Koonin suggests that group II introns entered an archaeal host via the α-proteobacterial endosymbiont. The consequence is an explosion in intron numbers, compensatory evolution of a number of mechanisms of transcript quality control (the nucleus [2], nonsense-mediated decay (NMD), ubiquitinylation), and side-effects such as the evolution of linear chromosomes with telomeres and telomerase.
The important point here is that group II introns, upon arrival in the archaeal host have, 'apparently, gone berserk within the host cell' (see also [2]). Like the mitochondrial seed hypothesis [4, 5], upon which Koonin's model is based, the argument is based on two premises. First, that group II introns are the ancestors of spliceosomal introns and the spliceosome, and second, that introns would proliferate in the genome of the host (under both models this can apply both to group II intron spread and later spliceosomal intron spread after the evolution of the spliceosome). Regarding the first point, there is plenty of circumstantial evidence that can be interpreted in favour of a common ancestry. Alternative interpretations are nevertheless possible, and I discuss this briefly in the referee report that accompanies Koonin [1]. I will not reiterate that point here; the current discussion is best served by assuming that group II introns did indeed enter the eukaryote lineage via the mitochondrial ancestor, later evolving into the spliceosome and spliceosomal introns.
The second premise is that group II introns would have proliferated in the host genome. Both papers [1, 2] argue that introns would have been selectively disadvantageous; below I describe why this is important, but here is the reasoning given for the selective disadvantage of intron insertion. If host transcription and translation are coupled, and excision of newly inserted introns is the slowest step in production of a functional message, the ribosome runs the risk of reading through into the intron before it has been spliced out. This can therefore result in formation of aberrant proteins.
A further key feature of this model is that the source of group II introns invading any host genome is the mitochondrion; spread of group II introns between individuals (i.e. intergenomic spread) is not invoked. The model permits any host genome within the population to receive group II introns from their mitochondria (endosymbiont gene transfer), so transfer is iterative and ongoing. However, the main source of intron proliferation is spread of introns already integrated into the host genome (intragenomic spread). According to the model, the small effective population size of this new 'prekaryotic chimera' means it is not possible to eliminate group II intron proliferation by purifying selection; drift will dominate, and individuals in which intron proliferation is occurring will be fixed, in spite of their reduced fitness.
That genetic elements can proliferate at the expense of the fitness of the host in which they are found is now widely accepted. However, this is only predicted to occur under certain circumstances. A model published by Hickey [6], illustrates under which circumstances this will occur, and furthermore shows that spread is not predicted under the conditions invoked in the Koonin/Martin model. Moreover, a documented case of recent group II intron invasion into archaea fits with Hickey's model, not the Koonin/Martin model. Finally, I will point out that the mitochondrial seed hypothesis (upon which the Koonin/Martin model is based, but which in contrast allows the host to be a sexual eukaryote) is compatible with the predicted behaviour of selfish genetic elements described by Hickey.
What follows is a brief summary of the relevant aspects of Hickey's model, which I will relate back to the Koonin/Martin model. The first point concerns the nature of intragenomic spread. Hickey defines the average copy number per cell (f) of a transposable element in a population as:
where a is the average copy number of the element per genome, counting only those genomes with at least one copy; b is the number of genomes containing one or more copies; N is the population size (number of genomes).
f can increase via an increase in either a or b. If f increases due to an increase in a, this is intragenomic spread; the element is increasing in frequency only because those genomes containing copies now contain even higher numbers of the element. In this case, which describes an asexual population, elements do not spread to new genomes. Consequently, those individuals with harmful elements (as per Koonin/Martin) will be at a selective disadvantage relative to those without, and element-carrying individuals are not predicted to spread within the population.
The parameter b is not irrelevant to the Koonin/Martin model. While the model does not include intergenomic spread between individuals, endosymbiont gene transfer of group II introns enables b to increase in the absence of intergenomic spread. Therefore, it is possible for group II introns to be fixed in the short term, by drift or by high rates of endosymbiont gene transfer.
This is unsurprising in that we know that group II introns exist in both bacterial and archaeal lineages, and are able to spread via horizontal gene transfer. However, what is not observed is massive proliferation within these asexual lineages, even in the presence of horizontal gene transfer.
