Identification of TCR repertoires in functionally competent cytotoxic T cells cross-reactive to SARS-CoV-2

SARS-CoV-2-specific CD8+ T cells are scarce but detectable in unexposed healthy donors (UHDs). It remains unclear whether pre-existing human coronavirus (HCoV)-specific CD8+ T cells are converted to functionally competent T cells cross-reactive to SARS-CoV-2. Here, we identified the HLA-A24-high binding, immunodominant epitopes in SARS-CoV-2 spike region that can be recognized by seasonal coronavirus-specific CD8+ T cells from HLA-A24+ UHDs. Cross-reactive CD8+ T cells were clearly reduced in patients with hematological malignancy, who are usually immunosuppressed, compared to those in UHDs. Furthermore, we showed that CD8+ T cells in response to a selected dominant epitope display multifunctionality and cross-functionality across HCoVs in HLA-A24+ donors. Cross-reactivity of T-cell receptors isolated from them exhibited selective diversity at the single-cell level. Taken together, when stimulated well by immunodominant epitopes, selective pre-existing CD8+ T cells with high functional avidity may be cross-reactive against SARS-CoV-2.

S evere acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is responsible for causing the COVID-19 pandemic. Recently, several COVID-19 vaccines, such as mRNA-based vaccines or adenovirus-vectored vaccines expressing SARS-CoV-2 spike (S) protein, have now been administered through global vaccination programs [1][2][3] . Neutralizing antibodies (Abs) need to be generated to protect the body by the SARS-CoV-2 vaccine. Furthermore, to inhibit SARS-CoV-2 more efficiently, CD4 + T cells and CD8 + T cells play important roles in combating SARS-CoV-2 4 . Recent studies on humoral immunity induced by the vaccines provide evidence regarding the duration of the neutralizing antibody response 5 . Moreover, it is known that CD4 + T cells and CD8 + T cells, in addition to neutralizing Abs, can play an important role in inhibiting SARS-CoV-2 4 . Anti-viral CD8 + cytotoxic T cells have the potential to eliminate virus-infected cells, even during the latency of viruses, such as HIV, CMV, EBV, and HSV [6][7][8] . Understanding anti-viral CD8 + T cell immunity may be useful in identifying biomarkers to evaluate the severity of or develop treatment strategies against COVID-19.
Pre-existing CD8 + T cells for seasonal coronaviruses that cross-react with SARS-CoV-2 can be detected in only about 20% of the unexposed healthy donors (UHDs) 11 , but may still be biologically relevant 16 . In UHDs, the majority of immunodominant epitopes are derived from non-S regions, ORF1a, and nucleocapsid, but not S protein regions 11,19 because of the low homology of amino acids in S protein. The cross-reactivity of CD8 + T cells for S protein has not been studied sufficiently. The identification of cross-reactive immunodominant epitopes in S protein in UHDs is among the key challenges for CD8 + T cell-induced vaccine development. In addition, as TCR crossreactivity under specific HLA-restriction has not been studied well, molecular analysis of CD8 + T cell cross-reactivity is required for understanding their roles.
Patients with hematological malignancies (HM) are vulnerable to infectious diseases, including SARS-CoV-2 infections, and exhibit higher mortality than those with solid tumors [20][21][22] . Not only the high risk but also the worse outcomes of infection observed for HM patients are likely due to the fact that they are severely immunocompromised by the underlying hematological malignancy and systemic chemotherapy. In contrast, despite extremely low B-cell or antibody production, a sufficient CD8 + T cell count can reduce the COVID-19-related mortality in patients with HM 17 . Therefore, a vaccine should engage in as many facets of the immune system as possible, such as CD8 + T cells and antibodies. Here, we explored the immunodominant epitope on S protein for the generation of SARS-CoV-2 crossreactive CD8 + T cells under HLA-A24 restriction. Using this epitope, we showed the cross-functionality of the epitopespecific CD8 + T cells in UHDs. We examined the mechanism underlying the cross-reactivity to seasonal coronaviruses at the single-cell TCR level. In addition, we found the decrease of these SARS-CoV-2 cross-reactive CD8 + T cells in the patients with HM to that in UHDs

Results
Identification of the immunodominant epitope in SARS-CoV-2 S in HLA-A24 + UHDs. The relationship between HLA alleles and COVID-19 has recently been reported 23 . We, therefore, assessed the interaction between HLA-A*24:02, as one of the major HLAs, and CD8 + T cells cross-reactive for SARS-CoV-2. Candidate targets for the CD8 + T immunodominant epitope in SARS-CoV-2 spike protein (SARS-CoV-2 S) were predicted using NetMHC 4.0. We selected and synthesized six candidate peptides for these epitopes from the entire S protein sequence and one epitope from nucleocapsid protein as a control (Fig. 1a) 9 .
