In work led by , we develop pseudovirus system to deep mutational scan full #SARSCoV2 spike: biorxiv.org/content/10.110
TLDR: can now safely measure how all mutations to viral entry proteins affect antibody neutralization.
🧵 below (1/n)
Conversation
Replying to
Background: viruses have entry protein(s) that enable them to infect cells.
Entry proteins are main target of neutralizing antibodies, so when they evolve rapidly can cause antibody escape.
Binding & fusion by entry proteins also determines cell tropism (2/n)
1
5
14
Over last ~3 yrs, we’ve seen importance of evolution of #SARSCoV2 entry protein, spike.
Spike has escaped antibodies, requiring vaccine updates.
Mutations increasing spike stability (D614G) or receptor affinity (N501Y) coincided w emergence of more transmissible variants (3/n)
1
5
18
In 2020, in our group developed yeast-display deep mutational scanning method to measure how 1000s of mutations to one important part of spike (receptor-binding domain or RBD) affect antibody & ACE2 binding cell.com/cell/fulltext/ (4/n)
1
6
19
These measurements help interpret #SARSCoV2 evolution.
Eg, in early 2021 identified E484 as major site of antibody escape in early viruses (twitter.com/jbloom_lab/sta), then predicted 486 as site of escape in Omicron before it emerged in BA.4/BA.5 (twitter.com/jbloom_lab/sta) (5/n)
Quote Tweet
Looking at antibodies that neutralize #Omicron helps understand portion of antibody response to current vaccines/ infections that still neutralizes #Omicron. These antibodies often target sites 346, 348, 444, 456, 468, 475, 486, 487. Watch those sites for future mutations! (8/n)
Show this thread
2
6
18
Recently, et al took yeast-display RBD deep mutational scanning to next level to explain convergent evolution of new #SARSCoV2 variants & understand specificity of antibodies boosted by Omicron breakthrough: twitter.com/yunlong_cao/st (6/n)
Quote Tweet
Sharing our investigation on the unprecedented convergent RBD evolution of BA.2.75 and BA.5 on sites including 346, 356, 444-446, 450, 460, 486, which have generated highly concerning variants such as BA.2.75.2, BR.1, BJ.1, and BQ.1.1. (1/n)
biorxiv.org/content/10.110
Show this thread
1
4
21
This deep mutational scanning has formed basis of “antibody escape calculator” that is useful tool for interpreting antigenic impact of newly identified mutations in #SARSCoV2 variants: twitter.com/K_G_Andersen/s (7/n)
This Tweet is unavailable.
2
6
19
RBD deep mutational scanning has even contributed to vaccine design by mapping specificity of antibodies elicited by next-generation coronavirus vaccine candidates: twitter.com/ScienceMagazin (8/n)
Quote Tweet
Nanoparticles displaying the #SARSCoV2 receptor-binding domain (RBD) protected only from SARS-CoV-2 challenge, whereas nanoparticles displaying RBDS of SARS-CoV-2 and 7 animal sarbecoviruses protected from SARS-CoV challenge, too, a new animal study shows. fcld.ly/levq544
1
4
16
However, like any technique, yeast-display RBD deep mutational scanning has limitations:
1. Only encompasses RBD, which is just part of spike
2. Measures antibody binding, not neutralization
3. Doesn’t generalize to viruses w entry proteins not amenable to yeast display
(9/n)
2
3
20
Some of these limitations have been overcome by excellent recent work using mammalian cell display, such as by (cell.com/molecular-cell) & (twitter.com/wchnicholas/st) (10/n)
Quote Tweet
Our new study led by @timothyjtan measured the effects of thousands of S2 mutations on SARS2 spike fusogenicity and expression. It involves the development of a new high-throughput method. (1/12)
biorxiv.org/content/10.110
Show this thread
1
3
17
However, we decided to take a different approach and move away from cell display altogether because we wanted to measure how mutations affect *neutralization* and cellular entry, not just binding. (11/n)
1
4
20
We therefore based our system on lentiviral pseudotyping, a safe way to study entry proteins of many viruses, including #SARSCoV2
Pseudotyping expresses entry protein on lentiviral particles, where it mediates single round of cellular infection blog.addgene.org/viral-vectors- (12/n)
1
4
16
Challenge is creating genotype-phenotype link between spike on virion surface & barcode in viral genome, since transfected cells take up multiple plasmids & lentiviruses are pseudodiploid. We solved this with two-step approach below. (13/n)
3
8
32
Using this system, we created replicate libraries of Omicron BA.1 & Delta spikes encompassing ~7000 spike amino-acid mutations observed in natural sequences or are at sites of positive selection. Most mutations in several multi-mutant backgrounds in our libraries. (14/n)
1
5
17
We wanted to use libraries to measure how spike mutations affect antibody neutralization.
