A RECORD MAD STRUCTURE FROM THE NSLS

Preprint of an article for the November NSLS Newsletter

Sylvie Doublié and Tom Ellenberger, Harvard Medical School.

Robert M. Sweet, BNL Biology Dept.

One year ago this fall, multi-wavelength anomalous diffraction (MAD) data were collected from a crystal of selenomethionyl-T7 DNA polymerase at the BNL Biology Department beamline X12-C. These data rapidly produced the largest new structure (108,000 Da) to be determined by this novel method, raising the prospect that MAD phasing can succeed in most macromolecular structure determinations, large or small.

The MAD experiment was performed after months of unsuccessful attempts to obtain phases by the conventional method of multiple isomorphous replacement (MIR). The success of a MIR structure determination depends on the selective binding of heavy metals such as mercury or platinum to a limited number of sites in the crystal. Heavy atoms scatter x-rays more strongly than the light atoms they replace, resulting in a perturbation of the diffraction pattern. Metal-induced changes in the intensities of diffracted x-rays can reveal the locations of the metals in the crystal, and this information yields the amplitudes and phases of the x-rays scattered by the heavy atoms.

In a further extension of the isomorphous replacement method, X-rays with energies near the energy levels of electrons in the heavy metals experience large phase and amplitude shifts as they resonate with these electrons, and this "anomalous" scattering provides additional phase information. Unfortunately, in many cases metal binding also shifts the positions of the macromolecules within the crystal lattice, causing large and unpredicatable changes in x-ray diffraction. (The crystals are no longer of the "same form", that is "isomorphous" with the original crystal.) In extreme cases the metals may render the crystals severely disordered and useless for diffraction experiments.

Heavy atom soaks were unsuccessful for crystals of the T7 DNA polymerase, so we used a trick developed by Wayne Hendrickson and coworkers at the Columbia University and the Howard Hughes Medical Institute beamline X-4 at NSLS. They biosynthetically incorporated selenomethionine into proteins in place of methionine. This was accomplished by overexpressing the protein in bacteria grown in defined media containing selenomethionine. The selenium atom replaces a sulfur in the methionine side chain, and this electron difference is readily detected by x-ray diffraction from crystals of the selenomethionyl-protein. Unlike the larger metals typically used for MIR, these selenium atoms are tolerated well by most proteins. In fact, the selenomethionine imposter is convincing enough to fool the cell's own protein biosynthetic machinery. Selenomethionyl-proteins generally behave like their natural counterparts during protein purification, and crystals of these modified proteins typically grow under the same conditions as crystals of the native protein.

In pioneering studies, Hendrickson and coworkers showed that diffraction data from single crystals of proteins containing selenium or another anomalous scatterer could produce accurate phases for a crystal structure. Their method of multi-wavelength anomalous diffraction (MAD) requires x-ray diffraction measurements at two to four x-ray energies near an atomic absorption edge of the heavy atom, chosen to maximize the real and imaginary components of anomalous scattering. MAD phasing is rapidly becoming the method of choice for determining new crystal structures of small to medium-sized proteins, and MAD has succeeded for a variety of anomalous scatterers including Se, Fe, Cu, Br, Tb, Pt, and Hg. Several of the NSLS beamlines devoted to macromolecular crystallography are equipped with x-ray optics that produce low bandwidth, high intensity x-rays at energies near the K absorption edge of selenium (lambda = 0.98 Å). This situation is ideal for the accurate measurement of anomalous scattering from weakly-diffracting crystals of selenomethionyl-proteins.

So why was a MAD experiment using the selenomethionyl-T7 DNA polymerase not our first choice for determining the crystal structure? We crystallized a complex of the polymerase bound to its processivity factor E. coli thioredoxin, a DNA primer-template, and a nucleoside triphosphate, totaling 108,000 Daltons in the asymmetric unit. This complex is larger than any of the protein crystal structures that had been successfully phased solely on the basis of anomalous scattering from selenium. Moreover, it would be necessary to locate most or all 15 selenium atoms in the crystal asymmetric unit. This is typically accomplished with a Patterson function calculated from the anomalous intensity differences of Friedel mates in the diffraction pattern. Looking at all possible vectors connecting 15 selenium atoms, one would have to sift through 210 interatomic vectors in the difference-Patterson map -- a daunting task! Nevertheless, the quality of phase information obtained from other MAD experiments at NSLS beamlines X12-C, X4, and X25 suggested to us that a multiwavelenght experiment with the selenomethionyl-polymerase complex was appropriate.

We performed a 3-wavelength MAD experiment at beamline X12-C, guided by the automatic MAD data collection method developed by J. Skinner and R. Sweet. Diffraction data were collected with the hybrid MAR Research imaging-plate / Nonius diffractometer to a resolution of 2.2A from a single crystal, cryo-cooled at 100 K. The x-ray data sets were collected at the inflection point of the Se K edge (lambda-1 = 0.9822Å), the absorption peak (lambda-2 = 0.9788Å), and a high-energy wavelength remote from the edge (lambda-3 = 0.95A; Figure 1). We collected an additional data set from the same crystal using a laboratory x-ray source (lambda-4 = 1.54Å) upon returning from NSLS. Phase information was derived from the anomalous differences at the absorption peak, and from dispersive differences involving all pair-wise combinations of data sets. Care was taken in aligning the plate-like crystal in the plane of the cryoloop so that Bijvoet pairs appeared on the same image or adjacent images, maximizing the accuracy of intensity difference measurements. Crystal diffraction quality was assessed in our laboratory at Harvard Medical School prior to transporting the frozen crystals to the NSLS.

The MAD experiment requires the energy-resolution and brilliance of a synchrotron beamline in order to measure small intensity differences arising from anomalous scattering, coupled with an ease and reproducibility of repeated wavelength changes. Beamline X12-C clearly met these requirements, producing high quality data of the polymerase complex. However, we still faced with the problem of finding the 15 selenium atoms in the crystal asymmetric unit. Direct phasing methods implemented in George Sheldrick's program, SHELXS-86, readily solved the problem. All 15 selenium atoms were located from dispersive differences calculated from data collected at the inflection point and at the high energy remote wavelength. The resulting MAD-phased electron density (Figure 2) far exceeded our expectations! The experimental electron density map, completely free of model bias, is in many regions indistinguishable from the 2Fo-Fc map calculated with phases from the final, refined model (Figure 3). The MAD experiment not only ended our laborious search for heavy atom derivatives, but it also produced electron density of exceptional quality, expediting model building and refinement of this large complex.

The structure of the T7 DNA polymerase complex provides the first glimpse of a replicative DNA polymerase poised for the incorporation of an incoming nucleotide into a growing DNA strand (Figure 4). The structure yields many insights concerning the high fidelity of template-directed DNA synthesis (selecting the correct nucleotide for incorporation), a means for detecting misincorporated bases, and the mechanism of metal-assisted nucleotidyl transfer by this and a large group of related polymerases. An article describing the crystal structure is in press at Nature (S. Doublie', S. Tabor, A.M. Long, C.C. Richarson, and T. Ellenberger. Crystal structure of bacteriophage T7 DNA polymerase complexed to a primer-template, a nucleoside triphosphate, and its processivity factor thioredoxin).