S-SAD on bovine insulin

The aim of this experiment is to determine the X-ray structure of bovine insulin from a cubic crystal form. The starting point is a data set collected at the BESSY synchrotron. The phase problem is solved by sulfur SAD phasing, i.e., the anomalous signal of the sulfur atoms of insulin is used to determine the protein phases. SAD stands for "single wavelength anomalous dispersion". A model is built manually in COOT and refined with REFMAC. Finally, the structure is analyzed and molecular figures are prepared for the protocol.

Create a new project in CCP4I2 named "insulin-[your_last_name]". For that create a new folder "insulin-[your_last_name]" in the folder practical.

Process the raw data images with MOSFLM as in practical 2 (Data reduction of diffraction data of a lysozyme crystal). The images can be found in P:\practical_protein_crystallography\data\insulin and will have a name ins_ssad_1_???.img.

Furthermore, omit the "Strategy" part as a full dataset is already collected. Output of the integration procedure will be by default a file ins_ssad_1_001.mtz.

After the integration is finished, close MOSFLM. The integrated data is saved automatically to the CCP4 database.

The program AIMLESS scales and merges the symmetry equivalent reflections.

• in the Task menu activate the submenu → "X-ray data reduction and analysis"
• choose → "Data reduction - AIMLESS"

A new window opens to input the data for the program AIMLESS. The "Input Data" is under normal circumstances selected automatically. The red background color, however, indicates you that you have to fill in some of the other fields.

• Always input meaningful job titles, e.g., "scaling and merging of mosflm data"
• In the field "Select unmerged data files", select the data output from the MOSFLM run.
• Use "insulin" as crystal name as well as dataset name
• In "Options for symmetry determination" select "Choose a known or previous solution"
• Below "Options for choice of space group or Laue group" select "Perform search then choose spacegroup". In the empty field on the right choose the space group I2₁3 (number 199). Write it like "I 21 3".
  You have to choose this spacegroup by hand, because you cannot distuingish between I23 and I2₁3.
• In the section "Optional existing FreeR set, define to copy or extend if necessary" choose 0.10 for "Fraction of reflections in generated freeR set".
  Ten percent of the reflections will be flagged as reflections to calculate the free R factor. These reflections will not be used for refinement. This so-called test data set should contain at least 1000 reflections, but also not much more, in order not to reduce the number of observations in the refinement too much.

Choose → "Run" to start the calculation.

Before actually starting the phasing procedure, it is helpful to examine the contents of the mtz file. Click on the symbol (triangle or plus sign) left of the AIMLESS job in the "Job list" (left big window). Click on "/insulin/insulin" (this name depends on the crystal and dataset name you have given in the AIMLESS run). The MTZ file may contain the following data. Later, you can open other MTZ files (like maps) in the same way. Have a look at the column labels, which are needed for different purposes (like storage of diffraction data, calculation of maps, phasing...).

H, K, L

Miller indices of a reflection

Iplus

the intensity of the +h,+k,+l reflection

SIGIplus

the standard deviation (error) of Iplus

Iminus

the intensity of the -h,-k,-l reflection

SIGIminus

the standard deviation (error) of Iminus

FREER

reflections set aside for calculation of the "free" R factor

F

structure factors

PHI

phase values

FOM

uncertainty of phase values (phase errors)

HLA, HLB, HLC, HLD

Hendrickson-Lattmann coefficients describing a phase value distribution

In the following also the intensities and standard deviations of the intensities of the Bijvoet pairs are given.

Thus, F is obtained by averaging all symmetry-related reflections. F(+) and F(-) are obtained by averaging F(+h,+k,+l) with its rotational symmetry mates and F(-h,-k,-l) with its rotational symmetry mates, respectively.

For structure determination you should have an idea how many protein molecules are located in the asymmetric unit. This is possible based on the typical solvent content of protein crystals, the size of the asymmetric unit and the size of the protein.

The necessary program is found under menu → "Import merged data, AU contents, alignments and coordinates" and then → "Define AU contents".

• Click on the → "plus sign" and select to → "Enter text".

• Give the name "chain A" and copy/paste or type in the sequence of chain A.

  If you want to type in the sequence, use always capital letters. The use of spaces is okay.

  Now → "Save".

• Repeat the last two steps for chain B.
• In the field "Solvent content analysis" select the correct "Experimental data".
• → "Run".

In this step, we have both defined the contents of the asymmetric unit and also calculated the so-called Matthews coefficient. Based on this value, we can estimate the probability (Matthews probability) for the presence of a certain number of protein molecules in the asymmetric unit (Number of copies). The corresponding solvent content (Solvent %) is also derived.

