Science Abe6522 Full
Science Abe6522 Full
Science Abe6522 Full
We are currently faced with the question of how the CoV-2 severity may change in the years ahead. Our
analysis of immunological and epidemiological data on endemic human coronaviruses (HCoVs) shows that
infection-blocking immunity wanes rapidly, but disease-reducing immunity is long-lived. Our model,
incorporating these components of immunity, recapitulates both the current severity of CoV-2 and the
Humans have regularly been threatened by emerging patho- population in which older age-groups with no previous expo-
gens that kill a substantial fraction of all people born. Recent sure are most vulnerable to severe disease. We use our esti-
decades have seen multiple challenges from acute virus infec- mates of immunological and epidemiological parameters for
tions including SARS, MERS, Hendra, Nipah and Ebola. For- endemic HCoVs to develop a quantitative model for endemic
tunately, all were locally contained. When containment is not transmission of a virus with SARS-CoV-2 -like characteristics,
immediately successful, as is likely for the novel betacorona- including the age-dependence of severity. Our model explic-
virus SARS CoV-2 (CoV-2) (1, 2), we need to understand and itly considers three separate measures for immune efficacy
plan for the transition to endemicity and continued circula- that wane at different rates (fig. S1).
tion, with possible changes in disease severity due to virus Building on ideas from the vaccine modeling literature,
evolution and build-up of host immunity and resistance. immunity may provide protection in three ways (7). In its
CoV-2 is an emerging virus that causes COVID. The virus most robust form, “sterilizing” immunity can prevent a path-
has a high basic reproductive number (R0) and which is trans- ogen from replicating, thereby rendering the host refractory
missible during the asymptomatic phase of infection, both of to reinfection. We term this property immune efficacy with
which make it hard to control (3). However, there are six respect to susceptibility, IES. If immunity does not prevent
other coronaviruses with known human chains of transmis- reinfection, it may still attenuate the pathology due to rein-
sion, which may provide clues to future scenarios for the cur- fection (IEP) and/or reduce transmissibility or infectiousness
rent pandemic. There are four human coronaviruses (HCoVs) (IEI). Indeed, experimental reexposure studies on endemic
that circulate endemically around the globe; they cause only HCoVs provide evidence that the three IE’s do not wane at
mild symptoms and are not a significant public health bur- the same rate (8, 9). Callow’s experimental study (8) shows
den (4). Another two HCoV strains, SARS CoV-1 and MERS, that reinfection is possible within one year (relatively short
emerged in recent decades and have higher case fatality ra- IES); however, upon reinfection symptoms are mild (high IEP)
tios (CFRs) and infection fatality ratios (IFRs) than COVID- and the virus is cleared more quickly (moderate IEI). Details
19 but were contained and never spread widely (5, 6). on the derivation of the model can be found in section 2 of
We propose a model to explore the potential changes in the supplementary materials (SM).
both transmission and disease severity of emerging HCoVs We reanalyze a detailed dataset that estimates age-spe-
through the transition to endemicity. We focus on CoV-2 and cific seroprevalence based on both IgM (acute response) and
discuss how the conclusions would differ for emerging coro- IgG (long-term memory) against all four circulating HCoVs
naviruses more akin to SARS and MERS. Our hypothesis is in children and adults (10) to estimate parameter ranges for
that all HCoVs elicit immunity with similar characteristics, transmission and waning of immunity (see Fig. 1A). The rapid
and the current acute public health problem is a consequence rise in both IgM and IgG seroprevalence indicates that pri-
of epidemic emergence into an immunologically naïve mary infection with all four endemic HCoV strains happens
First release: 12 January 2021 www.sciencemag.org (Page numbers not final at time of first release) 1
early in life, and our analysis of these data gives us an esti- dynamics for higher R0, which are more akin to those of CoV-
mate for the mean age of primary infection (MAPI) between 2 in the absence of control measures. If transmission is high,
3.4 and 5.1 years, with almost everyone infected by age 15 (see the model predicts a high case load and death rate in earlier
SM section 1 for details). The absence of detectable IgM titers years following emergence (Fig. 2 and fig. S5). In Fig. 2B we
in any individual over the age of 15 years suggests reinfec- see that, as might be expected, longer lasting sterilizing im-
tions of adults causes a recall response, indicating that while munity slows down the transition to endemicity.
