Molecular recognition in myxobacterial outer membrane exchange: functional, social and evolutionary implications



Through cooperative interactions, bacteria can build multicellular communities. To ensure that productive interactions occur, bacteria must recognize their neighbours and respond accordingly. Molecular recognition between cells is thus a fundamental behaviour, and in bacteria important discoveries have been made. This MicroReview focuses on a recently described recognition system in myxobacteria that is governed by a polymorphic cell surface receptor called TraA. TraA regulates outer membrane exchange (OME), whereby myxobacterial cells transiently fuse their OMs to efficiently transfer proteins and lipids between cells. Unlike other transport systems, OME is rather indiscriminate in what OM goods are transferred. In contrast, the recognition of partnering cells is discriminatory and only occurs between cells that bear identical or closely related TraA proteins. Therefore TraA functions in kin recognition and, in turn, OME helps regulate social interactions between myxobacteria. Here, I discuss and speculate on the social and evolutionary implications of OME and suggest it helps to guide their transition from free-living cells into coherent and functional populations.


Microbes live in hostile environments and are surrounded by predators and unpredictable dangers. Within these microbial jungles there are siblings and other cells that can cooperate to allow individuals to thrive in groups. How cells recognize beneficial partners from foes is critical for building functional communities, multicellular organisms and for self-preservation. Although cellular recognition is widely ascribed to eukaryotic cells, the analysis of how bacteria physically identify their neighbours is still an emerging field. In contrast, recognition from afar by diffusible or quorum signalling is better understood in bacteria. In this MicroReview, I will discuss a newly discovered recognition system in myxobacteria (Pathak et al., 2013). These δ-proteobacteria are useful model microbes for studying cell–cell recognition because they live in diverse communities and exhibit sophisticated social interactions. The molecular determinant of the recognition system described here is a cell surface receptor called TraA, of which identical or very similar alleles must be present in both partnering cells to direct a process whereby cells transiently fuse their outer membranes (OMs) and exchange proteins and lipids – a behaviour henceforth called OME (Wall, 2014). A background and biological framework are first provided, followed by details of OME recognition and implications in the sociobiology and evolution of myxobacteria. Comparisons and analogies are also made with other bacterial and eukaryotic systems to help interpret the biological role of OME.

Multicellularity in myxobacteria

To place OME in a proper context, an overview of myxobacteria is warranted. These Gram-negative bacteria are typically found in soil, herbivore dung and plant material. However, since they can form quiescent and environmentally stable spores that are easily dispersed, they are consequently found on nearly any surface or body of water on the planet (Reichenbach, 1999; Dawid, 2000). Myxobacteria feed on decaying organic matter or act as predators that kill and consume other microbes (Berleman and Kirby, 2009). Myxobacteria are noted for their social behaviours, which include gliding motility and multicellular development (Whitworth, 2008; Pathak et al., 2012b). In one of their more spectacular displays of cooperation, free-living myxobacteria will coalesce into functional cohorts that erect fruiting bodies in response to starvation (Fig. 1). Within these fruits, which consist of 105–106 cells, cells differentiate into dormant myxospores.

Figure 1.

Cell–cell recognition mediates the transition of heterogeneous to homogenous populations. Myxobacteria are common inhabitants of soil, environments in which microbial diversity is high. On the left different species are represented by different shapes and colours. Under appropriate conditions of high cell density and starvation, vegetative myxobacteria transition into aggregates that give rise to the formation of a fruiting body (Chondromyces crocatus; micrograph courtesy of Hans Reichenbach).

Our understanding of the ecology of myxobacteria, like most microbes, is limited. What we do know is that a soil niche can apparently contain hundreds of different myxobacterial species or subspecies, based on sequence analysis of the highly conserved 16S rRNA gene (Wu et al., 2005). In a species specific study of Myxococcus xanthus strain diversity, soil isolates that contained identical 16S rRNA sequences, were found in detailed analysis of polymorphic loci to be very different from one another (Vos and Velicer, 2006). That is, among the 78 M. xanthus isolates obtained, they parsed into at least 45 distinct strains (Vos and Velicer, 2009). Expanding on this theme of microbial diversity, modern DNA microarray technologies can now resolve over 33 000 different bacterial and archaeal species from a soil sample (Mendes et al., 2011). Based on classical methods, which may underestimate these difficult-to-culture organisms, myxobacteria have been found to constitute a significant fraction of the soil microbial population, ranging from 1000 to 500 000 cells per gram of soil and perhaps representing up to 1% of the bacterial community (McCurdy, 1969; Reichenbach, 1999; Dawid, 2000). Given the cosmopolitan environments that myxobacteria reside in (Vos et al., 2013), and their notable ability to assemble social groups, would seem to suggest they have evolved means to discriminate between related and non-related individuals to build social groups.

