BMB/Bi/Ch 173 Winter 2017

Size: px

Start display at page:

Download "BMB/Bi/Ch 173 Winter 2017"

Dwayne Morton
6 years ago
Views:

1 BMB/Bi/Ch 173 Winter 2017 Homework Set 3.2 Assigned 1/26/2017, Due 1/31/17 by 10:30am TA - Sara Weaver sjweaver [a] Caltech.edu Office hours Broad 3rd floor kitchen - Friday 1/27 1:30pm-2:30pm, Monday 1/30 5pm6pm, or by appointment Problem 4 (60 points) Single particle cryoem In your research you decide to solve a ribosome structure by single particle cryoem. 4.a. (5 Points) After meeting with Alasdair in the cryoem lab about your project, you set to work preparing negatively stained protein samples. To learn more about what a good negative stain experiment looks like, you review a few of the GraFix papers discussed in Prof. Jensen s lectures. Below is an excerpt from supplementary figure 1 in Kastner, B., Fischer, N., Golas, M., Sander, B., Dube, P., Boehringer, D., Hartmuth, K., Deckert, J., Hauer, F., Wolf, E., Uchtenhagen, H., Urlaub, H., Herzog, F., Peters, J., Poerschke, D., Lührmann, R., and Stark, H. (2007) GraFix: sample preparation for single-particle electron cryomicroscopy, Nat Methods, 5, Supplementary Figure 1 : A) Negative stain image of a 17S U2 snrnp fraction of a U2 snrnp preparation, with glutaraldehyde (GA) added after standard gradient centrifugation. B) Negative stain image of 17S U2 snrnp after GraFix of the same U2 snrnp preparation reveals a considerably more homogeneous population of particles What problems did the authors see with figure A that led them to develop the GraFix method? What general insights can you gain from a negative stain experiment? Hint: you don t need to research GraFix to answer the question. Just appreciate the figure and use what you ve learned in class. 2.5 points: In their single particle work, the authors found some of their protein complexes with many subunits were unstable during grid preparations. Their micrographs showed inhomogeneous fields of particles, where some particles had fallen apart. This made it difficult to only choose the complete complex from the mess of partially degraded particles. It also reduced the efficiency of their data, as they had to throw out so many particles. So they developed the GraFix fixation protocol to improve their yield of fully formed complexes. Homework Set 3.2 1

2 BMB/Bi/Ch 173 Winter 2017 In figure A, you can see the particles in the image are not all the same size. That inhomogeneity suggests that they are not all identical, so that would be bad for single particle. In figure B, the particles are homogeneous in size and approximate shape (taking into consideration that each particle is in a random orientation relative to the grid). 2.5 points: In general, negative stain allows us to assess the quality of a single particle protein prep. We re looking for a good distribution of particles. Particles should not be touching and should not appear aggregated. We re looking for particle homogeneity, so seeing particles of different sizes is a bad sign. While negative stain results don t directly translate to cryoem preparations, they often suggest problems with the protein sample that will remain during cryoem sample prep. Sometimes microscopists use negative stain data to answer their scientific question (if the resolution is sufficient). Negative stain reconstructions can also be used to create an initial model for a subsequent cryoem experiment. Negative stain is not a conclusive way to tell if your particle has a preferred orientation because orientation problem could be caused by interaction with the stain. A cryoem experiment will have the protein in ice, so the orientation preferences might not be the same. 4.b. (5 points) Below is an example image from your negatively stained E. coli 50S ribosome grid. Comment on its quality based on what you know about negative stain. How does this experiment inform your future cryo experiments? The scale bar is 100 nm. A little aggregation and the particles are too close, but overall it looks good. I would call this a monodisperse sample. The aggregation will probably decrease if you decrease the sample concentration. Additionally, the aggregation could be caused by protein interacting with the negative stain. The concentration is a problem because if you tried boxing out these particles, it would be hard to avoid getting nearby particles in the box. While you would hope that these extra densities from nearby particles would average out, this is the kind of artifact that can send your reconstruction off in the wrong direction, leading to an incorrect solution. The particles all look similar (in approximate size and shape), suggesting stoichiometric homogeneity. Homework Set 3.2 2

BMB/Bi/Ch 173 Winter 2017 4.c. (5 Points) Once your negative stain looks good, you decide to move onto plunge freezing your protein samples.

