Thread by @jbloom_lab

After posting our pre-print mapping #SARS_CoV_2 mutations that escape Regeneron antibodies, been getting questions about implications for therapies/vaccines. I'd like to contextualize results. Summary is don't be alarmed, but pay attention to viral evolution. Long version: (1/n)

https://twitter.com/jbloom_lab/status/1333897221561978880

Antibody treatments consist of a single antibody (or in some cases a cocktail of a few) that bind viral spike. Since an antibody binds to one small patch of virus, typically a single mutation is sufficient to escape binding by an antibody. (2/n)

We know from other viruses this can happen. One of the best examples is Regeneron's trial of an antibody to treat RSV in infants (academic.oup.com/cid/advance-ar…). They ran an entire large / expensive clinical trial that failed. Retrospectively, the reason was obvious... (3/n)

There are 2 groups of RSV & one group (RSV B) had mutations that escaped antibody. If this had been known a priori, Regeneron could have targeted RSV A or used different antibody. Goal of our work is to prevent similar failures by identifying escape mutations ahead of time. (4/n)

For #SARS_CoV_2, we know from our pre-print & many other recent studies that there are mutations to virus capable of escaping most antibodies. See also for instance work by @theodora_nyc @PaulBieniasz @vsv512 : elifesciences.org/articles/61312 and biorxiv.org/content/10.110… (5/n)

Furthermore, some escape mutations are already present. For instance, >1 of every 1000 virus isolates already has either Y453F (which escapes one Regeneron antibody) and N439K (which escapes the other). Importantly, this result isn't just shown in our new pre-print... (6/n)

... but also corroborated by others. Regeneron's own paper shows Y453F escapes REGN10933 (Fig S1 of science.sciencemag.org/content/369/65…) and Thomson et al (biorxiv.org/content/10.110…) shows N439K escapes REGN10987 (7/n)

The Regeneron cocktail will still work against either one of those mutants alone, but the fact that >1 in 1000 isolates already have mutations that escape one or the other antibody in cocktail shows escape is a real possibility. (8/n)

Our paper then finds many other additional possible escape mutations (in fact, it maps all of them). Does this mean we should throw up our hands? No!

What it means is we need to use antibodies in a way that manages viral evolution. (9/n)

If we know which mutations escape any given antibody, and we are sequencing viruses, then we can see when possible resistance mutations arise. This can all be managed if we combine surveillance with interpretation of what mutations are doing. (10/n)

For instance, HIV evolves much faster than #SARS_CoV_2, but can be treated with drug cocktails in ways that limit escape. Problem is much easier for #SARS_CoV_2 since it evolves slower, but it's still an evolving virus so we can't just put our head in the sand. (11/n)

For this reason, we think our results should make people optimistic rather than fearful. If we know what potential escape mutations are out there, it will inform how to use the many great antibodies being developed in a way that mitigates impacts of viral evolution. (12/n)

Our work also found a single amino-acid mutation, E406W, that escapes both antibodies in Regeneron cocktail. Fortunately this mutation is not accessible by a single-nt change, so it's not more alarming than single-antibody escape mutations already known (eg, Y453F, N439K) (13/n)

But existence of E406W does demonstrate that you can't figure out everything about escape mutations just by looking at protein structures, since mutation isn't in direct contact with antibody. How does E406W work? We don't know... (14/n)

But E406W effect is robust in binding and viral neut assays (we're investigating more too, and hope our pre-print will spur others to do same). More broadly, I think it's important to report this mutation even if we can't rationalize effect directly from structure. (15/n)

After all, proteins are complex, & not everything about mutations can be rationalized just by staring at structures: if it could, my PhD advisor @francesarnold wouldn't have won Nobel Prize for directed evolution as we'd just engineer proteins by structure gazing instead. (16/n)

Finally, what do our results imply for vaccines and natural immunity? These will be *much harder* for virus to evolve to escape as they induce many antibodies (& also T-cells) that target lots of parts of virus... (17/n)

*Maybe* some antigenic drift from vaccine immunity will happen (important area for future work), but if so it will certainly be slower than antibody escape. Again, we should study this, and if needed vaccines can be updated to handle it. (18/n)

So overall, #SARSCoV2 is already infecting millions of people and generating billions of mutations. Most mutations have no effect, but some will escape antibodies & be acted on more or less efficiently by natural selection depending on immune pressure. Fortunately, ... (19/n)

... lots of antibodies (& vaccines) being developed. As long as we do a good job interpreting which mutations are important (goal of our work) the scientific community can monitor evolution & adjust. So we don't need to be alarmed by viral mutations, but just pay attention (20/n)

Finally, thanks to @tylernstarr @khdcrawford @VUMC_Vaccines for frank feedback that emphasized need to provide additional contextualization to the scientific results in the pre-print (21/n)