Protein Crystallization Hits

When it comes to the design of optimization matrices, think 12, 4 or and 3. But NOT 10.
Here is why: Crystallization plates come in 3x4, 4x6, 8x12, 16x24 formats, not 10x10. I don't know what's the underlying reason for this non-metricity, but I suspect it may be related to the finding that the width of modern railroad tracks is based on history extending back to Roman (apparently it's not). Somehow we're stuck with this somewhat unwieldy dozenal (or is it duodecimal, or duodenary?) system.
Regardless, the protein crystallization plates we're dealing with are rectangular and typically have 24, 96 or 384  wells. Therefore designing crystallization matrices for such 4x6, 8x12 formats makes a lot of sense because they nicely fit into thes plates. When I start with an optimization matrix I usually pick the minimum and maximum concentration and then fill in the concentration for the remaining wells. The practical consequence is that for 96 well plates the numbers 11 and 7 assume a special meaning since those are the divisors used when calculating all concentrations in the well. Dividing by 12 or 8 does not give you equally spaced values because you're boxing in conditions 1 to 12 and there's not well zero. 

Multipliers for columns and rows to design protein crystallization gradient matrices that nicely fit a 8 x 12 screen into a standard 96 well protein crystallization plate.

If that's too much of a pain, I'd recommend checking out the E-screen builder. It does all of this on the fly and much more. Check it out.

Happy Thanksgiving,
Peter

SSGCID (Seattle Structural Genomics Center for Infectious Disease) is a NIAID-funded structural genomics effort targeting medically relevant proteins of class A and B pathogens. All the crystallography for SSGCID takes place here at Emerald Biostructures. While most structures of our targets can be solved by Molecular Replacement, a number of structure solutions require experimental phase information.

In order to obtain experimental phases, we recently have had very encouraging successes taking advantage of the strong anomalous signal that iodide has at CuKa-wavelength (f"=6.8e-), which is even stronger than the anomalous signal for selenium at the edge (f"=3.7e-). What was intended to be a test study yielded four new in-house phased structure within a month. As a start for this technique, we used a very bold approach: Crystals were transferred stepwise in a buffer containing up to 1M KI. The crystals were soaked in the high-iodide buffer for about 2 hours and then vitrified. Data were collected on our in-house diffractometer (Rigaku MicroMaxHF007 with a Saturn 944+ detector). Again, in a rather bold approach, we collected 360° of data in 0.25° frames with rather short exposures (~10s). Data were processed with XDS/XSCALE.

In our first case, a plasmid partition protein from Borrelia burgdorferi (BobuA.01478.a, pdb: 3k9g), crystallized in a tetragonal space group using NaCl as the precipitant, which was in two steps partially replaced with KI. A 25-fold redundant data set was collected, which showed an anomalous signal out to 2.25Å resolution. Seeding Phaser_EP with 2 strong sites, a total of 16 sites were eventually found and refined. After density modification with PARROT the electron density was readily interpretable (see picture at the end), and almost the entire model was built with BUCCANEER, which then could be extended with ARP/wARP. Further analyses revealed the importance of data redundancy: 180° of data yielded about the same results as 360°, while 90° of data would not yield the structure quite as readily. Other program packages, such as SHARP + ARP/wARP or PHENIX could also solve this structure easily.

In a second case, a deoxycytidine triphosphate deaminase from Anaplasma phagocytophilum (AnphA.00973.a, pdb 3km3), crystallized in a rhombohedral space group and yielded a 10-fold redundant data set up to 2.1Å resolution. Here, the anomalous signal was not as strong as in the previous case. Phaser_EP was seeded with 13 strong and weak iodide sites, eventually 16 sites were found and refined. Again, almost the entire model could be built automatically.

Within the past month, these two structures and another two structures could be solved using anomalous phase information from iodide. Structure #3 yields very well diffracting however twinned crystals; structure #4 only diffracted to 3Å, however phase combination with a weak molecular replacement solution easily yielded the structure. So far, we have only used rather short soaks with 1M KI. Further tests are planned to optimize for lower iodide concentrations.

