MM/PBSA Analysis of Molecular Dynamics Simulations of Bovine -Lactoglobulin: Free Energy Gradients in Conformational Transitions? Federico Fogolari, 1 * Elisabetta Moroni, 2 Marcin Wojciechowski, 3 Maciej Baginski, 3 Laura Ragona, 4 and Henriette Molinari 1,4 1 Dipartimento Scientifico e Tecnologico, Universita ` di Verona, Verona, Italy 2 Dipartimento di Fisica G. Occhialini, Universita ` di Milano-Bicocca, and INFN, Sezione di Milano, Milano, Italy 3 Department of Pharmaceutical Technology and Biochemistry, Gdansk University of Technology, Gdansk, Poland 4 Laboratorio NMR, Istituto per lo Studio delle Macromolecole, CNR, Milano, Italy ABSTRACT The pH-driven opening and clo- sure of -lactoglobulin EF loop, acting as a lid and closing the internal cavity of the protein, has been studied by molecular dynamics (MD) simulations and free energy calculations based on molecular mechanics/Poisson–Boltzmann (PB) solvent-acces- sible surface area (MM/PBSA) methodology. The forms above and below the transition pH differ presumably only in the protonation state of residue Glu89. MM/PBSA calculations are able to reproduce qualitatively the thermodynamics of the transition. The analysis of MD simulations using a combination of MM/PBSA methodology and the colony energy approach is able to highlight the driving forces implied in the transition. The analysis suggests that global rearrangements take place before the equilib- rium local conformation is reached. This conclusion may bear general relevance to conformational tran- sitions in all lipocalins and proteins in general. Proteins 2005;59:91–103. © 2005 Wiley-Liss, Inc. Key words: -lactoglobulin; loop dynamics; MM/ PBSA; Poisson-Boltzmann; colony en- ergy INTRODUCTION Mobile loops often correspond to protein sites implicated in functions. The characterization of conformational tran- sitions involved in events such as ligand binding, product release, and allosteric changes is therefore of primary importance. Along this line, we have shown for the first time that bovine -lactoglobulin (BLG), an 18-KDa -bar- rel lipocalin composed of 9 antiparallel -strands defining an internal cavity, modulates its binding properties through the reversible opening– closure of one loop (EF loop, resi- dues 85–90) acting as a lid that closes the protein interior binding site below pH 7 and opens it at higher pH. 1 EF loop conformational transition, first detected by Tanford and Nozaki 2 using optical rotatory dispersion, is triggered by the titration at unusually high pH (7) of the buried carboxyl group of E89, located within the loop at the protein open end. Molecular details of this conformational change have recently been revealed from X-ray structures by using crystals grown at pH values on either side of the transition, 3 sometimes referred to as the Tanford transi- tion, and explained by electrostatic calculations. 4 The titration of E89 side-chain at high pH (7) involves a remarkable rearrangement of the H-bond pattern involv- ing L87, E89, N90, E108, N109, S110, and S116 residues, and causes the fold back of the EF loop, with a consequent solvent exposure of glutamic acid side-chain (Fig. 1), changing its solvent accessible area from 4 Å 2 at pH 6.2 to 116 Å 2 at pH 7.1, and 136 Å 2 at pH 8.2. 5 Our studies 1 on the role of this conformational rearrange- ment on the functional properties of BLG revealed that this feature is common to all BLGs. Indeed, E89 is conserved in all BLGs from different species, and we have recently shown that its titration acid dissociation constant (pKa) is influenced by the nature and charge of nearby residues, thus regulating the binding pH. A few observa- tions reported for other lipocalins, such as tear lipocalin, 6 bovine retinol-binding protein, 7 nitrophorin, 8 and mem- brane enzyme PagP, 9 suggest that conformational rear- rangements, mainly involving loops at the protein open end, could be a general feature of the lipocalin ligand- binding mechanism, somehow reminiscent of lipases’ inter- facial activation. Given the important functional role of this conformational transition in lipocalins and, more generally, of conformational transitions in proteins, it is of primary importance to identify a reliable approach for the computation of free energy differences between several conformations of the same biomolecule. Indeed disentan- gling free energy contributions arising from different interactions allows for better understanding and biomolec- ular rational design of ligands. Implicit solvent methods 10,11 that compute conforma- tional free energies offer several advantages over other simulation methods. In particular, (1) averaging over Grant sponsor: NATO (Collaborative Linkage Grant to F. Fogolari, M. Wojciechowski, and M. Baginski). Grant sponsor: MIUR (Italy) ex 40% 2002. Grant sponsor: FIRB 2001. *Correspondence to: Federico Fogolari, Dipartimento di Scienze e Tecnologie Biomediche, Universita ´ di Udine, P.le Kolbe 4, 33100 Udine, Italy. E-mail: fogolari@dstb.uniud.it; http://www.dstb.uniud. it/fogolari Received 7 May 2004; Accepted 3 September 2004 Published online 2 February 2005 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/prot.20384 PROTEINS: Structure, Function, and Bioinformatics 59:91–103 (2005) © 2005 WILEY-LISS, INC.