Generation and performance of the SARS-CoV-2 antigen array
The antigen array comprised the following SARS-CoV-2 proteins or protein fragments: commercial full-length S protein (SFL), super stable S trimer (SFL3), RBD (SRBD), S1 (SS1sub) and S2 subunits (SS2sub), N (NFL) and M (MFL) proteins, as well as 23 in-house produced protein fragments representing predicted epitopes within the S, M and N proteins, NSP1-5, ORF3b and ORF8 (Additional file 1). Human serum albumin and glutathione-containing elution buffers served as negative controls, while human serum IgG, IgA and IgM served as positive controls. A representative image of testing the antigen array with serum from COVID-19 patient is shown in Additional file 2.
Inter-assay reproducibility was assessed by testing the antigen array with two serum samples in three replicates on different days. The mean coefficient of variability (CV) for IgG antibodies was 0.204 and for IgA – 0.439 that is acceptable variability for immunoassays. Linear dynamic range of the IgG assay was assessed by testing the array with serial dilutions (1:100; 1:400; 1:1600 and 1:6400) of 4 serum samples. An example of the results is shown in Additional file 2. Results showed that the signal intensities for the majority of antigens were within the linear dynamic range for the whole range of dilutions, except the strongest signals that reached the saturation at 1:100 dilution. Serum dilution of 1:200 was chosen for the detection of IgG, IgA and IgM responses in order to reach the maximum sensitivity while maintaining the dynamic range of the assay. IgE is the least abundant antibody class constituting only approximately 0.05% of the serum immunoglobulins (11), therefore the detection of IgE responses to SARS-CoV-2 required lower dilutions of serum and secondary antibodies – 1:5 and 1:500, respectively.
Dynamics of antibody responses to various SARS-CoV-2 antigens
Dynamics of antibody responses against various SARS-CoV-2 antigens was studied in 78 longitudinal samples from 34 COVID-19 patients, including 27 samples that were collected ≤14 days after the onset of symptoms (DPS).
IgGs against SFL3, SFL, SRBD, SS1sub and SS2sub, MFL and NFL were already present in 66.7 to 81.5% of the samples collected before 14 DPS (SFL3, NFL and MFL are shown Figure 1). The remaining patients became seropositive in the second timepoint (days 15- 70). The IgG levels reached the plateau between days 12 and 20 and stayed relatively stable at least over 3 months. The timing of seroconversion for the other antigens was ranging from day 5 to 59. However, the frequency and mean signal intensity was also significantly lower; hence the later seroconversion may be related to higher limit of detection of these antibodies, not later antigen exposure or slower response of B cells.
IgM responses were tested in 8 patients. As expected, IgM levels against SFL3, SFL, SRBD, SS1sub and SS2sub, MFL and NFL were lower than those for IgG and were rapidly decreased starting from day18 in all except one case. This suggested that the IgM levels do not have a plateau phase and are not comparable between the groups of samples collected at various time-points, therefore the IgM responses were not tested in the remaining samples.
The kinetics of IgA responses was more heterogeneous among the patients. One part of the patients had anti-SFL3, SFL, S1sub and S2sub IgAs at the first time point and the levels tended to decrease in the subsequent time-points, whereas 5 patients were seronegative until 40-65 DPS and seroconverted before the subsequent blood draw (95-120 DPS). In some of the patients, IgA production against various forms of S protein happened simultaneously with NFL and MFL, whereas in other patients the production of IgAs against various antigens appeared to be uncoupled.
The timing of seroconversion of IgE responses against NFL was comparable with that of IgG responses and the anti-NFL IgE levels remained relatively stable over the time course of this study. Anti-SFL3 IgEs seroconverted later, however it is possible that the signal reached the detection limit only once it was close to the peak level. IgEs against other antigens were found in a substantially smaller fraction of patients and the assessment of seroconversion was not meaningful.
Prevalence of IgG, IgA and IgE antibodies against various SARS-CoV-2 antigens
The prevalence of IgG, IgA and IgE antibodies against various SARS-CoV-2 antigens was analyzed in serum samples collected from 73 COVID-19 patients >14 DPS (Table 1), when the majority of patients were expected to be seroconverted. In cases where multiple serum samples were available, the sample collected between 15 and 85 DPS was selected for this analysis.
Table 1
Frequencies of IgG, IgA and IgE responses in sera of COVID-19 patients collected >14 DPS, %.
