The human genome (like all mammalian genomes) is loaded with sequences that don't perform any known function. And many of these sequences are junk. And it's not just mammals -- many animal genomes are loaded with junk, as are those of other eukaryotes. That's not to say that some of the sequences of unknown function do, in fact, have a function that we have yet to identify. However, much of the junk comes from the remnants of transposable element (parasitic sequences that hop around the genome).
That's not the impression you get from the introduction to a review of transposable elements (TEs) published in the most recent issue of Cell (doi:10.1016/j.cell.2008.09.022):
Following the discovery of transposons in plants and bacteria, the presence of mobile DNA in eukaryotic species gained widespread acceptance. However, the concept of "controlling elements" gave way to disparaging terms such as selfish DNA and "junk DNA." Nevertheless, the notion of transposable elements as merely molecular parasites, benign at best and powerful mutagens at worst, that hijack cellular mechanisms for their own selfish propagation, seemed incomplete to some biologists. Given that evolution tends to dispose of that which is useless and harmful for a species, it was curious that the genome should be cluttered with so much "junk." Now we understand that genomes have coevolved with their transposable elements, devising strategies to prevent them from running amok while coopting function from their presence. Repetitive DNA, and retrotransposons in particular, can drive genome evolution and alter gene expression. Evolution has been adept at turning some "junk" into treasure.
I don't see how the hypothesis that transposons are selfish genetic elements conflicts with the observation that host genomes have coevolved with these parasites. That is, in fact, what we expect to happen between hosts and parasites -- red queen and all.
As for cooption, that probably applies to a minority of TEs in eukaryotic genomes. The vast majority of them are just there, incapable of being removed by natural selection. Why? Because selection can only efficiently remove features that are deleterious enough to overcome the effects of genetic drift (the random change in allele frequencies due to sampling error).
Are TEs just hangers on? No, they can dramatically change the architecture of a genome. However, that does not meaning they are being exapted. It just means that they're there, and that's what they do. They can lead to the deletion of functional elements. They can rearrange chromosomes. They can even be used to control the expression of nearby genes. That last one is an example of exaptation. The others are merely things that happen as the result of TEs. They can be deleterious, neutral, or beneficial.
But TEs aren't kept around because they have the potential to do lots of stuff to the host genome. That's not how natural selection works -- it can't look into the future. The TEs are kept around because the deleterious effects of those mutations (rearrangements, insertions, deletions) aren't bad enough to cause the TEs to be purged from the genomes in which they reside.
Goodier and Kazazian. 2008. Retrotransposons Revisited: The Restraint and Rehabilitation of Parasites. Cell 135: 23-35 doi:10.1016/j.cell.2008.09.022
Excellent post! That's always been a pet peeve of mine from back when I was working on conjugative plasmids and transposons. Students were always buying into the "selected for potential future use" shtick, when that's clearly bollocks.
The argument advanced in this paper seems to be logically flawed and inconsistent with what we know about evolution. You see it, I see it, hundreds of other scientists see it.
So, how does a paper like this get published in a journal like Cell? Is the peer review process broken? Is science in trouble?
It seems worth noting that 'delayed gratification' (on timescales beyond the individual lifespan) is not a priori impossible for evolution; note maynard smith's 'twofold cost of sex.'
My point being, there is a valid a posteriosi issue, potentially rather complex, as to whether benefits of TEs could outweigh costs. Certainly the drift idea is a hell of a lot simpler and probably more important and plausible. I lack access to the paper, which i bet gives values for the fitness costs of TEs, but i doubt they are that much higher than those of mutations generated by the host proper. Indeed, ill bet the two phenomena are intertwined: assuming that the law of diminishing returns applies to both dna replication fidelity and anti-TE immunity (if there is any), then an organism probably would probably want to accept roughly comparable fitness burdens from TEs and replication error. In other words, say you are suffering x fitness cost from replication error, and another x from TEs trashing your exons, for a total mutational fitness hit of 2x. and say anti-TE Immunity is already about as good as it can possibly be. In that case, you arent going to want to spend 15% of your energy budget getting the burden of polymerase infidelity down to 0.1x. After all, that would only cut your total mutational fitness burden in half, from 2x to 1.1x. Whereas in a world free of TEs such an exchange could be attractive, because youd them be going down tenfold in total burden. You can also argue this whole thing vice versa.
I still have a hard time, though, understanding why mammalian replication error doesnt go arbitrarily low: it only needs to do so in the germ line, not in the soma, so it is hard to see how the energy costs could be that severe. I guess sperm is cheap but not dirt cheap. (the female side can be left out of account he most replication errors come from the make as k army suggests.)
Pardon me, thats 'male as j crow suggests.'
I lack access to the paper, which i bet gives values for the fitness costs of TEs
The paper is hardly that sophisticated.
Hum, unless there is some substance to above species selection, where just like sex/asex, species that happen to set relatively TE-free genomes may open the road to a more probable extinction than those who don't (for e.g. if the the replication cost is outweighted by the occasional beneficial recruiptment as expression modifier filling a need for a quick adaptation)(that seems to be the argument from the paper, and the abstract doesn't seem that messy). I don't think the observed genome pattern would fit the expectations, but the hypothesis is theoretically plausible and the view isn't too heretical (I don't buy it but I won't shit condescendingly on it neither).
