Identification of SARS-CoV-2-derived HLA class I- and HLA-DR-binding peptides
A novel prediction and selection workflow, based on the integration of the algorithms SYFPEITHI and NetMHCpan, identified 1,739 and 1,591 auspicious SARS-CoV-2-derived HLA class I- and HLA-DR-binding peptides across all 10 viral open-reading frames (ORFs, Fig. 1a, Extended Data Fig. 1a, b). Predictions were performed for the 10 and 6 most common HLA class I (HLA-A*01:01, -A*02:01, -A*03:01, -A*11:01, -A*24:02, -B*07:02, -B*08:01, -B*15:01,-B*40:01, and -C*07:02) and HLA-DR (HLA-DRB1*01:01, -DRB1*03:01, -DRB1*04:01,-DRB1*07:01, -DRB1*11:01, and -DRB1*15:01) allotypes covering 91.7% and 70.6% of the world population with at least one allotype, respectively22,23 (Extended Data Fig. 1c and 2a). To identify broadly applicable SARS-CoV-2-derived T-cell epitopes, we selected 100 SARS-CoV-2-derived HLA class I-binding peptides comprising 10 peptides per HLA class I allotype across all 10 viral ORFs for immunogenicity screening (range 3 - 20 peptides per ORF, mean 10, Fig. 1b, c, Extended Data Fig. 1d-m, Supplementary Table 1). In addition, 20 SARS-CoV-2-derived promiscuous HLA-DR-binding peptides across all ORFs from peptide clusters of various HLA-DR allotype restrictions representing 99 different peptide-allotype combinations were included (Fig. 1d, e, Extended Data Fig. 2b-k, Supplementary Tables 2 and 3). Of these HLA-DR-binding peptides, 14/20 (70%) contained embedded SARS-CoV-2- derived HLA class I-binding peptides for 7/10 HLA class I allotypes. The complete panel of 120 SARS-CoV-2-derived peptides comprised 10% of the total SARS-CoV-2 proteome (57%and 12% of nucleocapsid and spike protein, respectively; Extended Data Fig. 2l) and showed an equally distributed origin of structural ORF proteins (61/120 (51%)) encompassing spike, envelope, membrane and nucleocapsid proteins as well as non-structural or accessory ORFs (59/120 (49%)). The broad HLA class I and HLA-DR allotype-restriction of the selected SARS-CoV-2-derived peptides allowed for a total coverage of at least one HLA allotype in 97.6% of the world population (Fig. 1f). Recurrent mutations of SARS-CoV-224,25 affected only a minority of selected SARS-CoV-2-derived peptides with 14/120 (12%) sequences (1.7% at anchor position) including reported mutation sites (Fig. 1g, Supplementary Tables 4 and 5).
Validation and characterization of SARS-CoV-2-derived CD8+ and CD4+ T-cell epitopes
IFNγ ELISPOT screening of SARS-CoV-2 convalescents (SARS, group 1, n = 116, Extended Data Table 1, Supplementary Table 6) and donors never exposed to SARS-CoV-2 (PRE, group A, n = 104, samples collected prior to SARS-CoV-2 pandemic, Extended Data Table 1, Supplementary Table 7) validated 29/100 (29%) SARS-CoV-2-derived HLA class I- (3/10 HLA- A*01, 2/10 HLA-A*02, 3/10 HLA-A*03, 2/10 HLA-A*11, 5/10 HLA-A*24, 2/10 HLA-B*07, 4/10 HLA-B*08, 0/10 HLA-B*15, 5/10 HLA-B*40, 3/10 HLA-C*07) and 20/20 (100%) HLA-DR-binding peptides as naturally occurring T-cell epitopes (Fig. 2a-f, Extended Data Tables 2 and 3, Supplementary Fig. 1 and 2, Supplementary Table 8). Flow cytometry revealed that T-cell responses directed against HLA class I-binding peptides were mainly driven by IFNγ+CD8+ T cells, whereas HLA-DR-binding peptides were recognized by multifunctional (IFNγ+TNF+CD107a+) CD4+ T cells and in single donors additionally by CD8+ T cells (Fig. 2b, d). 12/29 (41%) and 11/20 (55%) SARS-CoV-2-derived CD8+ and CD4+ T-cell epitopes were dominant epitopes (recognized by ≥ 50% of SARS donors) with recognition frequencies up to 83% (A01_P01) and 95% (DR_P16), respectively (Fig. 2e, f, Extended Data Tables 2 and 3).
