Broad cross-T cell immunity between emerging SARS-CoV-2 serotypes waved by spike-signature mutations

ABSTRACT

Since its emergence in late 2019, severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) has continuously threatened human health through constantly arising variants with iterative immune escape abilities. While SARS-CoV-2 variants have been recently classified into six serotypes based on cross-reactive antibody responses, T cell response features against these serotypes remain largely unknown. We evaluated SARS-CoV-2 spike-specific T cell responses among convalescents infected by three different strains (prototype, BA.5.2/BF.7, and XBB/EG.5.1) against SARS-CoV-2 prototype and 15 subvariants covering all six serotypes. Generally, cross-reactive T cells could recognize variants within the same serotype, but they also mounted weaker responses to variants from subsequent serotypes. Serotype I (prototype) convalescents showed lower T cell responses against Omicron variants (Serotype II to IV), with cross-reactive T cell gaps between different serotype strains, i.e. Serotype II > III > IV. Serotype IV (BA.5.2/BF.7) convalescents exhibited weaker T cell responses to Serotype V (XBB/XBB.1.5/XBB.1.16/EG.5.1) strains and even lower responses to Serotype VI (BA.2.86/JN.1) strains. Similarly, Serotype V (XBB) convalescents showed significantly weaker cross-T cell responses to the Serotype VI (BA.2.86) strains than to the Serotype V strains. We also identified key serotype-signature mutations in T cell epitope hotspot regions that could attenuate CD8⁺ and/or CD4⁺ T cell recognition, potentially underlying SARS-CoV-2 serotype-associated T cell immune evasion mechanisms. Our findings reveal the pivotal role of population T cell immune barrier against emerging SARS-CoV-2 variants in a serotype-associated pattern and provide insights into the T cell-oriented universal vaccine development for coronaviruses.

GRAPHICAL ABSTRACT

Introduction

Over five years on, coronavirus disease 2019 (COVID-19) continues to impact global public health [1,2]. Since its emergence, severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) has rapidly evolved, giving rise to multiple new variants from Alpha (B.1.1.7) to Omicron (B.1.1.529), etc., which have been responsible for numerous global outbreaks [3–5]. Among them, the Omicron variant, which was first identified in South Africa in November 2021, rapidly became the dominant strain worldwide [6]. The high transmissibility and striking antibody evasion properties of the Omicron variant have caused a surge in COVID-19 cases and hospitalizations, overwhelming healthcare systems in many parts of the world [7]. The Omicron lineage has further diversified into subvariants including BA.2.75, BA.5.2/BF.7, XBB recombinants (XBB.1.5, XBB.1.16), and BA.2.86/JN.1 [8]. By mid-2025, JN.1 sub-lineages (KP.3, XEC, NB.1.8.1, etc.) dominate globally, exhibiting enhanced transmissibility and immune evasion [9,10]. Studies of the antibody response to the SARS-CoV-2 lineages and sub-lineages, preliminarily demonstrate the adaptability and antibody evasion capability [11,12]. However, assessing how mutations in these SARS-CoV-2 variants contribute to T cell immune escape remains a veritable challenge.

Viral serotyping is a reliable virological classification method that is based on the cross-reactivity of neutralizing antibodies (NAbs) within the sera of convalescent donors with related viral infections. This method is commonly used for dengue virus classification [13,14]. To better delineate the immune evasion characteristics of the SARS-CoV-2 variants, studies have classified SARS-CoV-2 into different serum subtypes based on the spike protein and receptor-binding domain (RBD) of the SARS-CoV-2 [15–18]. Hu et al. classified SARS-CoV-2 into six serotypes (Serotypes I to VI), with the strains arising before Omicron variant classified as Serotype I. The Omicron variant is further subdivided into five serotypes (Serotypes II to VI), which are represented by strains such as BA.1, BA.2.75, BA.5.2/BF.7, XBB/XBB.1.5, and the recently emerged BA.2.86/JN.1, respectively [17–20]. These were generally concordant with the serotypes of SARS-CoV-2, together with SARS-CoV defined by Wang’s group [16]. SARS-CoV-2 serotype definition can provide insights into development and usage of diagnostic tools, treatments, and vaccines [21,22].

SARS-CoV-2 Omicron variants evade the immune system mainly by acquiring mutations at key S protein residues, which render NAbs generated against previous strains ineffective [23]. The current classification of SARS-CoV-2 serotypes and the immune evasion characteristics of different variants mainly rely on the recognition of S protein RBD by antibodies [15,16]. It has been indicated that the S protein is also a dominant target for T cell responses, although the sites recognized by T cells are relatively broad [24–26]. Recent studies have shown that T cells can effectively recognize Omicron variants through conserved epitopes, including the hypermutated BA.2.86 variant [27–29]. However, the potential impact of accumulated S protein-loaded mutations on T cell immunity of different serotypes remains to be fully elucidated. Although T cell immunity targets multiple viral proteins (N, M, S, etc.), the S protein exhibits particular immune dominance in natural infection and harbours most mutations. Since the S protein is the dominant target for both humoral and cellular immunity, it is critical for understanding T cell-mediated immune evasion.

Herein, we systematically determined the S-specific cross-reactive T cell responses and evasion profiles of 16 SARS-CoV-2 variants, which included all six serotypes from I to VI. We observe that broad cross-T cell immunity exists between the variants within similar serotypes, but there are T cell immune gaps between different serotypes, which indicates an association of T cell response evasion and SARS-CoV-2 serotypes. These findings reveal the dynamics of population T cell immune barrier against SARS-CoV-2 evolution and provide insights for developing T cell-oriented universal vaccines that target conserved epitopes across serotypes.

Materials and methods

Sample collection

Peripheral blood mononuclear cells (PBMCs) were collected from volunteers infected with BA.5.2/BF.7 14–28 days (n = 65) and 6 months (n = 41) post-recovery. Among the 65 enrolled participants, 34 belonged to a longitudinal cohort, while 31 dropped out due to personal reasons. Seven additional subjects with matched baseline characteristics were recruited to enhance statistical power within the scope of ethical approval. PBMCs were also collected from XBB-infected individuals 14 days (n = 25) post-recovery, and those infected with the prototype 14–28 days (n = 48) post-recovery (Supplementary Table 4). The prototype convalescents were recruited through hospital referrals, while Omicron convalescent patients were recruited through open recruitment in the community. The SARS-CoV-2 infections were laboratory-confirmed by real-time PCR positive using nasopharyngeal swab specimens. Information such as infection time was collected by questionnaire and the infected status of SARS-CoV-2 was defined according to the self-reported symptom onset during the epidemic peaks of highest frequency of variants of SARS-CoV-2. We also performed next-generation sequencing on a portion of the samples and identified the infecting variants, which met our definition. To evaluate BA.2.86/JN.1 (Serotype VI) immune evasion in early 2024, PBMCs were collected from BA.5.2/BF.7 convalescents (n = 16), XBB/EG.5.1 convalescents (n = 4), and BA.5.2/BF.7 convalescents who were subsequently reinfected with XBB/ EG.5.1 (n = 12) (Supplementary Table 5). Infection status was based on the peak period of the epidemic, when the frequency of SARS-CoV-2 mutations was the highest (https://gisaid.org/, December 2022 to January 2023 for BA.5.2/BF.7, and May 2023 to September 2023 for XBB/EG.5.1). All donors reported no respiratory symptoms in the 3 months before the most recent infection of SARS-CoV-2. Isolated PBMCs were frozen in cell stock solution containing 90% fetal bovine serum (FBS) with 10% dimethyl sulfoxide (DMSO) and stored in liquid nitrogen for later use. All subjects provided written informed consent. The HLA typing was done at 4- or 6-digit resolution by Sanger sequencing or LABType SSO tests. The HLA distribution of the cohorts in the study could be representative of the population, covering most common HLA alleles of population (Supplementary Table 6). The study was approved by the Ethics Committee of National Institute for Viral Disease Control and Prevention, Chinese Center for Disease Control and Prevention (China CDC) (ethical approval no. IVDC2020-021; IVDC2023-005).

