A bacterial symbiont is converted from an inedible producer of beneficial molecules into food by a single mutation in the gacA gene

To determine why D. discoideum would carry bacteria that did not serve as a food source, we investigated the production of secondary metabolites by the two symbiotic P. fluorescens strains associated with the host-farmer D. discoideum QS161 (from now on referred to as the “farmer”). Both symbionts had identical 16S rRNA gene sequences; however, one strain, P. fluorescens PfB-QS161 (from now on referred to as “PfB”), served as a food source, and the other one, P. fluorescens PfA-QS161 (from now referred to as “PfA”), did not (SI Appendix, Fig. S1). We used a differential metabolomics approach to identify molecules with potential functional roles in the symbiosis (Fig. 1).

We grew the food source strain PfB and the nonfood source strain PfA in liquid media for 24 h. Initial LC-MS analysis of the ethyl acetate extract of the PfA culture revealed two major peaks that correspond to two different compounds: a previously unreported compound and the known antifungal pyrrolnitrin (Fig. 1) (17). The extracts of the PfB culture revealed one major diffusible small molecule—namely, the iron chelator pyochelin (Fig. 1) (18). The latter contains three stereogenic centers of which one (C2′′) can easily epimerize to yield a mixture of diastereoisomers known as pyochelin I and II. Furthermore, pseudomonads are known to produce two enantiomeric (mirror image) forms of pyochelin: (4′S, 4′′S) and (4′R, 4′′R), depending on whether they contain a cysteine epimerase domain in their biosynthetic gene cluster (19). Whereas fluorescent pseudomonads biosynthesize (4′S, 4′′S) pyochelin, P. aeruginosa, biosynthesizes (4′R, 4′′R) pyochelin (19). A measurement of the optical rotation revealed that PfB produces the enantiomer characteristic of fluorescent pseudomonads.

The unknown compound found in the PfA extract was produced in liquid culture at a concentration of ∼0.3 μg/mL (1.3 μM) after 24 h of cultivation at 30 °C; it was isolated as a pale yellow oil with the empirical formula C₁₄H₁₈O₃ as determined by high-resolution electrospray ionization-TOF mass spectrometry combined with both ¹H and ¹³C NMR spectroscopy (SI Appendix, Table S1 and Figs. S2 and S3; Fig. 1). The ¹H NMR spectrum in deuterated methanol displayed three downfield protons at 6.33, 5.86, and 5.78 ppm, a double doublet at 4.51 ppm, and 12 aliphatic protons between 0 and 3 ppm. The ¹³C NMR and gradient heteronuclear single-quantum coherence (gHSQC) spectra showed five aromatic, two olefinic, one oxygen-bearing aliphatic, and five aliphatic carbon signals. A combination of ¹H-¹H correlation spectroscopy (COSY), gHSQC, and heteronuclear multiple-bond correlation (HMBC) spectra allowed us to identify the molecule as the previously undescribed 3-ethyl-2-propyl-2H-chromene-5,7-diol (hereafter referred to as “chromene”; Fig. 1).

We confirmed the identity of both pyrrolnitrin and enantio-pyochelin by a combination of high-resolution mass spectrometry and ¹H NMR spectroscopy data, which were compared with published data (SI Appendix, Fig. S5 and Table S2). We then investigated the role of the main secondary metabolites produced by the nonfood strain PfA in the symbiotic association with the farmer strain.

In many symbiotic associations between bacteria and eukaryotes, the former produce secondary metabolites that protect their host from pathogens or display other beneficial effects. Pyrrolnitrin, which is produced by PfA, the nonfood source symbiont of D. discoideum, has potent antifungal activity against a variety of soil-borne fungi (20); it is also known to play a crucial role in other symbiotic associations between the producer P. fluorescens strain and plants. The decreased susceptibility of an infected host to soil-borne pathogens can be used as a form of biocontrol, and fluorescent pseudomonads play an important role as biocontrol agents (21–24). It is possible that Dictyostelia carrying pyrrolnitrin-producing strains like PfA could be similarly protected.

