2.1 Patients, control subjects and trial conduct

Between May 11, 2020 and August 20, 2020, when SARS-CoV-2 Wuhan Hu-1 was the only circulating virus strain, 133 patients diagnosed with COVID-19 disease were enrolled into this case–control study (Figure 1A). The 133 patients had rtPCR-confirmed (n = 116) and/or SARS-CoV-2 antibody-confirmed (Elecsys® Anti-SARS-CoV-2 assay Roche) (n = 131) COVID-19 disease.²⁵ From these 133 patients, one patient did not show any detectable antibody response against nucleocapsid-, S- or RBD-protein at 10 w and was therefore excluded from further analyses due to the probability of a false positive rtPCR test. Thus, the remaining cohort of 132 patients was analyzed 10 w (77.8 ± 24.6 days) and 10 m (9.5 ± 0.8 months) after infection (Table S1).

In parallel, 98 noninfected control subjects, who were reportedly asymptomatic for the last 10 w and who were SARS-CoV-2 negative by a certified SARS-CoV-2 antibody test (Elecsys® Anti-SARS-CoV-2 assay Roche) and had a negative rtPCR test for SARS-CoV-2 at the time of venipuncture were enrolled into the study. This cohort is identical to the one published in our previous report on the impact of COVID-19 on the immune system.²² Noninfected control subjects were well-balanced when compared to COVID-19 convalescent patients regarding demographic data, clinical background and drug intake (Table S2) and consisted of 44 males (44.9%) and 54 females (55.1%) with a median age of 51 years (range 14–77) (Table S1).

All individuals gave their written informed consent in accordance with the Declaration of Helsinki. The study was approved by the Ethics Committee of the Medical University of Vienna (EK No: 1302/2020). Venous blood was drawn from all subjects and was either EDTA-anticoagulated (for flow cytometric analyses), heparin-anticoagulated (for cryopreservation of PBMC), or silicon dioxide coagulated (to obtain serum for determining specific antibodies and cytokines).

Since infection-induced antibodies have the principal tendency to drop after an initial peak,²⁶ we have excluded all those COVID-19 convalescent subjects who presented with equal or elevated antibody levels at the end of the observational period of 10 m, to obtain a study population which had experienced only one SARS-CoV-2 infection. This resulted in a study population of 106 patients (Figure 1A). The 106 patients were further divided into three Groups according to their SARS-CoV-2 specific antibody decrease pattern as determined by k-means cluster analyses, which helps to partition a larger number of observations into distinct clusters with minimized within-cluster variances.²⁷ Accordingly, three clusters (Groups) could be assigned. Group 1 subjects were characterized by a low to moderate decrease of NC-specific (≤50%) with variable decreases of S-specific antibody levels; Group 2 showed a high decrease in NC-specific (>50%) but a low decrease in S-specific (≤40%; NC high/S low) antibody levels; Group 3 presented with high decreases of both NC(>50%)- and S(>40%)-specific (NC high/S high) antibody levels (Figure 1A).

2.2 Immunophenotyping by multiparametric flow cytometry

Immunophenotyping was performed by using optimal concentrations of directly conjugated monoclonal antibodies (Table S3A) to leukocyte and lymphocyte (sub) populations according to standard procedures²⁸ as described previously.²² Briefly, whole blood was analyzed freshly on the day of venipuncture (5–9 subjects per day) according to standard, quality-controlled procedures. For a small number of timepoints, some cell populations could not be resolved due to technical reasons, therefore the exact number of patients in the respective analyses is mentioned in each figure.