3 BMB/Bi/Ch 173 Winter c. (5 Points) Once your negative stain looks good, you decide to move onto plunge freezing your protein samples. Microscopists often use negative stain as a starting point to gauge a protein s behavior before moving onto cryoem. A rule of thumb in cryoem states that you protein should be 100x more concentrated for plunge freezing than for negative stain. Explain why you need a higher protein concentration for plunge frozen grids. In negative stain, all of the protein you apply remains on the grid. When you plunge freeze a grid, some protein is blotted away. This can be a big problem if your protein is attracted to the fibers on the blotting paper. Thus, you need a higher protein concentration in a plunge frozen sample to attain a similar particle distribution on the grid. 4.d. (10 Points) You decide to freeze your protein on glow discharged Quantifoil grids ( Your Quantifoil grid has a copper scaffold. A 12 nm sheet of carbon is placed on the scaffold. The carbon has 1.2 µm holes in it. Ideally proteins will be suspended in ice across the holes. You image the cryo grid on Caltech s T12 to screen your condition before you book time on the nicer F30 microscope. Your grid looks like the figure below. The scale bar is 100 nm. What s wrong with this grid? What would you do to try to fix it? Why can t you solve a ribosome structure using this grid? It turns out that your protein prefers to be on the carbon surface between the holes of your Quantifoil grid. As a result, you don t have much protein in each hole. Why don t you want to image proteins on the spaces between the Quantifoil holes? What can you do to improve your particle distribution? You don t want to image between the holes because the carbon will give a very strong signal and your protein signal will be obfuscated. It will be hard to separate them out later. Also more electrons are scattered by the dense carbon between the holes. Try adding detergent or other additives Try a different grid surface. You could add a thin (3 nm) layer of amorphous carbon or graphene oxide. Change the glow discharging parameters Homework Set 3.2 3

BMB/Bi/Ch 173 Winter 2017 4.e. (10 points) You improve your freezing protocol and take some beautiful images on the T12. Alasdair gives you the OK to book time on the F30.

4 BMB/Bi/Ch 173 Winter e. (10 points) You improve your freezing protocol and take some beautiful images on the T12. Alasdair gives you the OK to book time on the F30. You run an initial reconstruction on your data and get a structure with an apparent resolution of 15 Å. Two screenshots from the Chimera program are shown below. The electron density of the reconstruction is represented in silver. A bar graph of angular distribution is shown on a sphere around the electron density. Red bars mean there are a lot of views from that angle, purple is a medium number of views, and blue is few views. If there is blank space, the reconstruction doesn t contain views from that angle. Based on this reconstruction, what problem does your sample have? How can you tell? Please comment both on the electron density and the angular distribution graph. The protein has a preferred orientation problem. The reported 15 Å resolution is suspect because of the angular distribution we know that the resolution is anisotropic. You can see it in the angular distribution because a lot of the surface is empty (ie no bars). This means you are over representing views where the red bars are, and you re missing a lot of equatorial views. Additionally, you can see it in the electron density because the density is smeared out in the area that doesn t have a lot of views it s similar to the missing wedge problem in tomography. Now let s consider two of the possible scenarios that could lead to the result in 4.e. 4.f. (5 points) How could the image processing protocol have failed, resulting in the angular distribution observed in 4.e.? It s possible that the class averages are not well defined and some views were mis-assigned. It s also possible that when you picked particles you neglected a view. The initial model might have been bad, leading to a poor reconstruction. 4.g. (5 points) What phenomena could cause the angular distribution observed 4.e. during the sample preparation phase? Please also suggest how you could fix the problem when preparing your sample. Sometimes particles are attracted to the air/water interface or to the carbon surface. Sample prep: Try changing the buffer or grid surface. Sometimes adding salt or detergent helps. Sometimes glow discharging both sides of the grid helps. Changing grid surface coating can also help Homework Set 3.2 4

