Immunity to SARS-CoV-2 up to 15 months after infection

Information concerning the longevity of immunity to SARS-CoV-2 following natural infection may have considerable implications for durability of immunity induced by vaccines. Here, we monitored the SARS-CoV-2 specific immune response in COVID-19 patients followed up to 15 months after symptoms onset. Following a peak at day 15–28 postinfection, the IgG antibody response and plasma neutralizing titers gradually decreased over time but stabilized after 6 months. Compared to G614, plasma neutralizing titers were more than 8-fold lower against variants Beta, Gamma, and Delta. SARS-CoV-2-specific memory B and T cells persisted in the majority of patients up to 15 months although a significant decrease in specific T cells, but not B cells, was observed between 6 and 15 months. Antiviral specific immunity, especially memory B cells in COVID-19 convalescent patients, is long-lasting, but some variants of concern may at least partially escape the neutralizing activity of plasma antibodies.


INTRODUCTION
Coronavirus disease 2019 , caused by the novel severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), rapidly resulted in a pandemic constituting a global health emergency. The COVID-19 pathological process exhibits a wide spectrum of clinical manifestations, ranging from asymptomatic to mild, moderate, severe, and critical disease. The genome of SARS-CoV-2 encodes four major structural proteins that occur in all coronavirus species: spike protein (S), nucleoprotein (N), membrane protein (M), and envelope protein (E) (Naqvi et al., 2020). The S protein binds to the host receptor (ACE 2 [ACE2]) through the receptor-binding domain (RBD) in the S1 subunit, followed by the S2 subunit-mediated cell membrane fusion (Zhou et al., 2020).
The adaptive immune response is likely to be critical for the development of protective immunity to SARS-CoV-2 including viral clearance and the persistence of antiviral immunity (Poland et al., 2020). Generation of neutralizing antibodies that specifically target the receptor-binding domain (RBD) of the S protein is considered to be essential in controlling SARS-CoV-2 infection (Addetia et al., 2020;Khoury et al., 2021). A robust adaptive immune response with presence of RBD and S-specific neutralizing antibodies, memory B cells, and T cell response have been found in patients who have recovered from infection (Dan et al., 2021;Sherina et al., 2021;Wang et al., 2021c). Although circulating antibodies derived from plasma cells wane over time, long-lived immunological memory can persist in expanded clones of memory B cells (Wang et al., 2021c).
Although SARS-CoV-2 started spreading to Europe and the Americas in March 2020, the B.1 lineage carrying the D614G mutation (G614) quickly became the dominant lineage (Tao et al., 2021). The SARS-CoV-2 variant of concern (VOC) B.1.1.7 (Alpha) first identified in late September 2020 and rapidly growing in the United Kingdom in December 2020 then became the dominant lineage across much of Europe and America. During the same period, two additional VOCs rose in prevalence in South Africa (B.1.351, Beta) and Brazil (P.1, Gamma) but without causing as much damage in other continents as it caused in their place of emergence. Subsequently, a novel variant (B.1.617.2) increased in prevalence in India in winter 2021 and has since spread widely in multiple countries. Mutations in the S1 subunit in VOCs may lead to changes in the structure of the S protein and RBD (Davies et al., 2021a) and result in higher binding of the virus to the receptor (Ramanathan et al., 2021), increased risk of transmission, and severity of illness (Davies et al., 2021b), as well as a reduction in neutralization susceptibility by antibodies (Jangra et al., 2021;Liu et al., 2021;Wang et al., 2021b). The Delta variant is associated with an estimated 60% higher risk of household transmission than the Alpha variant and is becoming dominant worldwide (Mahase, 2021).
We have previously reported the longevity of the SARS-CoV-2 adaptive immune response (up to 6-8 months) in cohorts of Swedish and Italian patients infected with the SARS-CoV-2 G614 strain . The antibody and neutralizing titers were sustained at a relatively high level for at least 6 months after the onset of symptoms, whereas specific memory B and T cells were maintained for at least 6-8 months. In this study, the adaptive immune response in convalescent patients from the same cohorts was followed for up to 15 months. In addition, the specific antibody levels and neutralizing antibody titers were tested against VOCs.

Longevity of anti-SARS-CoV-2 antibody response
The majority of patients (99) were recruited between February 28 and December 3, 2020 during the first and second wave in Europe when the G614 variant (B.1 lineage) was the predominant circulating strain in both Sweden and Italy. The rest of the patients (37) were recruited between January 28 and June 16, 2021 during the third wave in which Alpha became a dominant variant and accounted for more than 50-60% of the cases at the beginning of March in Italy and Sweden (Carlsson and Sö derberg-Nauclé r, 2021;Di Giallonardo et al., 2021).