Peptides

According to the S protein sequences of the SARS-CoV-2 prototype and variants, peptides of 15–18 amino acids were designed and synthesized, each of which overlapped by 10 amino acids. The S1 protein peptide pool included a total of 101 peptides, with 30 peptides for the RBD, while 86 peptides spanned the S2 protein as the S2 peptide pool (Supplementary Table 7). S1 or S2 peptide pools for the SARS-CoV-2 variants were designed based on the mutation sites present in each variant. All the peptides were individually resuspended in DMSO at a concentration of 10–20 mg/mL. T cell epitope hotspots on the S protein are defined as immunodominant regions that contained more than two spatially overlapping MHC class I or II epitopes recorded in the IEDB database. These overlapping epitopes need to be limited to more than two different HLA alleles, representing regions of immune dominance at the population level.

In vitro T cell expansion

Briefly, PBMCs were cultured in 48-well plates at a density of 2–3 × 10⁶ cells/well for 9–12 days in the presence of S1/S2 peptide pools and interleukin (IL)-7 (25 ng/mL). Half of the culture medium was replaced with medium containing IL-2 (200 U/mL) every 3 days, and peptide stimulation was not repeated during this period. This approach simulates the physiological T cell activation microenvironment and preferentially supports the survival and expansion of antigen-specific memory T cells. Depending on the strain the convalescents had been infected with, the S1/S2 peptide pool of the corresponding strain was used to culture PBMCs. We were challenged by the fact that BF.7 and BA.5.2 were prevalent in the population at that time of the study and that both variants had only one different amino acid mutation in their S1 protein. For patients who recovered from BA.5.2/BF.7 for 14–28 days, we cultured the PBMCs with the S1 peptide pools of both strains when there were sufficient cells, and found that the test results were consistent (Figure 6). Therefore, we subsequently cultured the PBMCs of patients who recovered from BA.5.2/BF.7 for 6 months using the BF.7-S1 peptide pool (Figure 3). While the effects of the R346T mutation on memory T cells may change subtly over time, its impact on population-level responses is limited. The PBMCs were then harvested and subjected to an enzyme-linked immune spot (ELISpot) and intracellular cytokine staining (ICS) assays.

Figure 6.

Similar lower S1-specific cross-T cell responses to XBB reignited by BF.7 and BA.5.2 peptides in Serotype IV convalescents. PBMCs cultured with the BF.7-S1 peptide pool were exposed to the S1 (A) or RBD (B) peptide pools and assayed by ELISpot. The individual (left), averaged/log2-transformed (middle), and Serotype IV versus V comparison (right) data are shown (see Figure 3 legend for more detail). PBMCs cultured with the BA.5.2-S1 peptide pool were exposed to the S1 (C) or RBD (D) peptide pools and assayed by ELISpot. As for A and B, the individual (left), averaged/log2-transformed (middle), and Serotype IV versus V comparison (right) data are shown. In the whisker box plots, the top and bottom of the boxes indicate the interquartile range, and the horizontal line within each box indicates the median. In the line graphs, the line shows averaged and log2-transformed T cell response data. Each colour denotes a different serotype. Comparisons between two groups with statistically significant differences were marked at the top of the figure. “*” in the figure indicates that the T cell response to a given strain was statistically different from that of the BA.5.2/BF.7 variant. The error bars indicate 95% confidence intervals. The Wilcoxon matched-pairs signed rank test was used to compare differences between the T cell responses elicited by the different SARS-CoV-2 variants. A paired-sample t-test was used to analyze the log2-transformed mean value of each serotype.

Figure 3.

S1-specific cross-T cell responses to different serotypes in BA.5.2/BF.7 convalescents. PBMCs were cultured with the BF.7-S1 peptide pool, which was verified to yield equivalent results to BA.5.2-S1 (validation shown in Figure 6). Spike S1 (A) and RBD (B) peptide pool ELISpot assay results for each SARS-CoV-2 variant (left) and each serotype (right). In the whisker box plots (left), the top and bottom of the boxes indicate the interquartile range, and the horizontal line within each box indicates the median. In the line graphs (right), the line shows averaged and log2-transformed T cell response data. Each colour denotes a different serotype. Comparisons between two groups with statistically significant differences were marked at the top of the figure. The error bars indicate 95% confidence intervals. “*” in the figure indicates that the T cell response to a given strain was statistically different from that of the BA.5.2/BF.7 variant. (C-D) Paired comparison of T cell responses between serotypes. Each line connects paired measurements from the same individual. Lines with decreasing slope (coloured) indicate individuals showing weaker responses to Serotype V (C, D left) or VI (C, D right) compared to Serotype IV. Lines with increasing slope (gray) indicate stronger responses to the compared serotype. The proportion of individuals with decreased responses and P values are shown. Results of the intracellular cytokine staining (ICS) assay generated by exposing CD4⁺ (E) or CD8⁺ (F) T cells after stimulation with S1 peptide pools to each variant. The IFN-γ- and TNF-α-producing T cells are shown in this figure, with the statistical differences noted above. Each colour is used to show a different variant. Because the experimental procedure involved performing ELISpot assays first, with the remaining cells used for ICS analysis, some comparisons did not reach statistical significance. Therefore, ICS data should be interpreted with caution as supplementary evidence supporting the step-down trend, rather than as an independent confirmatory conclusion. (G) Representative flow cytometry plots from a single PBMC donor. Flow cytometry gating strategy identified all cytokine-producing cells within CD4⁺ and CD8⁺ populations, including cells with varying CD4/CD8 expression levels. The Wilcoxon matched-pairs signed rank test was used to compare differences between the T cell responses elicited by the different SARS-CoV-2 variants. A paired-sample t-test was used to analyze the log2-transformed mean value of each serotype. The P values in the figures (here and in subsequent figures) were not corrected for multiple comparisons, as correction would reduce the statistical power to detect significant differences. Given the sample size limitations, this approach may be better at detecting biologically meaningful differences. However, we acknowledge the increased risk of false positives and therefore also provide multiple comparison correction results in Supplementary Table 10 for reference.