T-cell responses showed high inter-individual as well as inter-peptide intensity variation(Supplementary Fig. 3). Overall, the intensity of HLA-DR-specific T-cell responses in the SARS group was significantly more pronounced compared to those directed against HLA class I T-cell epitopes (median 414 versus 56 calculated spot counts, Fig. 2g). All SARS- CoV-2-derived HLA-DR-binding peptides were found to be immunogenic, independently of the source ORF. SARS-CoV-2-derived HLA class I T-cell epitopes showed an equally distributed origin from structural (13/29 (45%)) and non-structural or accessory (16/29 (55%)) ORFs (Extended Data Table 2). However, ORF-specific differences regarding the proportion of validated HLA class I T-cell epitopes were observed, revealing the highest frequencies for ORF9 (50%, nucleocapsid protein), ORF1 (45%), and ORF3 (38%, Fig. 2h). The highest recognition rate in SARS donors was observed for HLA class I T-cell epitopes derived from ORF2 (55%, spike protein), ORF5 (52%, membrane protein), and ORF3 (45%), as well as for HLA-DR T-cell epitopes derived from ORF5 (95%, membrane protein), ORF8 (68%), and ORF4 (55%, envelope protein, Fig. 2i).
Cross-reactive T-cell responses to SARS-CoV-2-derived HLA class I and HLA-DR T-cell epitopes in unexposed individuals
Upon screening PRE group A, cross-reactive T-cell responses to 9/29 (31%) of the validated HLA class I and to 14/20 (70%) HLA-DR T-cell epitopes were detected. Recognition frequencies of single SARS-CoV-2 HLA class I and HLA-DR T-cell epitopes in PRE donors were lower compared to that of SARS group 1 (up to 27% for B08_P05 and 44% for DR_P01, Fig. 2e, f, Extended Data Tables 2 and 3). Recognition frequencies of HLA class I and HLA-DR T-cell epitopes in individual donors differed profoundly between the PRE and the SARS group within the different ORFs. ORF1-derived HLA class I (9%) and ORF8-derived HLA-DR (25%) T-cell epitopes showed the highest recognition frequencies in the PRE group, whereas noneof the T-cell epitopes from ORF5 (membrane protein) and ORF10 that were frequentlyrecognized in SARS donors were detected by T cells in PRE donors (Fig. 2i). In line with the lower recognition frequencies of single SARS-CoV-2 T-cell epitopes (Fig. 2e, f), donor-specific recognition rates of HLA class I and HLA-DR SARS-CoV-2 T-cell epitopes were significantly lower in the PRE group (HLA class I, mean 26 ± 9; HLA-DR, mean 10 ± 5) than in the SARS group (HLA class I, mean 52 ± 23; HLA-DR, mean 52 ± 23, Fig. 3a). Alignments of the SARS- CoV-2 T-cell epitopes recognized by unexposed individuals revealed similarities to the four seasonal common cold human coronaviruses (HCoV-OC43, HCoV-229E, HCoV-NL63, HCoV- HKU1) with regard to amino acid sequences, physiochemical and/or HLA-binding properties for 14/20 (70%) of the epitopes, thereby providing clear evidence for SARS-CoV-2 T-cell cross-reactivity (Fig. 3b, Supplementary Tables 9 and 10, Supplementary Data 1).