2.3 Determination of cytokine levels in human serum samples

Cytokines in human serum were determined as described.²⁹ Serum samples from healthy control subjects and COVID-19 patients at 10 w and 10 m were frozen at −80°C directly after sampling and analyses for cytokine levels were performed for all samples at the same time point. A panel of nine different bead regions in two batches with a total of 14 antibody pairs from Thermo Scientific (eBioscience), Biolegend or Miltenyi (Table S3B) were used for the analyses. Beads, 4 × 10⁶ microspheres (Luminex Cooperation, Diasorin, Austin, TX, USA), were coupled with 100 μg capture antibody (one antibody specificity per bead region), as recommended by the manufacturer. For analyses, human serum samples were thawed and centrifuged at 10,000 g for 10 min to remove insoluble precipitates. Subsequently, 30 μL of the serum aliquots were incubated in duplicates with 1.5 × 10³ beads per bead region coated with the respective capture antibodies in a total volume of 60 μL at 4°C overnight in the dark on a lab dancer in Multiscreen filter plates (MultiScreen® HTS BV, Merck Millipore, Burlington, MA, USA). On the next day, samples were washed with PBS (Gibco, Thermo Fisher, Waltham, MA, USA), and beads were incubated with 2.5 μg/mL per biotinylated secondary antibodies in a total of 25 μL recognizing the respective cytokines at room temperature for 1 h. Next, another washing step was performed followed by detection of bound antibodies with streptavidin-PE (Biolegend, San Diego, CA, USA) (2 μg/mL) in a total volume of 30 μL. Fluorescence intensities of individual bead populations were determined with a Luminex 100/200 apparatus (Luminex corporation, Austin, TX), related to standard curves obtained using known cytokine concentrations and absolute concentrations were calculated accordingly. The standard curves comprised 13 data points and ranged from 10,000 to 0.0564 pg/mL obtained by three-fold dilution steps.

2.4 Statistical analyses

For the determination of antibody decline the following formula was used: (antibody level 10 w—antibody level 10 m)/(antibody level 10w) × 100. Correlation analyses between the decline of NC- and S-antibody levels revealed three distinct clusters of antibody decline. The k-means algorithm²⁷ was used, to discriminate between similar and dissimilar antibody decline values and to confirm the grouping of each individual accordingly, by setting the expected number of clusters at three.

Prior to statistical evaluation, all metric variables underwent a distribution analysis. It was found that the majority of variables were best fit by a log-normal distribution and consequently log-transformed to obtain a normal distribution. Relative counts of T and B cells were best described by a Johnson type 2 distribution, suggesting a logistic transformation. Data from controls and COVID-19 patients (10 w and 10 m time points) were evaluated by Generalized Estimating Equation (GEE) models with unstructured correlation matrix. This procedure allows analyses of group comparisons and within-subject comparisons of time points in a simultaneous analysis. Groups and time-points were compared by linear contrasts with Sidak sequential correction of p values.

All analyses were performed using Stata 17.0 (StataSoft, College Stations, TX). In general, a p-value below .05 was considered significant. For easier presentation, in graphs we use * denoting p < .05, **denoting p < .01, ***denoting p < .001. Graphs (violin plots, line plots) were prepared by GraphPad 9.5 (GraphPad Software Inc., La Jolla, CA).

Additional material and methods are provided in Appendix A.

3.1 Characteristics of COVID-19 convalescent patients with short- and long-term follow up and noninfected control subjects

Here we investigated if there are long-term immunological imprints in a representative COVID-19 study population consisting of 133 individuals by comparing samples obtained 10 w and 10 m after disease onset (Figure 1A). Results obtained for noninfected control subjects were taken as reference. Of note, none of the COVID-19 convalescent patients were vaccinated between the 10 w and 10 m visit, since licensed vaccines became available only later. Furthermore, none of the 133 COVID-19 convalescent patients had reported symptoms of COVID-19 for the period between the two visits. One patient did not show any detectable antibody response against nucleocapsid-, S- or RBD-protein at 10 w and was therefore excluded from further analyses due to the probability of a false positive rtPCR test.

