In the Light of Evolution V: Cooperation (January 7th)

John C. Avise [1, 2] started off the colloquium by giving a very brief introduction to the In the Light of Evolution series, highlighting much of the history I discussed in my preview.

Nonacs began his talk by giving it an additional subtitle: “peaks of runaway niceness”. His main focus was to explain why some social systems seem to be more cooperative than might be explained by standard Hamiltonian inclusive fitness theory. Much of his own work has been in eusocial insects, so Nonacs started out by discussing monogamy as a preadaptation for cooperative breeding and/or eusociality. Because monogamous mothers give birth to offspring with maximum relatedness to each other, the logic is that monogamy should favor delayed dispersal by offspring, who remain as helpers. This “monogamy hypothesis” has not been borne out by the evidence, which appears to be a failure of inclusive fitness theory. Wanting to understand why observed mating patterns in cooperative breeders do not fit the monogamy hypothesis, Nonacs built a model designed to explore when monogamous cooperative breeding would be favored.

Nonacs’ model was pretty simple: it looked at the role that offspring who remain as ‘helpers at the nest’ play in survival of future offspring and predicted based on overall population genetics whether the model species would evolve to be solitary breeders, social breeders with monogamous parents, or social breeders with polygamous parents. Not surprisingly, the higher the survival benefits provided by helper offspring to their younger siblings, the more stable social breeding was predicted to be. When the survival benefits provided by offspring who delay dispersal was low, species were predicted to breed without intergenerational cooperation. But when the survival benefit is moderate, cooperative breeding is favored so long as all offspring are full siblings (in other words their mother is monogamous). When the survival benefits conferred by helper offspring are high, social breeding can be the favored strategy even when the mother mates with multiple males. The surprising result of Nonacs’ simulation is not so much that delayed dispersal can be favored by high relatedness and high benefits of altruistic behavior, but that the thresholds for the emergence of social altruism are far below that predicted by Hamilton’s Rule. In other words, Nonacs’ simulation predicts a greater degree of cooperative behavior than Hamilton’s Rule. His explanation for this was that Hamilton’s Rule is too simple and lacks mechanism, making it inappropriate to apply to many ecological scenarios.

Nonacs made sure to check the robustness of his model: the qualitative results were not altered by shifting from haploid to diploid males or by changing the helper trait from being dominant to being recessive. Faced with the prediction of more cooperation than expected by inclusive fitness benefits, he conducted an in silico experiment to see what would happen if indirect fitness effects were removed by switching helpers from one nest to another. Not surprisingly, this separation of helper offspring from their siblings destabilized the evolution of cooperative breeding, leaving parents to raise offspring alone. So while Hamilton’s Rule may not make the right predictions in this system, relatedness does matter. Seemingly this “saves” inclusive fitness theory, although I should point out that Nonacs’ reshuffling experiment simply served to make the population well-mixed and unstructured, conditions known to make the evolution of cooperation less probable.

To further understand the results of his simulation, Nonacs adopted a broader perspective and asked the question “when does Hamilton’s Rule make valid predictions?”. He compared the predictions of traditional inclusive fitness theory with observed patterns of reproductive skew, sex allocation governed by mothers and daughters, caste-rearing bias, and worker policing in eusocial insects. He concluded that for two of these phenomena (sex allocation and worker policing), the predictions of Hamilton’s Rule were well-supported by empirical observations. But for reproductive skew and caste-rearing bias, empirical evidence fails to support the Hamiltonian predictions: there tends to be more cooperative behavior than predicted by traditional inclusive fitness theory. Nonacs wondered why, and focused on mechanisms that could augment cooperation above levels allowed by kinship alone.

Nonacs focused on the phenomenon of the greenbeard, an altruistic individual whose identity as a cooperator can be confirmed by an externally-assessable trait. He suggested that having a greenbeard trait allows for the form of kin recognition most critical to the evolution of cooperation: the ability to recognize other individuals who are also reliable cooperators. For Nonacs, kin recognition is really just a collection of greenbeard traits, and if there are no greenbeard traits there can be no recognition and thus no selection. He explained “overly nice” traits as being fostered by the presence of greenbeard traits (especially multilocus greenbeards, which are less vulnerable to cheating), which lead to runaway social selection for altruism and higher-than-expected levels of cooperation. He argued that sex allocation and worker policing are not subject to greenbeard signaling and therefore are constrained to the levels of cooperation predicted by Hamilton’s Rule, whereas reproductive skew and caste-rearing bias could be augmented by the presence of greenbeard traits.

I really enjoyed Nonacs’ presentation because it started off the meeting on a very constructive tone: rather than adopting any particular orthodoxy, Nonacs showed how careful thinking about particular social phenomena can lead to the appropriate application of theory. By demonstrating that Hamilton’s Rule sometimes works and sometimes does not, Nonacs challenged theorists to think carefully about the assumptions that go into their models rather than assuming a particular construction applies to all systems everywhere. His emphasis on how spatial structuring is key to fostering cooperative behavior foreshadowed much of what was to come throughout the two-day colloquium.

