Evolutionary potential of the SARS-CoV-2 receptor binding domain (RBD)

 

Jesse Bloom

Fred Hutch Cancer Research Center / HHMI

 

Slides at http://slides.com/jbloom/vrc-2020

 

@jbloom_lab

SARS-CoV-2 virions are decorated by spike proteins that mediate viral entry

A key portion of the spike is the receptor-binding domain (RBD)

The RBD binds ACE2 receptor on cells

Neutralizing antibodies often bind RBD & block interaction with ACE2

Mapping how mutations affect RBD is crucial for understanding virus's evolution

-Asn-Ile-Thr-Asn-Leu-

We are accustomed to mapping sequence to structure 

RBD phenotypes

- protein folding

- affinity for ACE2

- binding by antibodies

Structure still matters implicitly, as it largely determines mapping from mutation to phenotype.

-Asn-Ile-Thr-Asn-Leu-
-Asn-Ile-Thr-Glu-Leu-
-Asn-Lys-Thr-Asn-Leu-

We will map change in sequence to change in biochemical phenotype

Outline

  • Map how RBD mutations affect folding and ACE2 binding

Outline

  • Map how RBD mutations affect folding and ACE2 binding
  • Map how RBD mutations affect antibody binding

Outline

  • Map how RBD mutations affect folding and ACE2 binding
  • Map how RBD mutations affect antibody binding
  • Use maps to predict if / how virus can escape antibodies
  • Escape from Regeneron antibody cocktail

Outline

  • Map how RBD mutations affect folding and ACE2 binding
  • Map how RBD mutations affect antibody binding
  • Use maps to predict if / how virus can escape antibodies
  • Escape from Regeneron antibody cocktail

There are a lot of possible mutations to the SARS-CoV-2 RBD

(201 sites) X (19 amino-acid mutations per site) = 3,819 mutations

We use yeast display to enable high-throughput experiments

RBD

fluorescent ACE2

yeast

fluorescent tag on RBD

Library of yeast, each expressing different RBD mutant with identifying 16 nt barcode

Click here for details on how library is made.

For single yeast variant, it's easy to use flow cytometry to make an ACE2 binding curve

For single yeast variant, it's easy to use flow cytometry to make an ACE2 binding curve

RBD

fluorescent ACE2

yeast

(1) label yeast at different ACE2 concentrations

For single yeast variant, it's easy to use flow cytometry to make an ACE2 binding curve

RBD

fluorescent ACE2

yeast

fluorescent signal

(1) label yeast at different ACE2 concentrations

(2) measure FACS signal at each concentration

[ACE2]

For single yeast variant, it's easy to use flow cytometry to make an ACE2 binding curve

RBD

fluorescent ACE2

yeast

fluorescent signal

(1) label yeast at different ACE2 concentrations

(2) measure FACS signal at each concentration

(3) binding curve of FACS signal versus [ACE2]

[ACE2]

Use deep sequencing to measure binding by all mutants at once

fluorescent signal

[ACE2]

(1) label yeast at different ACE2 concentrations

(2) sort into FACS bins

Deep sequence to assign RBD mutants to bins and infer dissociation curves

(3)

This strategy is TiteSeq, an elaboration of deep mutational scanning.

We similarly measure expression of folded protein for each mutant using tag on RBD

yeast

fluorescent tag on RBD

Phenotypic maps of how mutations affect RBD folding & ACE2 binding

Outline

  • Map how RBD mutations affect folding and ACE2 binding
  • Map how RBD mutations affect antibody binding
  • Use maps to predict if / how virus can escape antibodies
  • Escape from Regeneron antibody cocktail

We map escape mutations by sorting for RBD variants that don't bind antibody

We map escape mutations by sorting for RBD variants that don't bind antibody

We map escape mutations by sorting for RBD variants that don't bind antibody

We map escape mutations by sorting for RBD variants that don't bind antibody

We map escape mutations by sorting for RBD variants that don't bind antibody

In maps, tall letters indicate strong escape mutations

Escape map for one antibody (COV2-2499)

Interactive structural view of escape map (click here)

Maps easier to interpret if we color sites by region of RBD

core RBD

receptor-binding motif

ACE2-contact residue

core RBD

receptor-binding motif

ACE2-contact residue

Escape maps for three antibodies

Subtle differences between COV2-2165 & COV2-2832 (e.g., sites 486, 487) validate in neutralization assays (see here).