Although myxobacteria live in communities composed of many different microbial species and strains, the process of forming fruiting bodies appears to be selective (Fig. 1). This conclusion is based primarily on anecdotal findings from various investigators that fruits isolated from environmental sources typically consist of a single species (Dworkin, 1996). In particular, H. Reichenbach, who spent a career isolating ∼ 6000 myxobacterial isolates in the pursuit of novel secondary metabolites (Reichenbach, 2001), found that fruiting bodies almost always contain a single species (pers. comm.). A detailed study of fruiting body composition from wild isolates also found fruits consisted of a single species and haplotype analysis further found that individuals within a given fruit were highly related (Kraemer and Velicer, 2011). In a notable exception to this theme, Reichenbach and co-workers did discover that a number of Chondromyces crocatus isolates contained phylogenetically distant companion species (Jacobi et al., 1996). As isolating the C. crocatus and the companion from each other was difficult, these species appeared to have formed a symbiotic relationship and presumably evolved an inter-species recognition system (Reichenbach, 1999).

The ability of free-living cells to recognize others and assemble multicellular cohorts, of course, is not limited to myxobacteria. For instance, slime moulds, such as Dictyostelium spp., are confronted with the same dilemma. During vegetative growth, these single-cell amoebae similarly forage for food and, in response to starvation, coalesce into a multicellular slug that develops into a fruiting body (Strassmann et al., 2011). One mechanism Dictyostelium has evolved to ensure that related cells enter the fruit to sporulate, and to exclude unrelated cells from exploiting this environment, is the TgrB1-C1 recognition system (Hirose et al., 2011; Ho et al., 2013). These proteins are polymorphic cell surface receptors and through heterophilic interactions mediate self-recognition (Chen et al., 2013). Because of their relative ease of experimental manipulation, Myxococcus and Dictyostelium are model systems used to probe sociobiological and evolutionary questions (Strassmann et al., 2011).

OME discovery

The discovery of the OME can be traced back to pioneering work over 35 years ago (Hodgkin and Kaiser, 1977). In those early studies, a small subset of motility mutants was found to be rescued when mixed with other non-motile mutants. This extracellular complementation (also known as stimulation) requires that the ‘donor’ strain contains the wild-type version of the gene that is mutated in the non-motile stimulatable ‘recipient’. Stimulation requires physical contact between cells (Wall and Kaiser, 1998). These mutants fell into six complementation groups that corresponded to six genes called cglB/C/D/E/F and tgl (Hodgkin and Kaiser, 1979). These genes were subsequently identified, and the common link between them is that they contain either type I or type II signal peptides, indicating that they are localized in the cell envelope – more precisely, they are all likely to reside in the OM (Rodriguez-Soto and Kaiser, 1997b; Rodriguez and Spormann, 1999; Simunovic et al., 2003; Pathak and Wall, 2012). The mechanism of stimulation occurs by the transfer of the missing protein function from the donor to the mutant recipient (Fig. 2) (Nudleman et al., 2005). Thus once the recipient receives the missing function, gliding motility is restored. To easily visualize stimulation, the donor strain is also non-motile and non-stimulatable, so the resulting flares that emanate from the agar inoculum correspond only to the recipient cells that gained the ability to glide (Nudleman et al., 2005). As these stimulated recipients move and become physically separated from the donors that replenish their proteins, over time the transferred proteins are diluted below a threshold level by cell division and protein turnover, and thus stimulation ceases.

Figure 2.

Model for OME. The interaction between two cell poles is illustrated. Recognition is mediated by compatible TraA alleles that are expressed as cell surface receptors (blue triangles; left). TraA interactions are proposed to be homophilic and is followed by transient fusion of the OMs, which allows lipids and proteins (green and red colours on two cells) to be bidirectionally exchanged by membrane diffusion between cells (right, yellow OMs signifies lipid mixing between cells).

In the case of tgl, we have a deeper understanding of how stimulation works. Tgl is a lipoprotein that contains tetratricopeptide repeats and functions as a ‘pilotin’ for assembly of the PilQ secretin (Rodriguez-Soto and Kaiser, 1997a; Nudleman et al., 2005; 2006). Therefore in tgl mutants, the absence of Tgl prevents the formation of type IV pili motor (Wall et al., 1998; Wall and Kaiser, 1999). However, during stimulation, where OME allows the transfer of Tgl from the donor strain, PilQ is assembled in the tgl mutant, and functions as an OM gated channel that allows polymerization and retraction of the type IV pili that powers social motility (Pathak et al., 2012b). Thus, although type IV pili-dependent gliding motility is a complex machine that fails to assemble in a tgl mutant, the mutant, interestingly, is competent for assembly once Tgl is provided (Kaiser, 1979; Wall and Kaiser, 1998).

