Physiological Heterogeneity Triggers Sibling Conflict Mediated by the Type VI Secretion System in an Aggregative Multicellular Bacterium

ABSTRACT A hallmark of social microorganisms is their ability to engage in complex and coordinated behaviors that depend on cooperative and synchronized actions among many cells. For instance, myxobacteria use an aggregation strategy to form multicellular, spore-filled fruiting bodies in response to starvation. One barrier to the synchronization process is physiological heterogeneity within clonal populations. How myxobacteria cope with these physiological differences is poorly understood. Here, we investigated the interactions between closely related but physiologically distinct Myxococcus xanthus populations. We used a genetic approach to create amino acid auxotrophs and tested how they interact with a parental prototroph strain. Importantly, we found that auxotrophs were killed by their prototroph siblings when the former were starved for amino acids but not when grown on rich medium or when both strains were starved. This antagonism depended on the type VI secretion system (T6SS) as well as gliding motility; in particular, we identified the effector-immunity pair (TsxEI) as the mediator of this killing. This sibling antagonism resulted from lower levels of the TsxI immunity protein in the starved population. Thus, when starving auxotrophs were mixed with nonstarving prototrophs, the auxotrophs were susceptible to intoxication by the TsxE effector delivered by the T6SS from the prototrophs. Furthermore, our results suggested that homogeneously starving populations have reduced T6SS activity and, therefore, do not antagonize each other. We conclude that heterogeneous populations of M. xanthus use T6SS-dependent killing to eliminate starving or less-fit cells, thus facilitating the attainment of homeostasis within a population and the synchronization of behaviors.

M ulticellular organisms can arise by two fundamental strategies. In one case, multicellularity develops from a single cell, often a fertilized egg. Here, a single cell serves as a bottleneck through which multicellularity arises by clonal expansion. This ensures that all progeny cells are genetically identical and physiologically similar, at least within tissues, as cell growth occurs in the protective environment of the organism. Alternatively, multicellularity occurs by the aggregation of cells from the environment. Here, cells may have genetic, adaptive, age, and/or fitness differences. These differences lead to heterogeneous cell populations that may impede the cooperative interactions among cells required for multicellularity. Indeed, organisms that evolved multicellularity by an aggregative strategy develop into organisms with limited complexities, namely, fruiting bodies, as opposed to the more complex cell types and tissues associated with plant and animal species (1).
Myxococcus xanthus is a model bacterium that forms multicellular fruiting bodies by the aggregation of thousands of cells in response to starvation. During vegetative growth, M. xanthus lives in the soil as a microbial predator. Through gliding motility, M. xanthus cells explore their environment for food as social packs or as individuals. Their social behaviors require intercellular communication, motility, and an ability to synchronize the actions of thousands of cells to build spore-filled fruiting bodies (2). Engaging in social interactions is energetically costly for participant cells in terms of signal production, resource sharing, coordinating behaviors, and the differentiation of specialized cell types in which all cells do not necessarily share equal benefits (3). In addition, cooperative behaviors lead to conditions that can be exploited by individuals that use shared resources but do not equally contribute to their production. Therefore, mechanisms are needed to regulate social interactions and to limit exploitation by so-called cheater cells (4).
To address genetic heterogeneity and possible social conflicts, aggregative species need a way to recognize kin and/or discriminate against nonkin. Kin recognition in myxobacteria is mediated, in part, by a polymorphic surface receptor called TraA (5)(6)(7). Cells that share identical or nearly identical TraA receptors recognize and bind each other and exchange outer membrane proteins and lipids in a process called outer membrane exchange (OME). The consequences of OME depend on the nature of the cargo transferred between partnering cells. In some cases, OME rescues fitness defects associated with cells that contain damaged or missing cell components, such as proteins or lipids (8)(9)(10). In other cases, OME delivers toxin effectors, and recipient cells that do not contain cognate antitoxins are discriminated against (11,12). More broadly, many bacterial species use polymorphic toxins or bacteriocins to eliminate neighboring or competing cells that are not true clonemates. For example, the type VI secretion system (T6SS) is a commonly used system to deliver toxins to adjacent cells. These effectors are delivered through the contraction of an extended sheath that is anchored to the cell envelope and wrapped around an inner tube made up of the Hcp protein.
