Previous immunity shapes immune responses to SARS-CoV-2 booster vaccination and Omicron breakthrough infection risk – Nature Communications

Characteristics of the study population

The study cohort is part of a longitudinal vaccination study of Danish healthcare professionals described previously¹⁰. A total of 1325 healthcare professionals were included in the present study, of which 1145 (86.4%) were female, with a median age of 52 (IQR: 41–59) years. All individuals had received a third dose of the BNT162b2 vaccine (boost) at a median of 295 days after administering the first vaccine dose (IQR: 287–302 days, Table 1). At the time of 12 months sample collection, we identified 955 (72.1%) SARS-CoV-2 infection-naive individuals (nucleocapsid [N] protein negative and no positive RT-PCR result). A total of 463 (48.5%) of these had a positive RT-PCR result after the last sampling round, meaning that these individuals were infected after the collection of the 12-month sample and, consequently, after the boost (identified in the text as future infected individuals). We identified 370 individuals with hybrid immunity (a combination of SARS-CoV-2 infection and vaccination). Of these 370, 163 (44.1%) individuals were infected with the Omicron variant (identified in the text as infected with Omicron individuals), whereas 207 individuals were infected with an earlier variant before Omicron dominance in Denmark (identified in the text as infected before Omicron individuals). Of these, 100 were infected with the ancestral variant, 38 with the Alpha variant and 69 with the Delta variant (Table 1). Among participants infected before Omicron, 129 (62.3%) were infected only once, while 78 participants were reinfected by the 12-month sampling or after, confirmed by a positive RT-PCR result (identified in the text as reinfected individuals). The study design and timeline are illustrated in Fig. 1. Figure 2 depicts a flow chart with the participants and subgroups included in the analyses. The demographic characteristics of the different subgroups are described in detail in Supplementary Table 1. In the reinfection subgroup (identified in the text as reinfected and not reinfected) and future infection subgroup (identified in the text as future infected and not future infected) we evaluated whether certain immune responses are associated with reinfection and breakthrough infections, respectively; and in the immune imprinting cohort (identified in the text as reinfected and infected with Omicron) we evaluated the influence of previous infection on the boost responses.

Flow chart illustrating the inclusion and exclusion criteria to obtain the main cohort used for statistical analyses and the three subgroups obtained from the main cohort for further statistical studies (reinfection, future infection, and immune imprinting).

IgG levels dynamics after the booster diverge according to infection status, age group, and sex

We fitted a generalized linear-mixed model (GLMM) with five natural cubic splines (NCS) on the 12-month period to study the IgG dynamics over time (Fig. 3). Since the boost administration coincided with the gap between the 6- and 12-month rounds, the insufficient observed data did not allow us to model the immunologic event immediately after vaccination (Fig. 3, gray-shaded area). To show the expected immediate response after the boost, we fitted an additional two-part independent model. One GLMM with two NCS from baseline to 6 months and a linear-mixed model on the 12-month round only. Both models were used to theoretically project the antibody waning until boosting and project the peak reached after boosting (Supplementary Fig. 1, gray-shaded area). Since the GLMM fitted on the entire 12-month period allowed us to evaluate the influence of the initial immune response on the boost effect, statistical analyses were performed using this model.

Distribution of IgG levels, represented in log10(AU/ml), over time (days from the first vaccine) in infection-naive individuals (left), in individuals previously infected with a variant before Omicron (middle), and in individuals infected with Omicron (right). Circles and triangles represent the observed levels of circulating IgG levels in females and males, respectively. Solid and dashed lines represent the predicted levels of circulating IgG levels calculated by the model in females and males, respectively. Black, yellow, and blue colors represent individuals with age <40, 40–60, and >60 years, respectively. Horizontal black dotted line represents the threshold for assay positivity. Vertical dashed and dashed-dotted lines represent when the second and the third dose was administered, respectively (median days). Shadowed areas represent the 95% confidence interval. Centre for the confidence interval is the predicted (mean) values. Predictive values in the gray-shaded area (days 250–350) do not represent a realistic projection due to insufficient observed data to provide realistic predictive data. Source data are provided as a Source Data file.

