Coordinated adaptive immunity is essential for protection and clearance of viral infections, including SARS-CoV-2 (ref. 1). Virus-specific neutralizing antibodies are considered the main correlate of protection against SARS-CoV-2 infection, but wane over time2,3. T cell responses are more durable2,4,5 and increasing evidence supports their role in restricting SARS-CoV-2 infection and limiting the severity of COVID-196,7. Worldwide efforts have rapidly delivered SARS-CoV-2 vaccines, mostly designed against the spike (S) protein, which mediates host cell entry. Studies enumerating the T cell response to whole SARS-CoV-2 S protein using pools of overlapping peptides covering the entire protein sequence (peptide mixes) show that memory T cell responses to S protein in previously infected or vaccinated individuals are dominated by CD4+ T cells4,8,9,10.

SARS-CoV-2 CD4+ T cell epitopes have been identified, but mostly in assays that use high concentrations of stimulating peptides. In addition, their human leukocyte antigen (HLA) restriction has largely been inferred from in silico HLA-binding algorithms11,12. Detailed knowledge of the specificity of CD4+ T cell responses at the epitope level and their HLAII restriction is therefore currently lacking13. Furthermore, whether the CD4+ T cell epitopes are generated naturally through the HLA class II (HLAII) antigen processing pathway is currently unknown. Because of these limitations, the extent to which CD4+ T cells that recognize SARS-CoV-2 epitopes cross-react with other human β-coronaviruses (β-HCoVs) remains unknown. Low frequency CD4+ T cell reactivity to SARS-CoV-2 peptide mixes has been reported in some uninfected individuals14,15,16,17,18 and has been suggested to originate from previous exposure to other HCoVs (HKU1, OC43, NL63, 229E, SARS or MERS). This raises the possibility that pre-existing HCoV immunity could potentially contribute to controlling SARS-CoV-2 infection.

The extent to which T cells induced by ancestral SARS-CoV-2 proteins can protect against VOCs is a critical question. In particular, the highly transmissible omicron VOC contains several mutations within the receptor binding domain (RBD) of S protein—the main target of neutralizing antibodies19,20. These mutations reduce neutralization by the S-specific antibodies induced by ancestral SARS-CoV-2 variants or by the initial vaccines deployed in the pandemic, which were designed against the original S protein reference sequence21,22,23. Whilst neutralizing antibody titers can be partially restored by booster vaccination24,25, continued virus evolution has caused a high prevalence of secondary and vaccine breakthrough infections26. Ex vivo studies of previously infected or vaccinated individuals using peptide mixes have shown minimal reduction in the overall frequency of CD4+ T cell responses against the alpha (B.1.1.7), beta (B1.351), gamma (P1) and delta (B.1.717.2)23,27,28,29 or omicron (B.1.1.529) VOCs22,30,31. However, biologically relevant differences in epitope-specific recognition efficiency may have been missed32 and little information exists to understand the extent of CD4+ T cell epitopes evaded by current VOCs or to predict CD4+ T cell epitope loss in future SARS-CoV-2 variants.

Here, we performed a detailed analysis of CD4+ T cell immunity against SARS-CoV-2 in healthcare workers (HCW) infected in the first wave of the pandemic. We examined 159 CD4+ T cell clones and identified and characterized 21 HLAII-restricted T cell epitopes. Responses to most epitopes located in the S protein were also present in vaccinated individuals of appropriate HLAII genotype. Cross-reactivity of SARS-CoV-2 S-specific T cell clones was observed only against the closely related S2 region of the SARS virus, with no cross-reactivity observed for any other β-HCoV. Mutations were present in 10 of the 17 S protein epitopes within one or more SARS-CoV-2 VOC. Minor amino acid (aa) changes in seven epitope sequences, including those within the RBD region of omicron, were sufficient to reduce or evade recognition by S-specific CD4+ T cells. However, the breadth of responses to several CD4+ T cell epitopes seen in each individual suggested that current VOC mutations confer only limited evasion from CD4+ T cell surveillance.


SARS-CoV-2 infection induces broad CD4+ T cell immunity

Peripheral blood mononuclear cells (PBMCs) were collected from June to September 2020 from 20 HCW 3–6 months postinfection (PI) during the first wave of SARS-CoV-2 wild-type (WT, D614G) infection in the United Kingdom and 14 uninfected healthy (UH) volunteers. HCW-PI had detectable antibodies to S protein and nucleoprotein (N) at this timepoint, while UH volunteers had no detectable S- or N-specific antibody responses (Extended Data Fig. 1). To characterize T cell immunity against whole antigens, PBMCs and CD8-depleted PBMCs (hereafter CD8PBMCs) were tested against SARS-CoV-2 peptide mixes comprising 15mer peptides overlapping by 11 aa and spanning the entire open reading frame of S, membrane (M) and N proteins in ex vivo interferon-γ (IFN-γ) Elispot assays. Compared with control whole PBMCs, as expected, the response to a peptide mix of HLA class I (HLAI)-restricted epitopes was significantly lower in CD8PBMCs (Fig. 1a,b). In contrast, responses to S protein (tested as two pools, S1 and S2), M and N peptide mixes were increased in CD8PBMCs (Fig. 1a,b), confirming CD4+ T cell memory responses were predominant4,8. As previously reported14,15,16,17,18, we detected weak responses to individual SARS-CoV-2 protein pepmixes in CD8PBMCs from 8 of 14 UH volunteers (Fig. 1c). The magnitude of responses in UH volunteers was significantly lower than in HCW-PI (P < 0.01, two-sided Mann–Whitney U-test) (Fig. 1c) and may represent cross-reactive T cells primed by previous exposure to β-HCoVs.

