Cohort
Within the previously described southwest household cohort 13,14, we studied 28 families, with 61 adults (median age 44.8 years, IQR 41.4-50.0) and 50 children (median age 10.4 years, IQR 7.2-13.5). We selected the families with at least one child or adolescent and one COVID-19 case either assessed by polymerase chain reaction (PCR) or by seroconversion (see Methods for inclusion criteria and definition of seropositivity). Among those, 31 adults (50.8%) and 27 children (54.0%) were seropositive at the time of study inclusion. All infected individuals presented a mild disease course, and in 6.5% of adults and 33.3% of children, infection was classified asymptomatic. None of the participants was hospitalized. As shown in Table 1, the most frequent symptoms were fever, cough and dysgeusia. Sampling was performed during national contact restriction (lockdown) in the summer of 2020, and afterwards in spring 2021, corresponding to 4 and 12 months after infection, approximately.
Table 1
Demographics and key information on the study participants. See methods for definition of how samples were defined as being seropositive, asymptomatic or symptomatic. Abbreviations: BMI: Body Mass Index; IQR: Interquartile Range; NA: not applicable. *vaccinated subjects were excluded from the analysis in T2. **three seropositive children and $five seropositive adults from T1 were lost to follow-up.
| Timepoint 1 (4 months) | Timepoint 2 (12 months) |
Number of participants by age group (n) | Adults (61) | Children (50) | Adults (51) | Children (40) |
Median age (IQR) | 44.8 (41.4-50.0) | 10.4 (7.2-13.5) | 46.0 (41.6-51.1) | 10.7 (7.1-13.4) |
Number of females (%) | 29 (47.5) | 26 (52.0) | 24 (47.1) | 19 (47.5) |
BMI (IQR) | 25.4 (22.2-29.0) | 15.8 (14.5-19.4) | 26.3 (23.0-29.6) | 15.7 (14.5-19.4) |
Number of seropositive participants (%) | 31 (50.8) | 27 (54.0) | 16/41 (39.0)* | 22 (55.0) |
- asymptomatic (% of seropositive) | 2 (6.5) | 9 (33.3) | NA | NA |
Persistent seropositivity at T2 (% of seropositive at T1) | NA | NA | 16/22 (77.3)*$ | 22/24 (91.7) ** |
Symptoms at disease onset (of seropositive participants) | | | | |
- fever (%) | 18/31 (58.1) | 12/27 (44.4) | NA | NA |
- cough (%) | 16/31 (51.6) | 6/27 (22.2) | NA | NA |
- dysgeusia (%) | 16/31 (51.6) | 4/27 (14.8) | NA | NA |
- diarrhea (%) | 7/31 (22.6) | 2/27 (7.4) | NA | NA |
Median (IQR) days from positive PCR test result (or symptom onset) to the timepoint | 114.0 (109.3-120.0) | 363.5 (361.8-372) |
Vaccinated (%) | NA | NA | 10/51(19.6)* | 0 |
Number of households | 28 | 23 |
Participants with chronic disease: hypertension, diabetes mellitus, dyslipidemia (%) | 8/61 (13.1) | 0 | 7/51 (13.7) | 0 |
B |
Decrease in antibody titer but progressive neutralizing potency after SARS-CoV-2 mild/asymptomatic infection in children and adults.
We studied the specific antibody responses (IgG, IgA) to the S1 subunit of the spike protein (Fig. 1A, 1B), the receptor-binding domain (RBD) (Fig. S1A) and NCP (Fig. S1B). In line with previous analysis of this cohort 13,14, and of other cohorts 4, 4 months after infection serum titers of specific S1 antibodies were higher in children compared to adults (Fig. 1A). The difference between the two groups disappeared 12 months after infection (Fig. 1A). In contrast, the specific IgG antibody response to RBD protein was similar between adults and children over the observation period (Fig. S1A). Notably, a significantly higher response to NCP was observed in children 4 months after infection, but rapidly decreased, becoming significantly lower in children compared to adults 12 months after infection (Fig. S1B). For IgA, S1 specific responses did not differ between children and adults in the observation period (Fig. 1B). Overall, we observed a progressive decrease in the S1 (Fig. 1A), RBD (Fig. S1A) and NCP specific (Fig S1B) IgG antibodies in both adults and children. The extent of the decrease in S1 and RBD specific antibodies between 4 months and 12 months after infection was similar in children and adults (Fig. 1A and S1A), while children showed a much faster decrease in NCP specific antibodies (Fig. S1B). Interestingly, a progressive increase of S1 specific IgA antibodies was observed in children but not in the adult population (Fig. 1B). These data point to progressive class-switch over time in B cells of the pediatric cohort. With respect to the functionality of the antibodies, we found neutralizing antibodies in both cohorts; however, children showed a higher neutralization capacity than adults (Fig. 1C). Neutralizing antibodies persisted in both cohorts over time, with a minimal decrease in the adult population (Fig. 1C). Neutralizing antibodies mostly bind to RBD, with variable activity, depending on the epitope 15. Therefore, we used the neutralization potency index (NPI), defined as the relative proportion of plasma neutralization capacity in respect to the specific RBD antibody level 11,16. The NPI (Fig. 1D) progressively increased in children and adults. Importantly, the differences in serum S1 and RBD specific antibody levels between adults and children were not due to an increase in serum immunoglobulin levels (Fig. S2A, S2B).