These results indicated that iodide anomalous phasing can be used as the first shot for determining a new structure that requires experimental phases. With well diffracting crystals and high symmetry, structures can be built within hours after data collection. This technique has shown to work with a twinned data set; for another case a weak Molecular Replacement solution was of trememdous help. Adavantages of iodide soaks are:

• no growth of Se-Met labelled protein is necessary;
• strong anomalous signal in-house;
• high concentrations of iodide are possible during soaks;
• iodide can occupy both positively charged/polarized and hydrophobic pockets, less of a search than with heavy metals;
• it really works!

experimental electron density after in-house iodide phasing

-

Here's a big "Thank You!" to everybody calling in or sending emails of encouragement and congratulating us on the recent changes regarding our business. Early this week we had announced that the Emerald Businesses are separating from deCODE. See here for the actual News Announcement and details.

At Emerald we're all happy and excited about this change. The bottom line is that Emerald BioSystems will continue to provide the protein crystallography community with:

• sophisticated laboratory automation such as the Plug Maker for simple low-volume protein crystallization,
• bioinformatics software such as Gene Composer, check here for a free trial download or the online screen design tool E-screen Builder,
• tried-and-proven high quality protein crystallization reagent kits such as the Emerald Wizard Suite
• and protein crystallization plates such as the Combi Clover Jr. plates for sitting drop protein crystallization and microfluidic CrystalCards for plug-based crystallization.

And of course, this blog Protein Crystallization Hits will continue to be a growing resource for all things related to protein crystallization.

Thanks for the congrats and your encouragement!
Peter

I'd estimate that more than half of all protein crystallization experiments employ protein that has a poly-Histidine tag. These are usually hexahistidine tags at the N-terminus or at the C-terminus of a target protein to aid and simplify their affinity purification with metal chelate chromatography (IMAC). Since it is cumbersome and costly to remove such tags, many crystallographers keep the extra amino acids that come with the tag and subject the protein to crystallization trials without His-tag removal.

What is to be done if N-terminal His-tagged proteins with a TEV-proteolytic site don't yield crystals or the grow crystals with poor X-ray diffraction quality only? Dong et al. have come up with an ingenious protocol, in-situ proteolysis.

Simply put, crystallization experiments are prepared in the presence of trace amounts of a protease, for example chymotrypsin. This works for target proteins that have a recoginition site for TEV-protease, such as in MGSSHHHHHHSSGRENLYFQG or MGSSHHHHHHSSGRENLYFQGH where chymotrypsin cleaves at Tyr and Phe.

Since the authors were affiliated with high-throughput crystallography operations they were able to process many targets with their new in-situ proteolysis protocol. Altogether they subjected 55 different bacterial proteins (20 had previously failed to crystallize and crystals of the remaining 35 were no good) to the in-situ proteolysis protocol and ended up with 9 rescued targets. Not bad, converting 55 crystallization recalcitrant proteins into well diffracting crystals, yielding 9 new structures.

This is their in situ proteolysis protocol:

α-chymotrypsin (SigmaC3142), dissolved at a concentration of 2 mg/ml in a solution containing 1 mM HCL and 2 mM CaCl2, was added to the purified protein on ice immediately prior to crystallization trials at a ratio of 1μg chymotrypsin per 100 μg of histidine-tagged protein, dissolved at 10-20 mg/ml in 10 mM Hepes, pH 7.5 and 500 mM NaCl. Crystallization was performed in sitting drops at room temperature, adding 0.5 μl of the protease/protein mixture to 0.5 μl of the precipitant. Crystallization trials were set p immediately, without assessing the efficacy of the proteolysis, without stopping the roteolysis reaction, and without purification of any proteolyzed fragments.

In the supplemental material the group describes several examples of target proteins that crystallized only in the presence of the proteases or cases where the X-ray diffraction quality was improved. The description of the crystallization stories in the supplemental material make for a great read.

