Protein Crystallization Hits

There's so much progress in membrane protein crystallography generally, and in GPCR crystallography in particular, that I've started a new blog, called GPCR blog. I'm planning to cover topics relating to GPCR structural biology via the Emerald BioStructures website (note the URL: http://emeraldbiostructures.com/gpcrblog). For anybody who's interested in experimental GPCR crystallization conditions, these blog posts may be interesting to read: GPCR Crystallization Conditions  and GPCRs of known structure.

This is a good opportunity to attempt a quick look into the future and anticipate what's yet to come in our field. It is indeed amazing to see all the progress that's been made in the field of membrane protein research within the past ten years. When I decided to join the membrane protein structure research field - this was around the mid nineties - I was warned that this is a super high risk field, may derail my dreams of getting a PostDoc position and not land me a job in either academia or industry. And now, years later, there are so many more crystallographic membrane protein structures - getting close to hundred (depending how you count) and there are substantial efforts in the pharma and biotechnology industry to apply these difficult targets to crystallography based ligand discovery and to lead optimization in drug discovery programs. This is similar to what happened in the late 80ies to soluble protein structure research. Our game has changed dramatically since then, hasn't it?

It feels like a time warp looking at all the progress made in membrane protein structural biology.

So, what's next?

To me the next big steps in structural biology are about scale context in time and space. What does that mean?

1. A better understanding (with atomic resolution, of course) of the detailed dynamics within protein molecules as they go about their work. More precisely, an experimental understanding of how the motions of atoms and their bond rotations & translations taking place in the sub-nanosecond timescale create effects that manifest themselves in what we call "biochemical function" at the milli to second timescale (6-9 orders of magnitude scale difference).
2. The integration of structural data from atoms up, to explain the appearance of macroscopic structures such as cells and organisms (again ca. 6 orders of magnitude).

Of course both of these fields rely heavily on computational tools and require a lot of input by experimentalists as well, providing reality checks and help keeping the models grow better.

My 2 ct,
Peter

What's important to know when switching from crystallizing soluble proteins to crystallizing membrane proteins? I've compiled a list of points that I've made in the past when attempting to answer this question.

1. Go nano volume: Sample preparation involves the use of solubilizing detergent, and membrane proteins are notoriously unstable - unless you or your biochemist friend has worked out "conditions" (buffer, lipid, additives, temperature...) that keep the membrane protein from aggregating. This is all about getting the biochemistry right and often requires a lot more effort since standard conditions that are typically applied for soluble proteins may not be sufficient to keep the protein sample alive for a period of time that's compatible with crystal formation. Due to sample loss and cost in most cases you'll start with substantially less protein sample volume than what you're used to. Don't even think about uL-sized crystallization experiments.

2. Set up more crystallization experiments: Get used to screening more extensively, preparing more crystallization experiments and geting fewer crystal hits. Compared to soluble proteins, there are more parameters to screen for. This is due to the presence of an additional component, amphiphiles (detergent type, concentration, lipids...) and their complex behavior in solution. This dramatically increases the dimensionality of the already multidimensional protein crystallization phase space.

3. Spend more time at the microscope. The phenomenology of drop content is, how shall I say it without discouraging you, 'richer'. There are separate detergent rich phases that can look like oiled out protein, some phases are turbid and there are detergent crystals devoid of any membrane protein that may get you on the wrong track. Some membrane protein crystals may not even have proper facets.

You see, this is a funner game.

Of the Practical Membrane Protein Crystallization Tips listed here I think these 3 are the most useful:

4. Membrane proteins often require harsh detergents for their extraction out of the native membrane. Often crystals grow better with milder, shorter chain detergents.

5. Try to control detergent concentration (measuring it and reducing it). Often the detergent concentrates with the membrane protein and when low MWCO filters are used for sample concentration.

6. Start with crystallization screens that are rich in PEG as opposed to salts. For example the Ozma series of Emerald BioSystems' crystallization screens.

And finally:

7. Read this "Pedestrian guide to membrane protein crystallization" by Michael Wiener (Methods 34, 364-372, 2004).

8. By all means, explore non-traditional crystallization experiments that have worked for a number of membrane proteins. For example, utilizing bicelles or lipidic cubic phases (see primer here, and the Cubic LCP Kit).

Enjoy,

Peter

What would you give if you knew how the crystallizability of your target protein compares to 'what's our there'? There's a lot of talk about stability, crystallizability and their relationship and there are these hand waving arguments about supposedly problematic 'floppy regions' in proteins.

So here's a relevant paper that sheds solid data on this topic:

Price W.N. et al., Understanding the physical properties controlling protein crystallization based on analysis of large-scale experimental data. Nature Biotechnology 27(1), 51-57. 2009

 The authors thoroughly mined crystallization data from NESG (Northeast Structural Genomics Consortium) and, amongst a lot of other interesting results, present evidence for these key findings:
1. Overall thermodynamic stability (thermal melt) is not a good predictor of crystallization success
But here's the good news as well: Higher crystallization propensity is found for:
2. Proteins that form defined dimers and higher-mers (as opposed to monomers or aggregated protein)
We DLS lovers always knew this, of course ;)
I was not satisfied though with this hyperbole that the paper concluded with: "The dominant factor determining protein crystallization outcome is the prevalence of well-ordered surface epitopes capable of mediating stereochemically specific interprotein packing interactions" - to me this sounds like: "if the molecules pack well with each other, they'll form crystals".

Duh - "The dominant factor determining protein crystallization outcome is the prevalence of well-ordered surface epitopes capable of mediating stereochemically specific interprotein packing interactions."