For the entire 15 months follow-up, a total of 188 serum or plasma samples were collected from 136 COVID-19 patients (98 from Italy and 38 from Sweden) experiencing mild symptoms of critical disease (Figure 1, Table S1 and Table S2). Plasma from 108 historical negative controls collected before the SARS-CoV-2 pandemic were also analyzed. Plasma anti-RBD and anti-S antibody titers were measured by an in-house ELISA (ELISA) .  also Tables S1, S2, S3 and S4 For the cross-sectional analysis, we further divided samples into six groups based on the sample time points; collected at 7-14 days, 14-30 days, 31-90 days, 91-180 days, 181-365 days, and 366-452 days, post-onset of symptoms (Table S3). At the peak of the antibody response, 15-28 days after symptoms onset, anti-RBD IgM and IgA were increased in 77% (40/52) and 85% (44/52) of convalescent patients, respectively, but rapidly decreased between 1 and 3 months and were detected in less than 4.5% (2/44) and 11% (5/44) of patients tested between 6 and 15 months when assessing all COVID-19 subjects by cross-sectional analysis (Figures 2A,2B,2D,and 2E). Similarly, IgM and IgA anti-S proteins were detected in 88% (46/52) and 90% (47/52) of convalescent patients at 15-28 days, respectively, but less than 23% (10/ 44) of patients for both immunoglobulins from 6 to 15 months ( Figures 2G, 2H, 2J, and 2K). Using a onephase exponential decay model, we estimated the half-lives (t 1/2 ) of the RBD-and S-specific IgM antibodies to be 55 and 65 days, respectively (Figures 2A, 2G), and that of RBD-specific and S-specific IgA antibodies to be 56 and 55 days, respectively ( Figures 2B and 2H).
Plasma IgG antibodies binding to SARS-CoV-2 RBD and S protein increased in 94% (49/52) of COVID-19 convalescent participants tested 15-28 days after symptoms onset. The median RBD and S IgG antibody titers gradually decreased by less than 4-fold from the peak of the antibody response until 6 months (15-28 vs 181-365 days, Figure 2. Cross-sectional analysis of plasma anti-SARS-CoV-2 antibody titers patients over time Levels of anti-RBD (A-F) and anti-S (G-L) IgM, IgA, and IgG antibodies in plasma of COVID-19 patients, historical controls, and vaccinated individuals. Antibodies were measured in 185 samples from 136 COVID-19 patients, 108 historical controls (before the SARS-CoV-2 pandemic), and 23 vaccinated individuals. The RBD (A-C) and S (G-I) specific IgM, IgG, IgA antibody decay curves (in black) and half-lives (t 1/2 ) were estimated by a one-phase exponential decay model. Samples from patients were further divided in six study periods: 7-14 days (n = 19), 15-28 days (n = 52), 29-90 days (n = 35), 91-180 days (n = 35), 181-365 days (n = 33), and 366-452 (n = 11) after symptom onset (D-F and J-L) for comparison. Vaccinated individuals were sampled 14-35 days after the first dose and 14-36 days after the second dose. For each time interval, the proportion of positive samples is indicated below the X axis. Symbols represent individual subjects; horizontal black lines indicate the median and 95% CI. The dashed red line indicates the cutoff value for elevated anti-S and anti-RBD antibody titers (2.5 and 8.4 AU/mL for IgM, 0.5 and 0.08 AU/mL for IgA, and 0.03 and 14.81 AU/mL for IgG, respectively, giving a specificity of 96% for IgM, 99% for IgA, and 97% for IgG). The cutoff-value is not visible in some graphs because it is very close to the X axis. Mann-Whitney U test. *p % 0.05, **p % 0.01, ***p % 0.001, and ****p < 0.0001. See also Figures S1 and S2. 2L). The half-lives of anti-RBD and anti-S IgG antibody response estimated by a one-phase decay model were 134 and 113 days, respectively (Figures 2C and 2I), and were shorter in patients with mild/moderate (t 1/2 = 52 days and t 1/2 = 40 days) than severe/critical (t 1/2 = 372 days and t 1/2 = 239 days) disease (Figures S1A and S1B).