CoV specific immunity may wane it is not lost. Whether im- These results are robust to a more biologically realistic
munity would wane to naïve levels in the absence of high distribution for the duration of sterilizing immunity and the
pathogen circulation remains an open question. possibility that the generation of protective immunity re-
For most people to be infected so early in life—younger quires more than one infection (see SM section 3 and figs. S5
even than measles in the pre-vaccine era—the attack rate to S9).
must exceed transmission from primary infections alone. The Slowing down the epidemic through social distancing
model shows a high attack rate can arise from a combination measures that reduce R0 to close to one flattens the curve,
of high transmissibility from primary infections (i.e., high R0), thus delaying infections and preventing most deaths from
waning of sterilizing immunity and significant transmission happening early on, affording critical time for the develop-
First release: 12 January 2021 www.sciencemag.org (Page numbers not final at time of first release) 2
detectable IgM titers in adults. Strain-specific virulence fac- to be maintained in younger individuals. During the transi-
tors, such as the shared cellular receptor, ACE-2, to which tion to endemicity, primary CoV-2 infections will frequently
CoV-1, CoV-2 and the endemic strain NL63 all bind (13–16), occur in older individuals, and we need to determine if im-
may affect the CFR during the emergence phase but have lit- munity induced by infection or vaccination in adulthood is
tle impact on the severity of disease in the endemic phase. similar to that produced by natural infections in childhood.
Because the four endemic HCoVs have been globally circulat- Thus far, there have been few reinfections reported with CoV-
ing for a long time and almost everyone is infected at a young 2, and disease severity has varied (24); the only population-
age, we cannot ascertain how much pathology would result level study of reinfection of which we are aware estimates a
from a primary or even secondary case of any of these in an low rate of reinfection within the first six months after pri-
elderly or otherwise vulnerable person. mary infection and mild disease upon reinfection (25), but
The key insights come from how our model explicitly in- further analysis and monitoring are vital.
corporates different components of immunological protec- The findings presented here suggest that using symptoms
tion with respect to susceptibility, pathology and infectivity as a surveillance tool to curb the virus’s spread will become
(IES, IEP and IEI) and their different rates of waning. In our more difficult, as milder reinfections increasingly contribute
analysis we hypothesized that these components of immunity
First release: 12 January 2021 www.sciencemag.org (Page numbers not final at time of first release) 3
and seasonal influenza, which share similar age distributions study. J. Immunol. 186, 7264–7268 (2011). doi:10.4049/jimmunol.0903490
and immune responses. Medline
20. N. Le Bert, A. T. Tan, K. Kunasegaran, C. Y. L. Tham, M. Hafezi, A. Chia, M. H. Y.
REFERENCES AND NOTES Chng, M. Lin, N. Tan, M. Linster, W. N. Chia, M. I.-C. Chen, L.-F. Wang, E. E. Ooi, S.
1. S. Cobey, Modeling infectious disease dynamics. Science 368, 713–714 (2020). Kalimuddin, P. A. Tambyah, J. G.-H. Low, Y.-J. Tan, A. Bertoletti, SARS-CoV-2-
doi:10.1126/science.abb5659 Medline specific T cell immunity in cases of COVID-19 and SARS, and uninfected controls.
2. H. E. Randolph, L. B. Barreiro, Herd immunity: Understanding COVID-19. Immunity Nature 584, 457–462 (2020). doi:10.1038/s41586-020-2550-z Medline
52, 737–741 (2020). doi:10.1016/j.immuni.2020.04.012 Medline 21. R. Channappanavar, C. Fett, J. Zhao, D. K. Meyerholz, S. Perlman, Virus-specific
3. C. Fraser, S. Riley, R. M. Anderson, N. M. Ferguson, Factors that make an infectious memory CD8 T cells provide substantial protection from lethal severe acute
disease outbreak controllable. Proc. Natl. Acad. Sci. U.S.A. 101, 6146–6151 respiratory syndrome coronavirus infection. J. Virol. 88, 11034–11044 (2014).
(2004). doi:10.1073/pnas.0307506101 Medline doi:10.1128/JVI.01505-14 Medline
4. S. Su, G. Wong, W. Shi, J. Liu, A. C. K. Lai, J. Zhou, W. Liu, Y. Bi, G. F. Gao, 22. R. Eguia, K. H. D. Crawford, T. Stevens-Ayers, L. Kelnhofer-Millevolte, A. L.