Antigen
|
Mild, %
|
Moderate, %
|
Severe, %
|
Total, %
|
IgG
|
IgA
|
IgE
|
IgG
|
IgA
|
IgE
|
IgG
|
IgA
|
IgE
|
IgG
|
IgA
|
IgE
|
Full-lenght proteins or subunits
|
SFL3
|
94.7
|
16.7
|
66.7
|
100
|
38.5
|
92.3
|
100
|
54.5
|
90
|
97.3
|
30.1
|
79.7
|
SFL
|
97.4
|
25
|
12.1
|
100
|
42.3
|
50
|
100
|
54.5
|
60
|
98.7
|
35.6
|
33.3
|
SRBD
|
92.1
|
22.2
|
0
|
100
|
42.3
|
3.85
|
100
|
63.6
|
10
|
96
|
35.6
|
2.9
|
S1sub
|
89.5
|
22.2
|
0
|
100
|
42.3
|
34.6
|
100
|
63.6
|
20
|
94.7
|
35.6
|
15.9
|
S2sub
|
100
|
25
|
3.03
|
100
|
42.3
|
19.2
|
100
|
54.5
|
20
|
100
|
35.6
|
11.6
|
NFL
|
94.7
|
19.4
|
66.7
|
100
|
46.2
|
96.2
|
100
|
45.5
|
100
|
97.3
|
32.9
|
82.6
|
MFL
|
84.2
|
13.9
|
0
|
100
|
30.8
|
23.1
|
100
|
45.5
|
0
|
92
|
24.7
|
8.7
|
Predicted epitopes or protein fragments
|
S1
|
5.26
|
0
|
0
|
7.69
|
0
|
0
|
27.3
|
0
|
0
|
9.33
|
0
|
0
|
S2
|
31.6
|
0
|
0
|
26.9
|
0
|
0
|
72.7
|
0
|
0
|
36
|
0
|
0
|
S3
|
26.3
|
0
|
0
|
26.9
|
0
|
0
|
54.5
|
0
|
0
|
30.7
|
0
|
0
|
S4
|
5.26
|
0
|
0
|
15.4
|
0
|
0
|
36.4
|
0
|
0
|
13.3
|
0
|
0
|
S5
|
34.2
|
0
|
0
|
23.1
|
0
|
0
|
36.4
|
0
|
0
|
30.7
|
0
|
0
|
S6
|
13.2
|
0
|
0
|
15.4
|
0
|
3.85
|
0
|
0
|
0
|
12
|
0
|
1.45
|
S7
|
42.1
|
2.78
|
15.2
|
69.2
|
0
|
34.6
|
90.9
|
0
|
30
|
58.7
|
1.37
|
24.6
|
S8
|
13.2
|
0
|
0
|
15.4
|
3.85
|
0
|
18.2
|
0
|
0
|
14.7
|
1.37
|
0
|
S9
|
28.9
|
0
|
0
|
26.9
|
0
|
0
|
45.5
|
0
|
0
|
30.7
|
0
|
0
|
S10
|
15.8
|
0
|
0
|
15.4
|
0
|
0
|
18.2
|
0
|
0
|
16
|
0
|
0
|
S11
|
0
|
0
|
0
|
7.69
|
0
|
7.69
|
18.2
|
0
|
0
|
5.33
|
0
|
2.9
|
M1
|
18.4
|
0
|
0
|
38.5
|
0
|
0
|
54.5
|
0
|
0
|
30.7
|
0
|
0
|
N1
|
28.9
|
0
|
0
|
30.8
|
3.85
|
0
|
54.5
|
0
|
0
|
33.3
|
1.37
|
0
|
N2
|
42.1
|
0
|
0
|
53.8
|
0
|
0
|
63.6
|
0
|
0
|
49.3
|
0
|
0
|
NSP1
|
36.8
|
0
|
0
|
15.4
|
0
|
0
|
36.4
|
0
|
0
|
29.3
|
0
|
0
|
NSP2
|
7.89
|
0
|
0
|
11.5
|
0
|
0
|
27.3
|
0
|
0
|
12
|
0
|
0
|
NSP3
|
2.63
|
0
|
ND
|
11.5
|
0
|
ND
|
9.09
|
0
|
ND
|
6.67
|
0
|
ND
|
NSP4
|
2.63
|
0
|
ND
|
3.85
|
0
|
ND
|
0
|
0
|
ND
|
2.67
|
0
|
ND
|
NSP5
|
31.6
|
0
|
ND
|
34.6
|
0
|
ND
|
45.5
|
0
|
ND
|
34.7
|
0
|
ND
|
ORF3b
|
2.63
|
0
|
ND
|
3.85
|
0
|
ND
|
18.2
|
0
|
ND
|
5.33
|
0
|
ND
|
ORF8
|
7.89
|
0
|
ND
|
7.69
|
0
|
ND
|
18.2
|
0
|
ND
|
9.33
|
0
|
ND
|
ORF8_1
|
0
|
0
|
ND
|
0
|
0
|
ND
|
0
|
0
|
ND
|
0
|
0
|
ND
|
ORF8_2
|
5.26
|
0
|
ND
|
15.4
|
0
|
ND
|
27.3
|
0
|
ND
|
12
|
0
|
ND
|
IgG antibodies against the S2sub, SFL, SFL3 and NFL were found in >97.3% of the patients thus showing that these are the most suitable antigens for the production of serological tests for epidemiological studies. IgGs against the SRBD, S1sub MFL were also detected in 100% of the patients with severe and moderate disease, while the frequencies were ranging from 84 to 92% in patients with mild disease. From the individual predicted epitopes, S7 – a 32 aa. peptide (V551-L582) located in the SD1 subdomain of the RBD, showed the highest prevalence of IgG responses (58.7%) thus suggesting that this is the immunodominant B cell epitope within the RBD.
