Participant demographics

Ninety-eight participants with PCR-confirmed SARS-CoV-2 infections and 17 age- and sex-matched controls without history of COVID-19 infection or vaccination were enrolled into the study. Participants were stratified by self-reported COVID-19 severity, with 10 (10%) asymptomatic, 42 (43%) mildly ill (symptomatic, non-hospitalized), 35 (36%) severely ill (hospitalized, non-intensive care unit (ICU)), and 11 (11%) critically ill (hospitalized, ICU) participants. Of the COVID-19 positive cohort, 79 (81%) participants had biospecimens collected at 1 month, and 52 (51%) had biospecimens collected at all four timepoints (1, 3, 6, and 12 months), with an average of 3.6 samples per infected participant (Supplementary Fig. 1). The drop in biospecimen collection was due to participants being lost to follow-up. All samples (365) were assayed for anti-N and anti-RBD IgG titers, and 306 (84%) were assayed for anti-S IgG titers.

Participant ages spanned 18–87 years and the median age for the COVID-19 positive cohort was 42 years. 51 (52%) of the SARS-CoV-2 positive participants were female and 46 participants (47%) identified as Hispanic, with additional demographic and clinical information provided in Table 1. Enrollment in the study was restricted to unvaccinated participants. Between their positive PCR test and the end of the study, 65 (57%) participants received at least one COVID-19 vaccination. Sixteen (14%) participants experienced reinfections during the study, determined by either participant reporting a reinfection or a 1·5-fold increase in anti-N and anti-RBD titers relative to the prior sample collection timepoint.

Concordant antibody measurements across laboratory-developed and commercial assays

We analyzed anti-SARS-CoV-2 titers across four different assays and measured associations between antibody responses to RBD versus whole S protein at 1, 3, 6, and 12 months post-acute SARS-CoV-2 infection for participants with different disease severities. Anti-RBD titers, assayed with a commercial ELISA kit, were similar to anti-RBD titers assayed with the Siemens large-scale indirect chemiluminescent immunoassay (Supplementary Fig. 3A, Spearman correlation R = 0.88, p < 2 × 2e − 16). Anti-RBD titers also strongly correlated with anti-S titers by ELISA, validating the anti-S laboratory-developed ELISA (Supplementary Fig. 3B, Spearman correlation R = 0.83, p < 2 × 2e − 16).

Anti-SARS-CoV-2 antibody titers correlate with acute COVID-19 severity both initially and in convalescence

We next sought to determine whether there was a difference in convalescent SARS-CoV-2 antibody titers based on acute COVID-19 severity. We found that anti-N, anti-RBD, and anti-S IgG titers were significantly associated with acute disease severity at 1 month post-positive SARS-CoV-2 PCR (overall anti-N p.adj = 1 × 3E − 4, anti-RBD p.adj = 8 × 5E − 3, anti-S p.adj = 8 × 5E − 3, controlling for age and sex) (Fig. 2a–c), with the critically ill participants exhibiting the highest convalescent antibody titers. Notably, the association between acute disease severity and convalescent antibody titers was maintained over 12 months for anti-N IgG (3 months p.adj = 0.0012, 6 months p.adj = 0.0021, 12 months p.adj = 0.01), and for 3 months for anti-RBD IgG (3 months p.adj = 0.0094). While overall acute disease severity was significantly associated with anti-N titers up to 12 months, pairwise comparisons between severity groups at 12 months did not demonstrate statistical significance post-correction for multiple testing. Notably, we had lower power to resolve persistent association between severity and titers for anti-RBD and -S compared to anti-N because post-vaccination data were removed for these 2 assays.

Box and whisker plots of a Anti-N, b Anti-S, and c Anti-RBD (measured with the BioLegend anti-RBD ELISA kit) IgG titers for asymptomatic (A), mild (M), severe (S), and critical (Crt) groups at 1, 3, 6, and 12 months post-positive SARS-CoV-2 PCR. Sample sizes (n) for Anti-N are 1 month (n = 79), 3 months (n = 83), 6 months (n = 83), and 12 months (n = 66); for Anti-S are 1 month (n = 63), 3 months (n = 57), 6 months (n = 31), and 12 months (n = 11); and for Anti-RBD are 1 month (n = 75), 3 months (n = 61), 6 months (n = 37), and 12 months (n = 16). Global p-values were obtained using a cumulative link model testing the association between disease severity and antibody titers, controlling for age and sex. Pairwise comparisons were performed using the Wilcoxon test with FDR correction (* p.adj ≤ 0.05, ** p.adj ≤ 0.01, *** p.adj ≤ 0.001). Dotted lines indicate the 95% quantile of healthy controls.