In the current study, we used the PBMCs from healthy volunteers that were collected in early 2020 (from February to April in 2020) and during 2004-2010 (pre-pandemic era) as UHDs (Supplementary Table 1). None of the volunteers suffered from COVID-19, and they were never in close contact with a COVID-19 patient. In addition, all UHDs neither had a history of SARS-CoV-1 nor MERS infections. In addition, we assessed their plasma antibody production twice using two types of commercial kits that were often used in the previous studies 16,19,24 . All UHDs tested negative for antibodies against SARS-CoV-2. For initial screening of the optimal peptide for SARS-CoV-2 CD8 + T cell responses, five HLA-A24 + UHDs were tested. Short-term assays, such as 24-h activation induced marker assays described in the previous reports 11,13,15 , are useful for the detection of effector CD8 + T cells, whereas in-vitro expansion assays can be used to detect pre-existing CD8 + T cells against cancer or cross-reactive CD8 + T cells to SARS-CoV-2, even at low levels 14,[25][26][27][28] . PBMCs were cultured in the presence of each peptide, and the frequencies of epitope-specific IFN-γ-producing CD8 + T cells were assessed on day 21. Although these six peptides from S protein were expected to bind to HLA-A*24:02 with high affinity, five of these did not induce a T-cell response, and only Pep#3(QYI)-specific CD8 + T cell counts were elevated in cells derived from all UHDs ( Fig. 1b and Supplementary Fig. 1a, b). As positive response, we defined a 1.5-fold increase in T-cell response as compared to that in the controls ( Supplementary Fig. 1c). Intriguingly, Pep#3(QYI) exhibited high sequence homology with other coronaviruses, including four types of seasonal coronaviruses (Fig. 1c). Although all the six peptides of S protein were predicted to exhibit high affinity to HLA-A*24:02, five peptides were not homologous to other seasonal coronaviruses ( Supplementary Fig. 1d). These results suggest that CD8 + T cells may be cross-reacting, and that pre-existing memory CD8 + T cells may be responsible for the immune response.
We then focused on Pep#3(QYI)-specific CD8 + T cells from UHDs. In this assay, we isolated CD8 + T cells from PBMCs and cocultured them with irradiated autologous PBMCs pulsed with Pep#3(QYI), and the frequencies of epitope-specific IFN-γproducing CD8 + T cells were assessed on day 14 (Fig. 1d). We found that antigen-specific CD8 + T cells were present in 15 out of 18 UHD participants (83.3%) (Fig. 1e, f, g). The response rate of Pep#3(QYI)-specific T cells in HLA-A*24:02 + UHDs was higher than that in 20% of the UHDs in a previous study 11 . Even if the frequency was low, most of the responders possessed IFN-γand TNF-α-producing T cells (Fig. 1e, f).
Detection of the high affinity region around the Pep#3(QYI). Next, we synthesized four types of 11 mer-overlapped 15-mer peptides based on the S2 region (1200-1226) of S protein around the Pep#3(QYI) epitope, to examine the other immunodominant epitopes in this region (Fig. 3a). Since the relevant epitopes are relatively conserved across HCoVs and completely conserved across the SARS-CoV-2 variant strains, such as VOC-202012/01, 501Y.V2, and 501Y.V3, the peptide set is available for SARS-CoV-2 variant strains ( Supplementary Fig. 3). We cultured PBMCs from UHDs in the presence of a 15-mer peptide mix and evaluated the response of antigen-specific CD8 + T cells at day 14 ( Fig. 3b). Surprisingly, we detected SARS-CoV-2-specific CD8 + T cells in all the UHDs exposed to the peptides (Fig. 3c). Moreover, CD8 + T cells responded primarily to 15-mer pep#2 and #3, but in some cases, they also responded to 15-mer pep#1 and #4 (Fig. 3d-f). These findings indicated that the 15-mer peptide mix contained the other immunodominant epitopes in addition to Pep#3(QYI) for HLA-A24 + individuals and would be more useful for detecting or expanding pre-existing CD8 + T cells in HLA-A24 + individuals.
Multiple epitopes in SARS-CoV-2 S are highly immunogenic in HLA-A24 + UHDs. Next, to verify the parts of the epitopes of four types of the 15-mer peptide mix responsible for CD8 + T cell activation, a consecutive 9-mer peptide library was synthesized using a S protein sequence (Fig. 4a). We assessed peptide-specific T-cell response by restimulation with each 9-mer-peptide after priming with the 15-mer peptide mix (Fig. 4b). We found that the 9-mer peptides#1204 to #1218, particularly #1208 (Pep#3(QYI)), induced CD8 + T cell-specific responses comparable to 15-mer pep#2 to #3 (Fig. 4c, d). Accordingly, we conducted an in-silico screening of this region in SARS-CoV-2 and seasonal coronaviruses using IEBD and NetMHC4.0 (Fig. 4e) to examine which 9-mer peptides were recognized by these CD8 + T cells under the HLA-A24 restriction. We found six and eight types of 9-mer-epitopes with high IEBD and NetMHC scores expressed on SARS-CoV-2 as well as seasonal coronaviruses, which suggested that they bind to HLA-A24 (Fig. 4e). Thus, we found that some of the 9-mer peptides identified in our in-vitro experiments were the high affinity peptides predicted in silico under HLA-A24 restriction. However, other peptides may bind to other HLAs. Peptides derived from Spike (S) protein of SARS-CoV-2, shown in the list, were selected based on high-affinity binding (% rank) to HLA-A*24:02 via in silico prediction using NetMHC4.0. Pep#7 for nucleocapsid protein was selected as the control. b The data of IFN-γ + response of antigen-specific CD8 + T cells from one representative donor (UHD2). PBMCs were stimulated with seven types of peptides individually, cultured for 3 weeks in the presence of IL-2, and then assessed for IFN-γ production following recall with their cognate peptide by intracellular cytokine analysis (ICS). c Homology of the reported epitope in the Pep#3 region of SARS-CoV-2 compared with that in other human coronaviruses. Identical residues are same color-coded. d Culture protocol of Pep#3(QYI)-specific CD8 + T cell response. The sorted CD8 + T cells from PBMCs of HLA-A24 + UHDs were stimulated with irradiated autologous PBMCs pulsed with Pep#3(QYI). The Pep#3(QYI)-specific CD8 + T cells were examined by the frequency of IFN-γ + TNF-α + CD8 + T cells by ICS analysis on day 14. e Flow cytometry data are shown for representative donors (UHD5, 6, 13, 14, 18, and 19) (e) and the frequency of IFN-γ + TNF-α + CD8 + T cells is summarized (n = 18) (f). ※ responder (>1.5 fold change (restimulation+/restimulation−) g Pie chart representing the proportions of responders or nonresponders against Pep#3(QYI) in 18 HLA-A24 + UHDs.