Mechanism of OME

To date the mentioned six endogenous proteins are the only ones known to be transferred by OME; however, the total number of transferred proteins is likely to be in the hundreds (Wei et al., 2011). The primary reason for this assertion is that the only cis requirement for a heterologous reporter protein to be transferred is for it to contain a type II signal sequence for OM localization. Given that M. xanthus has over 400 predicted lipoproteins (Bhat et al., 2011), and that some transferred proteins contain type I signal peptides (Pathak and Wall, 2012), there are many other candidate proteins for OME. Because protein cargo transfer is relatively non-specific and OME occurs bidirectionally between cells (Fig. 2), OME is functionally distinct from other dedicated bacterial transport systems, for example conjugative systems and type III and VI secretion systems, which transfer specific molecules unidirectionally to target cells (Tseng et al., 2009; Hayes et al., 2010; Konovalova and Sogaard-Andersen, 2011). The mentioned reporter, SSOM-mCherry, is a fluorescent protein and along with fluorescent lipid markers, transfer of proteins or lipids can be monitored in live cells (Wei et al., 2011; Pathak et al., 2012a). In contrast, reporters that contain an inner membrane lipoprotein sorting sequence or are localized in the cytoplasm are not transferred (Wei et al., 2011). Because lipids are transferred during OME, the mechanism is proposed to involve the fusion of the OMs (Fig. 2) (Pathak et al., 2012a,b; Ducret et al., 2013).

OME is an efficient and relatively fast process, as partnering cells share exchangeable components to approximately equal levels in a matter of minutes (Nudleman et al., 2005; Wei et al., 2011; Pathak et al., 2012a). In other words, if one cell lacks a particular protein, after OME both resulting cells will have roughly the same amount of this protein. These striking features of OME have also been independently verified (Ducret et al., 2013). One reason that OME is so robust is that cells can undergo multiple rounds of OME among different partners before the next round of cell division.

Like stimulation, OME requires that cells be in physical contact and on a hard surface (Wall and Kaiser, 1998; Wei et al., 2011). Transfer does not occur in liquid or even when cells are densely packed on a soft agar surface. Based on these and other observations, transfer does not involve extracellular secretion or diffusion. Moreover, cell motility (A- or S-motility) plays a critical role in transfer, because transfer is barely detected in non-motile cells. The role of motility in transfer is, however, indirect, as only the donor or recipient – or even a third-party cell that is not involved in OME – needs to be motile to facilitate transfer (Wei et al., 2011). The role that motility plays in OME is not simply to create local areas of high cell densities, as concentrating non-motile cells by centrifugation does not alleviate the motility requirement (X. Wei and D. Wall, unpublished). Instead, motility likely facilitates proper cell alignments and contacts that lead to exchange. The role of cell alignment in OME was originally made by monitoring Tgl-dependent stimulation of type IV pili production (Wall and Kaiser, 1998), and was subsequently confirmed by monitoring SSOM-mCherry transfer (Wei et al., 2011). Cell alignment enriches end-to-end contacts, which may be critical for the formation of contact junctions between juxtaposed cells, and consistent with this, the TraA receptors are primarily localized to cell poles (Pathak et al., 2013). Cell motility may also facilitate engagement of cell surface receptors and, through pushing and pulling forces, may also help to destabilize the local environment of the OM to trigger fusion (Kozlov et al., 2010). The forces generated by gliding motility could also contribute to surface tension, which in some cases can catalyse receptor oligomerization (Lipke et al., 2012), a process that could have functional importance. Subsequently, gliding motility powers fission, which is necessary to separate cells after OME.

We now know that OME and stimulation, manifestations of the same biochemical process, require the TraA and TraB proteins in both engaged cells (Pathak et al., 2012a). These genes form an operon where their open reading frames overlap and mutations in either gene yields an identical phenotype that blocks OME. Since TraA and TraB both contain type I signal sequences they are predicted to localize in the cell envelope and may physically interact (Pathak et al., 2012a). As described in more detail below, our recent findings show that TraA functions as a cell surface receptor and the suggested interactions are homophilic (Fig. 2) (Pathak et al., 2013). The molecular anatomy of TraA is consistent with it functioning as a receptor as there is a PA14-like domain, followed by Cys-rich tandem repeats and a putative protein sorting tag called MYXO-CTERM. This domain organization is analogous, for example, to the yeast receptor proteins called FLO (Smukalla et al., 2008; Goossens and Willaert, 2010; Pathak et al., 2012a). In our current model, when two juxtaposed cells make contact and engage TraA receptors, we propose TraA functions as a fusagen to catalyse OM fusion (Fig. 2). Following fusion OM contents can laterally diffuse and exchange between cell membranes. An OM fusion mechanism is also consistent with end-to-end contacts because membrane curvatures, which occur at the cell poles, are known to lead to membrane destabilization that helps promote fusion (Martens and McMahon, 2008; Wickner and Schekman, 2008).

