Thread by @ENirenberg

I see people are determined to make a mountain out of this molehill so let's talk about this paper:


thread 🧵nature.com/articles/s4158…

First, a bit of context: when your cells make proteins, the ribosome read a sequence from mRNA in the form of three-letter codes (called codons). There are 64 possible codons which specify 20 amino acids or a STOP sequence to indicate the end of the protein.

Because the codons are read 3 letters at a time, it is possible to start reading them in the wrong spot (either 1 or 2 letters over) and end up with a protein that is completely different from the one you intended because this will change the amino acid downstream because...

the codons will now be completely different. If this occurs it is known as a frameshift (note: if a frameshift occurs 3 bases over, then this doesn't happen because the reading frame won't be affected, but you may duplicate or lose 1 amino acid).

In general, frameshifts are not what you want as an outcome of protein production- but not always. Frameshifts are uncommon but naturally occurring events in, for example, viral infections. These give rise to protein products that can also be targeted by the immune system.

For example, it was recently shown that influenza can produce a unique protein product through a frameshift that is very strongly recognized by killer T cells:
This resulting frameshifted product might have an important role in protection from flu.nature.com/articles/s4159…

Coronaviruses have evolved to require frameshifting as part of their replication cycle to make their proteins:

Anyway, let’s get back to this paper. mRNA vaccines are made with replacement of all of the U’s in the sequence of interest with m1ψ.jbc.org/article/S0021-…

This modification naturally occurs in our own ribosomal RNA. RNA has to be modified to prevent triggering of an antiviral immune response that would suppress production of the encoded protein (which would then prevent an effective antibody response):

https://x.com/ENirenberg/status/1666685106294243328?s=20

Prior work has shown that this results in faithful production of the protein product of interest (spike):

This is a really important point not to lose sight of: the mRNA vaccines are excellent at making the thing we want them to make.ncbi.nlm.nih.gov/pmc/articles/P…

This work argues however that this modification promotes frameshifting. Let’s take a look. First, the authors used a sample mRNA sequence that was designed to produce the end of the enzyme luciferase, which is what enables fireflies to glow.

If the Fluc+1FS RNA were made without frameshift mutations, then there wouldn’t be detectable luciferase activity from the frameshifted product. Here we can see that the m1ψ produced luciferase has the frameshifted product as evident by activity of the luciferase.

On the western blot examining the protein product resulting from different RNA modifications, while there are 2 additional bands for the m1ψ mRNA, it is noteworthy that there is faithful production of the in-frame product as well.

This broadly agrees with work from 2015 that showed that there is markedly enhanced protein production from m1ψ RNA.

This is, fundamentally, the point of using the modification: you want faithful production of your protein inside actual cells.sciencedirect.com/science/articl…

The authors however took it a step further, wondering whether the frameshifting was actually happening in real life. They vaccinated mice with the Pfizer or Oxford/AZ vaccine and looked for evidence of T cell responses against out-of-frame peptides, and…

the mice that received the Pfizer vaccine did show responses against the out-of-frame peptides (2b). But: they also had greater T cell responses to the in-frame peptides (2c) than those that received Oxford/AZ vaccine.

When they looked in humans, they found that the recipients of the Pfizer vaccine also had responses against out-of-frame peptides, but this varied a lot across individuals, suggesting that the extent of frameshifting may also vary a lot between individuals. Note: 2e is mouse data

Anyway, from here, the authors go on to identify “slippery” sites in mRNA that are prone to frameshifting, and they also show that the antibiotic paromomycin (which interferes with decoding of codons by the ribosome) promotes the fidelity of translation with the modified RNA.

The authors show that modification of these sequences to get rid of slippery sites seems to reduce frameshifting events. So, what’s the significance of these results?

In a practical sense as far as it applies to vaccination, IMO not much.

The sequences of Pfizer and Moderna differ significantly at the nucleotide level, but the nature of the responses and their safety profile is broadly similar, arguing against frameshift products in either case having a meaningful effect.

I do think that the major use of these findings is that they highlight the importance of understanding how immune responses against frameshifted peptides affect protection in the case of infection.

The major reason this would be relevant is because it adds a second dimension to codon optimization (because codons are redundant for any given amino acid, you can choose the “best” one for your particular application) in that...

one might want to optimize in such a way that it preserves major frameshifted peptides that may be important for immune responses.

Still, frameshifting within spike protein is not known to date to result in peptides of special importance to the immune response, so...

it’s not likely that it matters much here either way. The authors are (correctly) explicit here that there is no evidence of this contributing to any kind of safety issues with the mRNA vaccines.

Some clueless people claim that this somehow means there's a heightened...

risk of autoimmune disease because of these frameshifted products. For this to be true, there would need to be meaningful production of products that closely resemble our own proteins- something there. is no evidence for.

We in fact see that in stark contrast to COVID-19, there is no autoimmune signature for mRNA vaccines:


So, basically: a useful study but not good cause to worry about mRNA vaccines.nature.com/articles/s4146…

Another helpful point:

https://twitter.com/marc_veld/status/1732522104036192749

I realized I should perhaps include another point of clarification to help people understand the significance of this issue.

Within cells, any given mRNA is translated by many ribosomes (not just one). mRNA can form assembles of ribosomes called polysomes or polyribosomes:

m1ψ mRNA is particularly good at forming polysomes:

This helps to explain why the yield of proteins from mRNA bearing this modification is so much higher. The significance of this is simply that your cells have a ton of opportunities to make...ncbi.nlm.nih.gov/pmc/articles/P…

the correct protein from any given mRNA. If one ribosome frameshifts, it does not mean *every* ribosome will frameshift. Remember: the frameshifted product was a minor fraction of the overall proteins.

People should also take note that questions about the safety of mRNA vaccines have been repeatedly asked and answered. Data on billions of doses across the last few years affirms their excellent safety profile (reactogenicity notwithstanding). This requires no knowledge...

of cellular and molecular mechanisms. We eradicated smallpox with a vaccine that essentially pre-dates any modern understanding of the immune system. Despite the sexiness of the mechanisms, pharmacoepidemiology is and should always be king of safety evaluations.