Interpreting evolution of SARS-CoV-2 and other viruses

The Faroe Islands

"Measles had not prevailed on the Faroes since 1781, then it broke out early in April 1846."

"Of the 7782 inhabitants, about 6000 were taken with measles."

Lasrado et al (2023); Greaney et al (2021)

"Of the many aged people still living in the Faroes who had measles in 1781, not one was attacked the second time."

Panum was describing immune memory, which provides lifelong protection from measles.

What was reason for thinking coronavirus immunity would not be affected by viral mutations?

Coronaviruses have lower mutation rate than other RNA viruses

Coronaviruses are only RNA viruses with proofreading activity in their polymerase, and so have ~5- to 10-fold lower mutation rate than influenza virus

But rate is still high enough that evolution is not mutation-limited

The average single-nucleotide mutation to SARS-CoV-2 had occurred >10,000 independent times in human-transmitted SARS-CoV-2 by third year of pandemic.

Add some plot here

Evolution is not just mutation: also involves selection on phenotypic effects of mutations

We first asked: what is net effect of mutations to a coronavirus over decades of evolution in humans?

CoV-229E causes common colds and has been circulating in humans for a long time.

Typical person is infected every ~3 to 5 years.

We first asked: what is net effect of mutations to a coronavirus over decades of evolution in humans?

Evolution of CoV-229E spike

We experimentally generated CoV-229E spikes at ~8 year intervals so we could study them in the lab:

- 1984

- 1992

- 2001

- 2008

- 2016

Eguia et al

Evolution of CoV-229E spike erodes neutralization by human serum antibodies

Eguia et al

Strongest evolutionary selection in RBD

Sites of evolutionary change in the spike of CoV-229E over the last four decades

Sites of mutations in SARS-CoV-2 Omicron BQ.1.1 spike relative to Wuhan-Hu-1

Eguia et al

Ideally vaccines would elicit evolution-resistant neutralizing antibodies (like those made by person at right) rather than evolution-sensitive antibodies (like those made by person at left)

Evolution erodes neutralization by sera from different people at different rates

We now know SARS-CoV-2 evolves to erode neutralizing antibodies, just like CoV-229E

neutralization from original COVID-19 vaccine

Original vaccine induced hight neutralizing antibody titers against early viral strains

Lasrado et al (2023)

We now know SARS-CoV-2 evolves to erode neutralizing antibodies, just like CoV-229E

Lasrado et al (2023)

newer viral variants

neutralization from original COVID-19 vaccine

Main regions where neutralizing antibodies bind

Molecular basis of this antigenic evolution?

Lasrado et al (2023); Greaney et al (2021)

Molecular basis of this antigenic evolution?Mutations where antibodies bind

Main regions where neutralizing antibodies bind

Sites of mutations in recent (BA.2.86) SARS-CoV-2 strain relative to early 2020 strain

So overall, some human RNA viruses evolve to escape immunity, others don't

Why only some RNA viruses evolve to escape immunity is a deep question that has not been fully answered (see here for some theories).

But all these viruses have high mutation rates, so explanation has more to do with phenotypic effects of mutations than rate at which they arise.

Rate of viral antigenic evolution

Measles

Mumps

Influenza

SARS-CoV-2

1. Define a virus's evolutionary potential (eg Starr et al 2020; Starr et al 2022)

2. Quantify how easily a virus can escape a specific antibody or vaccine (eg Greaney et al 2021; Carr et al 2024)

3. Forecast and interpret viral evolution (eg, Dadonaite et al 2024)

Measuring phenotypic effects of viral mutations can enable us to:

1. Define a virus's evolutionary potential (eg Starr et al 2020; Starr et al 2022)

2. Quantify how easily a virus can escape a specific antibody or vaccine (eg Greaney et al 2021; Carr et al 2024)

3. Forecast and interpret viral evolution (eg, Dadonaite et al 2024)

Measuring phenotypic effects of viral mutations can enable us to:

All viruses have entry proteins, which are critical determinants of host range and immunity

target cell membrane

SARS-CoV-2 virion

Example: SARS-CoV-2's entry protein spike first binds ACE2 on target cells

spike protein

Image adapted from here

ACE2

After binding receptor, spike then fuses the viral and cell membranes to enable viral entry

viral membrane

cell membrane

spike

spike conformational change

Image adapted from here

ACE2

Antibodies to spike can block viral entry, such as by preventing receptor binding or fusion

antibody

Image adapted from here

Deep mutational scanning to measure effects of mutations to viral entry proteins

Deep mutational scanning involves pooled selections on libraries of protein mutants, read out by deep sequencing