An extreme form of membrane curvature found in myxobacteria are narrow tube-like structures that protrude out from the cell surface and are composed of the OM (X. Wei, C. Vassallo, D.T. Pathak and D. Wall, unpubl. data). It has been suggested that these intriguing OM tubes might serve as conduits for OME (Ducret et al., 2013; Remis et al., 2013; X. Wei, C. Vassallo, D.T. Pathak and D. Wall, unpubl. data). A time-lapse movie by Mignot and colleagues indicate that OM tubes can form following cell–cell contacts, and that lipids, but apparently not lipoproteins, can be transferred in these tubes (Ducret et al., 2013). Although this finding is striking, experimental evidence is actually mixed whether OM tubes play a critical role in OME. For example, the only known components required for OME are not required for OM tube production, as traA and traB mutants make OM tubes at wild-type levels (X. Wei, C. Vassallo, D.T. Pathak and D. Wall, unpubl. data). Conversely, TraAB overexpression causes cells to adhere together independent of OM tubes (Pathak et al., 2012a). Additionally, conditions that promote abundant OM tube production, namely liquid growth, prevents OME from occurring (Wei et al., 2011; X. Wei, C. Vassallo, D.T. Pathak and D. Wall, unpubl. data). Future work will need to elucidate how OM tubes are made and to determine if they are required for OME. In an analogous report, a Gram-positive bacterium, Bacillus subtilis, was found to produce cytoplasmic membrane derived tubes, called ‘nanotubes’, which connect cells in biofilms together (Dubey and Ben-Yehuda, 2011). Here, nanotubes are also proposed to transfer cell content, but in contrast to OME they transfer cytoplasmic material, and surprisingly, do so indiscriminately with distantly related bacteria such as Escherichia coli.

The function of TraB in OME is less clear. Aside from a C-terminal OmpA-like domain predicted to bind the cell wall, the remaining N-terminal ∼ 400-amino-acid region shows no apparent homology and functional tests of this protein has yet to be done in detail. Possible functions for TraB include that (i) it facilitates export of TraA to the cell surface, and/or (ii) it forms a complex with TraA that functions in OM fusion. TraB appears not to be directly involved in cell recognition, because heterologous express of different traA alleles, which alters strain recognition, all function with the same traB allele from strain DK1622 (Pathak et al., 2013). Last, orthologues of TraAB are restricted to the order Myxococcales, which is consistent with the idea that OME is a social behaviour characteristic of this group of bacteria.

Membrane fusion, mitochondria and homeostasis

Why do myxobacteria undergo OME? The simple answer is to deliver OM contents among cells, but given the cargo is relatively non-specific, why is this ability important? Eukaryotes offer insight to this question as examples of membrane fusion are bountiful and the machinery involved in these processes are also divergent (Martens and McMahon, 2008). In all cases, of course, the underlying purpose is cargo transfer, but important specifics vary. In some instances the purpose is to deliver genetic material, as found in envelope virus infection of host cells, mating in Saccharomyces cerevisiae, multi-cell fusion in Dictyostelium discoideum to form a macrocyst and sperm and oocyte fusion (Saga et al., 1983; Martens and McMahon, 2008). In other cases, the purpose of cell fusion is to deliver the contents of an entire cell to form large multinucleated syncytia, as in the case of muscle fibre development (Aguilar et al., 2013). In contrast, OME does not involve DNA delivery or multi-cell fusion. In eukaryotes, vesicle fusion can function in inter-cellular communication, for example in neurotransmitter and hormone delivery (Jahn et al., 2003). As mentioned below, OME may play a role in cell–cell communication.

The process that may be the most informative analogue of OME is the fusion/fission cycling of mitochondria. In a typical mammalian cell, there are hundreds of mitochondria (Chan, 2012). Each mitochondrion in turn contains about a thousand different proteins, of which the majority is encoded in nuclear DNA, whereas only a minority of the proteins and RNAs are encoded in the mitochondrial genome. This complexity of mitochondrial numbers and diverse components creates a logistical problem for how these organelles are faithfully made. The answer to this problem lies in their ability to mix and homogenize population contents by rounds of fusion and fission cycling (Youle and van der Bliek, 2012). To quote D.C. Chan, ‘If mitochondrial fusion did not exist, each of the several hundred mitochondria in a typical cell would act autonomously, and their biochemical and functional profiles would diverge’ (Chan, 2012). We similarly hypothesize that OME helps myxobacteria to transition towards population homeostasis (Pathak et al., 2012a). In another example of co-ordination and homeostasis, cells within the mammalian immune system exchange cell surface protein for the apparent purpose to create an integrated and uniform immune response (Davis, 2007).