5 BMB/Bi/Ch 173 Winter h. (10 points) Which image acquisition technique(s) could be used to overcome the problem in 4.e. if the solutions in 4.f. and 4.g. are insufficient? How does it work? Hint: there are several, but you only need to write about one. Single Particle Tomography (SPT): If the particle exhibits a preferred orientation in traditional SPA experiments, a tilt series can be collected to increase coverage of views in 3D reciprocal space Tilt pairs can be used. Two tilt pair strategies, orthogonal tilt reconstruction (OTR) and random conical tilt (RCT) reconstruction, were initially developed to determine initial models. Orthogonal tilt reconstruction (OTR) involves taking projection images of the same field of particles at -45 and +45. OTR is actually not recommended for cases of preferred orientation, but students were not penalized on this detail. Random conical tilt (RCT) involves taking projection images of the same field of particles at 0 and high tilt (often 50 ). Since you know the relationship between the tilt pair images, it becomes easier to determine the orientation of a given particle in the dataset. 4.i (5 points) Dual axis tilt tomography involves taking a tilt series, then rotating the sample 90 and taking another tilt series. This reduces the missing wedge in single tilt axis tomography to a missing pyramid. In single particle cryoem, the Orthogonal Tilt Reconstruction (OTR) method involves taking projection images at -45 and +45. As a result, the tilt pairs are 90 apart. At first glance, these techniques seem to have similar approaches. Why does OTR result in isotropic resolution in 3D reciprocal space, while dual axis tilt tomography experiments maintain a missing pyramid? The dual axis tomography tilt series still can only tilt to 60 or 70, whereas in theory the single particle OTR experiment has projection images from all angles. Problem 5 (40 points) Resolution in single particle cryoem 5.a. (10 points) What factors limit resolution in a single particle cryoem experiment? Radiation damage, quality of electron optics, defocus, CTF of microscope, signal to noise ratio, accuracy of alignment, particle homogeneity, MTF of camera, distortions, aberrations of microscope, beam-induced specimen movement, number of particles averaged, 5.b. (10 points) What is the Fourier Shell Correlation (FSC) curve? Conceptually explain how it is solved and what is it used for in single particle cryoem. During the 3D refinement, the dataset is randomly split into two halves. Each half undergoes the reconstruction. At the end, the two electron densities are compared pixel by pixel. For a given resolution shell, you compare the amplitudes and phases of each pixel with the corresponding pixel in the other reconstruction. To do the comparison, you take the real component of the dot product of the amplitudes and phases of one reconstruction with the amplitudes and phases of the other reconstruction. You do this for all shells. That sum is divided by the square root of the product of the sum of the shell of the power in the first reconstruction by the sum of the power in that shell of the other reconstruction. If the pixels are identical, the FSC equals 1. Conceptually, the numerator of the FSC measures how similar the two reconstructions are in a given resolution shell while the denominator normalizes that value to representing how much weight that resolution shell has in the overall reconstruction. 5.c. (5 points) What does a low value on the FSC curve mean? What does this tell you about your reconstruction at that spatial frequency? It means that the two reconstructions look very different at that spatial frequency. It s likely noise in each reconstruction that is varying randomly. Homework Set 3.2 5

6 BMB/Bi/Ch 173 Winter d. (5 points) Why was the ResMap software developed? Electron microscopy resolutions have anisotropic resolution. ResMap maps the local resolution onto the electron density isosurface so that one can evaluate which areas of the structure we are more certain of. 5.e. (10 points) In the previous problem, you generated a ribosome single particle dataset. For fun, you seed the 3D refinement of your ribosome data using a GroEL atomic structure as the initial model. If this reconstruction yielded an electron density with a 4 Å resolution, but it looked more like GroEL than the ribosome, what would you think? How would this happen? In this case, what does the resolution measure? Hint: A conceptual explanation is sufficient. You have experienced model bias! The reconstruction picked out noise that looked sort of like the GroEL projections, leading to the biased final structure. Since the FSC measures how similar the two halves are, you can still get a high resolution if your reconstruction converges to a consistent but incorrect solution. The 4 Å resolution reported by the reconstruction is likely the resolution of the crystal structured used as the initial reference. To know that this was a true GroEL structure magically appearing from our ribosome data, we would need to see a good FSC value past the resolution of the crystal structure reference, indicating that some real, novel information was found. Homework Set 3.2 6

II: Single particle cryoem - an averaging technique

II: Single particle cryoem - an averaging technique Radiation damage limits the total electron dose that can be used to image biological sample. Thus, images of frozen hydrated macromolecules are very