posted 09. July 2003 17:51                    

quote:

---

Evolutionarily conserved networks of residues mediate allosteric communication in proteins
Gürol M. Süel 1, 2, Steve W. Lockless1, 2, Mark A. Wall2 & Rama Ranganathan2

1. These authors contributed equally to this work.
2. Howard Hughes Medical Institute and Department of Pharmacology, The University of Texas Southwestern Medical Center, 5323 Harry Hines Boulevard, Dallas, Texas 75390-9050, USA.
Correspondence should be addressed to R Ranganathan. e-mail: rama@chop.swmed.edu

A fundamental goal in cellular signaling is to understand allosteric communication, the process by which signals originating at one site in a protein propagate reliably to affect distant functional sites. The general principles of protein structure that underlie this process remain unknown. Here, we describe a sequence-based statistical method for quantitatively mapping the global network of amino acid interactions in a protein. Application of this method for three structurally and functionally distinct protein families (G protein–coupled receptors, the chymotrypsin class of serine proteases and hemoglobins) reveals a surprisingly simple architecture for amino acid interactions in each protein family: a small subset of residues forms physically connected networks that link distant functional sites in the tertiary structure. Although small in number, residues comprising the network show excellent correlation with the large body of mechanistic data available for each family. The data suggest that evolutionarily conserved sparse networks of amino acid interactions represent structural motifs for allosteric communication in proteins.

Communication between distant sites in allosteric proteins is fundamental to their function and often defines the biological role of a protein family. In signaling proteins, it represents information transfer — the transmission of signals initiated at one functional surface to a distinct surface mediating downstream signaling. For example, ligand binding at an externally accessible site in G protein–coupled receptors (GPCRs) reliably triggers structural changes at distant cytoplasmic domains that mediate interaction with heterotrimeric G proteins1, 2. Studies in many other protein systems indicate that long-range interactions of amino acids also are important in binding (and catalytic) specificity. Substrate recognition in the chymotrypsin family of serine proteases3, 4, the tuning of antibody specificity through B-cell maturation5 and the cooperativity of oxygen binding in hemoglobin6-9 all depend not only on residues directly contacting substrate, but also on distant residues located in supporting loops and other secondary structural elements. Crystallographic studies in all of these systems5, 9-11 indicate that the distant residues participating in substrate recognition do so by acting through intervening positions to control the structure of the substrate-binding site. These long-range interactions are remarkable because many other sites, even if closer to active site residues, show little contribution to function. Taken together, these studies indicate that proteins are complex materials in which perturbations at sites — for example, substrate binding, covalent modification or mutation — may cause conformational change to happen in a fracture-like manner that is not obvious in atomic structures. From a biological point of view, these fractures represent the energy transduction mechanisms that mediate signal flow, allosteric regulation and specificity in molecular recognition.

---

posted 10. July 2003 17:10                   

quote:

---

This is work raises a very relevant question for protein evolution. If allosteric proteins in general function by engaging conduits of energetically coupled residues to transmit binding energies and convey signals, how can these proteins evolve in a stepwise fashion? If a mutation disrupts the conductivity of these allosteric conduits, then multiple mutations must occur simultaneously in order for the protein to maintain allosteric function.

---

But the paper didn't show that these networks can't tolerate mutation. Quite the opposite; the patterns of mutations that they do tolerate -- what the authors refer to as perturbation coupling -- allow us to identify the network on the basis of conservation between couplets of residues. In other words, if you find a mutation to glutamate (just pulling this as an example out of thin air here) at one position, you should expect to see a compensatory mutation (perhaps to a basic amino acid) at another position more frequently than what's expected by chance alone. By using a statistical analysis, they've identified residues of high coulpling in a large set of sequence data from many different proteins of the same family. When they add these couplets together, they end up with a network consisting of around 10-15% of the protein's residues. And lo and behold, this network correlates nicely to what's been established (by other means) as being the responsible party for the protein's function. But these networks do not have invariant sequences. Nor for that matter is any given residue within the network itself always going to be highly conserved. If this were the case, then the authors would not have had to have done the analysis that they did -- a simple sequence alignment would have sufficed, and they could have just picked out the conserved residues from there. Instead they had to account for the fact that the residues within the network are constantly mutating, but that this should constrain the mutations that should be selected for in other parts of the network.

I just wish I could do this kind of analysis on my protein. No more guess work in trying to find residues to mutate! But alas, I don't have mountains of sequence data. Maybe some day.

posted 11. July 2003 01:06                    

quote:

---

"Interesting hypothesis which I would like to see further worked out in detail to determine if indeed a stepwise fashion evolution is impossible. Looking forward to your further contributions here."