Rubenstein’s presentation was an appropriate follow-up to Peter Nonacs’ talk, as he extended the realm into which collective breeding and reproductive skew is unaccounted for by traditional inclusive fitness theory. Rubenstein works with a variety of taxonomic groups, but his talk was focused on birds, whose cooperative breeding behavior he has been studying for a decade. A great many species breed cooperatively, with offspring remaining in the natal nest to help raise successive generations rather than dispersing to found their own nests. Some species are plural cooperative breeders, meaning that there are multiple breeders in the colony. Others are singular breeders, where a single pair of individuals produce all the offspring cared for by helper individuals.

Rubenstein is interested in the role that environmental gradients play in fostering cooperation. Like Nonacs, he contended that traditional inclusive fitness models fail to explain why cooperative breeding exists in some species but not others. Instead, he turned to the role of spatial and temporal heterogeneity in making cooperation the optimal strategy for young adult individuals. Spatially, the clumping of resources may allow some individuals or groups to dominate particular resource patches, reducing the prospects of successfully establishing a mating territory for young male birds. This may favor philopatry and lead to the evolution of cooperative breeding. Temporally, variation in resources (in other words seasonality) creates periods of time in which resources are either exceptionally abundant or exceptionally sparse. Presumably larger groups would be able to better survive the famine periods (by collectively increasing their range of foraging and food location) and take advantage of periods of abundance (by being able to exert greater overall foraging effort), especially when raising dependent offspring. Although Rubenstein did not make this connection, this is essentially a version of the Resource Dispersion Hypothesis, which has been the focus of modeling work I have been working on with Jennifer Verdolin and graduate student collaborators.

To understand whether variation in resource availability drives cooperative breeding, Rubenstein turned to African starlings, one of the major bird groups he studies. Different starling species display different reproductive strategies, with some displaying no cooperative breeding while others show either singular or plural breeding strategies. Although absolute measures of resource heterogeneity are not available for each of the forty-five starling species he considered, Rubenstein was able to show that all cooperative breeders live in the savanna — a biome with strong seasonal variation — whereas those species without cooperative breeding predominantly live in forest biomes. Some species without cooperation also live in the savanna, suggesting that the correlation between biome and breeding system is not perfect. This imprecise match may also have to do with the rather broad categories of “savanna” and “forest”: it would be interesting to see if greater predictive power could be gained by actually measuring seasonality — or better yet seasonality of critical resources — in each of the species’ habitats. Nonetheless, Rubenstein shows strong evidence in favor of the hypothesis that variable environments promote cooperation.

To see if this trend extends beyond the starlings, Rubenstein looked at a survey of 9,310 avian species, an impressive collection encompassing 95% of the world’s birds. A geographical analysis indicates that Africa and Australia are the ‘global hotspots’ of cooperative breeding, a trend that could be due to either phylogenetic or environmental factors. While Rubenstein did find some phylogenetic bias, environment turned out to be of critical importance (Africa and Australia are home to biomes with some of the greatest seasonal variation). Interestingly, the component of the environment predicting cooperative breeding was different for passerine and non-passerine species. For passerines species, uncertainty in precipitation patterns promotes cooperative breeding, whereas in non-passerines it is variation in temperature that best predicts which species will be social breeders. My understanding of the difference between these two major groups of birds is too weak to forward an explanation for this difference, but it is interesting.

Rubenstein finished his talk by zooming in to a very small-scale example of how environmental heterogeneity can affect breeding strategies. In Superb Starlings both spatial and temporal variation can change the degree to which cooperative breeding is employed as a strategy. Nests tend to contain a higher fraction of males (an indication of cooperative effort, as males are the sex that remains as helpers at the nest) during periods when rainfall is low. There is also a demonstrable influence of spatial variation, an effect that can be demonstrated thanks to the historical presence of Maasai bomas, enclosed pastures which leave nutrient-rich circles of savanna even decades after they have been abandoned. What is interesting about the cooperative breeding behavior of Superb Starlings is that it is plastic: individuals modify their degree of cooperation based on environmental cues. This suggests that the large-scale trend towards cooperative breeding in response to environmental variation is not simply the result of long-term selection for survival under particular environmental conditions: individuals of these species may have evolved contingent behavioral strategies (‘should I stay or should I go’ rules) dictating their overall breeding strategy.

Like Nonacs’ talk, Rubenstein’s presentation did not so much exclude relatedness as an important factor for understanding but tried to put relatedness in its proper context: sometimes relatedness patterns can drive cooperation, but it is environmental conditions that dictate whether relatedness has the power to foster cooperation. This seems obvious from Hamilton’s Rule, as the costs and benefits that go into the inequality have to be in some part dictated by the abiotic and biotic environment. Unlike Nonacs, Rubenstein did not address whether Hamilton’s Rule can be used to predict cooperative breeding in birds. This does not surprise me, as it is probably near impossible to get reliable estimates of the costs and benefits associated with reproductive helping behaviors in so many bird species living in such varying environments. Does this mean that Hamilton’s Rule is instructive but not practical?

Microbes are of particular ecological importance, as they perform critical roles in all ecosystems. One of these roles is as a mutualist partner to other taxonomic groups. Because of size asymmetries, bacteria are often the ‘symbionts’ housed in their multicellular ‘hosts’, which sets up a conflict: if I am a bacterial cell, should I cooperate and provide net benefits to my host (the mutualist strategy) or should I defect and try to exploit my host’s resources even if this causes net harm (the parasite strategy)? Inter-species mutualisms are particularly interesting because from a conventional view they cannot involve kin selection: two species by definition do not share a gene pool, and therefore cannot gain inclusive fitness benefits from their interactions (more on that below).

