CCP4 Study Weekend 2000 on Low Resolution Phasing
University of York (UK), 7-8th January 2000
Organisers: Julie Wilson (York), Jonathan Grimes (Oxford) and Helen Saibil (Birbeck)
The meeting began with an overview of the standard tools used in small molecule crystallography and their relevance to macromolecular structure solution. Chris Gilmore described the limitations of these methods as the number of atoms increases and the resolution of the data is reduced, but showed how the situation can be simplified, for example by the use of ``globs'' of density rather than atoms, in order to use the direct methods approach. Chris also talked about the use of electron microscopy images and maximum entropy in providing low resolution phases. This led nicely into Helen Saibil's talk on cryo-electron microscopy, in which she introduced the different methods, i.e. 2D crystals, 1D helical assemblies and single particle EM. Helen described the various stages associated with achieving a 3D image from 2D projections and finished with a video showing how even at 30A the dramatic changes as GroEL binds either ATP or ADP can be seen.
We then moved on to a series of talks on the ab intio phasing of macromolecular crystal structures. Vladimir Lunin stressed the importance of good selection criteria in choosing the best phase set from a large population of random starting phase sets. He discussed selection procedures and concluded that, although as yet no infallible criteria was known, they can imply that certain ``sets'' of phase sets are more likely. Thus the alignment and clustering of electron density maps to produce such sets can lead to a determination of the molecular envelope after averaging. Alexandre Urzhumtsev continued with a review of constraints which can be applied to the electron density. He discussed the use of electron density histograms and connectivity theory as well as constraints given by packing considerations and known topological properties. Alberto Podjarny explained how a small number of large Gaussian spheres, generated randomly in the ``Few Atoms Method'', can be used to calculate structure factors whose magnitudes can be compared with the experimental magnitudes. After clustering, a generalized likelihood criteria is used to determine the best cluster which can then be averaged to provide a low resolution mask. The resolution can be increased gradually by generating smaller spheres within the mask obtained at each stage.
We then returned to electron microscopy with an account of the single particle crystallography of 50S ribosomal subunits given by Elena Orlova. She explained how the multiple orientations in this method are aligned and averaged before the ``common lines'' in the 2D images can be used to reconstruct the 3D image. The structure has been solved to between 9 and 5A by various groups, a resolution which Elena showed allows the fitting into the density of those parts for which X-ray crystal structures are available, for example the Principle Interface Protrusion and Protein L9. Vinzenz Unger followed with the use of 2D crystals in the solution of the membrane proteins, connexions, which form channels to allow small proteins through the membrane. The 5-10A electron microscopy maps clearly show separated helices at either end from which a first model based on the sequence could be built. Vinzenz explained how information on which residues are likely to have access to the channel, be exposed to the lipid bilayer or adjacent to other helices was used. Mutant studies have been used to test and improve the model. Niko Grigorieff then gave further details of the problems involved, particularly in the alignment of noisy data when no ``perfect'' model is available for comparison. He described the program FREALIGN, which maximises a liklihood function rather than using a correlation coefficient to align the images. The correlation between the refinement of two half data sets allows an independent assessment of the quality and provides a resolution cut-off.
The second day of the meeting began with a combination of the techniques of X-ray crystallography and electron microscopy as Steve Fuller described the use of X-ray structures in the interpretation of EM maps. As well as fitting X-ray structures into EM maps, Steve showed how, in the case of large viruses, a very low resolution X-ray structure (~30A) can be used as a model in the early stages of an electron microscopy solution. Bill Shepard followed with the use of anomalous scattering to generate contrast variation. MASC requires a single crystal rather than a change of solvent to determine the molecular envelope.
We then heard how it is possible to collect the very low resolution reflections, so vital in ab initio phasing, in-house. Gwyndath Evans described how, if the experiment is set up carefully, these reflections can be measured accurately. Then Jian-wei Miao dispensed with the phase problem completely by over-sampling the diffraction pattern in the case of non-crystalline particles and suggested the same approach could be used for crystals.
Then it was back to the molecular envelope and its use in phase extension and improvement. Pietro Roversi showed that the Fermi-Dirac distribution could be used to provide a continuous envelope rather than binary map for use in solvent flattening or the refinement of incomplete structures. Peter Main followed with a method to extend the phases provided by the envelope. In order to increase the resolution of an electron density map, the right amount of detail must be added to the correct place in the map and Peter showed that the wavelet transform gives some control over this. He showed that 10A starting phases could be extended as far as 6-7A.
Dave Stuart described the use of averaging both between different crystal forms and through non-crystallographic symmetry. He also showed how known sub-units can be fitted together into EM maps for use in the molecular replacement solution of large X-ray structures. This began the theme for the afternoon's talks. The problems, particularly of scale, encountered when fitting X-ray structures into EM maps were discussed by Alan Roseman. In the case of receptors with sugars, Michael Rossmann showed how difference maps were used to confirm the orientation of the receptors and finally Elizabeth Hewat warned of the additional difficulties involved when fitting into disordered maps.
The meeting ended with a discussion session led by Michael Rossmann. The recent advances in electron microscopy have been spectacular and it was agreed that the future would lead to further combination the technique with X-ray crystallography.