The discovery of robust pyrrolnitrin production by PfA (0.4 μg/mL = 1.6 μM in a 24-h liquid culture) was in some ways also surprising because this metabolite is potentially toxic to many eukaryotes, including D. discoideum. In one recent study on predator–prey chemical warfare involving P. fluorescens as the prey and another amoeba (Acanthamoeba castellanii) as the predator, pyrrolnitrin was toxic, and supernatants of amoeba cultures strongly induced the production of the molecule (25). Farmer resistance and nonfarmer sensitivity would indicate that the farmers adapted to harboring a potentially lethal symbiont.

To test whether the farmer has evolved the ability to resist the toxicity of its bacterial symbiont as a consequence of continuous exposure to its metabolites, we subjected the farmer D. discoideum strain to a range of concentrations of pyrrolnitrin. As a control, we investigated the effect of this compound on a nonhost, nonfarmer clone, D. discoideum QS160 (hereafter referred to as the “nonfarmer”), which was isolated in the vicinity of the farmer and was thus likely to have occasionally encountered the farmer D. discoideum strain as well as its symbiotic bacteria (PfA and PfB) and their secreted molecules. In D. discoideum, the number of spores produced from a given number of amoebas is a good indicator of the success of the social process, so we used this measure to evaluate the effect of pyrrolnitrin on both D. discoideum strains. To determine variation of spore production as a function of the concentration of pyrrolnitrin, we put vegetative amoebas in their exponential growth phase on a nitrocellulose filter paper soaked with starvation buffer containing different concentrations of pyrrolnitrin. Fruiting bodies form within 24 h; at this time point we counted the number of spores contained within the fruiting bodies.

The farmer strain and the nonfarmer strain performed differently in the presence of pyrrolnitrin (Fig. 2). Though the nonfarmer suffered from the toxic effect of pyrrolnitrin, the farmer did not, and in fact did better than controls. We interpret these results as an adaptation of the farmer to pyrrolnitrin’s toxic effects.

The presence of pyrrolnitrin in PfA suggested that chromene might have similar or related biocontrol activities. In a 24-h liquid culture, chromene was found to be present in comparable concentration to pyrrolnitrin (chromene: 0.3 μg/mL = 1.3 μM; pyrrolnitrin: 0.4 μg/mL = 1.6 μM).

Initial bioassays, however, revealed no activity against the model fungus Saccharomyces cerevisiae or the gammaproteobacterium Escherichia coli, and only modest activity against Bacillus subtilis (IC₅₀ = 82 μg/mL = 35 μM). Because we found no obvious indirect effects (inhibition of D. discoideum’s microbial competitors) for chromene, we investigated possible direct effects on the farmer D. discoideum.

Treatment with chromene increased spore production in the farmer strain (Fig. 3), whereas the same treatment resulted in a decrease in spore production by the nonfarmer D. discoideum strain. The onset of visible differences in spore production occurred at concentrations on the order of 1 ng/mL (∼4 nM). This concentration of chromene is likely to be ecologically relevant because it is ∼300-fold lower than the concentration of chromene in a 24-h culture of PfA.

Having determined the molecular basis for farmer-beneficial effects of the nonfood symbiont PfA, we investigated the genetic difference between this bacterium and the food source PfB.

Because Pseudomonas fluorescens strains PfA and PfB both have identical 16S rRNA gene sequences, it seemed plausible that they shared a recent common ancestor. If so, only a few alterations in their genomes might have resulted in their observed phenotypic differences. Both PfA and PfB were thus characterized by Next Generation Illumina sequencing, and the resulting reads were assembled using the genome of Pf-5, which was sequenced in 2005, as a scaffold (26).

Because PfA and PfB had different secondary metabolite profiles, it seemed plausible to expect differences in biosynthetic genes of the two strains. We focused on the known biosynthetic gene clusters for the metabolites pyrrolnitrin and enantio-pyochelin. Though only PfA produces pyrrolnitrin, and only PfB produces enantio-pyochelin, both PfA and PfB had identical coding sequences for the biosynthetic gene cluster for pyrrolnitrin and enantio-pyochelin. This finding indicates that the difference between the small molecules produced by the two strains results from differences in regulation rather than biosynthetic potential. As a result, we searched for regulatory genes that, when mutated, could lead to the observed phenotypes. The search focused on two-component systems as the most likely regulators of gene expression in bacteria.