Here's a summary of the different cases encountered:

• SCO6256 failed to crystallize and in the presence of chymotrypsin two different crystal forms were obtained, one of them diffracting well enough for phasing and 2.5 A structure determination
• SCO4942 did not yield crystals with conventional crystallization screening efforts but yielded 2.8A-diffracting crystals after subjecting to chymotrypsin treatment
• NE2398 crystals grew to reasonable size but did not diffract. When re-screened in the presence of chymotrypsin, 1.8A X-ray diffracting crystals grew.
• ATU0899 grew as small needle cluster crystals that did diffract to only 5A; after in-situ proteolysis crystals diffracted to 1.8A
• ATU2452 grew small needle clusters (sea urchins) and the chymotrypsin treated protein grew as nice diamond-shaped crystals that diffracted to 2.5A
• ATU0870 grew as multiple crystals with his-tag and tag removed diffracted to 2.5A. When crystallized in presence of chymotrpysin, 1.95A X-ray diffraction data was obtained.
• HP0029 grew as thin needles as his-tagged protein, best diffraction was 3.2A; removal of his-tag using TEV failed; with chymotrypsin treatment 1.47A data was recorded and structure determined
• ATU0299 grew as un-optimizable needles that diffracted to 1.8 A when crystallized as a mix with chymotrypsin
• ATU0434 crystallized in 8 cocktails but diffracted only to 3.5A; with chymotrypsin treatment crystals did not get any better, but with trypsin treatment crytsals grown diffracted to 2.3A.

Interestingly, the authors revisit the structures they obtained with the in-situ chymotrypsin treatment and they make hindsight guesses as to why the crystallization of the untreated proteins did not work. For several cases 'it was difficult to ascribe a mechanism by which proteolysis facilitated crystallization in most cases' but there were several exceptions. One in particular showed that extended N-terminus would have disrupted packing in the obtained crystal form.

Check out the optimized 96-formulation crystallization screen to be used for such in-situ proteolysis experiments.

When you're dealing with an otherwise hopless case, a 16% rescue rate is not that bad.

Cheers,
Peter

P.S. just found the update to this paper, with better statistics. May post on this later.

This week I came across an interesting article reviewing the application of protein crystallization 'chaperones' in the magazine Drug Discovery World 2009. Chun-wa Chung from Glaxo Smith Kline has given her review the provocative title "will crystallization 'chaperones' make all proteins crystallisable?". "Intriguing", I thought - "what's the deal with this chaperone assisted protein crystallography"?

'Crystallization chaperones' are protein molecules that are combined with target proteins to help the nucleation and growth of crystals. The resulting crystals contain both, the target protein and the molecular 'chaperone'. Chun-wa lists several examples where this technique has worked well, notably, ion channels and GPCRs. The current chaperone repertoire is limited to monoclonal antibodies (mAb), fragments (Fv, Fab) of monoclonal antibodies, single chain antibodies (sFv) and other non-antibody proteins such as DARPINs.

While complexes with these chaperones can be grown like any other hetero-dimeric protein complex, it is a tedious process to generate a suitable binder. The fittingly named company Molecular Partners apparently offers this as a service. But you really need a good reason to go this route. Good reason such as: "high-value drug target that failed in conventional high-throughput screening assays", think GPCR (G-protein coupled receptor). You better know that it's worthwhile the extra effort and the substantial detour to generate a specific molecular chaperone to crystallize a particular target protein. I'd be very interested in seeing a large-scale, side-by-side study to determine which path gives you crystals of sufficient quality faster and with less effort:

A) focus on your target protein at hand; learn the ins and outs of it, generate many constructs and set up a lot of crystallization trials with ligands, or

B) generate a specific binder for your target protein at hand and set up crystallization trials with the binder:target protein complex.

Chaperone. Why should I need one?

In her conclusion Chun-wa answers the initial question with a clear 'no' - crystallization chaperones will not make all proteins crystallisable. However, once the detour and extra effort become less of a burden, these molecular chaperones may be used more commonly than now and help crystallize proteins that have otherwise proven recalcitrant to traditional crystallization. Until then, molecular chaperones will be applied to challenging integral membrane proteins where all other 'tricks of the trade' (I'm so glad she's not invoked the 'art'-word) have failed.

Cheers,
Peter