Sachs looked at this question and others from a phylogenetic perspective. He sought to assess the patterns of host association in various bacterial lineages and to understand why some bacterial are free-living in the environment and others are host-associated. To get at these questions, Sachs (like Rubenstein before him) looked at several levels of taxonomic resolution. First he considered the whole bacterial domain tree and then focused on fourteen well-studied mutualistic systems. In his analysis he looked for patterns of host association, including transitions from environment to host as either a parasite or mutualist and transitions within hosts from parasite to mutualist or mutualist to parasite. It has been hypothesized that the transition from parasite to mutualist is more likely than a transition from environment to mutualistic association, because a host that can ‘capture’ its parasite (preventing it from spreading horizontally) will effectively domesticate it. Although Sachs did not really get into how this works mechanistically, it seems to me that this coevolution of reduced symbiont harm has to involve differential survival of bacterial populations with varying degrees of harmful traits (in other words a form of group selection empowered by obligate vertical symbiont transmission).

Sachs’ phylogenetic analysis showed that most major host-associated bacterial lineages are the product of a transition from free-living in the environment to parasitism. Of the forty-one total independent host-association transitions shown on his phylogeny, nine went from being environmental to mutualistic and only three transitioned from parasitism to mutualism. Further calling into the question the ‘parasite capture’ hypothesis, not all mutualistic bacteria are vertically transferred. While some bacterial symbionts (such as the mitochondria of eukaryotic cells or the Buchnera housed by aphids) are transmitted vertically, it is not possible to reconstruct whether their transition to mutualistic association was or was not preceded by a transition to vertical transmission. For many mutualist symbionts that are environmentally acquired by each subsequent generation of hosts (in other words horizontally transferred), Sachs suggested that no real theory explains their evolution. This is a fascinating gap in theory!

Transitioning away from his overall phylogenetic approach and towards his fourteen case studies, Sachs identified three models for maintaining symbiotic cooperation: byproducts, partner fidelity, and partner choice. Byproduct mutualisms are almost coincidental: one way to look at them is as self-reinforcing reciprocal commensalisms. What does that mean? I like to think of commensal relationships as ‘waste or wake’: one species produces a benefit to another at no cost to itself, usually through the production of waste or as a service left in the wake of a particular behavior. In a byproduct mutualism, it just so happens that the waste-or-wake products of each species benefit the other. There is no cost associated with this exchange for either side, which means that the potential for cheating is eliminated. Sachs identified Pseudonocardia species, which live in association with ants and produce an antibotic that can kill off a fungus that parasitizes the ants’ fungus gardens, as a good example of byproduct mutualists. While the antibiotic provides a benefit to the ants, it has no net cost to the bacteria because it benefits from the competitive advantage conferred by the secretion. Another example cited by Sachs were the Bacteroides species present in the mammalian gut: we provide the bacteria with food as a normal waste product of our digestion, and the bacteria provides us with additional nutrients as a normal waste product of its metabolism. While I see the features of these symbiont-host associations that compelled Sachs to label them as byproduct mutualisms, I wonder: is there no net reward to either host or symbiont for developing slightly-costly traits to promote the success of its partner? Do ants really only support these critical symbiotic bacteria that literally preserve their sources of livelihood (fungus farms) as a byproduct of natural behavior and physiology? Is there no reward for ants who produce at some minimal cost secretions that better foster the growth of mutualistic bacteria? Is there no selection for species of bacteria that produce at some cost more effective and/or plentiful antibiotics? It is pretty clear why Sachs does not entertain these questions, as they would require a form of group selection, and his theoretical taxonomy explaining the evolution of mutualisms completely ignores the possibility of group selection (see Sachs et al. 2004, an otherwise valuable article).

As mentioned before, partner fidelity can be maintained when transmission of the symbiont from generation to generation is vertical. Sachs discussed the problems of vertical transmission, which include both biological logistics and evolutionary dynamics. Logistically there is the question of how to transport the symbiotic bacteria in parallel with the germ line, which has been solved by a few species. In terms of evolution, there is also the question of how to suppress symbiont cheating. Capture of the symbiont by the host is to the benefit of the host and to the detriment of the symbiont: Sachs pointed out that obligate symbionts like Buchnera lose genome-wide function as a result of their capture. So there is a strong asymmetry in the incentive to cheat, as symbionts may resist capture while hosts try to reinforce capture. Eventually the two species often become mutually dependent in a way that prevents further cheating, usually through the complementary loss of function. But how two species evolve to this point of interdependence is not clear.

Perhaps most interesting was Sachs’ discussion of partner choice in symbiont-host associations. Sachs focused on two mutualisms that involve partner choice: plant-Rhizobia and squid-Vibrio. A critical feature of partner choice mechanisms is that they require spatial structuring of symbionts within the host. For plants that host Rhizobia this means creating root nodules, and for squid that host Vibrio this means the presence of “crypts” for housing the bacteria. The value of the crypts is that they allow the host to reward or sanction subpopulations of their overall symbiont population, increasing cooperation and suppressing cheating. Although this is not how Sachs portrayed it, I would view this as a host-mediated group selective process (artificial or engineered group selection?). Other host-symbiont systems which have been less well-studied — such as the mammalian gut — also spatially structure symbiont populations, opening up the possibility that host control over the size and location of symbiont colonies may be a convergent adaptation to the threat of parasitism. Not surprisingly, the innovation of symbiont spatially structuring seems to allow for horizontal transfer of the symbiont, as the host need not capture mutualistic strains if it already has a means of excluding parasitic strains. In only a few tough-to-explain cases does horizontal transfer of symbionts who live without structure in the host occur (see my explanation below).