---

quote:

---

"This is a very interesting paper. Thanks Josh for bringing it to my attention."

---

quote:

---

You write:

quote:
This is work raises a very relevant question for protein evolution. If allosteric proteins in general function by engaging conduits of energetically coupled residues to transmit binding energies and convey signals, how can these proteins evolve in a stepwise fashion? If a mutation disrupts the conductivity of these allosteric conduits, then multiple mutations must occur simultaneously in order for the protein to maintain allosteric function.

But the paper didn't show that these networks can't tolerate mutation.

---

quote:

---

"Quite the opposite; the patterns of mutations that they do tolerate -- what the authors refer to as perturbation coupling -- allow us to identify the network on the basis of conservation between couplets of residues. In other words, if you find a mutation to glutamate (just pulling this as an example out of thin air here) at one position, you should expect to see a compensatory mutation (perhaps to a basic amino acid) at another position more frequently than what's expected by chance alone .

---

quote:

---

But these networks do not have invariant sequences. Nor for that matter is any given residue within the network itself always going to be highly conserved.

---

quote:

---

If this were the case, then the authors would not have had to have done the analysis that they did -- a simple sequence alignment would have sufficed, and they could have just picked out the conserved residues from there.

---

quote:

---

Instead they had to account for the fact that the residues within the network are constantly mutating, but that this should constrain the mutations that should be selected for in other parts of the network.

---

quote:

---

I just wish I could do this kind of analysis on my protein. No more guess work in trying to find residues to mutate! But alas, I don't have mountains of sequence data. Maybe some day.

---

quote:

---

I'd say that these statements add to the list of data and conclusions that support a low-CSI view of biology. The involvement of small numbers of amino acid residues in allosteric mechanisms is a pretty obvious addition to the list. The "fracture-like manner" in which allosteric effects may come about also reflects a low-CSI "state" - this follows from the limited fraction of any given protein that may be involved in the "fracture"

---

quote:

---

If a mutation disrupts the conductivity of these allosteric conduits, then multiple mutations must occur simultaneously in order for the protein to maintain allosteric function.
I am not sure what the point of this statement is - it seems to be implying a greater susceptibility of allosteric proteins to mutation, but I don't think you can support this with reasonable estimates of mutation frequency.

---

quote:

---

The more interesting point vis-a-vis mutations derives from the low-CSI implications mentioned above. It turns out that, in all likelihood, networks that function by transducing conformational changes can evolve relatively easily, owing to the small numbers of amino acid residues that need to change to add to or alter any given network. This is a manifestation of the low-CSI state of proteins.

---

posted 11. July 2003 16:15                   

quote:

---

Indeed, and they don't study the intermediates, just functional sequences found in current organisms. If you change this particular glutamate before making the compensatory mutations, is the allosteric function maintained? Even though we have an extensive database for the sequence families they chose, it does not necessarily reveal the status of intermediates. However, this is the point by itself: if two residues must be altered together, then this is an obstacle for RM&NS.

---

quote:

---

This addresses Art's CSI point. The reason they found only a few residues to be involved is that they weren't simply doing sequence alignments with proteins containing highly conserved residues.

---

quote:

---

The question is whether the time period between mutation 1 and the constrained mutation 2 involves a protein of compromised allosteric function. Certaintly some mutations compromise function before the second occurs to restore it.

---

quote:

---

And by the way, did you mean that the authors were cataloguing the complexity of design as we see it, and arraying it in such a way that we can detect allosteric interactions? Who said mutations produced all of this? (tongue in cheek) Are we assuming the answer before asking the question?

---

If we were to assume that these sequences are all designed, rather than derived from a common ancestor (assume a designed common ancestor if you want), why we should expect to see the level of variation that we do see? Why not make all of the independent amino acids identical? That would seem to be easier. There certainly isn't any reason to make their distribution the same as that for all natural proteins, mimicking an equilibrium process based on random mutation. And furthermore, I would add that even if we assume that there was design at some time in the past, mutation and evolution of these sequences is going to be inevitable. So the analysis which assumes their evolution (and works quite nicely, surprise, surprise), would still be valid.

quote:

---

Even if you get enough sequences, they have to vary enough such that the whole protein doesn't look like it covaries with each and every other residue!