Vaccination increases anti-RBD titers in hospitalized and non-hospitalized participants, but decreases the ratio of anti-RBD:anti-S

As antibodies against SARS-CoV-2 RBD are associated with viral neutralization, we sought to test the extent to which vaccination increased participants’ anti-RBD antibody titers. We evaluated anti-N and anti-RBD titers at 3, 6, and 12 months post-positive SARS-CoV-2 PCR, relative to participants’ titers at 1-month post-infection. Participant samples were stratified by hospitalization status, and we compared mean antibody titers in samples categorized by vaccination status (i.e., before and after vaccination). In pre-vaccination samples from both hospitalized and non-hospitalized participants, a decline in anti-N and anti-RBD levels is evident over the 12 month time-course (Fig. 3a, b). As expected, anti-N titers were not impacted by vaccination (Fig. 3a). Anti-RBD titers increased significantly following vaccination, regardless of participants’ hospitalization status (non-hospitalized: p = 1 × 5E − 6 (3 months), p = 5 × 4E − 6 (6 months), p = 3E − 4 (12 months) and for hospitalized: p = 3E − 3 (3 months), p = 3 × 9E − 7 (6 months), p = 3E − 3 (12 months)) (Fig. 3b). The ratio of anti-RBD/ anti-N titers did not differ significantly between hospitalized versus non-hospitalized participants at any timepoint (Supplementary Fig. 4), although this ratio did increase in both groups following vaccination, as expected (Fig. 3a, b). Notably, anti-RBD titers were elevated to a greater extent after vaccination in non-hospitalized compared to hospitalized participants at 3- (p.adj = 4E − 2) and 12-month (p.adj = 4 × 7E − 2) timepoints.

Log2 fold change compared to 1-month for a Anti-N and b anti-RBD IgG levels. Hospitalized sample sizes (n): 3 months before 1st vaccination (n = 25), after 1st vaccination (n = 11); 6 months before 1st vaccination (n = 14), after 1st vaccination (n = 25); 12 months non-hospitalized (n = 5), hospitalized (n = 25). Anti-N non-hospitalized sample sizes (n): 3 months non-hospitalized (n = 36), hospitalized (n = 11); 6 months non-hospitalized (n = 23), hospitalized (n = 21); 12 months non-hospitalized (n = 11), hospitalized (n = 25). c Anti-RBD to anti-S ratio at 1, 3, 6, and 12 months post-positive SARS-CoV2 PCR, separated by whether the samples were collected before or after the 1st vaccination. Before 1st vaccination sample sizes (n): 1 month non-hospitalized (n = 36), hospitalized (n = 39); 3 months non-hospitalized (n = 36), hospitalized (n = 25); 6 months non-hospitalized (n = 23), hospitalized (n = 14); 12 months non-hospitalized (n = 11), hospitalized (n = 5). After 1st vaccination sample sizes (n): 1 month non-hospitalized (n = 1), hospitalized (n = 3); 3 months non-hospitalized (n = 11), hospitalized (n = 11); 6 months non-hospitalized (n = 21), hospitalized (n = 25); 12 months non-hospitalized (n = 25), hospitalized (n = 25). d Anti-RBD to anti-S IgG ratio before and after 1st vaccination for all samples; paired sample size (n = 49). Pairwise comparisons were performed using the Wilcoxon test with FDR correction (* p.adj ≤ 0.05, ** p.adj ≤ 0.01, *** p.adj ≤ 0.001). Dashed lines indicate 1-month titer levels.