The programs of the SHELX program suite are used for locating the anomalous scatterers and calculating the phases. SHELX uses the anomalous differences to determine the location of the anomalous scatterers by direct methods. Following that, the protein phases are calculated. The phases are improved by phase refinement via solvent flattening. The programs outputs an mtz file with the calculated phases and figure of merit and a pdb file with the heavy atom positions.

Use in the Task menu → "Experimental Phasing" → "Automated structure solution - SHELXC/D/E phasing and building" .

• For "Start pipeline with" choose "Substructure detection" and for "and end with" select "Den.mod + poly-ALA tracing"
• Uncheck "Input protein sequence" and fill in the number of amino acid of insulin in "Number of residues per monomer""
• In "Crystal #1 composition and collected anomalous dataset(s)" specify the type of heavy atoms which should be found ("Substructure atom") and how many you do expect.
• You expect all cystein residues to form cystin bridges. Fill in the number of expected S-S pairs into the field after "Number of S-S pairs searched for as 1 supersulfur".
• In the tab "Advanced options", lower the number of trials in substructrue detection ("Num. trials") to 1000.

When determining heavy atom positions by either the SIR method or by anomalous dispersion (SAD) using Patterson methods or direct methods, there is always a 50 % chance that the wrong hand solution is obtained, i.e., the inverted (mirror image) coordinates of the heavy atoms are obtained instead of the correct coordinates. When phasing with SIR, both heavy atom solutions will give a similar quality map, but the electron density of the wrong hand solution will produce a mirror image of the correct electron density map, i.e., the alpha helices are lefthanded. When phasing with anomalous dispersion as in the current example, the correct hand of the heavy atom solution will give better maps or interpretable maps. This is usually manually inspected by looking at the electron density maps for both solutions. SHELX therefore provides phases obtained from both solutions.

Below the SHELX result you will find the option to run the program COOT. Choose then "anomalous substructure coordinates" for "Coordinates". In "Electron density maps" you load the maps of both hands in parallel. By default the correct solution (hand) is chosen as "Map coefficients". It has the name "Best electron density map coefficients". To add another density map, click → "Show list" and then on the → "plus sign". Select the second "Best density - other hand" as "Map coefficient". Now → "Run".

Both density maps are loaded with a quite similar color. You can choose to show/hide a map via the "Display Manager" in the top menu of COOT. Just tick/untick the "Display" box next to the density you want to show/hide.

Analyze the electron density maps and the graphs of the SHELX run:

• How many S atom positions have been found by SHELX and what is the occupancy of these sites?
• Compare the electron density maps of the two solutions? Has a correct and good solution been obtained?
• Can you identify side chains in the electron density?

• Based on the positions of the disulfide bridges, try to find the sequence around one or two of the cysteines.

  Document your "sequencing procedure" for the protocol by making snapshots.

With an electron density map of this quality, automatic model is usually the first choice and works well, as we will see later. It is, however, more instructive to manually build a model based on the electron density map. In this practical, the aim is to build a model of insulin with the program COOT based on the experimental electron density map.

General tips and procedures:

• Building a protein structure from scratch is like building a house (or car, ...). You need the right tools, in a potentially logic order. Look at the flowchart below. It tries to summarize the general way you might want to follow in the practical. The order, however, is not as fixed as the scheme suggests. For example, if you want to renumber the residues first and then do the Mutation step, it will perfectly be fine.
• It is very helpful to activate the names of the COOT tools on the right side. Go to → "Edit" → "Preferences". Select the tab → "Refinement Toolbar" and in the section "Toolbar Style" click on → "Icons and Text".
• As soon as you add a new C-alpha trace, there will be a unique number associated with that chain in COOT. You can see the number in the Display Manager. It is advisable to remember that number (e.g., by writing it down) especially in the mergeing or renumbering step or for saving the coordinates (COOT will ask for the Select molecule number to save). Because each chain is now separated into a new molecule, you need to merge the S-SAD on bovine insulin Use → "Calculate" → "Merge Molecules..." to do so.

  After converting any C-alpha trace to an poly-ALA model or before beginning to model a new C-alpha trace, you should always delete the "Baton Atoms" (in the "Display manager")! Forgetting this will usually create a very confusing model that is extremely hard to fix.

• The electron density map from the SHELX run is calculated from imperfect phases. These were derived from the the positions of the anomalous substructure and the subsequent density modification. As a direct result of the phase imperfection the electron density might contain errors or is weak or even absent for side chains or complete residues. Only try to build the side chains or amino acids you can clearly see in the electron density.
• Give the chains the correct chain ID ("A" in the PDB file should be chain A of insulin) and the right numbering (the first amino acid should have the number 1). The name of the chains is either the character "A" or "B".
• Although the experimental map also reveals some water positions (electron density blobs aside of the main chain), we first want to refine the model and add the water molecules later.

