[Sticky proteins and complex relationships]
[(protein) Relationship drama: promiscuous proteins in small populations]
[Not all is good that sticks: non-adaptive complexity gain through compensatory protein adhesion]
[Man, I suck at titles]
NB: This post can be considered as part 2.5 of my In defense of constructive neutral evolution series; also recommended for some background are part 1, discussing selection, drift and Neutral Theory, and part 2, discussing Constructive Neutral Evolution; to answer a popular question, part 3 *will* materialise
Constructive neutral evolution is one mechanism of complexity increase without any associated increase in fitness – or, in other words, non-adaptive complexity gain. Basically, a random interaction between two proteins can lead to a fixed dependency if this interaction compensates for a mutation that was otherwise lethal – termed 'pressuppression'. In this way, previously unnecessary dependencies accumulate to make a very bulky, bureaucratic system that essentially does the same thing. We've all seen it in our institutions, and evolution is about as efficient.
Fernández and Lynch 2011 Nature paper, from here onwards referred to as "the paper".
Protein 'stickiness' can be enhanced by biochemical means. Proteins vary in stability, and themselves come in populations – generally, most are in the optimal conformation that is presumably functional, but some individuals are messed up. This happens well past the sequence and folding errors, and some perfectly 'normal' proteins can be in a suboptimal state at any given time. Clearly, this affects the overall efficiency of the protein – even if it's enzymatically awesome, the overall 'protein' as we biologists understand it (sans population aspect) would decline in efficiency if a large chunk of its population is in a misfolded state.
One aspect that pushes around the proportion of the protein in the 'right' conformation is how well it plays with water. It shouldn't be too surprising that hydrophobic regions induce instability. What was new to me, but perhaps old news to those who actually understood chemistry, is that the exposure of the polar(hydrophilic) protein backbone to water also has a destabilising effect – and not only that, but often more significant than that of exposed hydrophobic regions! This may seem counterintuitive – doesn't water like hydrophilic regions? And there lies our problem.
Water molecules are attracted to polar groups, and the amino acid backbone is quite polar. This means little water molecules wander in towards the backbone and form hydrogen bonds with it. The problem is twofold: first of all, the protein, like all molecules, likes to 'jiggle'. The more it can jiggle in its given conformation, the more favourable that conformation is thermodynamically since its satisfied by more states. Entropy, etc. (now we're *really* entering territory I know nothing about, since my phys chem experience is locked away by PTSD...). Hooking up this backbone with water molecules reduces its 'jiggle' room, and makes it less thermodynamically stable – making change to other conformations more probable, therefore possibly leading to more errors in the protein population.
Secondly, as detailed further in the paper, water likes to hang out with more of itself. Water molecules are happiest in foursomes, sharing four hydrogen bonds with their neighbours. When a creepy protein backbone emerges and lures an unsuspecting water molecule away into the protein's murky depths, the water molecule cannot form as many bonds with its fellows (or as many hydrogen bonds, period), and is really sad and lonely. Or, in proper terms, the system becomes less stable, since thermodynamics will favour an arrangement where these water molecules are all happily coordinated with each other, and not being molested in a corner by an amino acid polar group. In other words, exposing the polar backbone (Solvent-Accessible Backbone Hydrogen Bonds, SABHBs in the paper) to water induces what is called Protein-Water Interfacial Tension (PWIT).
One way this tension can be released and the backbone exposure ('coded for' by genes, by the way) can be compensated for is if a random other protein (or more of its own kind) are recruited to cover that exposed backbone. This would help stabilise the protein conformation, and allow this potentially deleterious drawback to be tolerated (and get fixed in the population). Ultimately, the second (and third, etc) protein can become exapted for something useful, although just an eventual dependency is good enough to make sure these proteins stick together permanently. The crazy web of interactions gets crazier.
(Disclaimer: I'm horrible at chemistry, this may all have been thoroughly wrong...read the paper.)
Fernández & Lynch's fig1a suffices perfectly but I like making diagrams, so I made one anyway. See text.
Fernández & Lynch's fig1a suffices perfectly but I like making diagrams, so I made one anyway. See text.