A canonical two-component system is composed of a histidine sensor kinase and a cognate response regulator. Upon stimulation, the sensor kinase undergoes autophosphorylation to yield a phosphorylated histidine, which in turn activates the response regulator by phosphotransfer to an aspartyl residue (27, 28). The response regulator then mediates the output, typically a change in gene expression (29).

We found a point mutation in PfB generating a premature stop codon that resulted in a truncated gacA gene product (Fig. 4). A cytosine-to-thymine mutation was responsible for the conversion of wild-type gacA glutamine-164 to a stop codon in PfB’s gacA gene.

The GacA–GacS system is a well-known global regulator in many pseudomonads. In the opportunistic pathogen P. aeruginosa, for instance, the GacS–GacA two-component system regulates the expression of acute and chronic virulence determinants (30). In P. fluorescens, the GacA–GacS system is a global regulator for antibiotic production (31, 32). Upon binding to an as-yet-unidentified external signal, this regulatory system upregulates the production of pyrrolnitrin, and downregulates the production of siderophores such as pyochelin. Spontaneous mutations disabling the GacA–GacS system have been reported in P. fluorescens that were mainly characterized by a change in colony morphology (33), not unlike the differences in morphology between the PfA and PfB strains in this study. Furthermore, such mutations in pseudomonads are known to occur in the rhizosphere of plants where they display a role in competitive root colonization (34). The point mutation observed in PfB introduces a premature stop codon upstream of the helix-turn-helix motif required for DNA binding, and inability to bind DNA prevents the GacS response regulator from activating transcription (35) (Fig. 4).

To verify that a mutation in this two-component system was sufficient to explain the chemical and edibility differences between PfA and PfB, we deleted the entire gacA gene in PfA. A clean deletion was obtained by allelic replacement (36). The gacA knockout of PfA displayed the PfB phenotype: it could now serve as a food source for the farmer D. discoideum (SI Appendix, Fig. S1) and the secondary metabolite profile was the same as for PfB (Fig. 5). However, a complementation of PfB with PfA supernatant (containing pyrrolnitrin and chromene) did not turn PfB into a nonfood source, so it seems unlikely that these two secreted molecules are responsible for the inedibility of PfA (SI Appendix, Fig. S10). Instead, the edibility difference between the strains must result from some other, unknown downstream effect of gacA, which is not surprising given that its inactivation affects expression of 10% of the genes in the genome (32).

A single mutation transforms a beneficial but inedible P. fluorescens strain into a food source for D. discoideum. The nonfood source produces secondary metabolites that provide chemical defenses (pyrrolnitrin) and increase spore production in the farmer (chromene and pyrrolnitrin). This mutation in the gacA gene generates independent but closely related symbiotic strains that perform strikingly different roles in the symbiosis.

Though the genetic difference between the two bacterial strains was determined, a detailed phylogenetic analysis is required to understand their evolutionary history.

It seems likely that the functional gacA gene is ancestral because there are many more possible loss-of-function mutations than gain-of-function mutations. In support of this hypothesis, we examined the gacA sequences from the 11 other P. fluorescens strains with sequenced genomes and also the two additional P. fluorescens strains we have isolated from other D. discoideum farmer clones (Pf-QS68, Pf-QS152). All gacA sequences appeared functional; none had the stop codon at position 164 or any other evident disabling stop codon or frameshift-causing insertion or deletion.

Thus, the stop mutation that causes edibility is clearly derived. The similarity of PfA and PfB suggests they are very closely related, yet both of them display a significant number of SNPs (136 in 5,561 homologous genes) showing they diverged well before we isolated them in the laboratory. We then estimated the maximum-likelihood phylogeny using DNA sequences of 20 conserved genes for all 15 strains, using Pseudomonas syringae as an out-group. The result (Fig. 6) shows very clear support (100% bootstrap) for a clade that includes the previously sequenced Pf-5 along with the four strains collected from D. discoideum. There is some uncertainty (50% bootstrap) about the relative positions of the outermost clones in the clade (Pf-5 and Pf-QS68). This best tree is consistent with either two gains of the trait of being carried by D. discoideum or with a single gain plus a loss in Pf5, and because of uncertainty at the key node, we also cannot rule out a less-supported tree with Pf5 basal to the other four, with a single gain.