Mitochondria dynamics also allow cells to respond to stress and mutational burden (Youle and van der Bliek, 2012). Mitochondria are exposed to high levels of reactive oxygen species (ROS), a by-product of oxidative phosphorylation, which damages proteins, lipids and DNA. This exposure, in conjunction with the high copy number of mitochondria per cell, makes their genomes particularly susceptible to mutations. Thus the deleterious effects of missing or damaged component can be diluted among a pool of healthy mitochondria. In essence this functions by complementation without the exchange of DNA (Nakada et al., 2001; Youle and van der Bliek, 2012). Consistent with this idea, stress induces mitochondrial hyperfusion (Youle and van der Bliek, 2012). Similarly, in myxobacteria, mutations that cause defects in OM function can be complemented by OME without DNA exchange (Pathak et al., 2012a). Whether myxobacterial cells mitigate deleterious defects caused by environmentally damaged OM components being diluted among the population is plausible but remains to be tested.

Mitochondrial OM fusion is catalysed by homotypic interactions between mitochondrial surface proteins called mitofusins (Mfn1 and Mfn2 in mammals; dynamin family members) (Chan, 2012; McNew et al., 2013). The ability of mitochondria to fuse is also facilitated by their mobility (Cagalinec et al., 2013). Similarly, OME fusion is proposed to be driven by homotypic interactions between TraA receptors and by cell motility (Fig. 3) (Wei et al., 2011; Pathak et al., 2012a). In conjunction with OM fusion, mitochondria usually also fuse their inner membranes in a mechanistically distinct step involving another dynamin protein called Opa1. In contrast, myxobacteria do not fuse or exchange inner membrane or cytoplasmic components.

Figure 3.

Colour-coded model for cell–cell recognition mediated by homotypic TraA interactions.

A. A mixed population of Myxococcus strains is shown (left). Genetic relatedness of cells and TraA alleles (particularly the PA14-like domain) is illustrated by cell body and receptor (triangle) colours respectively. In some cases, the genomic relatedness between cells might be relatively low, e.g. M. fulvus (shown as a tan cell) with other M. xanthus isolates, but their traA alleles (PA14-like domain) are highly related (as indicated by the same/similar colour of the triangles). Only cells with the same/similar TraA alleles bind through presumed homotypic interactions. Functional pairs of cells are shown at the right indicating that pairing can occur within each of the groups. Environmental strain names are also shown.

B. Two different myxobacteria with incompatible traA alleles cannot engage in OME (top). In contrast, when the traA allele in the aqua strain is replaced with that of the partner cell OME occurs (bottom). For details of recognition groups and allele replacement experiments, see Pathak et al. (2013).

Molecular recognition by TraA

OME involves intimate cell–cell contacts during which significant amounts of cellular goods are exchanged. The nature of these interactions suggested to us that OME may use self/non-self-recognition to identify partnering cells, as sharing large quantities of cellular resources with unrelated cells seemed implausible. To investigate specificity, OME was tested by use of fluorescent markers between distantly related bacteria, for example E. coli and Pseudomonas aeruginosa, and found not to occur. Next, OME was tested against a panel of different Myxococcus environmental isolates (Pathak et al., 2013). From a combination of mixing experiments, it was found that all tested Myxococcus isolates were competent to exchange with themselves, but they showed discriminatory behaviour towards other strains. Thus the 17 tested strains fell into five major recognition groups (Fig. 3A). Isolates within a recognition group exchanged with each other and failed to exchange with strains outside their group. From this panel some strains were very selective and only exchanged with themselves; other isolates transferred to themselves and up to seven other isolates (Fig. 3A) (Pathak et al., 2013). One group, referred to as ‘supergroup D’, exhibited a more complex behaviour. That is, supergroup D could be divided into subgroups that all exchanged with each other and some members also exchanged with other subgroups; however, no strain exchanged with all supergroup D members. Of the 213 transfer pairs tested, 188 were conducted in a reciprocal fashion where the fluorescent marker, which monitors unidirectional transfer, was switched between pairs. In every case tested (94/94 pairs) the reciprocal experiment gave the identical results indicating bidirectional strain transfer; in no case was marker transfer unidirectional (Pathak et al., 2013).

