It is well well-established that TCR diversity declines with age. In individuals within the first two decades of life, TCRβ diversity in the naïve T cell repertoires is estimated to be 60-120 million, but a decline to 8-57 million is observed in individuals over 70 years of age (96, 97). This age-related decline in TCR repertoire has been confirmed in the antiviral responses to human influenza A viruses (98, 99). This age-related decline in TCR repertoire has been observed in the antiviral responses to human influenza A viruses (98, 99). The numbers of antigen-specific CD8⁺ T cells across universal influenza epitopes were reduced in the elderly, although their effect/memory phenotype remained stable (98). Interestingly, the mortality rate of elderly COVID-19 patients is notably higher than that of young and middle-aged patients, while children mostly exhibit milder disease outcomes when infected with SARS-CoV-2 (2, 100, 101). Further studies are needed to clarify whether aged and less diverse TCR repertoires impact the ability of elderly patients to generate sufficiently robust T cell response to SARS-CoV-2. HLA variation is another factor of concern, influencing the composition of TCR genes by affecting both intra-thymus and extra-thymus clonal selection (102). A study by Francis et al. demonstrated that HLA variations significantly affect the CD8⁺ T cell repertoire shape and utilization of immune recall upon SARS-CoV-2 infection (84). Genetic differences in the HLA genes directly affect the binding affinity of MHC molecules to antigens, which in turn confer susceptibility or resistance to viral infections (32, 33, 103).

The size, frequency, and publicity of individual clonotypes within the TCR repertoire can provide insights into both successful and failed immune responses. During SARS-CoV-2 infection, the diversity and clonability of TCR repertoire peaked within 8-14 days (104, 105), and then returned to base levels within a week after virus elimination (106). In COVID-19 convalescent individuals, SARS-CoV-2 peptide continue to mediate long-term immune responses, with robust functional T cell responses persisting for up to 6 months post-infection (107). A study by Cohen et al. evaluated 254 COVID-19 patients longitudinally up to 8 months and found that virus-specific CD4⁺ and CD8⁺ T cells were polyfunctional and maintained with an estimated half-life of 200 days (108). Long-term immunity against SARS-CoV-2 is primarily driven by the clonal diversity of antigen-specific T cell responses (107, 109). Studies have demonstrated that highly diverse TCR repertoires can offer protection against a variety of antigens, including those from CMV, EBV, and HIV (110, 111), and such repertoires may be associated with a higher level of affinity, affinity, and overall functionality in immune responses (112, 113).

Vaccination against SARS-CoV-2 infection elicit strong T cell responses. Vaccine-induced CTL expansion seems to be relatively weak and results in fewer distinct clonotype clusters compared to CTLs induced by natural infection (114). Despite the rapid contraction of the circulating T Cell responses to SARS-CoV-2 mRNA vaccination, there is a persistent memory that were readily detectable in most individuals out to 235 days after vaccination (115). Repeated mRNA vaccination lead to large expansions of memory spike-reactive T cell clonotypes, most of which were CD8⁺ T cells, while also eliciting diverse spike-reactive T cell clonotypes not observed before vaccination (116). Infection-induced spike-specific CD8⁺ T cell memory plays an important role in the formation of circulating T cell bank size and clonal composition after vaccination (117), and mRNA vaccination promotes the expansion of memory CD8⁺ T cells (117, 118). As both virus- and vaccine-induced antigen-specific TCR repertoires undergo significant clonal contraction over time, coupled with an overall decline in immune response, booster vaccination may be the primary strategy to enhance long-term protection (109, 119).