For awhile I bought into the idea that mutualisms could be maintained by a simple logic of “I scratch your back and you scratch mine”. But given the opportunity for cheating one’s mutualist partner by over-exploiting them — especially amongst microbes that can so easily become parasitic — this idea is beginning to lose traction in my mind. Byproduct mutualisms seem like they could easily be made ‘more cooperative than expected’ by selection for individuals who take on some minor cost to better reap the benefits of their partner’s production of benefits. Such selection could lead to the partner fidelity and partner choice mechanisms discussed by Sachs, but could also be maintained by simple competitive advantages of partnerships between less exploitative symbiont-host pairs. And even partner fidelity and partner choice seem, as sole mechanisms explaining the evolution of mutualism, like inadequate explanations for the maintenance of mutualistic interactions. In theory, shouldn’t rapidly-evolving bacteria living in a host with a generation time that is orders of magnitude longer occasionally evade host-sanctioning mechanisms? And yet according to Sachs’ analysis, the transition from mutualist to parasite is the least common shift seen in the bacterial phylogeny. While I think Sachs’ phylogenetic analysis is really valuable, his explanation of the trends uncovered in the bacterial phylogeny are ad-hoc and require real theoretical and empirical validation.

Increasingly, I do not see how you maintain stable mutualisms without some form of group selection. I am not saying that I am sure that group selection is always at work, I just do not see justification for removing it from the list of possible mechanisms. I also see the argument that kin selection is not at work in mutualisms as being a bit distracting. It is true that two species do not share gene pools (although see Strassmann and Queller 2010 for an interesting possibility that sometimes we should consider inter-species partnerships as single organisms [superspecies?]). But this does not mean that kinship is irrelevant in mutualisms. Within each species that participates in the mutualism — particularly in bacterial species living symbiotically within a host — there is serious incentive to help kin by displaying restraint in exploiting one’s host. If an individual bacteria mutates to become parasitic, what prevents this mutant from converting the mutualism into a parasitism? Clearly there is an incentive on the part of not only the host but also the symbiotic bacterial population to suppress such cheating. Could there be selection for symbiont populations that show restraint? I think that the answer has to be ‘maybe’, and whether you call the mechanism producing such restraint ‘kin selection’ or ‘group selection’ is probably more a matter of ideological perspective than actual disagreement over mechanism. Regardless, such a mechanism selecting for restraint would not fall into the three categories discussed by Sachs. It seems a little naive to me to believe that byproducts, partner choice, or partner fidelity can generate and maintain mutualisms in absence of some other selective mechanisms. Our explanations for how and why so many mutualisms exist in nature need to be greatly improved.

In this mostly-theoretical talk, Queller attempted to clear up some of the confusion over how cooperation evolves. He began his talk by listing all of the mechanisms posited to explain the evolution of cooperation. He suggested that inclusive fitness theory and kin selection are often made equivalent when they are not: kin selection is a process, whereas inclusive fitness is an accounting method (which can sometimes be used to quantify the effects of kin selection, but might also quantify the effects of other selective mechanisms).

Queller advocates that we think of selection operating among kin, kith, and kind. Assuming that the role of kinship was well-established in the minds of his audience, he went on to explain what he means by “kith” and “kind”. Kith is an archaic word meaning friends, neighbors, and acquaintances. Queller suggests that kith selection can evolve by partner choice or other mechanisms that make cooperation conditional on reciprocation. Interestingly, such a definition allows us to consider inter-species mutualisms as well as interactions between individuals of the same species (again, see Strassmann and Queller 2010). What makes kith selection work is the covariance of the cooperative phenotype among ‘friends’, a quantity that can be measured and is labeled by Queller the “kith coefficient” (f). Considered in this manner, cooperation is predicted to evolve whenever:

m*f – c > 0

Where f is the kith coefficient, m is the positive effect of mutualistic behaviors, and c is the cost of those mutualistic behaviors. Queller pointed out that this is simply Hamilton’s Rule in a form that depicts unrelated individuals. Thus, kin and kith selection have some equivalence.

Next, Queller moved on to discuss kind selection. Kind selection is his way of depicting the greenbeard mechanism for the evolution of cooperation: cooperation occurs between individuals who reliably communicate their willingness to cooperate. He differentiated this sort of recognition from “true kin selection”, where individuals target their altruism at known relatives. Under kind selection, I need only know that an individual shares a known marker of cooperation in order to direct altruism at that individual. According to Queller, Dictyostelium discoideum is a good example of a species under kind selection, as it uses an adhesion molecule critical for cooperation as an honest signal of ‘willingness’ to cooperate.