          Initial electron density
                     |
                     |
             Activate map skeleton
                     |
                     |
    ╭──────> C-alpha baton mode  <------>  Reverse direction?
    │         (create CA trace)           (correct C->N error)
    │                |
    │                |
    │        Ca Zone -> Mainchain    <-->  Reverse direction?
    │         (convert CA trace to        (correct C->N error)
   Add          poly-ALA chain)           
  second             |
  chain              |
    │         Mutate & AutoFit   <------> Add residues?
    │         (assign sequence)           (extend termini)            
    │                | 
    │                |
    │        Renumber residues +
    │          Change Chain ID
    ╰─────< (correct register and
              names of chains) 
                     |
                     |
                Merge chains
                     |
                     |
                Initial model

Open the correct program via → "Model building and Graphics" → "Autobuild with ModelCraft, Buccaneer and Nautilus".

• In the field "Reflection data" select for "Reflections" the output of the AIMLESS job. It should have the name ("/insulin/insulin"). The "Free R set" should also be chosen from the same job.
• Untick "Get initial phases from refining the stating model" as you want to include the phases you have been obtained in the SHELX job. Define the "Phases" as ""Best" phases".
• In "Asymmetric unit contents" choose the appropriate "Asu content file".
• In "Optional pipeline steps" select the following steps and unselect the others: → "Classical density modification with Parrot" and → "Addition of waters".
• Hit → "Run".

Automatic model building is in general the first step towards the new molecular structure. Sometimes the initial density is not as good as in this case, so by using the information from the molecular model might help to improve the phases and thus obtaining a better and more interpretable electron density map.

The automatically built model might not place the model at the same position in the asymmetric unit as your model is situated. Open in COOT your model as well as the model built automatically. To compare both structures more easily, use a superposition method to place the automatic structure onto your model. In COOT go to → "Calculate" → "SSM Superpose...". As "Reference Structure" choose the model you built, "Moving structure" is the autobuilt model. Hit → "Apply".

Refining a structure needs both reciprocal space refinement as well as real-space refinement. Below is a flowchart this process:

                Initial model      
                     |
                     |
    ╭───────>  Reciprocal space 
    │           refinement
    │        (geometry, B factors)
    │                |
    │                |
    ╰─────────<  Real space      <------>  Add waters, ligands, ... 
                 refinement               (New things visible in
           (side chain conformations,      the electron density?)
           water positions, ligand       
             conformation, ...)
                     |
                     |
                Final model

In CCP4I2, open in the task menu "Refinement" and then "Refinement - REFMAC5".

• In "Main inputs" choose as "Atomic model" in the first run the initial model you built in step 7. If the model is not available in the drop down list, you should use the "Browse files" symbol next to the menu.
• "Reflections" and "Free R set" should be selected from the AIMLESS job.
• You want to "Use the anomalous data" to "only calculate an anomalous map".
• Under "Options" specify that you want to do 20 cycles of refinement.

Start the refinement. After the refinement has been started, the log output will show the results from the refinement run, even if the job is not yet finished. Check that there are no error messages and everything works as expected. Watch how the R-factor and free R-factors change.

Next examine the refined model in COOT.

• In CCP4I2 first define the coordinate file from REFMAC5 as "Coordinates".
• Then as "Electron density map" use the "Weighted map from refinement" and as "Difference density map" you need to select the "Weighted difference map from refinement".
• Inspect all residues from the N-terminus to the C-terminus of both chains to see if everything appears to be correct and fits the density well. The density should be better than the first experimental electron density from phasing, since the model phases are better than the experimental phases (if the model is mostly built correctly and has a reasonable completeness). If necessary, correct conformations or add missing residues.
• Helpful options when analysing the structure: Display symmetry related molecules via "Draw" → "Cell & Symmetry" and then activate "Symmetry On". With this option you can analyze, if density at crystal contacts is explained by symmetry related molecules. Analyzing interactions of a residue/molecule: "Measures" → "Environment Distances" → "Show Residue Environment?".

Rebuild parts of the model, if necessary.

