidin-coated and uncoated bacteria indicates that discoidin must be bound to the bacterium for it to persist after internalization. We quantified the persistence of bacteria after LIMBI by measuring the length of time that GFP-labeled bacteria remained intact within vegetative amoebae, using time-lapse imaging in an under-agar assay to create a pseudo–two-dimensional view (Fig. 4B) (21). By following bacteria from the time that they were internalized by amoebae to the time that they lost their structural integrity, we found that discoidin-coated bacteria persisted nearly seven times longer than uncoated bacteria (Fig. 4C).

LIMBI may provide an opportunity for bacterial DNA transfer into the host nucleus, and we examined this by inducing LIMBI of DscA-coated E. coli harboring a shuttle plasmid and selecting for drug-resistant amoebae. Each of the discoidin-coated E. coli strains introduced into D. discoideum by LIMBI led to genetic transformation of the amoebae at high frequencies (~2 to 20%), including those carrying plasmids with G418 (neor) or blasticidin (bsr) resistance genes (table S5). To visualize LIMBI transformation, we used plasmids that express D. discoideum histone H2b fused to the RFP mCherry and observed that all of the transformed amoebae expressed the mCherry-tagged histone within their nuclei (Fig. 4D and table S5). Immunoblots of proteins from these cells confirmed that the fusion protein was expressed (fig. S4A). Somewhat unexpectedly, ~10% of the amoebae expressed RFP-H2b 24 hours after LIMBI without selection for drug resistance, suggesting that transient expression also occurs (table S5). These results indicate that DNA can be transferred from the bacteria into the nucleus of the amoebae, implying that LIMBI is functionally distinct from the phagolysosomal digestion pathway that amoebae use to process bacteria as food.

We tested the generality of LIMBI in eukaryotes by testing plant lectins as elicitors and mammalian cells as recipients. We identified several plant-derived lectins that bound to E. coli, including Wisteria floribunda agglutinin, Dolichos biflorus agglutinin, soybean agglutinin (SBA), and concanavalin A (ConA) (22). We mixed lectin-coated bacteria with D. discoideum, using the same protocols developed with discoidin I, and found that each of the plant lectins appeared to be equally effective for LIMBI and LIMBI-mediated transformation (fig. S4, B and C, and table S6). We also introduced lectin-coated E. coli to mammalian cells in culture and found that DscA, ConA, and SBA were each equally effective for LIMBI and LIMBI-mediated transformation of mouse RAW264 macrophages (tables S5 and S6). Competition experiments with admixtures of GFP- and RFP-expressing bacteria showed that only lectin-coated bacteria persisted intact within RAW264 macrophages after overnight incubation (fig. S5). Our results show that LIMBI can occur in amoebae and in animal cells, using amoebal or plant lectins, suggesting that modified uptake and persistence of lectin-coated bacteria in eukaryotic cells is a general phenomena.

We have described physiological differences between D. discoideum carrier and noncarrier strains that appear to account for the salient features of bacterial carriage during social amoebae development. Because discoidin I is sufficient to induce carriage by noncarrier strains, its secretion by carrier strains provides a plausible mechanism for the transition of carrier amoebae from active bacterial killing and feeding to starvation and bacterial symbiosis. Prestarvation factor (PSF) is a ~70-kDa secreted protein that reports on the bacterial food supply at the growth-to-development transition (23, 24). PSF is sequestered by bacteria when bacterium/amoeba ratios are high, but as the bacteria are depleted by amoebal feeding, the level of free PSF rises and stimulates D. discoideum development through autocrine signaling. PSF has never been identified, and because discoidin I binds bacteria and affects the kill-versus-carriage dynamics of amoebae, we are testing whether discoidin I is PSF. Discoidin I’s H-type lectin domains are on the C termini of the trimer subunits, but the N-terminal discoidin domains have distinctive structural similarities to F-type lectins and have been proposed to bind fucose and sialic acid containing N- or O-glycans (14). If this is true, it is possible that the H-type lectin domains of discoidin I bind bacteria, leaving the discoidin domains free to interact with N- or O-linked glycans on the cell surface of D. discoideum amoebae to initiate LIMBI or engage in receptor-mediated signaling. If the uptake mechanism does involve amoebal glycoproteins as receptors, plant lectins might act as bacteria-bound ligands for those same receptors.

Lectinophagocytosis results in efficient bacterial uptake by eukaryotic cells through specific receptor recognition of lectins provided by the bacteria or the host cell (25, 26). Plant lectins can also mediate macrophage lectinophagocytosis (27, 28), and in some cases this is concomitant with the suppression of bacterial killing (29). It is well known that lectins target bacteria for destruction by the innate immune system of the host, but there are reports of lectins with functions unrelated to defense (30, 31). If lectin protection of bacteria within hosts proves to be a general feature of multicellular eukaryotes, it would provide a new perspective on the regulation of microbiomes in well-studied systems, including humans. For example, LIMBI may promote the maintenance of intracellular bacteria in the mucosal epithelium or myofibroblasts of the colon and contribute to inflammatory disease processes.

Supplementary Materials

Materials and Methods

Figs. S1 to S5

Tables S1 to S6

References (3246)

References and Notes

Acknowledgments: The authors thank W. F. Loomis, G. Shaulsky, and E. Ostrowski for providing insights during the course of this work. Funding: Dictyostelium Functional Genomics Program Project Grant from the National Institutes of Health (PO1 HD39691). Author contributions: C.D. and A.K. conceived and designed the experiments, C.D. directed and performed all of the experiments not done by others, T.F. and S.H. confirmed LIMBI and LIMBI-mediated transformation independently, T.F. performed the phagocytosis and short-term bacterial-killing assays, S.H. made the mCherry-H2b plasmid and performed the under-agar LIMBI imaging, and O.Z. quantified sentinel cells, performed the carriage assays of migrating slugs, and performed the amoebal bacterial-killing assays. A.K. wrote the manuscript with scientific and editorial input from C.D., O.Z., T.F., and S.H. Competing interests: The authors declare no competing interests. Data and materials availability: All data are available in the main text or the supplementary materials.

Stay Connected to Science

Navigate This Article