Micro-vs-macro evolution is a purely methodological distinction

On the internet, the terms macroevolution and microevolution (especially together) are usually used primarily in creationist rhetoric. As such, it is usually best to avoid them, especially when talking to non-scientists. The main mistake creationist perpetuate when thinking about micro-vs-macro evolution, is that the two are somehow different and distinct physical processes. This is simply not the case, they are both just evolution. The scientific distinction between the terms, comes not from the physical world around us, but from how we choose to talk about it. When a biologist says “microevolution” or “macroevolution” they are actually signaling what kind of questions they are interested in asking, or what sort of tools they plan on using.

In verbal and empirical theories, the micro-macro distinction is usually one of timescales. A person in the macroevolutionary paradigm, usually asks questions above the level of individual species, as Evolution 101 writes (emphasis mine):

instead of focusing on an individual beetle species, a macroevolutionary lens might require that we zoom out on the tree of life, to assess the diversity of the entire beetle clade and its position on the tree.

Empirically, macroevolutionary answers to these sort of questions are usually ones that don’t have access to detailed evolutionary histories or direct experiment. Instead, the method tends to be ones that use geology, fossils, and back-inferences from broad differences/similarities of existing species. As such, most macroevolutionary theories tend to be descriptive, instead of predictive. Most of paleontology can be classified under the macroevolutionary paradigm.

If someone explicitly says that they are looking at microevolution then this usually refers to a contrasting methodology that tends to be heavy on direct experimental manipulation. Most importantly, microevolutionists tend to have access to rich and detailed evolutionary histories. It is common to see studies on E. coli, slime molds, or even fruit flies! Of course, most studies are at intermittent levels, and no this isn’t called meso-evolution (except by silly people). If you are not clearly using the macroevolutionary nor the microevolutionary paradigm but still looking at evolution, you would just simply say ‘evolutionary’ without any prefix.

Formal and mathematical models

Mathematical and computational modeling of evolution is a huge field. To start off with some broad strokes, there are two main approaches to evolutionary modeling with their own communities: frequency-independent and frequency-dependent models. Of course, in a real biological setting, the truth lies somewhere in between, but models are idealizations of key principles, and so the two extremes are good to study. They serve as heuristic guides for the development of more accurate or insightful models.

For frequency-independent selection, they key concept is the fitness landscape — a way to map each genotype to a fitness. The population lives in this landscape, agents with higher fitness reproduce more, and the population slowly moves over the lattice of genotypes connected by mutations. If the model tracks the whole population (say as distribution over vertices in the fitness graph) then it would typically be called an evolutionary model (the word microevolutionary is seldom used explicitly in this field).

Under reasonable macroevolutionary assumptions such as very rare mutations and asexual populations, the population will tend to be homogenous and can be modeled as a single point in the fitness landscape with properly chosen selective sweeps to move the point from vertex to vertex (Gillespie, 1983; 1984). It seems like computer scientists prefer such models, with the both the Chaitin (2009) metabiology and Valiant (2009) evolvability models using this paradigm.

As an illustrative example, consider the simple model that Wilf & Ewens (2010; the first author was a noted mathematician, the second — a biologist) use to show that “there is plenty of time for evolution“. Their goal is to calculate the number of generations required to spell out a word like ‘Evolution’. They allow each letter to mutate with certain probability every generation, but stop mutating a letter once it is correct — this is the ‘selective’ feature of the sweep. To spell the word by randomly placing letters (blind search) would require ~5.4 trillion generations, but with selective sweeps the authors calculate that we’d only need about 57 generations on average. Chaitin’s separation between blind search (exponential time), evolution (between quadratic and cubic) , and intelligent design (linear) effectively achieves the same thing, but with an unreachable optimum (so he has unbounded growth in ‘complexity’) and directed mutations that can’t be decoupled from selection.

Among biologists, the most popular concrete approach for frequency-independent selection is Kauffman’s NK model of rugged fitness landscapes (Kauffman & Weinberger, 1989; Kauffman, 1993). Computer simulations tend to favor this sort of model. Wilf & Ewens’ model would be an NK model with K = 0, and fitness of 0 for incorrect letters and 1 for correct letters (to get the shorter walk). Of course, this is an extremely simple fitness landscape, and much more complicated “holey” landscapes are of interest to biologists. But, it isn’t clear how much biologists know about the structure of fitness landscapes or if the underlying assumption of local equilibrium is even reasonable. The recent literature, however, has been to moving towards non-static and frequency-dependent landscapes. As Simon Levin — one of the co-developers of the NK-model (Kauffman & Levin, 1987)– said: we should think of the fitness landscape not as rigid hills-and-valleys, but as a waterbed where the agents’ distribution deforms and creates new and different peaks as the environment co-evolves with the frequency of the agents.