Viral infection triggers massive T cells that can recognize specific antigens, resulting in skewing of the TCR repertoire toward these antigen-specific T cells (120). It has been demonstrated that certain V, D, and J fragments are over-expressed or under-expressed in COVID-19 patients with different clinical pictures (121). Compared with symptomatic patients, asymptomatic patients exhibit an overrepresentation of some TCR genes, including TRAV (AV17, AV12-1, AV19, AV35, and AV41), TRBV (BV12-5 and BV19), TRAJ16, and TRBJ2-1 (122, 123). Symptomatic patients, on the other hand, have higher frequencies of TRAV2, AJ8, AJ40, BV3-1, and BV5-1 (122). Severe patients tend to highly express TRBV5-6, BV14, BV13 and BV24-1 (124). Furthermore, a study showed that 25 sequences within the central parts of CDR3 region could be used to predict severe infection, emphasizing the significant impact of distinct clonal expansion on disease progression (125).

Interestingly, despite an estimated potential TCR diversity of 10¹⁵ (126, 127), TCRs that recognize a common ligand typically exhibit convergent sequence features in CDR3 residues that directly contact the peptide (111, 128), as well as CDR1 and CDR2 residues that can also contact the peptide and MHC, which are also known as “public” TCR motifs (129). In previous HIV-related studies, viral control has been linked to the presence of more cross-reactive public TCR clones, which may play a role in limiting viral escape pathways (130, 131). Several studies have identified “public” TCR in COVID-19 convalescents, characterized by conserved CDR motifs within and between individuals (71, 104, 132, 133). Ford et al. identified public CD8⁺ and CD4⁺ TCR motifs associated with SARS-CoV-2 spike specificity through TCR sequence similarity clustering (116). Analysis of over 4,000 epitope-specific TCR sequences showed that all SARS-CoV-2 exposures elicit diverse repertoires characterized by shared TCR motifs, confirmed by monoclonal TCR characterization (134). Here, we summarized TCR repertoires for 22 epitopes (each with a gene count >30) from the public VDJ database (135), as shown in Table 2 . Consistent with previous studies (71, 133, 136), we found that TRAV12-1 (53%, 462/872) and TRBV7-9 (18%, 196/1110) were used by most YLQ-specific TCRs ( Figure 2B ). Likewise, TCRs recognizing other SARS-CoV-2 epitopes (e.g., NYN, LLY, PTD) display notable bias in the usage of specific TCR gene fragments ( Figures 2C–E ). For example, TRAV12-1 gene was commonly employed by YLQ, NYN, and PTD specific TCRs. Half of the epitopes we counted, such as SPR, TTD, QYI, LTD, RLQ, NYN, FTS, PTD, VYF, RVA, and WPV were dominated by the TRBJ2-7 gene ( Table 2 ). This strong bias of TCR gene usage among epitope-dependent TCRs likely highlights the significance of germline-encoded features in TCR recognition, as previously reported in other antiviral immune responses (137, 138).

In addition to public TCRs, the TCR repertoire generated in response to a specific epitope varies between individuals, often referred to as “private” responses. The RLQ-epitope (spike_1000-1008, RLQSLQTYV) is another immunodominant epitopes located on the SARS-CoV-2 spike protein, triggering cellular responses in most HLA-A*02:01⁺ convalescents (71). Unlike the highly “public” TCRs generated in response to the YLQ-epitope (71), TCRs responding to RLQ epitopes tend to be predominantly “private” and exhibit greater diversity ( Table 2 ). The recognition of A*02-RLQ by receptors with diverse CDR3α and CDR3β pairings diminishes the publicity of A*02-RLQ responses between individuals, enabling them to recognize MHC and peptide in a manner that reduces the likelihood of identical or very similar V(D)J rearrangements in different individuals (28).