Epidemiology, genetic recombination, and pathogenesis of coronaviruses. Trends Greninger, J. A. Englund, M. J. Boeckh, J. D. Bloom, A human coronavirus evolves
Microbiol. 24, 490–502 (2016). doi:10.1016/j.tim.2016.03.003 Medline antigenically to escape antibody immunity. bioRxiv 2020.12.17.423313 [Preprint].
5. M. Salamatbakhsh, K. Mobaraki, S. Sadeghimohammadi, J. Ahmadzadeh, The 18 December 2020. https://doi.org/10.1101/2020.12.17.423313.
global burden of premature mortality due to the Middle East respiratory syndrome 23. M. Voysey, S. A. C. Clemens, S. A. Madhi, L. Y. Weckx, P. M. Folegatti, P. K. Aley, B.
(MERS) using standard expected years of life lost, 2012 to 2019. BMC Public Angus, V. L. Baillie, S. L. Barnabas, Q. E. Bhorat, S. Bibi, C. Briner, P. Cicconi, A. M.
Health 19, 1523 (2019). doi:10.1186/s12889-019-7899-2 Medline Collins, R. Colin-Jones, C. L. Cutland, T. C. Darton, K. Dheda, C. J. A. Duncan, K. R.
First release: 12 January 2021 www.sciencemag.org (Page numbers not final at time of first release) 4
doi:10.3201/eid2607.200282 Medline SUPPLEMENTARY MATERIALS
32. S. M. Kissler, C. Tedijanto, E. Goldstein, Y. H. Grad, M. Lipsitch, Projecting the science.sciencemag.org/cgi/content/full/science.abe6522/DC1
transmission dynamics of SARS-CoV-2 through the postpandemic period. Supplementary Text
Science 368, 860–868 (2020). doi:10.1126/science.abb5793 Medline Figs. S1 to S13
33. M. Chan-Yeung, R.-H. Xu, SARS: Epidemiology. Respirology 8 (suppl.), S9–S14 References (36–42)
(2003). doi:10.1046/j.1440-1843.2003.00518.x Medline MDAR Reproducibility Checklist
34. R. Verity, L. C. Okell, I. Dorigatti, P. Winskill, C. Whittaker, N. Imai, G. Cuomo-
Dannenburg, H. Thompson, P. G. T. Walker, H. Fu, A. Dighe, J. T. Griffin, M. 4 September 2020; resubmitted 10 November 2020
Baguelin, S. Bhatia, A. Boonyasiri, A. Cori, Z. Cucunubá, R. FitzJohn, K. Gaythorpe, Accepted 7 January 2021
W. Green, A. Hamlet, W. Hinsley, D. Laydon, G. Nedjati-Gilani, S. Riley, S. van Published online 12 January 2021
Elsland, E. Volz, H. Wang, Y. Wang, X. Xi, C. A. Donnelly, A. C. Ghani, N. M. 10.1126/science.abe6522
Ferguson, Estimates of the severity of coronavirus disease 2019: A model-based
analysis. Lancet Infect. Dis. 20, 669–677 (2020). doi:10.1016/S1473-
3099(20)30243-7 Medline
35. J. S. Lavine, COVID Immunity and Endemicity Code, Version 1.0.0, Zenodo (2020);
https://doi.org/10.5281/zenodo.4390505.
36. H. W. Schroeder Jr., L. Cavacini, Structure and function of immunoglobulins. J.
Allergy Clin. Immunol. 125 (suppl. 2), S41–S52 (2010).
First release: 12 January 2021 www.sciencemag.org (Page numbers not final at time of first release) 5
Table 1. Characteristics of coronavirus-immune interactions and relevant parameter ranges.
Characteristic and symbol Estimates from Value Citations
literature (range)
Primary infectious period ≥5.6 days 9 days (8)
(1/γ) ∼10 days (30)
First release: 12 January 2021 www.sciencemag.org (Page numbers not final at time of first release) 6
Downloaded from http://science.sciencemag.org/ on January 13, 2021
Fig. 1. A low mean age of primary infection suggests partially transmissible reinfections are common.
