Our current understanding of the soil microbiome, and crosstalk therein, is limited by the complexity of the potential interactions between individual denizens (Hirsch and Mauchline, 2012). The Gram-negative myxobacterium Myxococcus xanthus is a predatory member of the soil ecosystem with the capacity to lyse and consume a broad range of microbial species (Morgan et al., 2010). Despite the widespread distribution of myxobacteria and their suggested role as ecological regulators, there has yet to be a definitive study into their ecological impact, or the control mechanisms used (Shimkets, 1990; Dawid, 2000).

Within the soil there are numerous species competing for nutrients and space, creating complex food webs which culminate in a small number of ‘apex’ style carbon sequesterers (Lueders et al., 2006). Throughout such competitive ecosystems, individuals often seek to enhance their long term survival via mutual cooperation, which imposes a fitness cost on the individual, but increases the fitness of the population as a whole (Strassmann et al., 2011). The formation of a biofilm structure is the most commonly characterized form of bacterial cooperation, providing communal resistance to various environmental stresses (Watnick and Kolter, 2000; Jefferson, 2004). In order to orchestrate population-wide behaviors such as biofilm formation, quorum signaling (QS) systems are often used by bacteria to regulate gene expression (Miller and Bassler, 2001). There have been numerous systems characterized, each producing and detecting highly specific quorum signals (Dong and Zhang, 2005; Waters and Bassler, 2005).

A typical QS system involves members of a population producing a specific signaling molecule, which diffuses out of the producing cell and accumulates in the local environment. The concentration of QS molecule is thus broadly indicative of the number of signal-producing cells. At a particular threshold concentration of signal (representing a quorum of producing organisms), a biological response to the signal is triggered, typically through collective regulated changes in gene expression (Bassler and Losick, 2006).

Gram-negative bacteria primarily use acyl homoserine lactones (AHLs) as their quorum signals, whereas Gram-positive species often use processed oligopeptides for communication. All AHLs contain the same core structure containing a lactone ring, with a variable R-group carbon chain which can differ in chain length and with the addition of side chains (Miller and Bassler, 2001). AHLs are made by LuxI homologs (AHL synthases), are sensed by LuxR homologs, and are rapidly degraded in the environment through ring-opening lactonolysis (Yates et al., 2002). On binding AHL, LuxR binds to specific operator sites, activating expression of target genes, including but not limited to luxI and luxR (Williams et al., 2007). The resulting positive loop has led to QS molecules commonly being described as autoinducers.

Myxobacteria are renowned for their multicellular lifecycle, which features group motility, social development, and co-operative predation, however, their social interactions are not mediated by AHLs (Muñoz-Dorado et al., 2016).

The ability to control cellular motility enables M. xanthus to actively seek new nutrient sources and invade prey colonies (Berleman and Kirby, 2009). Motility of M. xanthus is encoded by two distinct genetic systems. A-motility (adventurous) can be performed by individual cells, while S-motility (social) requires cells to work in concert (Hodgkin and Kaiser, 1979). Upon starvation, populations of cells orchestrate their motility, forming a series of distinct patterns, culminating in the formation of a fruiting body containing 10⁵ cells (Zusman et al., 2007; Kaiser et al., 2010; Muñoz-Dorado et al., 2016).

The developmental program which results in fruiting body formation involves three chemically characterized intercellular signals. The A-signal is a diffusible population density-dependent signal which acts as a quorum signal. A-signal is not, however, an AHL, but rather a mixture of proteases, peptides, and amino acids (Kaplan and Plamann, 1996). C-signal is a surface-associated signal which is exchanged between cells upon cell–cell contact, while the E-signal is a mixture of fatty acid iso-15:0 and diacylmonoalkyl ether lipid TG1 (Ahrendt et al., 2016). A-signal acts as a signal of starvation and population density, initiating the developmental program. As starvation progresses, C-signaling induces its own production, leading to progressively higher signaling on cell–cell contact. As levels of C-signal increase, cells progressively ripple, stream, aggregate, and then sporulate within the nascent fruiting body (Kim and Kaiser, 1990; Muñoz-Dorado et al., 2016).

When myxospores are returned to a nutrient rich environment they germinate, reverting to predatory vegetative cells (Otani et al., 1995). The conversion from myxospore to vegetative cell comprises at least two distinct steps; a loss of refractility and spore coatings, followed by an increase in cellular metabolism and the emergence of a rod shaped cell (Elías and Murillo, 1991). Collective germination of the spores within a fruit results in a population of germinants, immediately proficient at co-operative predation.

Myxococcus xanthus colonies use the co-operative release of extracellular digestion factors and secondary metabolites to enable predation upon a broad range of prey (Rosenberg et al., 1977; Cao et al., 2015). Protease secretion and growth rate are cell density-dependent (Rosenberg et al., 1977), implying co-operative action, and secreted public goods include outer membrane vesicles (OMVs) which can kill and lyse prey organisms (Evans et al., 2012).

Although M. xanthus does not make AHLs itself, we would nevertheless expect it to be surrounded by the AHLs produced by other soil (potentially prey) bacteria. We therefore tested whether QS molecules made by (and indicating the presence of nutrients in the form of) typical soil bacteria, might affect the nutrient-dependent social behaviors of myxobacteria.

The addition of AHLs affects several behaviors exhibited by M. xanthus. Several lines of evidence suggest this is a consequence of signaling rather than due to the chemical and/or physical properties of the added AHLs. Motility, as manifested by colony expansion rates, was insensitive to nutrient levels in the substrate (Figure 1), but was stimulated by AHLs at very low concentrations (2 μM gave half-maximal stimulation), implying that AHLs are not being used as a nutrient source. In addition, the magnitude of the behavioral response in M. xanthus differed depending on the specific AHL added. This variation is non-linear with respect to AHL side chain length, showing that M. xanthus has differing sensitivity toward specific AHLs. Loss of activity upon heating [presumably due to enhanced pH-dependent ring-opening lactonolysis (Yates et al., 2002)], also indicates an effect due to specific signaling rather than non-biological properties.

