Is Sean Carroll Wrong?

One of the primary hypotheses of Sean Carroll's model of evo-devo is that cis-regulatory elements (CREs) are the primary drivers of morphological evolution (see here). This hypothesis is controversial in the evolutionary genetics community. Because it's hard to examine the effect of CREs on phenotypes at a genome-wide scale, the problem must be reduced into smaller elements. One such element that has been the subject of previous studies is the role of cis and trans mutations in the evolution of gene expression. I have previously discussed some results that shed light on this issue. The basic conclusion from those results is that the data indicate that both cis and trans changes play an important role in the evolution of gene expression.

A previous study examined the proportion of expression differences between two Drosophila species that were due to cis changes, and the fraction due to trans changes. That study found that the cis changes play a larger role, but trans mutations are also important. A new paper reports the results of a similar experiment in yeast. The authors of this study found that trans changes play a larger role than cis changes in expression differences between two strains of yeast. They do, however, acknowledge that cis changes also play an important role.

That brings us to the title of this post: is Sean Carroll wrong? Keeping in mind that these studies examine gene expression rather than morphology, it appears that Carroll is at least partially wrong. While CREs are important in teh evolution of phenotypes, they are not as important as Carroll sells them to be. One caveat is that trans effects on expression differences may, ultimately, be due to changes in CREs of transcription factors. Changing the expression pattern of a transcription factor may, ultimately, affect the transcription of multiple other genes. While this will show up as a trans effect in the assays to distinguish cis and trans changes, it is actually due to a mutation in a CRE of a regulatory gene located upstream in some gene expression pathway.


Hoekstra HE, and Coyne JA. 2007 The locus of evolution: evo devo and the genetics of adaptation. Evolution 61:995-1016 doi:10.1111/j.1558-5646.2007.00105.x

Prud'homme B, Gompel N, and Carroll SB. 2007. Emerging principles of regulatory evolution. Proc Natl Acad Sci USA 104:8605-8612 doi:10.1073/pnas.0700488104

Wang D, Sung H-M, Wang T-Y, Huang C-J, Yang P, et al. 2007. Expression evolution in yeast genes of single-input modules is mainly due to changes in trans-acting factors. Genome Res 17:1161-1169 doi:10.1101/gr.6328907

Wittkopp PT, Haerum BK, and Clark AG. 2004. Evolutionary changes in cis and trans gene regulation. Nature 430:85-88 doi:10.1038/nature02698

More like this

sorry about the crappy formatting. one other thing:

they claim you'd expect genes controlled by many factors to be more likely to be regulated by variation in trans (as opposed to genes regulated by a single input). I'm not sure that's true--depending on how the network it set up, you can imagine a multi-input signal being more robust (ie. if you need either TF1 or TF2 to start transcription, changes in the expression of TF2 won't really affect the gene-- TF1 is still present and able to activate).

It will be nice once Matt Hahn (Indiana U.) publishes what he's been talking about lately. He's got a subset of genes that very clearly show rapid coding-sequence evolution associated with speciation, and they're not the immune-etc. ones you normally think about. It's not the whole story, but it's clear that he's found a big missing piece that weakens the regulatory-overwhelms-coding-sequence assumption we've had since King & Wilson 1975.

My guess is that rather than cis-over-trans, or reg-over-CDS, or whatever, the ultimate pattern will fit the "whatever works" dictum that evolution invariably follows :-)

note that the authors of the study you mention are only looking at genes in single input modules-- a network structure in which a single gene controls the expression of a number of others. it makes sense that expression changes in this network architecture would be driven by trans-acting factors (likely the "single input" of the "single input module").

It might be more productive to think about the interplay between network organization and proportion of cis v. trans changes than the latter alone.

p-ter, I think you have it backwards. From the paper:

First, we chose yeast single-input module (SIM) genes, each of which is putatively regulated by a single transcription factor(TF). Since a SIM gene is regulated by a single immediate upstream TF, the chance for a change to occur in its trans-acting factors would, on average, be smaller than that for a multiple input module (MIM) gene, because a MIM gene is regulated by more TFs, and thus, likely by more trans factors than a SIM gene.

The way I understand it, SIM genes are individual genes that are each regulated by a single (not necessarily the same) transcription factor, not genes regulated by the same transcription factor.

hm, it's somewhat confusingly worded. A single input module is a structure where a number of genes are all regulated by a single factor (Alon's An Introduction to System's Biology has some good figures). It looks like they took genes from a number of SIMs:

This database includes 203 known TFs and their downstream target genes, from which we collected 1049 genes, each of which is uniquely regulated by only one of 72 TFs at a significance level of P < 10-3

That's an average of ~12 genes/module. one might expect genes in single input modules (as opposed to auto-regulatory loops or feedback loops) to be regulated in trans-- if several genes are controlled by a single factor, changes in regulation in that factor should be propogate through the network.

sorry about the crappy formatting. one other thing:

they claim you'd expect genes controlled by many factors to be more likely to be regulated by variation in trans (as opposed to genes regulated by a single input). I'm not sure that's true--depending on how the network it set up, you can imagine a multi-input signal being more robust (ie. if you need either TF1 or TF2 to start transcription, changes in the expression of TF2 won't really affect the gene-- TF1 is still present and able to activate).