Fig. 1: CD4+ T cell response to SARS-CoV-2.
figure 1

a, Representative Elispot assays for the production on IFN-γ in whole and CD8PBMCs isolated from HCW-PI, plated at 2 × 105 cells per well and incubated with mixes of 15 aa peptides (overlapping by 11 aa) from SARS-CoV-2 S1, S2, M and N proteins, control (Ctrl) HLAI and HLAII epitope mixes and DMSO solvent (Neg). b, Summary of Elispot assays for the production of IFN-γ in whole versus CD8PBMCs using n = 10 HCW-PI incubated with peptide mixes as in a. Results are shown as mean SFC per 106 PBMCs. Significance was determined by two-sided Wilcoxon test, *P < 0.05. c, Summary of Elispot assays for the production of IFN-γ in CD8PBMCs from HCW-PI (n = 20) and UH (n = 14) individuals incubated with peptide mixes as in a. Results are shown as mean SFC per 106 CD8PBMCs. Significance was determined by two-sided Mann–Whitney U-test, **P < 0.01. d, Summary of ELISAs measuring IFN-γ production by polyclonal CD4+ T cell lines generated by initial stimulation of CD8PBMCs with peptide mixes as in a, then stimulated 7–14 days later with individual 20mer peptides (overlapping by 10 aa) spanning the relevant SARS-CoV-2 protein. Individual rows show the response from HCW-PI, n = 11. NS, not significant.

Source data

We next examined the CD4+ T cell response to S, M and N proteins following SARS-CoV-2 infection at the epitope level. Polyclonal CD4+ T cell lines were initially established from HCW-PI by stimulating PBMCs with S, M and N peptide mixes, ensuring complete coverage of the proteins. These lines were then tested with individual 20mer peptides overlapping by 10 aa, or dimethylsulfoxide (DMSO) solvent as negative control, to determine the regions of reactivity against each protein. CD4+ T cells can be expanded using shorter 15mer peptides; however, the optimal response is often to a longer peptide33,34. For each individual, the polyclonal CD4+ T cell lines contained several responses against peptides distributed throughout the S protein, including the RBD, S1 and S2 regions, M and N (Fig. 1d). As noted by others, these data show that SARS-CoV-2 induced a broad CD4+ T cell response11,14.

HLAII type determines SARS-CoV-2 CD4+ T cell epitope responses

We next performed limiting dilution cloning from five individuals with different HLAII types and isolated CD4+ T cell clones specific for 21 epitopes (17 S protein, 2 M protein and 2 N protein epitopes). To identify the optimal peptide recognized by each T cell clone, we titrated four individual 20mer peptides (overlapping by 15 aa) covering the regions where T cell reactivity was detected in the screening assays. This defined peptides SSAN (aa 161–180) and FNCY (aa 486–505) as the epitopes for the CD4+ T cell clones c3 and c42, respectively (Fig. 2a). Peptides RGHL and RNSS were defined using the same approach in M and N respectively (Extended Data Fig. 2). The avidities of the S, M and N specific T cell clones were comparable with CD4+ T cell clones against other viruses previously measured by peptide titration (Fig. 2 and Extended Data Fig. 2)35,36,37.

Fig. 2: Characterization of new spike CD4+ T cell epitopes.
figure 2

a, ELISA assays for the production of IFN-γ from CD4+ T cell clones cocultured in overnight assays with autologous LCL loaded with individual 20mer peptides overlapping by 15 aa (10–5 to 10–11M). bd, ELISA assays for the production of IFN-γ from CD4+ T cell clones cocultured in overnight assays with autologous LCL prepulsed with epitope peptide or DMSO solvent and either tested alone (no antibody; No Ab), or in the presence of blocking antibodies against HLA-DP (αDP), HLA-DQ (αDQ) or HLA-DR (αDR) (b), autologous LCL and allogeneic LCLs with HLAII types partially matched to the autologous LCL, either prepulsed with 5 μM 20mer epitope peptide or DMSO solvent (Neg) (c) or autologous LCL either prepulsed with epitope DMSO solvent (Neg), peptide or 1 ng ml–1 S tetrameric protein (d). (ad) Results show mean IFNγ release ±1 s.d. and are representative of three experiments.