Outline

  • Map how RBD mutations affect folding and ACE2 binding
  • Map how RBD mutations affect antibody binding
  • Use maps to predict if / how virus can escape antibodies
  • Escape from Regeneron antibody cocktail

We grew spike-expressing virus in presence of antibody

Spike-expressing VSV (Case*, Rothlauf*, ..., Whelan. Cell Host & Microbe, 2020) in a real-time cell analysis assay (Gilchuk, ..., Crowe. Immunity, 2020).

See here for more details.

Antibody COV2-2499 selected escape mutations in 5 of 16 replicates 

mutation

G446D

Q498R

count

3

2

mutation

G446D

Q498R

count

3

2

Escape map has many mutations, why were these two selected?

mutation

G446D

Q498R

count

3

2

Only some mutations accessible by single-nucleotide changes

mutation

G446D

Q498R

count

3

2

Many mutations deleterious for ACE2 binding, but ones selected in virus aren't

ACE2 binding

weak                         strong

Selected mutations are single-nt changes that escape but retain ACE2 binding

effect on escape

ACE2 affinity

effect on escape

ACE2 affinity

Phenotypic maps also explain mutations selected by other antibodies

Antibody COV2-2165 did not select any escape mutations in repeated trials

0 escape mutants in 56 attempts

Initially appear to be many escape mutations from this antibody

But only a few are accessible by single-nucleotide changes

Accessible mutations impair folding or ACE2 binding, explaining why virus can't escape

RBD expression

ACE2 binding

RBD expression /

ACE2 binding

weak                         strong

Phenotypic maps explain if / how virus escapes antibodies

Outline

  • Map how RBD mutations affect folding and ACE2 binding
  • Map how RBD mutations affect antibody binding
  • Use maps to predict if / how virus can escape antibodies
  • Escape from Regeneron antibody cocktail

REGN-COV2 cocktail is two antibodies that bind distinct structural epitopes

But antibodies share an escape mutation (E406W) that escapes cocktail

The maps validate in neutralization assays, including E406W cocktail escape

How well do escape maps predict actual viral evolution under antibody pressure?

Regeneron used cell-culture selections to identify a handful of escape mutations

Regeneron cell-culture selected escape mutations, all in our maps 

Red are mutations from Baum, ..., Kyratsous. Science (2020) among single-nucleotide accessible mutations only

We looked at viral evolution in persistently infected patient given REGN-COV2

Numerous RBD mutations are selected after REGN-COV2 treatment at day 145

Collaboration with Jonathan Li, patient originally described in Choi et al, New England Journal of Medicine (2020)

 

Note extensive genetic hitchhiking and competition among viral lineages; also described in within-patient evolution of influenza in persistent infections: Xue et al, eLife (2017)

Regeneron cell-culture

Patient

Both

 

Most escape in patient not found in cell-culture selections, illustrates utility of complete escape maps

Four patient mutations escape REGN-COV2 antibodies

Conclusions

Conclusions

  • We can create phenotypic maps of how all mutations to SARS-CoV-2 RBD affect folding, function, and antigenicity

Conclusions

  • We can create phenotypic maps of how all mutations to SARS-CoV-2 RBD affect folding, function, and antigenicity
  • These maps are useful for understanding evolution under antibody pressure

Conclusions

  • We can create phenotypic maps of how all mutations to SARS-CoV-2 RBD affect folding, function, and antigenicity
  • These maps are useful for understanding evolution under antibody pressure
  • These maps are useful for viral surveillance

Tyler Starr

Allie Greaney

Sarah Hilton

Bloom lab (Fred Hutch)

Tyler Starr

Allie Greaney

Sarah Hilton

Bloom lab (Fred Hutch)

Li lab (Brigham & Women's)

  • Jonathan Li
  • Manish Choudhary

 

Crowe lab (Vanderbilt)

  • James Crowe
  • Seth Zost
  • Pavlo Gilchuk
  • Elad Binshtein

 

Whelan lab (Wash U)

  • Sean Whelan
  • Paul Rothlauf
  • Zhuoming Liu

 

King lab (Univ Wash)

  • Neil King
  • Dan Ellis

 

Veesler lab (Univ Wash)

  • David Veesler
  • Alexandra Walls

Other members of Bloom lab:

  • Kate Crawford

  • Adam Dingens

  • Will Hannon

  • Amin Addetia

  • Andrea Loes

  • Rachel Eguia

vrc-2020

By Jesse Bloom

vrc-2020

The evolutionary potential of the SARS-CoV-2 receptor binding domain

  • 879