| Goal |
| Materials |
| Methods |
| Results |
| Conclusions |
| Reference |
Have you ever wondered how
proteins recognize respective ligands without the help of any
sophisticated organs such as eyes, noses, or ears? It is
the magic of their complex tertiary structure, whose properties
usually determine if the interaction with a particular substrate
can take place or not. This leads us to another question:
What are these properties then?
During the past ten weeks,
we have been trying to approach the answer through the study of
guanine-binding proteins. We aim to quantify their binding
site by developing numerical templates which describe the protein
environment around guanine surfaces. It will enable us to
compare different protein structures and look for patterns in
protein-ligand interaction.
The raw materials that we
base our templates on are 84 guanine-binding proteins located
in the Protein Data Base (PDB). They all contain one form
of guanine moiety or another and most of them play important roles
in biological systems through specific recognition of various
ligands. For instance, a subclass of them, Ras proteins,
are involved in signal transduction from growth factor receptors
on the plasma membrane to gene activations within the nucleus.
The mysterious switch that turns them on or off is
GDP-GTP exchange. In their active state, ras bind guanosine
triphosphate; they are deactivated when GAPs (GTPase activating
proteins) simulate the hydrolysis of GTP to GDP. The
process is reversible by the interaction with certain receptors.
Top
We started from defining a plain guanine surface with 440 vertices (Figure-1 and Figure-2). For each vertex, twenty-seven properties were calculated. Twenty of them are amino acid properties, i.e. the shortest distance to every kind of ?-amino acids (alanine, arginine, asparagine, etc). Six are atom type properties, i.e. the shortest distance to aliphatic carbon, aromatic carbon, basic nitrogen, neutral nitrogen, acidic oxygen and neutral oxygen respectively. One is self-property, i.e. the shortest distance to the non-guanine part of their ligands. We performed data reduction by transforming the distance arrays to density arrays with the function y=e^-(x*0.9)^4. The purpose of doing so is to flag those parts touching the surface with density of 1 and those very far away from the surface with density of 0. Moreover, it allows a smooth transition in between (Figure-3). To avoid the overweighing of a particular kind of protein, we ran a sequence homology analysis on our 84 proteins and came up with 34 distinct (over 10 amino acid difference) structures. A list of them is shown in Figure-4. These individual templates were compared to each other for similarities in the binding environment. We tried to see how they correlate with the evolutionary relationship of Guanine-binding proteins, which is suggested by their folding tree. We also made average templates out of them for seeking general pattern. The major computer programs employed in our study are GRASP, MATHEMATICA and MIDAS. We obtained the folding tree by searching the DALI domain database.
We didnt see any specific recognition motif in terms of particular residue/ligand interactions. No residue is conserved among all the 34 non-homology structures studied. However, there is some common general pattern shared by them:
Some graphs have been made to display our results visually (Figure-5 and Figure-6). Due to the limitation of the space on this poster, we will take three types of proteins as the representative of all 34 distinct ones. They are Ras, RNAse T1, and DMSO reductase, all with quite different functions and foldings from each other. We detected certain similarity in ligand template and evolutionary relationship by comparing the folding tree (Figure-7) and our binding-surface trees (Figure-8). Two big families, agreed by both trees, turn out to be one made up of RNAase, DNA polymerase, Hgprtase and the other made up of Adp-ribosylation factor 1, Ras, Rap, Transducin, Gi Alpha1, Elongation factor.
The high density of basic
nitrogen and acidic oxygen is obviously related to their capacity
for hydrogen bonding. Since neutral ones fail to show such
trend, we may suggest that hydrogen bonding is stronger between
charged species than uncharged ones as the result of greater dipole
moment. The shape of aromatic residues determines its preference
for flat surface in order to achieve maximum interaction.
We dont have good explanation why aspartic acid is favored
over glutamic acid around a basic edge yet. Maybe this has
something to do with the lower entropic cost of holding aspartic
acid still. Its side chain is one-carbon shorter and thus
results in less overall entropy of the residue.
Top
Reference
1. Grasp (Graphical Representation
and Analysis of Structural Properties): Anthony Nicholls
2. Midas (Molecular Display
and Simulation System): UCSF
3. Mathematica: Stephen Wolfram
Figures
| 1 Ribonuclease T1 | 18 DNA Polymerase (1) |
| 2 Ribonuclease Sa | 19 DNA Polymerase (2) |
| 3 Ribonuclease Ms | 20 mRNA Capping Enzyme (1) |
| 4 Ribonuclease F1 | 21 mRNA Capping Enzyme (2) |
| 5 Ribonuclease A | 22 Human Adp-Ribosylation Factor 1 |
| 6 Ras | 23 Rat Adp-Ribosylation Factor 1 |
| 7 Rap2a | 24 Gi Alpha 1 |
| 8 Deoxynucleoside Monophosphate Kinase (1) | 25 Transducin-Alpha |
| 9 Deoxynucleoside Monophosphate Kinase (2) | 26 Elongation Factor Tu (1) |
| 10 Nucleoside Diphosphate Kinase | 27 Elongation Factor Tu (2) |
| 11 Guanylate Kinase | 28 Elongation Factor Tu (3) |
| 12 Adenylosuccinate Synthetase | 29 Elongation Factor G |
| 13 Thymidylate Synthase | 30 Purine Nucleoside Phosphorylase |
| 14 DMSO Reductase (1) | 31 Formate Dehydrogenase |
| 15 DMSO Reductase (2) | 32 Ftsz |
| 16 Histidine Triad Nucleotide-Binding Protein | 33 Jel 103 |
| 17 Glutamine Phosphoribosylpyrophosphate Amidotransferase | 34 Hypoxanthine Guanine Phosphoribosyltransferase |
![]() |
|
|
|
![]() |
| Figure-6: The density maps of acidic oxygen. The column height is proportional to the acidic oxygen density at that vertex. |
![]() |
![]() |
|
|