Source data

Next, to identify the HLAII allele restricting each epitope, we tested each clone against a peptide-loaded autologous lymphoblastoid cell line (LCL) in the presence of blocking antibodies against HLA-DR, HLA-DP and HLA-DQ, and after that against peptide-loaded allogeneic LCLs with partially matched HLA-DR, -DP or -DQ types. This approach indicated that HLA-DPB1*04:01 restricted the presentation of peptide SSAN to clone c3, while HLA-DRB1*01:01 presented peptide FNCY to clone c42 (Fig. 2b,c). We then defined the optimal peptide and presenting HLAII allele for all 21 S, M and N epitopes (Table 1), using the clones isolated against each epitope (Supplementary Table 1). Individual T cell clones specific for the same epitope recognized the same optimal peptide and HLAII combination (Extended Data Fig. 3). The T cell clones specific for all 17 S protein epitopes recognized autologous LCL pre-exposed to low concentrations of purified S protein (1 ng ml–1) (Fig. 2d and Extended Data Fig. 4), indicating that all S protein epitopes were efficiently generated through the exogenous HLAII processing pathway. All 21 S, M and N epitopes were presented by HLA-DR or HLA-DP alleles, with no HLA-DQ-restricted T cells found (Table 1).

Table 1 SARS-CoV-2 CD4+ T cell epitopes and responses

To address whether natural infection and vaccination elicited similar CD4+ T cell immunity, we repeated the ex vivo Elispots assays for the production of IFN-γ by CD8PBMCs from each HCW-PI using the defined epitope peptides appropriate to each individual’s HLAII type. Of the 11 epitopes presented by HLAII alleles present in several donors, responses to 9 were present in more than one donor (Table 1). Next, we examined the relative immunogenicity of SARS-CoV-2 vaccination by testing blood samples collected 1–5 months postvaccine from 9 of the 14 donors originally used as UH. All nine donors had S-specific antibodies, but undetectable N-specific antibodies (Extended Data Fig. 5), indicating they had responded to vaccination and had no history of natural infection. Based on the HLAII genotypes of these donors, we tested 14S epitopes and detected responses to 13 in ex vivo IFNγ Elispot assays (Table 1). Collectively, these results indicated that SARS-CoV-2 infection and vaccination induced broad CD4+ T cell responses to shared epitopes and that, in both contexts, HLAII genotype was a key determinant of the SARS-CoV-2 S-specific CD4+ T cell response.

Spike-specific CD4+ T cell clones do not cross-recognize β-HCoVs

Next we asked whether spike-specific CD4+ T cells elicited by SARS-COV-2 infection cross-reacted with closely related β-HCoVs known to infect humans (SARS, MERS, HKU1 and OC43), in which the S proteins share 34–76% aa similarity to SARS-CoV-2 S protein (Fig. 3a). Within the N-terminal S1 region, the highest similarity (64.7%) is between SARS and SARS-CoV-2, whilst the similarity of all other β-HCoV S1 regions with SARS-CoV-2 S1 is low (<32%) (Fig. 3b). The C-terminal S2 regions exhibit greater overall similarity, with SARS-CoV-2 and SARS having 90.0% similarity, and the similarity between SARS-CoV-2 and the other β-HCoV is up to 45%38 (Fig. 3b).

Fig. 3: Cross-recognition of SARS-CoV-2 spike-specific CD4+ T cell clones with β-HCoV.
figure 3

a, Dendrogram showing the evolutionary relationship and percentage aa similarity of SARS-CoV-2 S protein with S proteins from all known β-HCoVs. b, Correlogram of the percentage aa similarities between the S1 and S2 regions of the β-HCoVs. c,d, ELISA assays for the production of IFN-γ from CD4+ T cell clones cocultured with autologous LCL pre-exposed to DMSO solvent (Neg) or epitope peptide or S1 (c), or S2 (d) peptide mixes from β-HCoVs (15 aa overlapping by 11 aa) in overnight assays. Results show mean IFN-γ release ±1 s.d. and are representative of three independent experiments. Alignments show the aa sequences of SARS-CoV-2 S CD4+ T cell epitopes and the corresponding β-HCoV sequences. Amino acids highlighted in yellow are conserved with SARS-CoV-2.