Demographic parameters (Table S1) and COVID-19 unrelated clinical characteristics (Table S2) were similar between the 132 COVID-19 patients and the 98 non-SARS-CoV-2-exposed controls. Of the 132 patients who were infected with the Wuhan Hu-1 virus strain, three patients presented with severe or critical (2.3%), 11 with moderate (8.3%) and 115 with mild (87.1%) disease course, while three remained asymptomatic (2.3%) according to the current WHO classification (https://app.magicapp.org/#/guideline/j1WBYn). The 132 patients were further stratified according to their SARS-CoV-2 serology (Figure 1A). Twenty-six patients were excluded from further analyses due to suspected re-infection, detected by an increase in SARS-CoV-2 specific antibody levels. The remaining 106 patients were stratified according to the observed antibody decline pattern (Figure 1B–D) into three groups (Figure 1A).

With regards to preexisting health conditions, 87 patients (65.9%) in the COVID-19 study group reported comorbidities, among them 68 patients (64.2%) in the antibody decrease group, which was comparable in frequency to 64 individuals in the noninfected control group (65.3%). Among the different comorbidities, allergic diseases in COVID-19 patients (36.4%) and in noninfected controls (43.9%) were the ones with the highest prevalence followed by cardiovascular diseases, metabolic diseases and chronic lung diseases. All other comorbidities (diabetes mellitus, hematopoietic diseases, immunosuppressive conditions, liver, neurological or renal diseases) were present only in few individuals (<10%).

Overall, 60 (45.5%) COVID-19 patients and 48 (49.0%) noninfected control individuals reported regular medication intake to alleviate the above-indicated comorbidities (Table S2). None of the patients reported current or former therapies with lymphoablative biologicals (e.g., rituximab, alemtuzumab, etc.). One patient reported a current cytoreductive treatment, however, this patient was excluded from analyses following the enrolment scheme (Figure 1A). Three individuals received systemic corticosteroids, two in the COVID-19 convalescent and one in the noninfected control group. Therefore, the authors can exclude the influence of medication intake on their data.

3.2 Serum IgG antibody levels specific for S, RBD, and NC decline strongly few months after the first SARS-CoV-2 infection, especially in younger individuals

First, we investigated the levels of SARS-CoV-2-specific IgG antibody levels over the period of 10 m. For this purpose, IgG antibody levels specific for the complete SARS-CoV-2 spike (S) protein, its receptor-binding domain protein (RBD) and the SARS-CoV-2 nucleocapsid protein (NC) were measured. We found that IgG antibodies specific for the latter three antigens were absent in one subject (Figure 1A), which was excluded from the study, and significantly decreased between 10 w and 10 m for the majority of subjects as it is usually expected after an initial peak value after the first infection¹⁵ (Figure 1B). However, 26 out of the remaining 132 subjects showed stable or even increased levels of S-, RBD-, or NC-specific antibodies (Figure 1B). Since subjects with such unusually stable or increased virus-specific antibodies may have experienced a clinically silent infection we stratified the COVID-19 convalescent subjects into two subgroups (Figure 1A). The larger group (106 patients) showed decreased (Figure 1C), while the smaller group (26 patients) showed equal or increased antibody reactivity with S-, RBD- or NC-proteins at the second visit at 10 m. The latter group was excluded from further analyses, because it was the goal of the study to investigate the long-term effects of the first SARS-CoV-2 infection in our study population.

Analysis of the remaining 106 COVID-19 convalescent subjects revealed that anti-S protein serum IgG levels significantly decreased from a mean OD₄₀₅ of 1.651 ± 0.852 at 10 w to 0.853 ± 0.570 at 10 m (p < .0001). Significantly, at 10 w only 3 out of 106 (i.e., 2.8%) COVID-19 convalescent patients had anti-S protein-IgG levels below the cutoff value. This number significantly increased by more than six-fold to 19 out of 106 (17.9%) convalescent patients at 10 m (p = .0005).