(A) Mean proportion seropositive for IgG (green, top lines) and IgM (purple, bottom lines) against the four
endemic HCoV strains [dots connected by dashed lines; vertical lines represent the 95% CI; data from
Zhou et al. (10)]. The mean age of primary infection (MAPI) based on IgM data with 95% CI is shown in text
inside each panel (see SM for details). (B) MAPI as a function of waning of sterilizing immunity (ω, y axis)
and transmissibility of reinfections (ρ, x axis). The MAPI was calculated from the equilibrium dynamics of
the model shown in fig. S1 and supplementary equations 3 to 9 with a plausible basic reproductive number
(R0 = 5) and 0 < ω < 2 and 0 < ρ < 1. See SM section 2.1 for details. The white band in indicates the plausible
combinations of values of ρ and ω consistent with the MAPI for HCoVs estimated in (A). [See fig. S1 for
parallel figures calculated at extreme plausible values for R0 (i.e., R0 = 2 and R0 = 10).]
First release: 12 January 2021 www.sciencemag.org (Page numbers not final at time of first release) 7
Downloaded from http://science.sciencemag.org/ on January 13, 2021
Fig. 2. The time scale of the transition from epidemic to endemic dynamics for emerging coronaviruses
depends on R0 and the rate of immune waning. Transition from epidemic to endemic dynamics for emerging
HCoVs, simulated from an extension of the model presented in fig. S1 that includes age structure. Demographic
characteristics (age distribution, birth, and age-specific death rates) are taken from the United States, and
seasonality is incorporated via a sinusoidal forcing function (see SM section 2.2). Weak social distancing is
approximated by R0 = 2. (See figs. S9 to S11 for strong social distancing results, R0 < 1.5.) (A) Daily number of
new infections (black line, calculations in SM section 2.3). An initial peak is followed by a low-incidence endemic
state (years 5 to 10 shown in the inset). A higher R0 results in a larger and faster initial epidemic and more rapid
transition to endemic dynamics. The proportion of primary cases in different age groups changes over time
(plotted in different colors), and the transition from epidemic to endemic dynamics results in primary cases
being restricted to younger age groups. Parameters for simulations: ω = 1 and ρ = 0.7. (B) Time for the average
IFR (6-month moving average) to fall to 0.001, the IFR associated with seasonal influenza. Gray areas represent
simulations where the IFR did not reach 0.001 within 30 years. The time to IFR = 0.001 decreases as the
transmissibility (R0) increases and the duration of sterilizing immunity becomes shorter. Results are shown for
ρ = 0.7. See SM section 2.3 and figs. S4 to S7 for sensitivity analyses and model specifications.
First release: 12 January 2021 www.sciencemag.org (Page numbers not final at time of first release) 8
Downloaded from http://science.sciencemag.org/ on January 13, 2021
Fig. 3. The overall infection fatality ratio (IFR) of emerging coronaviruses once they reach endemicity is
strongly influenced by the IFR of young children in the initial epidemic. The age dependence of the IFR
determines how the overall IFR changes during the transition from epidemic to endemic dynamics for emerging
CoVs. (A) Age dependence of the IFRs for the three emerging CoVs. Primary infections with MERS and CoV-1
are consistently symptomatic and the IFR and CFR are therefore assumed to be the same. CoV-1 and CoV-2
have J shaped profiles, with a monotonic increase in IFR with age. The age-specific IFR for MERS is U shaped,
with high mortality in the young and old age groups. Details of the statistical smoothing are described in SM
section 6. (B) The overall IFR changes during the transition to endemic dynamics. These calculations assume
deaths due to reinfections are negligible. We relax this assumption to allow for a slower build-up of immunity
and possible death due to secondary infection in figs. S5 to S9 and show the qualitative results do not change.
First release: 12 January 2021 www.sciencemag.org (Page numbers not final at time of first release) 9
Immunological characteristics govern the transition of COVID-19 to endemicity
Jennie S. Lavine, Ottar N. Bjornstad and Rustom Antia
REFERENCES This article cites 36 articles, 9 of which you can access for free
http://science.sciencemag.org/content/early/2021/01/11/science.abe6522#BIBL
PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions
Science (print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the Advancement of
Science, 1200 New York Avenue NW, Washington, DC 20005. The title Science is a registered trademark of AAAS.
Copyright © 2021 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of Science.
No claim to original U.S. Government Works. Distributed under a Creative Commons Attribution License 4.0 (CC BY).