A number of experimental methods, such as phage display, are available for detecting and improving protein affinity for a substrate. The immune system is an example of optimized binding with antibodies that boast picomolar affinities for their cognate ligands. However, laboratory evolution and affinity maturation generally require detectable initial binding activity. A complementary role for computational design would involve constructing the binding sites de novo, which could then be enhanced by experimental methods (Chapter 17).
Designing substrate binding sites can be approached in two different ways. First, the surface amino acids of a protein can be re-engineered to facilitate the binding of a novel molecule. This requires identifying optimal sites on an existing scaffold and then choosing the appropriate amino acids to optimize binding. Alternatively, the protein fold itself can be designed from scratch to accommodate to the structural and energetic requirements for substrate binding. In such an approach, secondary structural elements such as a-helices and P-hairpins can be thought of as "Lego-blocks" that are used to build the appropriate fold. In this section, we discuss examples of protein, in which novel binding sites are constructed on existing protein scaffolds.
Some of the earliest successful designs involved introducing metal binding. Iron-sulfur (Fe-S) clusters are found in a wide variety of proteins and participate in electron transport and redox chemistry. To test whether an iron binding site can be engineered, the backbone of thioredoxin was scanned using DEZYMER (Hellinga et al. 1991; Hellinga and Richards 1991) for geometrically compatible positions for the four cysteine ligands to bind an iron (Figure 16.2) (Benson et al. 1998). The minimalist computational model optimized the bond lengths, angles, and steric compatibility with the surrounding protein matrix to construct a tetrahedral metal binding site. No backbone flexibility was included in the model. The final design used two Xxx ^Cys mutations and two natural cysteines to successfully bind iron with spectro-scopic properties similar to that of rubredoxin, a natural protein with similar iron coordination geometry. A number of other biologically important metal sites were similarly designed into thioredoxin including the iron center of superoxide dismutase (SOD) (Pinto et al. 1997), a cuboidal iron-sulfur center, a Cys2His2Zn center (Wisz et
al. 1998), and a FeHis3O2 site (Benson et al. 2000). The effect of the protein microenvironment on binding and redox properties was tested by engineering the same metal substrate in different locations of thioredoxin. Moving the FeHis3O2 site to grooves, shallow pockets, surface-exposed positions, and buried core sites had significant effects on its metal binding and redox properties. The role of second shell interactions was studied in the SOD designs, where positively charged side chain-stabilized O2- binding accelerated the rate limiting Fe-O2- forming step. A similar approach was used in the program METAL_SEARCH to introduce a zinc binding site into protein G|1 (Clarke and Yuan 1995; Klemba et al. 1995; Regan and Clarke 1990).
Computational tools that were initially developed on metal binding sites have since been extended to include other more complex substrates. Some examples include protein-protein interactions such as enhancing the affinity of the a2| 1 integrin for collagen (Shimaoka et al. 2000), increasing the specificity of calmodulin for myo-sin light chain kinase (REF) (Shifman and Mayo 2002; Shifman and Mayo 2003) or programming of the target sequence specificity of PDZ domains (Reina et al.
2002). A cleft between the two domains of a bacterial periplasmic binding protein (PBP) was computationally redesigned to bind a series of small molecules including l-lactate, serotonin, and trinitrotoluene (TNT) (de Lorimier et al. 2002; Looger et al.
2003). The diversity of substrates was made possible by the ideal cleft geometry of PBP, which presented multiple positions where binding affinity and specificity could be introduced. Mutating 12 to 18 positions and considering substrate rotational and translational degrees of freedom requires sampling an immense number of conformations (on the order of 1070-1080). In order to make this process feasible, the substrate configurations were first reduced to those compatible with a pocket consisting of alanines. These configurations were then optimized for van der Waals packing, hydrogen bonding, and other molecular forces between the substrate and protein. Sequences and side-chain rotamers were optimized using dead-end elimination (DEE). The algorithm reduces the number of permutations of n-interacting residues by eliminating the conformations not compatible with the global minimum based on pairwise comparison (Desmet et al. 1992; Looger and Hellinga 2001). During optimization, protein-substrate hydrogen bonding interactions were weighted strongly to address the energetic penalty of desolvating potential hydrogen bonding donors and acceptors upon binding. The PBP scaffold was functionalized with a fluorescent dye near the hinge between the two domains, allowing facile detection of binding by induced conformational change.
One of the challenges in predicting the effects of a mutation is ensuring that backbone flexibility is sufficiently accounted for. Backbone flexibility is necessary to describe the induced fit mechanism of substrate binding. As with the previous example using PBP, it is beneficial to first model flexibility at a coarsegrained level. For example in ROSETTA DOCK (Wang et al. 2007), which predicts protein-protein interactions, the sampling is carried out in three steps: (1) random perturbations of torsion angles in the backbone and rigid body motion between binding partners; (2) coarse-grained optimization of the perturbed structure with repacking of side-chain rotamers; and (3) energy minimization and local refinement of structures near the ligand binding site using off-rotamer side-chain sampling, small variations of backbone torsion angles, and rigid body sampling between proteins. Introducing sequence perturbations in step 1 could be useful in designing small molecule binding sites (Meiler and Baker 2006). As discussed later, this strategy was implemented in the design of enzyme active sites.
A number of challenges remain in the computational design of substrate binding sites in proteins. Binding sites are typically found on the surfaces of proteins, albeit in partially occluded pockets. Modeling a microenvironment where both crucial hydro-phobic and polar interactions are involved is difficult. The role of water molecules in binding and structure is also a difficult problem (Reichmann et al. 2008). Continuum solvation versus explicit water molecules have their respective advantages and disadvantages when it comes to modeling. One approach is to include water as part of side chains in a rotamer library (Figure 16.3) (Jiang et al. 2005). These solvated rotamer libraries include water molecules that are permanently hydrogen bonded to donor and acceptor atoms of amino acids. While this reduces the sampling required
Was this article helpful?
All Natural Immune Boosters Proven To Fight Infection, Disease And More. Discover A Natural, Safe Effective Way To Boost Your Immune System Using Ingredients From Your Kitchen Cupboard. The only common sense, no holds barred guide to hit the market today no gimmicks, no pills, just old fashioned common sense remedies to cure colds, influenza, viral infections and more.