Cheating, facilitation and cooperation regulate the effectiveness of phage‐encoded exotoxins as antipredator molecules

Abstract Temperate phage encoded Shiga toxin (Stx) kills the bacterivorous predator, Tetrahymena thermophila, providing Stx+ Escherichia coli with a survival advantage over Stx− cells. Although bacterial death accompanies Stx release, since bacteria grow clonally the fitness benefits of predator killing accrue to the kin of the sacrificed organism, meaning Stx‐mediated protist killing is a form of self‐destructive cooperation. We show here that the fitness benefits of Stx production are not restricted to the kin of the phage‐encoding bacteria. Instead, nearby “free loading” bacteria, irrespective of their genotype, also reap the benefit of Stx‐mediated predator killing. This finding indicates that the phage‐borne Stx exotoxin behaves as a public good. Stx is encoded by a mobile phage. We find that Stx‐encoding phage can use susceptible bacteria in the population as surrogates to enhance toxin and phage production. Moreover, our findings also demonstrate that engulfment and concentration of Stx‐encoding and susceptible Stx− bacteria in the Tetrahymena phagosome enhances the transfer of Stx‐encoding temperate phage from the host to the susceptible bacteria. This transfer increases the population of cooperating bacteria within the community. Since these bacteria now encode Stx, the predation‐stimulated increase in phage transfer increases the population of toxin encoding bacteria in the environment.

We also showed previously that in the face of attack by predator, a only minor subset of bacteria lysogenized with temperate Stxencoding lambdoid phage produce and release sufficient exotoxin to substantially reduce of predation the balance of the (Arnold & Koudelka, 2014;Lainhart et al., 2009;Stolfa & Koudelka, 2012). This finding suggests that Stx acts as a public good.
In cooperation mediated by public goods, producer cells both benefit from and bear the costs of production. Since exotoxins encoded by temperate lambdoid phages are produced only during lytic growth and their release depends on phage genes that induce bacterial cell lysis (Arnold & Koudelka, 2014;Lainhart et al., 2009), bacterial death accompanies exotoxin release. Therefore, similar to other examples of public goods-mediated bacterial cooperation (Paton, 1996;Voth & Ballard, 2005), Stx-mediated protist killing is a form of altruism known as self-destructive cooperation (Ackermann et al., 2008). Since bacteria grow clonally, in the case of Stx-mediated predator killing, the benefits accrue to the kin of the altruistic organism. However, since Stx is a diffusible public good that theoretically could benefit all potential protist prey, this cooperative behavior may be susceptible to 'cheating' in the face of selection pressure. Cheaters can reap the benefits of cooperation without bearing the costs of production. In the extreme, the benefits of the public good can accrue primarily to cells other than the producer. Since many bacterial predators do not discriminate between bacterial food sources, exotoxin-mediated predator killing may enhance survival of nearby, but unrelated, 'cheater' bacteria. It is unknown whether this is the case with Stx-mediated cooperation.
Cheating creates a dilemma for a population. If the cost of production exceeds the direct benefit to the producer, and cheaters benefit, but do not pay the cost of secretion, then they are expected to outcompete cooperating producers. Thus, without a mechanism to support cooperation, the fraction of cells producing the public good will decline in frequency resulting in the loss of the public good, a condition which would decrease the fitness of the entire population, producers and cheaters alike (Nowak, 2006).
Cooperation can be maintained in the face of cheating if specific mechanisms to allow preferential utilization of public goods by cooperating cells arise in the population. For example, cooperation is favored by the physical association of cooperative partners or if there is limited diffusion of the public good away from the cooperating producers (Brown & Buckling, 2008;Kummerli, Griffin, West, Buckling, & Harrison, 2009). This observation is consistent with kin selection theory, which postulates that cooperative interactions will evolve and be maintained if the benefits preferentially accrue to organisms that carry cooperation genes (Hamilton, 1964). Transmission of cooperation genes to nonproducers can also support cooperation by increasing genetic similarity. The increased genetic-relatedness of the individuals in the population increases both the direct and indirect benefit of public goods transactions (Gardner, West, & Wild, 2011;West, Griffin, Gardner, & Diggle, 2006). Strikingly, genes encoding cooperative traits are often found associated with mobile genetic elements such as plasmids or transposons (Dimitriu et al., 2014). These mobile elements facilitate rapid horizontal gene transfer within and between bacterial lineages suggesting that they can quickly alter the cooperative social structure of a population. Although the spread of plasmid-encoded antibiotics resistance factors serves as an important and well-studied case of horizontal gene transfer (HGT) mediated stabilization of public goods cooperation (Yurtsev, Chao, Datta, Artemova, & Gore, 2013), the ability of temperate bacteriophage to serve as a conduit for the horizontally transfer of public goods genes has not been thoroughly explored. The genomes of temperate prophage are found at surprisingly high frequency within the chromosomes of these organisms.
The evolution of phage resistance occurs relatively easily. Thus, since the ultimate goal of bacteriophage is to reproduce, and in doing so, these phages kill their host, the prevalence of phage DNA, inside host chromosomes suggests that phage are tolerated because their presence provides an evolutionary advantage to the host.
While Stx release enhances the survival of the kin of Stxencoding bacteria in the presence of bacterial (Arnold & Koudelka, 2014;Lainhart et al., 2009;Stolfa & Koudelka, 2012), it is unknown whether the fitness benefits of Stx antipredator activity extend to the unrelated bacteria in the population. Also, since Stx is encoded on a mobile bacteriophage, the movement of this phage between bacteria could greatly impact the social structure of the bacterial population. That is, predation could encourage the transfer of phage to susceptible bacteria. Such transfers could lead to exploitation of these bacteria by using them to enhance public goods production and/or by creating new lysogenic cooperating bacterial. To test these ideas, we determined if phage-mediated dissemination of a public good impacts the population of cooperating producers and cheaters in the face of predation. Given the contribution of Stx and other temperate lambdoid phage-encoded exotoxins to serious human disease, this information may aid in understanding the increasing occurrence and virulence of these diseases.