As a complementary approach, we analyzed the antibody titers from 42 patients who donated blood at two or more time points and estimated the half-lives of the RBD-specific and S-specific antibody response in IgM (t 1/2 = 71 and 73 days), IgA (t 1/2 = 32 and 28 days), and IgG (t 1/2 = 128 and 90 days) (longitudinal analysis; Figure 3). No significant difference in the anti-S and anti-RBD IgG titers were observed between the paired samples (n = 11) containing a sampling time point between 181 and 300 days (6-10 months) and a second one between 301 and 452 days (10-15 months), confirming that specific IgG antibody titers reached a plateau phase after 6 months (p = 0.8984 and p= 0.3125, respectively; Figures 3C and 3F).
We further compared the antibody response induced from natural infection to that induced by one or two doses of the Comirnaty (Pfizer-BioNTech) vaccine (Table S4). The vaccine induced no or low level of RBD-specific, S-specific IgM ( Figures 2D and 2J  IgG antibodies, further comparisons were made between paired samples (n = 11) at two time points ranging from 6 to 15 months (TP1: 181-300 and TP2: 301-452 days after symptoms onset; right panel). Symbols represent individual subjects; horizontal black lines indicate the median and 95% CI. The antibody decay curves (in black) and half-lives (t 1/2 ) were estimated by a one-phase exponential decay model. The dashed red line indicates the cutoff value for elevated anti-S and anti-RBD antibody titers (2.5 and 8.4 AU/mL for IgM, 0.5 and 0.08 AU/mL for IgA, and 0.03 and 14.81 AU/mL for IgG, respectively, giving a specificity of 96% for IgM, 99% for IgA, and 97% for IgG). The cutoff-value is not visible in some graphs because it is very close to the X axis. Wilcoxon signedrank test. See also Figure S1. The neutralizing activity against the G614 variant was measured by a microneutralization assay and was expressed as the neutralizing titers (R1:10) which inhibit 90% of the virus infectivity (NT 90 ). Similar to the dynamic of the anti-RBD and anti-S antibodies, the median NT 90 reached a peak of 1:160 between 15 and 28 days with 98% (51/52) of convalescent patients exhibiting plasma neutralizing activity ( Figures 4A and 4B). The plasma NT 90 gradually decreased by about 2-fold up to 6 months (15-28 vs 91-180 days, p = 0.0493) but was still observed in more than 87% (35/40) of the patients sampled at 181-365 (27/31) and 366-452 days (8/9) ( Figure 4B). Furthermore, no significant difference was observed in the median NT 90 measured at 181-365 (1:40) and 366-452 days (1:80) (p = 0.6674; Figure 4B) or between paired samples at two time points between 6 and 15 months (n = 9, TP1: 181-300 days vs TP2: 301-452 days, p = 0.5156; Figure 4C). A one-phase decay model showed a rapid initial decay using both cross-sectional (t 1/2 = 70 days, Figure 4A) or longitudinal analysis (t 1/2 = 44 days, Figure 4C) which slow down to a plateau phase extending from around four up to 15 months. A (C) For longitudinal analysis, samples were taken at two (n = 31) or more (n = 7) time points and further comparisons were made between paired samples (n = 9) at two time points ranging from 6 to 15 months (TP1: 181-300 and TP2: 301-452 days after symptoms onset; right panel). The NT 90 decay curves (in black) and corresponding half-lives (t 1/2 ) were estimated by a one-phase decay model (A, C). To test cross-neutralization, the level of anti-RBD IgM, IgA, and IgG titers (D, G), binding activity of IgG antibody to RBD from SARS-CoV-2 variants (E, H), and plasma neutralizing activity against variants (F, I) were tested in plasma collected from COVID-19 patients at 15-106 days (median day of 24) and 9-15 months (241-452 days, median day of 370). The data in D and G represent a subset of data presented in Figure 2. The dashed red line indicates the titer cutoff value (R1:10). The cutoff-value is not visible in some graphs because it is very close to the X axis. Symbols represent individual subjects; horizontal black lines indicate the median and 95% CI. Mann-Whitney U test. *p% 0.05, **p % 0.01, ***p % 0.001, and ****p < 0.0001. See also Figures S1 and S2.

OPEN ACCESS
iScience 25, 103743, February 18, 2022 5 iScience Article decrease of NT 90 followed by a plateau phase was observed in both patients with mild/moderate and severe/ critical diseases using cross-sectional analysis ( Figure S1C), whereas no decline was observed for the severe/critical patients group using longitudinal analysis ( Figure S1F) suggesting that neutralization activity may persist longer in this group. The NT 90 directly correlated (p < 0.0001) with the levels of RBD-specific IgM (r = 0.45), IgA (r = 0.37), and IgG (r = 0.49) as well as with the level of S-specific IgM (r = 0.44), IgA (r = 0.44), and IgG (r = 0.55) antibodies ( Figure S2).