• Alternate conformations. Based on the high resolution electron density map, alternate conformations of residues may become apparent. Add these with the option "Add Alt Conf..." in "Model/Fit/Refine" or on the toolbar on the right side. If there is only one alternate conformation for the side chain (usual case), choose "split at Calpha". If also the main chain atoms show an alternate conformation, choose "Split all of a single residue". Often, this requires also the previous/next residue to have an alternate conformation.
• For the Gln, Asn, and His residues also check if the side chain conformation is reasonable with respect to a 180° flip based on hydrogen bonding interactions. You may need to check interactions to symmetry related atoms for this purpose (option "Draw" → "Cell & Symmetry"). If necessary, change the conformation.
• Are residues with ill defined density present? What is the cause of the "bad" density?
• Adding water molcules: Before going through the refined model, water molecules should be added, such that their positions can be checked together with the protein residues. Use the option "Calculate" → "Other Modelling Tools..." → "Find waters ..." to add the water molecules. Choose the current difference density map and the current model. Pick peaks higher than 3.0 sigma. Some water molecules will be placed in protein residues that have not yet been built or in density due to alternate side chain conformations. Delete these waters when going through the chains ("Delete" in "Calculate" → "Model/Fit/Refine" or on the toolbar on the right side).

Save the molecule to a PDB file and do another refinement with REFMAC5 (20 cycles). After that, check the protein model and the water molecules again based on the electron density and the points mentioned above. The R-factors and density should further improve due to the addition of water molecules in the previous cycle. Search for further water molecules.

After you finished the refinement cycle:

In CCP4I2:

Choose the in the task menu "Validation and Analysis" the "Multimetric model geometry validation". Select the correct model and reflections (from the AIMLESS run).

The output will be printed in the "Results" section. Use a screenshot to get an image of the Ramachandran plot(s) for the protocol. Furthermore, you get the number of residues in the favored, allowed and high-energy backbone-conformations (outliers) from the Ramachandran plot.

Are there also rotamer outliers? If so, is the density well defined for these residues? The program also analyzes the orientation of the side-chains of asparagine, glutamine and histidine and suggests, whether they need to be flipped. Check the hydrogen bonding interactions of the side chain to validate, if a flip is necessary.

• Load the last REFMAC5 refined structure with COOT. This time, you need to add an "Anomalous density map". Choose "Weighted anomalous difference map from refinement" from the last REFMAC5 job.

• Analyze the disulfide bridges. Which cysteine residues are involved in disulfide bridges? Which disulfide bridges link the two insulin chains?

• Analyze the crystal packing: This is best done by using a C-alpha trace representation of the molecule (in the Display Manager) and the symmetry related molecules. Use "Draw" → "Cell and Symmetry" and in "Symmetry by Molecule" switch to "Symmetry on" and activate "Display as CAs" for the current model. Activate "Show Unit Cell". Increase the radius to 50 Å.

  How is the molecule packed? Save a snapshot for the protocol.

• Analyze the anomalous map based an the measured anomalous differences and the final model phases. What is the peak height of the weakest and the strongest sulfur atom? What is the peak height of the strongest anomalous density peak that is not caused by a sulfur atom?

• Analyze where the hydrophobic and where the hydrophilic residues are mainly located. Where are the water molecules located? Are water molecules found in the center of the protein?

For the next step in this practical you need to export the electron density maps. In COOT go to "File" → "Export Map..." and choose the 2Fo-Fc map ("Weighted map from refinement"), the Fo-Fc map ("Weighted difference map from refinement") as well as the anomalous map ("Weighted anomalous difference map"). Save them to the "insulin" folder as "refmac_2fo-fc_final.ccp4", "refmac_fo-fc_final.ccp4" and "refmac_ano.ccp4", respectively.

Save the model coordinates into the same folder as well. This can be done in COOT with "File" → "Save coordinates...". Use the file name "refmac_final.pdb".

Before you can start working with PyMOL, you need to export some more coordinates and maps with COOT.

In CCP4I2 reload the SHELX run by clicking once on its entry in the CCP4i2 database. Choose to start COOT (on the bottom of the results). In the section "Coordinates" select as "Atomic Model" the "Anomalous substructure coordinates".

The "electron density map" should be the "Best electron density map coefficients".

After COOT is loaded, export the "Best electron density map coefficients" as "shelx_phasing.ccp4" in the "insulin" folder. Also save the coordinates from the anomalous substructure as "shelx_substructure.pdb".

Now make yourself familiar with PyMOL. The best is just to load your structures and to start!

The following figures should be prepared:

1. An overview of the fold of insulin in cartoon representation. The two chains are represented in different colours and the secondary structure elements as well as the termini of the two chains are labeled. The disulfide bridges are shown, too.

   First, change the background to white. All figures should be prepared with white background. Most figures look better with the "Maximum quality" settings. Activate it in "Display" → "Quality".

   Load the pdb file "refmac_final.pdb".