The decline of anti-RBD IgG antibody levels, which decreased from a mean OD₄₀₅-value of 0.825 ± 0.689 at 10 w to 0.166 ± 0.164 (p < .0001) at 10 m was pronounced (Figure 1C). While at 10 w, already 26 out of 106 (24.5%) convalescent patients had anti-RBD serum IgG levels below the cutoff-value, this number more than tripled to 86 out of 106 (81.1%) convalescent subjects at 10 m (p < .0001) (Figure 1C). However, none of the 106 subjects was NC negative, neither at 10 w nor at 10 m, thus, confirming that NC-specific IgG represents the more reliable marker for confirming a previous infection with COVID-19 provided that no NC-containing vaccine³⁰ or inactivated-virus vaccine (Sinovac,³¹ VLA 2001³²) was used.

Since RBD-specific antibodies are strongly associated with virus neutralization,^{23, 24} more than 20% of convalescent subjects 10 w after first infection and more than 80% of convalescent subjects after 10 m of the first infection may have had insufficient levels of SARS-CoV-2-neutralizing antibodies. Indeed, these results were confirmed by a MIA, which serves as surrogate for a virus neutralization test.^{23, 33} Of the 106 patients with a drop in S-, RBD- and NC-antibodies at 10 w, only 52 patients (49.1%) showed protective neutralizing antibody levels above the cutoff for 25% of inhibition, while at 10 m this number plummeted to 10 patients only (9.4%, p <.0001). The mean percent inhibition dropped from 25.8% at 10 w to 10.8% at 10 m. Only 20 patients showed an inhibition of more than 50% at 10 w, this number dropped to one patient at 10 m.

Likewise, anti-NC antibody levels dropped from a mean cutoff index (COI) value of 72 ± 44 at 10 w to 25 ± 28 (p < .0001) at 10 m (Figure 1C). The percentage decrease in NC antibody levels (Figure 1D) significantly correlated with the decrease in S antibody levels (r = .3461, p = .0003).

The correlation plot of antibody responses appeared to form three distinct patterns of antibody decline (Figure 1D). In fact, k-means algorithm cluster analysis applied to the two-parameter correlation revealed three different groups of patients according to their NC- and S-specific antibody decline pattern. Group 1 had a low decrease in NC (≤50%) and a mixed decrease in S antibody levels (NC low), Group 2 presented with a high decrease in NC (>50%) but a low decrease in S (≤40%; NC high/S low) antibody levels, while Group 3 showed a high decrease in both NC and S (NC high/S high) antibody levels (NC decrease >50%, S decrease >40%) (Figure 1D). Remarkably, the two extreme Groups 1 and 3 significantly differed in their mean age by 10.3 years (56.2 ± 11.5 vs. 45.9 ± 12.3, p = .0072) (Figure 1E). Notably, at 10 w NC antibody levels did not significantly differ between Groups 1 and 3 (86.7 ± 37.4 COI vs. 66.7 ± 45.0 COI; p = .1137) (Figure S1A), while S antibody levels were higher in Group 1 when compared to Group 3 (2.062 ± 0.663 vs. 1.557 ± 0.700 OD₄₀₅; p = .0089) (Figure S1B). Neutralizing antibody levels similarly dropped in Group 1 (12 [52.2%] vs. 3 [13.0%] patients with inhibition above 25%; p = .0106) and Group 3 (34 [7.6%] vs. 7 [11.9%] respectively, p < .0001). Interestingly, of the patients in Group 2, only 6 (25%) showed an initial blocking activity of more than 25%, while this was not the case in any of the patients at 10 m (Figure S1C).

