...or how a learned to stop worrying and love evo-devo.
As my mind gets a chance to process some of the stuff I heard and talked about at the meeting I just returned from, I'll post some thoughts that will help me organize my ideas (hopefully better organized than that last sentence). This is the first (of perhaps few, perhaps many) of those (possibly incoherent) ramblings -- interrupted by as few paranthetical remarks as possible.
In this post, I'll try to tie together:
- A talk by Sean Carrol on the evolution of wing pigmentation.
- A talk by Peter Andolfatto on the evolution of Drosophila non-coding DNA.
- A talk by Carlos Bustamante on signatures of natural selection in human genes.
All of the work has been published (follow the links above), so I'm not going to spend much time reviewing the content. Instead, I'll discuss how these different results relate to each other, and how they are helping me formulate a coherent concept of evolution at the DNA level. There is, of course, more below the fold.
At the meeting, I attended a session on cis-regulatory evolution. As one of the presenters noted, everyone began their talk with a nod to King and Wilson's 1975 paper (full text available here if you have access to JSTOR). King and Wilson postulated that differences in protein sequence between humans and chimpanzees could not explain the phenotypic differences between the two species (I have written about this before). Instead, changes in the regulation of when and where the genes encoding those proteins are expressed are responsible for phenotypic evolution. This model spawned the field of evo-devo, which did not get a clever name until some twenty years after King and Wilson's seminal paper.
Sean Carroll is one of the leading researchers bridging the gap between evolutionary and developmental biology. His students' work on the evolution of Drosophila wing pigmentation (reviewed here and here by PZ Myers) revealed that changes in the cis-regulatory elements (CREs) flanking the yellow gene are partially responsible for the gain and loss of wing spots during evolution. The Carroll lab has not unraveled the entire story, but they have shown that changes in the expression of a gene (rather than its protein coding sequence) can lead to novel phenotypes. It would be interesting to find out why this trait is sexual dimorphic (only males have wing spots, so some sex-specific upstream transcription factors are probably involved in determining the phenotype) and what other genes are involved the pigmentation patterning (the transgenic D. melanogaster have dark pigmentation in the anterior-distal portion of their wing, but this pattern is not as crisp as in the species with endogenous pigmentation).
While the Carroll lab is focusing on an individual gene, Peter Andolfatto is taking a whole genome approach to understand the evolution of non-coding DNA. I wrote a summary of Andolfatto's paper for my old blog; go there if you're unfamiliar with this research. Some non-coding sequences are conserved across long evolutionary distances -- these sequences are probably constrained by purifying selection because they are essential for the proper expression of nearby (and maybe not so nearby) coding sequences. Andolfatto has shown that many other non-coding sequences are not conserved because they are under positive selection (and other sequences are turning over at a neutral rate). This inference was made based on the relationship between polymorphism and divergence at these loci. King and Wilson's hypothesis only distinguishes between the evolution of protein coding sequences and regulatory sequences -- they do not explicitly promote a selectionist explanation over a neutral one. Andolfatto's result, however, indicates that if changes to CREs are primarily responsible for phenotypic evolution, that evolution is probably driven by positive selection.
King and Wilson also failed to distinguish between gene regulation in cis and trans, although their thesis that changes in protein coding genes cannot explain morphological evolution seems to imply that they would favor cis-regulatory evolution (if they were familiar with the distinction). One of Sean Carroll's former students, Tricia Wittkopp, showed that changes to CREs could explain more differences in gene expression between Drosophila melanogaster and D. simulans than trans changes. This result seems to trivialize the role of protein coding sequences in the evolution of the phenotype. Sure there are changes in protein coding sequences (Wittkopp did observe evidence of trans-regulatory changes), but the majority of gene expression evolution is due to CREs. Intuitively, this makes sense; CREs only regulate a single gene, whereas transcription factors regulate multiple genes. The expression of a single gene can be modified by a change to its CREs, but a change in a transcription factor could have deleterious effects on the other genes it regulates.
So, what role does that leave for protein coding sequences? Changes in the protein coding sequence of some genes may be partially responsible for the evolution of the human brain (although this result is far from conclusive). Working in the same theoretical framework as Andolfatto, Carlos Bustamante and colleagues compared polymorphism and divergence (from chimps) at both synonymous and non-synonymous sites for over 11,000 human genes. These data were collected by resequencing these genes, not by scoring SNPs, so they should be immune from ascertainment bias. Nine percent of the genes had an excess of a non-synonymous substitutions between humans and chimps, indicating that these sequences were under positive selection. Interestingly, many transcription factors have this signature of positive selection. It looks like trans-regulation does play an important role in evolution.
Does phenotypic evolution occur via cis or trans changes? As with most scientific controversies, the truth lies somewhere in between the caricature of the two camps. Carroll and colleagues were the only group to look directly at morphology, but they only studied a single gene. Wittkopp looked at more genes than Carroll's group, but she was examining gene expression -- a bit less complex than morphology. Wittkopp did show that trans changes contribute to the evolution of gene expression, and Bustamante and colleagues showed that transcription factors are under positive selection. But so are CREs, as Andolfatto demonstrated.
Sean Carroll likes to point out how important CREs are in the evolution of form. Bustamante joked (with Carroll in the room) that his data paint a very different picture, with transcription factors playing an important role. Why do they take such a polarizing stance? Well, controversy fuels excitement. If we all just acknowledged that cis and trans changes are important, we could move on toward understanding how these two regulatory mechanisms work together to contribute to evolutionary change. Carroll's story is incomplete without understanding why the wing spot phenotype is sexually dimorphic -- probably due to some trans regulation. Furthermore, Bustamante and Andolfatto have shown that both transcription factors and CREs contain signatures of positive selection. Wittkopp was the only one to simultaneously examine cis and trans factors, finding evidence that both are evolutionarily important. Let's end the stupid bickering-- it's not doing anyone any good.
It's all just parameter estimation. What percent of variation that we are interested in is cis vs. trans ?
Great post. This might go a bit of topic but I would be very interested in seeing a study looking at different biological functional units and their evolutionary rates and selection pressures. What is the amount of information contained in different types of regulatory regions, chromatin regions, different protein interfaces, protein domains etc, and how do they evolve and what are the selection pressures. Something like a cross between this paper by Sean R Eddy and this one by M Lynch ? :) Not really sure myself what this means.
That Michael Lynch paper is one of my favorites -- it's a great example of the nearly neutral theory at work. But I don't think it has much to do with the gene expression.
The greatest difficulty in this analysis, IMO, comes in developing algorithms for identifying functional, non-protein-coding sequences. Finding protein coding genes is fairly straight forward, but CREs and non-coding RNAs don't have the same predictable form. In the end, I think it's going to come down to identifying sequences with signatures of selection (ie, what Andolfatto has done).
I saw a paper by Hoekstra and Coyne in May
in which they made the comment that, in a genetic pathway with unlinked genes, a cis regulatory mutation in a downstream gene is the same as a trans regulatory mutation. This is a point that had been bothering me for a little while... For a given degree of pleiotropy/network connectivity, I wonder, is there really a distinction between cis and trans to be made (except in a very tightly focussed, single gene sense)?
Brad, that's a valid point. I discussed the Hoekstra and Coyne paper here. I think the the distinction between cis and trans does only make sense when looking at individual genes, but it's still useful.
Ideally, we'd be able to figure out if some phenotype (in this case gene expression) is the result of a protein coding change or a CRE change. Working out interaction networks to that level of resolution takes a bit more work.