| Construction of antibiotic resistant strains
MG1655 recA (CAT, TET) was made by transforming a vector pACYC182 harboring tetracycline resistance gene into MG1655 recA .2

| Preparation of Tetrahymena and bacteria
Bacteria and Tetrahymena cells were prepared as follows: Cultures of the specified bacteria (Stx + : EDL933, Stx − : EDL933ΔStx, immune strain: MG1655 λimm933 ,phage-susceptible strain: MG1655::AMP) were grown to saturation at 37°C in M9 plus 0.08% glucose supplemented with antibiotics when appropriate. The cultures were centrifuged at 8,000g for 10 min, washed twice with M9 plus 0.08% sodium citrate, and resuspended in M9 plus 0.08% sodium citrate. Tetrahymena cells were diluted fivefold from saturated liquid cultures and grown for 3 days in proteose peptone plus FeCl 2 at 30°C. The cells were centrifuged at 5000g for 5 min, washed three times with 10 mmol/L Tris-HCl (pH 7.4), and suspended in M9 plus 0.08% sodium citrate in a volume sufficient to give 10 4 cells/ml. To each washed Tetrahymena culture, 10 8 cells/ml of the indicated bacteria were added and the cocultures were maintained at 30°C.

| Tetrahymena and bacterial viability in mixed cocultures
To create artificial microcosms, we cocultured the indicated bacterial strains without or with Tetrahymena. In all cases, the total number of bacterial cells in these microcosms was 10 8 cells/ml. When present, Tetrahymena were added at 10 4 cells/ml. In microcosms in which the ratio of toxin and nontoxin producing bacteria was varied, the total number of bacteria was maintained at 10 8 cells/ ml, only the ratio of toxin producing to nontoxin producing bacteria was altered. The microcosms were maintained by shaking at 30°C. At t = 0, and 6 hr, two aliquots were removed from these microcosms. One aliquot was used to determine the change in Tetrahymena cell count. The cell count was obtained by counting the number of Lugol stained cells, visualized in a hemocytometer (Lainhart et al., 2009). Tetrahymena cells killed by exposure to Stxexpressing bacteria are not visible, presumably because they have lysed. The second aliquot was used to determine the number of bacteria by plating the dilutions of cultures on agar plates containing 100 μg/ml ampicillin and determining the number of colony forming units (CFU). Each measurement was performed in duplicate and the data were averaged. Each experiment was repeated a minimum of three times. The data presented are the average of the three (or more) replicates.
The stx2A and uidA genes were amplified as a 20 μl reaction mixture using the PCR MasterMix in Bio-Rad iQ5 real-time PCR detection system. The following two-step thermal profile was used: 5 min at 95°C, 45 repeats of 10 s at 95°C, 45 s at 60°C. Standard curves for the real-time PCR analysis were made using genomic DNA containing the regions of interest.
Cells were grown to saturation at 30°C. DNA was extracted using InstaGene matrix, stx2A , uidA, and rbfE genes were amplified using the following thermal profile: 5 min at 95°C, 30 repeats of 10 s at 95°C, 30 s at 60°C, 30 s at 72°C, and 5 min elongation at 72°C. The same primers were used to amplify stx2A, and uidA as mentioned above whereas the following primers were used to amplify rbfE:

