Analysis of the structure of human hemoglobin

You are preparing a presentation on the structure of human hemoglobin. You are particular interested in the structural differences of the deoxy and the oxy form of the protein. In particular, you want to compare the structures from the two dominant methods for structure determination: X-ray crystallography and cryo-EM. Finally, you will compare the experimental structures

Prepare a short presentation about your results of the practical. The emphasis of this part lies on the usage of Cryo-EM derived structures and the use of the Coulomb potential maps. You will also compare a predicted model to the experimentally solved protein structures. The slides should mainly show pictures of the densities and models, but also explain the scientific background in short words.

Dealing with cryo-EM maps

In crystallography experiments, the Xray beam is scattered by the electrons located around an atom, while in Cryo-EM an electron beam is used and these are scattered by the electric potential created of the negatively charged electron cloud and the positively charged nucleus of the atom, also named Coulomb potential. Thus, both methods differ in the resulting map. X-ray crystallography will produce an electron density, while Cryo-EM yields a Coulomb potential map. Because of this difference, Cryo-EM maps might also give you not only the position of the atoms in the sample, but also information about its charge state (Bick et al., 2024).

Download the coordinates in PDB format as well as the potential map from the PDB. The identifier for the oxy form is 8WJ1, the deoxy form can be found via the ID 8WJ2. Unfortunately, you cannot use the built-in feature from COOT to download the corresponding structure as COOT has problems to read the structure from the mmCIF file correctly. COOT also does not download Cryo-EM maps automatically.

Does the map appear okay for you? Is there noise? Noise can appear as irregular shaped density at positions on the map, where you clearly not expect it. Does the model fit the map?

What are other means of quality criteria for cryo-EM? Does the article feature metrics to understand how good the measurements were? Have a look at the corresponding sites of the Electron Microscopy Data Bank (EMDB). Explain in short words what the Fourier shell correlation (FSC) is describing. Does the curve look good? Load also the supplementary information of the corresponding article and have a look at the representative 2D class averages. Can you already identify features of the structures (like a contrast between the electron rich helices and the space between the helices) or do the images appear blurry?

There is frequently a loss of contrast in maps at high resolution that can stem from, for instance, molecular motion, sample heterogeneity, imperfect imaging, and incoherent averaging of the image data. Therefore, you often need to sharpen or to blur cryo-EM maps to allow their interpretation. Although other programs are probably more suitable for this purpose (https://phenix-online.org/documentation/reference/auto_sharpen.html), we will try to do this procedure manually. Go to → "Calculate" → "Map sharpening/blurring" .

I will try to give an easy explanation for the procedure. Sharpening a map will put more emphasis on the high resolution data, whereas blurring will do the opposite and will give higher weight on the lower resolution data. If a map is "over-sharpened" it will contain a higher number of artifacts (i.e., noise). On the contrary, high-resolution features, like side chain density or such, might be absent in "under-sharpened" maps. Indeed, cryo-EM maps that are deposited with the PDB (or better with the EMDB) need to be inspected for over- and under-sharpening and subsequently modified in the correct way. To make the situation even worse, the level of sharpening and blurring need to be adjusted to the part of the map you are looking at.

Do you need to sharpen or to blur the map? What is a good value to see all features around the heme prosthetic group? Can you easily identify the iron in the heme group based on its "density" value?

Export the modified maps from COOT for later use in Pymol. Use the file extension ".map" or ".ccp4".

High-resolution hemoglobin structures

Start a new instance of COOT. To compare the Cryo-EM structures with the structures obtained via X-ray crystallography load the structures from the PDB. You will find the high-resolution structures of the oxy form with the ID 2DN1 and of the deoxy form with the ID 2DN2.

Is the heme group well defined in the electron density? Can you identify the iron based on its "density" value (rmsd level of the map)?

The paper mentioned in the introduction will give you a good idea about the scattering factors of different elements. Can you explain the differences between the density height of the Cryo-EM and the crystallography maps? How does sharpening or blurring influence the ability to detect heavy atoms (heavier than C,O,N)?

Prepare Pymol figures

For your presentation you want to prepare multiple images with Pymol. Before you load the density maps, run the command unset normalize_ccp4_maps first. If you use the GUI to load the maps, unset → "normalize_ccp4_maps".

Figure 1: Superposition of the biological assemblies of the models 8WJ1, 8WJ2, 2DN1, and 2DN2.

Sometimes the "easy" (aka. automatic) mechanisms are not the correct ones. When you have loaded the structures, you will recognize that only one of them has the expected tetrameric structure of a normal hemoglobin. You need to find and download the "biological assembly" for the other structures. Please download only the CIF-file based structure coordinates, because the PDB files will cause problems with Pymol.