ELISpot assays

The assay detects the S1 and RBD peptide pools of the cultured strain and other variants to evaluate the crossover and escape of S1/RBD-specific T cell responses. The interferon (IFN)-γ production of T cells within the PBMC sample in response to stimulation with the overlapping peptide pools or a single peptide was detected using human IFN-γ ELISpot assay kits (BD Biosciences, Franklin Lakes, NJ, USA), as described previously [30]. Briefly, a 96-well ELISpot plate was precoated with 10 μg/mL of anti-IFN-γ monoclonal antibody and incubated overnight at 4 °C. The plates were washed with sterile PBS and blocked with RPMI 1640 medium containing 10% FBS (Gibco, Grand Island, NY, USA) at room temperature for 2 h. PBMCs (10⁵ cells/well) were added to each well along with either an overlapping peptide pool (2 μg/mL of each peptide) or a single peptide (10 μg/mL). The same volume of RPMI 1640 medium alone or medium containing phorbol-12-myristate-13-acetate (PMA) was added to the unstimulated control or positive control well, respectively. The cells were incubated in a humidified incubator, at 37 °C, with 5% CO₂ for 18–24 h. IFN-γ spots were detected after the addition of a biotinylated detection antibody, followed by streptavidin-horseradish peroxidase (HRP) conjugate and substrate. The results were expressed as the number of spot-forming cells (SFCs) per 10⁶ PBMCs, which were counted using an ELISpot Reader System (CTL Corp, USA). ELISpot positive criteria: (a) SFCs in the stimulated peptide group were significantly higher than those in the negative control (mock, no peptide stimulation). (b) SFCs in the peptide-stimulated group were at least twice that in the negative control. (c) SFCs ≥ 50/10⁶ PBMCs (background subtracted).

ICS assay

The cytokine profiles (including IFN-γ, TNF-α, IL-2, IL-6, and IL-4) of CD4⁺ and CD8⁺ T cells were detected using an ICS assay after stimulation of PBMCs with S1-protein or RBD peptides. Briefly, cultured PBMCs were added to each tube (10⁶ cells/tube) along with either an overlapping peptide pool or a stimulating peptide. After 1 h incubation at 37 °C, 5% CO₂, the cells were washed with RPMI 1640 medium containing 10% FBS, and incubated with medium containing Golgi-Stop and Golgi-plug (BD) for 9–12 h. Then, the cells were washed with PBS and stained with APC-Cy7 (Live/dead fixable aqua dead cell stain kit) for 20 min at 37 °C. Cells were washed with fluorescence-activated cell sorting (FACS) buffer containing 5% FBS, and then stained with the following antibodies on ice for 30 min: anti-CD3 FITC, anti-CD4 Per-Cy 5.5, and anti-CD8 BV510. Cells were then washed, fixed with the IntraStain Fixative/Permeabilization reagent, and stained with the following monoclonal antibodies on ice for 30 min: anti-IFN-γ PE-Cy7, anti-TNF-α PE, anti-IL-2 APC, anti-IL-6 594/Texas Red, and anti-IL-4 BV421. After washing, the cells were analyzed on a BD FACSAria II flow cytometry (BD Biosciences), and the data were analyzed using FlowJo software. ICS positive criteria for percentage of secreted cytokines: The percentage of positive cytokines in the peptide pool stimulation group was at least 0.05% higher than that in the negative control (mock, no peptide stimulation), and at least twice that of the negative control.

Sequence alignment and evolutionary analysis

All sequences in this study were downloaded from NCBI Virus (https://www.ncbi.nlm.nih.gov/labs/virus/) and GISAID (https://www.gisaid.org/). GenBank accession numbers are listed in Supplementary Table 8. Multiple sequence alignment and evolutionary sequence analysis were performed using Clustal W algorithm, and phylogenetic trees were constructed using MEGA X, based on full-length S protein amino acid sequences. The Maximum Likelihood method was used with 1,000 bootstrap iterations, partial deletion, and a 95% site coverage cutoff. The tree with the highest log likelihood was shown. Jalview software was used to align, edit, and visualize the S protein amino acid sequences of the different variants. Structural visualization and mutation mapping were performed using PyMOL software (version 2.5.0). PDB ID used: 7xix. Mutation sites were mapped onto the spike trimer structure with colour-coding as described in the main text.

Statistical analysis

All statistical analyses were conducted using GraphPad Prism 9.5.0 or SPSS 26.0. GraphPad Prism 9.5.0 was also used to generate graphs. Specific cellular IFN-γ responses were presented as the median with interquartile range (IQR). The Wilcoxon matched-pairs signed rank test and the Mann-Whitney U-test were used to evaluate differences between SARS-CoV-2 variants. All data were log2-transformed to normalize the distribution and improve comparability. A paired-sample t-test was used to analyze the log2-transformed mean value of each serotype. All tests were two-tailed. The significance threshold was set at a P-value < 0.05; * P < 0.05, ** P < 0.01, *** P < 0.001, and **** P < 0.0001.

Results

T cell epitope hotspots around the serotype-signature mutation sites

The phylogenetic tree based on the representative SARS-CoV-2 strains from the six serotypes (Figure 1(A)) demonstrated the evolutionary relationships among the SARS-CoV-2 variants aligned with their serotype classifications. The prototype, Alpha, Beta, Gamma, and Delta variants were classified as Serotype I, while the Omicron subvariants were placed into the five other serotype classes (Serotype II: BA.1 and BA.1.1; Serotype III: BA.2, BA.2.12.1, BA.2.75, and CH.1.1; Serotype IV: BA.5.2, BF.7, BQ.1, and BQ.1.1; Serotype V: XBB, XBB.1.5, XBB.1.16, and EG.5.1; Serotype VI: BA.2.86 and JN.1). The evolutionary trajectories of the representative strains of each Omicron variant serotype in our evolutionary diagram (Figure 1(B)) based on previous studies, demonstrated that all variants other than BA.1 originated from BA.2, with XBB subvariants arising due to the recombination between BM.1.1.1 and BJ.1 [31]. The mutation sites in the RBD region of the S protein with key roles in SARS-CoV-2 evolution indicated that [32–34], certain sites (346, 444, 452, 460, 486, etc.) were repeatedly mutated and substituted in the variants, while variants within the same serotype differed by only a few additional mutations. For instance, BF.7 could be distinguished from BA.5.2 by the R346T mutation, while JN.1 differed from BA.2.86 by the L455S mutation.

Figure 1.

Evolutionary relationship, mutation sites, and T cell epitope hotspots evaluation of variants belonging to the six known serotypes of SARS-CoV-2. (A) Phylogenetic tree of representative SARS-CoV-2 variants. The phylogenetic tree was constructed using the maximum likelihood method with a bootstrap value of 1,000. Each serotype is shown in a different colour. (B) Evolutionary relationship between the Omicron variants. Each serotype is shown in a different colour, with filled boxes used to highlight the strains used in the T cell response evaluation experiments in this study. The red font highlights the key mutation sites that accumulated within the RBD region of the S protein during viral evolution. (C) Schematic diagram of the “serotype-signature mutation sites” present within the S proteins 1 and 2 of different serotypes. Lines above and below the mutation site indicate MHC class I and II epitopes (identified in the IEDB), respectively. The HLA class I and II epitope hotspot areas are marked with red and purple dashed boxes, respectively. (D) A sequence alignment diagram of the S protein mutation sites belonging to the different serotypes. Sequence alignment was performed using the Clustal W algorithm, with Jalview software used to generate consensus sequences and refine them. The “raindrops” in the diagram represent amino acid positions that have never been mutated. The “serotype-signature mutation sites” of each serotype are shown in various colours. The “cumulative mutation sites,” which appear in one of the intermediate variants and are fixed in subsequent strains, are highlighted in blue. T cell epitope hotspots on the S protein are defined as immunodominant regions that contained more than two spatially overlapping MHC class I or II epitopes recorded in the IEDB database. These overlapping epitopes need to be limited to more than two different HLA alleles, representing regions of immune dominance at the population level.

