Introductory paragraph
Understanding immune responses following SARS-CoV-2 infection in relation to COVID-19 severity is critical to predicting the effects of long-term immunological memory on viral spread. Here we longitudinally assessed systemic and airway immune responses against SARS-CoV-2 in a well-characterized cohort of 147 infected individuals representing the full spectrum of COVID-19 severity; from asymptomatic infection to fatal disease. High systemic and airway antibody responses were elicited in patients with moderate to severe disease, and while systemic IgG levels were maintained after acute disease, airway IgG and IgA declined significantly. In contrast, individuals with mild symptoms showed significantly lower antibody responses but their levels of antigen-specific memory B cells were comparable with those observed in patients with moderate to severe disease. This suggests that antibodies in the airways may not be maintained at levels that prevent local virus entry upon re-exposure and therefore protection via activation of the memory B cell pool is critical.
Main
Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection that causes coronavirus disease 2019 (COVID-19) can present with a wide range of disease severity from asymptomatic to fatal. Individuals of advanced age and/or those with comorbidities are overrepresented among those who develop severe disease1. However, the majority of SARS-CoV-2 infected individuals experience asymptomatic infection or only mild disease2. Whether patients with different disease severities generate similar protective immunity is still unknown. Here we present data generated from a clinically well-characterized cohort of individuals with SARS-CoV-2 infection (n=147) representing the full spectrum of COVID-19 severity ranging from asymptomatic infection to fatal disease. Individuals were sampled during acute disease and convalescence with longitudinal, matched blood and airway samples. The levels of systemic and airway antibody responses as well as the generation of SARS-CoV-2-specific memory B cells were measured. Plasma, peripheral blood mononuclear cells (PBMC), nostril swabs (NSW) and nasopharyngeal aspirates (NPA) were collected across all disease severities whereas endotracheal aspirates (ETA) were collected only from intubated patients receiving intensive care (Figure 1a). Disease severity was assessed on a seven-point scale derived from the respiratory domain of the sequential organ failure assessment (SOFA) score3,4, with additional levels for non-admitted mild cases (1) and fatal cases (7). Patients were grouped based on peak disease severity (PDS) (Supplementary table 1 and online methods). In addition, pre-pandemic healthy controls (PPHC) (n=30) as well as individuals who experienced influenza-like symptoms and were possibly exposed to SARS-Cov-2 but had a negative diagnostic PCR (PCR-) (n=9) were sampled in the same way and included as controls (Supplementary table 2 and online methods).
a) Overview of study cohort and controls, timeline of longitudinal sampling, the anatomical compartments analyzed, and the measurements performed. AS=asymptomatic. b) Scatter plots show individual levels of plasma IgG and IgA (from left to right) in SARS-CoV-2 infected individuals with different peak disease severity (PDS). Data in cyan and green refer to mild disease (PDS 1 and 2), yellow and orange refer to moderate disease (PDS 3 and 4), red and cayenne refer to severe disease (PDS 5 and 6) and grey refers to patients with fatal outcome (PDS 7). Black lines indicate medians and dotted lines indicate the average background level from PPHC. Mann-Whitney U was used to compare the groups. c) Spearman correlation for plasma immunoglobulins against the RBD versus days from onset of symptoms during the acute and the convalescent phases are shown. Circles with black lining refer to convalescent samples. d-e) Longitudinal measurements of plasma IgG against RBD on a subset of patients with moderate/severe disease and fatal outcome who had low antibody titers at the time of study inclusion. Levels are shown at the time of study inclusion, during the late acute phase and at convalescence, and shown with respect to days from onset of symptoms and as a group comparison. The black lines connect data points from the same individuals. Wilcoxon test was used to compare the groups separately.
We first assessed systemic IgG and IgA responses at the time of study inclusion that ranged between 0-54 days from onset of symptoms; median 16 days, and at the first follow-up visit during convalescence (46-168 days; median 108 days). Plasma IgG and IgA against the SARS-CoV-2 nucleocapsid (N) and spike (S) proteins as well as the receptor binding domain (RBD)5,6 of the S protein were measured by ELISA. Antibody responses against the internal N protein have been shown to be elevated in deceased individuals but whether these antibodies contribute to disease severity is unknown7,8. In contrast, responses against the viral surface protein S and, in particular, against the RBD result in virus neutralization9. Responses against the RBD are thus likely necessary for protection from re-infection or prevention of symptomatic disease.