The tree also strongly supports PfA and PfB being the closest relatives whose sister group is another farmer-carried clone, Pf-QS152. Although the result could change with additional clone sampling, the most parsimonious conclusion is the edibility of PfB evolved in a strain already being carried by D. discoideum. This analysis suggests the unusual result that an organism evolved to be eaten by another, which makes sense only through kin-selected benefits to clonemates, more of whom will be carried to new locations by the well-fed farmer D. discoideum clone.

The original description of primitive farming by the social amoeba D. discoideum distinguished two types of amoebas: farmers, which had fruiting bodies containing several types of bacteria, and nonfarmers, which were bacteria-free (11). Roughly half of the farmer-associated bacteria could serve as a food source, and farmers transported to bacteria-poor environments were able to sow food bacteria and eventually consume this bacterial food. Using a differential metabolomics analysis coupled with metabolite isolation and characterization, we showed that a nonfood source bacterial symbiont (PfA) produces secondary metabolites that are beneficial for the farmer. One of the metabolites, chromene, functions directly by potently enhancing spore formation at ecologically relevant concentrations in the farmer strain, and suppressing spore formation in the nonfarmer strain. The other, pyrrolnitrin, showed similar effects, and it is likely to counter microbial pathogens of other species. The farmer strain was resistant to the inhibitory effects of pyrrolnitrin and chromene, which indicated that the farmer had adapted to carrying the nonfood strain PfA. Chemically uncharacterized supernatants collected from different non-food bacteria carried by other D. discoideum clones also help their farmer clone and harm non-carrier clones (37). Taken together, these observations provide a strong argument for coevolution of both the farmer and symbionts; they also establish that the symbiosis is not just a food and farmer symbiosis, but rather displays some of the multipartite qualities of other farming symbioses, such as the fungus-farming ants and bark beetles with their crop fungi providing the food and the bacterial symbionts providing chemical defenses against fungal pathogens (5, 7–9). What distinguishes this farming symbiosis is the close relatedness of PfB and PfA—the bacterial food source and the small-molecule producer (although this close relatedness does not hold for bacterial symbionts of other Dictyostelium farmers). A single point mutation in the gacA gene is sufficient to completely alter the chemical repertoire of the nonfood source, thus effectively changing its role in the symbiotic association from secondary metabolite producer to food source.

Finally, these findings highlight the usefulness of investigating the small-molecule chemistry that underlies so much of bacterial–eukaryotic symbiotic associations as a source of both new chemistry (molecules and pathways) and biology (defense, development, cooperation, and evolution).

We grew both P. fluorescens strains at 30 °C in SM/5 liquid media containing 2 g d-glucose, 2 g Bacto Peptone (Oxoid), 2 g yeast extract (Oxoid), 0.2 g MgCl₂, 1.9 g KHPO₄, and 1 g K₂HPO₄ per liter. We grew the farmer (QS161) and nonfarmer (QS160) Dictyostelium discoideum from spores on SM/5 agar plates (SM/5 liquid media supplemented with 15 g agar per liter) in association with Klebsiella pneumoniae as bacterial food source, at room temperature (21 °C) in Petri dishes (100 × 15 mm). For cocultures, we grew the farmer QS161 in association with its bacterial symbionts PfA and PfB (in a 1:10 ratio). For the complementation assay, we grew the farmer QS161 in association with PfB and 200 µL of cell-free PfA supernatant.

A 1-L culture of the respective bacterial strain in SM/5 media was grown for 24 h at 30 °C and 200 rpm. After 24 h, we extracted 15 mL of the culture with 25 mL ethyl acetate (EtOAc). We separated the organic phase, dried it over Na₂SO₄, and concentrated it in vacuo. The cocultures were soaked with EtOAc, and the organic phase was separated by filtration and concentrated in vacuo. The crude concentrates were redissolved in 15% (vol/vol) methanol in water and applied to a Waters C18 Sep-Pak column (0.5 g) and washed with 15% (vol/vol) acetonitrile in water to remove polar molecules. The column was eluted with 100% acetonitrile, and solvents were removed in vacuo to yield a brown concentrate.

The concentrates were then subjected to HPLC-MS analysis using an Agilent HPLC system equipped with a diode array detector, and a 6130 Series quadrupole mass spectrometer was used with a Phenomenex Luna C18 (5 µm, 100 Å 100 × 4.6 mm, flow rate = 0.7 mL/min) column. We used the following gradient for HPLC-MS analysis: 0−1 min, isocratic 30% acetonitrile in water + 0.1% formic acid; 1−21 min, linear gradient from 30% acetonitrile in water + 0.1% formic acid to 100% acetonitrile + 0.1% formic acid.