Analysis of SARS-CoV-2-specific CD8 + T cells in patients with HM. We examined whether Pep#3(QYI) can induce epitopespecific CD8 + T cells in HLA-A24 + patients with HM (Supplementary Table 2), who represent a high-risk group that may be susceptible to COVID-19 due to immune-cell dysfunction related to disease progression or chemotherapy. We confirmed that all patients with HM recruited in this study were negative for IgG specific for SARS-CoV-2 (Supplementary Table 2). Pep#3(QYI)specific CD8 + T cells were observed in 14.8% (4 out of 27) of the patients with HM, and the induction ratio of Pep#3(QYI)-specific CD8 + T cells was significantly lower in patients with HM than in UHDs (p < 0.01; Fisher's/Chi-square tests) (Fig. 5a-c). This indicated that the levels of SARS-CoV-2 cross-reacting CD8 + T cells against seasonal coronaviruses were lower in patients with HM than in UHDs.
Next, we analyzed these patients by four types of 11 meroverlapped 15-mer peptides ( Fig. 5d-f). Surprisingly, we detected SARS-CoV-2-specific CD8 + T cells in 65.4% of patients with HM ( Fig. 5e, f). T-cell response in the presence of a mix of four peptides was more efficient than that induced by the 9-mer peptide, particularly in patients with HM. Thus, there were fewer cross-reactive CD8 + T cells in patients with HM than in UHDs (65.4% vs 14.8%, p < 0.001; Fisher's/Chi-square tests), which indicated that seasonal coronavirus-specific memory T cells may be reduced in response to disease progression or therapy-related toxicity. However, CD8 + T cells that have the potential to respond to the 15-mer mix, mainly to 15-mer pep#2 and #3, can be activated by rechallenging with the 15-mer peptide mix. Furthermore, in some cases, they also responded to 15-mer pep#1 and #4 ( Supplementary Fig. 4).
CD8 + T cells exhibit cross-reactivity between SARS-CoV-2and HCoV-derived S peptide. We evaluated whether the Pep#3(QYI)-specific CD8 + T cells generated from ten HLA-A24 + UHDs and two patients with HM cross-reacted with other coronavirus families, and quantified SARS-CoV-2-specific CD8 + T cells (Fig. 6a). The Pep#3(QYI)-specific CD8 + T cell lines were assessed based on the production of IFN-γ and TNF-α in response to the relevant peptide of each virus (Fig. 6b). The Pep#3(QYI)-specific CD8 + T cell lines from UHDs exhibited cross-reactivity across the HCoVs. In these two patients with HM, cross-reactivity of CD8 + T cells was also found to be preserved. The CD8 + T cell lines responded to betacoronaviruses (SARS-CoV-1, MERS, HKU-1, and OC43), with 83-100% responders, and to alphacoronaviruses (NL63 and 229E), with 58.3-66.7% responders ( Fig. 6c-e). Next, we assessed the level of the cross-reactivity of Pep#3(QYI)specific CD8 + T cells to coronaviruses in UHDs. We also evaluated the functional avidities of these CD8 + T cells by IFN-γ production. Peptide titration experiments defined a dose range of 0 to 10,000 ng/mL for SARS-CoV-2 and seasonal coronavirus with the highest affinity (e.g., HCoV-HKU1) (Fig. 6f). EC 50 is calculated as the peptide concentration required to reach one-half maximal response. EC 50 for SARS-CoV-2 S was comparable to that for seasonal coronaviruses (Fig. 6f). These data demonstrated similar functional avidity of the SARS-CoV-2 cross-reactive CD8 + T cell response relative to the CD8 + T cell response for seasonal coronaviruses. Furthermore, we assessed the cytotoxic activity across different coronaviruses as another measure of cross functionality (Fig. 6g, h). The cytotoxicity against A24/CIR pulsed with the target peptide from SARS-CoV-2 was compared to that from seasonal coronaviruses with the highest affinity for Pep#3(-QYI)-specific CD8 + T cells in each UHD cell line (Fig. 6g). However, no statistical difference was observed between them (Fig. 6h). These findings indicated that although some pre-existing memory CD8 + T cells in HLA-A24 + UHDs have the potential to recognize SARS-CoV-2, they can be sufficiently skewed toward SARS-CoV-2 under optimal conditions. TCR avidity for recognition of SARS-CoV-2 and HCoVs at the single-cell level. To understand why CD8 + T cells exhibit multiple responses to various coronaviruses and the underlying mechanisms, we studied the cross-reactivity of CD8 + T cells at the single TCR level. After restimulation with Pep#3(QYI), we isolated CD8 + T cells from Pep#3(QYI)-specific CD8 + T cell