The specific recognition or binding of targets by affinity reagents is a critical step in any molecular detection scheme. Therefore, the ability to reliably and efficiently generate affinity reagents is of great importance for applied biotechnology as well as fundamental research in biology. The oldest approach to generating affinity reagents involves immunizing vertebrate animals such as mice or goats with the target (antigen) of interest. The immune system of the animal then generates antibodies that bind specifically to the target. The antibodies generated in this fashion are typically heterogeneous in nature and bind to different regions (epitopes) on the antigen, and are referred to as polyclonal antibodies (PAbs). While the use of PAbs as affinity reagents is very common, variability of the immune response in animals and the heterogeneity of PAbs can cause significant lot to-lot variability in PAb reagents. Also, generation of PAbs is an expensive process. The problem of variability in PAbs can be overcome by the use of hybridoma technology to create mammalian cell lines that produce monoclonal antibodies (MAbs), i.e. defined antibodies binding to a single epitope on the antigen. However generation of cell lines is tedious, time consuming and expensive. The recombinant production of antibodies in bacterial systems has been reported. However, this has proved to be difficult in general due to the large multi subunit disulfide-bonded structure of antibodies, which also leads to low stability. Thus both MAbs and PAbs suffer from low stability and high cost of production.
To overcome the limitations of antibodies, several alternate protein architectures have been used as scaffolds to engineer molecular recognition with varying degrees of success. Analogous to complementarity determining regions (CDRs ) in antibodies, the amino acid composition in certain areas of the scaffold protein is varied to create protein variants that bind to various targets. Examples include mutating residues on a secondary structure element such as alpha helices or beta sheets, grafting an amino acid loop onto the scaffold protein and changing the amino acid composition of one or more flexible loops in the flexible protein. However, in spite of all the limitations of antibodies and in spite of the availability of several alternate scaffolds, antibodies continue to be the most widely used affinity reagents. In addition to historical reasons, this is largely because antibodies satisfy the most important criterion for a successful scaffold – the ability to generate high affinity binding molecules specific to the vast universe of possible targets. It is conceivable that mutagenesis of a single alternate scaffold may not provide the required diversity in topologies to create high affinity binders to all possible targets.
As an alternative to antibodies, we have used a pool of small proteins from hyperthermophilic archaea and bacteria as scaffolds to create affinity reagents. Based on structural data for these proteins, we randomized 10-15 surface accessible residues on each scaffold to create a library of mutant proteins. Taken together, the individual component libraries represent a combinatorial “super library” of scaffolds. We hypothesized that such a super library, where multiple different topologies on multiple scaffolds are randomized, is more likely to yield binders to any given target than a library from any single scaffold. Further, the scaffolds chosen have several desirable properties. Due to their hyperthermophilic origin, these proteins have very high thermal stability. They have low molecular weight (~ 100 amino acid in length or less), no disulfide bonds and can be inexpensively produced at high yield in E. coli. Here we present our results comparing this super-library with a single-scaffold library having much higher sequence diversity. A comparison of yeast surface display and mRNA display methods for library screening, in the context of this problem, is also presented.