Abstract

Ribozymes are catalytic RNA molecules. Hammerhead ribozymes are one of a set of ribozymes capable of cleaving RNA molecules, in cis and in trans, without the help of other molecules, such as proteins. A trans-acting hammerhead ribozyme has a few related forms, which can be customized to target specific sites within RNA strands, including the transcripts of genes. As such, hammerhead ribozymes can be used to down-regulate or even silence any gene with valid cut-sites. All, except very short transcripts, will have multiple valid cut-sites. However, the efficiency of silencing by a hammerhead ribozyme depends on multiple often conflicting factors. Also, the efficiency of cleavage of any one ribozyme is normally low. Hence, it is useful to automate the process of design of hammerhead ribozymes to efficiently and without error explore the large space of possible designs. Computers are simply better than humans in doing a large amount of repetitive work without error.
This thesis describes an original computational algorithm that allows for the automated design of hammerhead ribozymes; the algorithm was implemented as a web service, by our collaborators, and is now freely available via the Internet. The algorithm takes into account all the relevant mathematically-modeled factors, influencing cleavage efficiency & target specificity, including cut-site availability, cut-site accessibility, ribozyme secondary structure, annealing temperature and off-target effects. Given an input sequence or gene, the algorithm proposes a list of potential hammerhead ribozymes that can, in theory, cleave the given RNA sequence. The latter part of this thesis describes the wet lab work, which involved the in vitro and hence, in vivo testing of several ribozymes targeting the transcript of the PABPN1 gene. This gene is the primary cause of a human disease, OPMD (Oculopharyngeal muscular dystrophy).
iv
We describe the experimental methods used to test the cleavage efficacy and reaction kinetics of a small number of hammerhead ribozymes generated by the computational algorithm. We measured the in vitro transcript cleavage efficiency and enzyme kinetics, as well the in vivo gene knockdown effect of individual ribozymes. We also measured the enhanced effect of using combinations of two or more ribozymes all targeting the same transcript. Finally, we design a mutant PABPN1 gene that gives the same protein as the wild type gene, but one that generates a transcript supposedly immune to the catalytic activity of hammerhead ribozymes. We test this hypothesis, in vivo, and provide the results.
The results show that (a) every one of the ribozymes generated by the computational algorithm is a functional ribozyme; (b) the use of two or more ribozymes increases the overall cleavage efficiency; (c) these ribozymes are effective both in vitro and in vivo; (d) the immune transcript is not affected by any ribozyme. Hence, the computational algorithm is an effective design tool. The use of at least two hammerhead ribozymes is clearly more effective than one. And finally, it is possible to generate a hammerhead ribozyme (or more) that would cleave the transcript of a given gene (say PABPN1), while not cleaving the transcript of a mutated version of the gene, one that codes for the same protein as the original gene. This opens the door to potential therapeutic applications of our approach, perhaps where other gene editing approaches cannot be employed.