Thread by @jbloom_lab

In a new study (biorxiv.org/content/10.110…), we have completely mapped mutations to #SARSCoV2 that escape human antibodies, and shown that these "escape maps" predict how virus evolves under antibody pressure & inform design of escape-resistant antibody cocktails. (1/12)

Background: antibodies that target the #SARSCoV2 receptor-binding domain (RBD) are being developed as therapeutics & vaccines aim to elicit them. Among most potent RBD antibodies are set isolated by @VUMC_Vaccines, @seth_zost & Pavlo Gilchuk (nature.com/articles/s4158…) (2/12)

Key question is what mutations to #SARSCoV2 RBD escape antibody binding. Classic way to determine this is to grow virus w antibody & see what is selected. But this approach is incomplete: any given replicate stochastically selects at most 1 of possible escape mutations. (3/12)

Fortunately, Allie Greaney & @tylernstarr developed deep mutational scanning method to measure how all mutations to RBD affect ability to fold & bind receptor (cell.com/cell/fulltext/…). Here they extended method to *completely* map mutations that escape antibody binding. (4/12)

For each antibody, they get escape map like below. Tall letters indicate mutations that escape antibody. By inspection of map, we see for instance that antibodies 2499 & 2096 are escaped by some of the same mutations, but that 2499 & 2050 have orthogonal escape mutations. (5/12)

The paper has escape maps for 10 antibodies, and you can see interactive versions here (jbloomlab.github.io/SARS-CoV-2-RBD…). (6/12)

We can organize antibodies in space of escape mutations. Left plot divides RBD in regions & right plot arranges antibodies so ones w similar escape mutations are close, coloring by RBD region of escape. Antibodies that bind similar region not always close in escape space! (7/12)

Then it gets really cool. Leveraging spike-expressing VSV from @vsv512 & @paulrothlauf, Pavlo & @seth_zost performed 100s of escape selections with various antibodies. Some antibodies selected escape mutations, some didn't. Can we understand which mutations selected & why? (8/12)

Yes! Plot below shows all RBD mutations & how they affect antibody & ACE2 binding. Selected mutations consistently escape antibody w/o impairing ACE2 binding & accessible by single nt change. The antibody (2165) that isn't escaped? D420Y bad for RBD folding (see paper). (9/12)

So we can combine complete escape maps with deep mutational scanning measurements of how mutations affect RBD folding and function to understand which antibodies are easily escaped, and which mutations will escape them. (10/12)

Finally, we use complete escape maps to design cocktails of antibodies that resist escape despite binding same region of RBD. This is new paradigm in therapeutic antibody cocktails: if you know enough about escape, antibodies don't need to bind different regions. (11/12)

Lots of details can't fit in Tweets, so read paper if you have time.

Main credit to Allie Greaney, @tylernstarr, Pavlo Gilchuk, @seth_zost for this innovative work.

Also many others @fredhutch & @VUMC_Vaccines, e.g. @EladBinshtein @huddlej @khdcrawford @AdamDingens (12/12)