A window on protein-carbohydrate interactions

It is a truth universally acknowledged that a single new method, especially one relevant to an area that has hitherto been resistant to study, often opens up enormous possibilities for increased understanding.  And yet, new methods often don’t get the publicity they deserve.  Possibly this is because a good new method, like a good new friend, is only really possible to identify in retrospect.  Will your new friend keep in touch when you move away, or is s/he only a good companion in a narrow geographical range?  Will the new method work in the system you’re interested in, or is it restricted to a narrow range of applications?  I don’t know how important the most recent method from Linda Hsieh-Wilson’s lab (Rogers et al. 2011.  Elucidating glycosaminoglycan—protein-protein interactions using carbohydrate microarray and computational approaches.  PNAS  PMID 21628576) is going to turn out to be; but I can tell that it’s a clever approach to a previously intractable problem.

The problem is the question of how to track interactions between proteins and various kinds of glycosaminoglycans (GAGs).  GAGs are polymers of disaccharides in which one of the sugar units carries an amine; they’re usually linked to proteins to make proteoglycans, which make up much of the extracellular matrix.  They’ve been hard to study because they’re structurally diverse — they may be made up of several different sugars, the chain can be over 100 sugar units long, and each sugar group may be modified with sulfate groups at different positions.  Hard to study does not equal uninteresting, however: GAGs regulate all kinds of biological processes, including cell growth, viral invasion, blood coagulation, and the process of recovery (or not) from that cruelest of injuries, spinal cord injury.  GAGs have been shown to help assemble multimeric protein complexes of growth factors such as fibroblast growth factor, helping the growth factor to induce signaling in target cells.  But little is known about how these interactions are controlled, because — until now, perhaps — the specificity of these interactions has been really hard to look at.

Rogers et al.’s attack on the problem combines two very different approaches.  They’ve developed carbohydrate microarrays that include a number of different synthetic sugar tetramers of defined sulfation pattern, as well as natural glycosaminoglycans. Incubating proteins with these microarrays allows the authors to pick up a range of interactions between GAGs and either individual proteins or protein-protein complexes, using small amounts of material.  Having discovered which sugar tetramers bind to the protein, they computationally attempt to dock one of the tetramers into different potential binding sites on the protein structure.  This might give you many potential binding sites, which you can then computationally string together (if the connectivity looks reasonable) to identify a binding site for a longer glycan string.  By combining the information on what does in fact bind with what looks as if it should bind given structural information and energy minimization, Rogers et al. can make a relatively confident prediction about which GAGs should bind to the protein well and which poorly.  And, crucially, they can identify binding sites on protein-protein complexes.

The authors first set out to validate their model by looking at protein-GAG interactions that are already relatively well understood.  The first question they wanted to answer was whether they could identify the binding sites on a protein if they already knew which GAG they were dealing with. For example, chondroitin sulfate A, a variety of GAG, binds to two sites on VAR2CSA, a protein from the parasite that causes malaria.  (Chondroitin sulfate is a linear polymer of glucuronic acid and N-acetylgalactosamine, and the A suffix refers to the fact that this particular form of chondroitin sulfate carries most or all of its sulfate groups on the C-4 position of the galactosamines.)  The interaction between this protein and this GAG causes a distressing syndrome in which the parasite accumulates in the placenta in pregnant women, resulting in low-birth-weight babies and anemic mothers. Rogers et al. identified 7 of the 8 residues in the two sites on VAR2CSA that are likely to be in direct contact with the GAG.

For a more complete test of the method, they turned to the well-known interaction between fibroblast growth factors and heparan sulfates.  The interesting aspect of this interaction is that the GAG, heparan sulfate, stabilizes the binding of the growth factor to its receptor. So the extracellular matrix surrounding a cell (which is made up in large part of proteoglycans, proteins that carry GAGs) can help to control how the cell responds to a growth factor.  Rogers et al. picked a representative example of the fibroblast growth factor family, FGF2, and a receptor, FGFR1.  Because FGFR1 is a membrane protein, they had to engineer a soluble version to test for binding on their microarrays: they (genetically) chopped off the extracellular region and fused it to the back end of an antibody, known as the Fc portion.  They then incubated FGF2 alone, the FGFR1-Fc protein alone, or a 1:1 mixture of FGF2 and FGFR1-Fc with the different tetrasaccharides and natural GAGs on their microarrays.  They find, as they should, that FGF2 binds to heparan sulfate tetrasaccharides and to the anticoagulant heparin, which is a close relative of heparan sulfate and is often used as a surrogate for natural heparan sulfates.  FGFR1-Fc alone doesn’t bind to much: but when FGF2 is present, FGFR1-Fc now binds.  This presumably reflects the formation of FGF2-FGFR1-Fc complexes that bind to heparan sulfates.  Next, they used their computational docking method to look for the heparin-binding sites on both proteins, and on the FGF2-FGFR1 complex.  Although they couldn’t find the heparin-binding site on the complex directly through docking, they could find sites on both of the individual proteins; and when these were overlaid on a structure of the complex, they could see a contiguous binding domain that matches the binding domain identified by crystallography extraordinarily well.

If they hadn’t known the answer in the first place, what would they have found out?  They wouldn’t have been able to claim confidently that FGF2-FGFR1 complexes are stabilized by heparan sulfate — you can’t directly use the microarrays to show that the amount of FGF2-FGFR1-Fc complex increases in the presence of heparan sulfates — but they would at least have been able to say that there’s a good reason to look for three-way complexes of FGF2, FGFR1 and heparan sulfate, and to ask whether the GAG binding site that bridges FGF2 and FGFR1 actually stabilizes the growth factor-receptor complex.

Enough validation.  What happens if they look at an interaction that isn’t already well understood? The Hsieh-Wilson lab has been interested in the interactions between TNF-alpha and GAGs for a while, and have evidence that a specific variety of chondroitin sulfate actually antagonizes the binding of TNF-alpha to its receptor.  Using the same approach I just described for FGF (soluble version of receptor, measurement of binding of protein, receptor and both together to the GAG microarray, etc.) they find that the single proteins (TNF-alpha alone, or the TNF-alpha receptor alone) selectively bind to the correct form of chondroitin sulfate, CS-E.  But the complex between TNF-alpha and its receptor doesn’t bind.  Docking shows that the binding site on TNF-alpha for CS-E overlaps with its binding site for the receptor: so indeed, TNF-alpha shouldn’t be able to bind CS-E when the receptor is already bound, and binding to CS-E should antagonize receptor binding. As an encore, Rogers et al. identify a large number of novel GAG-protein complex interactions involving the important neurotrophin family of proteins, which could have major implications for the specificity of neurotrophin signaling in vivo, and therefore for the survival, growth and function of neurons.

Again, who knows how important this really will turn out to be.  But when you have an area that’s hard to study and you find a new window through which to look at it, sometimes you learn a lot.  We’ll see.

Rogers CJ, Clark PM, Tully SE, Abrol R, Garcia KC, Goddard WA 3rd, & Hsieh-Wilson LC (2011). Elucidating glycosaminoglycan-protein-protein interactions using carbohydrate microarray and computational approaches. Proceedings of the National Academy of Sciences of the United States of America, 108 (24), 9747-52 PMID: 21628576