The tip of the tube is composed of a VgrG trimer and PAAR domain proteins. Upon contraction, the T6SS tube penetrates the membrane(s) of target cells and intoxicates them if they lack the cognate antitoxin or immunity protein (13,14).
Much less is known about how bacteria deal with physiological differences among cells within a clonal population and how such heterogeneity influences cell-cell interactions. Even in genetically uniform communities, physiological differences emerge because of spatial variations in nutrient availability and physical and/or chemical differences within microenvironments (15). In addition, cells vary by age, the damage load that they carry (e.g., their level of oxidized proteins and extent of membrane oxidation), and adaptation differences, along with stochastic differences in gene expression or phase variation (16)(17)(18). Whereas physiological heterogeneity may offer fitness benefits to populations by providing a wider array of adaptive responses to changing environments (19)(20)(21), the ability of social microbes to coordinate behaviors largely depends on synchronized actions of cells responding collectively to cellular and environmental cues. Because cells within populations are physiologically different, we hypothesized that M. xanthus evolved a mechanism(s) to manage heterogeneity, thus facilitating a transition toward the synchronization of social behaviors required for multicellularity.
Here, we examined how genetic siblings interact within physiologically distinct M. xanthus populations. To address this, we created auxotroph strains and mixed them with their parental prototroph strain on minimal medium. In these mixtures, one population starves while the other grows. We asked whether these kin cells antagonize, cooperate with, or act indifferently toward one another. We found that antagonism occurred between kin M. xanthus strains only under conditions under which they were physiologically distinct. We show that this antagonism is dependent on the T6SS, and we identify the effector-immunity pair that mediates this interaction, which provides a model for how antagonism occurs.

RESULTS
Phenotypic analysis of auxotroph mutants. To test for social interactions between starving and nonstarving populations of kin M. xanthus cells, we created auxotroph strains. Auxotrophs were made by disrupting genes predicted to be essential for the biosynthesis of histidine, arginine, threonine, or tryptophan. An auxotroph phenotype of each mutant was tested by its inability to grow on defined A1 minimal medium and the rescue of the growth defect when the medium was supplemented with the corresponding amino acid. However, while conducting these studies, we found that it was necessary to add calcium chloride to restore cell motility on A1; the absence of motility was caused by insufficient Ca 2ϩ levels from the ultrapure water and the agar that we used to make A1 (Fig. 1A). Restoring motility was important considering its essential role in a number of M. xanthus social behaviors such as predation, development, and OME (22)(23)(24). All of the constructed auxotroph mutants exhibited the expected growth phenotypes on A1 (Fig. 1B).
