Introduction An important goal of structural biology is to predict and manipulate the interactions between proteins and small-molecule ligands. To this end, extensive research efforts have been aimed at relating the known structure of protein–ligand complexes to the thermodynamics of that interaction. The success of these attempts has been limited, in part because of their neglect of the role of protein and ligand conformational dynamics in determining ligand-binding thermodynamics. 1,2 d d 3,4 5 4,6–9 Results and Discussion Entropic cost of binding et al 10,11 12–15 11 9 1ABE 1ABF 5APB Figure 1 16 T S 6 17–19 17,18 19 Assignment of apoABP d 20 1 15 Figure 1 21,22 Small chemical shift differences are seen at sites distal to the binding site and hinge region. These differences suggest that subtle changes in conformation or dynamics throughout the molecule occur on binding. Domain orientation of apoABP 23 1 15 R R Figure 2 Dynamics of ABP 24,25 26 Figure 3 Figure 3 To confirm and further explore the basis of this result, we have performed molecular dynamics simulations for ABP in the apo state and in complex with galactose. By several measures, we see significant increases in backbone dynamics in the complex as compared with the apo protein. RMS deviations from the average structure for both the N and C domains are significantly larger for the ABP–galactose complex than for apoABP. In addition, fluctuations of backbone heavy-atom positions across the trajectory are generally larger in the complex than in the apo protein (data not shown). Furthermore, these dynamic changes are seen to be more pronounced in the N domain than the C domain, consistent with the experimental observations. 27 Figure 3 27–30 Figure 3 S 2 apoABP S 2 ABPgal Figure 4 24 Table 1 25 S 2 s s S 2 f 31,32 S 2 LZ S 2 f 2 s S 2 s Table 2 in vitro Table 2 et al. 29 S 2 LZ S conf 26 S 2 LZ S 2 LZ Figure 5 T S conf T S conf The experimental data available for this system is limited to probes of backbone dynamics. Given the good agreement between experiment and the molecular dynamics simulations, it is perhaps reasonable to infer something of the side-chain dynamics from these simulations. We observe considerably more variability in side-chain order parameters than is evident for the backbone, consistent with findings in other proteins. Furthermore, there is much greater variability in the change in order parameter observed upon ligand binding. Despite this, the changes in side-chain dynamics are broadly similar to those seen for the backbone, with the majority of residues showing a small increase in flexibility on ligand binding. As fast side-chain dynamics are correlated with local backbone dynamics only weakly, this suggests an additional source of favourable entropy change accompanying binding. A number of residues around the binding site show larger changes in dynamics on binding, reflective of both increases and decreases in flexibility. These changes reflect similar heterogeneous dynamic changes seen in the protein backbone of the loops that comprise the binding site. 30 et al m 33 m 34–36 37,38 39 40 41,42 34–38 Materials and Methods Protein expression and purification 6,20 Escherichia coli 15 15 4 2 13 15 2 2 13 2 15 4 6,20 NMR spectroscopy 2 2 d 1 z 1 15 13 43 1 15 1 J NH et al 44 45 15 1 2 1 15 et al 46 R 1 R 2 R 2 R 1 R 2 Spectral assignment 20 1 15 1 15 1 15 i i Domain orientation of apoABP 21,22 1 15 47 5ABP 9 48 Lipari–Szabo analysis of relaxation data 24,25 49,50 15 51 52 49 Table 1 Table 2 33 m m 49 R ex R 2 m R ex R 2 26 S 2 LZ Molecular dynamics 57 − 8 2 The production period took 20 ns for both systems. The coordinates were saved every 2 ps of MD simulation. 27 (1) S L Z 2 = 3 / 2 [ < x 2 > 2 + < y 2 > 2 + < z 2 > 2 + 2 < x y > 2 + 2 < x z > 2 + 2 < y z > 2 ] − 1 / 2 x y z et al. 29 S 2 LZ (1) t t