Menu:

Research

Our lab works on both the application of cryo-EM for biomolecular structure determination and also the development of computational tools for cryo-EM. We have a focus on the determination of structures of amyloid fibrils in collaboration with our colleagues at the ICS-6 in Jülich and the IPB in Düsseldorf. Amyloid fibrils play a significant role in many (but not only in) neurodegenerative diseases, such as Alzheimers or Parkinsons disease.

Cryo-EM Structure Projects

Amyloid-β(1-42) fibril structure

gunnar Fibrillar aggregates of the amyloid-β protein (Aβ) are the main component of the senile plaques found in brains of Alzheimer’s disease patients. We have determined the structure of an Aβ(1-42) fibril composed of two intertwined protofilaments determined by cryo-EM to 4.0 Å resolution, complemented by solid-state nuclear magnetic resonance (NMR) experiments. The backbone of all 42 residues and nearly all sidechains are well resolved in the EM density map, including the entire N-terminus, which is part of the cross-β structure resulting in an overall "LS"-shaped topology of individual subunits. The dimer interface protects the hydrophobic C-termini from the solvent. The unique staggering of the non-planar subunits results in markedly different fibril ends, termed "groove" and "ridge", leading to different binding pathways on both fibril ends, which has implications for fibril growth.

PI3K-SH3 fibril structure

gunnar The Src-homology 3 domain of phosphatidyl-inositol-3-kinase (PI3K-SH3) is a model amyloid system that plays a pivotal role in our basic understanding of protein misfolding and aggregation. We have determined the atomic model of the PI3K-SH3 amyloid fibril with a resolution of 3.4 Å by cryo-EM. The fibril is composed of two intertwined protofilaments that create an interface spanning 13 residues from each monomer. The model comprises residues 1–77 out of 86 amino acids in total, with the missing residues located in the highly flexible C-terminus. The fibril structure allows us to rationalise the effects of chemically conservative point mutations as well as of the previously reported sequence perturbations on PI3K-SH3 fibril formation and growth.

Structures determined in collaborations:

Actin Filaments

Work on actin filaments was done in collaboration with the groups of Ed Egelman and Vitold Galkin.

Ribosome Structures and Dynamics

Work on ribosome structures was done in collaboration with the group of Holger Stark.

Chaperones (GroEL and CCT)

The work on chaperones was done mostly together with the group of Wah Chiu.

Cryo-EM Computational Methods


gunnar

Analyzing molecular motions from cryo-EM Data

Single-particle cryo-EM images provide information on the conformational distribution of biomolecules,, and therefore on conformational motions of these molecules. We are developing the principal motion analysis (PMA) method to determine these conformational motions.


Cross-validation in structure refinement

gunnar We have developed a cross-validation approach for real-space refinement against cryo-EM density maps in analogy to cross-validation typically used in crystallography. Our approach is able to detect overfitting and allows for optimizing the choice of restraints used in the refinement. We treat a high-resolution Fourier shell as a the test set, that is not used for the refinement, but only for validation. Because cross-validation requires splitting the dataset into at least two independent sets, we further present an approach to quantify correlations between the structure factor sets. This analysis is also helpful for other cross-validation applications, such as refinements against diffraction data or 3D reconstructions of cryo-EM density maps.

Clustering of helical polymer images

gunnar Helical protein polymers are often dynamic and complex assemblies, with many conformations and flexible domains possible within the helical assembly. During cryo-EM reconstruction, classification of the image data into homogeneous subsets is a critical step for achieving high resolution, resolving different conformations and elucidating functional mechanisms. Hence, methods aimed at improving the homogeneity of these datasets are becoming increasingly important. We have developed an algorithm that uses results from 2D image classification to sort 2D classes into groups of similar helical polymers.

Density Sharpening: VISDEM

gunnar We have developed a method to improve the visualisation of density maps by using general statistical information about proteins for the sharpening process. In particular, the packing density of atoms is highly similar between different proteins, which allows for building a pseudo-atomic model to approximate the true mass distribution. From this model the radial structure factor and density value histogram are estimated and applied as constraints to the 3D reconstruction in reciprocal- and real-space, respectively. Interestingly, similar improvements are obtained when using the correct radial structure factor and density value histogram from a crystal structure. Thus, the estimated pseudo-atomic model yields a suffciently accurate mass distribution to optimally sharpen a density map.

Real-space model refinement with DireX

Single-particle cryo-electron microscopy (CryoEM) is able to determine three-dimensional density distributions (density maps) of large macromolecules. The resolution ranges from (in few cases) higher than 4 Å to often well below 10 Å. In case high-resolution structures (e.g. from crystallography) is available the goal is to refine the structure to fit into the low-resolution CryoEM density map. For this purpose we have developed the real-space refinement program DireX.
Read more about DireX and watch movies of the flexible fitting...

Deformable Elastic Network (DEN)

We developed the Deformable Elastic Network approach to combine prior structural knowledge with low-resolution data. The idea is to adapt an elastic network potential (defined on a reference structure) in a way that deforms only those degrees of freedom that need to be deformed to fit the data but not more. This approach significantly reduces overfitting, which is usually encountered in low-resolution refinement.



Structural Interpretation of FRET Experiments

Modeling based on multiple FRET Experiments

Single-molecule Förster Resonance Energy Transfer (smFRET) measurements use the non-radiative transfer of energy from a donor to an acceptor dye. By monitoring the relative fluorescence intensities from the two dyes, the energy transfer efficiency can be evaluated for individual molecules. The transfer efficiency is commonly employed to determine the interdye distance. This method has emerged as a powerful biophysical tool to measure the distance between two labeled sites within one macromolecule or between two labeled macromolecules. FRET measurements are most sensitive in the 10-100 Å distance range. With several of such measured distances it becomes possible to build a three-dimensional model and to describe conformational transitions. We developed methods for this purpose and applied them to different kinds of systems from proteins to DNA and RNA.

Simulation of Fluorescence Anisotropy Experiments

Time-resolved fluorescence anisotropy decay experiments on a protein-attached dye can probe local protein dynamics and steric restrictions, but are difficult to interpret at the structural level. Aiming at an atomistic description, we have carried out molecular dynamics simulations of such experiments. Our simulations describe an Alexa488 fluorescent dye maleimide derivative covalently attached via a single cysteine to the AB-loop of bacteriorhodopsin. Fluorescence anisotropy decay curves obtained from the simulations agree well with the measured ones. Furthermore, our simulations allowed us to test the usual and inevitable assumption underlying these types of spectroscopic measurements that the attached dye probe does not severely perturb the protein dynamics. For the case at hand, by comparison with a simulation of the dye-free protein, the perturbation was quantified and found to be small.