With the discovery that OME involved strain recognition, we next sought to identify the specificity determinant. Because TraA is localized to the cell surface and is an adhesin, it served as a logical candidate (Pathak et al., 2012a; 2013). To test for specificity, the traA alleles from the 17 previously analysed strains were sequenced and analysed. The alleles were found to be polymorphic within a hyper-variable region that encompasses a domain that exhibits distant homology to the widely distributed PA14 domain, a domain generally implicated in glycan binding (Rigden et al., 2004). Importantly, sequence conservation within this polymorphic domain showed perfect correlation with the recognition groupings described above (Pathak et al., 2013). Thus siblings or strains with similar traA alleles formed partners for OME, whereas strains with distantly related alleles did not exchange (Fig. 3A). The heterogeneous transfer patterns found within supergroup D, was also mirrored by diverse polymorphisms in this group; however, importantly, PA14-like domain sequence conservation still correlated among recognition partners. Definitive proof that TraA was the molecular specificity determinant was demonstrated by swapping traA alleles between strains, which resulted in predictable changes in partner recognition (Fig. 3B) (Pathak et al., 2013). Striking, traA allele replacements could also reengineer social interactions mediated by OME between M. xanthus and Myxococcus fulvus species. These genetic experiments demonstrate that TraA determines partner recognition and the simplest interpretation of these combined results is that TraA functions as a homophilic cell surface receptor (Fig. 3). However, direct biochemical tests for homophilic TraA interactions remain to be performed.

Other recognition systems

Myxobacteria and other bacteria reside in complex ecological communities and molecular recognition is a necessary step for functional communities to form (Fig. 1). In the context of OME, one ecological niche can in fact contain isolates that belong to different recognition groups (Fig. 3A) (Wu et al., 2005; Vos and Velicer, 2006; Pathak et al., 2013). TraA thus provides one means of cell recognition; however, there are also other factors to consider in how social cohorts, e.g. fruiting bodies, are formed. Other mechanisms include: (i) As the name ‘myxo’ implies, myxobacteria produce a slimy extracellular matrix that coats and tightly bundles cells together and restricts their dispersion (Dana and Shimkets, 1993). Although this ‘sticky’ matrix holds cells together, it is unknown to what extent molecular recognition may be involved and to what degree cells actually stay bound in their environment. (ii) Similarly, myxospores residing in fruits are tightly bound together by a strong cohesive extracellular matrix, and thus upon germination the population may remain intact to establish a community. (iii) A blunt mechanism for distinguishing self from non-self is resistance to bacteriocins or antibiotics. Thus myxobacteria that lack immunity to these factors will perish. Indeed, inter-species bacteriocin-mediated killing has been proposed to explain territorial behaviour and enrichment of species within fruiting bodies after different species were experimentally mixed (Smith and Dworkin, 1994). (iv) Similar to other bacteria, most notably Proteus mirabilis, two different strains of myxobacteria typically form demarcation lines that appear to prevent swarms from merging (Gibbs et al., 2008; Vos and Velicer, 2009). This response suggests there is a recognition system, and genetic studies indicate that this behaviour is specific to the social motility system and involves a Lon protease called BsgA (Gill et al., 1988). (v) Last, clonal expansion is a means for creating a homogeneous population. In soil, where resources are scarce and competition is fierce, such growth would likely be precluded. However, under certain conditions, for instance, by inoculation of a fresh dung pellet, a single myxobacterium could feed and clonally expand. On a grander scale, a giant 12 m clonal patch of D. discoideum was discovered in a Texas, USA, pasture, demonstrating that large-scale clonal expansion of microbes does occur in soil (Gilbert et al., 2009). These and perhaps other factors contribute towards myxobacteria forming coherent populations that perform multicellular tasks.

