Our current model system OppA

OppA is a periplasmic binding protein involved in the uptake of extracellular peptides by Gram-negative bacteria such as E. coli. The structure of the protein was determined at York some years ago [5], revealing that in common with other periplasmic binding proteins, OppA engulfs its ligand between two domains joined by a flexible hinge (Figure 4). It binds peptides between two and five residues in length with little regard to sequence. Tight binding (K ~1 ^M) is achieved by the protein utilising the hydrogen bonding potential of the peptide backbone and strategically placed charged groups to bind the N and C termini. The ligand side-chains are accommodated in large hydrated cavities in the protein. The initial studies on a series of peptides of sequence KXK showed that the structure of the protein remains essentially unchanged on binding different ligands [7].

This system satisfies all the criteria discussed. We can produce large quantities of pure protein that is highly soluble and can be recycled by dialysis after ligand-binding experiments have been performed. Peptide ligands can be ordered directly from suppliers and there is a large repertoire of natural and unnatural amino acids which can be incorporated into the ligand. The ligands are, in general, highly soluble in aqueous solution and although the enthalpy change for some ligands is rather low, it has always been possible to measure the thermodynamics ofbinding. Crystallisation and determination of the structure of the protein-ligand complex is also relatively routine, although because a conformational change is involved, co-crystallisation is necessary rather than soaking crystals with ligand.

Figure 5. Structure of the peptide ligands KXK when bound to OppA, together with selected side chains from the protein (E32, W397, R404, H405). The structures are deposited with pdb codes 1b05, 1b0h, lblh, 1b2h, 1b3h, 1b32, 1b3f, 1b3g, 1b3h, 1b31, 1b40, 1b46, 1b4h, 1b4z, 1b51, 1b52, 1b58, 1b5h, 1b5i, 1b5j, 1b6h, 1b7h, 1b9j, 1jet, 1jeu, 1jev, 1olb.

Figure 5. Structure of the peptide ligands KXK when bound to OppA, together with selected side chains from the protein (E32, W397, R404, H405). The structures are deposited with pdb codes 1b05, 1b0h, lblh, 1b2h, 1b3h, 1b32, 1b3f, 1b3g, 1b3h, 1b31, 1b40, 1b46, 1b4h, 1b4z, 1b51, 1b52, 1b58, 1b5h, 1b5i, 1b5j, 1b6h, 1b7h, 1b9j, 1jet, 1jeu, 1jev, 1olb.

Two main series of ligands have been studied to date. The first is the natural KXK series where X is one of the 20 naturally occurring amino acids. The crystal structures and calorimetry data have been obtained for the full series [8]. The second series is also of the type KXK, but with X chosen from commercially available abnormal amino acids [6]. This series allows minimal changes to be made to the ligand. For example, comparing the ligands where X = diaminopropionic acid, diaminobutyric acid, ornithine and lysine, the effects can be seen of moving the side-chain amino group along the binding pocket in steps of one carbon atom. Similarly the results with norleucine and norvaline can be compared with alanine. In this way a much more precise picture is obtained than by comparing rather different groups with each other. One interesting feature of OppA is that the pockets can enclose bulkier ligands than the protein would expect to handle normally. These are not physiological ligands, as a key function of the protein is to select only natural peptides for transport into the cell.

Figure 5 shows the structure ofthe ligands and selected OppA side-chains for all the published (and deposited) structures determined to date in the series

KXK. Although there is some variation in the detailed conformation of the flanking lysine side-chains across the series, the peptide backbone of the lig-and remains essentially constant through all the structures. Importantly, there are only minor adjustments in the conformation of one side-chain (Glu 32) in the OppA protein structure. The figure emphasizes how the major change is in the chemical nature and bulk of the central amino acid side-chain. What is not shown (see References 6 and 8 for details) are the changes observed in the solvent structure in the ligand binding cavities.

The program LUDI [9] is the best-documented and widely used empirical scoring potential for estimating binding affinity for protein-ligand complexes. In common with other scoring functions, the affinity of binding is estimated from contact areas and hydrogen bonding, no explicit account being made for dynamics. We computed LUDI binding estimates for the interaction between different peptides and OppA. The program gave binding affinities that were far larger (as much as 106) than measured, and this error was not systematic - there was no correlation between the computed and experimental values. Through our links with the software company MSI, we were able to inspect the source code for this program, and established that the program was sensibly taking the solvent to be part of the protein.

Detailed analysis of this structural and thermodynamic data is currently underway, including thermodynamic perturbation to compute AAG values, solvation free energy calculations, and attempts at deriving a new empirical scoring function. The results of these calculations will be presented elsewhere.

There are some general features of the structures, however, which are worthy of note. Figure 6 shows the observed trend in the binding affinity for the KXK series with Dab, Dap, Om, Lys as the central side-chain. This positions an amino group at different positions in the binding pocket. Although there are clearly some energetic differences associated with the changing number of solvent molecules in the binding pocket, an important drop in K is noticed for ornithine. Inspection of the structures shows that in this amino acid, the amino group is placed alongside a salt bridge between a glutamic acid and arginine in the protein. This emphasises (as with the penicillin acylase example discussed above) that the observed binding affinity is not only a consequence of the interactions between the ligand and the protein, but also includes a term for any changes in the internal energy of the protein itself.

Another interesting observation (discussed in more detail for the natural KXK series [8]) is how the protein structure has evolved to provide a balance between entropy and enthalpy such that the overall binding affinity of the lig-and for the protein remains approximately constant. This is a recognised phe-

Figure 6. Moving an amino group through a ligand binding cavity. On the left hand side is a detail from the structure of KdapK (pdb code: 1b6h) with solvent molecules shown. On the right hand side is the measured K for the series KXK where X is Dab, Dap, Orn and Lys [6].
Figure 7. Enthalpy-entropy compensation in ligand binding. Comparison of the structures of KNvaK (red, pdb code: 1b6h) and KNleK (blue, pdb code: 1b7h) together with the measured thermodynamic values.

Table 2. Chemical structures of the central side chain in the peptides KXK. K^ is the dissociation constant for the tripeptide binding to OppA as determined calorimetrically [6]

Table 2. Chemical structures of the central side chain in the peptides KXK. K^ is the dissociation constant for the tripeptide binding to OppA as determined calorimetrically [6]

nomenon [10,11], which occurs universally when weak non-covalent forces are involved. The often-quoted explanation is that a strong, enthalpically favourable interaction results in a general tightening up of the complex, thus reducing the vibrational freedom and the entropy of the system. A striking example of this is available from our results when comparing norvaline and norleucine (Figure 7). Our dataset provides an opportunity to explore this phenomenon in more detail.

0 0

Post a comment