A cohort of 312 individuals was vaccinated with the Pfizer-BioNTech BNT162b2 vaccine from the beginning of January 2021 to May 2021 in Singapore (Supplementary table 1). The median age was 50.9 years (range, 22 to 82) and volunteers were predominantly female (58.3%) and Chinese (72.4%). Participants’ characteristics differed across the different vaccination groups, which reflected vaccine prioritization for healthcare workers and elderly individuals. None of the participants had known or reported history of SARS-CoV-2 infection and were all negative for antibodies against the N protein using the commercial Roche serology assay. At the time of vaccination, Singapore had a low case count which corroborated with low sero-prevalence. Participants were also all negative for antibodies against the N protein using the commercial Roche serology assay. To monitor immune responses, longitudinal blood samples were acquired at baseline corresponding to the day of the first dose, 21 days later at the time of the second dose and up to 180 days post first dose (Fig. 1A).
Antibody response to SARS-CoV-2 mRNA vaccine
All volunteers (n=312) were analyzed for vaccine-induced anti-Spike (S) protein-specific antibody levels and neutralizing efficacy using various assays. The flow cytometry-based assay (SFB) is based on the recognition of SARS-CoV-2 Spike protein stably expressed on the surface of HEK293T cells, allowing the detection of antibodies binding to different epitopes present on the full Spike protein [39, 40]. The majority of volunteers seroconverted after the first dose (95% had higher antibodies than the cohort baseline, and above their individual baseline) (Fig. 1B and Supplementary table 2). After the second dose, all participants but one had developed anti-Spike protein antibodies by day 90. Immunoglobulin isotyping revealed that the proportion of vaccinees with detectable IgM (above both cohort and individual baseline) was >85% at day 21 but dropped to 12% by day 90 (12%) and was negligible by day 180 (Supplementary fig. 1), indicating rapid maturation of the antibody after vaccination. Interestingly, IgG1 dominated the antibody response, followed by IgG3 and IgG2, while IgG4 was barely detected (Fig. S1). However, by day 180, anti-Spike antibody levels had declined in 95% of the vaccinees (Supplementary table 3) on average by 39% (median binding percentage from 40.5% at day 90% to 24.1 at day 180). We also observed a sizeable proportion of low responders (individuals with responses below median cohort response at consecutive time points (37.2% at day 90 and 22.2% at day 180) (Supplementary table 4).
We next profiled antibodies specific to the receptor binding domain (RBD) of the S protein, which is the immunodominant target of anti-SARS-CoV-2 neutralizing antibodies [41] using a commercial assay (Roche S). The first vaccine dose induced antibodies in all but two vaccinees (Fig. 1C and Supplementary table 2). After the second dose, all individuals seroconverted by day 90. However, 36.5% of individuals mounted a poor secondary anti-RBD IgG response (Supplementary table 5). A significant decline in anti-RBD antibody levels was also observed at day 180 in 77.8% of the vaccinees (Supplementary table 3), on average by 30% (median value from 1140 U/ml at day 90 to 799.8 U/ml at day 180).
We next measured the level of neutralizing antibodies in these vaccinees using a surrogate virus neutralization test (sVNT) for the Wuhan strain, that has a good concordance with the live-virus neutralization test [42]. It was observed that >79.1% of the plasma had neutralizing antibodies above individual baseline after the first dose, 99% after the second dose and ~93% at day 180. However, between day 90 and day 180, serum neutralization efficacy declined in 77.5% of the participants and on average by 25% (from a median inhibition of 89.9% to 67.4%) (Fig. 1D and supplementary table 3). One third of the vaccinees mounted a poor secondary neutralization response (Supplementary table 6) and 19 out of 312 (6%) had no neutralizing antibodies (below baseline inhibition) at day 180 (Fig. 1D).
