SARS-CoV-2 evolutionary dynamics


Trevor Bedford (@trvrb)
Associate Professor, Fred Hutchinson Cancer Research Center
12 Nov 2021
Emerging Variants Meeting
Massachusetts Consortium on Pathogen Readiness
Slides at:

1. Current circulation patterns

2. Assessing adaptive evolution

3. Variant transmission dynamics

4. Potential for epistasis

5. Future perspectives

Current circulation patterns

Spread of VOC / VOI lineages across the world

Delta is outcompeting other variants and is on track to sweep

Locally emerging sublineages such as AY.4.2 with spike mutations 145H and 222V

Variants show excess mutations across the genome

But show most substantial excess in the S1 domain of spike

Assessing adaptive evolution

Rapid and parallel adaptive mutations in spike S1 drive clade success in SARS-CoV-2

Katie Kistler,   John Huddleston


Phylogeny of 10k genomes equitably sampled in space and time

Measure clade growth as a proxy for viral fitness

Clades with more S1 nonsynonymous mutations grow faster

This correlation is absent in control regions

This correlation is stronger in S1 than other genes

S1 is quickly evolving and highly correlated with clade growth

dN/dS through time further highlights adaptive evolution

Rapid pace of adaptive evolution relative to H3N2 influenza

S1 mutations cluster phylogenetically, suggesting bursts of evolution

Many S1 mutations show strong convergent evolution

Mutations in S1 arising via within-host pressures result increase viral fitness and are enriched in the viral population by natural selection

Although selection has not been primarily for antigenic drift, observed level of adaptability suggests its potential

Variant transmission dynamics

Multiple approaches to modeling fitness differences between circulating variants

Multinomial logistic regression models work well on frequencies

However, frequency of a variant may rise while cases fall

SARS-CoV-2 variant dynamics across US states show consistent differences in transmission rates

Marlin Figgins


Estimation of variant-specific Rt through time using state-level data

State-level case counts are partitioned based on frequencies of sequenced cases

Figgins and Bedford. Unpublished.

Differences in intrinsic Rt across variants, but all trending downwards

Figgins and Bedford. Unpublished.

Consistent differences in variant-specific transmission rate across states

Figgins and Bedford. Unpublished.

Consistent reductions in variant-specific Rt from vaccination

Figgins and Bedford. Unpublished.

Future work would ideally tie together granular empirical estimates of viral fitness from frequency data together with mutational and phenotypic predictors to learn what is driving variant success

Model of additive effects of individual mutations shows little fitness variation within Delta

Potential for epistasis

Mutations at spike 484 and 501 perhaps less fit on Delta background

Mutations at spike 484 and 501 perhaps less fit on Delta background (zoomed to Delta)

Future perspectives

Broad expectations based on comparison with flu

  1. R0 of Delta is perhaps ~5 compared to R0 of flu of ~2. At same rates of evolution and waning expect more COVID circulation.
  2. Rates of adaptive sequence evolution in SARS-CoV-2 have been about ~5X that of H3N2 flu. Expect this to slow down, but suggests that virus will be generally quite adaptable.
  3. IFR of COVID is roughly comparable to flu once you have prior immunity granting durable memory responses.

1, 2 and 3 suggest a virus that circulates like flu does at perhaps higher levels (given R0), but individual infections aren't much more severe in terms of mortality than H3N2 influenza. Expect perhaps 50-100K deaths in the US each year.

SARS-CoV-2 scientific (and prosaic) priorities

  1. Forecasting evolution and ensuring vaccine updates match circulating viruses
  2. Universal vaccine to engender broader immune response and/or to target more conserved epitopes
  3. Plentiful testing, contact tracing and a public health apparatus
  4. Treatments, ventilation, etc...


SARS-CoV-2 genomic epi: Data producers from all over the world, GISAID and the Nextstrain team

Bedford Lab: John Huddleston, James Hadfield, Katie Kistler, Louise Moncla, Maya Lewinsohn, Thomas Sibley, Jover Lee, Cassia Wagner, Miguel Paredes, Nicola Müller, Marlin Figgins, Eli Harkins, Denisse Sequeira