These data suggest that the plasma anti-SARS-CoV-2 antibody response and neutralizing activity decrease up to around 6 months but neutralizing activity is maintained in the majority of patients up to 15 months. Furthermore, plasma neutralizing activity was lower against Beta, Gamma, and Delta variants, particularly 9-15 months after infection.
No statistically significant differences were observed in the number of memory B cells between mild/moderate and severe/critical COVID-19 patients over the period ranging from 6 to 15 months suggesting that the intensity and duration of the B cell response are not dependent on the disease severity (p = 0.5835; Figures S3 and S6).
Noninfected individuals sampled 14-35 days after the first vaccine dose showed a B cell response significantly lower than that observed in convalescent patients between 3 and 15 months (vs 91-180 days, p = 0.0002; vs 181-365 days, p < 0.0001; vs 366-452 days, p = 0.0077 after onset of symptoms; Figure 5B). Subjects sampled 14-36 days after the second vaccine dose showed a median number of circulating RBD-specific memory B cells similar to that observed in convalescent patients sampled during the same period (vs 91-180 days, p = 0.3222; vs 181-365 days, p = 0.6337; vs 366-452 days, p = 0.8884; Figure 5B).

SARS-CoV-2-specific memory T cells
The S1 and S N M O (Orf3a and Orf7a) peptide pool-specific T cells expressing interleukin-2 (IL-2) and/or interferon-gamma (IFN-g) were measured by FluoroSpot assay. No or a negligible number of IL-2, IFN-g, or IL-2/IFN-g -producing T cells against the two peptide pools were detected in the negative controls. Overall, a T cell response against at least one of the SARS-CoV-2 peptide pools (S1, or S N M O protein derived) was detectable at a level above the cutoff in 95% (69/73) of the patient samples tested over the study period ( Figures 6A-6C and S4A-S4C). When divided by groups based on the sample time points, specific T cells were detected in the majority of patient samples at 2-4 weeks (10/11, 91%) and 1-3 months (7/9, 78%), and more than 98% (52/53) of patient samples tested between 3 and 15 months.
The numbers of S1 and S N M O peptide pool-specific IL2 and IL-2/IFN-g -producing T cells were higher in severe/critical COVID-19 than mild/moderate patients between 6 and 15 months (p = 0.0044 and p = 0.0202 for S1 specific T cells, p = 0.0302, and p = 0.0427 for S N M O specific T cells), whereas no significant difference was observed in S1 and S N M O specific IFN-g -producing T cells (p = 0.0610 and p = 0.1012, respectively) ( Figures S5 and S6).
The S1-specific T cell response measured in samples collected after one vaccine dose was equivalent to that observed in convalescent patients at 12-15 months when the specific T cell number decrease (p = 0.6333, p = 0.8352, p = 0.6893; Figures 6D-6F). Compared to the peak of T cell response in patients (3-6 months), individuals with two doses of vaccine present a similar level of S1-specific IL-2 and ll OPEN ACCESS iScience 25, 103743, February 18, 2022 7 iScience Article IL-2/IFN-g-producing T cells but a 2-fold higher number of IFN-g-producing T cells (p = 0.5177, p = 0.9915, and p = 0.0475, respectively; Figures 6D-6F).
Finally, we have investigated the presence of all three arms of immunity for each individual who has been evaluated for SARS-CoV-2-specific adaptive immunity in all assays i.e., 1) IgM, IgG, and IgA antibody response and/or neutralization activity, 2) RBD-specific IgG-producing B cells, and 3) S1 peptide or S N M O pool-specific IL-2 and/or IFN-g-producing T cells ( Figure 7A). The majority of individuals (72%, 55/ 76) had three arms of immunity active against SAR-CoV-2, particularly from 2 to 15 months (80%, 48/60), highlighting the increase of memory B and T cells between 0 and 2 months and the long duration of the adaptive immune response.
The interrelationship between the components of the immune system was then examined by a multiparameter analysis ( Figure 7B). We observed a positive correlation between the level of IgG, IgM, and IgA anti-RBD antibodies and the neutralization titers. The level of RBD-specific IgG-producing memory B cells was also positively associated with the IgG anti-RBD antibody and the S1 and S N M O pool-specific T cell response. However, no correlation was observed between the T cell response and the anti-RBD antibody response. Figure 6. Cross-sectional and longitudinal analysis of S1-specific memory T cell responses in COVID-19 patients (A-C) Dynamics of S1-specific memory IL-2, IFN-g, and IL-2/IFN-g-producing T cells with the corresponding second order polynomial fitting curve (in black).
