The dynamics of humoral immune responses following SARS-CoV-2 infection and the potential for reinfection


Serological decline after MERS CoV and SARS CoV infection

A few studies have assessed antibody titres to MERS CoV and SARS CoV in the months and years following primary infection. Robust immune responses with long-lived (>1 year) functional antibodies were seen following severe MERS CoV infections or in those people with prolonged virus shedding. This was also observed in a small study of MERS CoV infections, where neutralizing antibodies were detectable in six (86 %) out of seven persons who had previously had severe MERS (including five with pneumonia) for at least 34 months after infection.

However, in this small group there was evidence of antibody waning; 4/7 showed 4- to 16-fold reduction in nucleocapsid-binding titres and 4/7 show a twofold reduction in neutralizing titres over 34 months, with 4/7 assessed as having a low neutralizing titres throughout. After mild or asymptomatic MERS CoV infections, antibody responses were either limited or rapidly declined. Although the numbers are small, no neutralizing antibody response was seen in 4/6 and 3/6 mild MERS CoV infections for some, not even immediately after infection.

In a separate study of 280 contacts of 26 confirmed MERS CoV index cases, 12 contacts likely to have been infected were identified. Seven out of 12 contacts sampled within 4–14 days of index contact were virus genome-positive by RT-PCR but serologically negative (actively infected), whereas 5/7 were virus genome-negative, but had detectable binding and neutralizing antibody titres (infected and recovered).

Similarly, although SARS CoV was largely associated with symptomatic disease, antibodies decline over time. In a 3-year follow-up of hospitalized SARS CoV patients, SARS CoV IgG-binding titres were undetectable in 19.4 % of people by 30 months post-infection and neutralizing titres were undetectable in 11.1 % of people at this time. Consistent with this observation, a study of 98 SARS patients over 2 years showed that all had detectable antibody binding titres over 2 years, but that, in a subset, titres declined over this period.

Eighteen individuals with neutralizing antibodies had titres that peaked on day 30 and then decayed gradually so that by 2 years 1/18 had no detectable neutralizing antibodies, and the remaining patients had low antibody titres close to background levels. Similarly, in a study following 176 previously SARS CoV-infected people, the enzyme-linked immunosorbent assay (ELISA) optical densities (ODs) that indicate antibody titre reduced by 33 % within 1 year, 46 % by 2 years and ~75 % by 3 years.

Nevertheless, other long-term follow-up studies of SARS CoV showed that although antibody titres decline over over 2 and 3 years, neutralizing activity was present in 89 % (17/19) of the recovered patients at 36 months, although the ability of sera to neutralize virus declined from 96 % inhibition at month 3 to 48 % at month 36. Although antibody titres to SARS CoV can be detected in people 12 years after infection, over 70 % the people studied (n=20) had extremely low titres, and so at 3 and 12 years post-infection SARS CoV antibody titres are likely to be very limited for virus neutralization, with little or no ability to protect a person from reinfection. However, this requires experimental determination.

Although limited in size, studies of MERS and SARS CoV indicate that total binding antibodies and neutralizing antibodies decrease to a level where by 2–3 years everyone previously infected will have minimal detectable antibody response, but those suffering more severe disease have the highest titre antibody responses for longer. Although the time-dependent decay of neutralizing antibody titres implies a lack of protection from reinfection by MERS and SARS CoV, this cannot be concluded unequivocally, due to lack of epidemic spread allowing reinfection. It is, however, suggestive of the potential for a population-level reduction in protection from reinfection by epidemic CoVs over a short period of time, dependent in some on initial disease severity.

One indication of the strength of immune protection from coronavirus infection is to consider what is known for the endemic seasonal CoVs, namely the genetically related alphacoronaviruses, NL63 and 229E, and the genetically related betacoronaviruses, HKU1 and OC43. There is some evidence for antigenic cross-protection between the human CoVs in the same genetic group. A cross-sectional seroprevalence study for seasonal human alphacoronaviruses NL63 or 229E showed that 75 and 65 % of children in the age group 2.5–3.5 years are seropositive for NL63 and 229E, respectively, and most children are seropositive by 6 years.

In adults, respiratory infection by human seasonal CoVs accounted for 22 % (43/195) and 25 % of acute respiratory illness. Therefore, the ability of human seasonal coronaviruses to infect adults who have likely been infected as children can be accounted for by either virus escape from neutralization (drift), reinfection with a heterologous CoV of a different genotype (alpha- followed by betacoronavirus infections, or vice versa) due to lack of cross-protective antibodies, or reinfection with homologous coronavirus due to sub-protective or waning antibody responses.

The lack of extensive time-resolved virus genetic data linked to serology studies of extant and historic strains of the four seasonal human coronaviruses makes the contribution of virus genetic drift to escape from pre-existing protective immune response difficult to judge. The evolutionary genetics of coronaviruses, especially in animals, however, shows considerable genetic diversity within coronavirus species, largely driven by high rates of substitution and recombination.