Based on the genomic similarities between SARS-CoV-2 and HCoVs, cross-reactive T cells may underlie the extensive heterogeneity observed in COVID-19 disease. SARS-CoV-2 T cell reactivity was mostly associated with CD4⁺ T cells, with a smaller contribution by CD8⁺ T cells (9, 10, 139, 140). Nevertheless, some reports provide the basis for a limited representation of cross-reactive CD8⁺ T cell responses (19, 60, 84, 85, 141). Minervina et al. detected certain T cell clones in memory fragments at pre-infection time points, suggesting participation of pre-existing cross-reactive memory T cells in the immune response to SARS-CoV-2 (19). Francis et al. reported that the clonal diversity of T cell responses to HLA-B*07:02 allele correlates with pre-existing immunity, allowing efficient presentation of homologous epitopes from both SARS-CoV-2 and HCoVs (84). Cytotoxic T cell responses against SPR-HLA-B*07:02 (N_105-113, SPRWYFYYL) are often associated with mild cases of COVID-19 (80, 84, 142–145). Augusto et al., found that T cells from pre-pandemic samples from individuals carrying HLA-B*15:01 were reactive to the immunodominant SARS-CoV-2 spike-derived NQK epitope (60). Additionally, Shimizu et al. found that CD8⁺ T cells in response to a selected dominant QYI epitope display multifunctionality and cross-functionality across HCoVs in HLA-A24⁺ donors (141).

The HLA-B*07:02-restricted SPR epitope is one of the most cross-reactive epitopes, showing a high frequency in unexposed pre-pandemic samples (80, 145). Notably, the SPR peptide sequence is identical in SARS-CoV-2 and SARS-CoV-1, differing by only one residue in OC43-CoV-1 and HKU-1-CoV-1 (LPR), three residues in 229E (SPK) and four in NL63-CoV-1 virus (PPK), as shown in Figure 3A . Lineburg et al. reported that CD8⁺ T cells exhibit cross-reactivity between SARS-CoV-2 SPR and OC43/HKU-1-derived LPR peptide, but CD8⁺ T cells were unable to cross-recognize SPK and PPK peptides (85). The crystal structures of SPR-HLA-B*07:02 and SPK-HLA-B*07:02 elucidate these outcomes. Despite sharing common motifs in P1-2 (SP) and P6-9 (FYYL), differences in peptide sequences at P3-5 result in distinct conformations for SPK and SPR ( Figure 3B ). In the structure of SPR-HLA-B*07:02, four of the five aromatic residues create an interaction network, forming a compact and substantial binding surface for potential interaction with TCRs ( Figure 3C ). In comparison, the SPK peptide predominantly exposes the three carboxyl residues (P6-8) at the C-terminus ( Figure 3C ). Due to their high sequence identity, the LPR peptide (from OC43 and HKU-1) may adopt a conformation similar to that of SPR peptide (found in SARS-CoV-1 and SARS-CoV-2), providing the foundation for cross-reactivity among CD8⁺ T cells in HLA-B7⁺ individuals. However, despite sharing >55% (5/9) sequence identity, the distinctive structures account for the relatively low-level CD8⁺ T cell cross-reactivity observed between these peptides.

Most NQK-specific reactive T cells display a memory phenotype, exhibit highly polyfunctional, and cross-react to a peptide from seasonal coronaviruses (60, 61). The NQK peptide of SARS-CoV-2 differs by only one residue in SARS-CoV-1, OC43-CoV-1, and HKU-1-CoV-1, and by three residues in 229E-CoV-1 and NL63-CoV-1 virus ( Figure 3D ). Crystal structures of peptide-HLA-B15 complexes reveal that the peptides NQKLIANQF (from SARS-CoV-1) and NQKLIANAF (NQK-A8, from OC43-CoV and HKU1-CoV) have similar stabilization properties ( Figures 3E, F ). This structural similarity of the peptides underpins T cell cross-reactivity of high-affinity public T cell receptors, providing a molecular basis for HLA-B*15:01-mediated pre-existing immunity (60). A similar pattern of peptide binding to HLA was observed with another dominant epitope, HLA-A24 restricted QYI (141). The QYI exhibits high sequence homology with SARS-CoV-1, OC43 (YYV), HKU-1-CoV-1 (MYV), 229E-CoV-1 (TYI), and NL63-CoV-1 (NYI) ( Figure 3G ). The cross-reactivity of the QYI epitope depends on the structural pattern of the peptide-HLA-A*24:02 complex and the combinations of TCR sequences (141). The structures of QYI-HLA-A24, TYI-HLA-A24, and MYV-HLA-A24 are almost identical, with rmsd (root mean square deviation) of 0.15 Å (QYI and TYI), 0.41 Å (QYI and MYV), and 0.41 Å (TYI and MYV), respectively ( Figure 3H ). The side chains of the peptides at P1, P4, P5, P7 and P8 are directed to the solvent region, and their structural orientations are slightly different ( Figure 3I ). This particular region forms a raised structure and is expected to interact directly with TCRs by their side chains. It is reasonable to speculate that some CD8⁺ T cells targeting seasonal coronaviruses may exist as long-term memory cells within the population. If these cross-reactive T cells are stimulated by COVID-19 vaccine or viral antigen, they could be skewed toward SARS-CoV-2.