Auxotroph mutants are conditionally antagonized by a prototroph during mixed growth. To investigate social interactions between physiologically distinct populations, a histidine auxotroph (hisC) labeled with tdTomato was mixed with an unlabeled prototroph strain at a 1:1 ratio and plated on A1 and A1 supplemented with 2 mM CaCl 2 (A1 Ca 2ϩ ). Mixed cultures incubated on A1 Ca 2ϩ supplemented with histidine, rich medium (CTT), or starvation medium (TPM) were used as controls. Prototroph-auxotroph mixtures were harvested to determine the competitive index, which was expressed as a change in the strain ratio (prototroph to auxotroph) as determined by counting cells under a microscope. After 48 h, mixed colonies from the three control media maintained a near-1:1 ratio. In contrast, the competitive index from minimal media, both A1 and A1 Ca 2ϩ , was increased in favor of the prototroph ( Fig. 2A). The finding that the prototroph cells outcompeted the auxotroph cells was an expected outcome because the growth of the latter strain is impaired on A1 medium lacking histidine (Fig. 1B). However, there was a dramatic difference in strain ratios between A1 and A1 supplemented with calcium. Notably, auxotroph cells were almost undetectable by microscopy examination of mixed colonies derived from A1 Ca 2ϩ . A time course experiment showed an unchanged strain ratio during the first 12 h, followed by a steady increase in prototroph numbers on A1 that plateaued after 48 h, whereas on A1 Ca 2ϩ the change was more dramatic and resulted in the complete depletion of auxotroph cells by 72 h (Fig. 2B). As outlined below, when the tdTomato marker was instead placed in the prototroph strain, the auxotroph was similarly depleted in competition on A1 Ca 2ϩ , indicating that this marker does not interfere with this assay. Additionally, antagonism measured by CFU, which determines viable cell numbers, showed that the number of auxotroph cells was reduced by 10,000-fold when grown with prototrophs on A1 Ca 2ϩ , compared with auxotroph grown as a monoculture on the same medium (Fig. 2C). In contrast, there was no difference in the ability of this auxotroph to survive in a mixed culture incubated on A1 compared with a monoculture on the same medium (Fig. 2C). These results suggested that the auxotroph on A1 Ca 2ϩ was not merely outcompeted by the prototroph because of nutrient limitation but was also antagonized. Notably, antagonism was absent on A1 and when both strains were able to grow (on A1 Ca 2ϩ with histidine and CTT) as well as in the case when both strains were starved on TPM, which does not contain amino acids ( Fig. 2A). We note that the difference in the degree of auxotroph depletion as determined by microscopy (competitive index) compared with CFU was likely due to the presence of some nonviable auxotroph cells that were nevertheless scored under the microscope. Despite this quantitative difference, we found the competitive index to be an accurate and reproducible method for detecting antagonism and used it subsequently because of its rapid readout.
To test whether this antagonism was a specific phenotype of the hisC mutant, three other auxotroph mutants (thrC, trpD, and argE) were tested. Notably, all three mutants showed the same phenotype as the HisC auxotroph, indicating that a prototroph strain generally antagonizes amino acid auxotroph mutants when they are coincubated on the medium that is nonpermissive for auxotroph growth (Fig. 2D). For simplicity, we continued to use the HisC auxotroph in further experiments.
Antagonism depends on motility but not OME. As described above, we found that antagonism on A1 required Ca 2ϩ and asked what role this ion might play. Considering that the only phenotypic difference observed between wild type (WT) incubated with and that incubated without Ca 2ϩ was a reduction in motility in the absence of Ca 2ϩ (Fig. 1A), we tested whether motility facilitates antagonism. Here, two different nonmotile prototroph mutants, mglBA and pilQ gltC mutants, were each mixed with the HisC auxotroph on A1 Ca 2ϩ . In both cases, antagonism was abolished (Fig. 3),  Table 1.
suggesting that the prototroph requires motility to antagonize the auxotroph. Furthermore, we found the same outcome when both the prototroph and auxotroph strains were genetically nonmotile and mixed on A1 Ca 2ϩ (data not shown), demonstrating that motility plays a critical role in antagonism.
In our previous studies, we showed that OME can lead to sibling antagonism that is mediated by the exchange of toxins between cells (11,12). To address the possibility that OME is involved in prototroph-auxotroph antagonism, the HisC auxotroph was mixed with a ΔtraA prototroph that is defective in OME. This ΔtraA mutant showed  T6SS-Mediated Sibling Conflict in a Social Bacterium ® comparable antagonism as the WT control ( Fig. 3), indicating that the antagonism occurs independently of OME.