The frequency-dependent selection approach is dominated by evolutionary game theory with questions like the evolution of cooperation. The starting approach is to use replicator dynamics and look at evolutionary stable strategies. However, the field is in general saturated with all kinds of exciting models (both analytic and computational). A very friendly book-length intro is Nowak’s Evolutionary Dynamics and a brief survey aimed at mathematicians can be found in Hofbauer & Sigmund (2003). Frequency-dependent models almost never make the micro-macro distinction. Although you could argue that dynamic models in evolutionary game theory are micro, and static equilibrium concepts like ESS are macro, but I doubt many egtheorists would endorse this view. The dynamics view can be seen as a part of algorithmic game theory, but there is still plenty of progress to be made in bringing insights from theoretical computer science.

Fundamentally, though, both static and frequency-dependent models are just convenient (or tractable) approximations to a real underlying evolutionary dynamics. This should never be forgotten.

Strange example bridging the gaps

Finally, it is important to stress that the macro- and micro-evolutionary paradigms are not necessarily exclusive and do not have to correspond to a difference in timescales! This is best done with an example of a respected theoretical study that mixes everything together.

In the human immune system — when exposed to an antigen — B-cells produce antibodies. If it is your first exposure to the antigen then the antibodies produced will probably have very low binding affinity. However, after some exposure time, your B cells will start to produce antibodies with much higher affinities for the antigen and thus you will be able to better fight off the disease. The cool part, is that the antigen produced is tune via an evolutionary process!

There is differential survival, with only antibodies with the highest affinity being able to survive. Variability is introduced by a very high mutation rate in the complementarity determining region (CDR). (Tonegawa, 1983). The length of this evolutionary process is very short, typically a local equilibrium is found after only 6-8 nucleotide changes in CDR (Crews et al., 1981; Tonegawa, 1983; Clark et al., 1985), so you need only a few point mutations to quickly develop a drastically better tuned antibody.

With a protein sequence on N sites, we say that evolution is fast (and we have a sudden leap) if after our fitness landscape changes, we can get to a new local equilibrium in a number of generations that scales with . Kauffman & Weinberger (1989) showed how the NK model can be used to study this affinity maturation, and showed that to achieve this sudden leap we need high epistasis and low correlations between pointwise mutants. In particular, their model suggests that typical epistasis in the CDR is on the order of 40 proteins (out of the total 112 proteins in the CDR).

Kauffman & Weinberger (1989) developed a macroevolutionary mathematical model because they used Gillespie’s trick to replace a population by a typical individual by abstracting away from the underlying microevolutionary calculation of fixation probabilities. However, their model was studying evolutionary dynamics within the human immune system (so timescales of days to weeks) and was tuned by parameters gathered by empirical microevolutionary studies that tracked individual nucleotide changes (Crews et al., 1981; Tonegawa, 1983; Clark et al., 1985). Lastly, the study results can be used to inform a question typical of verbal macroevolutionary theory: Are there any examples of sudden leaps in evolution?.

As such, the above study used a formal macroevolutionary model, informed by empirical microevolutionary work, to help us understand a question typical of verbal macroevolution while looking at a physical process that operated on the incredibly short timescale of days to weeks. No wonder people are so confused by the micro-macro “divide”!

This post is based on three of my answers ([1], [2], [3]) on the Biology StackExchange.

References

Chaitin, G. (2009). Evolution of Mutating Software. EATCS Bulletin, 97, 157-164

Clark, S.H., Huppi, K., Ruezinsky, D., Staudt, L., Gerhard, W., & Weigert, M. (1985). Inter- and intraclonal diversity in the antibody response to influenza hemagglutin. J. Exp. Med. 161, 687.

Crews, S., Griffin, J., Huang, H., Calame, K., & Hood, L. (1981). A single V gene segment encodes the immune response to phosphorylcholine: somatic mutation is correlated with the class of the antibody. Cell 25, 59.

Gillespie, J.H. (1983). A simple stochastic gene substitution model. Theor. Pop. Biol. 23, 202.

Gillespie, J.H. (1984). Molecular evolution over the mutational landscape. Evolution 38, 1116.

Hofbauer, J., & Sigmund, K. (2003). Evolutionary game dynamics. Bulletin of the American Mathematical Society, 40(4), 479-519.

Kauffman, S., & Levin, S. (1987). Towards a general theory of adaptive walks on rugged landscapes. Journal of Theoretical Biology, 128(1): 11-45.

Kauffman, S. (1993). The origins of order: Self organization and selection in evolution. Oxford University Press, USA.

Kauffman SA, & Weinberger ED (1989). The NK model of rugged fitness landscapes and its application to maturation of the immune response. Journal of theoretical biology, 141 (2), 211-45 PMID: 2632988

Nowak, M. A. (2006). Evolutionary dynamics: exploring the equations of life. Harvard University Press.

Tonegawa, S. (1983). Somatic generation of antibody diversity. Nature 302, 575.

Valiant, L.G. (2009) Evolvability. Journal of the ACM 56(1): 3.

Wilf, H.S., & Ewens, W.J. (2010). There’s plenty of time for evolution. Proceedings of the National Academy of Sciences, 107(52): 22454-22456.

Advertisement