Source data

CD4+ T cell clones specific for six SARS-CoV-2 S epitopes were tested against peptide mixes comprising 15mer peptides overlapping by 11 aa from each β-HCoV or DMSO solvent as a negative control. As expected, all six S protein-specific clones showed similar recognition against S1 or S2 SARS-CoV-2 peptide mix and the respective cognate epitope peptide of that clone (Fig. 3c,d). The SSAN c3, VVLS c21 and FNCY c42 T cell clones specific for epitopes located within the more divergent S1 region of SARS-CoV-2 did not cross-recognize any other β-HCoV peptide mix (Fig. 3c). These SARS-CoV-2 20mer epitopes had between 2 and 10 aa differences compared with the corresponding regions of SARS (Fig. 3c)—the virus with greatest overall sequence similarity (Fig. 3a); the corresponding epitope sequences within the other β-HCoVs were even more divergent (Fig. 3c). Of the CD4+ T cell clones specific for three epitopes within the S2 region of SARS-CoV-2, which has greater aa similarity with other β-HCoVs (Fig. 3b), NFSQ c117 did not cross-recognize any other β-HCoV (Fig. 3d). STEC c41 and SFIE c55 both cross-recognized the SARS peptide mix, but not the MERS, HKU1 or OC43 peptide mixes (Fig. 3d); both epitopes had only a single aa difference between SARS-CoV-2 and SARS (Fig. 3d). We extended the work to an additional 11 clones: 5 specific for epitopes within S1 and 6 within S2. The epitopes within the S1 region of SARS-CoV-2 differed from SARS by 5–16 aa, whereas the epitopes within the S2 region differed by 0–2 aa (Extended Data Fig. 6). None of the five S1-specific clones, but all six S2-specific clones, cross-recognized the SARS peptide mix (Table 2); the peptide mixes from the other β-HCoVs, with lower aa sequence similarity to SARS-CoV-2, were never recognized (Table 2). These data indicated that S epitope-specific CD4+ T cells isolated following SARS-CoV-2 infection could recognize highly homologous epitopes within the S2 region of SARS, but did not cross-react with MERS or the extant β-HCoVs HKU1 and OC43, consistent with the S protein-specific CD4+ T cell clones described here being primed by SARS-CoV-2 infection.

Table 2 CD4+ T cell clone recognition of β-HCoV peptide mixes

Mutations in variants of concern impair CD4+ T cell recognition

Next, we examined the recognition of previous and current SARS-CoV-2 VOCs, including alpha (B.1.1.7), beta (B.1.351), gamma (P.1), delta (B.1.617.2), zeta (P.2), theta (P.3) and omicron BA.1 and BA.2 (B.1.1.529), by S protein-specific CD4+ T cell clones. Amino acid substitutions or deletions were present in 10 of the 17 S epitope sequences within VOCs (Table 3), with some common to several VOCs, such as the N501Y mutation in the FNCY epitope identified in alpha, beta and gamma variants (Table 3), and others unique to particular viral isolates, such as the N764K mutation within the STEC epitope in omicron (Table 3). To study the impact of these mutations on CD4+ T cell recognition, we tested the S-specific CD4+ T cell clones from the HCW-PI donors, who had been infected during the WT (D614G) wave of SARS-CoV-2, against the optimal WT epitope peptide and the corresponding mutated peptides from the VOCs; 20mer peptides were employed to encompass the peptide flanking regions, as mutations within the MHCII-binding core and proximal flanking regions can interfere with epitope binding to MHCII (ref. 39). Peptides were tested at concentrations of 10−5 to 10−11M to detect effects of these mutations at low concentrations that may not be evident at higher concentrations. Single central aa substitutions in epitope TLVK (N969K) in omicron and SGTN (D80Y) in zeta eliminated CD4+ T cell recognition, except at supraphysiological peptide concentrations for the latter (Fig. 4). Two separate point mutations in epitope QLIR, A1022S in beta and T1027I in gamma impaired CD4+ T cell recognition at equivalent peptide concentrations (Fig. 4). However, several epitope peptides containing point mutations, such as LSET (T307A) in theta, STEC (N764K) in omicron and TYVT (A1022S) in beta were recognized equally to the WT epitope (Fig. 4). A double aa deletion (Δ43–44) in theta had no effect on the recognition of LVDL and TRFQ epitopes (Fig. 4); however, a triple aa deletion (Δ42–44) in beta reduced CD4+ T cell stimulation compared with the WT peptide (Fig. 4).

Table 3 S epitope sequences in SARS-CoV-2 WT and VOCs
Fig. 4: Impact of mutations in SARS-CoV-2 VOCs on spike-specific CD4+ T cell clone recognition.
figure 4

Representative ELISA assays for the production of IFN-γ from CD4+ T cell clones cocultured with autologous LCLs loaded with purified SARS-CoV-2 WT S peptides or corresponding mutated peptides from VOCs at 10–5 to 10–11 M in overnight assays. Results show mean IFN-γ release ±1 s.d. and are representative of three independent experiments.