The frequency of IgA responses against SFL3, SFL, SRBD, SS1sub, SS2sub, MFL and NFL ranged from 24.7 to 35.6% and all the responses except for NFL showed the highest frequencies in patients with severe disease, followed by moderate and mild disease. IgAs against the predicted epitopes or other protein fragments were detected in a few cases, presumably because the antibody levels to monovalent antigens didn’t reach the limit of detection.
IgE antibodies were detected against NFL, SFL3, SFL, SRBD, SS1sub, SS2sub, MFL and S6, S7 and S11 epitopes, with NFL being the most frequently recognized antigen. Anti-NFL IgEs were detected in 100, 96.2 and 66.7% of patients with severe disease, moderate and mild disease, respectively. However, the IgEs were detectable only at serum dilution 1:5, hence the signal intensities and frequencies are not directly comparable with those of IgG and IgA.
Correlation of antibody levels with severity of the disease
To assess to what extent the antibody responses to various antigens are correlated, we performed Spearman correlation analysis (Figure 2). This analysis revealed a cluster of strongly correlated IgG responses: anti-S2sub, SRBD, S1sub, SFL, SFL3, S7, M and NFL (Spearman R >0.52; P <2.4⋅10−11), Similarly, IgA responses to M, NFL, S2sub, SFL3, SFL, SRBD and S1sub were highly correlated (Spearman R >0.49; P <2.7⋅10−10). IgE responses to S1sub were correlated to S2sub and SFL (Spearman R >0.33; P <2.2⋅10−4) and anti-NFL to SFL3 (Spearman R =0.53; P =4.5⋅10−10). We did not observe any significant negative correlation that would have suggested that the antibody responses to different antigens are mounted in a mutually exclusive manner.
Next, we assessed whether the antibody levels during the SARS-CoV-2 infection correlate with the disease severity. The mean levels of IgG, IgA and IgE antibodies against each antigen were compared between the groups of patients with severe and moderate vs mild disease in the samples collected >14 DPS (Figure 3A). The samples from severe and moderate groups were combined due to too small number of cases in the severe group that precluded the statistical analysis. For patients who had several sequential samples, the sample with the highest reactivity was included. This analysis revealed that IgG levels against the NFL, SRBD, SFL3, S1sub and SFL were significantly higher in patients with severe/moderate disease as compared with those with mild disease. Likewise, anti-SFL IgA and anti-S1sub, NFL and SFL IgE levels were significantly higher in the severe/moderate disease group (all adj. P <0.05). ROC curve analysis showed that anti-SFL3 IgGs could distinguish patients with severe/moderate disease from those with mild disease with AUC of 0.851 (P = 6.8⋅10−9), whereas anti-NFL IgE had an AUC of 0.809 (P = 1.6⋅10−6) (Figure 3B).
Next, the levels of antibodies against the most reactive antigens were compared between patients with moderate/severe vs mild disease in samples collected at 3-14 DPS. Results revealed that anti-SFL3 IgG and anti-SFL IgA levels were significantly higher in patients with moderate/severe disease than with mild disease (P < 0.05) (Figure 4A). However, ROC curve analysis showed that these markers could distinguish between these groups of patients with a moderate AUC of 0.67 (P = 0.05) and 0.629 (P = 0.05), respectively (Figure 4B).