Docking Molecular Structures into EM

Introduction Detailed models of proteins are required in order to elucidate the structure-function relationship methods for directly imaging large complexes at atomic level are not currently available (and are not likely to be soon).

Introduction atomic models provide detailed 3D data, but are taken out of their physiological context and are often available only as smaller fragments of larger units Datasets obtained by 3D-EM techniques reveal a good approximate of the macromolecular shape (in the form of surface) as well as the internal arrangement of mass within the surface.

Introduction pseudo-atomic-resolution models of macromolecular complexes can be generated by combining high-resolution data (XRC and NMR) for assembly components with lower-resolution data (EM) for the whole complex

Qualitative Docking: Manual Fitting the fit of the atomic model into the isosurface envelopes is judged by eye and correlated manually until the fit “looks best” “ a jigsaw-puzzle”

Manual Fitting If components are large molecules with distinctive shapes at the resolution of the construction, manual fitting can often be performed with relatively little ambiguity manual fitting is widely used: Baker TS and Johnson JE, 1996, Low resolution meets high – towards a resolution continuum from cells to atoms, Curr Opin Struct Biol 6: Baker TS and Johnson JE, 1996, Low resolution meets high – towards a resolution continuum from cells to atoms, Curr Opin Struct Biol 6:585-95

Manual Fitting Example 1 Rayment et al, 1993, Science, 261:58-65

Manual Fitting Example 1 muscle contraction occurs when thin actin filaments and thick myosin filaments slide past one another Impossible to create a structural model for how muscle contracts (contractile cycle) without knowledge of the 3D structures of the components and how they interact

Manual Fitting Example 1 Rayment et al. obtained a model for rigor complex of F actin and myosin head by combining the molecular structures (XRC) of the individual proteins with electron density maps obtained via cryo-EM XRC myosin S1 structure to 2.8 Å resolution XRC myosin S1 structure to 2.8 Å resolution coordinates for F actin obtained from x-ray structure of G actin by fitting and refining model to x-ray fiber data from F actin gels coordinates for F actin obtained from x-ray structure of G actin by fitting and refining model to x-ray fiber data from F actin gels data combined using low-resolution (~30 Å) electron density maps of Factin alone and myosin-actin complex calculated from images recorded from cryo- EM. Manual building using FRODO data combined using low-resolution (~30 Å) electron density maps of Factin alone and myosin-actin complex calculated from images recorded from cryo- EM. Manual building using FRODO

Manual Fitting Example 1 1.actin model positioned in F actin envelope 2.actin envelope replaced with actin- myosin envelope and myosin structure rotated and translated into place myosin head highly asymmetric, therefore easy to position molecule unambiguously in envelope myosin head highly asymmetric, therefore easy to position molecule unambiguously in envelope

Manual Fitting Example 1 clear that myosin head must be close to actin – leads to model for molecular basis for muscle contraction:

Manual Fitting Example 1 even though resolution of EM is only ~30 Å, the accuracy of the fitted results is higher – ambiguity of ~5 Å

Manual Fitting Example 1 collision at the site of actin-S1 interface viewed not as shortcoming of model, but an indication that there may be a conformational change, which may contribute to understanding structural basis of contractile cycle

Manual Fitting Example 2 J. Virol, 1993, 67: HRV major cause of the common cold over 100 serotypes structure of HRV14 known to atomic resolution – icosahedral, 300 Å capsid is built from 60 copies of four viral proteins – VP1, VP2, VP3 and VP4

Manual Fitting Example 2 EM graph of virus with complexed neutralizing antibodies immunogenic sites all occur at the rim of a 12 Å deep canyon that encircles each of the icosahedral fivefold axes. The canyon is the site at which the cell surface receptor proteins bind. Arrangement allows naturally-occurring mutants to circumvent immune recognition without affecting receptor recognition. HRV14-Fab17-IA HRV14

Manual Fitting Example 2 HRV14-Fab17-IA crystal structures for both whole virus and Fab-17IA were known EM envelope served as constraint for docking the Fab atomic model onto surface of HRV

Manual Fitting Example 2 translations of the Fab model by as little as 2-4 Å moved substantial portions of the model outside the EM envelope or caused the two models to overlap docking procedure lead to a pseudo-atomic model of the complex, from which interacting surfaces could be defined from this, a number of electrostatic interactions were identified as contributing to the binding affinity of Fab for the virus- confirmed by mutagenesis HRV14-Fab17-IA

Problems with Manual Fitting Manual fitting does not always proceed to happily- Docking results may be invalid if the molecule adopts a significantly different structure in crystal and large complex (but may be proof that a molecule HAS changed conformation) Docking results may be invalid if the molecule adopts a significantly different structure in crystal and large complex (but may be proof that a molecule HAS changed conformation)

Problems with Manual Fitting divergent models of the same complex docked by eye have also been reported: e.g. Hoenger et al, 1998, J. Cell. Biol. e.g. Hoenger et al, 1998, J. Cell. Biol. intracellular transport, flagella beating and other motile phenomena in cells are based on the interaction between microtubules and motor proteins such as kinesins or dynein

Hoenger et al, 1998, J. Cell. Biol. motor proteins consist of several domains – a head, a stalk and a tail controversial reports on stochiometry of binding of tubulin and kinesin – one head per tubulin subunit or two

Hoenger et al, 1998, J. Cell. Biol.

Quantitative Fitting objective scoring functions can be used to assess the quality and refine the initial manual fit, or perform and automated procedure: e.g. J. Structural Biology, 1999, 125, global search with correlation between the calculated electron density from atomic model and that observed by EM objective evaluation of the quality of the docking

Quantitative Fitting The global search is followed by a statistical analysis of the distribution of the fitting criterion results in definition of the confidence intervals that eventually lead to solution sets solution sets are small regions in parameter space that satisfy the data within the error margin defined by the chosen confidence interval

Solution Sets The size of the solution set can serve as a normalised goodness-of-fit criterion – the smaller the set the better the data that determines the position of the fitted atomic structure Solution sets allow for the use of standard statistical tests, such as Student’s t-test, to evaluate the differences between models in different functional states and to help model conformational changes

Difference Mapping Difference mapping between the density calculated from the fitted model and the reconstruction from electron microscopy can locate portions of the structure not present in the crystal structure or identify conformational changes e.g. actin-myosin S1 (Rayment 1993) – light chain positional change on binding to actin search volume for S1 fragments solution-set centre fit of complete S1 structure – note bad fit of white domain map of density difference between S1 and b cyan solution-set centre fit of motor domain of S1 structure solution-set centre fit of light chain into remaining density difference density between light chain and c

solution-set centre fit of complete S1 structure – note bad fit of white domain solution-set centre fit of motor domain of S1 structure search volume for S1 fragments map of density difference between S1 and b cyan solution-set centre fit of light chain into remaining density difference density between light chain and c