More Recent Comments

Tuesday, December 01, 2009

On Determining the Structure of a Protein


Michael Clarkson is a biotech postdoc who blogs at Discount Thoughts. One of his recent thoughts is Don't look for "the" structure. He is referring to the fact that the crystal structure of a protein doesn't actually represent the only structure that the protein can adopt.

The figure shown here illustrates one of the problems with referring to the structure of a protein. This is a representation of an NMR structure of bovine ribonuclease A. It shows that various parts of the protein exist in several different conformations. The actual protein structure is a composite of all these structures in equilibrium with each other.

These conformation could be considered "breathing" and you may think they're not important. However, there are many cases where the conformations of a protein are quite different. We are familiar with allostery, where the conformation of a protein changes when it's bound to a ligand, but the are also examples where two very different structures exist in equilibrium in the absence of ligand.

Read his blog posting and keep in mind that proteins are dynamic structures and not static rigid crystals.



15 comments :

TwoYaks said...

You mean proteins don't exist as crystals in vivo? The devil you say!

Next you'll tell me that most reactions don't occur in perfect solvents, or stochasticity is a feature of an organism's life history, or some other hogwash.

James Thompson said...

I don't understand. Now that we have The Human Genome Sequence, why can't we have The Structure for all 60,000 proteins?

DK said...

This is a representation of an NMR structure of bovine ribonuclease A. It shows that various parts of the protein exist in several different conformations. The actual protein structure is a composite of all these structures in equilibrium with each other.

Strictly speaking, one can never know for sure whether the different models correspond to the real alternative conformations or simply indicate a degree of uncertainty in the experiment.

E.g., you do NMR of a crystalline protein and you still end up with a bunch of models. And the worse the data are, the more "breathing" will be "observed".

But I don't think anyone ever seriously thought that proteins are not dynamic. Crystallographers who output single models commonly find that the structures vary significantly depending on crystal form and crystallization conditions.

So do look at "the" structure - but keep in mind that it is only as good an approximation of a real thing as the quality of the data.

Sparky said...

DK makes an important point: an NMR structure set is not a veridical representation of the actual conformational ensemble. Rather, it represents the "best" 20 or so structures output by some minimization protocol based on restraints estimated from various kinds of data (NOEs, RDCs, chemical shifts, etc.). Areas of variability in these ensembles therefore typically represent stretches of sequence that have fewer restraints (but one must always consider how the structures were aligned) — strictly, they reflect uncertainty rather than dynamics. Because conformational fluctuations tend to reduce the number of restraints that can be determined, regions of variability in the NMR ensemble often correspond to regions of variability in the real structural ensemble. But, since the variability appears precisely because the conformation is underdetermined by the data, one should never trust that any one of these structures is "real" (in the sense that it is significantly populated in the actual conformational ensemble). That said, the extremes of the NMR structure set may correspond to the limits of the most probable conformations.

Vene said...

James, because we don't have the ability to model the interactions of individual amino acids yet. Plus, many proteins are modified after they are translated. Plus, many genes are modified after transcription (yay for splicing).

Anonymous said...

Great post! But like James, i fail to understand why can't we have The Structure for all 60,000 proteins, when we have already found the Human Genome SEquence. I am a college sophomore with a dual major in Biology and Mathematics @ University of California, Santa Barbara. By the way, i came across these excellent biology flash cards. Its also a great initiative by the FunnelBrain team. Amazing!!!

James Thompson said...

I'm glad that people agree that structures are important! I agree with Caroline in wanting more structures, although I was also being a little bit flippant in my original comment. The idea of The Structure is just a simplifying approximation, and really structures exist as ensembles. This is just like The Human Genome doesn't really correspond to the complete genome sequence of any human. It's a pieced together assembly from a number of different individuals, and we're still missing pieces! Fundamentally, genomes and structures are too complex to have a single solution (such as The Genome or The Structure) explain all of what we see. Once we have more data, I hope that we'll adjust our concepts and nomenclature accordingly.

I think that DK's point is very important, and I especially like that the implication that our models will get more precise and more accurate as we get better data. I also think that there is more uncertainty in published crystal structures than most people realize. A few years ago Tom Blundell's group published a software suite called RAPPER that made this point very clear. Their software takes a published crystal structure and the electron density map used to determine that structure. Their software outputs alternate conformations of protein structures that fit the experimental data equally well, even when the crystallographic data is of a high resolution.

Their papers on the topic are very good if people are interested in reading about it. Here's a link to their web page and a citation for one of their papers:

http://mordred.bioc.cam.ac.uk/~rapper/

# M.A. DePristo, P.I.W. de Bakker, T.L. Blundell (2004) Heterogeneity and inaccuracy in protein structures solved by X-ray crystallography. Structure (Camb). 12:831-838

CrystalCowboy said...

DK makes an important point: an NMR structure set is not a veridical representation of the actual conformational ensemble.

Hear hear! I worked on a crystal structure of a protein once, after three or more NMR papers had been published about it. The protein was too large for a complete NMR structure, but the NMR papers made a number of specific predictions. About half of them were too vague to be useful. ("There's a methyl group somewhere near residue X'). The other half were just plain wrong.

The next NMR paper, following the crystal structure, was considerably more accurate.

Also, inherent flexibility in the structure was revealed by comparing multiple protein molecules in the crystallographic asymmetric unit.

Bryan said...

"Great post! But like James, i fail to understand why can't we have The Structure for all 60,000 proteins, when we have already found the Human Genome SEquence"

I think the confusion here may be coming from the difference of a sequence vs a structure.

From the gene sequence we can determine the sequence of the resulting protein. Like genes, proteins are a strand of monomers stringed together into a line. The *sequence* of the protein's monomers (amino acids) is easily determined from the sequence of its gene.

Proteins are different from DNA in one key way - after that strand of protein is synthesised it folds into a complex, 3D structure. That 3D structure is critical for the function of a protein - fold a protein wrong and its just garbage.

The 3D structure itself is a product of two separate "structures". The first is what we call "secondary structure" which is basically a series of simple 3D shapes (tubular coils, flat sheets, etc). The second is what we call tertiary structure; basically the folding of the secondary structures into a more complex 3D unit.

The structure being talked about here is that 3D structure. The exact shape of that structure is determined by many things - the protein sequence, modifications, whether the protein is in a membrane or solvent, post-translational modification, even things like pH and osmolarity.

Determining that 3D structure is very hard, and we're a long ways away from being able to do it simply from computational models. These days the standard ways are to either crystallise the protein and bounce x-rays off of it, or to use magnetic resonance to map the vibrations of the molecule.

And just to make things a little more complex, there is something called "quaternary structure", which is what you get when multiple 3D proteins come together into a complex.

DK said...

Folks, can't we just agree that crystallography rulez and NMR sux? :-)

The Other Jim said...

DK said...
Folks, can't we just agree that crystallography rulez and NMR sux? :-)


Nope. We can't. ;-)

James Thompson said...
...why can't we have The Structure for all 60,000 proteins?


Where is 60,000 coming from?

James Thompson said...

Where is 60,000 coming from?

That was the red herring that was supposed to convey my sarcasm. The number 60,000 was an estimate for the number of genes in the human genome several years before we had a finished draft sequence. Sorry, written sarcasm is apparently difficult to pick up.

The Other Jim said...

James Thompson said...
Sorry, written sarcasm is apparently difficult to pick up.


Got it now ;-)

Mark Pallen said...
This comment has been removed by the author.
Mark Pallen said...

You might also want to consider that there is no THE Origin of Species, but at least six different variants. See this online variorum