Sera of adults and children infected with the Wuhan wild type form of the virus (WT) efficiently neutralized the variants of concern (VOC) alpha, beta and gamma 4 months after infection (Fig. 2A) even though, neutralization potency gradually decreased from alpha to beta VOCs (Fig. 2A). The neutralization capacity remained stable over the observation period (Fig. 2B). In contrast, the neutralization breadth index (NBI), calculated as neutralization capacity against the alpha, beta and gamma specific VOCs in relation to the neutralization of the WT of virus 11,16, showed a significant increase over time (Fig. 2C). These data support the idea that in both children and adults, post-infection sera can recognize VOCs, and such recognition of VOCs can improve over time thanks to a progressive maturation of the immune response in recovered individuals.
Progressive increase in plasmablasts and IgA- positive cells over 12 months after infection.
The distribution and frequency of B cell subpopulations in children and adults show fundamental differences 17 that can influence the individual immune response to infection. Extensive B cell phenotyping of SARS-CoV-2 seropositive adults and children 4- and 12-months post infection (Fig. 3A-C, S3A) revealed higher percentages of B cells in children compared to adults (Fig. S3B). Within the B cells, children had higher transitional and naïve B cells, but lower class switched memory cells (Fig. 3A), reflecting a less experienced immune system. Within the class-switched B cells, IgG1 and IgG3 were more frequent and IgG2, IgG4, IgA1, and IgA2 were less frequent in children compared to adults (Fig. 3A). Over the 8-month observation period most of the B cell subpopulations remained stable (Fig. 3B, 3C), with the exception of an increased proportion of plasmablasts and IgA2 positive B cells in adults (Fig. 3B). In children, we observed an increase in plasmablasts, switched memory cells, marginal zone like B cells and IgA2 positive B cells (Fig. 3C). These data may support the concept of an ongoing B cell activation in infected individuals, with appearance of more mature B cells over time. As samples were acquired during a period of social distancing, we propose that the progressive differentiation of B cells is by in large due to the original infection, and not caused by re-infections in the examined individuals.
Persistent low frequency of circulating S1 specific memory B cells in children.
The frequency of circulating SARS-CoV-2-specific B cells was assessed in our cohort by S1 or RBD tetramer staining (Fig. 3D, S4A-B). In contrast to the serum S1-specific antibody titer, S1-specific B cells were less frequent in children compared to adults both at 4 and 12 months after infection (Fig. 3E). RBD-specific B cells were less frequent in children 4 months after infection, but the frequency was similar 12 months after infection (Fig. S4B). S1- and RBD-specific B cells persisted at a similar frequency during our observation period (Fig. 3E, S4B). While the frequency of tetramer specific B cells within the B cell pool was stable in children and adults, we observed a progressive maturation of children’s B cell response as indicated by the acquisition of the marker CD27 in the S1- and RBD-specific B cells (Fig. 3G, 3H, S4C). This indicates that in children a longer time period was needed for S1-specific B cells to acquire a memory B cell phenotype as indicated by CD27 expression. Stimulation via TLR9 favors blasting and antibody production from memory B cells, while naïve B cells are only poorly stimulated and do not undergo class switch 18. We studied the ability of S1- and NCP-specific memory B cells to produce specific antibodies upon TLR9 stimulation (Fig. 3F, S5). In vitro activation resulted in secretion of S1-specific antibodies in more than 80% of adults with a history of SARS-CoV-2 infection (Fig. 3F). In line with the lower frequency of S1-specific B cells in children, fewer children showed S1 antibodies in the supernatant of stimulated B cell cultures (Fig. 3F). This assay was highly specific, as we did not find any S1- or NCP-specific antibody production in vitro in non-infected individuals (Fig. S5A). The frequency of S1-specific memory B cells detected by flow cytometry correlated with immunoglobulin concentrations in the culture supernatant in adults (Fig. S5B), in line with the model that detection of specific B cells by tetramer staining corresponded to the presence of memory cells that can be rapidly re-activated. In children, the frequency of S1-specific B memory cells correlated with antibody secretion only at 12 months after infection (Fig. S5B). This may be due to the low frequency of S1-specific memory B cells 4 months after infection, or to their immature phenotype. In the supernatant of stimulated B cell culture, we also tested the presence of S1 IgA antibodies (Fig. S5C) and NCP IgG antibodies (Fig. S5D). The number of adults carrying S1 IgA memory cells that could be reactivated decreased over the observation period, while S1 IgA memory cells increased in children (Fig. S5C), similarly to the increase in serum S1-specific IgA. In contrast, the NCP response remained stable (Fig. S5D).
