Improving properties of llama heavy chain antibodies using in silico analysis

Abstract

Microbicides offer a promising way to prevent HIV infection in developing countries. Llama heavy chain antibody fragments (VHHs) that neutralize HIV are an ideal candidate for usage as active microbicide component. A VHH is the smallest intact antigen binding domain which allows it to reach for antigen surfaces that can not be reached by conventional antibody fragments and allows a more efficient tissue penetration. VHHs display a good performance at many different conditions. The usage of VHHs as microbicide component is dependent on many factors; high affinity and neutralization capacity, stability at different temperatures, functionality at a low pH, and efficient production in high quantities at low costs. The GRAS organism Saccharomyces cerevisiae is often used for the production of VHHs. However, the production yield in this organism often is too low for commercially viable large scale application. Understanding the causes of these low production yields and improvement of the yield is an important topic within this thesis. The characteristics of VHHs are studied in silico by typical bioinformatics and genomics approaches as well as in vitro and in vivo by mutational and cultivation studies to find correlations between protein sequence and stability or production level. We found correlations between low production yield and non-optimal codon usage, as well as framework four encoded by J-gene segment 7. To facilitate a more extensive study of all available VHH sequences a bioinformatics infrastructure was set up, resulting in the antibody variable domain database and the VHH 3DM, a molecular-class-specific information system that creates an accurate structure-based multiple sequence alignment. This sequence alignment revealed a set of eight key residues important for proper folding, which was validated in vitro and in vivo. These data were combined with known and predicted mutation effects resulting in a generic 'VHH-mutability' table, which flawlessly predicted the best producing variants of a set HIV neutralizing VHHs. The alignment, combined with structural data, also made a clear definition of the VHH CDR regions possible. A more detailed in silico and in vitro and in vivo studies of two similar VHHs demonstrated that the following optimisation possibilities; improved codon usage, framework four replacement, and yeast fermentation on ethanol, resulted in a higher production yield and secretion efficiency in yeast. Combined with electron microscopy and mammalian cell expression results some light was shed on the possible mechanisms causing the low production levels in yeast. With the availability of Lama glama and Lama pacos germline sequences the maturation pathway of VHHs was studied. For L. glama 23 V-genes were identified, as well as typical llama hotspot motifs. Evidence was found for gene conversion of D-gene segments. Knowledge of the maturation pathway will eventually assist the selection and design of even better VHHs against HIV. Combined with the in silico screening methods and demonstrated optimisation possibilities, exquisite opportunities are offered to select or design the optimal VHHs for the use in microbicides and many other applications.