We next scanned the variants for unique mutation sites (i.e. mutation sites that were shared by more than half of the strains of a certain serotype but not found in other serotypes) in the S proteins of the different serotypes. Compared with Serotype I, the S protein of Serotype II had ten unique mutation sites, Serotype III had five, Serotype IV had one, Serotype V had eight, and Serotype VI had 21 sites (Figure 1(C)). All the mutation sites present in the variants relative to the prototype (Figure 1(D)) can be divided into serotype-signature mutation sites specific to each serotype and cumulative mutation sites (which appeared in one of the intermediate variants and were fixed in subsequent strains). Notably, most of these unique mutation sites were also present within the identified MHC-class-I- and class-II-restricted T cell epitopes found within the Immune Epitope Database (IEDB), suggesting that these mutations could impact T cell immune responses. Specifically, these serotype-signature mutations affected multiple MHC-restricted epitopes, with Serotype VI showing the most extensive impact (41 MHC I and 63 MHC II epitopes) (Figure 1(C), Supplementary Tables 1 and 2).

To comprehensively evaluate the impact of these mutations on T cell immunity across the viral proteome, we analyzed epitope distribution and conservation across viral proteins from the IEDB database using the prototype reference strain (Figure 2). The results revealed S protein as the dominant target for CD4⁺ T cells (46.2% of total epitopes) and a major target for CD8⁺ T cells (24.0%) (Figure 2(A)). Conservation analysis across serotypes demonstrated that the S protein epitopes underwent a gradual decrease (CD4⁺ from 100% to 54.2%; CD8⁺ from 100% to 75.3%, from Serotype I to VI), while epitopes in N, M, E, ORF1ab, and other proteins remained highly conserved (Figure 2(B)). These conserved protein epitopes provide stable cross-reactive Tcell immune responses even when the S protein escapes significantly. These findings provide a foundation for identifying conserved epitopes that could serve as targets for cross-protective immunity and universal vaccine design. Given the immunodominance of the S protein and the significant serotype-dependent variation concentrated in the S1 region (positions 1–685), our subsequent analysis focused on S1-specific T cell responses.

Figure 2.

Comprehensive analysis of T cell epitope conservation in SARS-CoV-2 serotypes. (A) Distribution of CD8⁺ and CD4⁺ T cell epitopes in viral proteins based on the IEDB database (Serotype I, prototype). A pie chart shows the number and proportion of epitopes for each protein (S, spike protein; M, membrane protein; E, envelope protein; N, nucleocapsid protein; ORF1ab, open reading frame 1ab; others, other structural proteins). Notably, the S protein accounts for 46.2% of CD4⁺ epitopes and 24.0% of CD8⁺ epitopes, making it one of the dominant targets in T cell immune responses. (B) The heatmap shows the epitope conservation rates of the six SARS-CoV-2 serotypes (I-VI). By determining the proportion of epitopes unaffected by serotype-specific mutations, the conservation rate of these epitopes across the six serotypes was calculated. Colour intensity indicates the percentage of conserved epitopes: dark blue (highly conserved), light blue to light red transition (moderately conserved), and dark red (lowly conserved). The numbers in the cells represent the precise percentage of conservation. S protein epitope conservation gradually decreases from Serotype I (100%) to Serotype VI (CD8: 75.3%; CD4: 54.2%), while other structural proteins (M, E, N) and non-structural proteins (ORF1ab, others) maintain a high degree of conservation. This suggests that other structural proteins may provide stable cross-reactive T cell immunity despite spike protein mutations.

Step down of S1-specific cross-T cell responses to Serotypes V and VI in Serotype IV convalescents

We next assessed the S1-specific T cell responses of individuals after BA.5.2/BF.7 infection by ELISpot and ICS assay. We found that the majority of individuals mounted S1-specific T cell responses against Serotype IV (BA.5.2/BF.7) post-infection. However, we found that the magnitude of the cross-reactive T cell immune responses (measured as SFCs/10⁶ PBMCs) elicited against Serotype IV declined on exposure to each subsequent Serotype V (XBB.1.16/EG.5.1) and Serotype VI (BA.2.86/JN.1) (Figure 3(A)). After averaging and log2-transforming the data, we found that the T cell response values of Serotype IV (mean, 10.86 Log₂ SFCs/10⁶ PBMCs; 95% confidence interval [CI], 10.62–11.10) were significantly higher than those of Serotype V (mean, 10.80 Log₂ SFCs/10⁶ PBMCs; 95% CI, 10.57–11.02; P = 0.0054) and Serotype VI (mean, 10.40 Log₂ SFCs/10⁶ PBMCs; 95% CI, 10.13–10.68; P < 0.0001) (Figure 3(A)). The T cell response values of Serotype V were also significantly higher than those of Serotype VI (P < 0.0001). RBD peptide pools stimulation showed similar patterns (IV: 10.32 vs V: 10.12 vs VI: 9.87, all P < 0.0001) (Figure 3(B)). These results suggest that each subsequent variant gained an escape advantage over its predecessor serotype.

To clearly compare the decreased trend from Serotype IV to V and VI, we performed a pairwise decrease curve analysis on the averaged data obtained for each serotype. The results of the S1 peptide pool experiments showed that, among the 41 donor samples tested, 31 showed lower responses to Serotype V than Serotype IV (75.6%, P = 0.0012) and 39 showed lower responses to Serotype VI than Serotype IV (95.1%, P < 0.0001) (Figure 3(C)). Similar decreases in T cell responses were observed in assays conducted with the RBD peptide pools, with 78.0% (32/41) and 87.8% (36/41) of individuals showed reduced responses to Serotypes V and VI, respectively (P < 0.001) (Figure 3(D)), indicating progressive immune escape with emerging serotypes.