Nevertheless, their notion that Glycine, Alanine and Phenylalanine residues are 'good' for crystallization may serve as a useful guide when designing protein surface mutations to enhance your targets' crystallizability. Methinks that this will form a centerpiece in CAPCE (computer aided protein crystal engineering).

Cheers,

Peter

Some protein structures are good, others less so. Here's an all time classic that explains how we're supposed to tell the two apart:

Klywegt G. & Brunger, A. "Checking your imagination: applications of the free R value". Structure, 4(8) 897- 904 (1996)

Gerard and Axel "review applications of the free R value, and discuss practical issues and caveats related to the use and interpretation of this statistic." Then they go on to "... present a survey of free R values of published X-ray crystal structures of macromolecules." Dry stuff, I admit it. Mind you though, checking data quality and putting your own data in context with everything else that's out there is a key element of what we do. And it's a reminder of what we need to do to make an impression that lasts for generations to come. In light of the recent fabrications our field has seen I have come to embrace such self-policing efforts that can hardly be valued enough.

Here's an interesting vignette about the quality of protein crystal structures (Acta Cryst (2007)D63, 941-950). Brown and Ramaswamy conclude their retrospective analysis in the PDB hat protein structures published in high impact journals are worse than those in low-impact journals. Here it is in their own words:
" The most striking result is the association between structure quality and the journal in which the structure was first published. The worst offenders are the apparently high-impact general science journals."

Wow. How can that be? My own explanation is that sometimes high-impact structures, published in high-impact journals, are determined and authored by researchers that have little experience in X-ray crystallography as they have focused their work on very difficult biology / biochemistry projects for a long time. The X-ray diffraction & structure determination is then done with simple-to-use software tools, enabling biochemists that have not even yeard of Bragg's law to 'do a protein structure'.
Take my room mate for example, with whom I shared 10 stressful days at Cold Spring Harbor, trying to learn the basics of Macromolecular Crystallography back in 2000. On the last day at BNL we walk by a beamline that has the cover of Cell posted on their hutch - that was his structure. 

Beware of protein structures in NATURE and SCIENCE? The lower the impact factor the better the structures.

I'm not saying that there's anything wrong with newcomers publishing protein structures - don't get me wrong here! The point is that high-impact structures are often the result ofbreakthroughs in biochemistry and crystal growing rather than advances in structure determination technologies. And therefore researchers trained in biochemistry get to determine structures to get their first author paper. Researchers with little experience in X-ray crystallography. The solution to this? Get a buddy. Make friends with a crystallographer who has seen the pitfalls of interpreting noisy maps, overrefining and trusting output file statistics. In my view, having a seasoned crystallographer co-authoring a structure paper adds to credibility. More so than the name of the publishing journal.

<this is where the original post ended. Thanks to Roger, here's a twist to this interesting story: 

Turns out that when several factors (such as size, resolution, asymmetric unit volume & deposition date) are matched in controls and are corrected for, the quality of protein structures published in high impact journals is not much worse than those published other journals as described by

Randy Read and Gerard Kleywegt who followed up on their earlier research and show in Case controlled structure validation Acta Crystallogr D Biol Crystallogr. 2009 February 1; 65(Pt 2): 140-147 this table: 

Only slightly poorer validation statistics of protein structures published in high vs. those in low impact journals. Take into account size, resolution, asymmetric unit volume & deposition date (done by matching controls) and the differences disappear.(Original Table)

Higher impact journals seem to publish protein structures with lower resolution, but the effect is small. What do you think? Does that bother anyone? (original image)

They show that there are serveral factors that were not analyzed in the earlier study -e.g. the size of the structure, date of publication etc. - and once included, the effect 'high-impact journals having lower quality protein structures' became very small.

Thanks for pointing out this paper, Roger! > 

Cheers,
Peter

 There's really no excuse anymore not to know what you want to know. With all the massive amount of information out there on the internet it is fairly easy to learn new techniques and keep up with improved lab procedures. It's 'just' a matter of finding the gems.

Here's an example regarding Cryocrystallography, provided by Jim Pflugrath (with whom I had the honor of hands-on learning how to flash-cool harvested protein crystals, amongst other things) and Any Howard. Jim and Andy provide an overview and step-by-step info on cryocrystallography in the form of powerpoint slides & PDFs that they had apparently used during the  2008 ACA summer school . I particularly enjoyed the slides explaining proper adjustment of the cryostream, how to rescue iced-up crystals, quick-and dirty annealing and bulk solvent density matching. There are many useful tips and tricks there. Information that often does not make it into the published literature (despite new attempts to do precisely that). And some of the useful practices covered in the slides may have not yet trickled down into every crystallography lab. So check out these slides.  

I'm grateful that the organizers kept these slides and a lot more up on the web, for everybody to learn about the intricate details of cooling crystals. And there's more: Check out the 2008 lecture notes  covering topics such as safety, lattices & symmetry, scattering & diffraction, synchrotron radiation, conventional X-ray sources, Phasing, molecular replacement,  density modificaiton, anomalous scattering, twinning, noncrystallographic symmetry. And notes on practicals covering cryocrystallography, data processing, syncrotron data collection, direct methods (BnP),  model building with Coot, interpreting international tables, construct design and radiation damage.

What a feast.  

Cryocrystallography by Powerpoint. If it were that easy...

Here's a big 'Thank you!!!" to the ACA school and the contributors for providing and putting up this content on protein crystallization and crystallography on the web. For free, for everybody to see. 

Thank you!

Peter