The epitopes with most mutations were GGNY and FNCY, both present within the RBD region—a frequent target of mutation in VOCs19,20. Although single point mutations arising in earlier VOCs in epitopes GGNY (L452R in delta and B.1.324) and FNCY (N501Y in alpha, beta, gamma and theta) did not affect recognition, the numerous mutations accumulated in omicron in GGNY (G446S, L452R, R457N) and FNCY (Q493K, G496S, Q498R, N501Y, Y505H) eliminated recognition by the CD4+ T cells specific for these epitopes (Fig. 4). Overall, recognition of 7 of the 17 epitopes in S protein by the CD4+ T cells specific for these epitopes was affected by mutations present in one or more VOCs (Fig. 4 and Table 3). The affected epitopes were restricted through a range of HLAII alleles (Table 1) and, in every case where recognition of an epitope was lost, the same person possessed CD4+ T cell responses against other epitopes that were not impacted by mutation (Fig. 1d).

T cell clones allow only a small number of TCRs to be studied. In vivo, epitope-specific CD4+ T cell responses comprise a multitude of different TCRs40. To test whether other TCRs specific for the same epitopes may be unaffected by VOC mutations, we examined ex vivo memory CD4+ T cell (CD4+ TM cells) populations specific for S protein epitopes in HCW-PI or in UH 1–5 months postvaccination. First, to analyze the clonal TCR composition of the SARS-CoV-2 infection-induced CD4+ TM cells specific for S protein epitopes we used three representative peptide-HLAII (pHLAII) tetramers (Tet) containing epitope sequences from the WT S protein: one epitope that contained no mutations in the current VOCs, SSAN (DPB1*04:01/SSAN), one epitope where mutation in the omicron VOC abrogated T cell clone recognition, GGNY (DPB1*02:01/GGNY) and one where the current VOC mutations did not affect T cell clone recognition, STEC (DRB1*15:01/STEC) (Extended Data Fig. 7a,b). The pHLAII Tet were used in flow cytometry assays alongside a panel of antibodies specific for 21 TCR Vβ segments commonly expressed on CD4+ T cells41. As expected, in every case, the S epitope-specific CD4+ TM cells were polyclonal, with evidence of oligoclonal expansion (Fig. 5a and Extended Data Fig. 7c). Several TCR Vβ segments were over-represented within the pHLAII Tet+ CD4+ TM cells compared with the total CD4+ TM cell repertoire, with smaller frequencies of pHLAII Tet+ CD4+ TM cells expressing other TCR Vβs also detected (Fig. 5a).

Fig. 5: The impact of mutations in SARS-CoV-2 VOCs on ex vivo CD4+ T cell recognition.
figure 5

a, Flow cytometry analysis of total CD4+ T cells and pHLAII tetramer+ (pHLAII Tet+) cells in PBMCs stained with the indicated pHLAII Tet and antibodies against 21 defined TCR Vb segments. b, Elispot assays for the production of IFN-γ in CD8PBMCs plated at 4 × 105 cells per well and incubated with purified SARS-CoV-2 WT S peptides or corresponding mutated peptides from VOCs at 10–6 to 10–8 M. Results are shown as mean SFC ± 1 s.d. per 106 CD8PBMCs.

Second, we performed IFN-γ Elispot assays using CD8PBMCs from HCW-PI or UH 1–5 months postvaccine to test the response of CD4+ TM cells to five epitopes, four of which (SGTN, LVDL, GGNY and QLIR) had shown decreased CD4+ T cell clone recognition against VOC peptides and one, STEC, that had maintained recognition. Compared with the WT peptide, the frequency of CD4+ T cells that produced IFN-γ in response to the zeta SGTN peptide containing the point mutation (D80Y) was lower at each peptide concentration tested (Fig. 5b). Likewise, the recognition of LVDL, GGNY and QLIR similarly showed reduced T cell responses for at least one of the variant peptides tested. Notably, no ex vivo IFN-γ Elispot response was detected to the mutated GGNY omicron peptide (G446S, L452R, R457N) (Fig. 5b) in the same individual that possessed a polyclonal population of pHLAII Tet+ cells specific for the WT peptide (Fig. 5a). In contrast, similar magnitudes of IFN-γ producing CD4+ T cells were detected against the STEC peptide in omicron (N764K) and WT (Fig. 5b). Collectively, these data showed the acute sensitivity of SARS-CoV-2 S-specific CD4+ T cells to small aa changes in their epitope sequence and that our findings using T cell clones were representative of circulating polyclonal epitope-specific CD4+ T cell responses.


Our detailed analysis of HCW previously infected with WT SARS-CoV-2, including the isolation and extensive use of CD4+ T cell clones, provided several new insights into the CD4+ T cell response to SARS-CoV-2. Focusing on the S protein, the lack of cross-reactivity with β-HCoVs indicated that all the SARS-CoV-2 S-specific CD4+ T cell clones originated from the naïve repertoire rather than pre-existing β-HCoV-specific CD4+ TM cells. A key finding was that the CD4+ T cell response in every individual targeted several viral epitopes. This broad response is important because we showed mutations in SARS-CoV-2 VOCs compromise CD4+ T cell recognition of some, but currently not all, S epitopes. The breadth of the CD4+ T cell response therefore limits the impact of mutations in current VOCs on overall CD4+ T cell surveillance.