Killing is T6SS dependent. Considering the magnitude of auxotroph depletion in mixed cultures, we reasoned that antagonism is the consequence of a specific killing mechanism. To investigate this, we conducted competition experiments using prototrophs that contained mutations in predicted or known toxin systems. Included in this analysis were T6SS mutants. M. xanthus has 13 core T6SS genes within one gene cluster ( Fig. 4A) (25) and an additional orphan vgrG gene, vgrG2, which shares 43% amino acid sequence identity with VgrG1. Generally, VgrG proteins are structural components of the T6SS located at the tip of the Hcp tube and are essential for effector delivery; alternatively, VgrG proteins can be fused directly to an effector domain to form specialized effectors (26). Whereas other putative toxin system mutants retained the antagonistic phenotype (data not shown), all T6SS mutants, with the exception of a ΔtagF mutant, were unable to antagonize the auxotroph (Fig. 4C). TagF is a conserved T6SS protein that functions as a negative posttranslational regulator of T6SS in Pseudomonas aeruginosa (27), and our results show it to be nonessential for T6SS activity in M. xanthus. This is consistent with the previous finding that the ΔtagF mutant still secretes Hcp (28), a process that is a commonly used indicator of T6SS activity (29). Additionally, our findings that both vgrG genes were important for antagonism (Fig. 4C) suggest that each vgrG gene has a specific, nonredundant function. To eliminate the possibility that the in-frame vgrG deletions have polar effects, wild-type versions of these genes were ectopically expressed in each mutant strain. In each case, the vgrG killing phenotype was restored, showing that the mutations were complemented and that the vgrG in-frame deletions were not polar. We conclude that T6SS mediates antagonism between phenotypically divergent M. xanthus strains.
The VgrG2 operon encodes an effector-immunity pair. In bacterial species with an orphan vgrG gene(s), the open reading frames (ORFs) found immediately downstream sometimes encode T6SS effector-immunity proteins (30). Considering this, we analyzed two ORFs downstream of M. xanthus vgrG2 (Fig. 4B). By sequence analysis, MXAN_5572, here called TsxE (the putative effector), is predicted to contain amidohydrolase 2 (AH2) and LysM (LM) domains, whereas the protein product of the putative immunity gene MXAN_5571 (TsxI) has no conserved features (31). LysM domains are  involved in peptidoglycan binding and are found in many cell wall-degrading enzymes, including some T6SS effectors (32), whereas AH2 belongs to a superfamily of metallodependent hydrolases that act on many different substrates (33). This vgrG2 gene cluster is found in the genomes of several members of the Myxococcus genus and in some distantly related Myxococcales species such as Nannocystis exedens (Fig. 4B). Interestingly, the N. exedens tsxE homolog contains a PAAR domain, which along with VgrG composes the tip of the Hcp tube. PAAR domains can also be fused to effector domains or can noncovalently bind T6SS effector cargo (34). To test the possible role of tsxEI in the antagonism, we first inactivated tsxE in a WT strain. Importantly, this tsxE mutant lost its ability to kill the HisC auxotroph on A1 Ca 2ϩ medium (Fig. 5A) (note that the hisC auxotroph was outcompeted because it cannot grow, but it was not antagonized; compare to Fig. 2A). When mixed with the parent strain (WT), the tsxE mutant remained resistant to killing (Fig. 5A). Complementation of the tsxE mutant with an ectopic isopropyl-␤-D-thiogalactopyranoside (IPTG)-inducible tsxEI construct restored the killing phenotype (Fig. 5A). Next, we deleted both the putative effector and immunity gene (ΔtsxEI) and conducted competition assays. As expected, this mutant also was unable to kill the auxotroph on solid A1 Ca 2ϩ (data not shown). Importantly, the parent strain killed the ΔtsxEI mutant even on rich medium (Fig. 5B). Moreover, the parent strain antagonism of the ΔtsxEI mutant was abolished by ectopic expression of tsxI (Fig. 5B). Direct microscopic examination of the ΔtsxEI mutant labeled with tdTomato and mixed with a green fluorescent protein (GFP)-labeled parent strain revealed that the mutant underwent morphological changes whereby the cells became rounded and their numbers decreased (Fig. 5C). In contrast, a ΔT6SS strain did not antagonize the ΔtsxEI mutant (Fig. 5B). Together, these results suggest that the T6SS exports toxins from WT cells into neighboring cells, killing the siblings that lack the cognate immunity protein.