Durable T cell response in adults and children 12 months after SARS-CoV-2 infection.
After infection with SARS-CoV-2, virus-specific CD8+ and CD4+ T cells can be studied by tetramer staining in peripheral blood 19–22. The detection of specific T cells by tetramers is limited by the restriction of the HLA-type. As our cohort was not selected by a specific HLA type, we assessed the presence of SARS-CoV-2 specific CD4+ and CD8+ T cells by re-stimulation of T cells with a SARS-CoV-2 peptide mix in vitro, followed by analysis of activation markers CD69 and CD137 (Fig. 4A-B, Fig. S6A-C). This activation-induced quantification assay allowed detection of SARS-CoV-2 S1-reactive T cells with very high sensitivity, as only one child out of 48 seropositive participants had T cell numbers below the cut-off threshold (see methods for details). The assay was also highly specific, as S1-reactive T cells were detected in only one non-infected, seronegative participant out of 46 and in none of 25 subjects whose material was preserved before 2019 (Fig. S6D and ROC curve in Fig. S6E,F). Specific CD4+ and CD8+ T cell responses were detectable both in children and adults 4 and 12 months after infection, albeit at a lower frequency in children (Fig. 4C,D). The frequency of specifically activated CD4+ cells remained constant over the 8-month observation period in adults, while a significant reduction of specific CD4+ T cell frequencies was observed in children (Fig. 4C). CD8+ T cells were detectable at a lower frequency in children compared to adults as well, but persisted in both age groups over time (Fig. 4D). Similar results were found after stimulation with a peptide mix covering C-terminal part of S protein with homology with endemic coronaviruses (SARS-CoV-2 S2-peptide mix, data not shown). In contrast, when T cells were stimulated with a S1- and S2-peptide mix of four different common coronaviruses (“pan-corona”) (Fig. S7A,B), T cell activation was similar between children and adults. Thus, a stronger CD4+ and CD8+ T cell responses in adults as compared to children is a specific feature of infections by SARS-CoV-2.
Similar distribution of effector memory T cells in children and adults, but altered cytokine profile in children.
As for the B cell compartment, the T cells in children were largely naïve, and showed low frequencies of central and effector memory T cell subsets (Fig. S8A,B), in line with previous observations 23,24. We analyzed the phenotype of SARS-CoV-2 specific T cells in both cohorts. Despite the differences in the bulk CD4+ T cells, SARS-CoV-2 specific CD4+ T cells showed a similar distribution of naïve, effector memory and terminal effector memory cells between children and adults, only CD4+ central memory T cells were slightly lower in frequency in children 4 months after exposure (Fig. 5A-C). SARS-CoV-2 specific CD8+ T cells showed a similar distribution of the naïve and memory subpopulations in adults and children (Fig. 5D-G). The large majority of S1-specific CD4+ and CD8+ T cells were effector memory T cells, as expected in a virus specific immune response 24. The phenotype of specific CD4+ and CD8+ T cells remained constant 4 months and 12 months post infection in adults (Fig. 5B, 5F). In children, we observed a progressive decrease of effector memory CD4+ (Fig. 5C) and CD8+ (Fig. 5G), and a progressive increase of CD8+ terminal effector T cells (Fig. 5G).