Next, we performed ICS assays to measure the production of IFN-γ and TNF-α by T cells exposed to the different variants, by using representative strains of each serotype (prototype, BF.7, XBB.1.5, BA.2.86, and JN.1) due to limited PBMC numbers. A declining trend in IFN-γ and TNF-α-secreting CD4⁺ and CD8⁺ T cells was detected in response to BA.2.86 and JN.1 than in response to BF.7 and XBB.1.5 (Figure 3(E and F)). For IFN-γ-secreting CD4⁺ T cells, the proportion was higher when stimulated with BF.7 (median, 0.68%; IQR, 0.43%–1.23%) and XBB.1.5 (median, 0.75%; IQR, 0.42%–1.29%) compared to BA.2.86 (median, 0.49%; IQR, 0.33%–0.81%) and JN.1 (median, 0.49%; IQR, 0.29%–0.91%) (Figure 3(E), Supplementary Table 9). Similarly, the proportion of TNF-α-secreting CD4⁺ T cells induced by BF.7 (median, 1.66%; IQR, 0.99%–3.05%) and XBB.1.5 (median, 1.83%; IQR, 1.16%–2.37%) was also higher than that activated by BA.2.86 (median, 1.09%; IQR, 0.69%–1.81%) and JN.1 (median, 1.27%; IQR, 0.67%–1.84%). Similar patterns emerged for CD8⁺ T cells, although the differences were less pronounced than those observed for CD4⁺ T cells (Figure 3(F), Supplementary Table 9). There was no significant difference in the levels of cytokines secreted by both CD4⁺ and CD8⁺ T cells in response to BF.7 and XBB.1.5, although one representative donor had reduced cytokine secretion in response to XBB.1.5 (Figure 3(G)).

Degraded S1-specific T cell responses to Serotype VI in Serotype V convalescents

Subsequently, we evaluated the S1-specific T cell responses to 12 variants among individuals after XBB infection. The results of S1 peptide pool stimulation ELISpot experiments showed that individuals mounted a significantly lower response to BA.2.86 of Serotype VI (median, 790 SFCs/10⁶ PBMCs; IQR, 335–1275) than each strain in Serotype V, including XBB (median, 810 SFCs/10⁶ PBMCs; IQR, 385–1465; P = 0.0004), XBB.1.5 (median, 900 SFCs/10⁶ PBMCs; IQR, 455–1410; P = 0.0022), XBB.1.16 (median, 1030 SFCs/10⁶ PBMCs; IQR, 440–1340; P = 0.0001), and EG.5.1 (median, 860 SFCs/10⁶ PBMCs; IQR, 365–1465; P = 0.0006) (Figure 4(A)). Log2-transformed analysis confirmed higher responses to Serotype V (mean, 9.68 Log₂ SFCs/10⁶ PBMCs; 95% CI, 9.18–10.19; P < 0.0001) than Serotype VI (mean, 9.30 Log₂ SFCs/10⁶ PBMCs; 95% CI, 8.73–9.87) (Figure 4(A)). The results of the RBD peptide pool assay were consistent with the above findings. The T cell response values for XBB (median, 610 SFCs/10⁶ PBMCs; IQR, 285–760; P = 0.0003), XBB.1.5 (median, 550 SFCs/10⁶ PBMCs; IQR, 200–960; P = 0.0089), XBB.1.16 (median, 630 SFCs/10⁶ PBMCs; IQR, 175–825; P = 0.0007), and EG.5.1 (median, 560 SFCs/10⁶ PBMCs; IQR, 160–820) were higher than those for BA.2.86 (median, 450 SFCs/10⁶ PBMCs; IQR, 140–590) (Figure 4(B)). The log2-transformed data also showed the same decreasing pattern (V: 8.70 vs VI: 8.30, P = 0.0028) (Figure 4(B)). Pairwise analysis revealed 84.0% (21/25) and 76.0% (19/25) of individuals showed reduced responses from Serotype V to VI in S1 and RBD assays, respectively (both P < 0.001) (Figure 4(C and D)). These findings indicate that most XBB convalescents could not mount effective S1-specific T cell immunity against BA.2.86 from Serotype VI.

Figure 4.

S1-specific cross-T cell responses to different serotypes in XBB convalescents. Spike S1 (A) and RBD (B) peptide pool ELISpot assay results for each SARS-CoV-2 variant (left) and each serotype (right). In the whisker box plots (left), the top and bottom of the boxes indicate the interquartile range, and the horizontal line within each box indicates the median. In the line graphs (right), the line shows averaged and log2-transformed T cell response data. Each colour denotes a different serotype. Circles represent individuals reinfected with XBB after BA.5.2/BF.7 (n = 19), and boxes represent primary XBB infections (n = 6). Comparisons between two groups with statistically significant differences were marked at the top of the figure. The error bars indicate 95% confidence intervals. “*” in the figure indicates that the T cell response to a given strain was statistically different from that of the BA.2.86 variant. (C-D) Paired comparison of T cell responses between serotypes. Each line connects paired measurements from the same individual. Lines with decreasing slope (red) indicate individuals showing weaker responses to VI compared to Serotype V. Lines with increasing slope (gray) indicate stronger responses to the compared serotype. The proportion of individuals with decreased responses and P values are shown. Results of the intracellular cytokine staining (ICS) assay generated by exposing CD4⁺ (E) or CD8⁺ (F) T cells after stimulation with S1 peptide pools to each variant. Only the IFN-γ- and TNF-α-producing T cells are shown in this figure, with the statistical differences noted above. Each colour is used to show a different variant. (G) Representative flow cytometry plots from a single PBMC donor. The Wilcoxon matched-pairs signed rank test was used to compare differences between the T cell responses elicited by the different SARS-CoV-2 variants. A paired-sample t-test was used to analyze the log2-transformed mean value of each serotype.

ICS assays showed reduced IFN-γ⁺ and TNF-α⁺ CD4⁺/CD8⁺ T cells in response to BA.2.86 compared to XBB.1.5/EG.5.1 (Figure 4(E and G), Supplementary Table 9). We found that the proportion of IFN-γ⁺ CD4⁺ T cells was higher when stimulated using XBB.1.5 (median, 0.95%; IQR, 0.65%–1.40%; P = 0.0469) and EG.5.1 (median, 1.08%; IQR, 0.88%–1.23%; P = 0.0234) S1 peptide pools than those spanning BA.2.86 (median, 0.78%; IQR, 0.61%–1.06%) (Figure 4(E)). The proportion of TNF-α⁺ CD4⁺ T cells induced by XBB.1.5 (median, 1.65%; IQR, 1.22%–3.41%) and EG.5.1 (median, 2.01%; IQR, 1.36%–3.16%) was also higher (albeit not significantly) than that activated by BA.2.86 (median, 1.51%; IQR, 1.28%–3.40%). Similarly, the proportion of TNF-α⁺ CD8⁺ T cells responding to the S1 peptide pools of XBB.1.5 (median, 0.94%; IQR, 0.63%–1.01%) and EG.5.1 (median, 0.93%; IQR, 0.80%–1.15%; P = 0.0117) was higher than that responding to BA.2.86 (median, 0.79%; IQR, 0.59%–1.10%) (Figure 4(F and G)). Though not all comparisons reached significance due to limited sample size (n = 9), ICS results support the ELISpot trends. In summary, evaluation of XBB convalescents revealed that Serotype VI (exemplified by BA.2.86) was more effective at evading T cell immunity than Serotype V (represented by XBB variant series). Subgroup analysis (Supplementary Figure 2) confirmed these patterns were consistent regardless of infection history (reinfection vs primary infection). Although the primary infection group did not reach statistical significance due to sample size limitations, the median T cell response also showed a trend of BA.2.86/JN.1 being lower than that of the XBB series subvariants.