(G-I) For longitudinal analysis, samples were taken at two (n = 10) or more (n = 5) time points and further comparisons were made between paired samples (n = 8) at two time points ranging from 6 to 15 months (TP1: 181-300 and TP2: 301-452 days after symptoms onset; right panel). The results were expressed as the number of spots per 300,000 seeded cells after subtracting the background spots of the negative control. The horizontal black lines indicate the median value and 95% CI of the group. The cutoff value (dashed red line) was set at the highest number of specific T cell spots for the negative controls (>7 to nine spots/300,000 seeded cells depending on the T cell population). The cutoff-value is not visible in some graphs because it is very close to the X axis. Mann-Whitney U test. **p % 0.01, ***p % 0.001, and ****p % 0.0001. See also Figures S4, S5  iScience Article Taken together, SARS-CoV-2-specific memory B and T cells remained present in the majority (>95%) of patients followed up between 6 and 15 months. A reduction of the specific T cell response, but not B cell memory, was observed at 12-15 months.

DISCUSSION
The magnitude, duration, and quality of immunological memory are crucial for preventing reinfection. In this study, we extended our assessment of the longevity of the SARS-CoV-2-specific antibody, B and T cell immune response in cohorts of convalescent patients in Italy and Sweden that experienced mild to critical symptoms of COVID-19. In many viral infections, such as those caused by Dengue and Zika virus, serum IgM responses precede the appearance of IgG and IgA antibodies (Roltgen and Boyd, 2021). In iScience Article contrast, IgG antibodies to SARS-CoV-2 S and RBD appear at approximately the same time as serum IgM and IgA antibodies, usually within the first 2 weeks after symptom onset which might be because of a longer asymptomatic period. In accordance with previous studies, although the RBD-and S-specific IgG titers peaked 14-30 days after infection and gradually decreased over time, the IgG antibody response stabilized after 6 months and was still detected in the majority of convalescent plasma donors at 6-15 months (Addetia et al., 2020;Jangra et al., 2021;Khoury et al., 2021;Liu et al., 2021). The sustained persistence of RBD-IgG titer over time suggests the generation of long-lived bone marrow plasma cells. Anti-S antibody titers were previously shown to correlate with the frequency of S-specific bone-marrow plasma cells of SARS-CoV-2 convalescent patients 7 to 8 months after infection (Turner et al., 2021). Only a low proportion of individuals had anti-S-and anti-RBD IgM or IgA antibodies more than 6 months after infection confirming the faster decrease of the specific IgM and IgA antibody response observed in other studies (Anand et al., 2021;Dan et al., 2021;Roltgen et al., 2020;Xiang et al., 2021) although a few studies reported a high prevalence of specific IgA up to one year Wang et al., 2021c). In accordance with a previous study, the persistence of IgG antibody level was associated with disease severity and patients with milder disease appeared to have more rapid IgG anti-RBD antibody waning . It has been reported that antibodies against SARS-CoV and Middle Eastern respiratory syndrome (MERS)-CoV, could still be detected 1-3 years after infection onset despite lack of reexposure to this virus . After a rapid decline of antibodies against SARS-CoV in the first two years, specific IgG have been detected in some patients up to 12 years after infection (Guo et al., 2020). A longer follow up will be necessary to evaluate if it is also the case for SARS-CoV-2-specific IgG antibodies as both viruses present different etiology.