In line with previous reports7,8,10-12, IgG and IgA responses against N, S and RBD were robust in acute disease in the majority of individuals with moderate to severe disease while they were substantially lower in individuals with mild disease (Figure 1b). However, in our cohort, this might also partially be due to earlier study inclusion of patients with mild symptoms (median days from onset of symptoms 11 as compared with 13 and 21.5 in the moderate and severe groups respectively, and 13 in the fatal group) (Supplementary table 1). In fact, while IgG levels remained high in the patients with moderate to severe disease in the convalescent phase, levels had increased in the individuals with mild disease. In contrast, IgA levels from the acute phase, against all antigens, waned substantially during convalescence in most patients (Figure 1b and supplementary fig. 1a). While the levels of IgG against the RBD (Figure 1c), as well as against N and S (Supplementary fig. 1b), exhibited a positive correlation with days from onset of symptoms during the acute phase, this was less pronounced for IgA levels. Altogether, these data confirm that the generation of IgA likely precede that of IgG. Earlier reports of individuals with asymptomatic infection or mild disease also showed robust early IgA responses13-15.
Comparably to another study16, the levels of both IgG and IgA against all antigens tested, during both acute disease and convalescence, correlated directly with disease severity (Supplementary fig. 2a), indicating that individuals with severe disease or even fatal outcome mounted the highest antibody responses against SARS-CoV-2 antigens. In our cohort, this is further supported by the fact that the patients with moderate/severe disease and fatal outcome for whom we initially observed low IgG titers against RBD, had an early study inclusion (on average 13 days from onset of symptoms); but showed significantly higher titers later during the acute phase (on average 19 days) (Figure 1d-e). Similar kinetics of anti-RBD responses have previously been noted in different patient cohorts17-19.
a) Scatter plots of the levels of IgG and IgA to RBD in the upper airway. The black lines indicate median values. Mann-Whitney U was used to compare the groups. b) Heat map generated by sorting data according to PDS. The heat map also includes data from PPHC and PCR-as a reference (indicated with PDS 0). Missing data and not available/applicable samples are shown in black. c) Comparison of the levels of RBD IgG/A in different upper and lower airway compartments. Data on plasma are shown as a reference. The black lines connect data points from the same individuals. Wilcoxon test was used to compare the groups separately. d) Representative example with gating strategy of SARS-Cov-2-specific memory B cells from the four different categories analyzed, and example of further characterization on RBD binding and B cell isotype (IgG+ or IgA+ assumed to correspond to IgD-IgM-IgG-B cells). e) Scatter plot shows the quantification of S-specific memory B cells in matched acute (filled) and convalescent (filled with black lining) PBMCs in relation to days in the subset of individuals analyzed color-coded according to PDS. f) Pie charts showing the cumulative characterization of RBD binding and memory B cell isotypes. The dotted lines on a) and e) indicate the average background level from PPHC. In a) the line overlaps with not detected (ND) for IgG levels.
In line with previous reports, the Charlson Comorbidity Index (CCI) and the Body Mass Index (BMI) were associated with disease severity1 in the cohort (Supplementary fig. 3). However, the presence of comorbidities did not impact the ability to generate antibodies against any of the antigens analyzed (Supplementary fig. 2b).
As the respiratory tract is the initial site of viral infection and replication, we next measured the levels of IgG and IgA in the upper and lower airways and compared with levels in plasma at matched time points. Due to limited sample volumes, we focused our analyses on IgG and IgA responses against the RBD since these responses are likely most critical for virus neutralization. We found that RBD-specific antibodies could be readily detected in NSW and NPA during the acute phase in several patients across all disease severities (Figure 2a-b and supplementary fig. 4a). In agreement with our observations in plasma, antibody levels in the upper respiratory tract were higher in patients with moderate or severe disease as compared with individuals with mild disease (Figure 2a-b). Both IgG and IgA levels declined significantly in the convalescent phase, with IgG declining to almost undetectable levels (Figure 2a-b). This demonstrates that airway antibody levels wane much faster than those in plasma during convalescence. Low but detectable levels of antibodies to SARS-CoV-2 have previously been reported also in saliva during convalescence20. However, whether these low antibody levels at mucosal sites will be sufficient for protection is not known. We found that RBD-specific IgG and IgA levels in the respiratory tract correlated well with those in plasma during the acute phase but to a lesser extent during convalescence (Supplementary fig. 4b and supplementary fig. 5). When comparing matched NSW, NPA and ETA collected at the same time point during acute disease from intubated patients, significantly higher levels of IgA against the RBD were found in NPA as compared with NSW and ETA (Figure 2c). While this could be partially influenced by differences in sampling method and sample volume, these data suggest that antibody abundance and possibly virus neutralization via IgA differs along the respiratory tract and may be more pronounced in the nasopharynx compared to the lower airways. Hence, nasopharynx antibodies (both IgG and IgA) showed a strong correlation with plasma antibody responses (Supplementary fig. 4b and supplementary fig. 5). We also assessed the presence of B cells in the respiratory tract of COVID-19 patients by analyzing the lymphocytes that could be retrieved from NPA and ETA as compared with NPA from three healthy controls (HC). Lymphocyte frequencies were lower in both NPA and ETA from several COVID-19 patients as compared with NPA from HC. However, the relative proportion of B cells was not different. In addition, we observed that in the COVID-19 patients, the frequencies of B cells were higher in NPA as compared with ETA which possibly contributes to the higher antibody levels at this site (Supplementary fig. 6).