Due to the large number of predictive values (model outcomes) provided by the different GLMMs, predictive values are reported only on females due to simplicity and power rationale. All predicted values for all age groups, infection status, and sex can be found in the respective tables (Supplementary Tables 2–17) in the Supplementary Information. The time points chosen for direct comparison between groups or time periods were selected based on when the peak level was reached, depending on the age group, infection status, and sex.

IgG dynamics over time were characterized by an initial peak in IgG levels after the second dose (prime), followed by a rapid waning. Administration of a boost resulted in the IgG levels being restored to similar levels as observed following the prime in all three groups defined as (i) infection-naive individuals, (ii) individuals infected before Omicron and iii) individuals infected with Omicron (Fig. 3, left, middle and right panel, respectively). Significant differences in IgG dynamics over time were observed between the three groups (p < 0.001, Fig. 3). Individuals infected before Omicron demonstrated a consistently higher IgG response compared to infection-naive individuals regardless of the age group (e.g., 23,821 Arbitrary Units [AU]/ml [95% confidence interval (CI): 19,433–29,241 AU/ml] in females infected before Omicron aged <40 years compared to 14,999 AU/ml [95% CI: 13,163–17,131 AU/ml] in infection-naive females aged <40 years, after the boost; Supplementary Table 2). Individuals infected before Omicron aged >60 years presented higher IgG levels after the third dose compared to the younger age groups (e.g., 23,821 AU/ml [95% CI: 19,433–29,241 AU/ml] in females aged <40 years, 21,710 AU/ml [95% CI: 18,343–25,733 AU/ml] in females aged 40–60 years, and 41,708 AU/ml [95% CI: 30,683–57,099 AU/ml] females aged >60 years, Supplementary Table 2). As expected, individuals infected with Omicron showed the highest IgG levels following the boost due to the proximity of the last infection to the sampling.

A significant interaction between days from the first vaccine dose and sex was observed (p = 0.011). IgG levels after the boost in infection-naive males were higher than in infection-naive females, contrary to the IgG peak levels observed after the prime. This interaction was only observed in infection-naive individuals but not in individuals with hybrid immunity, where the IgG levels were higher in females than males both after the prime and boost (Supplementary Table 2). The biggest difference in IgG levels between sexes was observed in individuals infected before Omicron (e.g., 23,891 AU/ml [95% CI: 19,433–29,241 AU/ml] in females aged <40 years old compared to 17,202 AU/ml [95% CI: 12,312–23,959 AU/ml] in males aged <40 years old, Supplementary Table 2). Males infected before Omicron had similar IgG levels after the boost compared to the infection-naive males (e.g., 17,202 AU/ml [95% CI: 12,312–23,959 AU/ml] in males infected before Omicron aged <40 years compared to 16,547 AU/ml [95% CI: 13,413–20,496 AU/ml] in infection-naive males aged <40 years, Supplementary Table 2).

Peak IgG levels reached after the prime in infection-naive individuals were higher than the peak IgG levels observed in individuals infected with Omicron regardless of age (e.g., 26,107 AU/ml [95% CI: 21,799–31,208 AU/ml] in infection-naive females aged <40 years compared to 19,355 AU/ml [95% CI: 14,566–25,612 AU/ml] in females infected with Omicron aged <40 years, Supplementary Table 2).