3.3 COVID-19 impacts on blood leukocyte populations as late as 10 m after infection

Next, leukocyte populations were compared using a 15-parameter flow cytometry approach (Figure 2 and Tables S4A,S4B). In contrast to 10 w, COVID-19 patients had significantly lower total leukocyte counts at 10 m (6.533 ± 1.800 × 10⁶ cells/l vs. 5.749 ± 1.661 × 10⁶ cells/l, p < .0001) and involved neutrophils, monocytes and lymphocytes alike. Lower neutrophil counts in COVID-19 patients compared to noninfected controls were already observed at 10 w, but counts continued to decline until 10 m (4.156 ± 1.419 × 10⁶ cells/l vs. 3.694 ± 1.367 × 10⁶ cells/l, p < .001) (Figure 2). Similar to noninfected control subjects at 10 w, monocyte and lymphocyte counts were found to be lower in COVID-19 convalescent patients at 10 m as compared to 10 w (500 ± 177 × 10⁶ cells/l vs. 452 ± 189 × 10⁶ cells/l, p = .0068; and 1.897 ± 693 × 10⁶ cells/l vs. 1.624 ± 535 × 10⁶ cells/l, p < .0001) (Figure 2, Table S4A). Low lymphocyte counts were equally due to low T, B and NK cell numbers (Figure 2; Table S4A). The described changes were similar for Groups 1–3 for total leucocyte counts and had a similar tendency for total lymphocyte and granulocyte counts, while monocytes were reduced in Groups 2 and 3 but not Group 1 (Figure S2). Reduction of T cells at 10 m affected both the CD3⁺CD4⁺ helper and the CD3⁺CD8⁺ cytotoxic T cell subsets. Notably, the significant signs of T cell activation observed at 10 w²² affecting both the CD3⁺CD4⁺ and CD3⁺CD8⁺ T cell subsets and showcased by neo-expression of HLA-DR and CD38,²² largely disappeared at 10 m (Figure 2 and Figure S3A). Figure S3B shows representative dual parameter dot-plots for one control subject and a COVID-19 subject at 10 w and 10 m. Interestingly, at 10 m, the 10 patients within our study who required hospitalization due to severe COVID-19, still presented with elevated numbers of HLA-DR⁺ and HLA-DR⁺CD38⁺ T cells at 10 m, mainly within the CD3⁺CD8⁺ subset (Figure S3C) which is in accordance with other studies.³⁴ The described changes in T cell types and subsets thereof were similar for Groups 1–3 (Figure S2), which was clearly different for NK and B cells. In fact, CD56⁺CD3⁻ NK cell numbers were significantly lower at 10 m (265 ± 139 × 10⁶ cells/l to 216 ± 113 × 10⁶ cells/l, p < .0001) in COVID-19 convalescent subjects when compared to 10 w (Figure 2). However, this drop was mainly attributable to Group 2 (338 ± 220 × 10⁶ cells/l vs. 244 ± 129 × 10⁶ cells/l, p = .0021) and Group 3 (242 ± 99 × 10⁶ cells/l vs. 200 ± 107 × 10⁶ cells/l, p = .0015) (Figure S2A).

Also, B cell numbers seemed to decrease in the entire COVID-19 study group (Figure 2). However, while they plateaued in Group 1 at 141 ± 81 × 10⁶ cells/l vs. 140 ± 81 × 10⁶ cells/l, p = .8237, they significantly dropped only in Group 2 from 189 ± 142 × 10⁶ cells/l to 148 ± 87 × 10⁶ cells/l, p = .0338 and in Group 3 from 198 ± 92 × 10⁶ cells/l to 161 ± 183 × 10⁶ cells/l, p < .0001 (Figure S2B).