To make it clear how kin and kind selection differ, Queller contrasted the two across several criteria. Kin selection requires that two individuals share the same allele, whereas kind selection only requires that two individuals express the same trait. Thus, kin selection is a process of “identity by descent” whereas kind selection is a process of “identity by state”. Under kin selection relatedness is the same across the genome, whereas under kind selection it is higher at the locus that expresses the greenbeard trait. Kin selection can be multigenic, whereas kind selection involves one gene or linked gene complex. Kin selection can give rise to complex forms of cooperation, whereas kind selection can only foster simple forms of cooperation. Kin selection produces additive fitness effects, whereas kind selection effects are not additive. Finally, kin selection is not frequency-dependent, whereas kind selection is. All of these comparisons went by at lightening speed, so I am not totally certain I got them all right. And moreover I must admit that I do not fully understand them all.

Based on Queller’s presentation, it is clear to me that I need to do some more reading and thinking about inclusive fitness. In particular I need to understand how to apply the Price Equation, which appears to unite all of the forms of selection discussed by Queller. For whatever reason the placement of all forms of selection for cooperation under the umbrella of inclusive fitness seems like a sleight-of-hand to me. As Sober and Wilson (1998) have pointed out, simply accounting for all forms of selection using inclusive fitness theory (the “averaging fallacy”) obscures the diversity of mechanisms that can produce cooperation. I cannot say with certainty that this is what Queller is doing in making kith and kind selection comparable to kin selection, but it feels like this might be the case. I need to gain a deeper understanding of these issues to make a credible judgment.

For years Strassmann and Queller have studied cooperation in wasps, so it must have come as a surprise when they turned to the social amoeba Dictytostelium discoideum as a new platform for scientific inquiry. In her talk, Strassmann explained why ‘Dicty’ was too good a study organism to turn down. She indicated that the major issue with studying social behavior in wasps was the inability to connect genes with behavior, so she and her long-time collaborator went looking for a new study system. They decided to work on Dicty because it:

• is social;
• is simple;
• can display complete self-sacrificing altruism;
• is possible to study in its natural context (soil);
• has a rich and well-resolved phylogeny displaying variation in social behavior;
• is amenable to experimentation;
• can be subjected to experimental evolution;
• can be studied using gene knockouts to determine the function(s) of particular genes;
• has a genome that has been fully sequenced; and
• is studied by a diverse group of potential collaborators.

It might be a little strong to say that Strassmann is a Dicty evangelist, but she does make a strong case for the system.

Strassmann explained that Dicty, which can live as a single-celled solitary individual, has three life cycles: vegetative, sexual, and social. It is the social cycle that is of interest to those of us who study cooperation, because under environmental stress individual autonomous Dicty cells can come together to form a cooperative slug. This slug displays somewhat astonishing coordination as it ‘crawls’ away to a place suitable to form its fruiting body. This fruiting body allows some members of the slug to form spores and escape, but only because other cells sacrifice themselves to become part of the stalk that elevates the spores about the soil substrate. This social behavior is easy to induce in the lab, making the study of cooperation possible.

A key question about Dicty is how it maintains its cooperative behavior given that it forms from an independent assortment of cells which may or may not be related to each other. Although the social form of Dicty seems like an analog of a multicellular individual (and is indeed being studied in an attempt to understand the evolution of multicellularity), it differs from most ‘true’ multicellular organisms in a key way: it is chimeric, containing more than one strain with potentially competing genetic interests. Chimeric slugs can be studied by taking wildtype Dicty strains and exposing them to insertion mutations. The resulting insertion mutants have been shown to cheat by disproportionately seeking representation in the back of the slug, which becomes the ‘spore’ component of the fruiting body. Interestingly, the number of mutations that can cause cheating is great, suggesting that the pathways responsible for cooperative behavior are complex enough to be disrupted in multiple ways.

In the laboratory, researchers are beginning to unravel the mechanisms by which Dicty maintains it cooperative social behavior. Strassmann depicted “noble resistors”, “control by pleiotropy”, and “control by high relatedness” as three pathways by which Dicty stays cooperative in the face of free-loading by mutant strains. Some of these cheater mutants are so extreme in their defection that they cannot form fruiting bodies in the absence of other more cooperative strains. Not surprisingly, these strains do not exist in nature, suggesting that they are eliminated by some selective process. In fact, when she went out to collect wild fruiting bodies, Strassman found no extreme cheaters and few chimeric slugs, suggesting that in the wild spatial structuring is leading to overall high relatedness in fruiting bodies. She depicted this as resulting from a form of frequency-dependent selection, although I was not totally clear on what she meant by this term in this context. Dicty can also apparently exercise some form of kin or kind recognition, and is more likely to try to cheat other strains when it is found in a chimeric slug.

Finally, Dicty is not just about social evolution within a species or strain: it also has the potential to inform our understanding of mutualism. Strassman depicted the work of one of her lab members, Debbie Brock, who has shown that about a third of Dicty clones are bacterial farmers. While the ancestral niche of Dicty is that of a predator on bacteria, some clones can actually prudently consume bacteria, forming a mutualism that may help them in their direct and indirect competition with other Dicty strains.

If you want to learn more about Dictytostelium discoideum, check out the DictyBase website.

