Analysis of NMR titration data and docking results in the study of biomolecular complexes

Thanks to several Utrecht NMR Research group lab members

This practical consists of two parts:
    Note: To run this practical make sure that SPARKY is installed on your system, otherwise download it from the SPARKY website.

The data required to run this tutorial can be downloaded as a tar file: titration.tar (7.0 MB). Untar it using the following command tar xvf titration.tar This will create a directory called Titration-Data containing all the necessary files.

A helpfull link for this practical:

NMR is a very powerful technique to map the interacting site of a protein upon complexation to its partner. The technique is based on analysing 15N-HSQC spectra. In an HSQC spectrum, one can observe connections between the amide protons and the nitrogens connected to it. The nitrogens have to be of the 15N isotope for this NMR experiment; so one needs 15N labelled protein. An HSQC spectrum is like a nice "fingerprint" of the protein because each peak correspond to an NH of each residue. You will get as many peaks as the number of residues in the protein (except for the first amino acid and the prolines). Side chain NH groups (from Asn, Gln, His and Trp) can also give a peak in the spectrum. The chemical shift in both nitrogen and proton dimension is sensitive to the chemical environment of the two nuclei. Therefore this experiment is a nice technique to map the interacting site of a protein with its partner. One first records an HSQC of the 15N labelled protein alone and then HSQC spectra of the protein complexed with increasing amounts of its unlabelled partner. The peaks showing a change in the chemical shift in the HSQC of the complex compared to the HSQC of the protein alone have a different chemical environment. This shift can be due to two phenomena:
  • When adding the partner, chemical shifts of the residues at the interface will be perturbed by the proximity of the partner.
  • The addition of the partner may induce structural changes, or conformational changes in the protein itself.
The analysis of the shifts and the mapping of these shifts on the protein structure will give information on the binding interface and the conformational changes of the protein. In this NMR practical, you will have to analyse the HSQC spectra of a protein, UbcH5B, upon complexation to its physiological partner, the Not4 Ring finger. The assignment of the protein has already been done. You will assign the HSQC spectrum of the protein in its bound form and plot the shifted residues on the structure of UbcH5B in order to visualise the binding area of UbcH5B.

UbcH5B and Not4 fulfill an important function in the ubiquitination pathway when they are in complex with one another. Therefore the study of the complex is important to gain insight into the molecular basis of this interaction. Due to various reasons it can be difficult to solve the molecular structurer of the complex by NMR. But, as you know by now NMR is a powerful tool to study the interaction with titration experiments, for instance. In this practical you will gain insight into the binding interface of UbcH5B by mapping shifted residues on the protein structure. We performed the reciprocal experiment and determined the shifted residues in Not4 upon binding of UbcH5B. Now we know the interaction interface of both proteins and we can use this information to predict the complex. In the second part of this practical you will work on this prediction process by brief introduction to our dat-driven docking program HADDOCK.

Titration analysis of UbcH5B upon complexation with Not4 ringfinger

We will use the software SPARKY to analyse the spectra of UbcH5B. Type Sparky to start the assignment software. To open the project for this exercise go to file -> Project/Open and select titration.proj. The UbcH5B HSQC spectrum will be presented on the screen. The other spectra can be displayed by selecting them under the menu-item Windows. For closing the spectra don't click on the upper right corner; but use the shortcut vh or View -> Hide View; then they can be accessed at any time again inside the project. All the spectra are connected so when you will move through one of them, the other ones will also move. This is made possible in View -> More -> Synchronize. You will see that the 15N and 1H axis are connected. Specific View options of each spectum can be accessed by the shortcut vt or View -> View options. The menu can also be accessed by right-clicking inside the spectrum. In this project we have the following HSQC's:
  • ubch5, the 15N labelled UbcH5B protein alone.
  • ubch5_not_1-8, added 1/8 of the unlabelled Not4 Ring finger protein
  • ubch5_not_1-4, added 1/4 of Not4
  • ubch5_not_1-2, added 1/2 of Not4
  • ubch5_not_1-1, the 1:1 complex of UbcH5B and Not4
Make an overlay of the 1-8, 1-4 and 1-2 HSQC spectra; Open the window for ubch5_not_1-8 and select view -> More -> Overlay Views. Select ubcH5B_not_1-4 in From view and ubch5_not_1-8 in Onto view. Click add. Select ubcH5B_not_1-2 in From view and ubch5_not_1-8 in Onto view. Click add and close the window. You should see the peaks shifting from red to blue in the window of 1-8. One can zoom in and out in the spectra by using the shortcuts zi and zo, or by changing the pointer to zoom capabilities (F11).

Now open the 1:1 complex spectrum (Windows -> ubch5_not_1-1). Copy the HSQC peak list of ubch5 alone onto the ubch5_not_1-1 spectrum. To do so, in the ubch5 window, select all peaks (pa), copy them (oc). In the ubch5_not_1-1 window, paste the peaks (op). Center the peaks (pc). You will see that most of the peaks have been centered to their new position, but the peaks corresponding to a significant shift couldn't be automatically centered in the new spectrum. You will have to adjust them manually. To do so, go through the spectrum and adjust the peaks by changing the pointer to (select)(F1). Move the box close to the corresponding peak (use the overlay spectra to follow the shift) and center it (pc).

When the assignment has been done; the assignment table can be opened; Peak -> assignment -> Assignment table. Select standard deviation and deselect chemical shift to see which residues shift the most. Save this file as shift.tbl in /home/courseX/practical/.

In your linux window, type xmgr.
Go to File -> Read Block Data and select shift.tbl. Select X from column 1 and Y from column 2. Accept, close.