lines by gating CD107a + CD8 + cells and performed single-cell analysis of TCR repertoires. To analyze the clonalities of Pep#3(QYI)-reactive CD8 + T cells, a total of 227 T cells were screened, and 44 TCR clonotypes were identified from five UHDs and one patient with HM ( Supplementary Fig. 5). Furthermore, we focused on the dominant TCR types from among the clonotypes and identified four types of TCRα and TCRβ pairs from four donors (three UHDs and one HM). To assess the specificity and functions of cloned TCRs, we transduced the TCRα and TCRβ genes into the SKW3-CD8AB (human T-ALL) cell line and demonstrated their peptide-specific response (Fig. 7a). Intriguingly, the four types of TCR repertoires varied from each other in their epitope recognition (Fig. 7b). TCR-T (TCR-T-1) cells from UHD2 responded well to all the epitopes derived from coronaviruses, whereas the TCR-T (TCR-T-2) cells from UHD8 responded only to SARS-CoV-1 and SARS-CoV-2. These data are interesting because UHD8 had not been infected with either of them. Dykema et al. also reported the same phenomenon in a previous study 27 , wherein SARS-CoV-2-specific, monoreactive CD4 + T clonotypes were detected in the UHDs and indicated the possibility of cross-reactive response to unknown coronaviruses or other pathogens. The TCR-T-2 data, however, may also be a chance finding. Furthermore, the TCR-T (TCR-T-3) cells from UHD6 responded to SARS-CoV-2, SARS-CoV-1, NL63, and 229E, while those (TCR-T-4) from HM16 responded to SARS-CoV-2, SARS-CoV-1, MERS, HKU1, and OC43, but not to NL63 and 229E (Fig. 7b, c). Some types of single TCRαβ-transduced T cells responded broadly to the relevant epitopes on HCoVs, whereas others responded only to two or four peptides. These findings on the SARS-CoV-2-epitope-responding T cell repertoires imply that TCRαβ may be selective at the single-cell level.
We next evaluated the functional avidities of these TCRtransduced T cells by upregulation of CD69 expression (MFI). Peptide titration experiments defined a dose range of 10 fg/mL to 10 μg/mL for SARS-CoV-2 and each seasonal coronavirus (Fig. 7d). TCR-T-1 exhibited the highest avidity for OC43 compared to that for SARS-CoV-2 and others. However, other TCR-T cells (TCR-T-2 to TCR-T-4) from UHDs exhibited the highest avidity for SARS-Co-V-2 (Fig. 7e). Comparing TCR-T-2 Fig. 3 Design of the 15-mer peptide mix-responding CD8 + T cells. a Design of 15-mer peptide library, including Pep#3(QYI) epitopes, in the S2 region of SARS-CoV-2. b Analysis of the 15-mer peptide-responding CD8 + T cells. PBMCs from UHDs were cultured in the presence of a 15-mer peptide mix. Two weeks later, the cultured cells were restimulated with or without the peptide mix for 16 h in the presence of monensin and brefeldin. The frequency of IFN-γ + TNF-α + CD8 + T cells was measured after gating the CD8 + T cells by an ICS assay at day 14. c Data show the frequency of IFN-γ + TNF-α + CD8 + T cell response to the 15-mer peptide mix in UHDs (n = 20). ※ responder (>1.5 fold change (restimulation +/restimulation−). d As shown in (b), the cultured cells were restimulated with or without the peptide mix or each 15-mer peptide for 16 h in the presence of monensin and brefeldin. The frequency of IFN-γ + TNF-α + CD8 + T cells was measured after gating the CD8 + T cells by an ICS assay. Flow cytometry data from representative donors. e Summary of the 15-mer peptide-responding CD8 + T cells. The peptide-specific CD8 + T cell frequency was calculated by subtracting the frequency of IFN-γ + TNF-α + CD8 + T cells cultured without peptide restimulation (CD8T (pep−)) from the frequency of IFN-γ + TNF-α + CD8 + T cells cultured with peptide restimulation (CD8T (pep+)). Data indicate the ratio of the frequency of each 15-mer peptide-responding IFN-γ + TNF-α + CD8 + T cells to that of the 15-mer peptide mix responding IFN-γ + TNF-α + CD8 + T cells (n = 20). f Frequency of donors responding to each 15-mer peptide. (monoreactive) to other cross-reactive TCRs (TCR-T-1, -3, and -4), we could not find large differences in their functional avidities (Fig. 7e). Irrespective of the previous infection status, Pep#3(-QYI)-specific CD8 + T cells from UHDs displayed sufficient cross-functionality at the single cell level (Fig. 7e). From these data, we concluded that Pep#3(QYI)-specific CD8 + T cells from UHDs did not exhibit low avidity.