Functional neutralizing antibodies specific to SARS-CoV-2 (anti-S and anti-RBD) that are produced following infection or vaccination are considered important for viral neutralization and viral clearance (Bergwerk et al., 2021;Poland et al., 2020). In the absence of definitive correlates of protective immunity, the presence of neutralizing antibodies against SARS-CoV-2 provides the best current indication for protection against reinfection (Addetia et al., 2020;Bergwerk et al., 2021;Khoury et al., 2021). As previously reported, the neutralizing ability of polyclonal plasma correlated positively with anti-S IgG or anti-RBD IgG (Roltgen et al., 2020;Sherina et al., 2021;Wajnberg et al., 2020). Plasma neutralizing activity reached a plateau after 4-6 months and was maintained in the majority of patients up to 15 months which was consistent with a previous study showing no significant difference in anti-RBD IgG antibody level and plasma neutralizing activities against Wuhan strain between 6 and 12 months (Wang et al., 2021c). In addition, the two-phase pattern with an initial rapid decay of neutralizing antibodies which slowed down to a flat slope (Pradenas et al., 2021) might explain why early studies with shorter follow-up (2-4 months) reported a fast decline of the antibody response (Ibarrondo et al., 2020). The long-lasting neutralizing activity might be caused by the accumulation of somatic mutations in IgG antibody genes and the production of antibodies with increased neutralizing potency (Pradenas et al., 2021;Wang et al., 2021c). Although no significant difference in neutralization activity was observed beyond 6 months, it might be important to predict how the long-term neutralizing antibody decay will affect the clinical outcomes following both natural infection and vaccination. However, a recent study in fully vaccinated health care workers showed that the occurrence of breakthrough infection was more associated with the peak of antibody and neutralization titers induced by the vaccine rather than on the subsequent antibody decay suggesting that the degree of protection might depend more on the initial immune response and the generation of memory B and T cells (Bergwerk et al., 2021).
Although there is evidence that SARS-CoV-2 seropositivity is associated with protection against the same strain (Harvey et al., 2021;Lumley et al., 2021), reinfection has been observed in some patients, which may represent nondurable protective immunity or infection with different variants (Breathnach et al., 2021;Hansen et al., 2021;Sabino et al., 2021). Notable mutations identified in the S1 subunit of the Alpha (Del69-70, N501Y, and P681H), Beta (K417N, E484K, and N501Y), Gamma (K417T, E484K, and N501Y), and Delta (L452R, T478K, P681R, and occasionally E484Q) variants may lead to small changes in the structure of the S protein and RBD (Davies et al., 2021a) leading to a higher infection rate and reduction of acquired immunity by neutralizing antibodies. As previously reported, we observed a substantial reduction of neutralization activity in the plasma of convalescent patients collected at early and more particularly, late phase of convalescence against the Beta, Gamma, and to a lower extent Delta (Wang et al., 2021a). The N501Y mutation in the Alpha variant RBD does not alter the neutralizing ability of plasma antibodies from naturally infected individuals , whereas the Beta (B.1.351) and Gamma (P.1) carrying the N501Y and E484 mutations are more resistant to the neutralizing activity from convalescent and vaccine immune sera as well as neutralizing antibodies Planas et al., 2021;Wang et al., 2021b,  iScience Article 2021c). The substitution at position E484 in the RBD of pseudovirus or recombinant virus is known to confer resistance to neutralization by convalescent human sera (Jangra et al., 2021;Liu et al., 2021). The L452R mutation found in Delta was shown to impair neutralization by antibodies (Starr et al., 2021), whereas the T478K mutation in the RBD is unique to Delta and close to the E484K mutation. Recent studies suggest that immunity from infection with prior lineage of the SARS-CoV-2 virus provides a lower protection against reinfection with Delta (Dhar et al., 2021), although it could still reduce disease severity and hospitalization (Sheikh et al., 2021).
Memory B and T cell immune responses with SARS-CoV-2 have been detected in convalescent individuals but clear correlates for protective immunity have yet to be defined. RBD and S-specific memory B cells are very rare in unexposed individuals but start to appear within 2 weeks after SARS-CoV-2 infection (Dan et al., 2021;Rodda et al., 2021;Sherina et al., 2021). RBD-specific and S-specific memory B cells steadily increased over the following months and were still present more than 6 months after initial infection (Dan et al., 2021;Sherina et al., 2021). To date, only one study has examined the RBD-specific memory B cell response up to 12 months after onset of symptoms (Wang et al., 2021c). Using ELISpot, we measured the number of B cells secreting IgG antibodies specific for SARS-CoV-2 RBD which are the most numerous and more likely to produce neutralizing antibodies. In accordance with our previous results (Wang et al., 2021c), we observed that while serum anti-RBD antibodies peaked 15-30 days postinfection and gradually decrease, RBD-specific memory B cells increased with time, reaching a maximum at 3-6 months postinfection, to a similar level induced in individuals receiving two doses of Pfizer vaccines. Previous studies also showed that RBD specific IgM + memory B cells formed the largest fraction of total memory B cells in the first month after symptom onset, but declined in frequency at later time points, whereas RBD specific IgG + memory B cells predominate at later time points with a peak at 3-4 months (Hartley et al., 2020). More importantly, the memory B cell response persists without further significant decrease for up to 15 months post-symptom onset, irrespective of disease severity. This study highlights that a decline in serum antibodies in convalescent patients may not reflect waning immunity, but rather a contraction of the immune response, with the development and persistence of virus-specific, long-lived memory B cells in the bone marrow. It has been shown that previous infection with SARS-CoV-2 increases both the number of RBD binding memory cells and neutralizing antibody titers after a single dose of vaccine (Goel et al., 2021;Planas et al., 2021;Wang et al., 2021c) suggesting the induction of plasma cells differentiation from the memory B cell compartment (Turner et al., 2021).