Considering the essential role of anti-RBD in viral neutralization, we next tested if the relative titers of antibodies against RBD versus total S protein differed according to hospitalization or vaccine status. Prior to vaccination, hospitalized participants had a significantly higher anti-RBD to anti-S IgG ratio at 1-month post-PCR test compared to non-hospitalized participants (p.adj = 0·048). Similarly, post-vaccination, hospitalized participants had an elevated ratio of anti-RBD to anti-S IgG at 3 months post-PCR test relative to non-hospitalized participants (p.adj = 0.017) (Fig. 3c). In addition, paired comparisons of mean titers pre- and post-first vaccination across all disease severities demonstrated that the anti-RBD to anti-S IgG ratio significantly decreased in convalescent participants post-vaccination (p = 1 × 1e−4) (Fig. 3d).

Anti-N SARS-CoV-2 antibody titers decay faster, but remain higher, in hospitalized compared to non-hospitalized participants over 12 months post-SARS-CoV-2 infection

We next analyzed the decay kinetics of anti-N IgG titers (for all samples) and anti-RBD IgG titers (for pre-vaccination samples) over the course of 12 months post-confirmed infection. Anti-N and anti-RBD titers declined over 12 months for participants of all disease severities (Fig. 4a, b). However, the rate of decay was generally greater in hospitalized relative to non-hospitalized participants. Specifically, as a measure of antibody longevity, we examined the percent of participants at 3, 6, and 12 months who sustained SARS-CoV-2 antibody titers against N and RBD above 50% of their individual 1-month levels (Fig. 4c, d). The proportion of participants with anti-N titers remaining above 50% of their 1-month levels declined with decay constant of 0.604 month⁻¹ in hospitalized versus 0.237 month⁻¹ non-hospitalized participants, and anti-RBD titers declined with decay constants of 0.628 month⁻¹ in hospitalized versus 0.345 month⁻¹ non-hospitalized participants. Of note, despite the faster antibody decay rate in hospitalized participants, given their substantially higher anti-N and anti-RBD titers at 1 month, antibody titers generally remained higher over the study period in hospitalized relative to non-hospitalized participants (Fig. 4a, b; overall anti-N p.adj = 1 × 3E − 4, anti-RBD p.adj = 8 × 5E − 3, controlling for age and sex).

Longitudinal decay of a anti-N and b anti-RBD antibody titers in participants stratified by initial disease severity. Participant timepoints are connected by light gray lines. The healthy controls 95% confidence interval is represented by a dashed line. Percentage of participants whose c anti-N titers and d anti-RBD titers remained above 50% of their respective 1-month post-infection titers over 12 months. Percentage of participants whose e anti-N and f anti-RBD antibody titers remained above the 95th quantile of controls over 12 months in all participants. Sample sizes are provided in the tables accompanying each respective plot. Dashed lines in c–f represent 95% confidence intervals.

For anti-N, there was a significant difference in the frequencies of non-hospitalized versus hospitalized participants retaining titers above 50% of initial titers (asymptotic logrank two-sample test p value = 0.01). By 3 months, 50% of non-hospitalized and 22.2% hospitalized participants retained at least 50% of their 1-month anti-N titers (Fig. 4c). By 6 months, only 25% of non-hospitalized and 3.3% of hospitalized participants retained anti-N titers above 50% of 1-month levels, consistent with a faster antibody decay rate for hospitalized participants. For anti-RBD, by 3 months, 56.5% of non-hospitalized and 30% hospitalized participants had titers above 50% of their 1-month titers (Fig. 4d). This percentage decreased to 14.3% of non-hospitalized and 0% of hospitalized participants by 6 months and to 0% by 12 months for both non-hospitalized and hospitalized participants, indicating that none of the participants retained 50% of their initial RBD by 12 months of infection. However, there was no significant difference in the rate of decline in anti-RBD titers between non-hospitalized and hospitalized participants.