It is of note that T6SS-mediated killing was reduced~100-fold when strain mixtures were placed on TPM starvation agar compared with rich medium (Fig. 5C and D). These results were consistent with our above findings that prototroph-auxotroph antagonism was blocked on TPM starvation agar ( Fig. 2A), which suggested to us that starving cells downregulate T6SS function. As the killing was not completely abolished on TPM agar, we speculate that ΔtsxEI cells were initially killed by the WT early during coincubation, before the onset of the starvation response and T6SS downregulation. Additionally, killing of ΔtsxEI cells was reduced~100-fold when the WT strains were nonmotile (Fig. 5E), again confirming our above findings that motility plays a key role in the T6SS-mediated attack. Taken together, these results indicate that TsxEI comprise a T6SS-dependent effector-immunity pair.
TsxI levels decline in starving cells. The finding that nutrient-deprived auxotrophs were killed by growing prototroph siblings suggested that TsxI function decreases in the former population. One explanation for this could be that cellular levels of TsxI decrease in response to starvation, thus rendering these cells susceptible to attack by siblings that retain TsxEI. To examine this, we epitope tagged the TsxI immunity protein to generate a TsxI-FLAG protein produced from its native promoter. The TsxI-FLAG fusion was functional as determined by competition assays with the parent strain (Fig. 6A). Next, we compared levels of TsxI in auxotroph and prototroph strain backgrounds when cells were plated separately on A1 Ca 2ϩ agar (Fig. 6B). Cells were harvested at various time points, and TsxI-FLAG levels were monitored. In addition to TsxI-FLAG, we also monitored the levels of Hcp, a core component of T6SS (29). As a control, the levels of the major pilin PilA were also assessed. For Western blot analysis, samples were normalized by a Bradford assay to ensure that equal amounts of total protein were loaded in each lane. In the first 6 h after plating on A1 Ca 2ϩ medium, there was no obvious difference in TsxI or Hcp levels between the auxotroph and prototroph strains (Fig. 6B). However, between 36 and 72 h the relative amount of TsxI-FLAG and Hcp was much higher in the prototroph than in the auxotroph. As found previously, the relative levels of the PilA control protein increased in the starving cells over time compared with the growing prototroph cells (35). In contrast, the relative levels of TsxI and Hcp increased when the auxotroph was plated on rich CTT medium, compared with when those cells were plated on A1 Ca 2ϩ (Fig. 6C). Importantly, adding histidine to A1 Ca 2ϩ restored the levels of TsxI-FLAG and Hcp in the HisC auxotroph ( Fig. 6D; compare with Fig. 6B). Together, these results suggest that the levels of TsxI-FLAG and Hcp are elevated in growing cells compared with starving cells, in which they decrease over time. Additionally, TsxI and Hcp levels in growing cells increased over time compared to a 6-h time point (Fig. 6B). Because prolonged growth on a surface could alter the growth phase of some cells within a colony (15) and hence alter the relative Unlabeled tsxE mutant (DW2650) was mixed with either the WT tdTomato or hisC tdTomato strain. The tsxE mutation was complemented when the gene was ectopically expressed from an inducible promoter (DW2659; tsxE P IPTG -tsxEI). Strains were plated on A1 Ca 2ϩ medium, and the competitive index was determined at 48 h. (B) WT tdTomato was mixed with unlabeled ΔtsxEI mutant (DW2655) or this mutant complemented with an inducible tsxI gene (DW2667; ΔtsxEI P IPTG -tsxI), and the ΔT6SS tdTomato strain (DW2671) was mixed with a ΔtsxEI strain. Strains were plated on CTT, and the competitive index was determined at 48 h. (C) WT VipA-GFP (SA4137) and ΔtsxEI tdTomato (TdT) strains were mixed at a 1:1 ratio and plated on the indicated 1% agarose pads. As a control, WT VipA-GFP was mixed with a WT tdTomato strain. Micrographs were taken at 20 h using a 60ϫ objective lens, except for those in the panels second from the top, which were taken using a 100ϫ objective lens, and the blue box area is enlarged at the bottom row to show rounding of ΔtsxEI cells. Black bars, 10 m; white bar, 5 m. (D) WT tdTomato was mixed with the ΔtsxEI mutant and plated on the indicated media, and the competitive index was determined at 72 h. (E) Unlabeled WT or indicated motility mutants (same as in Fig. 3) were mixed with the ΔtsxEI tdTomato strain and placed on CTT, and the competitive index was determined at 48 h. All strains were mixed at a 1:1 ratio, and assays were done in triplicate; error bars indicate the SD.