Notably, when the data from different serological assays were analyzed according to age, we observed a significant negative correlation of the age of the individuals with the antibody response at day 21 (after the first dose) and also with the antibody response at day 90 (after the second dose) (Supplementary fig. S2). Sample distribution revealed two clusters of individuals, low responders (1) <60 and (2) ≥ 60 years. This was confirmed by principal component analysis (PCA) of the three combined serological assays at days 21, 90 and 180 (Fig. 2A). Thus, the data age-stratified and analyzed (<60 versus ≥60). For all assays, vaccinees ≥ 60 years had a lower response compared with younger vaccinees after the first dose (Fig. 2B). After the second dose, the antibody responses were boosted in all groups. However, the increase was less pronounced in the older age group who displayed lower antibody levels against the whole spike protein and RBD and had serum neutralizing capacity at day 90 (Fig. 2B). At day 180, while the elderly had more antibodies against the whole S-protein compared with the younger population (Fig. 2B, left panel), they still had lower levels of anti-RBD antibodies and lower neutralizing antibody capacity (Fig. 2B, middle and right panel respectively). Similarly, there were more older individuals (participants with responses below median cohort response at consecutive time points) among the low responders (Supplementary tables S4, S5 and S5).
We next assessed the waning of antibodies between age groups by measuring the difference in antibody levels between days 180 and 90 in paired samples (Fig. 2C). Although antibody levels were lower at the cohort level, decline in antibody levels was significantly more pronounced in the older population than in the younger one (Fig, 2C left and middle panels, and supplementary table 3). The waning of neutralization capacity did not differ between both age groups (Fig. 2C right panel and supplementary table 3).
We also examined the binding efficiency of the vaccinated plasma to the spike protein of the Delta variant using the SFB assay. In a previous study, we reported that the levels of IgG against the Wild type or its variants strongly was strongly correlated with their capacity to inhibit pseudovirus and live virus neutralization expressing the respective various Spike proteins [43, 44]. Here, we show that any time points the antibody response was lower against the Delta variant than the wild-type original strain (Fig. 3A). However, the difference in recognition was only different between age groups after the first dose at day 21 (Fig. 3A, left panels). This was not observed at later time points similar in both age groups (Fig. 3B, middle and right panels).
Memory B cell response to SARS-CoV-2 mRNA vaccine
To measure the induction of RBD-specific circulating memory B cells by the vaccine, a B cell ELISPOT assay was performed on a subset of randomly selected age-matched individuals (n=78, from which we had 36 paired samples for the four time points). After the first dose, there was no significant increase at day 21, even though 47% of the individuals of the paired groups had higher memory B cells than their individual baseline (Supplementary table 2). After the second dose, a significant increase in RBD-specific memory B cell percentage was observed at day 90 (Fig. 4A). Analysis of paired samples confirmed these observations (Fig. 4B), where 76.5% had positive responses above their baseline levels. By day 180, the numbers of RBD-specific B cells continued to increase (Fig. 4A), with 85.3% of individuals having responses above their baseline levels at day 180 (Supplementary table 2). Generally, all individuals had produced RBD-specific circulating B cells in the six-month period.
When the data were age-stratified, we observed that the memory B cell response was lower in vaccinees ≥ 60 years at day 21, compared with the younger vaccinees after the first dose, (Fig. 4C). However, after the second dose, the memory B cell response increased in vaccinees ≥ 60 years at day 90 and 180 (Fig. 4C), corresponding to an overall increase in the number of total memory B cells (Fig. 4D). After two doses, the memory B cell response continued to increase for both age groups over time (Fig. 4C). At day 180, the difference between the two age groups disappeared, with both age groups having similar levels of memory B cell response (Fig. 4C-D). By comparing differences in the memory B cell response between time points (Fig. 4E), we found that younger individuals responded faster, with a greater increase right after the first dose at day 21. Therefore, the second dose is critical for the older age group as observed by the significant increase at day 90 (Fig 4E). .