   Create two objects ("fold", "ss-bonds") from the loaded structure by the command 'copy'. Hide all. Show object 'fold' in cartoon representation and object 'ss-bond' with ss-bonds related cysteine residues as sticks. Color both chains of 'fold' in different color, e.g., marine and red. Change color of 'ss-bonds' residues to yellow. Label every alpha and beta element, also termini.

   Orient the molecule in a way you like and in which you can see all the labels. Use the "Editing" mode of PyMOL to move labels, if necessary.

   Save the actual orientation matrix with the command "get_view". That matrix will be printed into the console output of PyMOL.

   You can ray-trace the figure with the command "ray". It might take a while to render on your computer. Save the rendered figure (further on called scene) with the name "scene_A" via "File" → "Export Image As" → "PNG...".

2. A C-alpha representation of the fold in the same orientation. Use the same settings as before, except that no labeling of the secondary structure elements is given. Use the orientation matrix from figure (A). To do that, mark the lines with the "set_view" command and copy these into the PyMOL command input area.

   Copy a new object ('calpha') from 'fold'. Disable the 'fold' object. Hide labels. In the new object (rename it to "fold2") hide cartoon and show ribbon. If you activated the "Maximum Quality", your ribbons will drawn in a roundish fashion, however, ribbons are often shown with sharp edges. Use the command "set ribbon_sampling, 1" to get the sharper edges and use "set ribbon_sampling, 10" for the rounded ribbons. Choose the representation you like most.

   Save scene as "scene_B".

3. A view of an all atom representation (sticks) of the whole structure. The two chains are distinguished by different colors of the carbon atoms. The water molecules should also been shown.

   Create new object 'sticks-all' from 'calpha'. Disable every other object. Show sticks for 'sticks-all'. Enable wall-eye stereo. You need to adjust the view afterwards.

   Save scene as "scene_C".

   Besides wall-eye stereo representation, there are other ways of depicting 3D figures (like cross-eye stereo or anaglyph stereo). Thus, mention in the protocol, which type of representation you have chosen.

4. Figures of residues and their density. This includes a disulfide bridge, residue with alternate conformations, disordered regions, a well defined region and water molecules.

   Disable every other object. Create new objects for the residues to be displayed. Load the maps "refmac_2fo-fc_final.ccp4" and "refmac_fo-fc_final.ccp4". You need to create for both maps three electron density objects (with the command "isomesh"). Show the 2Fo-Fc electron density in different colors. Use green and red for Fo-Fc density. This is also the default of PyMOL.

   Save scenes.

5. The sidechain of an Asn, Gln or His residue, for which the 180° sidechain flip conformation, can be unambiguously determined based on the hydrogen bonding pattern. Show the residue and the hydrogen bonds (as dashed lines with the distance in Å) in both conformations.

   Disable former objects. Create a new object with the chosen amino acid and surrounding (4 Å distance) amino acids. Show sticks and make hydrogen bonds with 'measurement' wizard. Move distance labels so that they are visible.

   Save scence.

6. An anomalous |F+ - F-| Fourier map with the final model phases. Show the S atoms found by SHELX and the cysteines of the final refined model together.

   Disable former objects. Load the coordinate file "shelx_substructure.pdb". Show the sulfur atoms of the SHELX anomalous substructure and of the dedicated cysteines residues as spheres, but scale down sphere radius to 0.3 (with the "alter" command). Color sulfurs in yellow. Load and show the anomalous density map "refmac_ano.ccp4" around the "shelx_substructure".

   SHELX does not necessarily place the sulfur atom of the anomalous substructure in a way that they belong to the final model of insulin. Actually, their position are arbitrarily chosen. You can create the symmetry related sulfurs via the "Action" menu → "Generate" → "symmetry mates". Choose a "within" radius appropriate for you.

   Save scence.

7. The experimental electron density map together with the final model. Show only part of the model such that the quality of the map can be judged well.

   Disable former objects. Create a new object, show ribbons and lines. Zoom into a region for which you want to display the electron density. Use COOT to decide, which region you want to show. Maybe show a disulfide bond or a region inside the protein with well defined aromatic residues.

   Hide ribbons, show cartoon. Fine tune helix and strand appearance. Change the color of the helices and strands as you want. Create another object for residues you want to show as sticks, show them as sticks.

   Load in density map "shelx_phasing.ccp4" and show density.

   Save the scene.

Faust, A., Panjikar, S., Mueller, U., Parthasarathy, V., Schmidt, A., Lamzin, V.S., Weiss, M.S.. (2008) 'A tutorial for learning and teaching macromolecular crystallography' Journal of Applied Crystallography, 41, (6), 1161-1172. Paper