®
TsxI and Hcp levels, we tested their relative levels between exponential-and stationaryphase growth in liquid CTT and found no difference (Fig. 6E). This result suggests that growth on a solid surface specifically induces T6SS expression (compare Fig. 6B and E) and is consistent with findings from P. aeruginosa (27).
To test if the decreased levels of TsxI-FLAG and Hcp in starved cells are due to these proteins being degraded when growth is blocked, we incubated a tetracycline-sensitive HisC auxotroph on rich medium supplemented with tetracycline, a bacteriostatic antibiotic that inhibits translation. In the presence of a growth-inhibitory concentration of tetracycline, the relative levels of TsxI-FLAG and Hcp decreased, compared to the relative level of PilA (Fig. 6D, left lane), thus recapitulating the results seen in starving cells. These observations imply that under conditions during which protein synthesis is hindered-by amino acid starvation or by blocking translation with an inhibitor-TsxI-FLAG and Hcp proteins are relatively unstable and are degraded, whereas other proteins, such as PilA, are stable, and thus, their relative cellular levels increase. These observations also explain why starving auxotrophic cells are susceptible to T6SS attack when mixed with growing prototrophic cells.

DISCUSSION
Physiological heterogeneity is generally regarded as a beneficial property of populations, because their overall fitness can increase through a "bet hedging" strategy, whereby phenotypically distinct subpopulations are optimized to quickly adapt to particular changes in the environment to which other subpopulations cannot efficiently respond (21,36,37). For example, bacterial populations can contain a subpopulation of dormant persister cells that are able to resist environmental insults, such as exposure to antibiotics, and then those cells can repopulate the biofilm when the insult is removed (20,38). However, in cases where population fitness depends on cooperative and synchronized cell behaviors, such as for social or aggregative multicellular organisms, population heterogeneity represents a challenge that may hinder the ability of a The WT strain was grown in liquid rich culture medium, and cells were harvested during the exponential phase (Exp) and stationary phase (Stat). All blots were sequentially probed with antibodies to FLAG (TsxI), Hcp, and PilA. Experiments in panels B and C were done in triplicate, the experiment in panel E was done in duplicate, and representative data are shown.
T6SS-Mediated Sibling Conflict in a Social Bacterium ® population of cells to coordinate behaviors (39). To help unify populations for multicellular behaviors, organisms have evolved mechanisms to synchronize their actions. One example is quorum sensing, during which individuals within a population produce and respond to signals made by the population to allow a uniform or coordinated response.
In this study, we show that another form of heterogeneity, involving growing cells mixed with nongrowing siblings, results in antagonism whereby the more-fit growing cells kill their weaker nongrowing siblings. As a consequence, the population moves toward being homogeneous, i.e., all cells are growing, and physiological strife is removed. Since physiological heterogeneity is high within biofilms (15), we hypothesize that the unification of cells' physiology facilitates coordinated multicellular behaviors in M. xanthus.
