Towards novel modular polyketide synthases – lessons learnt from evolution

Modular polyketide synthases (PKS) have attracted a lot of attention as a target for creating novel pharmaceutical products. Although there has been success in manipulating decorating enzymes, most attempts to modify the polyketide backbone have resulted in very low product yields making them of no commercial interest. As there is a vast diversity of natural polyketides, it is interesting to examine how modular polyketides evolve in nature. Most attempts to modify PKS in vitro have exchanged single domains. Examination of the ratio of synonomous/non-synonomous codon changes between different PKS modules shows that there is very strong purifying selection. This would explain why the non-natural junctions created by in vitro approaches are so deleterious and suggest a strategy of preserving natural junction sequences, which could be implemented by appropriate bioinformatic methods. The study of the evolution of modular PKS has been dominated by the observation that the ketosynthase domains (KS) form cluster-specific groupings in phylogenetic trees whereas the acyl transferase (AT) domains are grouped according to substrate incorporated. This discordance has sometimes been interpreted as evidence of extensive horizontal gene transfer. Our studies showed that this phenomenon could be more easily explained by a high frequency of gene conversion within clusters. Once this was understood, it was possible to reexamine the mechanisms of modular PKS evolution. We suggest that a major mechanism utilises homologous recombination systems for single cross overs between clusters. This would generate radically new polyketide backbones. Modelling of homologous recombination in silico provides a novel strategy for generating new polyketide backbones with natural junctions, which are likely to synthesise products in reasonable yields. A software package CompGen was developed to model homologous recombination and predict the chemical structures of novel compounds produced by recombinants. Examples of predicted recombinants and their products are presented in the r-CSDB database. Methods are available to allow the efficient production of recombinants in the laboratory.