The measurement of binding affinities by NMR chemical shift perturbation

Abstract

We have carried out chemical shift perturbation titrations on three contrasting proteins. The resulting chemical shifts have been analysed to determine the best way to fit the data, and it is concluded that a simultaneous fitting of all raw shift data to a single dissociation constant is both the most accurate and the most precise method. It is shown that the optimal weighting of ¹⁵N chemical shifts to ¹H chemical shifts is protein dependent, but is around the consensus value of 0.14. We show that chemical shift changes of individual residues can be fit to give residue-specific affinities. Residues with affinities significantly stronger than average are found in close contact with the ligand and are suggested to form a rigid contact surface, but only when the binding involves little conformational change. This observation may be of value in analysing binding and conformational change.

Keywords: Affinity; Binding; Chemical shift; Conformational change; Dissociation constant; NMR.

Fig. 1

¹⁵N HSQC titration data for a SH3b b Barnase c HisJ. In each case, peaks are colored from red to violet with addition of ligand. Signals undergoing large shift changes are labelled

Fig. 2

The locations of large chemical shift changes on titration of a, b SH3b c, d barnase e, f HisJ with their ligands. In each case, the right panel is a rotation of 180° around a horizontal axis. Blue is used to denote negative (upfield) chemical shift changes in H or N, and red for positive (downfield) changes. For SH3b and barnase, the ligand is denoted by sticks. For HisJ the ligand is completely buried and is in the center of the protein. The shift changes shown include approximately 20% of the amino acids in the protein that have reliably fitted shift changes. Residues undergoing large shift changes are labelled

Fig. 3

Typical chemical shift titration data: HisJ, showing change in chemical shift with addition of the ligand lysine. The curves are individual best fits to the data. a Shift changes in ¹H. Data are shown for T77 (red: fitted to K_d of 62 ± 6 μM), D14 (blue: fitted to K_d of 49 ± 6 μM), and G129 (black, fitted to K_d of 54 ± 12 μM). b Shift changes in ¹⁵N. Data are shown for F194 (red: fitted to K_d of 56 ± 5 μM), G171 (blue: fitted to K_d of 60 ± 11 μM), and T198 (black: fitted to K_d of 58 ± 13 μM). All data are shown as positive shift changes for ease of presentation. The actual shift changes for T77, D14, F194 and T198 are negative

Fig. 4

K_d values fitted for HisJ binding to lysine, for a ¹H b ¹⁵N. Data are shown for all residues that could be fitted. A small number are not shown, mainly because of overlap or because they are prolines. Note that the K_d values are truncated at 0.2 mM: many of the truncated residues are much larger than this

Fig. 5

K_d values fitted for HisJ binding to lysine, after filtering out unreliable values, for a ¹H b ¹⁵N. Nuclei were removed if they had a fitted K_d of > 200 μM, total ¹H shift changes of < 0.03 ppm, or total ¹⁵N changes of < 0.15 ppm. It was not necessary to filter out very small K_d values because these were removed by the chemical shift limits

Fig. 6

Residues fitting to K_d values significantly different from the mean (defined as outside the 99% confidence intervals), for a SH3b b barnase, c HisJ. Residues with K_d values significantly smaller (stronger) than the mean are in orange, and residues significantly larger (weaker) are in green, and are labelled. In each case, the right panel is a rotation of 180° around a horizontal axis. For SH3b and barnase the ligand is shown as sticks; for HisJ the protein surface is drawn slightly transparent, allowing the bound histidine ligand to be seen in the center of the protein. For SH3b the ligand shown is GGGGG, using the crystal structure with PDB ID 5leo. The ligand used here is YGGGGG, where the additional tyrosine is at the N-terminal (top, in the view shown in a) end of the peptide

Fig. 7

A simple model for flexible protein/ligand binding. Both the protein and the ligand may have internal flexibility (a). When they bind, the resulting complex has lost this flexibility (d). The process of binding may thus be separated conceptually into a rigid docking of part of the ligand (a, b), with an affinity K₁, followed by an induced-fit type rearrangement accompanied by loss of flexibility (b–d). The equilibrium constant for this process is K₂, which has values between 0 and 1, but may typically be expected to be greater than 0.5. Alternatively, the binding can be modelled as a loss of internal degrees of freedom (a–c) followed by a rigid docking (c, d)