| Forward primer 5′-CTACAGGTGAAGGTGG AATGGT-3′
Reverse primer 5′-GTAGCCTATAACGTCATGCCAAT-3′ (Desmarchelier et al., 1998). The PCR products were separated on a 5% native PAGE gel. To measure the frequency of lysogenization, the microcosms containing EDL933, MG1655 recA (CAT, TET) , and Tetrahymena were maintained at 30°C for 24 hr. At t = 12, and 24 hr, Tetrahymena cells were separated from bacteria by differential centrifugation at 1,000g for 2 min, washed 5× with10 mmol/L Tris-HCl (pH 7.4), lysed (Bolivar & Guiard-Maffia, 1989), and internalized bacteria were diluted 1:200 fold in LB supplemented with 50 μg/ml chloramphenicol and 25 μg/ml tetracycline. Cells were grown to saturation at 30°C, DNA extracted, and stx2a and uidA genes were amplified as a 20 μl reaction mixture using the PCR MasterMix in Bio-Rad iQ5 real-time PCR detection system. Control experiments established that the amplification efficiency of the stx2A and uidA genes are identical under our conditions. The stx2A:uidA ratio represents the fraction of MG1655 recA (CAT TET) that had stx2 + prophage. The number of lysogens in the population of MG1655 recA (CAT, TET) was calculated by multiplying the ratio of stx2A:uidA with total number of bacteria.

| Statistical methods
Error bars presented in the figures represent standard deviations of the means of multiple (≥3) replicate experiments. t test was used to test the significance of differences between the mean of the measured initial amounts and the amounts of bacteria and/or Tetrahymena after treatment in each experiment.

| Stx as an antipredator public good
To verify that Stx can act as a public good, we examined the effect of EDL933, a stx + phage-bearing Shiga toxin encoding E. coli (STEC) bacterial strain or its stx − derivative, EDL933Δstx, on the survivorship of MG1655 λimm933 , an stx − E. coli strain that cannot be infected by stx2 + phage, in cocultures without or with predatory Tetrahymena. The EDL933 and MG1655 λimm933 strains were distinguished by marking them with different antibiotic resistance genes.
However, in cocultures containing Tetrahymena, the presence of the Stx-encoding EDL933 strain increased the survival of the nontoxinencoding MG1655 λimm933 strain 1.5-fold more than cocultures containing EDL933Δstx (Figure 1a). These observations indicate that the benefits of Shiga toxin's antipredator activity accrue to both the toxin producing and nontoxin-producing cells in the bacterial population. These findings are consistent with the idea that Stx functions as a public good. These results also indicate that nontoxin-producing cells in the bacterial population can cheat, that is, take advantage of the fitness benefits of Stx, without paying the fitness costs of its production.
To verify that the observed survival enhancement of MG1655 λimm933 seen in Figure 1a is a consequence of the antipredator activity of Stx released by EDL933, we examined the growth of Tetrahymena in these cocultures. We found that, as com-

| Role of Stx phage infection in regulating bacterial survival during predation
To probe the effect of phage infection on bacteria survival and predator killing, we repeated the above experiment, but replaced the

| Impact of phage-susceptible and phageimmune bacterial strains on the predation resistance of bacterial populations
The placement of Stx on a mobile phage provides a method for both increasing production of this public good and amplifying the allele that encodes this product. That is, the presence of phage-susceptible MG1655 λimm933 cannot be infected by Stx-encoding phage released from EDL933. Hence, the findings in Figure 3a,b suggest that in cocultures containing a smaller amount of EDL933 and a larger amount of MG1655 λimm933W , the amount of toxin produced is apparently too low to affect predator viability. On the contrary, in cocultures containing a smaller amount of EDL933 and a larger amount of 'susceptible' E. coli MG1655, the predator viability decreased, suggesting that the stx + phage released by EDL933 is using phage susceptible E. coli as a surrogate to amplify the amount of exotoxin produced by the population.   In addition to showing that the ability of exotoxins to kill bacterivorous predators is exploited by unrelated bacterial 'cheaters', the results in Figure 3 indicate that MG1655 can be used as a surrogate, to amplify the amount of exotoxin produced by the population and facilitate enhanced population survival by killing predatory