Neutralizing antibody levels are enhanced after booster dose

GLMMs with NCS were used to study the dynamic changes in neutralizing antibodies (nAbs) over time. Significant differences in the nAbs dynamics were observed in all three groups (p < 0.001, Fig. 4). Administration of the boost substantially increased the nAb levels compared to the peak generated after the prime in all groups (Fig. 4). Individuals with hybrid immunity (particularly those infected most recently with Omicron) presented the highest levels of nAbs compared to infection-naive individuals (Supplementary Table 3). Significant changes in the nAbs trends were observed in the different age groups over time (p < 0.001, Fig. 4), where aging was associated with lower nAbs levels after the prime, while similar nAb levels were observed between the different age groups after the boost (Supplementary Table 3). As observed for IgG, the peak of nAbs following the prime was lower in individuals infected with Omicron compared to infection-naive individuals (e.g. 14,323 International Units [IU]/ml [95% CI: 11,011–18,601 IU/ml] in infection-naive females aged <40 years compared to 11,597 IU/ml [95% CI: 8230–16,402 IU/ml] in females infected with Omicron aged <40 years, Supplementary Table 3). Moreover, an association between nAb levels over time and sex was observed (p = 0.037), where females displayed higher levels of nAb following the prime. However, after the boost, males showed a more substantial increase in nAbs than females. This dynamic change resulted in comparable levels following the booster shot (Supplementary Table 3). A two-part independent model of nAb waning and boost projection are illustrated in Supplementary Fig. 2.

Distribution of neutralizing antibody levels, represented in log10(IU/ml), over time (days from the first vaccine) in infection-naive individuals (left), in individuals previously infected with a variant before Omicron (middle), and in individuals infected with Omicron (right). Circles and triangles represent the observed levels of circulating neutralizing antibody in females and males, respectively. Solid and dashed lines represent the predicted levels of circulating neutralizing antibody calculated by the model in females and males, respectively. Black, yellow, and blue colors represent individuals with age <40, 40–60, and >60 years, respectively. Horizontal black dotted line represents the threshold for assay positivity. Vertical dashed and dashed-dotted lines represent when the second and the third dose was administered, respectively (median days). Shadowed areas represent the 95% confidence interval. Centre for the confidence interval is the predicted (mean) values. Predictive values in the gray-shaded area (days 250–350) do not represent a realistic projection due to insufficient observed data to provide realistic predictive data. Source data are provided as a Source Data file.

Hybrid immunity maintains IgA responses

IgA levels were modeled using GLMMs with a binomial distribution due to the assumptions of non-normally distributed data. An increase in the probability of a positive IgA response after the boost was observed in all groups (Fig. 5). However, the IgA response dynamics differed significantly over time according to the infection status (p < 0.001, Fig. 5). Individuals infected before Omicron maintained a probability of having a positive IgA response above 25% over time, which was enhanced after the boost. Individuals infected more recently (infected with Omicron) showed the greatest increase in the probability of a positive IgA response following boosting and infection (Supplementary Table 4). Infection-naive individuals exhibited a poorer IgA response after the boost compared with individuals with hybrid immunity (e.g., 33% [95% CI: 22–46%] in infection-naive females aged <40 years compared to 82% [95% CI: 73–90%] in females infected before Omicron aged <40 years, Supplementary Table 4). Individuals who became infected with Omicron in the future had a lower probability of a positive IgA response after the administration of the prime compared with individuals who remained infection-naive (e.g., 67% [95% CI: 57–77%] in infection-naive females compared to 44% [95% CI: 29–59%] in females infected with Omicron aged <40 years, Supplementary Table 4). No significant differences were observed between females and males (p = 0.581, Fig. 5). A two-part independent model of IgA response waning and boost projection are illustrated in Supplementary Fig. 3.

Distribution of positive IgA response (probability) over time (days from the first vaccine) in infection-naive individuals (left), in individuals previously infected with a variant before Omicron (middle), and in individuals infected with Omicron (right). Blue and pink backgrounds represent the conditional density estimation of positive and negative IgA responses, respectively. Solid and dashed lines represent the predicted probability of positive IgA responses calculated by the model in females and males, respectively. Black, yellow, and blue colors represent individuals with age <40, 40–60, and >60 years, respectively. Vertical dashed and dashed-dotted lines represent when the second and the third dose was administered, respectively (median days). Shadowed areas represent the 95% confidence interval. Centre for the confidence interval is the predicted (mean) values. Predictive values in the gray-shaded area (days 250–350) do not represent a realistic projection due to insufficient observed data to provide realistic predictive data. Source data are provided as a Source Data file.