Given the active role of T cells in combating viral infection, some mutations could potentially lead to the loss of CTL epitopes that evade recognition by CD8⁺ T cells ( Table 3 ). Mutations within CD8⁺ epitopes in S protein (YLQ-L270F, LLF-L822F, and SII-S691P/S691C/I692T), M protein (GLM-L90F), N protein (MEV-M332I/L331S/L331F) were noted in one study during the course of acute infections, resulting in loss of epitope-specific responses (20). Prolonged SARS-CoV-2 infection in immunocompromised hosts may create a great opportunity for T cell escape. For instance, in the case of chronic SARS-CoV-2 infection, the NSP3 T504P mutation has been reported to result in the loss of CTL response (90, 146). These findings are limited to a few cases and suggest the need for more prospective cohort studies to systematically assess the risk of T cell escape in certain patient populations.

As the T cell responses target epitopes across the SARS-CoV-2 genome, the footprint of T cell escape is more broadly distributed than antibody-driven changes. In multiple SARS-CoV-2 lineages, some mutations within the immunodominant ORF1a (TTD- T1637I, T1638I, P1640S, P1640L, P1640H) ORF3a (FTS-Q213K) and N protein (QRN-P13L/S/T, KTF-T362I, P365S) CD8⁺ T cell epitopes resulted in a complete loss of recognition (88). Among these mutations, the N protein P13L mutation is present in Omicron within B*27:05-restricted CD8⁺ epitopes. Given the hypothesis that VOCs arise in chronic infections, it is tempting to speculate that the presence of P13L in Omicron reflects selection due to T cell stress during chronic infection, in addition to the constellation of spike mutations that are likely driven by antibody pressure. The mutant YLQ-P272L epitope could not be recognized by over 120 YLQ-specific TCRs, which may allow viral variants to escape from vaccine-induced T cell responses (28, 29). Spike-encoded L452R and Y453F led to the loss of HLA-A24-restricted CTL responses (91). In addition to evading antibodies and enhancing ACE2 binding affinity (147), the role of T cells in driving this change is uncertain. The extent to which these observations also represent incidental effects of mutations driven by other stresses on T cell responses is currently unknown.

T cell escape can occur through several mechanisms. Amino acid changes within epitopes or flanking regions can disrupt antigen processing (148), and changes to anchor residues can interfere with MHC/TCR binding to epitopes (29, 92, 149). Both these mechanisms can result in irreversible loss of T cell responsiveness to a particular epitope. To better understand the mechanism of CTL immune escape achieved by alterations in the binding between ligands and receptors, researchers have resolved several crystal structures of TCR-pHLA ternary complex and pHLAs loaded with original or mutant SARS-CoV-2 peptides, including YLQ, YLQ-P272L, RLQ, RLQ-T1006I, NYN, NYN-Y453F, KIA, KIA-K417T peptides ( Table 4 ). All of eight TCR-pMHC ternary complexes loaded with SARS-CoV-2 peptides are symmetrically docked on pHLAs in a canonical diagonal orientation ( Figure 4 ).