Go to Graph -> Autoscale, select autoscale axis: all. Accept and close. Go to Graph -> Symbols, select histogram instead of no symbol, sym size: 49, line properties: none, accept and close.

Go to print -> Print to file and give a file name.

Do the same thing for the nitrogen dimension by choosing in Read -> Block Data y from column 3.

Define the residues having the biggest shifts upon complex formation.

You have calculated the chemical shift difference for the nitrogen and proton dimension separately. Now the combined chemical shift difference has to be calculated. This can be done using the equation:

combined shift difference = [(proton shifts)2 + (nitrogen shifts/6.51)2]0.5

A significant shift can be defined as a values higher than the average of all combined shifts plus one standard deviation. We have calculated this for you. The eventual significantly shifted peaks for UbcH5B in complex with Not4 are:
    Ala2, Leu3,Arg5, Ile6, Glu9, Leu10, Ans11, Asp12, Thr58, Asp59, Tyr60, lys63, Ser94, Ala96, Thr98, Ile99

Open UbcH5B.pdb (in the directory pdb-files/) using rasmol (type: rasmol UbcH5B.pdb in your window). Select the residues having the largest shifts and color them. Conclude. You can verify your success in assigning the shifted peaks by looking at the eventual results here

Prediction of the complex between UbcH5B and Not4 using the titration data and the docking program HADDOCK.

Now that you have assigned the shifted peaks in UbcH5B upon binding to Not4 you have information about the interaction site. An important aspect in the study of biomolecules and their interactions with other molecules is to get insight into the interactions at a structural atomic level. For UbcH5B we would like to study the interaction with Not4 in this way. However, as already mentioned in the introduction it is not always possible to solve the structure of the complex by NMR using classical NOE distance restraints. Aspects like obtaining a good sample of the purified complex in high concentration or detecting NOEs in the case of weak and transient complexes can be difficult.
But, if we know the 3D structures of the native components of a complex and have some information about their interaction surfaces, then we can try to predict the structure of a complex using a computational docking approach. In our department we developed the software package HADDOCK for this purpose. You can learn more about HADDOCK by having a look at the available online presentation.

For UbcH5B and Not 4 the structures of the proteins in their native form are available. For UbcH5B you have mapped the interaction interface. We have performed the reciprocal experiment with Not4. The significantly shifted peaks for this protein are as follows:
    Leu16, Cys17, Met18, Cyus41, Asp48, Glu49 and Arg57

We now have information about the interaction interface of both molecules and we can use this to predict the complex. We cannot use all residues however. If any residue in the interaction interface of molecule A wants to have an interaction with any residue in the interaction interface of molecule B it must be accessible at the surface. Therefore we have to filter the shifted residue list on surface accessibility. with the program NACCESS you can determine the surface accessibility of all residues of UbcH5B and Not4. We have done this for you on the NMR ensemble of both molecules. The average relative solvent accessibility for each residue can be found the the files ubch5b_rsa_ave.lis and not4_rsa_ave.lis in the data directory. We define a residue as surface accessible if it has relative surface accessibility higher than 50%. To display all residues that fulfill this definition, type the following command string in your terminal (in the directory of the *_rsa_ave.lis files):
    awk '{if (($5+$6)>=50 || ($7+$8)>=50)) {print $0}}' (filename)_rsa_ave.lis

From left to right you will see the following columns of data: residue name, residue number, all atoms relative surface accessibility plus standard deviation, backbone relative surface accessibility plus standard deviation, and side chain relative surface accessibility plus standard deviation. Go through the list for both molecules and determine which shifted residues are not surface accessible.
The surface accessible shifted residues of both molecules are used in the docking process to position the two interfaces and bring the molecules together. This is done by translating the residues into Ambiguous Interaction Restraints (AIR). An AIR is defined as:
  • An ambiguous distance restraint with a maximum value of 2.0 A between any atom of an active residue i on molecule A and any atom of all active and passive residues of molecule B.
In the definition of those residues, one distinguishes between "active" and "passive" residues:
  • The "active" residues are those experimentally identified (e.g. from chemical shift perturbation) to be involved in the interaction between the two molecules and solvent accessible.

  • The "passive" residues are all solvent accessible surface neighbors of active residues.

For our two proteins the AIR restraints were defined using the following residues:<
    For UbcH5B:

  • active: Leu3, Arg5, Thr58, Asp59, Lys63, Ser94, Ala96, Thr98

  • passive: Met1, Lys4, His7, Lys89, Asp29, Met30, Phe31, His55, Phe62, Lys66, Gln92, Pro95

    For Not4:

  • active: Leu16, Cys17, Met18, Glu19, Phe40, Asp48, Glu49, Asn50, Leu52, Ala55, Cys56, Arg57

  • passive: Glu13, Pro15, Pro20, His34, Arg44, Thr47, Pro54

The docking yielded 200 structures. A plot all solutions with their intermolecular energy as function of the RMSD from the lowest energy solution looks like:

A few clusters can be distinguishedby eye. The solutions in the two most populated clusters have almost equal intermolecular energies but different RMSD values. We picked three solutions from the clusters. You can find them in the pdb-files directory as ubch5b-not4_1.pdb, *_2.pdb and *_3.pdb. Inspect the structures using rasmol or pymol. See if you can identify the most likely solution by coloring the active and passive residues and using the definition of an Ambiguous Interaction Restraint (see if any of the restraints are violated).

You should be able to identify one structure that does not match the criteria. The other two are (almost) equally probable. To distinguish we can make use of other available data that give some information about the interaction. For instance, we know from mutagenesis studies that Glu49 of Not4 interacts with Lys63 of UbcH5B (Winkler et al., 2004). Evaluate the two remaining solutions using the mutagenesis data.

Look at the figures of the solutions to see if you concluded correctly.

Links to figures.