Pierrick Bourrat’s new paper in Philosophy and Theory in Biology criticizes aspects of the influential ‘export of fitness’ framework developed by Rick Michod and colleagues and extended by Samir Okasha (Bourrat, P. 2015. Levels, time and fitness in evolutionary transitions in individuality. Philos. Theory Biol., 7: e601. doi: 10.3998/ptb.6959004.0007.001). According to this view, an evolutionary transition in individuality, for example from unicellular to multicellular life, involves a transfer of fitness from the lower level units (e.g. cells) to the higher level unit (e.g. nascent multicellular organism). Fitness is defined as the product of viability and fecundity, and the emergence of a division of labor between reproductive (germ) and non-reproductive (somatic) units at the lower level exports fitness to the higher level. Full disclosure: Rick Michod was my Ph.D. co-advisor, and he has had a huge influence on my thinking about this topic.
For the sake of readability, I’ll refer to cells and (multicellular) organisms, though the principles are generally agreed to be similar for (at least) the fraternal transitions in individuality. It just gets a bit cumbersome to keep referring to ‘lower-level units,’ ‘higher-level units,’ and ‘propagules.’
Michod’s framework is largely focused for the potential for conflict among cells and the need to prevent or mediate these conflicts, a need that is generally met by a unicellular bottleneck in Michod’s models. Conflict mediation, though, is not sufficient for the emergence of a higher-level individual; a division of labor between germ and soma is also required. Up to this point, Bourrat seems to have no problem with Michod’s framework. However, Michod claims that germ-soma differentiation results in cells with zero fitness, and this is where Bourrat disagrees:
Michod and colleagues’ model relies on an assumption regarding fitness of the higher-level entity (hereafter the “collective”) that is inconsistent with the claim made about the fitnesses of the lower level entities (hereafter the “particles”). I claim that an ambiguity surrounding the concept of viability (fertility) of the particle/ viability (fertility) of the collective exists in Michod’s and Okasha’s writings and creates the illusion that the fitness of the particle can be nil while the fitness of the collective is high.
The crux of Michod’s claim that specialized cells have zero fitness is that germ cells invest all of their effort into reproduction, resulting in zero viability, and somatic cells invest all of their effort into viability, resulting in zero reproduction. Since zero times anything is zero, the product of viability and reproduction for either cell type is zero. Bourrat objects that Michod is conflating viability and fecundity of cells with viability and fecundity of the multicellular organism:
…the terms ‘viability’ and ‘fertility’ at the collective level do not respectively equate to the terms ‘viability’ and ‘fertility’ at the particle level. In fact, during the phase of development of the multicellular organism – that is, when only the somatic or viability function of the organism is performed – the somatic line has offspring, and to do so the somatic cells need both to survive and reproduce. Similarly, during the phase of reproduction of the multicellular organism the germ cells both survive and reproduce…the viability and fertility of the multicellular organism result from the effort of the cells in those variables. Yet, if the invested effort of a particular line of cells in the viability (fertility) of the multicellular organism is nil, it does not follow that the viability (fertility) of the cell is nil.
This is mostly true, although in Volvox at least, the somatic cells born of the final round of cell division do not have any offspring. I agree with Bourrat that Michod’s claim that differentiated cells have zero fitness is either metaphorical or mistaken, but for different reasons. It is trivially easy to show that germ cells do not have zero fitness: they (usually) have offspring. Even for soma, Michod’s conclusions rely on a particular definition of fitness; the somatic cells resulting from the final round of cell division have zero offspring and zero direct fitness, but if the germ cells, with a relatedness of r = 1 to the somatic cells, reproduce, the inclusive fitness of the somatic cells is nonzero. Michod’s model confounds EFFORT invested in the components of fitness with the components themselves, especially when components are realized in the context of the multicellular group. It is reasonable to assume that viability and fecundity are functions of the effort invested in each, but it is not reasonable to assume that the components are equal to the effort. Furthermore, all cells invest in their own viability; as Bourrat points out, a germ cell with zero viability is useless. I’m not sure if Michod’s view on this has changed; in his 2011 book chapter, he refers to ‘low cell fitness’ rather than ‘zero cell fitness’ (p. 181).
Bourrat takes an entirely different approach to the question of somatic cell fitness. Rather than arguing on the grounds of inclusive fitness, he argues that the claim that somatic cells do not reproduce (and thus have zero fitness) fails when reproduction does not require material overlap (‘formal reproduction,’ in Peter Godfrey-Smith’s terms):
For formal reproduction to happen, relatively to a particular grain of description, the only requirement is the presence of a causal link between the presence/absence of a somatic cell in the parental organism and the overall number of cells produced (number and quality of offspring).
Fig. 1 from Bourrat 2015. Toy illustration of the formal causal relation between the presence/absence of one somatic cell in a multicellular organism of the genus Volvox and the number of multicellular offspring produced.
Here I think that Bourrat (and, to be fair, the inclusive fitness account) has missed an important point of Michod’s models. In one important sense, the somatic cells (but not the germ cells) do have zero fitness: they are evolutionary dead ends in that novel mutations that occur in somatic cell lineages will never be passed on to future organismal generations (unless they cause the somatic cells to produce offspring, in which case they are, by definition, no longer somatic cells). The claims about formal reproduction may be right, at least for some concepts of fitness, but Bourrat’s example shows that fitness concepts based on formal reproduction fail what I consider the most important function of fitness: to predict, at least probabilistically, the short-term outcome of selection.
