Why would anyone wish to design a live, attenuated vaccine for COVID-19? They cannot be administered in pregnant women or immunosuppressed subjects, but they do have one major advantage. They are the only form of vaccine that can be given intranasally. And given through this route, they stimulate the production of secretory IgA antibodies, which are the only isotype which protect the upper respiratory tract against COVID-19. (The lower respiratory tract is protected by circulating IgG).
Unlike other types of vaccines therefore, live attenuated vaccines would be expected to stop not just illness from COVID-19 in the vaccinated subject, but also transmission of the virus to contacts.
But how is the live virus attenuated? There are various means of doing so- by growing it at a lower temperature, or in a non-human cell line, but the one method used for 3 vaccine candidates using live attenuated virus during the current pandemic use a technique called codon de-optimisation.
The genetic code is redundant. That is to say, a given amino acid can be encoded by more than one codon. Yet, amongst these multiple codons, there is one that is favoured above all others- a phenomenon called codon bias.
Several live vaccines have attenuated the causative virus by reverse genetics (targeting the putative DNA triplet bases which codes for a certain amino acid in the peptide chain), through replacing the normally favoured codon in the viral DNA or RNA with a less favoured codon. In some cases, this goes a bit further and replaces a favoured codon pair with a less favoured codon pair. This is called codon de-optimisation.
While in theory, the replaced codon codes for exactly the same amino acid, in practice, this disrupts the tertiary structure of the peptide chain and leads to a dysfunctional viral protein being translated in the host cell. It is thought that the tRNA carrying the "non-favoured" anti-codon somehow interferes with the translation machinery.
Interestingly, when non favoured codon pairs are introduced during codon de-optimisation, it invariably introduces more CpG dinucleotides (nnCpGnn). The latter are under-represented in favoured codon pairs. It has however been shown that this excess of CpG nucleotides is not mechanistically responsible for the disruption of translation, which is thought to be induced by the unfavourable "fit" caused by the tRNA carrying the "unfavoured" anticodon, as described above.
The corollary to this is that codon optimisation (i.e. using the favoured codon) can improve the yield of useful proteins produced for medical usage in E.coli by phages.
A related concept is stabilisation of a recombinant viral protein vaccine by introducing two proline residues around a beta turn in the peptide molecule. A beta turn is a portion of the peptide chain when there is a sudden change in direction , say from an alpha helix to a beta pleated sheet, or between two alpha-helices. The artificially introduced proline residues at the beta turn stabilises the whole protein molecule and prevents misfolding.
This technique has been used in COVID-19 by Novavax for their recombinant Spike protein vaccine. It has also been used for the mRNA vaccines produced by Moderna and Pfizer-Biontech. The mRNA molecule in these vaccines is destined to be translated into the full length S-protein inside the cells of the vaccinated person, with the difference from the wild type S-protein being the 2 stabilising proline residues.
