Protein Crystallization Hits

Most of the detergents that are in use today were available 5 or 10 years ago. Maybe the purity has increased over time, but the list of detergents extracted from successfully crystallized membrane proteins really hasn't seen any substantial additions in the past few years. In the previous blog "Applying the 80/20 rule to membrane protein crystallization pre-screening" I showed a list of detergents and referenced statistics that show that with just 8 detergents (dodecyl maltoside, decyl maltoside, nonyl glucoside, octyl glucoside, LDAO, C12E8 C12E9 and undecylmaltoside) about 80% of the membrane protein crystallization 'landscape' is captured.  If you're not working on beta barrel proteins you may as well remove the polyoxyethylene ethers and do with six detergents only, three of them being maltosides.
Only rarely new detergents are conceived, synthesized and tested with membrane proteins. Pil Seok Chae, Sam Gellman and others have done just that, and more: they've come up with a great new detergent class. This is their recent paper on their new series of detergents, called MNGs (maltose-neopentyl glycol amphiphiles):

Chae, P., Rasmussen, S., Rana, R., Gotfryd, K., Chandra, R., Goren, M., Kruse, A., Nurva, S., Loland, C., Pierre, Y., Drew, D., Popot, J., Picot, D., Fox, B., Guan, L., Gether, U., Byrne, B., Kobilka, B., & Gellman, S. (2010). Maltose–neopentyl glycol (MNG) amphiphiles for solubilization, stabilization and crystallization of membrane proteins Nature Methods, 7 (12), 1003-1008 DOI: 10.1038/nmeth.1526

The MNG molecule looks like this:

Figure: Welcome to the detergent family, MNGs:  maltose-neopentyl glycol amphiphiles. Reserving a spot in the front row on the detergent shelf in the freezer may be a good idea.

The molecule architecture, with the neopentyl in it center, was deliberately designed to provide a more rigid environment for membrane protein crystallization. Looking at the list of preferred detergents, the choice of maltosides (sic!) as a head group seems to be a non-brainer  (this is in retrospect, mind you), but the concept of a quarternary carbon is unique. The successful applications of this new detergent series range from stabilization (activity of beta 2 adrenergic receptor after detergent exchange from DDM), thermostability (of melibiose permease),  solubilization (leucine transporter, light harvesting complex and photosynthetic reaction center) and crystallization (cytochrome b6f & b2AR). While an improvement of crystal quality for cytochrome b6f was not seen, the authors mention that MNG-3 aided in improving crystals of agonist bound beta 2 adrenergic receptor. 

Overall the behavior of the MNGs seem to be on par with DDM (dodecyl maltoside), the currently highest ranked detergent in terms of use for successful crystal growth leading to high-resolution X-ray structures.

I raise may glass to Pil Seok and Sam: well done!
Peter

In an ideal world, this would be a fairly simple way to grow protein crystals for X-ray structure determination purposes:

Start out with ca. 400 ul of filtered target protein solution, in 20 mM Hepes buffer neutral pH and maybe 50 mM NaCl , freshly purified to ca. 95% purity and concentrated to 10 mg/ml.

1. Snap freeze four x 25 ul aliquots of protein samples in thin-walled PCR tubes and store at -80C. These sample will be used in a week for protein crystallization optimization.

2. Document the quality of the protein sample via SDS-PAGE and determine its concentration (ca. 5 ul). Measuring UV absorption (OD280) and calculating the protein concentration is good enough.  You'll need ca. 5 ul for such an OD measurement (depending on cuvette path length and dilution). Check this online tool to get an estimate for the extinction coefficient based on sequence.

3. Now let's get out the multichannel pipettor and set up all of the remaining 200 ul of sample solution as 1ul + 1 ul sitting drop vapor diffusion crystallization experiments. A great first pass would look like this:
Plate 1: Wizard 1 & 2 (96 drops)
Plate 2: Wizard 3 & 4 (96 drops)
By the way, Emerald offers these reagents as the Wizard Suite. This is a 192 point, non-overlapping, sparse matrix that has a proven track record to yield crystals in first-pass protein crystallization exploration trials.

And while we're at it: go with the Compact Jr. plates. These plates can be sealed with clear tape and drops form nicely on the hyrdophobic polypropylene surfaces.

4. Store the crystallization trial at room temperature, minimizing temperature fluctuations and vibration. Observe right away, after 1 day, 3 days and 1 week.

5. The initially clear drops with now contain precipitate, some clear drops and a few with clustered microcrystals or needles in them. If the corresponding well-solution does not contain similar crystals, chances are that you've grown protein crystals! Since these crystals are likely to be too small to diffract them using your home-source X-ray generator and detector system, you'll need to optimize crystallization conditions and grow larger crystals.