Lower S1-specific T cell responses against subsequent Serotypes II–IV Omicron variants in SARS-CoV-2 Serotype I (prototype) convalescents

It was more difficult to find Serotype II convalescents (represented by BA.1) than BA.5.2/BF.7 and XBB convalescents in China. Thus, we used stored PBMC samples from prototype convalescents to evaluate S1-specific T cell immune responses against Serotypes I (prototype), II (BA.1), III (BA.2), and IV (BA.5.2) using ELISpot assays. The results showed that the T cell immune response to the prototype (median, 3430 SFCs/10⁶ PBMCs; IQR, 2095–5028) was significantly higher than that to BA.1 (median, 3190 SFCs/10⁶ PBMCs; IQR, 1790–4700; P < 0.0001), BA.2 (median, 3460 SFCs/10⁶ PBMCs; IQR, 1840–4573; P = 0.0281), and BA.5.2 (median, 2945 SFCs/10⁶ PBMCs; IQR, 1890–4128; P < 0.0001) (Figure 5(A)). The averaged and log2-transformed data showed that the T cell immune response against Serotype I (mean, 11.63 Log₂ SFCs/10⁶ PBMCs; 95% CI, 11.39–11.86) was higher than that against Serotypes II (mean, 11.49 Log₂ SFCs/10⁶ PBMCs; 95% CI, 11.23–11.75; P = 0.0019), III (mean, 11.52 Log₂ SFCs/10⁶ PBMCs; 95% CI, 11.28–11.77; P = 0.0091), and IV (mean, 11.41 Log₂ SFCs/10⁶ PBMCs; 95% CI, 11.17–11.64; P < 0.0001) (Figure 5(B)). Pairwise reduction curves showed a decreasing trend on transitioning from Serotype I to II (P < 0.0001) in 34/48 (70.8%) samples, from Serotype I to III (P = 0.0281) in 30/48 (62.5%) samples, and from Serotype I to IV (P < 0.0001) in 38/48 (79.2%) samples (Figure 5(C)). The baseline data for S1–specific T cell responses against prototype (Serotype I) and BA.5.2 (Serotype IV) peptide pools were previously reported in the supplementary materials of our earlier study [35]. Here, we have integrated these with additional data for BA.1 (Serotype II) and BA.2 (Serotype III) peptide pools to comprehensively characterize T cell immune responses in prototype convalescents against four serotypes (I–IV). These results indicate that although the prototype convalescents could generate S1-specific T cell immune responses against Omicron variants, they exhibited varying degrees of immune escape.

Figure 5.

S1/S2-specific cross-T cell responses to different serotypes in prototype convalescents. (A) After culturing the PBMCs with the S1 peptide pool of the prototype, T cell responses to the prototype, BA.1, BA.2, and BA.5.2 were detected by ELISpot. S1 peptide pool ELISpot assay results for each SARS-CoV-2 variant. Each colour denotes a different serotype. In the whisker box plots, the top and bottom of the boxes indicate the interquartile range, and the horizontal line within each box indicates the median. “*” in the figure indicates that the T cell response to a given strain was statistically different from that of prototype. (B) A line graph showing averaged and log2-transformed T cell response data. Statistically significant differences are shown at the top of the figure. The error bars indicate 95% confidence intervals. (C) S1 peptide pool ELISpot comparisons of the T cell response to Serotypes I to II (left); Serotypes I to III (middle); and Serotypes I to IV (right). Lines with decreasing slope (coloured) indicate individuals showing weaker responses to Serotype II, III, or IV compared to Serotype I. Lines with increasing slope (gray) indicate stronger responses to the compared serotype. The proportion of individuals with decreased responses and P values are shown. (D) After culturing the PBMCs with the S2 peptide pool of the prototype, S2-specific T cell responses to the prototype, BA.1, BA.2, BA.5.2, XBB, and BA.2.12.1 were detected by ELISpot. The three variants BA.2, BA.5.2, and XBB belong to Serotypes III, IV, and V, respectively, but their S2 protein sequences are completely conserved. There is no significant difference in the T cell immune response of all tested strains. In the whisker box plots, the top and bottom of the boxes indicate the interquartile range, and the horizontal line within each box indicates the median. The Wilcoxon matched-pairs signed rank test was used to compare differences between the T cell responses elicited by the different SARS-CoV-2 variants. A paired-sample t-test was used to analyze the log2-transformed mean value of each serotype.

In contrast to the serotype-dependent attenuation of S1-specific T cell responses, we also detected S2-specific T cell immune responses against Serotypes I (prototype), II (BA.1), and III-V (BA.2, BA.5.2, and XBB, with the same S2 protein sequence). The results showed that there was no significant difference in the T cell immune responses among all tested strains (Figure 5(D)). The limited mutations on the S2 protein did not have a significant effect on the cross-T cell immune responses, which indicates the step-down of cross-T cell immunity with emerging SARS-CoV-2 serotypes is mainly driven by S1, and conserved regions like S2 may contribute to population-level cross-protective immunity.

Similar S1-specific T cell responses reignited by BF.7 and BA.5.2 peptides in Serotype IV convalescents

To validate the R346T mutation’s minimal impact, we compared S1-specific T cell responses using separate BF.7 and BA.5.2 S1 peptide pools. After the in vitro culture with BF.7-S1 peptide pool, the ELISpot assay results indicated that, regardless of whether S1 or RBD peptide pools were used, the T cell responses against Serotype IV were higher than those against Serotype V (Figure 6(A and B)). The log2-transformed data showed that, when cultured with the BF.7-S1 peptide pool, the T cell response against Serotype IV (mean, 11.36 Log₂ SFCs/10⁶ PBMCs; 95% CI, 11.20–11.51) in the S1 peptide pool experiment was higher than that against Serotype V (mean, 11.30 Log₂ SFCs/10⁶ PBMCs; 95% CI, 11.14–11.46; P = 0.0186) (Figure 6(A)). The T cell response against Serotype IV (mean, 10.86 Log₂ SFCs/10⁶ PBMCs; 95% CI, 10.72–11.02) in the RBD peptide pool experiment was similarly higher than Serotype V (mean, 10.66 Log₂ SFCs/10⁶ PBMCs; 95% CI, 10.46–10.86; P < 0.0001) (Figure 6(B)). The ELISpot detection results of PBMCs cultured with BA.5.2-S1 peptide pool were also consistent (Figure 6(C and D)).

We next compared the average T cell response results for each Serotype IV variant and calculated the percentage decrease in T cell activation on transitioning from Serotype IV to V. Among the 65 samples cultured with BF.7-S1 peptide pool, 38 (58.5%) had a reduced T cell response on transitioning from Serotype IV to V (P = 0.0188) in the S1 peptide pool experiment; this number was 49 (75.4%) in the RBD peptide pool experiment (P < 0.0001) (Figure 6(A and B)). For the 62 samples cultured with the BA.5.2-S1 peptide pool, 43 (69.4%) had a reduced T cell response on transitioning from Serotype IV to V (P = 0.0007) in the S1 peptide pool experiment; this result was 51 (82.3%) in the RBD peptide pool experiment (P < 0.0001) (Figure 6(C and D)). These results confirmed that BA.5.2/BF.7 convalescents can generate partial cross-reactive T cell responses against XBB, but with reduced efficacy against Serotype V.