Frequency of SARS-CoV-2 T-cell responses in COVID-19 convalescents and unexposed individuals
Epitope screening in SARS and PRE donors enabled the identification of SARS-CoV-2-specific T-cell epitopes recognized exclusively in convalescents after SARS-CoV-2 infection and of cross-reactive T-cell epitopes recognized by both, convalescents and SARS-CoV-2 unexposed individuals (Fig. 2e, f). To allow for standardized evaluation and determination of T-cell response frequencies to SARS-CoV-2, we designed broadly applicable HLA class I and HLA-DR SARS-CoV-2-specific and cross-reactive T-cell epitope compositions (EC, Fig. 3c, Extended Data Table 4). These EC were utilized for IFNγ ELISPOT assays in groups of convalescents (SARS group 2, n = 86, Extended Data Table 1, Supplementary Table 6) and unexposed donors (PRE group B, n = 94, Extended Data Table 1, Supplementary Table 7). Of the SARS donors, 100% showed T-cell responses to cross-reactive and/or specific EC (Fig. 3d, e), whereas 81% of PRE donors showed HLA class I (16%) and/or HLA-DR (77%) T-cell responsesto cross-reactive EC (Fig. 3d). In line with the findings obtained with the screening group(SARS group 1), the intensity of HLA class I T-cell responses was significantly lower compared to HLA-DR T-cell responses, both for specific (median calculated spot count HLA class I 379, HLA-DR 760) and cross-reactive EC (median calculated spot count HLA class I 86, HLA-DR 846, Fig. 3f, g). In line with the differences in recognition rates observed between SARS group 1 and PRE group A, the intensity of T-cell responses to cross-reactive EC was significantly lower in the PRE group (median calculated spot count HLA class I 14, HLA-DR346) compared to the SARS group (Fig. 3g).
Relationship of SARS-CoV-2 T-cell and antibody responses
Anti-SARS-CoV-2 IgG antibody responses in SARS donors were analyzed in two independent assays. The S1 IgG ELISA assay revealed 149/178 (84%), 7/178 (4%), and 22/178 (12%) donors with positive, borderline, and no anti S1 antibody response, respectively (Fig. 4a). Of the borderline/none responders, 18/29 (62%) were also negative in a second, independent anti-nucleocapsid immunoassay (Fig. 4b). However, SARS-CoV-2-specific CD8+ and/or CD4+ T- cell responses were detected in 10/18 (56%) of these “antibody double-negative” donors (Fig. 4c). The intensity of SARS-CoV-2-specific and cross-reactive HLA-DR T-cell responses correlated with antibody levels (Fig. 4d, e), whereas no correlation was observed with HLA class I T-cell responses (Extended Data Fig. 3a, b). No correlation between antibody titers directed against the nucleocapsid of human common cold coronaviruses (HCoV-229E, HCoV- NL63, HCoV-OC43), as determined by bead-based serological multiplex assays and the intensity of cross-reactive CD4+ and CD8+ T-cell responses in the SARS group, was detected (Extended Data Fig. 3c-h).
Association of SARS-CoV-2-directed antibody and T-cell responses with clinical characteristics in COVID-19
Finally, the association of anti-SARS-CoV-2 antibody and T-cell responses with disease severity as assessed by a combinatorial symptom score (SC) of objective (fever ≥ 38.0°C) and patient-subjective disease symptoms was determined (Extended Data Table 1). Alike in critically ill patients26, independently of age high antibody ratios significantly associated with disease severity in our collection of convalescent SARS donors (n = 180), which in general were in good health condition and had not been hospitalized (Fig. 4f, Extended Data Fig. 4a). Neither the intensity of SARS-CoV-2-specific nor of cross-reactive T-cell responses to HLA class I or HLA-DR EC correlated with disease severity (Fig. 4g). Rather, diversity of T-cell responses in terms of recognition rate of SARS-CoV-2 T-cell epitopes was decreased in patients with more severe COVID-19 symptoms (Fig. 4h, Extended Data Fig. 4b), providing evidence that development of protective immunity requires recognition of multiple SARS- CoV-2 epitopes.