T-cell-derived IFN-γ levels are boosted after the third dose and correlate with IgG and IgA levels

Cellular responses were assessed as IFN-γ release from T-cells stimulated with S1 peptides. IFN-γ levels were significantly higher in infection-naive individuals and individuals infected with Omicron after boost administration (12-month sampling) compared to levels before boosting (6-month sampling) (p < 0.001 in both groups, Fig. 6a, c). Individuals infected before Omicron presented higher IFN-γ levels at the 6-month sampling compared with the other groups (p < 0.007, Kruskal–Wallis test). However, no significant difference in IFN-γ levels between the two sampling rounds was observed in this group (p = 0.100, Fig. 6b). Individuals infected with Omicron showed the highest IFN-γ level response after the boost compared to the other groups (p < 0.001, Kruskal–Wallis test). Comparable results were observed when IFN-γ levels were modeled using a linear-mixed model (Supplementary Fig. 4 and Supplementary Table 5). Significant differences in IFN-γ levels between the different age groups were observed (p = 0.006, Supplementary Fig. 4), characterized by higher levels of IFN-γ in younger individuals (Supplementary Table 5). Males generated lower cellular responses compared to females (p = 0.033), but the dynamics were similar between the sexes (Supplementary Table 5).

a–c IFN-γ levels collected before (6-month round; green) or after (12-month round; yellow) of the third dose administration, represented in log10(mIU/ml), in infection-naive individuals (n = 115 and n = 650 biologically independent samples before and after the third dose administration) (a), individuals previously infected with a variant before Omicron (n = 23 and n = 122 biologically independent samples before and after the third dose administration) (b), and in individuals infected with Omicron (n = 23 and n = 98 biologically independent samples before and after the third dose administration) (c). Circles represent observed data. Data reported as median and interquartile range (box), whiskers represent 1.5 times the interquartile range. Dashed horizontal line indicates assay positivity threshold. P-values were calculated using Mann-Whitney U test (two-sided). Correlation between IFN-γ levels with IgG (d) and IgA levels (e) at 6-month round. Correlation between IFN-γ levels with IgG (f) and IgA levels (g) at 12-month round. Blue, red, yellow colors represent infection-naive individuals, individuals infected before Omicron, and individuals infected with Omicron, respectively. Black color represents overall Spearman Rank test results. Circles represent observed data. Shadowed areas represent the 95% confidence interval. P-values were calculated using Spearman Rank test (two-sided). p < 0.05 was considered statistically significant. Source data are provided as a Source Data file.

Significant correlations between IgG and IFN-γ levels were observed for all groups at both the 6- and 12-month sampling (Fig. 6d, f), being more pronounced at the 12-month sampling after the boost (overall R = 0.47, p < 0.001, Fig. 6f). Correlation between IgA and IFN-γ levels was only evident in individuals infected before Omicron at the 6-month sampling after the prime (R = 0.46, p = 0.026, Fig. 6e). However, at the 12-month sampling, following the boost, a significant correlation between IgA and IFN-γ levels was only observed in infection-naive individuals (R = 0.21, p < 0.001, Fig. 6g). The overall correlation between IgA and IFN-γ levels was detected at both samplings (6-month sampling: overall R = 0.26, p = 0.001, Fig. 6e; 12-month sampling: overall R = 0.36, p < 0.001, Fig. 6g).