---

True 'dat. But of course the idea is that the more proteins you get, the more variety you'll have. My protein is at least 600MY old, and can be found in all animalia which have so-far been sequenced. (Which means a whopping 6 or so ) So if I had more sequences, I should have lots of variation to study. Or at least this is what you think if you're an evolutionist.

posted 11. July 2003 18:40                    

quote:

---

True 'dat. But of course the idea is that the more proteins you get, the more variety you'll have. My protein is at least 600MY old, and can be found in all animalia which have so-far been sequenced. (Which means a whopping 6 or so [Mad] ) So if I had more sequences, I should have lots of variation to study. Or at least this is what you think if you're an evolutionist.

---

posted 17. July 2003 12:49                    

quote:

---

Sorry, I thought you were making a stronger claim than you actually were. As far as using this approach to further study the evolution of proteins, I think it's a great idea, but the subject is extremely difficult to tease apart for a variety of reasons. That's why I get irritated with people like Dembski who use a "tornado in a junkyard" approach, which is simplisitic to the point of being totally useless. But anyway, about some of your specific comments:

---

quote:

---

The question is whether the time period between mutation 1 and the constrained mutation 2 involves a protein of compromised allosteric function. Certaintly some mutations compromise function before the second occurs to restore it. That's probably true in some cases, but again I think it's highly context dependent. The fact that these residues vary as to which ones they interact with (again, it's hard to tell how much) shows that there is probably an interconnected pathway through sequence space which can be traversed to reach any of the sequences used in the analysis. In fact, it can probably be traversed to reach a large number of protein sequences with highly diverse functions. But I guess that's the question we're asking, isn't it?

---

quote:

---

If we were to assume that these sequences are all designed, rather than derived from a common ancestor (assume a designed common ancestor if you want), why we should expect to see the level of variation that we do see?

---

quote:

---

Why not make all of the independent amino acids identical? That would seem to be easier. There certainly isn't any reason to make their distribution the same as that for all natural proteins, mimicking an equilibrium process based on random mutation.

---

quote:

---

And furthermore, I would add that even if we assume that there was design at some time in the past, mutation and evolution of these sequences is going to be inevitable. So the analysis which assumes their evolution (and works quite nicely, surprise, surprise), would still be valid.

---

posted 29. July 2003 10:26                    

quote:

---

After reading the paper you cite, I'd have to say: 1. For the proteins examined, the combinations that are involved in allosteric effects are indeed quite malleable. This follows directly from the data, which shows a degree of variability that takes some pretty clever tweaking to tease apart.

---

quote:

---

2. The matter of accesibility and bridges in the instances discussed in the paper is not an assumption, but a reasonable and logical conclusion that flows from the data and analysis in the paper you have pointed us to. (A useful exercise would be to examine all of the dataset for one of the proteins of interest, focus on a few of the residues suggested to be important for allosteric effects, and catalogue the variability in the putative interacting amino acid partners. I haven't done this, but I suspect that one would see sufficient flexibility to easily accommodate "Darwinian" processes.)

---

quote:

---

Now for the low-CSI state of things:

---

posted 30. July 2003 18:26                   

Sorry for the delay in getting back to you, Josh. I know about the "perturbation constraints" of grad school.

I wrote:

"If we were to assume that these sequences are all designed, rather than derived from a common ancestor (assume a designed common ancestor if you want), why we should expect to see the level of variation that we do see?"

And you replied:

"Are you asking me to offer wildly speculative models? "

No! I was just asking a rhetorical question, in the hopes of pointing out why the authors, along with everyone else, find the evolutionary point of view useful. In fact, the very nature of the study depends on the idea that random mutation is randomizing the covariance between two residues that are not coevolving. If random mutation had not been at work, they would have had lots of false positives. The author's explain their reasoning, for example...

quote:

---

First, if site l contributes nothing to either the folding or function of the protein, the corresponding amino acid frequencies in the MSA should be unconstrained and, therefore, should approach their mean values in all proteins.

---

quote:

---

We made a perturbation at this site by extracting the subalignment containing only tyrosine at this site (Tyr296), a manipulation that retains 34.6% of sequences from the parent alignment. Both the full MSA and the Tyr296 subalignment are sufficiently diverse and over-represented so that unconserved sites, such as position 19, show amino acid frequencies near to their mean values found in all proteins (Fig. 1d).

---

quote:

---

In contrast, position 125, a moderately conserved site in the third transmembrane helix of GPCRs, shows several changes in its distribution in response to the Tyr296 perturbation (Fig. 1e) corresponding to a larger coupling energy (Gstat125,296Y = 1.86kT*).

---