A case in point is the recent discovery of group II introns in two species of archaea, Methanosarcina acetovorans and Methanosarcina mazei [7, 8]; in both cases, group II introns have become established as the result of horizontal gene transfer from bacteria. This provides an ideal analogue to the 'primitive prekaryote' host genome, a garden-variety archaeon. Assuming rates of gene transfer and proliferation are similar to Methanosarcina spp. and that these species are sufficiently 'garden-variety', one would expect a similar rate of proliferation as in the Koonin/Martin model. However, none of the group II introns are inserted in archaeal open-reading frames, hence do not result in ribosome readthrough. Instead, these genes have a tendency to insert into the reverse transcriptase genes encoded by other group II introns, generating nested introns. I have no idea as to the effective population size of these two archaea, but that these elements have neither gone berserk (4 in M. mazei, 21 in M. acetovorans), nor inserted into archaeal protein-coding genes does not serve to strengthen the model presented by Koonin.
Rather, this is what one may expect for selfish elements in asexual lineages, even with horizontal gene transfer. Upon insertion, such elements will end up in linkage disequilibrium with the other genes in the genome (there is no meiotic recombination and no outbreeding). Consequently, in the long term, even if drift fixes the presence of an element that imparts a cost to the host, elements that evolve to be less harmful will be at a selective advantage; the fitness of the host and the element are identical. Thus, rather than spreading wildly, the result will be more cautious mechanisms of maintenance or spread. That the group II introns from Methanosarcina spp. have not inserted into coding regions is probably consistent with this. Likewise, assuming some cost associated with element presence, individuals that completely lose the elements will be at an advantage [6, 9].
If an element does proliferate wildly in an asexual population, the cost to all individuals may become so high as to lead to population extinction. Indeed, it has recently been argued that element overload is one probable cause of extinction of obligately asexual lineages that have evolved from sexual lineages [10]. Survival would entail loss of those elements with the highest cost (i.e. that proliferate greatly in number). Consistent with this is the observation that retrotransposons appear to have been completely lost from Bdelloid rotifers, which evolved from sexual ancestors some 80 million years ago [11]. The same picture is seen for Giardia lamblia, which is not known to be sexual [12], and which, incidentally, is very intron-poor, with only three introns identified to date [13, 14].
The bottom line is that even if endosymbiont transfer can ensure an increase in the proportion of individuals carrying one or more genomic copies of an element, there will still be a selective advantage for attenuation, even if complete loss does not occur.
Assuming introns-late, these elements have must have proliferated at some stage during early eukaryote evolution. Again, Hickey's paper explains under which circumstances this can occur, even when the element has a deleterious effect on the fitness of the host. In an attempt to avoid reproducing the entire paper here, the key point is that this will happen in a diploid, sexual, outbreeding population.
Individuals carrying the element, despite having a lower fitness relative to element-free individuals, will nevertheless dominate the population. Considering just a single locus, when the frequency of the element in the population is close to zero, the element will double in frequency within the population because most zygotes receive a single copy (they begin as heterozygotes) but pass on twice as many copies per gamete (transposition makes them homozygous for the element). If they are only half as fit as element-free individuals (i.e. they pass on only half as many gametes), the number of copies of the element that are passed on is the same for a normal nontransposable gene. Consequently, provided the cost to the host is < 0.5, and with high efficiency of transposition, the element will spread. Hickey's model illustrates that, as the element increases in the population, selection against individuals can rise significantly above 0.5, yet this will not impede further spread.
What this is telling us then is that meiotic sex and outbreeding are prerequisites for introns to have proliferated massively under introns-late. For Koonin's theory to work over a long enough time scale for several complex systems of quality control to emerge, I would argue that he should at least have invoked the emergence of facultative meiotic sex. The problem with this is that sex, with its two-fold reproductive cost, must be invoked under a scenario where there is a population of primed selfish elements 'waiting' to spread. While the level of sexuality can be increased in a facultatively sexual population, this is not so for an asexual population [9]. As the ever-present difficulty with models for the origin of sex is accounting for the short-term selective advantage for sex, this would represent a rather backwards way of approaching the problem!