Early T cell responses during COVID-19 have been correlated with rapid viral clearance and reduced disease severity . In accordance with previous results, memory IL-2 and/or IFN-gÀproducing T cells specific to M, N nonstructural proteins, as well as S protein, were generated in the majority of convalescent patients following SARS-CoV-2 infection Rydyznski Moderbacher et al., 2020;Sekine et al., 2020;Tan et al., 2021). Functional SARS-CoV-2 specific T cells were detected at low levels within 2-4 weeks from onset of symptoms and reached a peak between 3 and 6 months at levels similar to that of fully vaccinated individuals (Addetia et al., 2020). Specific IFN-g producing Th1 CD4 + and cytotoxic CD8 + effector cells are considered important in supporting immunity against SARS-CoV-2 (Sattler et al., 2020). Polyfunctional T cells, secreting more than one cytokine, are also typically associated with superior control of pathogens and may be important to prevent reinfection by SARS-CoV-2 (Boyd et al., 2015;Owen et al., 2010). The presence of dual positive IFN-g and IL-2 S1-specific T cells that preferentially retain characteristics of both effector function and proliferative potential in vivo is indicative of strong and sustained S-specific T cell immunity. Similar to previous results (Dan et al., 2021), a decrease in specific T cell response, and more particularly in dual positive IFN-g and IL-2 S1-specific and S N M O-specific T cells, was observed after 6 months. Nevertheless, memory T cells were observed in the majority of patients between 12 and 15 months and a longer follow-up period with more participants will be necessary to evaluate the sustainability of the T cell response beyond 15 months postinfection.
Lasting immunity following acute viral infection and vaccination requires maintenance of both serum antibody and antigen-specific memory B and T lymphocytes, ranging from lifelong for smallpox or measles (Crotty et al., 2003), to more transient for common cold coronaviruses (Edridge et al., 2020). The majority of individuals in our study had three arms of immunity active against SAR-CoV-2, particularly from 2 to 15 months. We also found that the ability to mount high levels of RBD-specific IgG + memory B cells was associated with RBD-specific IgG antibodies and S1-specific and S N M O-specific T cells.  Bert et al., 2020). Our data suggest that T and B cell memory following SARS-CoV-2 infection might reach a more stable plateau, or slower decay phase, beyond 15 months postinfection. Nevertheless, long-term follow up study will be necessary to evaluate if the immunity to SARS-CoV-2 will be lifelong or will be more similar to that against endemic common cold coronavirus, where serum antibody responses decline and susceptibility to homologous virus reinfection occurs within 1-2 years (Edridge et al., 2020).
In conclusion, we observed that circulating memory B and T cells, and neutralizing antibodies are present in the majority of convalescent patients around 15 months after SARS-CoV-2 infection, demonstrative of a long-lasting immune response. Recent studies have shown that although the Pfizer-BioNTech and AstraZeneca vaccines were effective in reducing the risk of infection and COVID-19 hospitalization caused by Delta, these effects on infection appeared to be diminished when compared to those with the Alpha variant (Sheikh et al., 2021). Our neutralization data suggest that immunity develops during earlier waves of infection may not be fully protective against reinfection with Delta and other VOCs (and most likely against the recently isolated Omicron), indicating that convalescent patients may still benefit from vaccination. Furthermore, although two doses of vaccines could trigger rapid and robust immune responses as observed in naturally infected individuals, the longevity of immunity in vaccinated individuals should be followed up in future studies.

Limitations of the study
We are reporting the simplest statistical model that better fits our data. The magnitude of the antibody response over time reflects antibodies produced first by short-lived plasma cells and then long-lived plasma cells. More complex kinetics such as the power law model, which models a scenario in which the rate of antibody decay slows down over time could have been used. In addition, a higher number of individuals with more sampling time points over the 15 months period would have provided a more precise understanding of the kinetics of durability of SARS-CoV-2 antibodies. Other potential limitations to our study include the limited number of samples at the late time points (12-15 months) and a higher proportion of samples from severe patients at earlier time points (0-3 months) compared to later time points (12-15 months) (Table S3). This might have influenced the estimation of half-life of both antibody and neutralization responses although we also made a comparison between mild/moderate and severe/critical patients ( Figure S2). Furthermore, although based on available clinical data, no evidence suggests reinfection in any of the donors tested, the long duration of the immune response, including the increase in neutralization activity in some recovered participants, could theoretically be related to a natural boost after a reexposure to the virus.