Specifically, we showed that physiological differences in growth rates were policed by T6SS-mediated killing. We further identified an effector-immunity pair (TsxEI) that mediates this antagonism. A ΔtsxEI mutant is itself killed when mixed with the parent strain on rich medium, indicating that T6SS function is constitutively active in growing cells. Immunoblot analysis revealed markedly decreased levels of the T6SS protein Hcp as well as of TsxI in starving cells compared with nonstarving growing cells. Based on these findings, we suggest that M. xanthus responds to nutrient limitation by decreasing T6SS and Tsx(E)I levels. In response to amino acid starvation, M. xanthus, like many bacterial species, induces a stringent response that allows cells to adapt to nutrient depletion (40)(41)(42). During the stringent response, cells cannibalize their own proteins to use as an energy source and to provide the building blocks to synthesize an altered proteome and adapt to nutrient limitation. We suggest that TsxI is degraded, along with many other host proteins, during the stringent response, which in turn renders those cells susceptible to attack by TsxE delivered by T6SS from cells that are growing (Fig. 7). In contrast, when the entire population starves, there is a uniform reduction in TsxI levels and T6SS activity, and thus, starving cells are not killed. This scenario provides M. xanthus with a mechanism to monitor the growth fitness of adjacent cells; if the population is heterogeneous with respect to growth rates, then less-fit cells are eliminated. Additionally, killing and lysis of siblings may also release nutrients to help those cells that are more fit. Another potential role of this mechanism is to help cells navigate the crucial decision between growth and development under conditions in which nutrients are becoming limiting. Under natural conditions, not all cells may experience nutrient depletion at the same time or to the same extent and will thus respond differently. Cells that, in response to nutrient depletion, enter development prematurely will, as a result, lower their TsxI levels, rendering them vulnerable to T6SS attack by their neighbors. Sibling killing and cannibalism can serve as a way to delay the onset of development as an energetically costly and committed process. Similarly, in Bacillus subtilis, Spo0A, a master sporulation regulator, triggers differentiation of matrixproducing cells in biofilms and initiates a cannibalism pathway, mediated by two toxins, Skf and Sdp. Non-matrix-producing cells remain susceptible to intoxication and  (43). This cannibalism prolongs vegetative growth and thus delays the onset of sporulation. In the case of myxobacteria, once the perception of starvation is established by the majority of cells, the population can jointly initiate development.
Interestingly, fruiting body development in M. xanthus involves death and lysis of more than 50% of the population (44,45). Programmed cell death has been suggested as the mechanism, although this has not yet been shown and the molecular details are unknown (44,46). Our data suggest that when cell populations are uniformly starved under laboratory conditions, T6SS does not play a major role in competition because its activity is diminished during development. However, in the wild, environmental and physiological heterogeneity is likely higher, and thus, T6SS could play a role in sibling competition. In separate work, we also showed that the exchange of polymorphic toxins by OME leads to sibling death (11,12). Future studies are needed to understand if population heterogeneity plays a role in developmental lysis and to determine what molecular mechanism(s) controls this process.
Our experiments demonstrate that killing requires two VgrG proteins with nonredundant functions, where, hypothetically, one VgrG acts as a structural component of the delivery tube and the other acts as a specific adaptor for TsxE delivery. In this scheme, VgrG1 likely serves as a structural component, as its gene is part of the core T6SS gene cluster, whereas vgrG2 is adjacent to tsxEI. Interestingly, other myxobacterial genomes contain many vgrG genes (TIGR03361); for example, the Minicystis rosea DSM 24000 genome contains 61 paralogs (47). Presumably, a high number of vgrG genes allows those organisms to deploy many different toxins by the T6SS. We also found that efficient T6SS-dependent killing in M. xanthus requires gliding motility. Because the T6SS apparatus is assembled and is dynamic in nonmotile M. xanthus cells (data not shown), motility may facilitate the frequency and precise deployment of T6SS toxins to target cells. In M. xanthus, cell motility is also required for other cell-cell contactdependent processes, such as OME and C-factor signaling during development (23,24). Since a role for cell motility has not been described for T6SS function in other species, the connection between these processes may be restricted to myxobacteria, which again have many complex social behaviors.