Crystal structures of CD8 + T cells binding to HLA-A*24:02 and TCR complex. The crystal structures of HLA-A*24:02 and three peptides, namely, SARS-CoV-2, 229E, and HKU1, were determined (Fig. 8a, Table 1). These three complex The 9-mer peptides were identified at the equivalent position of the HLA-A*24:02 structure (Fig. 8b). Although electron densities of the peptide backbones were observed, those of some side chains were poor, particularly, Lys4 for three peptides, Trp5 and 7 for CoV2 and 229E, and Pro6 for HKU1. The conformation of these peptide backbones overlapped well. Of the nine positions, the side chains at positions P2, P3, P6, and P9 were involved in the HLA-A*24:02•peptide interactions at the bottom of the binding pocket. In contrast, the side chains at positions P1, P4, P5, P7, and P8 were directed to the solvent region and oriented differently in the complex structures (Fig. 8c). Thus, these side chains are expected to directly interact with TCR.
In the Protein Data Bank, 31 structures of the HLA-A*24:02•peptide complex have been deposited so far. Peptide lengths are in the range of 8-mer to 11-mer. Of these, two complex structures are with the 9-mer peptide 29 . A comparison between these 9-mer peptides and the peptides in this study revealed that all the peptides anchored to the binding pocket at positions P2 and P9, and conformational variations were observed at positions P4-P7 (Fig. 8d). This region forms a bulge structure and corresponds to the putative interaction site with TCR. Furthermore, we investigated the influence of peptide sequence on the interface and the intrinsic plasticity of the TCR/ pMHC trimolecular and pMHC biomolecular complexes via structural modeling performed by replacing the HLA moiety of the HLA•peptide•TCR complex structure 30 (PDB code 3VXM) with the structure mentioned in the present study.

Discussion
Elucidating the role of pre-existing, cross-reactive T cells in COVID-19 is important for understanding the development and severity of COVID-19 and determining the optimal T-cell engagement strategies for development of diagnostic tools and vaccines. An expansion of cross-reactive T cells with low avidity may lead to exacerbation of infection. Therefore, it is critical not only to identify pre-existing T cells but also to find an approach to expand functionally competent, cross-reactive T cells with high avidity. In the current study, we focused on an amino acid sequence conserved between SARS-CoV-2 and seasonal coronaviruses in the S2 region, and selected 9-mer peptides. One of the selected peptides, Pep#3(QYI), efficiently induced SARS-CoV-2 cross-reactive CD8 + T cells in HLA-A24 + UHDs. These Pep#3(QYI)-specific CD8 + T cells from UHDs exhibited polyfunctionality and similar cytotoxicity against seasonal coronaviruses. Furthermore, we identified immunodominant epitopes, including Pep#3 (QYI), which were covered by three 15-mer overlapping peptides. Moreover, to elucidate the mechanism, we assessed the cross-reactivity and functional avidity of Pep#3 (QYI)-specific TCR at the single TCR level.  (UHD 1, 8, and 11) are shown. c Frequency of donors responding to each peptide of the 9-mer peptide library. d Response to 9-mer peptide-specific CD8 + T cells selected from the 15-mer peptide mix. Data indicate that the fold change (FC) of CD8 + T (pep+) to CD8 + T (pep−) was 1.5 times higher (orange) and 3 times higher (red) in the 9-mer peptide, respectively. (n = 12) Gray indicates not done. e In-silico analysis of the affinity for HLA-A24 in both SARS-CoV-2 S protein 1200-1226 and the relevant region of seasonal coronaviruses using IEBD and NetMHC4.0. (%rank < 0.5 (red), 0.5 < %rank < 2 (orange), %rank > 2 (white)).
Although cross-reactivity with pre-existing T cells has been reported [10][11][12][13][14]19,31 , whether SARS-CoV-2 cross-reactive T cells are protective is still an issue of debate. SARS-CoV-2 crossreactive CD4 + T cells have been reported to exhibit low avidity, and some studies have demonstrated that they are nonprotective 24,27 . However, there are three studies that address the cross-reactivity of pre-existing CD8 + T cells in UHDs. Schulien et al. performed an analysis of pre-existing and induced CD8 + T cells in three patients pre-and post-SARS-Cov-2 infection and showed the importance of anti-viral effector CD8 + T cells 16 . T cell cross-reactivity can be defined to recognize more than one distinct peptide-MHC structures by single TCR 32 . Mallajosyula et al. demonstrated that T cells that recognize peptides conserved among coronaviruses are abundant in UHDs 33 . Moreover, they verified that these are more abundant in COVID-19 patients by spheromer technology. They also detected shared TCR motifs of SARS-CoV-2 specific CD8 + T cells between UHDs and mild COVID-19 patients 33 , suggesting a protective role of pre-existing CD8 + T cells in COVID-19. A recent study reported immunodominant, HLA-B7-restricted SARS-CoV-2 nucleocapsid protein epitope cross-reactive CD8 + T cells 28 . The authors found the presence of a shared CDR3β motif in epitope-specific CD8 + T clonotypes in exposed and unexposed donors, which suggested that pre-existing immunity in HLA-B7 + individuals favored clonal expansion. In the current study, we demonstrated the pattern of Pep#3 (QYI)-specific CD8 + T cell TCR cross-reactivity across HCoVs under HLA-A24*02 restriction at the single-cell level, which exhibited substantial functional avidity. Nevertheless, further molecular studies that examine TCR characteristics of HLA-A24 + UHDs and infected patients are required.