Currently, Pymol (and all the other programs mentioned here) might have problems with the superposition of these CIF files of the biological assemblies. Please run the following Pymol code before you start to superpose the structures with 8WJ1 or 8WJ2!

remove chain "A-3"+"B-3"

You can superpose the structures either with the Pymol command "super" or by clicking → "Action" → "align" → "to molecule (*/CA)".

What are the corresponding RMSD values? Compare all four models with each other. How do the structures compare to each other? What is the most striking difference between the oxy and deoxy form?

Have a look into the Pymol setting "cartoon_cylindrical_helices". Sometimes it makes sense to reduce the visual complexity of a picture in a subtle way. You can modify the radius of the helices via the setting "cartoon_helix_radius".

Depiction of superpositions as "Cartoons" or other smoothed representations can give false impressions of how regions superimpose. You should avoid these and use a "ribbon" representation, but with the setting set ribbon_sampling, 1.
Figure 2: Should contain the panels A-D. Show oen of the heme groups of same subunit of the models 8WJ1, 8WJ2, 2DN1, and 2DN2 in a perspective in which also the proximal and distal histidine residues and their distances to the iron or the oxygen are visible. In an inset, show close-up pictures of the same heme groups with their respective electron density. Do not show other residues. Indicate the sigma-level of the density maps in the figure description.

General hints for the pictures:
To move the density maps according to the new position of the structure you need to use the command "matrix_copy". Look up the description of the command on the Pymol Wiki. You certainly want to show the iron atom inside of the heme group. You can do this with show sphere, (selection) where selection refers to the selector for the iron atom. The size of the sphere of an iron atom is very large though and might obscure other features, which are situated behind. You can change its size with the command set sphere_scale, [number], (selection). Choose a value for [number], which suits the expressiveness of the picture best.
You should reduce the complexity of the image by keeping the focus on the heme group and the histidine side chains.
- orient the heme group in the middle of the Pymol window and keep only a small space on all sides
- reduce the transparency of all the other objects you want to show in the background, e.g., cartoon of the surrounding structure
- use a bright (or brighter) color(s) for the objects in the background
Figure 3: Panel A+B: The so-called B factors can give an impression of the flexibility of the molecules.

You have heard that crystal contacts can alter the flexibility of regions that are stabilized by neighboring molecules in the lattice. Cryo-EM structures should in principle give an unbiased view on the flexibility, so comparing the structures from two independent method can be used to evaluate the results independently. Display the normalized B factors of the superposition of the oxy (panel A) and deoxy forms (panel B).

General hint:
The comparison of B factors between structures is generally problematic as they are for example connected to the resolution of the data. One easy way to deal with this problem is to normalize the data. Pymol can read Python scripts and you often can find scripts in the web to do such jobs. You can find a simple script here.

Browse to this (Python script) and have a look at the script to learn about its usage. Often this information is encapsulated within triple double quotes. You should also try to understand the code at least on a superficial way to make sure that it will not do harm to your computer.

Click with the right mouse button on → "Raw" and download the Python file (it should have the name b_normal.py) to your local disc. Put it somewhere close to your Pymol scripts.

You can load its functionality into Pymol with the command run [location of the script]\b_normal.py. After that you can access the command "b_normal".

Run help b_normal to re-read how the function is to be used. You need to assign the colors by B-factors (search the Pymol Wiki) and if you done this beforehand, you may need to run rebuild to re-render the actual image.

Make sure that you choose a representation that delivers the actual information. Using a "cartoon" might be too distractive.

Check out the possibility to use cartoon putty, which use the B-factor information to define the thickness of the "putty".

Predicting hemoglobin with AlphaFold 3

So, finally you want to try out the AlphaFold 3 server, because you have heard that it is capable of predicting the position of certain metals, ligands and prosthetic groups. Go to https://alphafoldserver.com/ and log in with your Google Account. Create one, if you don't have one at the very moment.

Input:
2 copies of protein (subunit alpha -> Uniprot P69905)
2 copies of protein (subunit beta -> Uniprot P68871)
4 copies of ligand (heme)

What do you think about the predicted model in terms of its reliability? Are the orientations of the alpha and beta domains correct? How does this correspond to the pLDDT values given by AF3? How did AF3 handle the incorporation of the heme? Which state does the model resemble more (oxy- or deoxy)? Superpose the AF3 model with both crystal structures. Were you able to decide which form AF3 should model?

Although it is a showcase from Google Deepmind itself, browse to https://deepmind.google/technologies/alphafold/impact-stories/ and have a look at two or three of their stories and try to understand how AF3 was used in actual research.

Summary

Use your final slide to describe the commonalities and differences between Cryo-EM and X-ray crystallography in a table. The table should contain at least the type of probe (X-ray, electrons) and the type of the resulting map. Is it possible (in theory) to distinguish between different types of ions? What do you think, where Alphafold or other prediction methods can help you in the future? Try to incorporate this into your final slide as well.