Velicer gave his talk a new secondary title, which was something like “Social Diversity in the Myxobacteria from 10⁷ to 10^-7 scales”. I am not totally sure that I got these scale limits correct, but the point was that we can look at the social diversity of these bacteria at broadly varying scales. Like the amoeba Dictyostelium, Myxobacteria that are starving form a fruiting body. Using a quorum-sensing molecule to assess whether there are enough comrades to form a socially-cooperating group, Myxobacteria can also form what Velicer termed a “wolf pack” that swarms as it produces antibiotic and lytic compounds that allow it to digest other bacteria. Myxobacteria present an interesting system for studying social evolution because they can exist in both chimeric and non-chimeric forms, both possible routes to the evolution of multicellularity.

According to Velicer, Myxobacteria form a diversity of fruiting bodies, although only a few can easily be cultured in the lab. Interestingly, it is not fully known why Myxobacteria make fruiting bodies — which is a contrast with Dictyostelium, in which the role of fruiting in dispersal has been well-understood. Whatever its purpose, the fruiting body is clearly a cooperative structure, and the usual questions “why not just go solo?” and “how is cooperation maintained against the threat of cheaters?” can be asked using the Myxobacteria as an experimental system. And given how much diversity there is in this group, we can also ask what drives the diversity we observe.

Velicer discussed some of the approaches that are being used to answer these questions. One approach is experimental evolution, which has been used to look at the effect of relaxing selection, how cheaters and cooperators have coevolved, the role of cheater suppression mechanisms, and the origin of kin discrimination. In addition, large-scale surveys of the diversity of Myxobacteria have shed light on how cooperation is maintained in this group. Natural populations show diversity in their social interactions, the timing of development and overall morphology, the rate at which they swarm, the cell density at which they go social, and the range and efficiency of their predatory swarm. Velicer described a series of surveys that looked at the diversity of Myxobacteria, the overall result of which was that divergence increased with distance. This result suggests that these bacteria might — like their multicellular brethren — be subject to allopatric speciation, and therefore may not be ‘cosmopolitan’ as is often suggested for bacteria. On a smaller scale, Myxobacteria in the lab suggest that individual clones may be highly-sensitive to differences with other clones, sometimes to the extent to which certain pairs of clones will not form social swarms together. There also can be social asymmetry, with particular clones exploiting others consistently. Velicer suggested that if genetically-similar neighbors do not get along, sympatric speciation may be possible in the Myxobacteria. There are also ‘super cheater’ strains that can maintain up to a 50-fold advantage over cooperative strains, which certainly begs the question “how do cooperators resist cheaters?”. Velicer suggested that there is strong frequency-dependent selection against cheaters. To my ear, this sounds like group selection: get caught in a region with too many cheaters and none of the local strains can go social, which puts them at a strong disadvantage when competing against other regions where sociality leads to greater predation success.

Between Myxobacteria and Dictyostelium, it is clear that there are ample opportunities to observe, manipulate, and evolve cooperative behavior. I expect that both of these social microbes will yield important insights into how cooperation evolves in the coming years. Velicer’s presentation of the spatial dimensions of cooperation in Myxobacteria excited me, because I believe that a lot of the ‘big questions’ in this field have the potential to be answered by looking at the spatial constraints on cooperation.

Ben Kerr delivered a talk with a simple message that reverberated through the rest of the meeting: restraint is the hallmark of cooperative behavior. Using the Prisoner’s Dilemma as a frame, Kerr talked about prudent and rapacious strains and how both active mechanism (such as partner choice or altruistic spite) and passive mechanisms (such as limited dispersal) can promote the evolution of cooperation.

Kerr began his talk with a discussion of colicins, chemicals released by some strains of Esterichia coli that are designed to poison competitors and favor comrades. Colicins are a public good in that they help any individual who is immune to them, which ought to make them terribly susceptible to cheating mutants who fail to produce the toxin. This is particularly true because producing colicins is a strongly altruistic trait: those individuals who produce the toxin also lyse themselves, essentially committing suicide to help their comrades. Kerr suggested that we think of colicin production as a greenbeard trait, with the three genes for colicin production, immunity to colicin, and lysis closely linked. This was a little unclear to me; although I see that it helps to have these three genes closely linked, I still do not totally see how it is a true greenbeard because the ‘identifying’ trait is the gene for immunity while the ‘altruistic’ trait is the gene for colicin production, so I do not see how this system is protected against mutants who lose the colicin production gene but not the immunity gene.

Kerr went on to discuss how frequency-dependent selection can act to preserve both altruistic and non-altruistic strains at some equilibrium. Clearly a toxin-producing strain has a competitive advantage over a strain that does not produce toxins and is sensitive to their negative effects. So why doesn’t the toxin-producing strain competitively exclude sensitive strains? Kerr suggested that resistant strains mediate the competition between toxin producing and sensitive strains, because these resistant strains out-compete toxin-producing strains (because they do not incur the costs of toxin production but also do not suffer the ill effects of toxin exposure) but can be overrun by sensitive strains (which do not pay the costs of resistance). This sets up what is known as a “non-transitive” competitive system, more commonly known as a “rock-paper-scissors” system after the children’s game with similar properties.

Kerr has shown through both spatially-explicit simulations and experiments that limited dispersal is the key to the evolution of cooperation. In the experiments — which take place in 96-well plates that are periodically mixed via robotic manipulation — altruism evolves most strongly when different wells are limited to local and infrequent dispersal. Kerr has also shown similar results in the virulence of bacteriophages: if dispersal is limited, bacteriophages show lower virulence than when they disperse more globally.

