Skip to main content

Biological News

The Uses of Disorder

We spend a lot of time thinking about proteins in this business – after all, they’re the targets for almost every known drug. One of the puzzling things about them, though, is the question of just how orderly they are.
That’s “order” as in “ordered structure”. If you’re used to seeing proteins in X-ray crystal structures, they appear quite orderly indeed, but that’s an illusion. (In fact, to me, that’s one of the biggest things to look out for when dealing with X-ray information – the need to remember that you’re not seeing something that’s built out of solid resin or metal bars. Those nice graphics are, even when they’re right, just snapshots of something that can move around). Even in many X-ray studies, you can see some loops of proteins that just don’t return useful electron density. They’re “disordered”. Sometimes, in the pictures, a structure will be put up in that region as a placeholder (and the crystallographers will tell you not to put much stock in it), and sometimes there will just be a blank region or some dotted lines. Either way, “disordered” means what it says – the protein in that region adopts and/or switches between a number of different conformations, with no clear preference for any of them.
And that makes sense for a big, floppy, loop that makes an excursion out from the ordered core of a protein. But how far can disorder extend? We have a tendency to think that the intrinsic state of a protein is a more or less orderly one, which we just refer to (if we do at all) as “folded”. (You can divide that into two further classes – “properly folded” when the protein does what we want it to do, and “improperly folded” when it doesn’t. There are a number of less polite synonyms for that latter state as well). Are all proteins so well folded, though?
It’s becoming increasingly clear that the answer is no, they aren’t. Here’s a new paper in JACS that examines the crystallographic data and concludes that proteins cover the entire range, from almost completely ordered to almost completely disordered. When you consider that the more disordered ones are surely less likely to be represented in that data set, you have to conclude that there are probably a lot of them out there. Even the ones with relatively orderly regions can turn out to have important functions for their disordered parts. The study of these “intrinsically disordered proteins” (IDPs) has really taken off in the last few years. (Here’s another paper on the subject that’s also just out in JACS, to prove the point!)
So what’s a disordered protein for? (Here’s one of the key papers in the field that addresses this question). One such would have a number of conformations available to it inside a pretty small energy window, and this might permit it to have different functions, binding to rather different partners without having to do much energetically costly refolding. They could be useful for broad selectivity/low affinity situations and have faster on (or off) rates with their binding partners. (That second new JACS paper linked to above suggests that it’s selection pressure on those rates that has given us so many disordered proteins in the first place). Interestingly, several of these IDPs have shown up with links to human disease, so we’re going to have to deal with them somehow. Here’s a recent attempt to come to grips with what structure they have; it’s not an easy task. And it’s not like figuring all this stuff out even for the ordered proteins is all that easy, either, but this is the world as we find it.

17 comments on “The Uses of Disorder”

  1. luysii says:

    I don’t see why a string of 100 amino acids linearly linked together should have a single shape, or just a few shapes (which the proteins which make us up obviously do). Why should one or a few conformations out of the 3^100 possible ones (ignoring those eliminated by self-intersection) be so much stabler than the rest. For more along this line see

  2. MattF says:

    I think there’s an analogy here to discoveries in solid-state physics in the ’70’s. Classic theories of material properties were based on the assumption that solids were crystals– i.e., that solids had long-range periodic order and short-range properties where a minor effect. But in the ’70’s researchers started making disordered materials (‘glasses’) and discovered, to their surprise, that the properties of glasses were generally very similar to the properties of ordered materials. It turned out that most of the physics was in those previously-ignored short-range correlations and short-range interactions.

  3. Anonymous says:

    p53 is a nice example of an extensively disordered protein whose disordered regions are likely critical to its work. Only its globular DNA-binding core is ordered; the flanking wings are a wild collection of disordered regions capable of binding hundreds of different partners.
    Peter Tompa (Hungarian Academy of Sciences) has written a nice monograph on the topic: Structure and Function of Intrinsically Disordered Proteins.

  4. Cellbio says:

    Thought provoking post….suggests that the intrinsic state of many proteins is context dependent. Specific cell, cellular localization and state of cell activation will dictate a set of binding partners and modification states, inducing structure and efficient function of the complex.
    I guess that this aspect of protein structure/function, or our failure to be able to model the complexity in vitro, is what has plagued our target based approaches. As this field develops, I wouldn’t be surprised to learn that many drugs work through mechanisms that induce or inhibit formation of specific conformations of the target and influence partner protein binding, stability and function. I haven’t worked in NHR field, but this might be a primary mechanism of action that is already appreciated? I guess it also plays in less appreciated ways in other target classes that rely heavily on crystal structure, like kinases.

  5. drug_hunter says:

    I think of the emerging science of protein disorder as enabling a much wider range of strategies for engaging receptors. Creative ligand design strategies can help uncover protein conformational changes. In this effort, xray structural information is a great (albeit incomplete) way to shed light on these phenomena. It is always a good day when the crystallographers show us a new protein conformation with an inhibitor bound in a previously unimagined way.

  6. Daen de Leon says:

    It suggests a drug class which might lock in a particular required conformation in these “wobbly” regions, like a folded napkin under a wonky table leg (wobblin’ inhibitors?).