Decreased humoral responses to the SARS-CoV-2 vaccine are related to reinfections

Differences in IgG, IgA, nAbs, and IFN-γ levels following priming between individuals infected before Omicron who did not experience reinfection (not reinfected) and those who did (reinfected) were studied (Fig. 7). Of note, only individuals who had the first infection before the administration of the second vaccine dose were included to establish reliable comparisons (demographic characteristics in Supplementary Table 1). IgG, IgA, and nAb levels were significantly lower in individuals who experienced reinfection in the future compared to those who did not (p = 0.007, p = 0.024, and p = 0.035, respectively, Fig. 7). There was no significant difference between the groups regarding IFN-γ levels (p = 0.340, Fig. 7). Multiple linear regression analyses showed comparable results (p < 0.001, p = 0.042, and p = 0.035 for IgG, IgA, and nAbs, respectively, Supplementary Table 18). Similar observations were detected using GLMMs. The dynamics of IgG and nAb levels showed significantly different trends between reinfected and not reinfected individuals (p = 0.044 and p = 0.016, respectively; Supplementary Figs. 5 and 6, Supplementary Tables 6 and 7). A tendency was observed regarding the dynamics in the probability of positive IgA responses and IFN-γ levels between reinfected and not reinfected individuals (p = 0.099 and p = 0.118, respectively; Supplementary Figs. 7 and 8, Supplementary Tables 8 and 9).

Distribution of IgG levels (a) and IgA levels (b), both represented as log10(AU/ml) (n = 59 and n = 55 biologically independent samples in the not reinfected and reinfected groups, respectively); neutralizing antibody levels (c), represented as log10(IU/ml) (n = 57 and n = 55 biologically independent samples in the not reinfected and reinfected groups, respectively); and IFN-γ levels (d), represented in log10(mIU/ml) (n = 7 and n = 5 biologically independent samples in the not reinfected and reinfected groups, respectively); during the waning period (day 15 after the second dose and day -1 before the third dose) in individuals infected before Omicron who did not get a second SARS-CoV-2 infection (Not reinfected) and those who did get a second SARS-CoV-2 infection (Reinfected). Green and yellow colors represent not reinfected and reinfected individuals, respectively. Circles represent observed data. Data reported as the median and interquartile range (box), whiskers represent 1.5 times the interquartile range. Dashed horizontal line indicates the threshold for assay positivity. P-values were calculated using Mann–Whitney U test (two-sided), where p < 0.05 was considered significant. Source data are provided as a Source Data file.

Lower humoral responses after the third dose are associated with future infections

To evaluate the association between immune responses and the risk of future infections, IgG, IgA, nAb, and IFN-γ levels were studied in a subcohort of infection-naive individuals who remained uninfected (not future infected) after the 12-month sampling and infection-naive individuals who became infected after sample collection assessed by a positive RT-PCR test (future infected) (Fig. 8, demographic characteristics in Supplementary Table 1). The IgG, IgA and nAb levels were significantly lower in individuals who would experience a future infection than those who did not after the 12-month collection round (p = 0.009, p = 0.031, and p = 0.028 for IgG, IgA and nAb levels, respectively, Fig. 8). No significant difference in IFN-γ levels was observed between the groups (p = 0.510, Fig. 8). Multiple linear regression analyses showed comparable results (p = 0.018, p = 0.040, and p = 0.902 for IgG, IgA, and IFN-γ, respectively, Supplementary Table 18), although a tendency was observed for nAbs (p = 0.103). When using GLMMs, similar trends were detected, where the IgG and nAb levels and probability of positive IgA responses dynamics over time differed significantly between individuals with a future infection and those who were not future infected (p = 0.023, p = 0.028, and p = 0.028 for IgG, nAbs and IgA, respectively, Supplementary Figs. 9–11, respectively, Supplementary Table 10, 11 and 12). Significant differences in the IFN-γ levels dynamics were also observed (p = 0.006, Supplementary Fig. 12, Supplementary Table 13).