In line with previous studies on HCW-PI assessing T cell responses to whole antigens ex vivo4,8,9,10, we detected robust CD4+ TM cell responses against peptide mixes from SARS-CoV-2 S, M and N proteins. Our data showed that every donor possessed a broad CD4+ T cell response after SARS-CoV-2 infection that targeted several epitopes. In total, we defined 21 HLAII-restricted epitopes. This considerably expanded the number of experimentally verified CD4+ T cell epitopes and also specified the HLAII restriction of some previously reported epitopes11,12,13,42,43. Some of our epitopes were presented by HLAII alleles found at high frequency in global populations, such as the HLA-DRB1*04 subtypes, DRB1*01:01, DRB1*15:01 and DPB1*04:0144. CD4+ T cell responses to these epitopes are therefore likely to be widespread following infection or vaccination. Our data considerably strengthens the evidence that SARS-CoV-2 epitopes are skewed towards HLA-DP and HLA-DR restriction alleles14,25. This HLAII usage is distinct from other human viruses investigated using CD4+ T cell clones37,45.The reason remains unclear, but it will be important to understand, as our data showed that HLAII genotype was a key determinant of the CD4+ T cell response to S protein following either SARS-CoV-2 infection or vaccination.

Several endemic β-HCoVs infect humans and cause mild disease. T cells elicited by previous infection with these viruses could modulate the course of disease, if they cross-reacted with SARS-CoV-2. All the CD4+ T cell clones we studied were generated using SARS-CoV-2 peptides from donors previously infected with SARS-CoV-2, who also had evidence of historical infection with the extant β-HCoVs HKU1 and OC43. All S protein-specific CD4+ T cell clones efficiently recognized SARS-CoV-2 S protein peptide mix, but not the peptide mixes from OC43 or HKU1 S proteins, which have less than 40% aa similarity to SARS-CoV-2. The only observed cross-reactivity to β-HCoV was against epitopes located within the S2 region of SARS, which has 90% aa similarity with SARS-CoV-2, but to which the UK HCWs studied here had never been exposed. These results strongly suggested that, for our cohort of previously SARS-CoV-2 infected donors, CD4+ T cell clones specific for the S protein originated from naïve CD4+ T cells primed by SARS-CoV-2 infection rather than from CD4+ TM cells primed by previous β-HCoV infections. Consistent with our data, CD8+ TM cells specific for an epitope in the SARS-CoV-2 N protein originated from the naïve CD8+ T cell repertoire in SARS-CoV-2 infected individuals46.

Our data do not contradict previous studies that have reported the presence of pre-existing cross-reactive S-specific CD4+ T cells in unexposed individuals14,15,16,17. Where cross-reactivity has been investigated at the level of epitopes, it was focused on small regions of the S protein that are highly conserved between SARS-CoV-2 and other β-HCoVs13,14,15,18,47 Our data showed that following SARS-CoV-2 infection, the CD4+ T cell response to the S protein was broadly targeted across the entire protein. Therefore, most CD4+ TM cell responses to SARS-CoV-2 S protein targeted epitopes with low sequence similarity to the S proteins of other β-HCoVs and, as expected, did not cross-react.

We only investigated the cross-reactivity of CD4+ T cells specific for the S protein, which is highly targeted by mutation. Other viral proteins are more conserved across β-HCoVs and T cell cross-reactivity may therefore be more probable. Accordingly, pre-existing T cell immunity against the highly conserved N and ORF1ab-encoded nonstructural proteins (NSP) has been reported in SARS-CoV-2-exposed HCWs with no evidence of virus infection15,16,48. The β-HCoV cross-reactive T cells in those donors were frequently directed against epitopes located in the early-expressed replication transcription complex encompassing NSP7, NSP12 and NSP13 48. Ultimately, the extent to which pre-existing cross-reactive T cell immunity contributes to controlling SARS-CoV-2 infection in an individual might be determined by a complex combination of factors, including the conservation of the epitopes presented by their HLA genotype, their TCR repertoire and their history of previous β-HCoV exposure13.

Examining S proteins sequences from VOCs, we identified aa substitutions or deletions in 10 of the 17 S protein CD4+ T cell epitopes; some were common to several VOCs, while others were unique to particular viral isolates. Combining data from the CD4+ T cell clone experiments and the experiments using ex vivo PBMCs, which contain polyclonal epitope-specific populations, we found variable effects of the epitope mutations on CD4+ T cell recognition. Triple aa deletion and several aa substitutions within individual epitopes had the greatest impact on CD4+ T cell recognition49. The effect of point mutations was complex. A single mutation could markedly decrease CD4+ T cell recognition or have no effect. Notably, decreased CD4+ T cell sensitivity to several epitopes was only apparent at lower peptide concentrations and was not evident using the high concentrations of peptides previously employed22,23,29,50. This highlights the requirement for further careful experimental definition of the currently known immunodominant SARS-CoV-2 CD4+ T cell epitopes, the identification of the essential aa required for HLAII binding and TCR engagement and the need to consider each epitope individually. The epitope mapping presented here provides a rational basis for VOC risk stratification. Additional mutations recently acquired in the BA.4 and BA.5 omicron variants, such as Δ69–70 and F486V highlight the continued evolution of SARS-CoV-2 and potential for further T cell epitope mutation.

