Things like HGT, duplications, recombination, etc create a qualitatively different process of genetic change compared to something like point mutations. In the case of HGT and recombination, they involve a population-dependent filter of genetic change (in that material that killed a host won't be part of the pool for things to be swapped in), and in all such cases they're a mechanism by which robustness can be introduced (via introduction of copies). Also, because they have a scale that is much larger than a point mutation, there are entropic considerations which mean that their effects at introducing drift don't scale as badly for larger sequences as point mutations do. There are also internal evolutionary processes, such as the evolution of transposon copy number, that may be very important to the big picture.

That said, there are systems where point mutations clearly play a very central role as the dominant evolutionary mechanism. My first thought is in the adaptive immune system, where after a large-scale process (VDJ recombination) produces initial naive B-cell sequences, point mutation rates are increased by about a factor of 1000 (somatic hypermutation). The result is that B-cell sequences are refined over the course of ~20 generations to improve affinity with the antigen - this is usually something like 2-5 alterations. The problem seems like it has very different demands than, say, viral escape from the immune system since in the one case a specific structure must be pinned down but in the other case, the virus just needs to change its surface wildly while remaining functional (so large-scale rearrangements are a very powerful tool for this).

On long timescale evolutionary problems I don't think I can readily say how important point mutation truly is (at least as a constructive force). My guess is that its still important for local refinements, but that the big changes are being strongly driven by the other mechanisms. The test would be to look at how closely point mutation rates across species are held to the error threshold - if everything is operating right at the error threshold it strongly suggests that point mutations are playing an important role and are being maintained by selection pressures.

Mutational hotspots are another thing to look at for this question, since they would indicate areas where point mutation is either coincidentally higher or more interestingly areas where there may be some selection pressure driving point mutation rates up.

Thanks for the reply, Dr. Guttenberg. Are you a direct descendent of the man who is credited with invention of the printing press?

My "devil in details" question was rhetorical, as you probably know. I wanted to open a discussion, and the format here in ResearchGate is all about "questions" so I phrased it as a question.

Whether long or short timescale, or strong or weak selection, we can't say whether one mechanism is "more improtant" than another, unless we specify exactly which gene and organism and selection pressure we are discussing. For example in HIV evolution, the escape from an antiretroviral drug is the same for all viruses exposed to that drug. We repeatedly observe for example the M184V change in reverse transcriptase in thousand of different patients exposed to the same drug. Convergent evolution in a very basic example. Escape from a particular human host immune response however is quite different. Each human produces dozens of selection pressures on the virus, from 10 or more CTL responses to dozens of immunoglobulin responses. And for each, the viruses in the infected individual has several different escape routes.. In viruses, all of this occurs via point mutations, small insertions and deletions, and recombination between highly similar viruses. The point is, that even within the same gene in the same virus in the same host, the similar but different selection pressures put on the virus by drug or host immune responses, can result in different genetic escape routes.

Instead of arguing about which type of evolution is "most important" it is interesting to discuss how ALL types of genomic change can have differing levels of contributions under different circumstances. And discussing how it is critical to consider differing levels of genomic change when comparing the fossil record and morphological changes in organisms to the DNA data and phylogenetic classifications of organisms.

In the context of all eukaryotes, it makes a lot of sense to look at humans as 98% or 99% identical to chimpanzees and gorillas. In the context of human migration out of Africa and how we may have interbred with Neanderthals, it makes more sense to look at chimpanzees as a distant outgroup, and focus on what changes have been most important, perhaps gene duplications and so on, rather than the point mutations.

If we attempt to look at genome rearrangements and gene duplication between primates, fish, plants, insects, etc. we are lost.

My point for opening this discussion, is that I see reviewers of papers most often asking the authors essentially to "use a bigger computer", to analyze their data set (complete mitochondrial genomes of amphibians for example) . The reviewers will look at a neighbor-joining analysis and say "use maximum likelihood" or they will look at a maximum likelihood analysis and say "use Bayesian methods". They almost never say "check to see if the data is truly appropriate". DNA distances, by any method of analysis, are only good over some range of values. Outside that range, other data may be needed.

I came across a paper today with a real nice analysis of the evolution of the complete mitochondrial genomes of snakes. It illustrates some differences in the rate of evolution of different genes in the same genomes, and also differences in the rate of the same gene at different times in the history of snake evolution.

Comparative mitochondrial genomics of snakes: extraordinary substitution rate dynamics and functionality of the duplicate control region.

Jiang ZJ, Castoe TA, Austin CC, Burbrink FT, Herron MD, McGuire JA, Parkinson CL, Pollock DD.