6. There are many different ways to optimize crystallization conditons, and depending on prior knowledge you may want to carry out seeding experiments, include additives or change the treatment of the protein sample (i.e. filtration). Here's a simple optimization schema: create a grid screen around all conditions that gave you crystals. This rational crystallization optimization schema works great since it separates the effect of pH, salt, precipitant and protein concentration). You should use the online ScreenBuilder design tool to create an optimization screen. Your fellow researchers at Emerald BioSystems  are happy to prepare and send such a customized optimization screen to you.

7. Set up 96 - follow-up optimization 1 ul + 1 ul crystallization experiments with the saved protein material that you thaw in your fingers. If everything goes according to plan, crystals of different sizes will grow.

8. Harvest a crystal, cryoprotect, diffract and determine its structure ;)

Enjoy,

Peter

Protein crystals often form at air/water interfaces, on the container wall or on dust that's present in the crystallization drop. Why not expand this concept and throw some sand, horse hair or seaweed into crystallization setups? That's exactly what Thakur et al. have done and describe in their PLOS paper. While Allan d'Arcy has previously published on the use of natural seeding materials for nucleation of protein crystallization, this paper describes a systematic study to improve the success of sparse matrix protein crystallization screening with heterogenous nucleating agents. They find that hydroxyapatite, cellulose, horse hair and dried seaweed promote crystal formation. This apparently works even better once you combine all of these particles in a cocktail. The authors encourage crystallizers to further explore and identify even better materials for this purpose.
How about a few springkles of sawdust, dandruff or couscous? I suppose that this is one of those cases where "who crystallizes is right".

All the best,
Peter

When I look at different structural biology labs with their different crystallization processes and equipment they employ I see a wide diversity of techniques used and processes installed. At the extreme end of the spectrum there is the individual investigator working on the structure determination of a single or several related proteins at a time. At the other side are the high-throughput crystallography labs, churning through thousands of constructs and producing hundreds of new X-ray structures. Most labs are somewhere in between and have fine-tuned their operation based as a result of the given resources, expertise and available technology.

At one point in this spectrum there is a peculiar transition, akin to a phase change where interesting things happen. Here the one-investigator-one-target model is replaced by a process-management model. Evidently, the success rates of the high-throughput operations are a lot lower than that of smaller labs. On the other hand, the cost /structure has been continuously driven down by the high-throughput model, making it possible to generate X-ray protein structures at well below $100,000.

Why is there this decline in success rates? One reason is that the 'one-size-fits-all' approach of high-throughput structure determination explicitly accepts failures. Some targets are just not compatible with the particular crystallization regime chosen, the latter of which would be typically varied when handled as a single entity. I suspect though that during the transition to high-throughput, researchers manage processes in a more abstract way and something gets lost. For example they typically don't know precisely what is happening to each individual protein in the pipeline. I think that the intimate knowledge of a protein's characteristics in expression and purification does contribute to success in crystallization.

In addition there may also be a motivational issue. You'd do everything in your capacity to grow better crystals from CLO's (Crystalline Looking Objects) because that's the one target you're focusing on (think back to your days as a PhD student). A process manager however would be more likely to triage a weak hit and instead order a new construct or crank the high-throughput machine a little faster and create more output that way.

I wonder if there's a way to combine the advantages of both regimes. Can we bring back this bond that exists between the person carrying out the crystallization experiment and the target protein? Any ideas?

Cheers,

Peter

The main use of protein structures in pharma and biotech is of course guiding medicinal chemists in their effort to synthesize better compounds to bind to target proteins. To do this, it is necessary to get structures of protein-ligand complexes. In my discussions with industrial crystallographers I sense that these days more than half of all drug discovery projects are supported by X-ray crystallographic efforts. Hence drug discovery efforts can be effected in lead discovery, for instance via fragment-based screening or in the later phase of lead optimization. Both methodologies require the ligand to associate with the protein either before crystallization or after crystallization. The reality is though that the corresponding experiments, co-crystallization or ligand soaking don't always work. There are lots of reasons why they may not work. What's the back-up plan when soaking or co-crystallization experiments fail?
Cross-linking crystals and soaking them with ligands has worked for PDE10 (human phosphodiesterase 10a) - see this reference. Initially only apo PDE10 crystals could be grown. Ligands did not show up in the corresponding X-ray structures, neither after soaking apo crystals with ligands nor by crystallizing PDE10 together with the ligand. Once the PDE Crystals were cross-linked with glutardialdahyde though, and ligands were soaked in, X-ray structures with ligand bound were obtained.

Another trick in our toolbox,

Peter