Altogether, the data presented so far confirm that moderate and severe COVID-19 result in high levels of circulating antibodies and show that despite IgG being well-maintained during convalescence, antibody levels in the airways decline significantly after the acute phase. Generally, antibodies present in circulation and at local sites are the result of secretion from short-lived plasmablasts and/or terminally differentiated plasma cells in the bone marrow or mucosal sites21. The response to a secondary infection once antibody titers have waned below protective levels mostly relies on the presence of resting antigen-specific memory B cells that are rapidly activated upon antigen re-exposure21. We therefore investigated the induction and maintenance of antigen-specific memory B cells similar to other studies22-24. We focused on the direct comparison between individuals with mild disease and patients with moderate/severe disease, along with individuals who had reported mild influenza-like symptoms but were SARS-CoV-2 PCR-. Patients with moderate/severe disease who had high circulating IgG and IgA levels were specifically selected for the analysis to be able to compare the opposite ends of the COVID-19 disease spectrum. Matched PBMC from acute disease and convalescence were analyzed side-by-side using fluorescently labelled S and RBD probes to detect antigen-specific B cells22-24. Patients with moderate/severe disease had switched memory B cells specific to S in the acute phase and the memory B cell pool had further expanded in the convalescent phase (ranging from 0.009 to 1.35%; mean 0.42% during convalescence) (Figure 2d-e). Individuals with mild disease showed lower memory B cells during acute disease than the patients with moderate/severe disease, but the levels had increased by the time of convalescent sampling (ranging from 0.17% to 0.64%; mean 0.35% during convalescence) and were comparable between the groups (Figure 2d-e and supplementary fig. 6). Further phenotyping of the S-specific memory B cells indicated that the majority of these cells may be specific for epitopes on S outside of the RBD (Figure 2f). However, it is possible that binding of B cells to RBD could be underestimated as RBD is also present in the S protein. S-specific memory B cells in the circulation were predominantly IgG+, rather than IgA+ (Figure 2f). Low frequencies of S and RBD-specific memory B cells were observed in the PCR-individuals (Figure 2e). However, these were not significantly different from the levels observed in PPHC (Supplementary fig. 7). Collectively, since the majority of individuals infected with SARS-CoV-2 are either asymptomatic or experience only mild COVID-19 symptoms2, the possibility of generating antigen-specific memory B cells without experiencing severe disease, would be very important in the prospect of establishing potential immunity at the population level.
In summary, here we show that COVID-19 disease severity not only determines the magnitude of systemic but also airway antibody levels with efficient generation of virus-specific memory B cells against SARS-CoV-2 also occurring upon mild disease. While plasma IgG levels were generally well detectable at convalescence in all groups, there was a significant decline in airway antibodies after the clearance of infection. This suggests that antibodies in the airways may not be maintained at levels that prevent local virus entry upon re-exposure. However, our data indicate that the majority of infected individuals have the ability to generate anamnestic responses via the memory B cell pool and thereby may be protected or mitigate disease severity. Whether sufficient number of memory B cells will be maintained long-term and to what extent they will prevent the spread of SARS-CoV-2 at the population level remain to be understood. This is critical knowledge to acquire in the near future to evaluate together with the memory B cell response generated after the introduction of SARS-CoV-2 vaccination. Ultimately, the requirements for establishment of long-term protection and immunity will need to be determined.