Cell phenotyping using flow cytometry was not performed, and it was not possible to distinguish whether the T cells measured at early time points were effector or memory cells. Although Fluorospot cannot differentiate phenotype in a mixture of cell population, it is a more sensitive method than flow cytometry, partly because it can detect antibody or cytokine production from cells in 24-48 h. Furthermore, our findings are consistent with other studies predominantly based on mild to moderate disease cohorts, and where flow cytometry assays were used to measure the RBD-specific memory B cells and which similarly observed a reduction in T cell but not B cell SARS-CoV-2 specific response over a period of 8 months (Dan et al., 2021).

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:  Table S1 Blood samples (convalescent, healthy) Sweden, this paper Table S1 Serum (convalescent, healthy) Sweden, this paper Table S1 Buffy coat (healthy donor) Sweden, this paper

Lead contact
Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Harold Marcotte (harold.marcotte@ki.se).

Materials availability
RBD and S1-S2 proteins can be generated and shared on a collaborative basis.

Data and code availability
The article includes all datasets generated or analyzed during this study.
d Data reported in this paper will be shared by the lead contact upon reasonable request.

Study design and participants
Previously enrolled study participants were asked to return for a follow-up visit at the Fondazione IRCCS Policlinico San Matteo in Pavia, Italy and Karolinska Institutet, Stockholm, Sweden. Eligible participants were 18 years of age with a history of participation in prior study visit(s) of our longitudinal cohort study of COVID-19 recovered individuals . All participants had a confirmed history of SARS-CoV-2 infection, who had tested PCR-or serology-positive for SARS-CoV-2 . Of those patients, 7 Italian and 4 Swedish patients returned for a follow-up sample between February 1 and June 10, 2021. In addition, 21 Italian and 28 Swedish convalescent patients were recruited between November 18 and June 16, 2021. Study inclusion criteria included subjects over 18 years of age, who were willing and able to provide informed consent, confirmed positivity of SARS-CoV-2 by real-time RT-PCR targeting the E and RdRp genes according to Corman et al. (2020) protocols and monitored until two subsequent samples with negative results.
Blood sample was taken from 53 patients at one single time point between day 15 and 452 after symptoms onset while 7 patients had blood taken at two time points for a total of 67 samples including 20 between 241 to 452 days (9-15 months). Results obtained from the new recruited and new follow-up samples were merged with previously published data assessing the immune response to SARS-CoV-2 up to 6-8 months . Following merging, a total of 188 blood samples were collected from 136 patients, 98 Italians and 38 Swedish, for the entire 15-months follow-up ( Figure 1). Ninety-four donors had blood drawn at one single time point ranging from 7 to 452 days after symptom onset while 35, 5, and 1 donors had blood taken at two, three and four time points, respectively. For detailed participant characteristics see Tables S1).
Disease severity was defined as mild (non-hospitalized), moderate (hospitalized, with lower respiratory tract infection, with dyspnea or not, but without oxygen support), severe (infectious disease/sub intensive ward with a need for oxygen and/or positive chest computed tomography scan, severe lower tract infections, with any oxygen support) and critical (intensive care unit (ICU) patients, intubated or with extracorporeal membrane oxygenation procedures) .
The demographic and clinical characteristics of the patients for the entire 15 months follow-up are detailed in Table S1 and summarized in Table S2. The Italian patients, 58 (59.2%) males and 40 (40.8%) females, had a median age of 66.0 years (range 22-89). The degree of clinical severity of COVID-19 in the cohort was mild (n = 8), moderate (n = 17), severe (n = 67) and critical (n = 6). The Swedish patients had a median age of 44 years (range 18-75) with 16 (42.1%) males and 22 (57.9%) females and all 38 of them had mild symptoms.
In addition, serum samples from 108 anonymized individuals (16 to 80 years of age), collected before the SARS-CoV-2 pandemic (1995 to 2005) were used as historical negative controls for the ELISA. PBMCs from four healthy controls (median age 41 years, range 39-50) and seven additional buffy coats collected in Sweden before the SARS-CoV-2 pandemic (2011-January 2020) were included as negative controls for the B and T cell assays. Patients and samples tested in different assays are summarized in a flow chart (Figure 1).