Cancer patients have an increased risk of severe illness from COVID-19, with a high mortality rate 20,34,35 . Particularly, adult patients with HM having COVID-19 were found to have a 34% risk of death. Patients aged >60 years had a significantly higher risk of death 22 . Furthermore, patients with impaired B-cell function, such as myeloma, lymphoma, or CLL, may not respond to standard vaccines and may require special attention 36,37 . Using a 15-mer peptide mix, SARS-CoV-2 cross-reactive CD8 + T cells were successfully detected at 75%, 50%, and 56% higher rates in AML, CML, and MM, respectively, compared to those detected using Pep#3(QYI). Thus, the 15-mer peptide mix may act as CD8 + T-cell-induced vaccines or immune monitoring tools for the efficacy of COVID-19 vaccines in high-risk HLA-A24 + patients with HM. Since these findings indicate the importance of HLA in COVID-19 pathogenesis, other HLA and peptide interactions should also be studied.
The restricted formula for linking HLA-peptide and TCR confers heterotypic cross-reactivity across SARS-CoV-2. Furthermore, to elucidate the molecular basis of cross-reactive TCRα/β recognition between SARS-CoV-2 and seasonal coronaviruses, we examined the structure of SARS-CoV-2 peptides and the relevant peptides of seasonal coronaviruses bound to the HLA-A*24:02 molecule (Fig. 8). The obtained data on the HLA-A*24:02•peptide complex structures imply that the side chains at positions P1, P4, P5, P7, and P8 interact with the TCR. The superimposed model (where the HLA moiety is replaced by the structure presented in our study) indicates that the peptide positions mentioned above are close to the contact surface of the TCR. The side chains at positions P1/P4/P5 and P7/P8 can interact with the loop moieties of TCRα and TCRβ, respectively. In the sequence alignment among TCR1-4, insertions/deletions were found in these loop regions, suggesting variety of affinities between the HLA-A*24:02 peptide complex and TCRs. Furthermore, our data suggest that the selective cross-reactivity of T cells in SARS-CoV-2 epitopes depends on combinations of the structure mode of the peptide on the HLA-A*24:02 complex and the sequences/loop region of the TCR conformation. Further studies are required to broadly underpin the cross-reactivity by analyzing the structure of the cross-reactive TCR complex with HLA-A*24:02•SARS-CoV-2-S2 and HLA-A*24:02•seasonal coronavirus epitopes.
Overall, we found that the potential of cross-reactivity may depend on the interaction between TCR and the structure of specific HLA, i.e., HLA-A*24:02. Moreover, TCRα/β from Pep#3(QYI)-specific CD8 + T cells exhibited varying patterns of cross-reactivity against HCoVs. Lastly, we determined the optimal 15 mer-mixed peptides that may stimulate SARS-CoV-2specific CD8 + T cells, even for most patients with HM in addition to all UHDs; the results indicated the possibility of stimulatory epitopes. We speculated that some CD8 + T cells that act against seasonal coronaviruses might persist as long-term memory cells in UHDs. If these cross-reactive T cells were stimulated by a type of vaccine, including immunodominant epitopes, these could be skewed toward SARS-CoV-2. Development of a vaccine for modulating cross-reactive CD8 + T cells may help reduce the rates of the disease.

Methods
Human samples and preparation. Peripheral blood samples were obtained from healthy blood donors at our Institute (early 2020) or from the buffy coats of healthy  CD8 + T cell lines were restimulated with the indicated peptide or DMSO for 16 h and analyzed using IFN-γ-APC and TNF-α-PE by ICS analysis. b Representative flow cytometry data from UHD1 and HM21 are shown. c Frequency of donors responding to each peptide of the coronaviruses. d The numbers of donors responding to each peptide of the coronaviruses. e The peptide-specific CD8 + T cell frequency was calculated by subtracting the frequency of CD8 + T (pep−) from that of CD8 + T(pep+). The ratio of each coronavirus-derived peptide-specific IFN-γ + TNF-α + CD8 + T cell frequency to that of CoV-2 pep#3 is depicted. (n = 12). f Avidity of antigen-specific CD8 + T cells. Flow cytometry data show IFN-γ expression for the indicated concentrations of peptides from SARS-CoV-2 and seasonal coronavirus (UHD1). Based on these data, half maximal response (EC 50 ) of Pep#3(QYI)-specific CD8 + T cells was measured in the response to cognate peptide and the relevant peptide of seasonal coronaviruses. (n = 9) (n.s., Mann-Whitney U-test). g, h Cross-functionality of Pep#3(QYI)-specific CD8 + T cells. Peptide or -specific cytotoxic activity of Pep#3(QYI)-specific CD8 + T cells from 5 UHDs (UHD5, 6, 13, 15, and 18) was compared. Cytotoxic T cell assays were performed using Pep#3(QYI)-pulsed A24/CIR or the seasonal coronavirus derived peptide-pulsed A24/CIR as target cells (E/T ratio = 50). The cytotoxicity assay was performed in duplicates. Data represent the mean of duplicate (g). The white bar and other symbols represent the mean of values from five UHDs and each data, respectively. (mean ± SEM) (n.s., Mann-Whitney U-test) (h). procured from RIKEN BRC (Japan). All cell lines were tested according to the manufacturer's protocol and proved to be mycoplasma free (Mycoplasma Detection Kit; Minerva Biolabs). A24/CIR has been routinely tested for HLA-A24 expression by our hands. According to the International Cell Line Authentication Committee register (https://iclac.org), SKW-3 cells were misidentified as KE-37 cells. We verified that our cells lack TCRs and functionally work like T cells when transduced with T cell receptors and CD8AB.