We prepared log-growth amoeba by plating 2 × 10⁵ spores of each clone in association with K. pneumoniae on SM/5 agar plates. Amoeba log growth occurs ∼32–36 h after plating spores. At this time, we collected amoebas and washed them free of bacteria using ice-cold aqueous KK2 buffer (2.2 g KH₂PO₄ and 0.7 g K₂HPO₄ per liter). We determined amoeba density with dilution using a hemacytometer and a light microscope. For the filter-pad assay, we prepared dilutions of either chromene or pyrrolnitrin in KK2 + 0.5% DMSO from stock concentrations prepared in 100% DMSO. We used 150 × 15 mm Petri plates lined with two layers of Whatman no. 3 (Schleicher & Schuell) soaked with either control buffer or the various test dilutions laid with a grid of equidistant 13-mm square AABP 04700 (Millipore) black filter squares. To test social-stage spore production, we spotted the filters individually with 1.25 × 10⁶ amoebas in starvation buffer, and we prepared duplicate samples for each clone for each experiment. We allowed the clones to hatch, grow, and develop under direct light to limit potential movement of slugs before final culmination to fruiting bodies. Fruiting body formation for all clones was complete after ∼24 h. We allowed the spores to mature in the fruiting bodies for an additional 24–48 h before collection. To determine spore number, we placed each test filter with fruiting bodies in an Eppendorf tube containing 1.0 mL KK2 + 0.1% Nonidet P-40 alternative (Calbiochem). The tubes were vortexed briefly to disperse the spores, and the spores were counted without dilution using a hemacytometer. We calculated spore number for experimental treatments as a percent change compared with control based on spore number recovered from starvation buffer control samples.

Genomic DNA of PfA-QS161, PfB-QS161, Pf-QS68, and Pf-QS152 was obtained from a 5-mL overnight culture at 30 °C in SM/5. Extraction of DNA was performed using GenElute Bacterial Genomic DNA Kit (NA2100; Sigma Aldrich). Next Generation Illumina sequencing with 100-bp single-end reads was performed at Harvard Medical School Biopolymers Facility. Reads were mapped to the reference genome sequence of Pf5 using Burrows–Wheeler Aligner software with default parameter settings (38). The SNPs and consensus genome sequences were called using SAMtools (39). A maximum-likelihood tree was constructed from concatenated DNA sequences of 20 genes (dnaG, gcp, infB, ksgA, nusA, nusG, rplA, rplC, rplE, rplF, rplK, rplN, rpoB, rpsB, rpsC, rpsD, rpsG, rpsH, secY, and ychF), a subset of the topologically congruent genes as reported by Bapteste et al. (40), from newly sequenced strains and other P. fluorescens strains with completed genome sequences that are retrieved from NCBI’s BioProjects database (www.ncbi.nlm.nih.gov/bioproject/browse/) using the HKY model by MEGA5 (41). P. syringae pv. tomato strain DC3000 was included as an outgroup. We used DNA sequences rather than amino acid sequences because the group including the four D. discoideum carried clones, and Pf-5 differed by only one amino acid in these 20 genes.

Antibiotics were used in culture media at the following concentrations: gentamicin 15 μg/mL for E. coli and for P. fluorescens. To counterselect E. coli donor cells in gene-replacement experiments, nalidixic acid was used at a concentration of 15 μg/mL; mutant enrichment was performed with gentamicin at a final concentration of 15 μg/mL. Small- and medium-scale preparations of plasmid DNA were made with the QIAprep Spin Miniprep kit (Qiagen, Inc.). PCR was performed using a KOD hot start polymerase kit (Novagen, EMD Millipore). DNA fragments were purified from agarose gels with QIAquick Gel and PCR Purification Kit (Qiagen, Inc.). All constructs obtained by PCR techniques were confirmed by sequence analysis (Genewiz). A clean gacA deletion mutant of P. fluorescens PfA-QS161 was constructed using the suicide plasmid pEXG2 (42). Primer sequences and a detailed description of cloning procedures are provided in SI Appendix.