Distribution of IgG levels (a) and IgA levels (b), both represented as log10(AU/ml) (n = 492 and n = 460 biologically independent samples in the not infected and infected groups, respectively); neutralizing antibody levels (c), represented as log10(IU/ml) (n = 488 and n = 459 biologically independent samples in the not infected and infected groups, respectively); and IFN-γ levels (d), represented in log10(mIU/ml) (n = 340 and n = 308 biologically independent samples in the not infected and infected groups, respectively); generated after the third dose (day 15 after the third dose) in infection-naive individuals who remain infection-naive after the 12-month round (Not infected) and those who get a future Omicron infection after the 12-month round based on a positive RT-PCR result (Infected). Green and yellow colors represent not infected and infected individuals, respectively. Circles represent observed data. Data reported as the median and interquartile range (box), whiskers represent 1.5 times the interquartile range. Dashed horizontal line indicates the threshold for assay positivity. P-values were calculated using Mann–Whitney U test (two-sided), where p < 0.05 was considered significant. Source data are provided as a Source Data file.

Humoral and cellular vaccine responses are influenced by previous immune imprinting

To assess the impact of possible imprinting from previous SARS-CoV-2 variants on the vaccine immune response, we studied IgG, IgA, nAb, and IFN-γ levels in a subcohort of individuals exposed to an earlier SARS-CoV-2 variant who experienced reinfection with Omicron (reinfected) and infection-naive individuals who experienced a primary infection with Omicron (infected with Omicron). In both groups, infections occurred before the sample collection following the boost (demographic characteristics in Supplementary Table 1). IgG, IgA and nAb levels were significantly increased after the boost in individuals infected with Omicron compared to those who were reinfected (p < 0.001, p = 0.013, and p = 0.019 in IgG, IgA, and nAb levels, respectively, Fig. 9). There was no significant difference between groups regarding IFN-γ levels (p = 0.270, Fig. 9). Multiple linear regression analyses showed comparable results (p < 0.001, p = 0.036, p = 0.055, and p = 0.679 for IgG, IgA, nAbs, and IFN-γ, respectively, Supplementary Table 18). When modeling using GLMMs, significantly different dynamics over time were detected for IgG and nAbs (p < 0.001 for both, Supplementary Figs. 13 and 14, respectively), characterized by a greater increase in IgG and nAb levels in individuals infected with Omicron compared to reinfected individuals following boosting (Supplementary Tables 14 and 15, respectively). Borderline significant differences in the dynamics over time of the probability of positive IgA responses between groups were observed (p = 0.064, Supplementary Fig. 15). Individuals infected with Omicron were lacking a positive IgA response before boost administration and subsequent Omicron infection. Consequently, these individuals showed a marked increase in IgA response compared to reinfected individuals (Supplementary Table 16). IFN-γ level dynamics differed significantly over time (p = 0.001, Supplementary Fig. 16, Supplementary Table 17), with a clear increase in IFN-γ levels in individuals infected with Omicron.

Distribution of IgG levels (a) and IgA levels (b), both represented as log10(AU/ml) (n = 22 and n = 163 biologically independent samples in the reinfected with Omicron and infected with Omicron groups, respectively); neutralizing antibody levels (c), represented as log10(IU/ml) (n = 22 and n = 161 biologically independent samples in the reinfected with Omicron and infected with Omicron groups, respectively); and IFN-γ levels (d), represented in log10(mIU/ml) (n = 13 and n = 98 biologically independent samples in the reinfected with Omicron and infected with Omicron groups, respectively); generated after the third dose (day 15 after the third dose) in individuals infected before Omicron who were reinfected with Omicron (Reinfected with Omicron) and individuals infected with Omicron for the first time (Infected with Omicron) at the 12-month round. Green and yellow colors represent reinfected with Omicron and infected with Omicron, respectively. Circles represent observed data. Data reported as the median and interquartile range (box), whiskers represent 1.5 times the interquartile range. Dashed horizontal line indicates the threshold for assay positivity. P-values were calculated using Mann-Whitney U test (two-sided), where p < 0.05 was considered significant. Source data are provided as a Source Data file.