For infectious bronchitis virus (IBV) of chickens, the existence of many serotypes, with little cross-protective immunity between them, is attributed to both small and large numbers of amino acids substitutions in the spike gene, supporting the view that immune escape through genetic drift in animal coronavirus is common. Work on the human endemic coronavirus OC43 suggests that genetic drift is similarly important, with considerable genetic diversity in the spike gene suggesting circulation of distinct OC43 variants.

Further, genetic drift mapping to sugar-binding domains in S protein of CoV OC43 suggests that drift may contribute to persistence of this genotype in the human population. Similar studies on other endemic CoV genotypes are lacking and whether the use of different cell receptors constrains genetic variation in the spike gene of some human coronavirus more than others is not known, but infection due to immune escape through genetic drift seems important for coronaviruses. Waning of the neutralizing antibody response also seems to contribute to coronavirus reinfection. Whether coronaviruses encode specific proteins whose action is to limit the adaptive immune response or the spike protein is poor at initiating long-lived plasma cells is not known, but potential consequences of declining humoral immunity can be observed.

Reinfection by seasonal human coronaviruses in the community:

A small number of studies have attempted to detect reinfection by endemic CoVs in the community. In a cohort study of community-acquired and childhood pneumonia admissions to hospital in Kenya, reinfections by human coronavirus NL63 were detected over a 6-month period (December–May 2010) in 46 out of 163 patients (28 %). Most reinfections resulted in low virus titres and decreased disease. However, for a small number (11 %), reinfection resulted in higher virus shedding compared to the previous infection, with the caveat that the peak viral genome load in the first infection could have been missed in the sampling window.

Reinfection by seasonal human coronaviruses in controlled human infection models (CHIMs)

Another way to distinguish between infection due to virus escape from neutralization, or infection in the presence of sub-protective antibody responses, is to attempt to experimentally infect adult volunteers with seasonal human coronavirus, either in the presence of their pre-existing immunity or by rechallenge with a homologous virus.

Inoculation of healthy adult volunteers with the endemic coronavirus 229E led to infection in 10/15 people and clinical symptoms in 8 of those 10 infected people, even though most must have already experienced 229E infection previously in their lives. All those infected had increased antibody titres within 3 weeks of infection, which declined rapidly by 12 weeks and returned to baseline by 52 weeks. When rechallenged at 1 year, 66 % (6/9) became reinfected, but none developed clinical symptoms.

Thus, SARS-CoV-2 seroconversion occurs on a time course that is consistent with other epidemic CoVs and antibodies to spike RBD were the most reliable for case counting in this study. At 2 weeks post-symptom onset, antibody titres were statistically higher in critical compared to non-critical patients, possibly due to different rates to a maximal antibody response or reflecting similar disease severity observations from MERS-CoV and SARS-CoV patients as described above.

A large-cohort serology study of 285 COVID-19-positive patients of which 262 had a record of disease symptoms, and 39 were severe infections from 3 hospitals in Hubei province, determined the antibody response to nucleoprotein and a peptide from spike protein of SARS-CoV-2. This showed that all patients seroconverted by 17–19 days after symptom onset and that severely ill patients had a significantly higher IgG titre compared to non-severe cases 7–14 days post-symptom onset, but that by 15–21 days there was no difference in the mean antibody titre between these groups. However, a considerable range of antibody titres from low to high was clearly seen in the non-severe group, while IgG titres entered a plateau within 6 days after the first positive samples. Similar results continue to accumulate in other serological studies from China.

It is clear that most people infected with SARS-CoV-2 display an antibody response between 10 and 14 days after infection. In some mild cases, detection of antibodies requires a long time after symptoms, and in a small number of cases, antibodies are not detected at all, at least during the time scale of the reported studies (Fig.1). There is a paucity of information about the longevity of the antibody response to SARS-CoV-2, but it is known that antibodies to other human coronaviruses wane over time, and there are some reports of reinfection with homologous coronaviruses after as little as 80 days. Thus, reinfection of previously mild SARS-CoV-2 cases is a realistic possibility that should be considered in models of a second wave and the post-pandemic era. Obtaining longitudinal serological data where both binding titres and functional neutralization titres are stratified by age groups and previous disease severity status should be undertaken as a matter of urgency.

Further, people with low antibody titres after mild disease should be followed up for evidence of reinfection and recurrent disease by regular clinical monitoring and diagnostic virus detection by RT-PCR. If reinfection is detected, serial viral load and measures of antibody status at the time of reinfection should be established. Detailed human immunology characterization and animal studies will be necessary to determine if prior infection leads to an altered disease course if reinfection occurs.


Author:Paul Kellam ,Wendy Barclay


Please enter your comment!
Please enter your name here