In conclusion, our study demonstrated the fine sensitivity of SARS-CoV-2 S-specific CD4+ T cells to aa variation in epitope sequence and the potential for SARS-CoV-2 evolution to evade the CD4+ TM response. The breadth of SARS-CoV-2 S epitopes targeted in every individual indicated current VOC mutations are likely to have only limited impact on overall CD4+ T cell surveillance. However, continued evolution of SARS-CoV-2 could lead to further epitope loss and continued monitoring of emerging VOCs is important.


Donor characteristics and ethical statement

The study was approved by the North West–Preston Research Ethics Committee, United Kingdom (20/NW/0240) and all participants gave written informed consent and received no compensation. Blood was collected in June to September from 20 healthcare workers (HCW), ten males and ten females aged 28–64 years, at 3–6 months postinfection during the first wave of SARS-CoV-2 WT (D614G) infection in the United Kingdom. Control samples were collected from SARS-CoV-2 UH individuals (n = 14; 6 males and 8 females aged 24–63 years) confirmed to be seronegative for SARS-CoV-2 (Extended Data Fig. 1), or collected before the pandemic from healthy donors as part of an ethically approved study (South Birmingham Research Ethics Committee 14/WM/1254). Subsequent samples were collected from those UH individuals who remained uninfected following vaccination with Pfizer BioNtech BNT162b2 or AstraZeneca ChAdOx-1 (n = 9; 3 males and 6 females aged 39–63 years) 1–5 months after vaccination.

Sample preparation

Plasma and PBMCs were isolated from heparinized blood using standard Ficoll-Hypaque centrifugation. The resulting PBMC layer was washed twice with RPMI and either used directly or cryopreserved before use. Where stated, PBMCs were depleted of CD8+ T cells (CD8PBMCs) using anti-CD8 Dynabeads (Invitrogen) to over 94%, as measured by flow cytometry. Autologous LCLs were generated by transformation with B95.8 EBV as previously described51 and HLAII typing by next generation sequencing was performed at the Anthony Nolan Histocompatibility Laboratories.

Synthetic peptides and protein

For stimulation of T cell responses against whole protein sequences, peptide mixes containing 15mer peptides overlapping by 11 aa covering the full length of WT (D614G) SARS-CoV-2 S1 and S2 (PM-WCPV-S-2), M (PM-WCPV-VME-2) and N (PM-WCPV-NCAP-2), the S1 and S2 regions of SARS (PM-CVHSA-S-1), MERS (PM-MERS-CoV-1), HKU1 (PM-HKU1-S-1) and OC43 (PM-OC43-S-1), and control HLAII CEFX peptide mix (PM-CEFX-3) were purchased from JPT. HLAI peptide mix was generated inhouse by combining 30 known EBV CD8+ T cell epitope peptides45.

For analysis of responses at the epitope level, individual 20mer peptides overlapping by 15 aa covering the full sequences of SARS-CoV-2 S, M and N were purchased from Alta Biosciences. Upon epitope identification, purified epitope peptides (>85% purity) and corresponding peptides (20mers and one 15mer for which the variant 20mer peptide could not be synthesized) from SARS-CoV-2 VOCs were synthesized by Alta Biosciences and Genscript. All peptides were resuspended in DMSO at a concentration of 5 mg ml–1. The concentration of all purified epitope peptides was confirmed using a Pierce Quantitative Colorimetric Peptide assay (Thermo Scientific) that binds to the peptide amide backbone, and peptides were adjusted to equivalent concentrations. The SARS-CoV-2 S protein is a soluble prefusion stabilized form, containing proline substitutions at positions F817P, A892P, A899P, A942P, K986P and V987P52. Epitope sequences recognized by the isolated CD4+ T cell clones are free of these stabilizing mutations.

Interferon-γ Elispot assay

Whole or CD8PBMCs (0.2–0.4 × 106 cells) were resuspended in standard medium (RPMI supplemented with 8% batch-tested FCS, 100 IU ml–1 penicillin and 100 μg ml–1 streptomycin) and added to duplicate or triplicate wells of IFNγ Elispot Pro Kit (Mabtech) plates containing 1 μg ml–1 peptide mix, 5 μg ml–1 purified epitope peptide or titrated concentration as stated, DMSO (negative control) and PHA or anti-CD3 (positive controls). Samples were incubated at 37 °C for 16–18 h. Plates were developed in accordance with the manufacturer’s instructions and read on a Bioreader 5000 Pro F Gamma (Bio-Sys GmbH). To quantify antigen-specific responses, mean DMSO values were deducted from all test wells and the results were expressed as spot forming cells (SFC) per 106 cells.