That poses another challenge: deep sequencing reads out *relative* abundances of mutations after antibody selection, but neutralization is an absolute-scale measurement. (15/n)
1
5
14
To solve this, we developed an “absolute neutralization standard.”
This standard (VSV-G pseudotype) is *not* neutralized, so we can use it to convert relative abundances of spike mutants from deep sequencing to absolute scale of neutralization. (16/n)
1
4
19
Inclusion of the absolute neutralization standard enables us to quantitatively measure neutralization of many thousands of spike variants at different antibody concentrations just using deep sequencing. (17/n)
1
4
15
As proof-of-principle, below are escape measurements for RBD-targeting antibody LY-CoV1404 vs BA.1 spike.
To explore data, look at interactive plots at dms-vep.github.io/SARS-CoV-2_Omi rather than just static ones below. (18/n)
1
7
17
Because our deep mutational scanning directly measures neutralization, the predicted change in IC50 of mutants from the deep mutational scanning is very well correlated with actual IC50s measured by traditional pseudovirus neutralization assays. (19/n)
1
5
20
To demonstrate how the pseudovirus deep mutational scanning can also map non-RBD antibodies, we collaborated with Lihong Liu & David Ho to map escape from NTD-targeting antibody 5-7 (below & interactive plots at dms-vep.github.io/SARS-CoV-2_Omi) (20/n)
1
2
14
We then collaborated with & Dennis Burton to map two antibodies targeting the S2 stem helix. The deep mutational scanning again validated well, & explained why only one of the two antibodies cross-neutralized MERS-CoV (21/n)
1
3
15
Our deep mutational scanning also measures how well each spike mutant mediates pseudovirus infection of 293T-ACE2 cells. Here is distribution of functional scores representing infection efficiency across spike mutants with different types of mutations. (22/n)
1
3
12
We used global epistasis models to deconvolve these measurements into estimates of how each mutation affected pseudovirus infection. These measurements can be interactively explored here: dms-vep.github.io/SARS-CoV-2_Omi (23/n)
1
3
14
Although most mutations bad or neutral for spike-mediated pseudovirus infection, some are predicted by deep mutational scanning to improve infection.
We directly tested a few of these mutations & validated that they all modestly increased pseudovirus infection. However… (24/n)
1
3
15
… mutations we validated to increase pseudovirus infection not favorably selected in evolution of actual #SARSCoV2. Eg, we speculate P1140T & P1143S destabilize prefusion trimer, promoting rapid cell entry in culture but hurting actual transmission via impaired stability. (25/n)
1
3
16
This finding consistent w known fact spike mutations that improve pseudovirus infection (eg, deleting last 19 amino acids of cytoplasmic tail) not selected in actual #SARSCoV2, indicating some divergence between spike-mediated pseudovirus infection & actual viral fitness. (26/n)
1
2
13
To quantify relationship between deep mutational scanning & natural evolution, we elaborated idea developed by (biorxiv.org/content/10.110) to quantify enrichment/depletion of each mutation in natural evolution vs expectation from neutral mutation rate (27/n)
1
3
15
We can plot how each deep mutational scanning dataset correlates w actual enrichment of mutations in #SARSCoV2 evolution (keep in mind both experiments & enrichment estimates have noise). Also see interactive version of plot dms-vep.github.io/SARS-CoV-2_Omi (28/n)
1
2
15
Current pseudovirus deep mutational scanning correlates better w enrichment in actual evolution than other datasets from our lab & . But really is there is no “best” dataset. They all measure relevant molecular phenotypes that holistically contribute to fitness (29/n)
1
2
17
All data & analysis notebooks described above are available at dms-vep.github.io/SARS-CoV-2_Omi & thanks to all the paper authors mentioned above as well as Will Hannon, @timycuu & Ariana Ghez Farrell (30/n)
1
4
13
Finally, I’d like to add few forward-looking thoughts. Lentiviral pseudotyping is safe way to study entry proteins of numerous viruses. This includes both circulating human-adapted viruses like #SARSCoV2 & potential pandemic viruses: blog.addgene.org/viral-vectors- (31/n)
1
10
22
Show replies
Show additional replies, including those that may contain offensive content
Show