Initially we used yeast display of receptor binding domain (RBD) of SARS-CoV-2 spike

cell sorting

RBD

fluorescent ACE2

yeast

fluorescent tag on RBD

Yeast display allows us to separate RBD mutants by how strongly they bind ACE2

Starr et al (2020)

In 2020, we identified N501Y as affinity-enhancing mutation; by 2021 it was widespread in SARS-CoV-2

Starr et al (2020)

RBD

fluorescently labeled antibody

yeast

fluorescent tag on RBD

We also used yeast display to measure how all RBD mutations affect antibody binding

Greaney et al (2021)

See https://jbloomlab.github.io/SARS2-RBD-escape-calc/ for interactive version of this antibody escape calculator (data from our lab and Yunlong Cao's), which has been used >100,000 times

site in RBD

antibody escape

We identified E484 and K417 as key sites of antibody escape; by 2021 mutations at both sites were common in SARS-CoV-2

484

417

Greaney et al (2021)

These anecdotes suggest experiments can help predict viral evolution

But we wanted more relevant experiments: yeast display only measures binding

cell sorting

Viral infection is more biologically relevant measure of mutation effects

However, we are cognizant of biosafety concerns of mutating actual virus

actual SARS-CoV-2 virion: pathogen capable of spread in humans

pseudotyped lentiviral particle: not a pathogen, cannot spread in humans

actual SARS-CoV-2 virion: pathogen capable of spread in humans

pseudotyped lentiviral particle: not a pathogen, cannot spread in humans

Therefore we use pseudoviruses rather than actual replicative SARS-CoV-2

Pseudoviruses are modified viruses that can only undergo single round of infection

Crawford et al (2020)

We need to link phenotype (spike mutant) and genotype (sequence encoded by virus)

To link genotype to phenotype link, we encode spike in viral genome with barcode

Dadonaite et al (2023)

We then create genotype-phenotype linked libraries by two-step process

Dadonaite et al (2023)

Result is library of spike mutant pseudoviruses with identifying barcodes

Captures full cell-entry function of spike
Each mutant identified by short barcode
Pseudoviruses safe at biosafety-level 2
Broadly applicable to viral entry proteins

Dadonaite et al (2023)

Measured how spike mutations to XBB.1.5 spike affect three molecular phenotypes

Measured how spike mutations to XBB.1.5 spike affect three molecular phenotypes

Measured how spike mutations to XBB.1.5 spike affect three molecular phenotypes

Neutralization by soluble ACE2 is proportional to ACE2 binding affinity

Measured how spike mutations to XBB.1.5 spike affect three molecular phenotypes

Full workflow for measure effects of mutations on antibody neutralization

Deep mutational scanning correlates well with traditional neutralization assays

We next estimated actual fitness (growth rate) of SARS-CoV-2 variants in humans

Fitness from multinomial logistic regression

With Trevor Bedford & Ben Murrell

Fitness from multinomial logistic regression

With Trevor Bedford & Ben Murrell

We analyze changes in clade growth rather than raw clade growth

change in clade growth

clade growth

Deep mutational scanning measurements correlate with SARS-CoV-2 variant growth

Combining phenotypes offers best predictions (because mutations can involve tradeoffs)

Can partially predict evolution in real-world from experiments!

Extending approach to other viruses

SARS-CoV-2 spike
influenza hemagglutinin (HA)
HIV envelope protein
Lassa virus glycoprotein
Nipah virus RBP and F proteins
RSV G and F proteins
Rabies G protein
Chikungunya virus E proteins

All enveloped viruses have entry proteins

SARS-CoV-2 spike
influenza hemagglutinin (HA)
HIV envelope protein
Lassa virus glycoprotein
Nipah virus RBP and F proteins
RSV G and F proteins
Rabies G protein
Chikungunya virus E proteins

All enveloped viruses have entry proteins

Pseudovirus of clade 2.3.4.4b H5 HA

HA molecular phenotypes relevant to pandemic risk

HA molecular phenotypes relevant to pandemic risk

HA molecular phenotypes relevant to pandemic risk

HA molecular phenotypes relevant to pandemic risk

HA molecular phenotypes relevant to pandemic risk

These data can inform surveillance of H5 influenza viruses spreading in animals

We prospectively identified A160T as strongly reducing neutralization by sera elicited by candidate vaccine strains

(L122Q, A160T, T199I)

(L122Q, P162Q, T199I)