Crystallography: phasing out heavy atom derivatives

REFERENCE: Liu et al. "Structures from Anomalous Diffraction of Native Biological Macromolecules." Science (2012) 336, pgs 1033 - 1037.                 

        The field of protein crystallography has partially pulled back the veil that lies between the macroscopic and microscopic worlds.  Coupling crystallography with X-ray diffraction yields structure models that allow scientists to consider the shape, surface, and, in short, the angstrom-length details of a protein.  Each spot on a diffraction pattern arises from a constructive interference event between X-ray waves within the protein crystal.  Amplitude and frequency information, two of the three variables that define a wave, are provided by these spots, but the final parameter of phase must be experimentally determined before a protein model can be built.  “Solving the phase problem” has thus far relied on soaking proteins in heavy atoms, use of experimental phases from a closely related protein, or incorporation of selenomethionine into the polypeptide chain.  Scientists would rather determine structures using crystals of native protein but such a method has thus far been unavailable.  In a recent study published in the journal Science, Liu et al. discuss their new technique that uses anomalous dispersion from native atoms within the protein to obtain phase information; the authors further support their technique’s worth by solving the structures of four different proteins that vary in size and subcellular localization.

                Anomalous dispersion relies on a heavy atom’s ability to absorb and emit X-rays.  When the wavelength of the X-rays approaches the characteristic emission wavelength of the atom, then the absorption drops off sharply giving rise to anomalous scattering.  With access to tunable synchrotron beamlines and protein crystals bearing heavy atoms, crystallographers using multiple anomalous dispersion (MAD) or single anomalous dispersion (SAD) can gather data sets at their heavy atom’s absorption edge to identify the heavy atoms’ locations and eventually provide the phases necessary to build a protein model.  Selenium (Z = 34) has traditionally been added to proteins to create anomalous signals, but obtaining phase information from crystals of native protein would be optimal. Intrinsic iron (Z = 26) has occasionally been sufficient for phase determination, but the next heaviest atom regularly found in proteins is sulfur (Z = 16).  .  

                Since 1981, 57 novel structures have been published that employed light atom (Z <20) SAD to determine phases.  In comparison, over 5000 structures have been deposited into the Protein Data Bank (PDB, www.pdb.org) in the past fifteen years that used heavy atom SAD.  This discrepancy arises from the low anomalous strength of light atoms, low amount of sulfurs per protein molecule, radiation damage, and diffuse scattering of the X-rays.

                Determined to overcome these complications, Liu et al. began with their previously reported technique that improved data from poorly diffracting selenomethione protein crystals.  To decrease radiation damage and increase signal to noise, a procedure was designed to merge data from multiple crystals.  Building on this, authors then optimized the X-ray beam energy based on knowledge that anomalous signal from light atoms increases with increasing wavelength.  Together, these two methods increased signal to noise and minimized radiation damage.  Finally, the crystal and beam path were placed inside a helium-filled cone and the beam size was matched to crystal size, which also boosted signal to noise and reduced incoherent X-ray scattering.

                Four test proteins were crystallized: netrin G2, TorT/TorS_s, HK9s and CysZ.  Three of the four contained at least 20 sulfur atoms, but HK9s only had three plus an additional chloride.  At least 5 crystals were used per protein.  The authors then define three criteria each crystal’s data set must meet before being merged with others: unit cell parameters less than 3σ, overall diffraction dissimilarity less than 5%, and the relative anomalous correlation coefficient greater than 35%.  This is an improved rubric from their previous work.  Of the 31 crystals screened, only one data set was removed after failing the above standards.  For each protein, data sets were added together one at a time and tested for structure determination.  With each successive data set, anomalous signal, resolution, and electron density maps improved, thus validating their procedures.

                As a way to demonstrate the versatility of light atom anomalous dispersion, the test proteins were diverse.  They ranged from127 to 1148 ordered residues, their crystal symmetries varied from monoclinic to tetragonal and resolutions are as high as 2.3 Å.  Netrin G2 and TorT/TorS_s were both previously unknown structures, while HK9s and netrin G2 were not amenable to previous structure determinations.  CysZ is a membrane protein, while the rest of soluble domains of membrane proteins.  Finally, TorT/TorS_s is reported to be at a complexity level exceeding 90% of the current PDB and its structure was determined here to reasonable resolution.  The authors do admit that the technique will benefit from improvements at beamlines, scaling and weighting procedures, but they do believe that multicrystal SAD phasing will be extremely useful for determining de novo structures of native proteins and nucleic acids.