Specificityty of L-arabinose-binding protein: Computer modelling study

Abstract

The natural affinity and stereoselectivity of carbohydrate binding proteins and enzymes toward their substrates have long been of great interest. Recently, it has been found that such properties are strongly manifested by the periplasmic sugar binding proteins of Gram negative bacteria. These proteins serve as specific receptors in bacterial active transport and chemotaxis. Their function and substrate binding specificity can be understood by studying the modes of binding of substrates and various inhibitors. L Arabinose binding protein (ABP), the first periplasmic sugar binding protein for which both the primary and tertiary structures have been established, has been chosen as a model system for this purpose. A molecular approach to studying such biomolecular interactions requires detailed knowledge of the three dimensional structures of the interacting molecules and their complexes. X ray diffraction is the most powerful technique for determining such structures. The structure of ABP has been solved at different resolutions by Quiocho and co workers. However, these studies are limited to only one complex of ABP with L arabinose. Notably, the predicted binding mode of L arabinose differed between the low and high resolution structures. It is now well established that unless macromolecular structures are determined at high resolution (<2 Å), the information gained-especially regarding molecular details of protein–ligand interactions-may not be completely reliable. Since most carbohydrate binding proteins have been solved only in the 2–3 Å resolution range, newer methods are required to take full advantage of available data. Moreover, to understand protein specificity and the relative binding affinities of inhibitors, the structure of a single protein–ligand complex is insufficient. Solving many complexes by X ray diffraction is difficult; however, computer modelling can generate realistic three dimensional structures of various protein–ligand complexes using data from a single high quality structure. The available X ray data of ABP (PDB entry 1ABP) correspond to an unrefined 2.4 Å study. Therefore, the present work attempts to generate three dimensional models of ABP–ligand complexes using this low resolution structure via a computer modelling method. These studies help to understand molecular details of ABP–ligand interactions and provide theoretical explanations for the relative binding affinities of different sugars. This thesis is presented in five chapters:

Chapter I – Introduction A concise introduction to periplasmic binding proteins, followed by a review of:

structure and specificity of ABP, physicochemical studies aimed at elucidating ABP specificity.

Chapter II – Inhibitor Nomenclature and Modelling Method This chapter describes:

the nomenclature of sugar inhibitors, stereochemical criteria used to select possible binding modes, potential energy functions and parameters used, the computer modelling method developed to generate 3D protein–ligand complexes.

Chapter III – Binding Modes of L Arabinose The binding modes of and anomers of L arabinose were investigated:

Starting from the unrefined 2.4 Å data, both anomers could be fitted unambiguously into the ABP binding pocket. The predicted binding modes agree with the 1.7 Å refined structure of Quiocho & Vyas (Nature, 310, 318 (1984)). The predicted hydrogen bonding scheme also shows good agreement with the high resolution study.

This validates the modelling method for predicting molecular details of protein–substrate interactions.

Chapter IV – Binding of D Galactose, D Fucose, and D Glucose The binding of and anomers of these sugars was studied:

These sugars preferentially bind in the form to ABP, unlike L arabinose which binds equally well in both anomeric forms. Binding modes differ slightly in hydrogen bond patterns. Residues Arg151 and Asn232 form bidentate hydrogen bonds with L arabinose and D galactose, but not with D fucose or D glucose. Calculated conformational energies and available kinetic data both indicate: D Gal > D Fuc > D Glu (binding affinity). Weak inhibitor D glucose binds to one domain only, resulting in a weaker complex.

Chapter V – Binding of Other Pentoses and Hexoses Sugars studied include:

L ribose, L lyxose, D xylose, L fructopyranose, L altrose.

Modelling results show:

Changes in orientation of C 2 and C 3 hydroxyl groups from equatorial to axial significantly reduce binding affinity (e.g., L ribose, L lyxose). An equatorial C 4 OH also reduces affinity, especially when a bulky C 5 substituent (CH OH) is present. Sugars with bulky C 1 substituents (L fructopyranose) or axial C 5 substituents (L altrose) are predicted not to bind ABP. Strong binding correlates with the ability to form bidentate hydrogen bonds with Arg151 and Asn232. Moderate or weak inhibitors seldom form such interactions.

Conclusion The present studies provide:

Preferred binding modes of various sugars to ABP. The role of configuration and substitution in determining binding affinity. A theoretical basis for the relative affinities of different sugars.

These results also demonstrate that computer modelling can effectively predict molecular details of protein–ligand interactions even when only low resolution X ray data are available.