I really appreciate Kerr’s work for several reasons. The first is his very careful combination of predictive simulations (which allow for the comparison of the predictions made when we make different assumptions about a given social system) with innovative experiments that allow him to perform experimental evolution under very controlled conditions. I guess if you were to level one criticism of Kerr’s modus operandi, it would be that he creates highly contrived systems that verge on being ‘biological computers’ rather than real analogs of known systems. I do not really buy this criticism; although there is no doubt that a high-throughput robot manipulating 96-well plates is far from “natural”, we really have no idea how microbes spatially segregate or how limited their dispersal is, so the contrived system suggests some viable scenarios that might represent some real systems. The only thing about Kerr that disappoints me is how soft he is in describing the significance of his work. At the close of his talk he suggested that both kin selection and multilevel selection are important factors in the evolution of restraint. But this seems too kind to the kin selection camp, as kinship only emerges post-facto after a process that is clearly group selective.

Despite my gripes with his soft stance on multilevel selection, I am a huge fan of Ben Kerr’s work. It is not just that he does some really interesting, innovative experiments: he also explains what he does with exceptionally-strong conceptual clarity. His talk was elegantly constructed, with really helpful sophisticated visuals and a logical flow that makes fairly complex ideas seem utterly understandable. Based on his presentation, I have to imagine that he is an amazing teacher as well as a creative researcher.

Like a lot of other speakers, Werren came with a new subtitle to his talk, “Does conflict beget innovation?”. He talked about “selfish genetic elements” (SGE’s), which he also labeled as “parasites in the genome” and “ultraselfish genes”. Basically these are genetic elements which try to advance themselves within the genome of their host by consistently over-representing themselves at before and particularly at reproduction. They create a conflict between their genetic interests (to get as many copies of themselves into the next generation) and the collective genetic interests of the genome in which they reside (to create viable offspring that can create more copies of genes in subsequent generations). So you might wonder: how did conflict and selfishness find their way into a meeting that is so focused on cooperation? The answer of course is that selfishness and conflict are the counterbalance to cooperation, and in the case of genomes (which generally are highly cooperative) we can better understand cooperation by understanding the prevalence and severity of threats to genomic cooperation. Whereas many social systems we study are mildly-cooperative in the face of severe selfishness and cooperation, genomes are generally cooperative but face a slight nagging drag as a result of selfish genetic elements.

Werren pointed out that while genomes are by necessity highly cooperative, that effective cooperation creates the possibility for diverse means of genomic freeloading. Genomic sequencing has illuminated the diverse ways in which genetic parasites ‘infect’ their host genomes. Werren reviewed these diverse mechanisms, starting with the well-known mobile elements. Mobile elements are genes that can move ‘horizontally’ in the genome by copying themselves into new locations in the genome, frequently causing harmful mutations by inserting into functional genes. Despite this potential harm these genes are highly prevalent, making up a whopping 45% of the human genome and 85% of the corn (maize) genome. A related but less damaging SGE is the biased gene converter, which also inserts itself throughout the genome but is more selective in locating itself outside of functional gene regions.

Making copies of yourself throughout the genome is not the only trick of SGE’s. They can also ‘cheat’ in the way that they get transferred to the next generation. For instance, the phenomenon of meiotic drive occurs when a particular gene can bias meiotic segregation that results in more than 50% of gametes carrying the SGE. Similarly, post-segregation distorters produce toxins that kill off all gametes that do not contain the SGE sequence; obviously this is quite negative to the host as a 50% decrease in gametes reduces the probability of successful union of sperm and egg. A particularly nasty post-segregation distorter called PSR resides in the genome of sperm but hops from sperm to sperm, killing the rest of the genome that accompanies it. These SGE’s might be fighting at a scale too small for us to really appreciate, but they do fight nasty!

Werren also pointed out that heritable microbes can lead to genomic conflict. The Buchnera vertically transferred by aphid species and intracellular microbial symbionts (mitochondria and chloroplasts) — both discussed by previous speakers — have a strong interest in having themselves over-represented in the gametes through the production of more females than males. This is because in the next generation, only females will have the potential to pass on their particular strain of symbiont. For the overall host genome, however, producing more females than males may lower overall reproductive success. This kind of “reproductive parasitism” is common, and no manipulator of hosts is better known than Wolbachia, a species that can further its own success by a variety of mechanisms that harm its host. Some strains of Wolbachia feminize their host, turning males into females. Some compel their hosts to become parthenogenetic, preventing the dilution of their transfer that would result from sexual reproduction. Some simply kill all males produced by their hosts, while others use cytoplasmic incompatibility to cause host-produced sperm to kill eggs that are free of Wolbachia infection. Overall, given how many ways there are to parasitize a genome, it is difficult to see how complex genomes have survived the continuous onslaught they have faced. Werren did not really address this question (“How has cooperation within genomes persisted?”), but it is clear that the benefits of having a mostly-cooperative genome are so great that a fair amount of freeloading can be tolerated. I do wonder what the ‘breaking point’ for genomic conflict is, and what mechanisms of evolutionary change have prevented genomic conflict from rendering organisms more simple.

