Evolution of multicellularity
In the early 1990s a set of studies on the evolution of cooperation and conflict were concluded: Rainey finally got round to writing this up about 10 years later thanks to a sharp prod from Greg Velicer. The paper is here and Greg’s paper was published along side.
Attention was on wrinkly spreader (WS) genotypes that emerge from the ancestral genotype when propagated in spatially structured (unshaken) microcosms. The success of WS is attributable to cooperation among individual cells: by linking to each other (via the over production of adhesive polymers) and ultimately to the edge of the glass vial the group of cells forms a self-supporting mat. Belonging to the group is costly to individual cells, but these costs are out-weighed by the benefits of group membership (group membership ensures access to the oxygen replete air-liquid boundary (there is no oxygen under the mat)).
The picture above shows an intact mat (the cumulative product of the cooperative interactions of millions of cells). When the mat becomes too heavy (or too weak) it collapses into the broth (it is not buoyant) as seen in the middle microcosm. It is important to note that a WS mat is far more than the sum of the individual parts. The photo of the microcosm on the right was taken immediately after disturbing (with a brief shake) a vial with an intact mat. The mat breaks into many pieces (just visible on the bottom) and does not spontaneously reform. While a mat will eventually re-emerge it will do so by a process of growth and development from just a single cell.
Although the evolutionary emergence of WS mats is a wondrous thing, such mats are short-lived. Selection continues to act at the level of individual cells and in so doing, favours mutant types that “cheat”, i.e., cells that no longer produce the adhesive polymers, but nonetheless take advantage of the benefit that accrues from being part of the mat. From here on I refer to these types as “non-sticky” or “dispersing cells” and the mat-forming WS cells as “sticky” types. In the absence of any mechanism to repress them, the non-sticky cells prosper (they are in effect a cancer), ultimately weakening the fabric of the mat to the point where it collapses.
Having witnessed the evolution of groups we began to think of the possibilities for the further evolution of these groups. At this point things became interesting. The typical group selection experiment would involve imposing selection for some group property, for example, mat strength, which is readily done by placing glass beads on the mats to find the strongest mat. The strongest mat (if that is the group trait of interest) is then allowed to reproduce (the experimenter takes that group and places some cells in a new microcosm. Some interesting things can arise in response to these types of experiments (Mike McDonald has even witnessed the evolution of a division of labour), but to me these kinds of group selection experiments fail to hit spot.
The reason for dissatisfaction stems from a desire to observe the evolution of mats under circumstances where mats participate in the process of evolution by natural selection as units in their own right. For this to occur there must be variation among mats, mats must be capable of giving rise to mat babies and the offspring mats must show some semblance to parental types. See Lewontin (1970) who derived this formulation and Peter Godfrey Smith’s excellent book for more. While mats show substantial genetic variation — this variation, depending on the ecology of the environment (see Ecological Scaffolding) maybe discrete — mats lack the capacity to leave off-spring mats. As a consequence, mats are an evolutionary dead-end.
Many would not see it this way, pointing to bits breaking off mats as a means for mats to leave off-spring copies. While it would be imprudent to ignore this possibility, we think this is an unlikely route to an endogenous means of reproduction, because pieces of mat that break off fall to the bottom of the vial and the cells die. Others might point to the trait group models of David Sloan Wilson and say this provides a way around the problem, to which I would reply, no it doesn’t! The trait group models require the entities that comprise the mat (“cooperating” and “defecting” types) periodically disperse and then re-group (otherwise defecting types take over). This might be possible in populations of organisms that have already evolved the capacity to decide when to interact and when not to interact, but in the context of these primordial groups the cells are unable to simply turn off production of the adhesive polymer, swim away, and then reform groups with different frequencies of cooperators and cheats. Such models simply do not reflect biological reality. Moreover, it is very difficult to see how selection can act with potency given the very limited heritability that operates when mats are formed by blending of genotypes.
This led us to suggest that if simple undifferentiated groups (such as the WS mats) constitute a raw material for the evolution of more fit groups, then it is necessary to face up to a paradox: for WS mats to become more complex they must be able to participate in the process of evolution by natural selection, but to participate in the process of evolution by natural selection they must be more complex.
Is there a way forward? We think there is. Here we recount a specific instance that draws on ecology, but in the section on Ecological Scaffolding we elaborate this in more general terms. Careful observation of the dynamic of WS evolution reveals the WS mats (which appear to be evolutionary dead-ends) and non-sticky types genotypes naturally cycle — in a frequency dependent manner. Mats are selectively favoured in static microcosms, but their success establishes conditions that favour the evolution of non-sticky genotypes arise readily by mutation and can leave the group. Moreover, non-sticky types can give rise to mat-forming WS genotypes by compensatory mutation. Remarkably this process can go on for some considerable time, as has been shown in spectacular fashion by Bertus Beaumont and more recently by Michael Barnett. If we pause for a moment and look at this dynamic from a different perspective we see a life cycle, where the WS mat approximate the soma and the non-sticky types approximate a germ-line. Admittedly this is a clumsy life cycle, reliant on mutation to switch between the different stages, but this life cycle does get us out of the paradox. WS mats, thanks to the production of non-sticky dispersing cells, can leave collective copies. They can therefore participate in the process of evolution by natural selection.
A useful way to view this in a plausible natural setting is to consider a pond studded with reeds. Each reed allows establishment of a single microbial mat (the soma-like phase), with the spacing of reeds ensuring variation at the level of mats. Mats that collapse, for example, through physical disturbance, are extinguished, allowing the possibility that an extant mat might, via production of a dispersing (germ-like) phase, increase its representation among the population of mats. Thus the possibility of a selective process unfolds at the level of mats.
This simple idea has a number of interesting implications. One being a possible route for the evolution a soma / germ-line distinction; another being a mechanism for the control of “cheats”. This arises from the mutual dependency of germ-line on soma and soma on germ-line.
In terms of multi level selection theory, recognition that the some-like / germ-like distinction might, under some circumstances, be viewed as a development cycle allows us to disband the naive assignment of fixed fitnesses to cheats and cooperaters. More importantly though, we assign fitness to the developmental programme.
The germ of this idea was published in 2007 as a short essay. Ben Kerr and I published a fuller account here in which the ideas are more thoroughly developed. Caroline Rose and Katrin Hammerschmidt performed an outstandingly heroic experiment in which they show these ideas to have substance. An additional experiment performed by Caroline and Katrin is currently available as a (soon to be updated) pre-print. Remarkably, during the course of the experiment the initially clunky life cycle evolved to the point where it came under the control of a simple but effective developmental switch. The work has continued in subsequent years thanks to Daniel Rexin and Philippe Remigi and now, with the resources of the MPG the work further expands on both theoretical and experimental levels.
We are especially interested in the process of lineage selection, the emergence of life cycles and all that goes with them including the reproductive division of labour and the evolution of development. Joanna Summers in particular is doing some wonderful work on the evolution of endogenously controlled life cycles.