Longevity of anti-SARS-Cov-2 antibody titers in participants relative to uninfected control participant titers

To evaluate when anti-SARS-CoV-2 antibody titers decline to levels of uninfected control participants, we compared anti-N and anti-RBD IgG titers in infected participants over 12 months post-confirmed infection relative to the 95% quantile of titers in uninfected control participants (Fig. 4e, f). We observed that the frequency of participants with anti-N titers over control levels declined faster for non-hospitalized compared to hospitalized participants (asymptotic logrank two-sample test p value = 0.002). At 1 and 3 months post-infection, over 50% of non-hospitalized and hospitalized participants had antibody titers exceeding the 95% quantile of control levels (Fig. 4e, f). By 6 months, 46.5% of non-hospitalized and 71.8% of hospitalized participants maintained anti-N antibody levels above control titers. Finally, by 12 months, less than 25% of non-hospitalized and 50% of hospitalized participants retained anti-N antibody levels above control titers, demonstrating a faster decline in the frequency of non-hospitalized participants whose anti-N titers remain above control levels. Interestingly, although anti-RBD titers decayed with time (Fig. 4b), they remained above the 95% quantile of pre-COVID-19 control levels over 12 months, regardless of acute disease severity (Fig. 4f).

Age correlates with SARS-CoV-2 antibody titers at early post-infection timepoints

There was a moderate, but significant positive correlation between age and antibody titers at 1 month for anti-N (Spearman correlation R = 0.32, p = 0.0049), and for anti-RBD antibodies (Fig. 5a, b, Spearman correlation R = 0.35 with p = 0.0023. These correlations remained significant despite controlling for severity. However, it remains possible that the correction may not completely remove the association between age and severity, as further explored in the discussion. In contrast, there was no correlation between sex and antibody titers at any of the tested timepoints (Fig. 5c, d).

Correlation between age and a anti-N IgG levels and b anti-RBD IgG levels while controlling for disease severity. Sample sizes (n) are as follows: anti-N 1 month (n = 79), 3 months (n = 83), 6 months (n = 83), 12 months (n = 66); anti-RBD 1 month (n = 75), 3 months (n = 61), 6 months (n = 37), 12 months (n = 16). Correlation between sex and c anti-N IgG levels and d anti-RBD IgG levels while controlling for disease severity. Sample sizes (n) are as follows: anti-N 1 month female (n = 40), male (n = 39); 3 months female (n = 45), male (n = 38); 6 months female (n = 43), male (n = 40); 12 months female (n = 38), male (n = 28); anti-RBD 1 month female (n = 39), male (n = 36); 3 months female (n = 35), male (n = 26); 6 months female (n = 19), male (n = 18); 12 months female (n = 9), male (n = 7). F female, M male.

PASC status is not associated with anti-N and anti-RBD convalescent antibody titers

In our study, participants who had at least one persistent symptom per body system for longer than 1 month after a positive COVID-19 test at 2 or more visits were classified as having post-acute sequelae of COVID-19 (PASC). Using this criterion, 16.3% of the cohort developed PASC, which included long-term symptoms such as brain fog, fatigue, insomnia, headaches, myalgias, cough, and dyspnea (Table 2). Stratifying participants by disease severity groups, 0/10 (0%) of asymptomatic, 6/42 (14.3%) of mildly ill, 8/35 (22.9%) of severely ill, and 2/11 (18.2%) of critically ill participants reported PASC symptoms. There was no significant correlation between PASC status and anti-N or anti-RBD antibody titers at initial or subsequent timepoints over the course of 12 months.

No association between acute stage immunosuppression treatment and convalescent anti-N and anti-RBD antibody titers

We also evaluated whether treatment with steroids during acute infection affected anti-N and anti-RBD antibody titers over the course of a year in hospitalized participants. We did not find that steroid treatment correlated with a difference in antibody titers in hospitalized participants at any time point (Fig. 6).

a Anti-N IgG levels and b anti-RBD IgG levels in hospitalized participants across 1, 3, 6, and 12 months post-SARS-CoV-2 infection, stratified by treatment with (+) and without (−) steroid. Anti-N titers sample sizes (n): 1 month −steroid (n = 7), +steroid (n = 35); 3 months −steroid (n = 8), +steroid (n = 28); 6 months −steroid (n = 8), +steroid (n = 31); 12 months −steroid (n = 7), +steroid (n = 23). Anti-RBD titers sample sizes (n): 1 month −steroid (n = 7), +steroid (n = 32); 3 months −steroid (n = 7), +steroid (n = 18); 6 months −steroid (n = 4), +steroid (n = 10); 12 months −steroid (n = 0), +steroid (n = 5). Pairwise comparisons of titers in participants treated with and without steroids were performed using the Wilcoxon test (p > 0.05 for all comparisons).