Epitope prediction and peptide selection. The S protein CD8 + T epitopes of SARS-CoV-2 and other HCoV for HLA-A*24:02 were predicted using the Immune Epitope Database and Analysis Resource (IEDB) (https://www.iedb.org/) 39  In-vitro expansion of SARS-CoV-2-specific T cells. For the peptide screening test, PBMCs (3-5 × 10 6 cells) were stimulated with the indicated peptide (each peptide 10 µg/mL) (summarized in Supplementary Table 3) in culture medium in the presence of IL-2 (100 U/mL) (Shionogi CO., LTD) and restimulated with 30Gyirradiated autologous PBMCs pulsed with each peptide weekly. Then, these cells were analyzed on day 21. For determining CD8 + T cell response against Pep#3(-QYI), CD8 + T cells (0.5-1 × 10 6 cells) were isolated from PBMCs using CD8 MACS beads (Miltenyi Biotec), and the isolated cells were stimulated with irradiated Pep#3(QYI)-pulsed autologous PBMCs (1:1 ratio) in the presence of IL-2 (100 U/mL) and further restimulated with irradiated autologous PBMCs pulsed with Pep#3(QYI) on day 7. Then, the cells were analyzed at day 14. For determining CD8 + T cell response using 15mer peptides, PBMCs (3-5 × 10 6 cells) were stimulated with the 15mer peptide mixture (each peptide 10 µg/mL) in culture medium in the presence of IL-2 (100 U/mL) and restimulated with irradiated autologous PBMCs pulsed with peptide mixture on day 7. Then, these cells were analyzed at day 14. Cultures were supplemented with R10 media (RPMI-1640 medium supplemented with 10% FBS, 55 mM 2-ME, and 1% penicillin/streptomycin) containing IL-2 twice or thrice weekly. The harvested cells were washed out with R10, restimulated with or without the indicated peptide (10 µg/mL) in the presence of brefeldin A (50 μg/mL) (SIGMA) and monensin (750 ng/mL) (SIGMA) for 16 h, and analyzed using flow cytometry. To determine the threshold, we prepared PBMCs or sorted CD8T cells from UHDs and cultured them without peptide in the presence of IL-2. Two weeks later, the cells were restimulated with Pep#3 (control cells were not restimulated), and IFN-γ production was analyzed by intracellular cytokine staining. Then, we calculated the fold change of the % of IFNγ + CD8 + T cells in the restimulated group compared to that in the control group. Since the fold change in these experiments did not exceed 1.5 (mean ± 2SD: 0.94 ± 0.4) ( Supplementary Fig. 1c), we set the threshold at 1.5.  peptide 10 μg/mL) and analyzed for cytokine production via intracellular cytokine staining. We used a graded dose of the peptide, ranging from 0.1 μg/mL to 10 μg/mL, to assess TCR avidity for each stimulating peptide. TCR avidity was assessed by %IFN-γ in CD8 + T cells. Furthermore, EC 50 was estimated by calculating the peptide concentration required to reach one-half maximum %IFN-γ reached in our assay.
Single-cell TCR sequencing and construction of retroviral TCR vector. Singlecell TCR sequencing was performed as below. All primers used are listed in Supplementary Table 5. CoV-2 pep#3-specific CD8 + T cell lines were restimulated with anti-human CD107a-BV421 in the presence or absence of 10 µM CoV-2 pep#3 for 6 h and then stained with anti-human CD8-PE and Aqua. The peptidespecific CD8 + CD107a + cells were sorted as single cells into 5 μL of RT-PCR mix in a 96-well plate, using FACS Aria III (BD Biosciences). The RT-PCR mix comprised 1 µL of 5× Prime STAR GXL Buffer, 0.45 µL RT-PCR primer mix, 0.4 µL of 2.5 mM dNTP, 0.05 µL of 40 U/µL RNase Inhibitor, 200 U/µL PrimeScript II Reverse Transcriptase (Takara Bio Inc,), 0.05 µL of 1.25 U/µL PrimeSTAR GXL DNA Polymerase (Takara Bio Inc.), and 2.95 µL nuclease-free water. The program used for the one-step RT-PCR was as follows: 45°C for 40 min, 98°C for 1 min, and 35 cycles of 98°C for 10 s, 55°C for 15 s, and 68°C for 1 min. The resultant PCR products were diluted 10-fold with nuclease-free water and used as templates for the 2nd-PCR. The 2nd PCR was performed using PrimeSTAR GXL with the pMXs-BamHI-InFusion primer and CA-rev2 primer for TCRα or CB-rev2 primer for TCRβ in a 10 µL reaction volume. The program for the 2nd-PCR was as follows: 98°C for 1 min and 44 cycles of 98°C for 10 s, 55°C for 15 s, and 68°C for 1 min.
The 2nd-PCR products were treated with ExoSAP-IT PCR Product Cleanup Reagent (Thermo Fisher Scientific K.K.) and sequenced using the CA-rev3 primer for TCRα or CB-rev3 primer for TCRβ. Sequencing results were analyzed using the V-QUEST tool of the IMGT database (http://www.imgt.org/IMGT_vquest/ input) 42 . For cloning TCR, VDJ and C regions were amplified from the 2nd-PCR products and PBMC-derived cDNAs of the donors, respectively. PrimeSTAR Max was used for amplification with the pMXs-BamHI-InFusion and CA-rev3 or CB-rev3 primer. Each fragment was cloned into BamHI-digested linearized pMXs-IRES-GFP for TCRα or -TdTomato for TCRβ by the InFusion HD reaction.