Consistent with this idea, in cocultures containing
Tetrahymena. We wished to confirm that the enhanced killing is a consequence of lytic growth of Stx-encoding phage released from EDL933 in MG1655.

| Effect of predation Stx phage production
The Stx-encoding phage in EDL933 carries only one copy of the stx2A gene in its genome (Plunkett, Rose, Durfee, & Blattner, 1999). Likewise, the E. coli strains also contain only one copy of β-glucuronidase gene uidA (Sanjar et al., 2014).

| Effect of predation on lysogen formation
In addition to functioning as surrogates to increase exotoxin pro- In the MG1655recA/MG1655recA 933W population, each stx2A gene corresponds to one prophage (Plunkett et al., 1999) and each bacteria contains one uidA gene (Hayashi et al., 2006). Thus, we calculated the efficiency of lysogen formation by using qPCR to measure the ratio of stx2A :uidA genes in the population after coculturing conditions (Aijaz & Koudelka, 2017). Hence, the number of phage released, and thus their titer, is quite low. In contrast, confinement of EDL933 to the Tetrahymena phagosome significantly increases its spontaneous induction frequency (Aijaz & Koudelka, 2017). Hence, our inability to 'see' these new lysogens forming in cocultures containing MG1655recA and EDL933 in the absence of Tetrahymena is seemingly due, in part, to a low phage titer that leads the formation of a small, and apparently undetectable number of new lyosgens.
In a bacterial population subjected to predation, only a small fraction of the Stx phage-bearing bacteria are induced and produce toxin, a sacrifice that reduces predation and increases the survival of the population (Arnold & Koudelka, 2014;Stolfa & Koudelka, 2012).
Thus, although Stx is an exotoxin, it improves the competitive fitness of bacterial populations that carry Stx-encoding phage by a novel mechanism, that is, providing the bacterial host a means to defend against the predation by single-celled protozoan bacterivorous predators. These observations argue that carriage of Stx-encoding phage embodies a form of self-destructive cooperation (Ackermann et al., 2008), with the fitness cost of this behavior being "paid" by death of the individual cell via and the fitness benefits accruing to the kin of sacrificed bacteria. This conclusion indicates that the phage-borne Stx exotoxin behaves as a public good.
Our results indicate that the fitness benefits of Stx production are not restricted to the kin of the phage-encoding bacteria (Figures 1,2 while not investing in its production. Cheating may help explain why not all bacteria in a given population encode this exotoxin. Consistent with this idea, we found that survival of nontoxin encoding, cheating bacteria varies with the proportion of cheaters within the population (Figure 3). Thus, above a certain threshold level of Stx-encoding bacteria, enough Stx is produced to reduce predation sufficiently so that the selfish nonproducers thrive in the presence of predators. More importantly, the increased survival of non-Stx producers in cocultures with Stx producers is only seen in the presence of a predator that is sensitive to killing by Stx. This observation confirms that the antipredator activity of Stx serves as the public good.
Alternatively, if the population contains a below-threshold level of Stx-encoding bacteria, the selfish nonproducers do not benefit from the presence of Stx-encoding bacteria (Figure 3). The negative dependence of cheater survival on cheater population size leads a dynamic equilibrium distribution of selfish and cooperating individuals in a population (Powers, 2011). Exotoxin-encoding bacteria and phage are ubiquitously distributed in the environment, but their occurrence is sporadic or episodic and their environmental persistence varies (Casas et al., 2006;O'Brien et al., 1984). Since cheating and cooperation facilitate the coexistence of cooperative (Stx + ) and selfish (immune, Stx − ) individuals in the face of predation, these processes may, therefore, help explain the environmental occurrence and persistence of Stx-encoding bacteria and phage.
The cheating by nontoxin-encoding bacteria threatens the persistence of the cooperative behavior engendered by carriage of Stxencoding genes. Social evolution theory suggests that in the face of cheating, cooperation can be maintained when its benefits are directed preferentially to organisms carrying cooperation genes (Charnov, 1977;Hamilton, 1963Hamilton, , 1964Hamilton, , 1970Smith, 1964). The problem of how to limit benefits to closely related organisms is particularly acute in cooperation that is mediated by a diffusible public good such as Stx. It is well established that the evolution and maintenance of cooperation is strongly influenced by population structure.
