You might think that all of the normal cells in your body are cooperating with each other. Not counting microbial infections, the occasional carcinogenic defection from the target morphology, mutant cells in your normal brain, or your baby’s cells that colonized you, it might seem that the vast majority of your own cells should be incentivized to cooperate with each other because they share the same genome.
It turns out that this is not the case, and despite their identical genomes, your body is home to cells, tissues, and organs that have not forgotten their roots as individual agents that compete for resources – both metabolic and informational. But I think there’s more to this. Our computational work suggests that evolution sets up artificial scarcity within the body, in order to achieve robust embryogenesis, which may have implications for other large-scale systems composed of agential parts (discussed at the very end of this post).
The concept of competition between organs during development was first outlined in detail in a classic but little-known work by the great developmental biologist Wilhelm Roux called “The Struggle of the Parts” (now finally translated from the German, by David Haig and Richard Bondi).
In this paper, Rick Gawne, Ken McKenna, and I review the data – much of it only known to farmers and livestock breeders, on what happens when you release the competition. One thing that happens is compensatory growth – when certain body parts are removed from the equation, others over-grow – both during embryogenesis and in maturation to adulthood, revealing how their growth was suppressed by the active presence of other tissues. What we know can be summarized as follows:
Basically, both cooperative and competitive interactions between body features (“characters” in the terminology of the field) give rise to the standard anatomy of a given species because limiting resources help coordinate relative growth rates. The cells and tissues are competing for molecules that are used as fuel, but also for informational molecules used as growth signals. Why would this be a useful architecture for an organism? In the case of metabolism, perhaps there’s just not enough food energy to be present in unlimited quantities. But informational molecules are normally used in very tiny concentrations in tissues, and in principle could be made in whatever quantity is needed – why limit those?
Peter Smiley (an undergraduate student in my lab at the time the work was done) and I decided to find out. Does evolution like competition within bodies? We set up an evolutionary computer simulation, of embryonic development driven by cells that have different possible behaviors – this is an example of the digital embryogeny field. In this paper, we describe what happens when evolution has a choice of using competitive interactions or not. In our simulation, embryos were rewarded for growing to specific shapes. Each cell had a genome which would encode policies of how and when it would divide, subject to the availability of permissive signaling molecules. There were a number of reservoirs of molecules the genes could refer to for this permission, some of which were present in limiting (finite) quantities and some of which were unlimited. If a cell had a genome that referred to the infinite reservoir, they could in effect divide whenever they wanted to (assuming the other rules, such as having space to divide into, allowed it). If a cell’s genome referred to a finite reservoir, it could divide as long as that reservoir wasn’t empty (and many other cells could draw from that same reservoir). Cells did not otherwise communicate in any way.
What we found is that evolution effectively selected for genomes relying on finite reservoirs – it specifically forced the cells to compete for a limiting resource, when other, unlimited, resources were present. The analyses below show how the usage of finite reservoirs help the fitness of the virtual embryos.
The infinite reservoir-using embryos routinely had lower fitness than ones who used finite ones (or both). We also found that if we take embryos evolved to use finite reservoirs, and we suddenly make their reservoirs infinite, their fitness drops precipitously:
which shows that they were actively relying on the fact that their reservoirs were limiting, to achieve normal morphogenesis. We also found that the use of finite reservoirs resulted in more consistent development, even though our selection scheme did not reward for that specifically – this consistency showed up “for free”, emergent from the dynamics of competition ensured by limited reservoirs. See here how the variability (vertical axis is the standard deviation over repeats of each embryo’s development) reduces when using finite reservoirs:
So, why are finite reservoirs so useful? How do successful developmental strategies use them? One way to do it is to let evolution choose the size of the limited reservoirs and see how many different sizes it uses. By analyzing different successful embryos that evolved, we found that: 1) one of the strategies is to use finite reservoirs to stop growth at the right time; but, 2) there is more than one strategy used; as often happens, evolution finds many ways to the same goal (which is, by the way, a definition of intelligence by William James). These ranged from harmonic growth with arms that leave only one stem cell exposed (not land-locked by neighbors), at any one time (and thus limiting growth in ways not relying on reservoirs being emptied), to other strategies with a major reliance on finite reservoirs; 3) many of these successful embryos did not use the simple strategy of using 1 reservoir molecule for each cell to be produced (i.e., some successful genomes had reservoirs far smaller than the total # of cells).
One basic point is that finite reservoirs are a way to coordinate – a kind of global register, or stigmergic scratchpad, which allows cell A to be affected by what cell B does, even at large distances, without explicit signaling mechanisms. Limiting resources are a useful coordination mechanism.
We also studied the evolutionary course of specific genes, and the developmental biology of successful embryonic strategies, defining a kind of digital genomics (asking which rules are present in the genomes) vs. transcriptomics (asking which rules are actually activated, in an embryo with a given genome). We found that by the end of evolution, only about 5% of the genome was actually used, to achieve the target anatomy. I’ve not been able to find the corresponding number for biological organisms – what % of the genome do we actually use, in development and maintenance? I’m not sure it’s known.
The basic strategy of using competition to produce fit embryos required very few genes – a small percentage were responsible for most of the jumps in fitness, and was often focused on the finite-reservoir alleles. Moreover, most work during development was done by the physiological/transcriptional level (how the existing genes were used by the cellular automata rules), not the genetic level (the prevalence of finite-reservoir calls in the genomes). For example, once the genome got to 20% use of finite reservoir genes, the actual usage of these by the embryo (how often the embryos activated those genes) had already reached the optimal value of 60%. Development was clearly able to use the genomic hardware in ways optimal for survival.
We could see how the evolutionary dynamics reliably produce organisms that preferentially pit their parts against each other, generating conflict, competition, and artificial scarcity of resources for their components in order to meet the goals of the higher level of organization. Even though evolution could give cells all the resources they need, with no efficiency pressure, instead it preferred the strategy of self-induced limitations of resources and competition. This is important, given the common assumption that evolution can be expected to drive within-agent cooperation: apparently, a system’s goals can sometimes be better satisfied by parts that are in conflict over resources. While we can’t extrapolate this idea carelessly, it’s probably wise to develop caution toward simple schemes that seek to bind active agents into some sort of larger collective. We need to beware of the development of large-scale, hard-to-detect, collective agents which benefit by enforcing unnecessary competition among their members.
Featured image by Midjourney.
Leave a Reply to NiCo Tymmesa Cancel reply