Normal Mode Analysis (PKU)
- 1 Short Task Description
- 2 WEBnm@
- 3 elNemo
- 4 References
Short Task Description
This week, we will perform a normal mode analysis of our wildtype protein. We will calculate visualize large coordinate movements to identify e.g. domains and try to watch ligand interaction. For this, we will use the webnm@ and elNemo webservers. See the task description for details, a journal of commands and scripts, if necessary, can be found here.
Normal Mode Analysis
In normal mode analysis (NMA) the protein is modeled as harmonically oscillating system to e.g. describe conformational changes. Normal modes are much faster to calculate than a molecular dynamics simulation, especially as usually very few low-frequency modes suffice to describe a proteins motions.<ref name=ElNemo>Karsten Suhre and Yves-Henri Sanejouand; ElNémo: a normal mode web server for protein movement analysis and the generation of templates for molecular replacement. Nucleic Acids Research Volume 32, Issue suppl 2</ref> Often, models do not include the sidechains of a protein so the effect of mutations can not be detected. Also, non-harmonic motions are not observable which limits the insights gained from NMA in comparison to the much more detailed molecular dynamics simulation.
There are 14 unmutated structures of human phenylalanine hydroxylase, some of them with different ligands e.g. the co-factor tetrahydrobiopterin, the co-factor and thienylalanine (thienylalanine is a receptor antagonist to the substrate phenylalanine) or the co-factor and norleucine. As can be seen from <xr id="tab:rmsd"/> of RMSD values below, unfortunately they are all in the same conformation. There are also phosphorylated and unphosphorylated structures of PheOH of rattus norvegicus (1PHZ and 2PHM in <xr id="tab:rmsd"/>) but they also are virtually identical with a RMSD of 0.330. When we submit 1J8U and e.g. 1MMT, the structures with the highest RMSD, we get an overlap of the high frequence modes and the 'conformational change' of over 0.82, which is neither surprising nor of much value for the analysis as the structural changes are very small and limited to some displaced loops. Therefore we cannot compare different conformations, but use the 1J8U structure with and without ligand in the analysis.
<figtable id="tab:rmsd"> RMSD of human phenylalanine hydroxylase structures in PDB to the 1J8U reference structure
|PDB ID||RMSD to 1J8U|
|1PHZ (rattus n.)||0.380|
|2PHM (rattus n.)||0.409|
This mode shows a high atomic displacement for the start (residue 20-40) and parts of the end begining with residue 240-300 in <xr id="fig:mod7dis" /> this also can be seen in <xr id="fig:mode7" /> where the image of a beating heart comes to mind. We therefore classify this as a pumping motion which opens the protein to the substrate.
Like in Mode 7 this mode opens the catalytic site which can be either to be able to bind, or to release the substrate or the product in respect. In difference to <xr id="fig:mod7dis" /> there is a rather small part (residue 25-40) of the molecule in this mode which is moving, as one can see in <xr id="fig:mod8dis" />. This is also easy to be seen in <xr id="fig:mode8" /> where only the lower left part of the protein is moving
Like in Mode 8 only a small part (residue 290-300 in <xr id="fig:mod9dis" />) are moving a lot, but still the protein is in some kind of pumping dynamic (<xr id="fig:mode9" />) with mostly the lower right side moving. This creates the idea of compressing the room inside the protein, in order to rearrange or even retract the substrate.
This mode is the first of the provided modes, we would not associate with a pumping or beating heart, but rather with a stretching motion, there is no clear tendence to be found for the protein which affects binding or catalytic domain (<xr id="fig:mode10" />). In difference to the structural changes, the displacement plot in <xr id="fig:mod10dis" /> shows a clear sign of strong movement in the beginning (residue 25-40) as well as in the end (residue 270-300) with a bit smaller movement shown by some peaks in the middle of the sequence.
This mode again moves only the end and the beginning of the protein, with similar compartmentalization like the modes 7 to 10 (<xr id="fig:mod11dis" />), but this time, the movement classifies as a torsion of the protein, which shears along the diagonal of the protein in <xr id="fig:mode11" />. This could either be a random movement or the torsion, which splits the substrate into pieces.
This mode unlike any other we had before offers a very small atomic displacement range (<xr id="fig:mod12dis" />) as the peak is only about 3 at residue 40, but the structure is moved changed quite a bit in this mode. This mode deforms the binding site in the center of the protein and also stretches the protein to the diagonal of the protein (<xr id="fig:mode12" />)
</figure>In this plot (<xr id="fig:correleationWebnma" />) one can see, that the correlated movements are very sparse in our protein. Only a few small correlations apart from the diagonal can be seen. One example is around residue 240 or 300. Apart from these two, there is no real correlation among the domains or intuitive way of finding or seperating domains from this plot alone.
ElNemo computes ten perturbed models for the first five non-trivial normal modes of the input protein as a standard and perturbations for the lowest 25 modes on request in a separate queue. There is no fixed limit to the size or number of modes but too long running jobs will eventually be killed. The calculation uses a cutoff value to determine which atom-atom interactions are kept in the elastic network model and another parameter to determine how many residues are treated as rigid body to speed up calculation appropriately.
1J8U (without ligand)
Modes 7, 9 and 10 show rather random motions of the protein that are hard to evaluate. Mode 11 and especially mode 8 could show an opening of the catalytic subunit to allow access to the active site. Mode 8 is also the only one, where the termini of the sequence do not move. As the crystal structure is truncated on both ends, the strong movement other modes show there is probably arbitrary .
</figure> <figure id="fig:1J8U_nm7_fluc">
</figure> <figure id="fig:1J8U_nm8_fluc">
</figure> <figure id="fig:1J8U_nm9_fluc">
</figure> <figure id="fig:1J8U_nm10_fluc">
</figure> <figure id="fig:1J8U_nm11_fluc">
1J8U and BH4 ligand
We included the ligand tetrahydrobiopterin and the FE(II) molecule in the normal mode analysis. Interestingly, except mode 9 the modes including the ligand appear to be inverted to the ones without ligand. The fluctuation matrices are almost identical but inversely colored, that means, 'opening' motions are now closing motions and vice versa. Modes 7 and 8 are additionally switched. As several crystal structures with different ligands are available (cf. our introduction), we know that different ligands influence the conformation only little. This applies for the wildtype protein at least, as there are disease causing mutations that respond to large doses of tetrahydrobiopterin with restored activity.
</figure> <figure id="fig:1J8U+BH4_nm7_fluc">
</figure> <figure id="fig:1J8U+BH4_nm8_fluc">
</figure> <figure id="fig:1J8U+BH4_nm9_fluc">
</figure> <figure id="fig:1J8U+BH4_nm10_fluc">
</figure> <figure id="fig:1J8U+BH4_nm11_fluc">
To see if independent movement of the domains is observable, we submitted another structure that contains the catalytic domain and the binding domain of phenylalanine hydroxylase, 1PHZ. Indeed show the matrices of fluctuation of the C alpha backbone of the low frequency modes almost exclusively the movement of the domains as unit and little intra-domain movement. In <xr id="fig:1PHZ_nm7_fluc"/> the large uncolored area in the nine lower right squares indicates how the catalytic domain moves collectively as rigid unit, the smaller mostly uncolored area in the upper left square shows the less rigid but still coordinated binding domain. See the animation in <xr id="fig:1PHZ_nm7"/> for a nice visualization.