An interesting final tidbit of Werren’s talk was his suggestion that there may be a connection between meiotic drive and speciation. Apparently variation in the different suppressors used by meiotic drive SGE’s has been shown to cause hybrid sterility, and might create a means of rapidly isolating sympatric populations.

Robinson had the unique task of entertaining a large group of evolutionary biologists after a really nice banquet. Whether that is an enviable task or not is probably a matter of preference, but the audience seemed to return to its normal level of attentiveness after a full meal. Robinson is a ridiculously prolific publisher in the area of what I would call ‘behavioral genetics’, and he unloaded a motherload of information on the subject; some of his talk I followed, some I did not. Whether that reflects my relative ignorance of this area or post-dinner fatigue is hard to determine.

Robinson started out by discussing the idea of social convergence. He works in arthropods — mostly hymenopterans — but sees connections with the social traits of vertebrates. Specifically he identified language, agriculture, and warfare as these social mega-traits that are shared by both groups. In the Hymenoptera, these complex social adaptations are accomplished through division of labor. For instance, bees maintain two castes (nurse and forager); other species, in particular many ants, maintain more complex caste systems. Robinson is interested in the evolution of both social life and the individual behaviors that give rise to it, and wants to understand the adaptive nature of these behaviors.

To get at this large question, Robinson and his collaborators have looked at how different evolutionary forces (in particular kin and group selection) act on processes such as feeding, nutrition, metabolism, and reward. Put another way, they wish to make connections between proximate mechanisms of sociality and the ultimate selective forces that shape these mechanisms. Robinson pointed out that what makes the Hymenoptera interesting is that reproduction and care for young are almost completely decoupled. While a foundress does provide some care, she switches to being a queen focused solely on reproduction as soon as her first brood of offspring are able to assume the task of caring for their sisters. To get at how such a system might have evolved, Robinson has looked at gene expression in the different castes present in a nest: gyne, queen, foundress, and worker. A cluster of “care genes” has been shown to be most similarly expressed in foundresses and workers, less similarly expressed in queens, and quite differently expressed in the reproductive gynes who disperse to found new colonies. This suggests that the evolution of sociality has relied on using environmental cues to differentially regulate care behavior through differential gene expression in different castes. Although Robinson did not specify how such a system evolved, I thought it was interesting that once again it is gene regulation that is implicated in a major evolutionary transition.

Backing up to a more macroscopic view, Robinson then discussed how the genetic basis of sociality varies between the “primitively eusocial” and “highly eusocial” insect species. By looking at the percentage of genes involved in social behavior, Robinson was able to show that although there are commonalities between these two major groups (most genes involved in eusociality are metabolic in nature), their social evolution has relied on changes to very different groups of genes. This suggests that eusociality in these two groups is actually a convergent trait, independently evolved along each lineage. Given that the two groups display different levels of eusociality, it would be interesting to see how the independent genetic pathway to social behavior followed by each group may have constrained one group’s potential for eusociality more than the other’s.

Next, Robinson zoomed in to the metabolic pathways that are so commonly responsible for mediating social behavior. He described the “behavioral maturation” process of worker individuals, who transition over their lifetime from nurse to defenders of the hive to foragers. This transition parallels a decrease in the amount of fat stored in each individual: nurses contain more fat reserves than foragers. There was more to this story but this was definitely part of Robinson’s talk that I was not completely following.

The last major section of Robinson’s talk focused on the reward systems that dictate different behaviors in different insect castes. He focused on the effect of octopamine in the foraging caste of bees. Keep in mind that foragers have lower fat reserves and are thus more sensitive to changes in their nutritional status. In non-social organisms, the octopamine pathway is strictly there to assure that the individual is sufficiently motivated to eat when reserves are low. But in the social insects, this pathway has been co-opted such that octopamine does not only regulate foraging behavior but also the intensity of dance behavior, the chief means by which the bee hive communicates to optimize collective foraging. Robinson calls this a “me to we” transition.

Along with collaborators, Robinson has been able to unravel how this “me to we” transition might have occurred. Taking advantage of the fact that cocaine stimulates dopamine production in a broad variety of taxa, Robinson’s group fed bees cocaine to see how it affected their dance behavior. In a really wild experimental set up they were able to manipulate the bee’s perception of distance, their baseline reward for foraging, and the amount of augmented perception of reward they received through exposure to cocaine. According to Robinson, their results provide strong evidence for a “hijacking” of the individual reward system to produce social behavior. Other work they have completed shows that the aggression displayed by guard castes results from a down-regulation of brain activity in these individuals, another form of social co-option.

In an interesting final summary, Robinson suggested that the next steps in his research area were to determine who evolutionary mechanisms — specifically what kinds of selective pressures — might be responsible for the transition from “me to we” in social insects. He contrasted the kind of cooperation evolved in eusocial insects with that of humans; unlike the social insects, we have not become reproductive specialists. Robinson suggested that this represented a “me to me-me-me” transition. This transition seemed to emerge from a very different change in behavioral regulation: whereas the story in eusocial insects is one of limiting the capabilities of individual social actors, human cooperation seems to have been fostered by a great expansion of individual capabilities. What I find interesting about this reality is that despite our broad individual potentials, we nonetheless end up specializing. The degree of specialization in human societies puts the eusocial insects to shame, and I doubt that the processes that drive specialization in humans are all that analogous to those that drive social insect specialization.