3.4 COVID-19 mainly depletes the pool of RTE T cells but not the central T cell memory as late as 10 m after infection

Given the significant reduction of T lymphocyte numbers observed at 10 m, we investigated effects on naïve and memory T cell populations. We found that the most naïve T cells, the RTEs, defined as CD3⁺CD45RA⁺CD62L⁺CD31⁺ T cells,³⁵ which were similar in numbers to those of the noninfected control subjects at 10 w, were almost cut to half at 10 m (230 ± 150 × 10⁶ cells/l, vs. 126 ± 107 × 10⁶ cells/l, p < .0001), which affected the respective CD4⁺CD45RA⁺CD62L⁺CD31⁺ helper T cell subset (142 ± 97 × 10⁶ cells/l vs. 68 ± 59 × 10⁶ cells/l; reduction: 52.1%; p < .0001) more drastically than the CD8⁺CD45RA⁺CD62L⁺CD31⁺ cytotoxic T cell subset (88 ± 66 × 10⁶ cells/l vs. 58 ± 54 × 10⁶ cells/l, reduction: 34.1%; p < .0001) (Figure 3A; Table S5A). The accuracy of flow cytometric RTE determination was self-validated by the fact that patients of younger age still had significantly more circulating RTE numbers when compared to older ones (p < .0001) (Figure S4). Representative results for individuals from different age groups, are shown in Figure S5 and revealed that the COVID-19-dependent drop in RTE numbers at 10 m is observable across all age groups investigated. Significantly, in 93 of 104 (89.4% the CD3⁺CD4⁺) and in 83 of 104 (79.8% the CD3⁺CD8⁺) of the convalescent patients, the PB RTE subsets were found to be reduced at 10 m when compared to 10 w (Figure 3A,B).

To clarify whether thymic output per se was the reason for the reduced numbers of circulating RTEs in COVID-19 convalescent individuals, TREC analyses on a subset of patient samples were performed. However, we found no evidence for significant reductions of TREC numbers during the observational period (Figure S6).

As already hypothesized above, the significant decrease of circulating RTEs at 10 m was paralleled by an increase of absolute numbers of CD3⁺CD45RO⁺CCR7⁺ central memory CD4⁺ and CD8⁺ T cell subsets, that is, from 144 ± 88 × 10⁶ cells/l to 176 ± 113 × 10⁶ cells/l, increase: 22%; p = .0041; and from 15 ± 14 × 10⁶ cells/l, to 39 ± 56 × 10⁶ cells/l, increase: 160%; p = .0001, respectively (Figure 3C,D; and Tables S5A,S5B). Of note, in 55 of 100 (55.0%) convalescent patients the CD4⁺CD45RO⁺CCR7⁺ and in 67 of 98 (68.4%) convalescent patients of the CD8⁺CD45RO⁺CCR7⁺ central memory T cell subsets were found increased at 10 m when compared to 10 w (Figure 3D). Notably, absolute numbers of overall CD45RO⁺ memory T cells were not increased at 10 m (Figure S7), indicating concurrent contraction of other memory T cell subsets, for example, distinct effector memory cell subsets.

Both, the CD3⁺CD4⁺CD127⁺ effector memory T cell increase and the Foxp3⁺ CD3⁺CD4⁺CD127⁻CD25⁺ T regulatory cell decrease observed at 10 w had the tendency to return to almost normal levels at 10 m. (Figure S8; Table S5B). In clear contrast, CD3⁺CD4⁺CD45RA⁺CD127⁺ cells remained elevated throughout the observational period (140 ± 130 × 10⁶ cells/l and 132 ± 115 × 10⁶ cells/l vs. 79 ± 106 × 10⁶ cells/l, p < .0001 and p < .0001 vs. HC, respectively) (Figure S8; Table S5A) and may reflect expanded CD4⁺TEMRA (terminal effector memory T cells which re-express CD45RA) cells.^{36, 37} Interestingly, the changes of T cell subsets were similar in the three Groups of patients with differential antibody decline, with regards to the significant loss of RTE (CD45RA⁺CD62L⁺CD31⁺) and the significant gain (except Group 2) of central memory T cells (CD3⁺CD45RO⁺CCR7⁺) of both CD4⁺ and CD8⁺ subsets (Figure S8,S9). In parallel, CD3⁺CD4⁺CD127⁺ effector T cells returned to normal levels in all three groups at 10 m.