TCR clonotyping and αβ TCR cell line generation. The concentrated TCR Vα and TCR Vβ viruses were used to infect 2 × 10 6 SKW-3-CD8AB cells with 5 µg/mL polybrene (Millipore) via centrifugation for 1 h at 1500 × g at 35°C. The virus was removed, and the medium for cell culture was replaced with fresh medium. TCR-transduced SKW-3-CD8AB cells were enriched by sorting the GFP + tdTOMATO + cells. Epitope identification and avidity analysis of TCR-transduced SKW-3-CD8AB cells. The transduced SKW-3-CD8AB cells were co-cultured with A24/CIR pulsed with the indicated peptide in a 1:1 ratio for 16 h. CD69 was measured to detect early T cell activation by flow cytometry. Graded titrations of the peptide (from 10 μg/mL to 10 fg/mL) were used to assess TCR avidity. TCR avidity was assessed via CD69 MFI. EC 50 was estimated by calculating the peptide concentration required to reach one-half maximum of CD69 MFI reached in our assay.
Cytotoxicity assays. CD8 + T-cell lines were cultured for three weeks and then harvested as effector cells. T-cell cytotoxicity assays were performed using the following methodology: CFSE (Molecular Probe)-labeled A24-CIR cells were pulsed with the indicated peptide or DMSO for 2 h and washed twice. Effector cells were cultured with 1 × 10 4 target cells, at effector/target cell ratios of 12.5, 25, and 50, for 6 h and then stained with TO-PRO3 (Molecular Probe) immediately prior to their analysis to identify dead cells. Spontaneous target cell death (SD) was determined by labeling the target cells that were cultured alone. As a positive control for total cytotoxicity (TD), the labeled target cells were permeabilized with BD Cytofix/Cytoperm reagent (BD Pharmingen). Specific lysis was calculated using the following formula: (Sample − SD/TD − SD) × 100. The cells were analyzed using flow cytometry.
Protein expression and purification. HLA-A24 α chain and β-2-microglobulin with an N-terminal histidine tag and a TEV protease cleavage site were coexpressed in a cell-free expression system 43,44 . The cell-free reaction mixture was prepared with 0.15 mg/mL of each peptide (CoV2, 229E, and HKU1). The HLA-A*24:02•peptide complexes were purified via chromatography using a HisTrap column (Cytiva) and subjected to TEV protease digestion. Protein solutions were then applied to the HisTrap column to remove the histidine-tag. The flow-through fractions were subsequently purified using HiTrap Q and Superdex200 gel filtration chromatography (Cytiva). The HLA-A*24:02•peptide complexes were concentrated in 20 mM Tris-HCl buffer (pH 8.0) containing 50 mM NaCl.
Crystallization and data collection. Crystallization was performed using the sitting-drop vapor-diffusion method at 293 K. The HLA-A*24:02•CoV2 peptide complex crystals were grown in 1.0 M Na/K phosphate (pH 8.2) at a concentration of 6.6 mg/mL. The HLA-A*24:02•229E peptide complex crystals were grown in 0.2 M sodium citrate and 20% PEG3350 at a concentration of 6.9 mg/mL. The HLA-A*24:02•HKU1 peptide complex crystals were grown in 0.1 M Tris-HCl (pH 8.5) and 25% PEG3350 at a concentration of 2.4 mg/mL. Data collection was carried out at 100 K with 10% glycerol as a cryoprotectant. Data were collected at BL26B2 [45][46][47][48] and BL32XU 49 of the SPring-8 synchrotron beamlines with a wavelength of 1.000 Å. The diffraction data were processed and scaled using the ZOO/ KAMO: Automated Data Processing System for Microcrystals 50,51 and the CCP4 software suite 52 .
Structure determination and refinement. The structures were determined using the molecular replacement method (MR) with Phaser 53 . The coordinates of the HLA-A*24:02 complex with the newly identified peptide (PDB ID: 4F7M) were used as the search model. The model was corrected and further refined using Phenix 54 and Coot 55 The Ramachandran plot indicated that 97.4%/2.6% for the CoV2 complex, 97.4%/2.6% for the 229E complex, and 97.7%, 2.3% for HKU1 complex are in favored and allowed regions. Data collection and refinement statistics are summarized in Table 1. Ribbon models in the figures are depicted using PyMOL software (http://www.pymol.org).
Statistics and reproducibility. Statistical analysis was performed using StatMate V (Nihon 3B Scientific Inc.). Fisher's/Chi-square tests were used to compare ratio difference between two groups. Data were analyzed using Mann-Whitney U-test for two independent groups and Turkey or Neuman-Keuls tests for nonparametric multiple comparisons. p < 0.05 was considered statistically significant. The number of biological replicates or sample size are given in figure legend.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The structures of HLA-A*24:02•peptide (CoV2, 229E, and HKU1) complexes were deposited in the RCSB Protein DataBank (PDB) under accession codes 7EJL, 7EJM, and 7EJN respectively. The TCR sequence data were deposited in the BioProject database (PRJNA779816). All source data underlying the graphs presented in the main figures are available in Supplementary Data 1.