Mutations in key amino acids in Serotype VI cause CD4⁺ (Th1 and Th2) and CD8⁺ T cell escape

Serotype VI (represented by BA.2.86/JN.1) had the most mutation sites compared to other serotypes within the SARS-CoV-2 Omicron variants, and may contribute most T cell immune escape. To explore the T cell escape features of BA.2.86/JN.1, we used 43 overlapping long peptides spanning all the serotype-signature mutation sites of the BA.2.86/JN.1 S protein. In the recent study, we screened these 43 long peptides by ELISpot using PBMCs from the SARS-CoV-2 convalescents, and identified 22 T cell-reactive long peptides [35]. To directly verify whether Serotype VI characteristic mutations impair T cell recognition, we selected 22 pairs of paired peptides for functional validation (Supplementary Table 3). Each pair of peptides showed differences in only one or a few amino acids at the Serotype VI characteristic mutation site (e.g. L455S, F157S, A264D, etc.). Here, we used the ICS assay to measure the secretion of five cytokines, including the Th1 cytokines (IFN-γ, TNF-α, IL-2) and Th2 cytokines (IL-4, IL-6) by CD4⁺ and CD8⁺ T cells in response to corresponding 22 paired peptides from prototype and BA.2.86/JN.1. Among the tested functional markers, SARS-CoV-2-specific T cells mainly produced IFN-γ and TNF-α.

For CD4⁺ T cell response, 12 peptides induced IFN-γ production in at least one individual, among which 10 peptides induced lower levels of IFN-γ production when originating from BA.2.86/JN.1 than from the prototype (Figure 7). 11 peptides induced TNF-α production in at least one individual, among which six peptides induced lower levels of TNF-α production when originating from BA.2.86/JN.1 than from the prototype. Among the peptides inducing both IFN-γ and TNF-α production by CD4⁺ T cells, five peptides (S_45–62, S_143–160, S_260–277, S_319–335, and S_454–468) showed escape characteristics for both cytokines. Specifically, for each of these five peptides, more than 50% of individuals who mounted a positive response to the prototype peptide exhibited reduced cytokine production when stimulated with the corresponding BA.2.86/JN.1 mutant peptide. The associated mutation sites were S50L, F157S, A264D, I332V, and L455S (Supplementary Table 3). We also observed a decrease in IL-2, IL-4, and IL-6 production on transitioning from the prototype to BA.2.86/JN.1 in response to some peptides. The S_260–277 peptide showed a significant decrease in the production of all five cytokines, especially IL-4 and IL-6, which may indicate Th2 response evasion [36].

Figure 7.

Cytokine secretion of CD4⁺ and CD8⁺ T cells stimulated by BA.2.86/JN.1-featured mutation sites. 43 overlapping long peptides were designed to include the serotype-signature mutation sites found in the S protein of BA.2.86/JN.1. Our previous studies used ELISpot screening assays to select 22 positive candidates among these 43 long peptides [35]. 18 of the 22 ELISpot-positive peptides also produced positive ICS results. The bar graphs presented show the results of the ICS assays performed with two sets (prototype and BA.2.86/JN.1) of the 18 positive long peptides; five cytokines (IFN-γ, TNF-α, IL-2, IL-4, and IL-6) secreted by CD4⁺ (left) and CD8⁺ (right) T cells were detected. Result positivity was adjudicated according to the criteria described in the relevant methods section. A peptide was defined as a positive long peptide if it produced a positive response in at least one single donor. If > 50% of responding individuals showed reduced cytokine production to a peptide from BA.2.86/JN.1 than to a peptide from the prototype, that BA.2.86/JN.1 peptide was defined as an escape peptide. The black bars in each figure represent the response to the prototype, the red bars represent the response to BA.2.86, and the blue bars represent the response to JN.1. Peptide S_228–245 only elicited a CD4⁺ T cell response and peptide S_345–362 only elicited a CD8⁺ T cell response. Data were statistically analyzed using paired Wilcoxon signed-rank tests. Although individual comparisons did not reach statistical significance due to limited sample size, the overall pattern supports the escape trend.

For the CD8⁺ T cells, 11 peptides induced IFN-γ production in at least one individual, among which 10 induced lower levels of IFN-γ production when originating from BA.2.86/JN.1 than from the prototype (Figure 7). Nine peptides induced TNF-α production in at least one individual, among which seven induced lower levels of TNF-α production when originating from BA.2.86/JN.1 than from the prototype. Eight peptides induced IFN-γ and TNF-α production by CD8⁺ T cells, among which five (S_143–160, S_150–167, S_157–171, S_319–335, and S_448–464) were “decreased” for both cytokines; the corresponding mutation sites in the S protein were F157S, I332V, V445H, N450D, L452W, and L455S. The three peptides S_143–160, S_150–167, and S_157–171 also caused a decrease in IL-2 and IL-6 production by CD8⁺ T cells. These results indicate that most of the serotype-signature mutation sites present in the S protein of the Serotype VI variants induce both CD4⁺ and CD8⁺ T cell escape. In addition, peptides S_228–245 and S_345–362 only elicited a CD4⁺ or CD8⁺ T cell response, respectively.

Discussion

Our data provided the first preliminary exploration of the six classified SARS-CoV-2 serotypes from the perspective of T cell immunity, revealing how serotype-signature viral mutations can potentially lead to T cell immune escape. Based on the current recognized dominant antibody binding regions and T cell epitopes, we can map the panorama of the relationship between characteristic mutation sites on the S protein and immune escape (Figure 8). These key sites are mainly concentrated in the S1 region, forming multiple hotspots in the antibody binding region i.e. NTD and RBD [37,38] (Figure 8(A)), and overlapping with numerous superimposed T cell epitope hotspots (Figure 8(B)). Our results show that as new SARS-CoV-2 serotypes emerge, cross-reactive T cell immune responses against the spike gradually decline, particularly responses against the most active mutation regions.

Figure 8.

The distribution of SARS-CoV-2 serotype-signature mutation sites in antibody binding regions and T cell epitope hotspots. (A) Mapping of mutation sites on the SARS-CoV-2 S protein to antibody binding regions. The upper panel shows five types of antibody binding modes (NTD-1 to NTD-5) in the NTD region. The lower panel shows eight types of antibody binding modes (RBD-1 to RBD-8) in the RBD region, which are divided into four functional areas: receptor binding motif, outer surface, inner surface, and side^36,37. Colour-coded mutations represent serotype-signature mutation sites (green for Serotype II, orange for Serotype III, blue for Serotype IV, brown for Serotype V, red for Serotype VI, and purple for a unique mutation L455S in JN.1, corresponding to the Figure 8B). Black marks represent cumulative mutation sites shared among multiple serotypes. The numbers 2–6 in brackets represent Serotypes II-VI. Among them, [2] S371L/F, [4] F486 V/S/P, [5] V213E/G, and [6] L452W/R/Q indicate that there are both serotype-signature mutation sites and cumulative mutation sites at this position. Structural visualization and mutation mapping were performed using PyMOL software (version 2.5.0). PDB ID used: 7xix. Mutation sites were mapped onto the spike trimer structure with colour-coding as described in the main text. (B) Analysis of the association between mutation sites and T cell epitopes across the full-length S protein sequences of six serotypes (I to VI), with the NTD (amino acids 13–305) and RBD (amino acids 319–541) regions marked. Red bubbles represent MHC class I T cell epitopes, blue bubbles represent MHC class II T cell epitopes, and the size of the bubbles reflects the number of epitopes affected by the mutation site. Different colours represent different serotype-signature mutation sites. The figure shows how the serotype-signature mutation sites potentially affect T cell immune recognition, especially those sites with larger bubbles may be more important for immune escape mechanisms.