To evaluate the functionality of SARS-CoV-2 specific T cells, we analyzed cytokine secretion in the supernatant of re-stimulated T cells (with S1-specific peptide mix) from patients 4 months post infection. We assessed Th1 cytokines (IFN-γ and TNF-α), that are important for virus clearance, pro-inflammatory cytokines (IL-1β, IL-17a) and cytokines regulating inflammatory reactions (IL-10, IL-13). IFN-γ was the primary cytokine in SARS-CoV-2 specific T cells. Comparison of the cytokine profiles in adults and children showed reduced INF-γ, TNF-α, IL-10, IL-17a and IL-1β in T cells from children versus adults, but similar levels of IL-13 (Fig. 5H). The reduced inflammatory cytokine secretion may contribute to the reduced symptoms observed in children compared to adults. While INF-γ, TNF-α and IL-10 levels correlated with the percentage of S1 specific T cells among PBMC; this was not the case for the inflammatory cytokines IL-1β and IL-17a (Fig. S9). Notably, children´s T cells were able to secrete cytokines comparably to adults in response to pan-Corona antigens (Fig. S10).
Persistence of cellular immune response and reduction of serological response up to 12 months post infection.
To gain a global view of the persistence of SARS-CoV-2 specific immunity and the progressive changes in the B and T cell compartment, we generated a heat map representing the log2 fold of change of the ratio between the parameter measured 12 and 4 months after the infection (Fig. 6A). This allowed for comparison of the changes in each parameter on a unified scale. This global analysis revealed that while specific humoral immunity (characterized by antibody response against S1, RBD and NCP) decreased over time, cellular immunity, such as S1 and RBD specific B cells and S1 specific CD4+ T cells, was mostly stable. However, there were exceptions: S1 specific IgA responses in serum, increased in children and some adults; S1 specific CD8+ T cells decreased in several individuals, primarily in children. Both the specific B and T cell compartments showed dynamic maturation over 12 months, especially in children, with acquisition of CD27 on specific B cells, and the acquisition of the terminal effector memory phenotype in specific CD4+ and CD8+ T cells in both children and adults. Hence, reduction of humoral immunity paralleled the maturation of the cellular immunity even in the absence of re-infection or vaccination. This concept is further supported by progressive increase in peripheral plasmablasts, marginal-zone like cells and IgA2+ class switched B cells. Additionally, the dynamics of the immune response in children and adults was mostly similar, with the exception of progressive IgA class switch and later expression of CD27 in specific S1 cells.
Looking for interdependence, we performed a matrix analysis using the Spearman test for significant correlations between the parameters 4 and 12 months after infection. Four months after infection (Fig. 6B), neutralizing antibodies positively correlated with the specific humoral response as well as the presence and maturation status of SARS-CoV-2 specific B cells in children. The presence of a specific antibody response and of specific B cells positively correlated with total plasmablasts and switched memory cells, and inversely correlated with naïve cells. These data suggested that the reduced cellular immunity might result from immaturity of the immune system in children. In adults, all specific antibody responses correlated with each other, as well as the specific B cell responses 4 months after infection. This suggested an independent development of plasmablast and memory cells in the adult immune system. Interestingly, class switched CD27-negative unconventional memory B cells, which contain B cells that develop at extrafollicular regions 25 inversely correlated with specific antibody responses and the memory phenotype (CD27+) of specific B cells in adults. As described previously, the extrafollicular response is expanded in COVID-19 12, and extrafollicular cell fate favors plasmablast development with low affinity at the expense of the germinal center dependent maturation.
Twelve months after infection (Fig. 6C), correlation plots between children and adults converged, with a strong correlation between specific antibody as well as B cell responses. Notably, in children, S1 specific CD4+ T cells correlated with S1 and RBD specific memory B cells and their maturation into CD27+ memory cells, in line with the known role of CD4 T cells in B cell development. At 12 months after infection, we were able to measure RBD specific IgG in the saliva of the participants of our cohort. We found that in children, the presence of RBD IgG in the saliva positively correlated with serum RBD and S1 IgG as well as their neutralizing capacity, and it inversely correlated with unconventional memory cells in children. In the adult cohort, RBD IgG in the saliva also correlated with specific serum antibody responses, but also with specific B and CD4+ T cells. These data suggested that high serum immunoglobulin concentrations may correspond to protective levels of virus-specific IgG at mucosal surfaces, which represent entry sites for SARS-CoV-2, both in children and adults. In the adult population, we found a correlation between the specific serum antibody responses to S1- and RBD-specific memory B cells (Fig. 6B, 6C) both 4 and 12 months after infection, suggesting a parallel development of plasma cells and memory cells. On the contrary, in children, the serum antibody response did not significantly correlate with the presence of S1- or RBD-specific memory cells at 4 and 12 months (Fig. 6B, 6C). These data suggested a predominant plasmablast response in children.