BMC Evol Biol. 2007 Jul 26;7:123.

PMID: 17655768

http://www.ncbi.nlm.nih.gov/pubmed/17655768?report=docsum

This is a very nice example, of not just looking at a phylogenetic tree and saying "the snakes seem to evolve faster" as many people have noted in the past, but digging deeper and asking exactly what is the pattern? Is the entire mitochondrial genome evolving faster, or is it just some genes?

I've looked a little bit at the variable mutation rate sort of picture in terms of theoretical models at least. Its clearly important as far as how efficiently evolution can proceed to have innate differences in local rates of genetic change. There's unfortunately an effect which masks this, which is that places that are sensitive to lethal mutations will show a reduced degree of variation just because those attempted mutations don't lead to surviving organisms. Do you know a way to distinguish between these two effects in e.g. sequence data?

What I mean by the efficiency of evolution by the way is basically the question of, how fast can you move a small section of the genome if it has a larger section that has to remain unchanged? There are some theoretical bounds that can be established based on entropy arguments and the error threshold. Consider a genome containing L 'important' sites, by which I mean locations where a point mutation or other genetic change might be lethal. This forces the per-base point mutation rate to scale like log(sigma)/L where sigma is the strength of selection. So if you have a genome that is very long compared to the segments that are of evolutionary importance at a given moment in the species' history then that species will be unable to respond quickly unless it develops a variable rate of mutation that focuses things into the areas that are immediately important. This same argument should more or less apply to other genetic mechanisms such as recombination and linkage disequilibrium and suggests that this localized tuning is really necessary for things with long genomes to keep up.

One question is how these hotspots themselves evolve to be 'hot'. If its based on the local sequence, then its inherently a self-destructive process as the high degree of mutation in the area will tend to erase the very sub-sequences that get targetted for having a high degree of mutation. If its tied to the larger-scale structure (e.g. location on the chromosome compared to histones or something) then that could have a longer persistence. On top of that there's the a meta-evolutionary character of having to evolve a mutation rate that might lead to some interesting constraints.

I briefly investigated a computer model where we had the mutation rate locally increase after a point mutation occurred (one could think of this as the local breakdown of homologous recombination, since the rate of homologous recombination is exponentially dependent on the degree of mismatch). The result was that for long genomes you could achieve significant speedups in how quickly the system would find an evolutionary optimum, but only if there were spatial correlations in the genome that were somehow mirrored in the structure of the fitness function (that is to say, you had to have larger scale 'regions' that each did a coherent thing, rather than each base having a random effect). The difference here is, the system didn't need to know a-priori that some regions were more important than others in order to figure out how to assign the right hotspots. Places with lots of conserved bases naturally tended to settle down to a low mutation rate, whereas neutral places and places that benefited from high rates of local mutation would naturally develop a high mutation rate that would persist due to the positive feedback effect. The main problem we had was pinning down a way to test if this sort of mechanism was actually at work in real systems.

There's a very tangentially related question I'm curious about. Namely, what did the total 'computational capacity' of the world's biosphere look like at various points in the history of life. That is to say, you have huge population counts of bacteria and the like and much smaller population counts of mammals, reptiles, etc, but at the same time much of what is going on in many of these organisms is neutral evolution. Has the evolutionary computational budget changed over time, or is it still basically dominated by unicellulars?

(aside, I sadly have no idea if I'm descended from that particular Gutenberg)

I suspect you are largely confusing rapid evolution with rapid mutation. In HIV-1 for example, the envelope gene evolves roughly twice as fast as the polymerase gene. But if you look more closely you find that this is because of more negative selection on the polymerase protein and more positive selection on the envelope protein, and not evidently due to different mutation rates. Recombination can also be important, because it can couple a less mutated pol gene to a more mutated envelope gene. It is quite possible that there are some hot spots for mutation in the HIV genome, but proving that they are not just hot spots for selection is very difficult.

The exact pattern of mutations vs pattern of selection is also a bit difficult to untangle. We observe far more G to A changes than A to G changes for example in the viral RNA genome (they would be C to T changes on the other strand which only exists as DNA). But the genome overall is A rich and C poor and has a relatively normal amount of G. The T-cell leukemia viruses, which replicate in the same human cells (T-cells), and are also retroviruses, have a different skew in base composition.

From simple organisms like viruses, to complex organisms like bacteria (with and without phages and plasmids to facilitate horizontal gene transfer) there are thousands of experimental systems where various aspects of mutation and evolution have been or can be tested. Theoretical models and computer simulations with those models can be critical for teasing out the various components which are always entangled in the real systems.