Methods
Study design, patient enrollment and sample collection
One hundred and forty-seven (147) PCR-confirmed SARS-CoV-2 infected patients were enrolled at the Karolinska University Hospital and Haga Outpatient Clinic (Haga Närakut), Stockholm, Sweden during March-May 2020 (acute phase) and during April-September 2020 (convalescence). Patients were enrolled at various settings, ranging from primary to intensive care. In order to recruit asymptomatic and mild cases, household contacts of COVID-19 patients were screened with PCR and enrolled if positive. A small subset of these individuals who experienced influenza-like symptoms and were possibly exposed to SARS-Cov-2 but had a negative diagnostic PCR (PCR-) (n=9 of whom 3 were household contacts of confirmed patients, 4 had symptoms including fever and 2 had symptoms but no fever) were sampled in the same way and included as controls alongside with 30 pre-pandemic healthy control samples (PPHC) from 2016-2018.
Respiratory failure was categorized daily according to the respiratory domain of the Sequential Organ Failure Assessment score (SOFA)3. The modified SOFA score (mSOFA) was calculated when arterial partial pressure of oxygen (PaO2) was not available. In this case peripheral transcutaneous hemoglobin saturation (SpO2) was used instead4. Estimation of the fraction of inspired oxygen (FiO2) based on O2 flow was calculated as per the Swedish Intensive Care register definition25. Patients were categorized based on the peak respiratory SOFA or mSOFA value with the 5-point respiratory SOFA score being extended with additional levels to include and distinguish admitted asymptomatic and non-admitted/admitted mild cases and to include fatal outcome added as a seventh level. Ten (10) patients with fatal outcome had peak disease severity score 6 prior to death and 2 patients had scores of 4 and 5. For convenience, the resulting 7-point composite peak disease severity (PDS) was condensed into a broader classification consisting of mild (1-2), moderate (3-4), severe (5-6), and fatal (7). Demographics and additional data were collected from medical records, including clinical history and risk factors such as BMI and co-morbidities. Total burden of comorbidities was assessed using the Charlson co-morbidity index (CCI)26 (Supplementary table 1). Additional clinical information on this patient cohort can also be found in Falck-Jones et al27.
Blood was collected in EDTA-containing tubes from all patients except those admitted to the intensive care unit (ICU) for whom blood was pooled from heparin-coated blood gas syringes discarded in the last 12 hours. For some ICU patients, additional venous blood was also collected in EDTA tubes. Nostril swabs (NSW) and nasopharyngeal aspirates (NPA) were collected from the majority of the patients whereas endotracheal aspirates (ETA) were only collected from patients with mechanical ventilation intubated in the ICU. Admitted patients were sampled during acute disease at up to four timepoints and ICU patient material was collected up at to ten timepoints. For this study, unless otherwise stated, the measurements referring to acute disease were performed with samples collected at the time of study inclusion and during convalescence when patients returned for a follow-up visit.
The study was approved by the Swedish Ethical Review Authority, and performed according to the Declaration of Helsinki. Written informed consent was obtained from all patients and controls. For sedated patients, the denoted primary contact was contacted and asked about the presumed will of the patient and to give initial oral and subsequently signed written consent. When applicable retrospective written consent was obtained from patients with non-fatal outcomes.
Enzyme-linked immunosorbent assay (ELISA)
The presence of IgG or IgA binding against the SARS-Cov-2 Nucleocapsid (N) and Spike (S) trimer or the Receptor Binding Domain (RBD) monomer5 in plasma and airway samples was assessed by enzyme-linked immunosorbent assay (ELISA). Recombinant proteins were received through the global health-vaccine accelerator platforms (GH-VAP) funded by the Bill & Melinda Gates Foundation. Briefly, 96-half well plates were coated with 50ng/well of the respective protein. Plates were incubated with a selected duplicate dilution that did not provide background noise against ovalbumin used as a negative control (data not shown) (i.e. 1:20 for plasma samples, 1:2 for NSW and NPA, and 1:5 for ETA in 5% milk/PBS buffer). Duplicate 7-point serial dilutions were initially performed for measuring plasma IgG against RBD during acute disease and the half maximal effective concentration (EC50) was calculated using GraphPad Prism 9. However, since for several samples with low antibody concentration (mostly from the asymptomatic/mild category) the EC50 was below the highest dilution used (of 1:20) and therefore below the limit of detection (Supplementary fig. 8a), the maximal optical density (OD) at 1:20 dilution was used instead for this and for all the other measurements subsequently performed. The relation with maximal OD and EC50 was also verified in a subset of patients with high IgG and IgA against S (Supplementary fig. 8b). Detection was performed with mouse and goat anti-human IgG or IgA HRP-conjugated secondary antibodies (clone G18-145 from BD Biosciences and polyclonal from ThermoFisher, respectively) followed by incubation with TMB substrate (BioLegend) which was stopped with a 1M solution of sulfuric acid. Blocking with 5% milk/PBS buffer and washing with 0.1% Tween-20/PBS buffer were performed between each step. Absorbance was read at 450nm and background correction at 550nm using an ELISA reader. Data were reported as maximal absorbance i.e. OD, as stated above, and plotted using GraphPad Prism 9. All of the antibody measurements in plasma and respiratory samples from SARS-CoV-2 patients were run alongside with samples from two different control groups as described above. Interestingly, low but readily detectable IgA reactivity against S was detected in the pre-pandemic healthy controls and in the PCR-individuals (Supplementary fig. 8b). After having verified the specificity and sensitivity of our ELISA assay for IgA detection with limiting sample dilutions (Supplementary fig. 8c), we hypothesize that this might be due to cross-reactivity on the shared portions of the S protein between SARS-CoV-2 and other common cold coronaviruses. Reports have shown that cross-reactivity between coronaviruses exists28,29.