None of this means that Michod’s models are not useful. To be sure, the claim that germ-soma specialized cells have zero fitness is either wrong or metaphorical (Godfrey-Smith makes a similar point). But we need not throw out the baby with the bathwater; the ‘export of fitness’ framework is still (IMO) a useful heuristic for thinking about evolutionary transitions in individuality. The division of labor between reproduction and viability can be usefully conceptualized as cells increasing their effort at one component of fitness at the expense of the other, resulting in cell groups that require both cell types to survive and reproduce. Whether or not such specialization creates a disconnect between cell-level fitness and organism-level fitness is the subject of the remainder of Bourrat’s paper, and one that I will address in a future post.
Here are my thoughts on a quick reading of the first half of the Bourrat paper. Rick Michod
1. All the life history models using vi, bi, V and B we developed were based on effort at reproduction and effort at survival. Since survival and reproduction were assumed to be monotonic functions of effort at survival and reproduction, respectively, it was easier to work in terms of survival and reproduction themselves.
2. The models do assume an isomorphism between cell level and group level efforts. So that efforts by a cell at cell level v or b do translate into V and B for the colony. Like I have said many times this is a simplification, likely to hold initially in a transition, but not in complex multicellular forms. The assumption seemed to work well for flagella action, the total flagella force for the colony could reasonably seem to be the sum of the forces generated by each cell. The assumption is less clear for effort at reproduction.
3. It is hard for me to follow Bourrat, there aren’t any real models developed so it is hard to follow his reasoning and many of the views he attributes to me I don’t agree with, nor do I see why he attributes them to me.
4. From my quick reading of Bourrat, he seems to think (e.g., p. 5 bottom) that I think or claim that germ cells in a colony don’t survive while in the colony. Of course, this isn’t true, the whole point of being in a colony with somatic cells is that germ cells will survive by virtue of the effort put into survival by the somatic cells. Somatic cells and germ cells make a good team while in a colony as they put effort into the fitness components ignored by the other cell type. So it should be clear that the vi and bi variables are NOT the realized survival and reproduction while in the group, as I think Bourrat is interpreting them. Instead, they are the survival and reproduction cells would have were they to leave the colony and exist alone. A germ cell were it to leave the colony would not survive if they weren’t putting any effort into survival, but in the colony can survive by virtue of the effort put into survival by the somatic cells.
5. Evidently, reading Bourrat, Godfrey-Smith also is confused about vi and bi in my models. I may not have always been clear on the interpretation of vi and bi as being survival and reproduction cells would have while on their own, not in a group, but clearly vi couldn’t ever be survival while inside the group as Bourrat is claiming. That would contradict everything I am trying to say about somatic cells and germ cells being a good team and together having a high fitness.
6. Likewise a somatic cell were it to leave the group would not reproduce if it were not putting any effort into reproduction. However, within the group, a somatic cell can contribute to the colony reproducing by helping the colony survive to the stage of reproduction.
7. Although somatic cells can contribute to the group reproducing through contributing to survival to the reproductive stage, I do not agree with the conclusion arrived at by Bourrat on p. 6 using the formal reproduction criterion that somatic cells can be seen to reproduce. That would certainly be a surprise to Volvox biologists as somatic cells don’t even divide. If attributing reproduction to somatic cells is a consequence of the “formal reproduction” criterion, I would see this a good reason to not use the idea of formal reproduction in this case.
8. The reason I liked the “export of fitness” description of the transition is because of the central role of altruism in the transition. The costs of altruism reduce the fitness of the particle and the benefits of altruism increase the fitness of the group. So the evolution of altruism can be seen as exporting fitness from the lower to the higher level. I certainly do not think or mean to imply that fitness is a conserved quantity and that if it is decreased somewhere it has to be gained somewhere else.
I have not read the paper yet, but discussed some of this with Pierrick Bourrat when I met him last summer. As to points #4 and #5 from Rick above, I do think that some people have been confused about whether the idea that germ cells “have zero fitness” refers to cells in the colony or out of it. I agree with Rick that the former idea is clearly false, so it makes sense to me to give the author the benefit of the doubt and go with the more reasonable interpretation (zero fitness refers to the would-be cell-level fitness of isolated germ cells) even if the wording in some past papers may not always be one hundred percent unambiguous.
ETIs are–at core–about entities coming to have fitness when they didn’t before. Often as a level gains fitness, the lower level (its components) loses the ability to evolve by NS (i.e. loses “fitness” in some sense). I don’t necessarily like calling this observation “transfer or export of fitness”.. perhaps I am being too literal-minded, but that phrase implies a conserved quantity to me. And–as Rick said above–it is clear that that is not what was meant by the phrase. I think that Rick and colleague’s viability-fecundity “transfer of fitness” models are an interesting abstraction and important contribution. They show how the higher level can emerge as an independent, evolving entity concurrently with the lower level losing the ability to function as an independent evolutionary unit. And they show how this process can be related to fundamental trade-offs and constraints. Of course, there are many devils in the details which come up when you work on extending the framework in various directions 🙂 I look forward to reading the Bourrat paper, as I doubt that the entire disagreement can be chalked up to an ungenerous reading of what “fitness going to nil” means.
–Deborah Shelton (former grad student of Rick Michod)