General workflow

Single Particle Electron Cryo Microscopy (cryo EM) is the main method used in this thesis to analyze the structure and finally the dynamics of molecular machines. The main advantage of this method is the high degree of directness. This means that raw data can readily be interpreted by eye and the room for the interpretation is rather narrow. In principle, it can even be seen as single molecule method since signals from individual molecules can be clearly distinguished and thus be sorted very accurately.

The general workflow and the possibility to use it for conformational dynamics is outlined in this section and will be explained in detail in the materials and methods section (see 2.4). From a very abstract point of view many similarities to biochemical purifications can be found (see figure 1.7). Very exhaustive and excellent reviews can be found in the textbooks of Joachim Frank and Michael F. Moody [55, 131]

1. As in any biophysical method the workflow starts with the preparation of the pu-rified sample for the method. Electron microscopes operate under high vacuum for which the sample has to be stabilized. In biomolecular EM the two frequently used methods are freezing the molecules in a thin film of buffer (cryo conditions) or em-bed and dry the sample with surrounding heavy metal stain (negative staining), both on the carbon surface of small metal grids as support.

2. After embedding of the sample, it is introduced into a TEM and images of several regions of the grid are taken. Each image depicts many copies of the molecule in random orientation and distribution. The images are close to perfect parallel pro-jections of the molecule, meaning the 3D-information is integrated into a 2D image.

These 2D images, however, have a very low Signal-to-Noise-Ratio (SNR) and due to the high energy of the electron beam, the exposure time of the sample to electrons is kept to a bare minimum. The orientation of a molecule on the carrier grid can be described by six degrees of freedom: the three translations along the coordinate axes and the three rotations around them. To determine these parameters, the images will be subjected to exhaustive computational image processing.

1.3 Single particle cryo EM as tool to analyze conformational dynamics 17 3. In a first step, the particles have to be identified (picked) within the image and an

image stack containing the individual particles is created.

4. Due to the low interactions of the electron beam with biological specimen, the images have to be taken with underfocus, leading to a broad and complicated point spread function (PSF) smearing the image information over a large area. This can be restored in-silico in a process called CTF-correction.

5. The corrected particle images need to be further prepared for analysis. To increase the SNR for initial analysis, they are down sampled (coarsed) and Fourier filtered.

Further, the images have to be normalized, since they arise from different micro-graphs and different parts of the grid.

6. To further improve the SNR, individual particle images are averaged. Since the par-ticles were randomly orientated on the grid different particle images depict different views of the particles and have to be sorted for their orientation and properly super-imposed. The processes necessary are calledalignment andclassification. Alignment orients the particles images in a way that they superimpose a set of given refer-ences. Classification identifies similar images and sorts them into a given number of classes. Thus, the particle images showing the same orientation, can be averaged to so-called class averages.

7. Once average images of similar orientation are calculated, a 3D-reconstruction can be attempted. Beforehand, the orientation of the averages in space with respect to each other (as Euler angles) has to be determined. This can be done mathe-maticallyin-silico in a process called angular reconstitution or experimentally with a technique called random conical tilt (RCT). After the angles are determined, a 3D model can be reconstructed.

8. Initial structures are low in resolution. To improve the structure, several refinement cycles are performed. In every cycle, the accuracy of alignment and orientation angle assignment is improved.

9. Once convergence of the refinement is reached, the model needs to be validated, if this was not done before. Furthermore, the resolution of the structure should be determined. The resolution describes which degree of detail can be seen in the structure meaning how far structural features need to be apart to be distinguishable.

10. Finally, interpretation of the model can be attempted. To interpret the computed 3D structures, they can be segmented and structural models gained from other methods can be fitted into the density model. Alternatively, conformational sorting can be performed as outlined in the following section.

18 1 | Introduction

Figure 1.7: Single Particle cryo EM Workflow. The general steps in single particle cryo EM are illustrated using the Anaphase Promoting Complex (APC) as an example. (I) In the microscope projection images are taken of the ice embedded molecules. Particles appear as dark densities in light background which is inverted for image processing. (II) In a first image processing step, molecules have to be selected in a process called particle picking. Here, the molecules are encircled in green. The identified particle images are cropped from the micrograph (right). (IIIa) Every particle is distorted with a point spread function. The Fourier transformation of this is called Contrast Transfer Function (CTF) which is depicted here. It can be seen that in certain areas of this function the contrast is negative, meaning inverted. This is fixed during CTF correction by phase flipping (IIIb) The corrected images are filtered to optimally prepare them for the alignment. (IV) In an initial 2D processing, images that show the same orientation of the molecule in space are superimposed in a process called alignment and grouped and averaged in a process called classification. (V) In the next step, for all good class averages (left) their relative orientation in space is determined (right). After orientation they are projected back into 3D space to calculate a first 3D model. (VI) The first 3D model is gradually improved in iterative refinement steps. Thereby, the 3D model is projected into 2D images which are used as references for an alignment against the full dataset. After alignment, the images are averaged with respect to the best matching reference and a 3D model is reconstructed from the averages. The whole procedure is reiterated with decreasing projection distance until it converges. Shown is a 7.4 Å model of the APC/C (EMDB-2651).

1.3 Single particle cryo EM as tool to analyze conformational dynamics 19