Engineering, evolution and folding of enzymes with natural and non natural activities

Resum

Ancestral proteins refer to proteins from extinct organisms that no longer exit on Earth. However, phylogenetic and bioinformatic analyses of modern protein sequences can lead to plausible approximations to the sequences of their ancestors. The methodology is called ancestral sequence reconstruction and allows for further preparation of the ancestral proteins in the laboratory (resurrection) and their characterization. It has been shown that resurrected ancestral proteins are very powerful tools to address important issues in evolution, and also to provide solutions to practical problems in biotechnological/biomedical scenarios due to their remarkable properties, quite different, in many cases, from those of their modern counterparts. In this doctoral thesis, both aspects have been investigated as summarized below. First, we have explored the relationship between in vitro and in vivo protein folding, in an evolutionary context. We have experimentally characterized the in vitro folding of a set of resurrected Precambrian and modern thioredoxins and found that, contrary to previous claims in the literature, the folding rates are not evolutionarily conserved. Thus, ancestral thioredoxins fold much faster in the test tube than their modern counterparts. Extensive mutational analyses have allowed us to identify mutation Ser74Gly as responsible for aggravating folding in E. coli thioredoxin. The evolutionary acceptance of this mutation is interpreted as an example of degradation of ancestral features at the molecular level. We propose that unassisted and efficient primordial folding was linked to fast folding encoded at the sequence/structure level. Once an efficient assistance machinery had emerged, mutations that impaired ancient sequence/structure determinants of folding efficiency could be accepted, since those determinants were no longer necessary. We conclude that in vitro and in vivo folding landscapes are disconnected and question the biologically relevance of the in vitro folding rate determinations, except as related to heterologous folding efficiency (see below). In the second part of this thesis, we have addressed a pivotal and common problem in biotechnology, the inefficient heterologous expression of proteins. As a model system thioredoxin from Candidatus Photodesmus katoptron, an uncultured symbiotic bacteria of flashlight fish, has been used. Our results demonstrate its slow in vitro folding (it takes several hours to reach the native state) and inefficient expression in E. coli, leading mostly to insoluble protein. By using a few back-to-the-ancestral mutations at positions selected by computational modelling of the unassisted folding landscape we were able to rescue its inefficient expression. Our results support that the folding of proteins in foreign hosts may be akin to some extent to unassisted folding due to the absence of coevolution of the recombinant protein with the natural chaperones of the new host. More generally, our results provide an approach based on sequence engineering to rescue inefficient heterologous expression with a minimal protein perturbation. Finally, the Appendix is dedicated to briefly summarize our results on the use of ancestral proteins as powerful tools for engineering de novo enzymes with non natural activities. We have resurrected and exhaustively characterized an ancestral TIM-barrel protein belonging to family 1 glycosidase, which displays remarkable properties such as thermostability and enhanced conformational flexibility. Surprisingly, the ancestral glycosidase binds heme tightly and stoichiometrically, and its binding enhances catalysis. This finding is unexpected because none of the reported crystallographic structures of the ∼1400 modern glycosidases shows a bound porphyrin. Moreover, these features reveals promising applications in custom catalysis and biosensor engineering.