  7. Anonymous says:

    Haven’t read the paper yet but have had first hand experience with what a difference a few residues can make to a construct. Proteins are finicky.

  8. pete says:

    Seems to me that this will definitely add to the vocabulary of protein “functional cassettes” to be considered in attempting to trace the paths of gene-family evolution.
    Here’s a recent PLoS One paper ( that finds conservation in viral nucleoprotein C-termini: regions that are predicted to yield disordered conformations but are known to be critical for several viral functions.
    So much more to learn about protein- (and gene-) structure-function…

  9. Sean says:

    The first JACS mentions that there is a range of order of proteins deposited in the PDB. The range is then normalized and followed up by saying there is a disordered continuum. I must be missing the insight here…

  10. luysii:Why should one or a few conformations out of the 3^100 possible ones (ignoring those eliminated by self-intersection) be so much stabler than the rest?
    You are right, they are not. It’s more like evolution has weeded out the other more or less equi-energetic structures during the fine-tuning of structural, metabolic and signaling pathways. Just imagine the quality control nightmare if every synthesized protein folded into even a slightly different structure. It’s still an intriguing question though.

  11. Density says:

    Disorder really means ‘not enough density’ to model atoms there.
    There can be several reasons for this, or at least several complications for this. That segment may be moving between a few or a lot of different states. The ability to see this distribution of states depends on the quality of the crystals and resulting density. Ultra high resolution structures, better than around 1A diffraction, often permit assigning more than one conformational state to sidechains.
    As the amount of available data (electron density) decreases that gets harder to do and you don’t know if it is because of local wiggle or larger motion.
    Sometimes a protein that won’t crystallize at all can be coaxed to do so by adding another protein or even antibody. This was done with a GPCR structure. GPCRs are interesting because they operate by shifting through several conformational states. Heptares claims to have made a breakthrough in crystallizing them by stabilizing fewer states.
    Sometimes poor, or incomplete refinement, contributes to poor density. There are structures in the PDB with huge B factors and 0 occupancy!
    Bottom line, if you don’t see density have a discussion with the crystallographer(s) and modeler(s).

  12. barry says:

    the most obvious use of disorder is in building a hinge region. More generally, sometimes I need to build a “key” and sometimes I need to build a “lanyard” on which to display that “key”. Since I’m committed to building them both from linear oligomers of amino acids, I must have some regions ordered and some not.

  13. dvizard says:

    @#11 Density: While this is technically true, the fact that proteins are actually crystallized to do X-ray structures on them automatically means that we will *overestimate* the order of proteins, if anything, since the crystals are on the lowest end possible flexibility states of a protein.

  14. Markus K says:

    I have a couple of issues with the concept of disorder presented in the way that it’s currently sold. First is the lack of experimental, wet lab data. The vast majority of the papers on the subject are bioinformatics and datamining, using algorithms to predict disordered regions in protein sequences. Maybe it’s the experimental scientist in me talking, but I’d make no hasty decisions on pure computational data. Low-density regions in loops in x-ray structures is hardly proof.
    Second, what would an almost completely disordered protein be like? From what I can tell it’d wrap those disordered regions around anything it could find to afford a more efficient fold and basically bind to anything and everything that crossed its path. To have proteins like these floating around seems very counterproductive to me.
    Third, where do we draw a line between disorder and a set of folded states? If protein has any structural transitions will it have to be considered partially disordered? The few experimental papers on the subject talk a lot about how their proteins are disordered, when in the majority of cases the disorder can be explained by transitions between two or more stable, folded states. Do we interpret albumins binding a huge array of different substrates as intrinsic disorder or a large ensemble of folds?
    I’m not willing to discount the disorder dogma altogether, definitely not. Proteins are always in a constant flux of movement and transition between states is sure to involve something that could be called disorder as the transitioning element wades through the medium to reach its next state. This constant flux is also responsible for the lower diffraction densities of loop regions compared to rigidly folded ones.
    It is possible that we also have truly disordered proteins in our proteome, but as long as the majority of the data proving it is bioinformatics, and not experimental, I’ll take the news and papers about intrinsic disorder with a spoonful of salt.

  15. Anonymous says:

    I think sequences within proteins required for recognition (as opposed to catalysis) can often work better without structure. No entropy penalty upon binding. Things like induced fit, autoinhibition can happen in “low complexity” regions but also in conserved kinase loops.
    NMR can directly measure flexibility on a large range of timescales in large proteins. It can distinguish cases of ensembles versus unstructured and measure conversion raes between them. Very powerful when combined with crystal structures.
    Applies to RNA too…

  16. Sean Ekins says:

    Thank you Derek, I enjoyed this blog and it reminded me of a collaboration a few years back which I was involved with looking at Nuclear Hormone Receptors (NHR)and Intrinsic Disorder (
    We looked at human NHRs and analyzed up and downstream connections in a pathway database and also looked at disorder predictions for all species with NHR sequences available. Our results indicated that prediction methods were useful at identifying regions of known disorder in human NHRs and variability of intrinsic disorder within NHRs was likely an important evolutionary force in shaping protein-protein interactions and NHR function.
    My interest initially was around understanding promiscuity of proteins and interactions with small molecules and the work from the Dunker group (which you mention) was influential. I think there is plenty more work to be done to understand the role of disorder.

Comments are closed.