Recent studies have shown that T cells maintain essential cross-recognition to BA.2.86 through conserved epitopes [27,28,39], although BA.2.86 has formed a new serotype. While we observed general cross-reactivity between serotypes, our data revealed a progressive decline in spike-specific memory responses, a step-down pattern reflecting the complexity of SARS-CoV-2 T cell escape. This pattern stems from epitope diversity: the S protein contains hundreds of potential T cell epitopes, far exceeding the number of key neutralizing antibody epitopes. Even when serotype-signature mutations disrupt some epitopes, numerous unaffected epitopes maintain cross-recognition [40–42]. This epitope diversity, combined with highly conserved regions (S2/N/M/E/ORF), provides stable compensatory T cell immunity across serotypes (Figure 2, Figure 5(D)) [43]. Unlike antibodies recognizing conformational epitopes, T cells recognize linear peptide-MHC complexes and tolerate single-residue changes more readily. Quantitatively, T cell responses showed 10–20% step-down attenuation across serotypes, while neutralizing antibody escape proved steeper [44–46]. This differential explains why vaccinated individuals maintain protection against severe outcomes despite breakthrough infections with antibody-evasive variants [47]. Most participants in this study received 2–3 doses of the inactivated prototype vaccine, with an interval of more than 6 months between vaccination and infection (Supplementary Table 4). This common baseline immunization from vaccines may have complex effects on T cell cross-reactivity following subsequent natural infection [48,49]. On the one hand, vaccine-induced memory T cells may enhance the recall response to homologous or conserved epitopes through the immune imprinting effect [50,51]. On the other hand, the immune memory bank established by the initial vaccination may provide some cross-protection against subsequent variants [52]. However, the serotype-specific T cell response decay pattern we observed was highly consistent across all participants, suggesting that although vaccines and infections together shape the pool of memory T cells, the differential responses between serotypes primarily reflect the cumulative effect of characteristic mutations in the infected variants [53].

We focused on identifying the specific mutation sites within the S protein of different serotypes and traced the corresponding changes in the T cell epitopes. Different serotypes harboured unique mutations, which involved hotspots of MHC I and MHC II epitopes [54,55]. The continuous variation in T cell epitopes indicates that the emergence of new SARS-CoV-2 variants may be driven by T cell-mediated selection pressure, akin to what is commonly observed in HIV and influenza viral escape mutations [56–58]. Nevertheless, this may also represent a byproduct of antibody escape mechanisms and viral adaptation to cell receptor binding. Phylogenetic analysis revealed complex evolutionary pathways among serotypes while highlighting their common ancestry and unique genetic characteristics. Both “serotype-signature mutation sites” and “cumulative mutation sites” may interfere with T cell and antibody recognition [59–61], clustering in epitope hotspots that overlap HLA binding/TCR docking sites and antibody-targeted regions (Figure 8(A)). Notably, BA.2.86/JN.1 exhibits significant divergence from BA.2 (>30 mutations in the S protein) [62], impacting cross-T cell recognition to varying degrees. Additionally, coherent evolutionary relationships occur among strains, such as BA.1 to BA.2, and BA.2 further into BA.4/BA.5, XBB, and BA.2.86. The gradual evolution and accumulation of the mutations may relate to the formation of step-down cross-T cell responses with emerging serotypes. Key sites (such as S371L/F, L452W/R/Q, F486V/S/P) serve as both serotype-signature and cumulative mutations among multiple serotypes, indicating the importance of these sites in the viral evolution process. Our ICS assay data showed mutations, such as A264D, R403K, and L455S, disrupted Th2 cytokine secretion and localized to the antibody epitope [63,64]. Therefore, we speculate that the gradual progressive T cell immune escape may indirectly exacerbate antibody-mediated protection decline.

Conserved epitopes across serotypes provide an empirical basis for the development of T cell-oriented universal vaccines by serving as targets for cross-variant protection. Simultaneously, serotype-specific components can provide targeted peak immune responses by enhancing rapid immunization renewal, thereby balancing the breadth and intensity of immunity. As of mid-2025, the dominant strains globally remain the JN.1 sub-lineage (KP.2, KP.3, XEC, etc.), which have accumulated 4–8 spike protein mutations based on JN.1. Serological studies confirm that these variants still belong to Serotype VI [19,65], consistent with our mutation mapping (Figure 1(C)), which shows that newly emerged substitutions cluster within or adjacent to established Serotype VI signature sites, rather than defining novel antigenic profiles. The serotype-associated T cell immune escape pattern revealed in this study underscores the necessity of continuous monitoring. Future surveillance should consider incorporating T cell response analysis into the standard procedures for assessing emerging variants, rather than focusing solely on neutralizing antibody data, to provide a more comprehensive immunological basis for vaccine updates and variant surveillance.

Our study has several limitations. First, sample sizes were limited for certain analyses, particularly ICS assays, and some comparisons should be interpreted as trends rather than definitive conclusions. However, the combined evidence from ELISpot, ICS, single peptide analysis, and multi-serotype cohorts enhances the reliability of our core findings. Larger prospective cohorts with individual-level longitudinal tracking are needed for validation. Second, S1 peptide pool analyses detected no HLA associations (Supplementary Figure 1), likely because pooled epitopes dilute single-allele effects. Future studies require larger samples and mechanistic validation (e.g. peptide-MHC tetramers, structural analyses) to refine epitope-HLA-TCR relationships. Third, most participants had vaccination histories, and we lacked unvaccinated convalescent comparisons to fully disentangle vaccine-infection interactions. Future studies should include vaccine-naive individuals and even animal experiments. Despite these limitations, our multi-method approach robustly documents serotype-associated T cell escape patterns.

SARS-CoV-2 continues to evolve and evade human immune responses. Our study on the viral serotype-related T cell immunity evasion profile of SARS-CoV-2 provides new insight into the dynamic nature of viral evolution and the ongoing selective pressure exerted by host immunity. These findings highlight the pivotal role of the population T cell immune barrier in containing emerging variants and provide critical references into T cell-oriented universal vaccine development for coronaviruses, emphasizing the need for immunogen designs that preserve broadly conserved epitopes across serotypes.

Author contributions

JL and GFG proposed and designed the study. YYG and JMT collected the samples. YYG, JMT, PPG, XW, and MJY conducted the experiments. BLS and JYS provided technical support and experimental assistance. YYG and PPG analyzed and interpreted data. YYG, JMT, PPG, and JL wrote the initial draft of the manuscript. All authors contributed intellectually and approved the manuscript.

Funding Statement

This work was supported by the National Key Research and Development Program of China (2022YFC2604100, 2023YFC3041500), the National Natural Science Foundation of China (92269203), and Major Project of Guangzhou National Laboratory (GZNL2025C01001).

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplemental Material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/22221751.2025.2610856.