3.5 Significant reduction of B cells and B cell subsets between 10 w and 10 m after first COVID-19

Similar to the decline of overall B cell numbers at 10 m (Figure 2), a significant reduction of overall circulating CD19⁺CD21⁺CD27⁺ memory B cells (51 ± 36 × 10⁶ cells/l vs. 39 ± 25 × 10⁶ cells/l, p = .0027) was observed at 10 m compared to 10 w (Figure 4A,D). B memory cell reduction was due to a significant decrease of non-class switched CD19⁺IgD⁺CD27⁺ memory B cells (29 ± 20 × 10⁶ cells/l vs. 23 ± 14 × 10⁶ cells/l, p = .0034) (Figure 4A,D; Table S6). These reductions were paralleled by normalization of CD19⁺IgM⁺CD38⁺ transitional B cell (5 ± 4 × 10⁶ cells/l vs. 8 ± 6 × 10⁶ cells/l, p = .0001) (Figure 4B; Table S6A) and CD19⁺IgM⁻CD38⁺ plasmablast levels (2.7 ± 3.7 × 10⁶ cells/l vs. 1.4 ± 1.1 × 10⁶ cells/l, p = .0121) (Figure 4B and Table S6A) compared to healthy controls. Another interesting finding was the fact that the subset of CD5⁺ B1-like marginal zone B cells was found to be severely reduced at 10 m in absolute (41 ± 37 × 10⁶ cells/l vs. 24 ± 22 × 10⁶ cells/l, p < .0001) (Figure 4C; Table S6A) and relative (Table S6B) numbers.

While Groups 2 and 3 presented with significant reductions of non-class switched B cells, a significant reduction of class-switched memory B cells was exclusively seen in Group 3 (Figure S10A–C). In contrast, the drop in CD5⁺ B cells was observed across all three Groups (Group 1, 34 ± 30 × 10⁶ cells/l to 24 ± 19 × 10⁶ cells/l, p = .0462; Group 2, 43 ± 55 × 10⁶ cells/l to 23 ± 24 × 10⁶ cells/l p < .0001; and Group 3, 42 ± 31 × 10⁶ cells/l to 25 ± 22 × 10⁶ cells/l, p < .0001), respectively (Figure S10D). The normalization of transitional B cell and plasmablast levels was similar among Groups 1–3 (Figure S10E,10F).

3.6 Systemic serum cytokine levels shift to a type 2-dominated pattern 10 m after COVID-19

Finally, we assessed serum Th1, Th2, Th17 and inflammatory cytokine levels. Figure 5A shows that levels of certain cytokines, which may be considered to be indicative of Th1-dominated inflammation (e.g., IL-1β, IL-8 and IL-12), were higher in sera of COVID-19-convalescent subjects 10 w after infection as compared to the noninfected control group. These alterations resolved 10 m after infection. However, in parallel to this resolution, Th2-dominated serum cytokine patterns emerged, as shown by the significantly elevated IL-4 and IL-10 serum levels (7.3-fold and 1.7-fold, respectively, p < .001 and p = .03, respectively) in sera obtained 10 m as compared to 10 w after infection (Figure 5A). The above-described changes at 10 m were accompanied by significantly lower IFN-γ (p < .0001) and IL-21 (p < .0001) levels compared to noninfected controls, while the overall low IL-17 levels did not change much. Accordingly, the respective IFN-γ/IL-4-ratios of COVID-19 convalescent patients significantly changed from 6.82 to 0.21 between 10 w to 10 m after infection (p < .0001) compared to the balanced ratio of 1.21 as observed in noninfected controls (p = .0486 and p < .0001, respectively). (Figure 5B). No evidence for reduced IFN-γ production on the single cell level was found (Figure S11). Interestingly, Group 1 patients presented with a less pronounced decline in IFN-γ levels, while they showed an increase in IL-10 levels at 10 m (Figure S12).