Physical segregation of producers from nonproducers increases the likelihood that a diffusible public good secreted by cooperating cells will be utilized by other cooperators. However, organisms in our liquid cocultures used here are well-mixed, a condition that mimics the conditions in which the target predator (Tetrahymena), producers and natural cheaters naturally interact.
How then might the cooperative behavior that is conferred by Stx-encoding phage be reinforced in bacterial populations? Our data suggest two mechanisms. First, since Stx is produced only during phage lytic growth, bacterial death accompanies Stx release. Thus, Stx production is a form of self-destructive cooperation, where a subset of individuals in a population die in order to help others (Ackermann et al., 2008). Since infectious phage are released along with toxin, the phage can function as a lethal anticompetitor tool against susceptible bacteria that do not carry them (Brown, Le Chat, De Paepe, & Taddei, 2006;Joo et al., 2006). The amplification makes phage carriers capable of efficiently invading well-mixed populations, even when initially rare (Brown et al., 2006). In the specific case of Stx-encoding bacteria, the phage susceptible bacteria are also apparently used as surrogates to amplify toxin production, thus further enhancing population survival (Figures 3,4), a feature that has impacts on Stx production and toxicity in animals (Goswami, Chen, Xiaoli, Eaton, & Dudley, 2015 Stx-encoding bacteria are present at low levels, the predator will more likely encounter and consume the dominant members of the community, that is, the bacteria that do not encode Stx. Whereas when the Stx producers are present in higher density, the predator consumes them, but the predator is, in turn, killed by the Stx produced. Genes encoding cooperative traits are often found in association with mobile genetic elements such as plasmids or transposons (Dimitriu et al., 2014). These mobile elements facilitate rapid horizontal gene transfer within and between bacterial lineages suggesting that they can quickly alter the cooperative social structure of a population. Horizontal gene transfer via transduction has traditionally been considered a rare event. However, recent studies have reported that transduction might occur at higher frequencies than previously thought (Evans et al., 2010;Kenzaka, Tani, & Nasu, 2010).
Protozoan predation increases the frequency of transfer, persistence, and spread of plasmids among bacteria (Cairns, Jalasvuori, Ojala, Brockhurst, & Hiltunen, 2016). Temperate bacteriophages have two lifecycles, lytic and lysogenic, so they have the potential to serve as a vessel in transferring genes encoding public goods across bacterial species. Our results suggest that protozoan predators, in particular those who feed indiscriminately, strongly enhance transfer of Stx-encoding temperate phage from one susceptible host to another. Predation may enhance phage transfer by two nonmutually exclusive mechanisms. First, predation may increase the number of phage. We showed previously that consumption of bacteria by a protozoan predator stimulates prophage induction (Aijaz & Koudelka, 2017;Arnold & Koudelka, 2014;Lainhart et al., 2009;Stolfa & Koudelka, 2012).
Second, predation may increase contacts between phage and recipient bacteria. The filter-like feeding behavior of Tetrahymena can result in the confinement of phage, its donor and recipient strains within a single phagosome, thereby enhancing the rate of phage infection in new hosts. The increased concentration of phage within the organelle would increase the multiplicity of infection, a condition that favors lysogen formation over lytic phage growth. Therefore, we believe that phages have contributed in the evolution of cooperative behavior of microorganisms.
We suggest that the constant struggle between predator and prey, cooperators and cheaters, shape the evolution of phages and if these phages carry toxins which can harm humans, as is the case with Stx2 encoding phage, then they shape the evolution of new human pathogens.

ACK N OWLED G EM ENTS
The work in this manuscript was supported by a grant from the National Science Foundation (MCB-0956454) to GBK. We thank the members of the Koudelka lab for critical reading of the manuscript. We certify that there is no conflict of interest with any financial organization regarding the material discussed in the manuscript.

CO N FLI C T O F I NTE R E S T
None declared.