Improving the Crystallization Process for Optimal Drug Development

Macromolecular X-ray crystallography is an important and powerful technique in drug discovery. Studying the specific interactions of a particular drug with its protein target at the atomic level can help improve the drug design process.

Hundreds of different crystallization conditions need to be tested against a macromolecule of interest to grow crystals of sufficient size and quality and enable successful structure determination. With advanced synchrotron X-ray sources, crystals as small as 10 μm or less can now be used for X-ray diffraction, but typically various conditions are tested to obtain crystals of 50 μm or larger in size, to make the data collection easier and more robust.

The crystallization process can often prove inefficient, with only one or a few crystals produced from several hundred test conditions. Occasionally these crystals will be of sufficient quality for X-ray diffraction analysis and permit determination of the structure, which eliminates the need for further optimization of the conditions. However, in most cases the original “hits” may not form large enough crystals, only form precipitates, or generate poor or no diffraction at all. In these cases the crystallization conditions of the original hits need to be optimized. Obtaining these optimized and high-resolution diffracting crystals from targets of interest is vital to determining accurate molecular structures of these molecules.

Crystallization process

The optimal crystallization condition relies on multiple and complex screening combinations. This can include the optimization of variables such as pH and concentrations of reagents and proteins in the hit condition, for example, by creating a gradient of the components across the crystallization plate.

Figure 1 – a) dragonfly and b) mosquito Crystal.

A dedicated automated robot such as the dragonfly™ (Figure 1a) from TTP Labtech (Melbourn, U.K.) can be used to set up these gradients. Optimization of 4–9 different components including diluent can be set up using the device’s 5- or 10-pipetting head format into a 48- or 96-well crystallography plate. Even with highly viscous fluids, the dragonfly is capable of accurately dispensing a minimum volume of 0.5 μL from a 10-mL reservoir tray with zero cross-contamination due to its positive displacement dispensing. The process is quick and easy, taking less than 5 min to make a gradient of four components in a 96-well plate.

However, if a dedicated optimization robot such as dragonfly is not available, the mosquito® Crystal (Figure 1b) (TTP Labtech), by itself, is capable of setting up gradients of components directly into the drops. A protein sample and each component of an original hit condition can be “multiaspirated” at different volumes in the same pipet tip, prior to being dispensed simultaneously, with additional mixing if required. The y-axis of the plate can be varied by loading eight different concentrations of a protein solution into a column of a protein reservoir strip, and concentrations of different components or pH can be varied in the x-axis.

This same concept can be applied for either the hanging- or sitting-drop method. Compared to optimization of hits manually, which are subject to inherent manual pipetting errors, this method is much faster and more accurate, and will save on reagents.

Experiment 1: Optimizing proteinase K crystals

In an experimental study, developed by Michael Collazo, Lab Manager at the UCLA Macromolecular Crystallization Facility (Los Angeles, CA), proteinase K crystals were optimized in a hanging-drop format using the mosquito Crystal. This was achieved by setting up a gradient in the drops and placing them over a reservoir of the original condition (0.1 M tris, pH 7.5, 1.2 M AmSO4) using the multiaspirate function. In this example, the protein concentration was varied in the y-direction (50 mg/mL to 30 mg/mL in 0.1 M tris HCl buffer, pH 7.5, in eight rows of a 96-well plate) and the concentration of ammonium sulfate in each drop was varied from 0.3 M to 0.8 M across 12 columns in the x-direction. One-microliter drops (1:1 of protein to condition) were set up and left to equilibrate over the reservoir solution, the plates were imaged for crystal formation, and the conditions producing the best crystals were determined (Figure 2).

Figure 2 – Proteinase K crystals: a) obtained in one of the original drops (200–250 μm) and b) an optimized single crystal, well defined and larger in size (400 μm).

Unlike conditional manual optimization or optimization with a dedicated robot (e.g., the dragonfly), here the drops are equilibrating against an “alternative” reservoir, which is the original hit condition. This is because the reagents are multiaspirated from columns of a source plate and mixed directly with the protein, rather than having the components mixed in a source plate well reservoir.

Experiment 2: Optimizing crystals of a novel proteobacteria protein

In a second experiment, crystals of a novel protein found in carboxysome-containing autotrophic proteobacteria were optimized using this method. These proteobacteria are potentially useful as bioremediation organisms (naturally occurring organisms that break down hazardous substances into less toxic or nontoxic substances). Modifying this protein could enhance productivity of the organism and optimize its suitability for contaminated waste treatment.1,2

The protein was cloned and purified in collaboration with Nicole Wheatley of UCLA. The original crystals were obtained in a hanging-drop format containing a 1:1 mixture of 25 mg/mL protein (in a buffer of 10 mM tris, pH 7.6, 50 mM NaCl) to a reservoir condition containing 4 M (NH4)2SO4 and 1 M bis-tris, pH 5.5 (Figure 3a). For this optimization, the concentration of the protein was kept constant, while the pH was increased on the y-axis of the optimization plate and the concentration of the AmSO4 was increased on the x-axis. Previously, both manual and optimization under oil were set up, but only very poor crystals were obtained where the needles were too thin to even harvest.

Figure 3 – Crystals of a novel protein: a) obtained in the original drop, resulting in 2.5 Å resolution, and b) optimized crystals (resulting in 2.1 Å resolution).

Diffraction data were collected on some of the best single and sharp-edged optimized crystals at the UCLA X-ray crystallography core facility at a wavelength of 1.54 Å. The optimized crystals resulted in improving the resolution from 2.5 to 2.1 Å (Figure 3b). While the refinement resulted in larger crystals of 625 μm compared to 200 μm in the original hit, the best diffracting crystals were not necessarily the largest.

Conclusion

The above examples demonstrate that the refinement of the crystallization conditions is improved by automation. Although dragonfly is the instrument of choice for fast and easy high-throughput optimization, the mosquito Crystal can successfully be used for both initial screening of the crystallization conditions and also for the optimization of the conditions, required to obtain high-resolution diffracting crystals.

References

  1. Pol, A.; van der Drift, C. et al. Isolation of a carbon disulfide utilizing Thiomonas sp. and its application in a biotrickling filter. Appl. Microbiol. Biotechnol.  2007, 74(2), 439-46.
  2. Liao, B.; Ji, G. et al. Profiling of microbial communities in a bioreactor for treating hydrocarbon-sulfide-containing wastewater. J. Environ. Sci. (China)  2008, 20(8), 897–9.

Michael Collazo is Laboratory Manager, UCLA Macromolecular Crystallization Facility, Los Angeles, CA, U.S.A.; Soheila Vaezeslami, Ph.D., is Field Application Scientist, and Sarah Burl, Ph.D., is Scientific Communications Officer, TTP Labtech Ltd., Melbourn Science Park, Melbourn, Hertfordshire SG8 6EE, U.K.; tel.: +44 1763 262626; fax: +44 1763 261964; e-mail: sarah.burl@ ttplabtech.com ; www.ttplabtech.com. The authors thank Nicole Wheatley, graduate student at the Dr. Todd Yeates Laboratory, UCLA, for supplying the novel proteobacteria protein used in this work.

Comments