Furthermore, one can think of a decreasing cost of hosting TEs, that is, when a lot of the genome is already made of junk ghost-TEs, then the probability of a translocation event to be deleterious is diminished (though this is not an argument in support of the expressed view).
Btw, don't you think your use of the "exaptive" terminology is way more adaptive than you wish to admit? (exaptations are adaptations and part of the adaptationist myth...)(just kidding)
So, how does a paper like this get published in a journal like Cell? Is the peer review process broken? Is science in trouble?
Or would it be that science is still a place where debate is tolerated and where dogma can still be challenged? A paper does not equal the Big Truth(TM)... Thanks Vishnu L.M. knows what evolution is...
I dont know the first thing about sex, but it does seem like species selection must be invoked to explain the well known phylogenetic pattern: many obligate-asexual eukaryote taxa, but almost no ramified lineages. (To that extent, its a mystery that anyone could deny that species selection exists at all.) But of course species selection is probably not the only form of selection thats impotent on a one-generation timescale yet active on a longer one. If one accepts Maynard Smiths twofold cost of sex, it follows that for almost all ecologic situations, robust benefits of sex must arise on about a 100-generation timescale at most - otherwise any asexual mutant would go to fixation over that time. And yet it is very hard to see how sex could even begin to overcome the twofold cost on a 1-to-2 generation timescale. The red queen view or the kondrashov view must be right (or both, as hamilton believed) - or something like that. So even though evo is not teleological it can sometimes LOOK a little teleological. It may not be common for longer-term selection to, in this way, be powerful enough to 'override' momentary selection, but it probably does happen.
Gene conversion is the cohesive force allowing species to exist.
What about selection?
repetitive DNA is an evolutionary device to catalyze formation of new genes by suppressing gene conversion.
But transposons enjoy meiotic drive, effectively. It's not very convincing to argue for some other mechanism of transposon maintenance, without even addressing their meiotic drive.
Gene conversion acts to homogenize the DNA sequences within a species and by doing so forms the cohesive force holding together the gene pool of a species. Gene conversion is by far the highest frequency event that changes the sequence of a DNA molecule. This frequency is about 1% per bp per generation, several orders of magnitude more frequent than mutation. Even beneficial mutations would be removed from the gene pool if there were no way to block gene conversion. Therefore, evolution requires the presence of transposons to create non-homologies between otherwise similar sequences, thus freeing them to diverge into new genes. In this sense transposons may be viewed as isolating mechanisms, creating boundaries that allow evolution of new genes.
This theory explains the abundance of interspersed repeats eukaryotes genomes- they must be numerous enough to provide a global mechanism with a chance of providing any gene a chance to become isolated from gene conversion and adapt beneficical mutations. It also explains why each mammalian order independently evolved its own version of the SINE transposons during the mammalian radiation and why the dispersal of the SINE elements coincided with the burst of mammalian evolution 50-60 MYR ago.
Well, thank you... those are quite interesting things. I must admit, before encountering you I was aware of homologous recombination, but not really too cognizant that it resulted in gene conversion, nor aware that it is such an extremely common event. I see your basic point: since gene conversion is random and has a high frequency near 1% /bp /generation, it of course imperils rare alleles.
If a new allele had a selection coefficient much above 1%, then this sort of drift should pose little danger to it, I think. But it's obvious that taxa could suffer and potentially vanish, courtesy of interspecific competition or whatever, if they were much behind other taxa in their ability to peserve and fix new advantageous alleles that have selection coefficients below 1%.
It does seem like the advantage you propose is probably realized only after a long delay of many, many generations - thousands? Meanwhile, transposon-free mutants (or mutants with a much lower burden of transposons), assuming such a mutant invades the situation, are rallying. Exponentially so, in the early part of the curve. The base in that exponential expression is of course one plus the cost of bearing the transposons, which is probably rather near zero. So it is pretty hard to make a gross estimate of whether this seems plausible. Maybe I'm an idiot but I just googled for 15 minutes trying to find the rate of nonsense insertions in the genes of complex animals due to transposons, and couldn't find it! To that, costs from unequal chiasmus should be added, but I guess they are probably lower since an early-embyonic abortion doesn't tax the parents fitness too much. Anyway, this will probably be my last entry, as I am already stretching my knowledge of genetics and especially transposons pretty thin - any further and I'd just be BSing.
OK just one more thing - it is possible to think of other ways of introducing nonhomologies, which lack certain disadvantages of transposons. One would be to just have a polymerase that can somehow recognize noncoding regions and go ahead and rip out the sequence and insert a repetitive sequence or random sequence in its place. But then, of course, the problem is that it might take several cistrons to set this up. Unless the fitness advantage per generation of this regime (vs using transposons) is greater than the total summed rate (over all these cistrons) of mutation to loss of function, then there is not enough selection pressure to maintain the apparatus, and it would decay into pseudogenes. Adding the cistrons in at high coy number doesn't help, since there is zero pressure for redundant copies not to decay. Whereas transposons, on the other hand, look after themselves, and have their own incentives to maintain themselves at a high copy number - reasons which are quite independent of the organismal fitness, and yet this behavior might possibly in the end raise organismal fitness after all.