Other bacterial species use molecular recognition to build microbial communities. Perhaps the best understood microbial community is found in our mouth. Kolenbrander and colleagues have developed species and molecular maps that describe specific spatiotemporal relationships that allow dental plaque to form by oral bacteria (Kolenbrander et al., 2002). Recognition between species and within species is mediated by cell surface receptors and adhesins that allow an orderly assembly of species into coaggregates that form biofilms. In E. coli there are cell surface adhesins that bind to related or sibling cells to allow the formation of aggregates or biofilms. One example is TibA, which is a multifunctional autotransporter, belonging to the type V secretion system, that recognizes neighbouring cells through homophilic receptor interactions (Sherlock et al., 2005; Cote and Mourez, 2011). CDI (contact-dependent inhibition) is another type V-based secretion system, which makes an adhesin that binds other E. coli cells. Specifically, the CdiA protein recognizes the cell surface BamA receptor on neighbouring cells, resulting in the transfer of a toxin located on its tip. The CDI system is widely distributed in proteobacteria, and the toxin domain and immunity factors are polymorphic (Aoki et al., 2010). Therefore, analogous to OME, the CDI systems parse into different functional groups. Interestingly, interactions between polymorphic toxin/immunity groups are antagonistic, whereas CDI interactions among siblings can be cooperative (Aoki et al., 2010; Garcia et al., 2013; Ruhe et al., 2013). Strikingly, the amino acid sequences on the BamA loops where CdiA binds are invariant among hundreds of E. coli isolates (Ruhe et al., 2013). In contrast, between closely related bacterial species these loops are polymorphic. Thus, CdiA binding and growth inhibition functions are species specific and provide a means for intra-species competition. In another example, P. mirabilis uses a self-recognition mechanism to create clonal boundaries and establish territories between rival strains (Gibbs et al., 2008). The molecular players involved in this process constitute two different identity systems that are secreted by a common type VI secretion pathway that likely requires cell–cell contact (Wenren et al., 2013).

Social outcomes from OME

OME is a social behaviour that has broad implications for how myxobacteria interact. As described above, one postulated function is to allow the growth of a coherent population. Another recently discovered function is that OME can transfer novel capabilities. Specifically, myxobacteria produce bacteriocins that kill other strains (Hirsch, 1977), and in strain mixing experiments protection from inter-strain killing was found to occur by a Tra-dependent mechanism that is allele specific (Pathak et al., 2013). We thus hypothesize that protection is conferred by the transfer of immunity factors. Details of how killing and immunity work remain to be elucidated.

We have also suggested that OME plays a role in cell–cell communication. This idea is based on the findings that TraAB functions are required for the decision of a mixed strain population to swarm or develop (Pathak et al., 2012a). That is, swarm expansion of a motile strain is blocked when it is mixed with a non-motile strain and this inhibition is dependent on OME. In other words, if either the motile or non-motile strain contains a tra mutation, there is a ‘relief’ of inhibition (Pathak et al., 2012a). From these findings, we hypothesized that the non-motile cells produce a signal that is transferred by OME to motile cells that blocks their motility or development (Pathak et al., 2012b). The advantage of this proposed system is that it allows disparate cells, based on genetic or physiological differences, to function as a team in multicellular behaviours. In contrast, when heterogeneity is removed from a population the function of OME is circumvented. For example, tra mutants grown as monocultures under ideal laboratory conditions swarm and develop with no overt defects (Pathak et al., 2012a). This finding suggests that the proposed five cell–cell signalling pathways involved in M. xanthus development (Shimkets, 1999; Kaiser, 2004) function independent of OME. However, since these signalling molecules and pathways are not fully elucidated and were defined by extracellular complementation experiments similar to stimulation of motility (Pathak et al., 2012b), more detailed studies are needed to test whether there might be redundant functions between OME and developmental signalling.

TraA path towards multicellularity

The described properties of TraA provide a conceptual framework for how myxobacteria might transition from single cells to multicellular life (Fig. 4). The ability of TraA to act as a recognition adhesin allows cells to identify siblings and related cells. Upon TraA recognition, the cells transiently fuse their OMs and mix components. Since the OM serves as the cell's gateway to the external world, and is uniquely exposed to environmental stresses that can damage this cellular compartment, the ability of cells to mix components allows the population to adapt to stresses and physiological differences between cells can be minimize. In turn OM homeostasis will help synchronize the populations' behaviours by ensuring individual cells produce and respond to inter-cellular signals in a similar manner. Combined with a role in cell communication, OME can help transition a mixed population into a cooperative and coherent unit (Fig. 4).

Figure 4.

Working model for how TraA contributes towards multicellularity. See text for details.

Evolution of cooperativity

According to the theory of evolution, life is a struggle and those that survive are the fittest. Because natural selection acts primarily on the individual, and not at the level of the species or population, how cooperative behaviours, which are found in many species, have evolved represents a challenge to the theory of evolution (Axelrod and Hamilton, 1981). Recognizing this dilemma and that genetic change drives evolution, Hamilton, in seminal work, developed a generally accepted mathematical formula for how cooperative social interactions can exist between related individuals in the context of evolutionary theory (Hamilton, 1964a,b). To help explain cooperative behaviour at the molecular level, Hamilton postulated the existence of ‘supergenes’ (Hamilton, 1964b). Later, Dawkins fleshed out this idea and proposed a new name, the ‘greenbeard’ effect (Dawkins, 1976; 1982). The greenbeard concept suggests that a single gene in an individual can lead to the expression of a trait and that this trait can be recognized in other individuals, which may or may not be kin, to confer preferential, cooperative or altruistic interactions. Although Hamilton and Dawkins thought it was unlikely that a single gene functioned in this way, the idea was valuable because it provided a mechanistic framework for how cooperativity could evolve.