Our work raises a number of questions for future studies. For example, how are T6SS and TsxEI regulated in response to starvation? As the levels of TsxI and Hcp decrease relative to other cellular proteins when de novo protein synthesis is blocked by tetracycline, protein turnover likely plays a key role, possibly along with transcriptional control. Because amino acid starvation triggers antagonism, the stringent response could play a role; however, it appears not to be the sole determinant, because we found that a relA hisC double mutant was still antagonized on minimal medium (data not shown). Another question concerns the mechanism of action of TsxE. Sequence analysis suggests that TsxE contains a LysM peptidoglycan-binding domain and an amidohydrolase domain, indicating the cell wall as the target. If this is correct, then TsxI would likely localize in the periplasm to neutralize TsxE. The tsxEI genes are conserved among several closely related Myxococcus species and occasionally also in more distant species such as N. exedens. As genes that encode T6SS effectors are not always linked to vgrG genes and given the predatory and complex social nature of myxobacteria, there might be other T6SS-dependent effector-immunity pairs encoded by the M. xanthus DK1622 genome. Additionally, although TsxEI mediate sibling killing as reported here, it is plausible that they may also serve other roles in microbial competition/predation with other species.
In conclusion, multicellularity offers many advantages to organisms, including increased size to avoid predation, the ability to exploit new environmental niches, and the development of specialized cell functions that allow complex and powerful behaviors. Although many eukaryote species have exploited these advantages, bacterial and archaeal domains have rarely done so. Although the reasons for this remain obscure, one likely hurdle for this evolutionary jump is the evolution of a system whereby individual cells interact with each other in a harmonious and cooperative manner.
These behaviors are, at least at face value, at odds with a Darwinian view that individuals (cells) are selfish and seek their own fitness gains. Myxobacteria are attractive model organisms to address this fundamental question because they have evolved a conditional form of multicellularity based on aggregation. One challenge to this strategy is physiological heterogeneity within populations, and here, we showed a case whereby cells resolve this conflict by killing less-fit individuals.
Plasmid and strain construction. To create insertion mutations,~500 bp of the gene of interest was PCR amplified and the amplicon was ligated into pCR2.1 TOPO TA or pCR2.1 TOPO XL vector (Life Technologies) and electroporated into the DH5␣ Escherichia coli strain. Constructs were verified by PCR, restriction analysis, and/or sequencing and were then electroporated into M. xanthus.
Markerless deletions were created using a two-step homologous recombination method. Briefly, DNA fragments upstream and downstream of the gene of interest were PCR amplified and cloned into the pBJ114 vector using Gibson assembly (New England Biolabs). Verified constructs were electroporated into M. xanthus; transformants were selected for Km r and subsequently counterselected for the loss of galactose resistance (plasmid excision) (49). Markerless deletions were confirmed by PCR using primers flanking the deletion site and by phenotypic analysis.
Inducible gene expression constructs were created by PCR amplification of the gene of interest and ligation of the amplicon into pMR3487 vector (50) downstream of the IPTG-inducible promoter using XbaI and KpnI restriction sites and T4 DNA ligase. Verified constructs were electroporated into M. xanthus, and transformants were selected by their ability to grow on medium supplemented with oxytetracycline and confirmed by phenotypic analysis.
A FLAG epitope tag was engineered at the C terminus of TsxI by PCR, and the amplicon was ligated into pCR2.1 TOPO TA. The construct was verified and electroporated into M. xanthus as described above. DK1622 P IPTG -tdTomato DK1622 tdTomato ( Fig. 2C and D, 5A to D) 12 DW2602 DK8615 trpD::pCR TOPO XL (MXAN_6062) Tryptophan (Trp) auxotroph (Fig. 2D) This study DW2603 DK8615 hisC::pCR TOPO XL P IPTG -tdTomato Histidine (His) auxotroph, tdTomato ( Fig. 2A and B,  antagonizing strain (prototroph or toxin producer) and the target strain (auxotroph or toxin-immunity mutant). Typically, 200 to 800 cells were counted per replicate. Alternatively, changes in the strain ratio and cell morphology were observed by direct microscopy on agarose pads with one-half CTT (CTT with 0.5% Casitone) or TPM. Antagonism was also assessed by determining CFU. Here, competing strains carried different antibiotic resistance markers, and after coincubation, CFU were determined by serial dilutions and plating of strain mixtures on the appropriate selective medium.
Error bars in competition experiments indicate the standard deviation (SD) based on at least three biological replicates performed on different days.