The addition of AHLs shifted the balance between vegetative cells and myxospores toward predatory vegetative cells. In addition, the predatory activity of vegetative cells was increased by AHLs. It is possible that the increase in predatory activity is purely a consequence of augmented motility rates increasing contact between prey and predator. This is plausible, given the antimicrobial activity of OMVs was unaltered in any way by AHLs, although we only tested activity against E. coli prey, and potentially AHLs could enhance the activity of OMVs against other prey organisms. Regardless of mechanism, AHLs indicative of prey stimulate predatory activity, and increase the predatory proportion of the population.

It is already known that M. xanthus is able to sense and respond to the presence of prey. The lipid phosphatidyl ethanolamine (PE) is a chemoattractant and its release from lysed prey potentially recruits further predators to the vicinity (Bonner and Shimkets, 2006). Following the invasion of a prey colony M. xanthus uses a chemosensory pathway to direct colony movement in a manner that allows for complete lysis of prey (McBride and Zusman, 1996; Berleman et al., 2006), and there is also the potential for prey to attract M. xanthus through elasticotaxis (Fontes and Kaiser, 1999).

It is thought that PE makes a suitable chemoattractant for M. xanthus because of the myxobacterium’s slow gliding speed. The low diffusion rate of PE in aqueous media makes a short range but stable concentration gradient to which M. xanthus can slowly respond (Bonner and Shimkets, 2006). AHLs are short lived due to pH-dependent lactonolysis (Tan et al., 2015), which means that long range signaling is precluded, and AHL gradients can only be established if there is continued production, i.e., by high densities of living prey (Nilsson et al., 2001).

The sensation of a signal by an organism other than the intended recipient means that the response of M. xanthus to AHLs could potentially represent a form of microbial eavesdropping, and AHLs as kairomones. Amongst metazoans, eavesdropping on chemical signals by predators has only been documented relatively recently (Hughes et al., 2010) compared to a wealth of literature regarding eavesdropping on acoustic signaling (Strauß and Stumpner, 2015). Bacterial eavesdropping on quorum signals has been described previously, with antagonistic and synergistic consequences (Chandler et al., 2012; Michie et al., 2016; Papenfort and Bassler, 2016), however, the work presented here seems to be the first example of a bacterial predator responding to signals produced by prey microbes. AHLs that stimulatex predation are produced by many known M. xanthus prey organisms, including Serratia marcescens, Pseudomonas syringae (Pham et al., 2005), P. aeruginosa (Evans et al., 2012), and Rhizobium sp. (our unpublished data), and we would therefore expect predatory eavesdropping to be widespread in soils. AHLs have been detected in biofilms at concentrations in excess of 600 μM, (Charlton et al., 2000), so the concentrations used in this study (up to 25 μM, with half-maximal effects at 2 μM) are likely to be encountered by M. xanthus in the wild.

It is not clear mechanistically how M. xanthus could respond to AHLs. No myxobacterial genome sequence encodes a LuxI AHL synthase (Pfam domain PF00765), and only one myxobacterium (Haliangium ochraceum) encodes a full-length LuxR homolog (Hoch_3098) containing the N-terminal autoinducer binding domain (PF03472). The P2TF database of DNA-binding proteins lists 12 proteins encoded in the M. xanthus genome which contain the C-terminal LuxR DNA-binding domain (PF00196). Six of those contain no recognizable sensory domain while five have N-terminal two-component system receiver domains, including the regulator of fruiting body development FruA.

The suppression of sporulation by AHLs does, however, open possibilities for identification of the genes responsible. Following random mutagenesis, enrichment of mutants with disrupted AHL-sensitivity could be achieved using individual-colony screens (Yu et al., 2016) in the presence of AHLs, followed by heat-killing of vegetative cells. It would also be interesting to determine which M. xanthus genes are differentially regulated on exposure to AHLs by using transcriptome sequencing methods. This would provide insights into whether enhanced predation is just a consequence of increased motility and an imbalance of vegetative cells:myxospores, or whether expression of ‘predatory’ genes is actively induced.

If the hypothesis of predatory eavesdropping by M. xanthus is correct, we would expect to see increased fitness of M. xanthus when predating an AHL-producing prey compared to a non-AHL producing prey. Experiments investigating such a phenomenon will be non-trivial to undertake though, as AHLs affect the biology of the producing organism, usually making them more resistant to environmental stressors (and presumably more recalcitrant to predation). It would also be difficult to disentangle AHL-induced increases in predatory activity from the alternative interpretation of AHL-enhanced susceptibility of prey to predation. We would also predict that other types of QS signaling molecules commonly secreted by prey (including peptides, farnesol, AI-2, A-factor, etc.) might affect myxobacterial predatory behavior, and that the addition of xenic signals might stimulate directional swarming toward the signal source.

Our understanding of trophic levels and predator-prey relationships within polymicrobial communities is currently limited, but receiving increasing research attention. The demonstration of a predator responding to signals secreted by prey, suggests that predator-prey interactions can be much more sophisticated than mere chance encounters between predator and prey, and phenomena such as QS, which exhibits a clear fitness advantage in the laboratory, may well have profound downsides in the wild.

DL designed and performed the experiments. DW conceived and supervised the project. DL and DW analyzed the data, wrote and edited the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.