Getting more complex and gaining.... nothing

 The origin of complexity is a highly debated issue in biology. For instance, many functions in the cell are carried out by intricate macro-molecular complexes formed of a multitude of subunits. When tracing the evolution of such complexes, as we did with mitochondrial Complex I, one often finds that the number of subunits have increased through time. However, the addition of subunits not always seems to correlate with the acquisition of novel functions, which would provide a selective advantage for the increase in complexity.  Can we think of a mechanism promoting a trend for increasing complexity in the absence of a selective advantage provided by a novel function?.

 A recent paper by Finnigan and colleagues show a plausible mechanism and present evidence that this may have been responsible for the acquisition of a novel subunit in fungal vacuolar ATPases (depicted below).

 This molecular machines that pump protons across membranes have a membrane ring (in green in the figure) formed by 6 units. In vertebrates two different subunits (originated from the duplication of an ancestral gene) form the 6-units ring in a 1:5, stoichiometry. In fungi a more recent duplication brought about one more subunit type so that the ring is formed by the products of three different genes in a 1:1:4 organization. Using ancestral sequence resurrection (I love that name!), a technique that consists of reconstructing most likely ancestral sequences and then synthesizing them in the lab, they show that a single mutation acquired early in each paralogue, was sufficient for making the two of them indispensable. Thus, such model could explain a trend to increase complexity in multi-paralogue complexes (those comprised by some subunits derived from duplicated genes) without a requirement for an initial selective advantage.  

 In a way, I see this model as a special type of sub-functionalization. That is, the two new paralogues would in sum make the same function that was performed by the ancestral gene. In the absence of more examples we do not know how widespread is this mechanism, but the fact that it does require few likely events and that it actually constitutes a "ratchet" (as noted by W Ford Doolittle), that is once you gain that complexity you don't go back, one would expect to have occurred in several of many multi-paralogue complexes, at least in some lineages. 

 Perhaps this could explain an intersting finding we did some years ago when looking at the evolution of the mitochondrial electron transport chain in fungi (mostly formed by multi-protein complexes): the amount of duplications in members of this complexes was of the same level as other proteins. This is in contrast to the gene-dosage effect hypothesis that states that complexes would tend to duplicate only when the stochiometry is conserved (that is in when the whole complex duplicates, e.g in whole genome duplications). 

 Finally, another remark that I always do when seeing ancestral sequence resurrection working is that the fact that ancestral reconstructions display the expected biochemical activities (e.g by complementing extant sequences) is an indication that the models of evolution we use are not that wrong after all.