Mutations within the epitope may destabilize or reassemble the pHLA complexes (20, 92). For example, our previous study showed that the K417N and Y144^del mutations lead to failure in the formation of the KIA-HLA-A2 and GVY-HLA-A11 complexes, respectively, blocking the first step in antigen presentation (92). Additionally, structural analysis shows that the positively charged side chain of N-terminal lysine forms a π-cation interaction with the indole ring of W167 of HLA-A2, and K417N/T mutations at this position (K417T) are expected to abolish the π-cation interaction at this position (92). The loss of this key peptide-HLA interactions may provide SARS-CoV-2 variants with an opportunity to evade cellular immunity.

Mutations within the epitope may disturb or change the peptide-dependent TCR contacts (29, 149). For the dominant YLQ epitopes, the original YLQ peptide forms six intra-peptide bonds to stabilize TCR binding, while the mutant YLQ-P272L-HLA-A2 (YLQLRTFLL) structure exhibits fewer internal contacts ( Figure 5A ). The P5 arginine of the peptide underwent a significant conformational shift during the binding of TCR YLQ36 ( Figure 5B ). Superimposed structures of YLQ-P272L-HLA-A2 and YLQ36-YLQ-HLA2 complex revealed that the leucine of YLQ-P272L might protrude within 1˚ of the YLQ36 CDR3α loop, creating a steric hindrance between them ( Figure 5C ). This steric hindrance could potentially jeopardize the interaction between CDR3α and YLQ peptide, resulting in the loss of YLQ36 T cell recognition. Similarly, the spik-Y453F mutation within the NYN epitope is another dominant mutation that triggers HLA-A24-restricted CTL immune escape (91). In order to explore the escape molecular mechanism, we determined the crystal structures of original NYN-HLA-A24 (NYNYLYRLF) (92), mutant NYN-Y453F-HLA-A24 (NYNYLFRLF), and a ternary structure of TCR^NYN-I-NYN-HLA-A24 (149). Structural analysis showed that after mutation or TCR^NYN-I binding, the conformation of the NYN peptide changed significantly, especially P4-Tyr, P5-Leu, and P6-Tyr/Phe ( Figures 5D, E ). The hydrophobic phenylalanine in the SARS-CoV-2 variants may disrupt contact network of the original tyrosine with TCR^NYN-I, suggesting that despite competent presentation by HLA, the mutant Y453F peptide failed to establish a stable TCR-pHLA ternary complex due to reduced peptide: TCR contacts( Figure 5F ). Unlike the P272L and Y453F dominant mutations, the RLQ-T1006I mutation was found to be tolerated in HLA-A24-restricted CTL activation (28, 30, 151). The T1006I substitution leads to a structural rearrangement of the HLA peptide-binding groove and alters the conformation of peptide residues P3-Gln and P6-Gln ( Figure 5G ). TCR RLQ7 CDR1α engages the N-terminal region of the original RLQ peptide via two direct and three water-mediated hydrogen bonds ( Figure 5H ), while TCR RLQ7 forms two new compensating hydrogen bonds with mutant RLQ-T1006I peptide (Asp31α with P4-Ser, Asp31α with P5-Leu) and an additional hydrogen bond with HLA-A2 (Glu29α with Arg66-HLA-A2) ( Figure 5I ). Similar stabilities of the two complexes may provide a reasonable explanation for the limited CTL immune escape.

Disruption or inhibition of antigen processing by mutation is a third mechanism of T cell immune escape. The components of the antigen processing pathway have preferences for their optimal amino acid sequences (152). It has been demonstrated in HIV-related studies that differences in amino acid sequence in the region flanking the epitope impaired the intracellular processing and presentation of epitope (153, 154). In the context of SARS-CoV-2 infection, a study proposed that variants may disrupt the HLA-I antigen presentation pathway by depleting proteasomes and altering the activity of ubiquitination enzymes, thereby preventing infected cells from presenting antigen proteins (148). A deeper understanding of the mechanisms by which SARS-CoV-2 evades host immune surveillance through the ubiquitin-proteasome system and HLA-I presentation warrants further investigation.