Thread by @jbloom_lab

We've experimentally measured how all amino-acid mutations to the #SARSCoV2 spike RBD affect ACE2 binding and expression of folded protein in a deep mutational scanning study led by @tylernstarr & Allie Greaney: biorxiv.org/content/10.110…

Why is this important? (1/n)

The RBD (receptor binding domain) enables #SARSCoV2 to bind to human cells. Evolution to bind human ACE2 was key to the emergence of this virus. Now it's also key to mitigating the virus: the most potent antibodies bind to RBD, and most vaccine candidates contain RBD. (2/n)

Before our study, the structure of the RBD protein was known but we didn't know how mutations affected it's function. Now we've measured how virtually all (3800 of 3819) amino-acid mutations affect binding to ACE2 and expression of the folded protein. (3/n)

If you're interested in the evolution of the virus or are studying antibodies / vaccines, you can go to this interactive heatmap (jbloomlab.github.io/SARS-CoV-2-RBD…) and simply look up the *experimentally measured* effect of any mutation. (4/n)

Some observations:

While many mutations to RBD are deleterious, lots are tolerated--including a surprising number that increase ACE2 affinity (blue in heat map in prior Tweet). But no evidence that strong affinity enhancing mutations are being selected in human isolates. (5/n)

We can map tolerance to mutations with respect to RBD folding / expression & ACE2 binding on structure (red = intolerant in images below). As expected, most selection for binding in ACE2 interface--but again, lots of mutations still tolerated (again see heat maps or paper) (6/n)

We can examine mutational tolerance of epitopes of different antibodies, and compare it to mutational tolerance of actual ACE2 binding interface (here are some antibodies, see paper for more). (7/n)

Key finding from that comparison is that no antibodies have epitopes as constrained as actual ACE2.
So epitope focusing can elicit more escape-resistant antibodies. And our data help point the way, as we show what mutations would be tolerated in resurfaced immunogens! (8/n)

Lots of other observations about natural mutations, structure-function, and sarbecovirus evolution in the paper itself, so take a look. (9/n)

Finally, how we were able to characterize so many mutations to this important protein? We leveraged great prior work. In 2016, @jbkinney, Walczak, & Mora showed yeast display & deep mutational scanning could measure dissociation constants at scale: elifesciences.org/articles/23156 (10/n)

To enable these deep mutational scanning titration assays to be applied to larger protein like RBD, we leveraged barcoding approach originally developed by @JShendure and subsequently extended to use PacBio by @lea_starita @kmatreyek @dougfowler42: nature.com/articles/nmeth… (11/n)

@tylernstarr had been optimizing approach for other proteins when #SARS_CoV2 hit. We had as colleagues @veeslerlab @coronalexington who had been building basic knowledge about coronavirus spikes for years. They helped advise @tylernstarr how to adapt his approach to RBD. (12/n)

Allie Greaney jumped on board to help @tylernstarr do the experiments amazingly fast, and Sarah Hilton and @khdcrawford helped with code and validation. (13/n)

Now we hope data we generated will help others studying evolution of virus and especially those developing vaccines. It's all publicly available, so if you're curious about a specific mutation to RBD, need a stabilizing mutation, or need to change a surface--go look it up! (14/n)