Flow cytometry
Staining of cells from airway samples was performed fresh. Briefly, samples were centrifuged at 400 g for 5 min at room temperature and cells were washed with sterile PBS. Mucus was removed using a 70 μm cell strainer and cells were subsequently stained with the appropriate combination of fluorescently labelled monoclonal antibodies as illustrated in Supplementary table 3a. Staining of PBMC was performed on previously cryopreserved samples. The appropriate combination of fluorescently labelled monoclonal antibodies binding to different cell surface markers and with fluorescently labelled S and RBD proteins used as probes for antigen-specific B cells is illustrated in Supplementary table 3b. Probes were prepared from biotinylated proteins using a 4:1 molar ratio (protein:fluorochrome-labelled streptavidin) considering the molecular weight of protein monomers and of the streptavidin only. The probes were prepared using streptavidin conjugated to PE and APC for S and with BV421 for the RBD. The gating strategy for the identification of antigen-specific memory B cells is shown in Figure 2d. Briefly, after identification of lymphocytes in single suspension, live B cells, (i.e. cells not expressing CD3/,CD14/CD16/CD56) were gated. From this gate, B cells were further isolated by expression of CD19 and CD20 and then switched memory B cells were identified as IgD-IgM-. From these, S-specific switched memory B cells were identified by binding to both S protein probes. Further characterization was then carried out by analyzing IgG expression (IgA+ switched memory B cells are assumed to mirror IgD-IgM-IgG-B cells) and fluorescently labelled RBD. Stained cells from airway samples were acquired using a BD LSRFortessa while stained PBMC were acquired using a BD FACSAria Fusion both interfaced with the BD FACSDiva Software. Results were analyzed using BD FlowJo version 10.
Statistical analyses
All statistical analyses were performed using GraphPad Prism 9. Spearman correlation was used to assess the interdependence of 2 different non-categorical parameters across individuals whereas Wilcoxon matched-pairs signed rank or Mann–Whitney U tests as appropriate, were used to assess differences or similarities for one single parameter between 2 different groups. Kruskal-Wallis with Dunn’s multiple comparisons test was used when assessing comparison between multiple groups.
Data Availability
The data that support the findings os this study are available from the corresponding author on reasonable request.
Author contributions
Experimental study design: A.C., Ka.L. and A.S-S. Clinical concept design: A.S-S., M.B., N.J., J.A., J.S. and A.F. Acquisition and sample processing: M.Y., S. F-J., S.V., B.Ö., E.Å., L.A., R.F-J., M.Ö. and F.G. Generation of data: A.C., M.Y., S. F-J., S.V. and P.C.G. Provision of custom reagents: D.P., M.M., L.C., and N.P.K. Analysis and interpretation of data: A.C., Ka.L. and A.S-S. Critical revision of the manuscript: all authors. Statistical analysis: A.C., K.L. and S.O. Ka.L. and A.S-S. contributed equally to the study.
Competing interests
The authors declare no competing financial interests.
Acknowledgments
We thank the patients and healthy volunteers who have contributed to this study. We would also like to thank medical students and hospital staff for assistance with patient sampling and collection of clinical data, the Biomedicum BSL3 core facility, Karolinska Institutet and Fredrika Hellgren for assistance with English editing. This work was supported by grants from the Swedish Research Council, the Swedish Heart-Lung Foundation, the Bill & Melinda Gates Foundation, the Knut and Alice Wallenberg Foundation through SciLifeLab and Karolinska Institutet.
References