Now I'm about the last person to willingly blog about biochemistry, and this seems to have little only a distant relevance to evolution, particularly the non-adaptive kind that fascinates yours truly. It will make sense in a bit. Recall from a few seconds ago (hey, already difficult for some of us) that protein instability leads to reduced protein efficiency. This reduction is generally tolerated, however, until it's bad enough to have a higher chance of being removed. Recall from [what should be] introductory population genetics that selection acts probabilistically, with true slightly deleterious mutations have a lesser, but still significant, chance of fixation than strongly deleterious mutations, which selection has a higher chance of taking care of before drift quietly fixes it. (more detail in older post here) Since proteins are, quite unsurprisingly, also governed by fundamental principles of population genetics, drift becomes involved there too.
As populations get smaller, drift becomes a more dominant force relative to selection, and the window of 'effectively neutral' mutations – slightly beneficial and slightly deleterious, but unlikely to be dealt with by selection – increases. More mess is tolerated. This means more protein inefficiencies are allowed to fix in the population, those induced by backbone exposure among them. Since there are now more proteins that are no longer happy with themselves (or, rather, have an increased Protein-Water Interfacial Tension), they are more likely to stick together for biochemical stability. And here Constructive Neutral Evolution can come in too, allowing further deleterious mutations that are now presuppressed by the recruited proteins. In a way, this greases the presuppression process, rather than competing with it as this BBC news piece made Ford Doolittle appear to suggest.
Now, this is all great in theory, but is there any real data in support of this? For one thing, there is a clear increase of interactome (set of all interactions in an organism) complexity correlating with decrease in effective population size, suggesting a link between lax selection and accumulating complexity. Furthermore, the proteins in organisms of these smaller populations have more blistering backbone exposures to water. Supporting the relationship with population size further yet with the advantage of more phylogenetically independent events (but less interactome data), bacterial intracellular endosymbionts consistently exhibit higher protein backbone exposure (hydration) than their free-living counterparts. Selection appears to disfavour not only polar backbone exposure (also described as 'poorly wrapped proteins' in the paper), but once again, the rise of interaction complexity as a whole. (Fernández and Lynch 2011 Nature, in case you somehow managed to miss that)
Obviously I like this paper because it adds another mechanism to the arsenal of evolutionary processes happening independently of adaptation. But moreover, I don't think one can find too many examples of biochemistry mixed with population genetics. You hardly find cell and developmental biologists thinking about population genetics, and perhaps many biochemists have never even been exposed to such a subject. When fields that should never come that close together do, some really nice explosions of insight can occur (my sad attempt at chemical metaphors). We really need to talk to other more, and maybe even wander over to other departments from time to time. It's sometimes (often) frustrating to communicate with those strange ones from afar, but just like ethnic xenophobia, its interdisciplinary counterpart must also be overcome.
Figure 2a annoyed me a little as it ignored phylogenetic relationships, which is a big no-no when comparing properties of taxa. The figure is technically fine, especially since there aren't any correlation analyses there, but it's hard to discount phylogenetic history as being the cause behind the correlation of the traits without actually the characters on a tree. Anyway, since I like playing with data and running statistical analyses on things, especially when I didn't actually have to go through the pain of obtaining the data myself, I mapped some characters (interactome complexity from fig2a) on a phylogeny:
Unfortunately, even the most basic statistical operations become an epic headache when trees are involved, and very quickly things become painfully complicated, for the human as well as the computer. Especially when you're handed a dataset of mixed categorical and continuous characters, as I learned the hard way last night. After fighting Mesquite for a good many hours, I finally had to resort to extracting the Ne*µ (effective pop size * mutation rate; roughly put, both lead to increased selection efficiency) estimates from Lynch & Conery 2003 – relying on an intersection of two datasets meant that our taxon sampling was quite sad by the end of this enterprise. Anyway, I ran a pairwise comparison test (Maddison 1999 J Theor Biol) on the data, which probably isn't the best thing ever, but I got something resembling significance: p = 0.019. Depending on how statistically noisy your field is, you may even deem this acceptable. In any case, not too bad given my crude (and somewhat clueless) analysis and limited taxon sampling:
I mostly did this because I thought it'd take a couple hours max. If hours meant days, that wasn't too far off... but hey, I learned something!
Acknowledgments: thanks to Lucas Brouwers for helping me wade through the heavy biochemical stuff, and to Mike Lynch for explaining the key idea of the paper a while earlier. Otherwise I would've probably been too daunted to even read it, let alone blog about it...
Oh, and my Twitter people for random phylogenetics advice ;-)
Fernández, A., & Lynch, M. (2011). Non-adaptive origins of interactome complexity Nature DOI: 10.1038/nature09992
[will add some supplementary refs once I return to internet on Monday...]