Haig subsequently provided a tangible explanation for how greenbeard genes could function (Haig, 1996). He suggested that proteins that function as homophilic receptors could express a perceivable feature, are able to recognize other cells that bear the same gene and could confer beneficial social treatment to those cells. In addition, for specificity to occur, the homophilic receptor must also be polymorphic in the population and those polymorphisms must have functional consequences. As outlined above, traA has those properties (Pathak et al., 2013), and a few other greenbeard examples have been reported that have some or all of those properties (Queller et al., 2003; Smukalla et al., 2008; Hirose et al., 2011; Ruhe et al., 2013). For clarity, it should also be mentioned that an expanded definition of greenbeard has been used, which includes complex genetic systems, e.g. the Ti plasmid, and competitive or ‘Darwinian-like’ interactions, which encompass bacteriocins (Gardner and West, 2010).

More than 15 years ago, Haig also postulated that myxobacteria and Dictyostelium species use greenbeard genes as a ‘security surveillance’ mechanism to identify kin or greenbeard-related cells from non-kin cells for selective inclusion in fruiting body formation (Haig, 1997). In Dictyostelium, the tiger gene products appear to function in such a discriminatory role to ensure the social behaviour of forming a fruiting body and sporulation is not exploited by free-living ‘cheater’ cells (Ho et al., 2013). Although it is not clear whether TraA functions in security surveillance, TraA does appear to function as a ‘sentinel’ for recognizing other greenbeard members (Fig. 3).

An inherent problem with cooperation is that this behaviour can be threatened by cheaters. For instance, a population that secretes public goods into their environment can easily be exploited by a mutant that benefits from the goods, but does not contribute towards the cost in their production (Sandoz et al., 2007). OME represents a new paradigm for how bacterial goods can be shared. Here, the goods are ‘restricted’, as only recognition partners can participate, whereas non-members are excluded (Fig. 3A). Because both partner cells must express cognate TraA alleles and the interaction involves direct reciprocity (Nowak, 2006); cheating cannot easily occur in the sense that both cells must share their OM resources. Although OME is reciprocal, under certain conditions, resource might be unevenly shared. For example, if one cell was starving or had a damaged OM, the exchange of goods could favour one cell and be detrimental to the other. Such an act could be deemed altruistic. However, from a traA gene's eye perspective this interaction is not altruistic as it can improve the fitness of that traA allele pool, as the beneficiary cell has an identical/related traA allele to the benefactor cell. Last, cheating can occur in this scheme if a mutant cell does not make a full complement of OM components and instead obtains those resources from other cells by OME.

One theoretical advantage OME offers is a means to recruit a wider cohort of cells during stressful times. That is, under typical environmental conditions it seems unlikely that > 105 sibling cells would be available for fruiting body recruitment. OME thus provides a greenbeard mechanism to override certain strain differences, for example, by relieving lethal actions of bacteriocins (Pathak et al., 2013), to allow a larger reservoir of individuals to be recruited. In other words, in the absence of OME two different strains might instead compete with each other leading to their mutual demise. An extension of this idea suggests that if there were no TraA selectivity, the size of the pool could be further increased. Although this line of thought has merit, natural selection counteracts this supposition. To quote D. Haig, ‘Natural selection … is a process by which things of one kind become things of a different kind’ (Haig, 2012). Thus the introduction of TraA selectivity increases the chances that individuals are related.

An extension of the TraA concept suggests that its function could also lead to intra-genomic conflict, whereby it facilitates cooperation and beneficial treatment to non-kin cells. In evolutionary terms this is called an ‘outlaw’ gene, and such a function would presumably be selected against (Dawkins, 1982). The clearest example of how TraA function could elicit genomic conflict is whereby OME overrides the killing activity of presumed bacteriocin(s) produced in that same cell (Pathak et al., 2013). Although under certain conditions TraA could have outlaw properties, such functions would likely be conditional, as under different conditions M. xanthus isolates may need to put aside their competitive differences to unite.

In summary, TraAB mediated OME is one of the social behaviours that myxobacteria engage during their multicellular life. Future studies are needed to elucidate mechanistic details and social consequences of OME that were outlined here. These studies may also offer insights into how other types of cells interact.


I thank Amy Fluet for expert editorial assistance and members of my lab for their enthusiastic pursuits that led to discoveries described here. This work was supported by NSF Grant MCB-848141 and NIH Grant GM101449 to D.W.