T cell responses to SARS-CoV-2 mRNA vaccine
T cell stimulation was determined by quantifying cytokines (IL-2 and IFN-γ) directly secreted by Spike–specific CD4 and CD8 T cells in whole blood after overnight incubation with peptide pools covering 75-80% of Spike protein [45]. This was done in a subset of volunteers (n=155) randomly selected from the cohort but age matched (n=81 < 60 and 82 ≥ 60). At baseline, majority (~95%) of the individuals had very low production of IL-2 or IFN-γ (less than 10 pg/ml) after Spike peptide pool stimulation (Fig. 5A and 5B). After one dose, most of the vaccinees had a T cell response that increased further after the second dose. 100% of the vaccinees had a peptide-mediated IL-2 response above individual baseline after the first dose at day 21, days 90 and 180 (Fig. 5A and supplementary table 2). A robust IFN-γ response above baseline was also observed (~93 to 95% after the first and second doses) and sustained up to day 180 (Fig. 5B and supplementary table 2).
We next performed a detailed analysis of the T cell subsets by Elispot in a smaller subset of the volunteers due to cell availability. We used peptides covering potential CD8 or CD4 T epitopes (see materials and methods). For the CD8 assay using Spike protein peptide pools covering potential 9mers CD8 epitopes [46], we showed that, at baseline, some vaccinees already showed a high level of spots (Fig. 5C), suggesting a cross-reactive CD8 T cell response from exposure to other circulating coronaviruses. After the first dose, 54% had an increase in spots above their individual baseline values at day 21. After the second dose, 75% of the vaccinees had a response above their individual baseline at day 90 (Fig. 5C and supplementary table 2). By day 180, only 40.3% still had a CD8 T cell response (above their own baseline values, Supplementary table 2). Overall, 88.9% (64/72) mounted a CD8 T cell response during the 6-month follow-up. However, a comparison between responses at day 180 and 90 revealed that the response waned in 48% of the vaccinees (Supplementary table 3).
We next stimulated PBMC with a 15mer peptide pool corresponding to potential CD4 epitopes and measured the response by Elispot. CD4 Th1 (IL-2 and/or IFN-γ) responses were low at baseline, except for a few individuals (Fig. 6D). After one dose, 69% of the vaccinees and 84.6% after the second dose had a response higher than their baseline by day 90 and 83.33% by day 180 (Supplementary table 2). Overall, the CD4 Th1 cell response was significantly different after the first dose and further significantly boosted after two doses. Over 96.2% (75/78) mounted a response during the six-month follow-up. Comparison between responses at day 180 and 90 revealed that the CD4 Th1 response waned in 46.6% of the vaccinees (Supplementary table 3). A CD4 Th2 cell response was observed but was not as strong as the Th1 response (Supplementary fig. 3). At day 21, 59% of the vaccinees had values above their own baseline, a percentage which remained constant at day 90 but started to wane by day 180 (Supplemental tables S2 and S3).
Age-stratified analysis showed that post-vaccination Spike peptide pool-mediated IL-2 response was similar in both age groups at all time points (Fig. 6A). The IFN-γ response was lower at baseline in the <60 group. However, after the first dose (day 21), it reached similar level to that of ³60 group. At days 90 and 180, the older age group had T cells producing significantly more IFN-γ than the younger individuals (Fig. 6B). We did not observe any age effect on the CD8 ELISpot response (Fig.6C). CD4 Th1 was significantly lower at baseline for the older age group but the responses were similar at days 21, 90 and 180 (Fig. 5D). Post-immunization Th2 cell responses were also similar at the different times (Supplementary fig. 3 and supplementary table 3).
We next assessed the waning of T cell responses between age groups by measuring difference in response levels between days 180 and 90 in paired samples (Fig. 6E and supplementary table 3). Although T cell responses were lower at the cohort level, the decline was not significantly different between age groups. On the contrary, IFN-γ T cell response was even higher in the older age group (Fig 6E, middle left panel).