Polyclonal T cell generation and analysis

Polyclonal CD4+ T cell lines, generated by stimulation of CD8PBMCs with 1 μg ml–1 S, M or N peptide mix (JPT, 15 aa peptides overlapping by 11 aa), were cultured in RPMI supplemented with 5% batch-tested human serum (Gibco) and 50 IU ml–1 IL-2. Following 1–4 biweekly repeat stimulations with the same peptide mix, 50,000 polyclonal T cells were incubated for 16–18 h in V-bottom microtest plate wells with 1 μg ml–1 individual 20mer peptides overlapping by 10 aa covering S, M or N or DMSO (negative control). IFNγ release into the supernatant was tested by enzyme-linked immunosorbent assay (ELISA; Invitrogen) according to the manufacturer’s protocol. Responses against the test peptides were considered positive if greater than twice the mean of the control wells.

T cell clone isolation and assays

CD4+ T cell clones were isolated from polyclonal cultures as previously described53. Briefly, polyclonal CD4+ T cells were selected on rechallenge with 1 μg ml–1 appropriate peptide mix using an IFNγ cell enrichment kit (Miltenyi Biotech) followed by MACS separation using anti-PE beads. Enriched cells were cloned by limiting dilution seeding to establish T cell clones originating from single cells. Growing microcultures were screened for reactivity against individual 20mer peptides and selected clones were expanded using standard methods35.

CD4+ T cell clones (2,000 or 5,000 cells per well) were incubated in V-bottom 96-well microtest plate wells with 5 × 104 cells per well of autologous LCL or allogeneic LCLs with HLAII types partially matched to the autologous LCL. In some assays, LCLs were either pre-exposed for 1 h to 5 μM 20mer epitope peptide, 1 μg ml–1 peptide mix, or for 3 h to 1 ng ml–1 soluble prefusion-stabilized SARS-CoV-2 S protein52 or equivalent volumes of DMSO (negative control) before washing and addition to the wells, or added with a titrated concentration of individual 20mer peptides. Half maximal effective concentration (EC50) values were defined for each peptide as the concentration eliciting 50% maximum IFNγ release produced in response to the optimal WT peptide. T cell clone responses to mutated peptides from VOCs were quantified as the fold increase in concentration required to yield T cell activity equivalent to the EC50 of the WT peptide. In blocking assays, peptide-loaded or DMSO-exposed LCLs were incubated for 1 h with purified monoclonal antibodies against HLA-DR (L243, Biolegend), HLA-DQ (SPV-L3, Biotium) and HLA-DP (B7/21, Life Technologies) before addition of T cells. In all other assays, T cells were added immediately. The supernatant medium harvested after 16–18 h was assayed for IFNγ by ELISA.

pHLAII tetramer staining and TCR Vβ repertoire analysis

PBMCs from HCW-PI or UH individuals postvaccine were stained with optimized concentrations of peptide-HLAII tetramers (pHLAII Tet) (NIH tetramer core) containing S epitopes SSAN (DPB1*04:01/SSAN), GGNY (DPB1*02:01/GGNY) or STEC (DRB1*15:01/STEC), appropriate for the HLAII genotype of each individual. A total of 1.5–2 × 106 PBMCs were used per tube to enable collection of low frequency pHLAII Tet+ events. PBMCs were washed in PBS and stained with pHLAII Tet in batch-tested human serum for 1 h at 37 °C with regular resuspension. After incubation, cells were washed in PBS and stained at RT for 30 mins with BV510 anti-CD14 (MΦP9), BV510 anti-CD19 (SJ25C1), LIVE/DEAD Fixable Aqua Dead Cell Stain (Invitrogen), BV650 anti-CD3 (OKT3) and PerCP anti-CD4 (RPA-T4), plus appropriate combinations of the following TCR Vβ antibodies covering over 70% of normal human expressed TCR Vβ repertoire41: FITC anti-TCR Vβ3 (CH92), FITC anti-TCR Vβ4 (WJF24), FITC anti-TCR Vβ5.1 (IMMU 157), FITC anti-TCR Vβ5.2 (36213), FITC anti-TCR Vβ8 (56C5.2), FITC anti-TCR Vβ13.1 (H131), FITC anti-TCR Vβ20 (ELL1.4), FITC anti-TCR Vβ22 (IMMU 546), APC anti-TCR Vβ1 (REA662), APC anti-TCR Vβ5.3 (REA670), APC anti-TCR Vβ7.1 (REA871), APC anti-TCR Vβ7.2 (REA677), APC anti-TCR Vβ13.6 (REA554), APC anti-TCR Vβ14 (REA557), APC anti-TCR Vβ23 (REA497), AF647 anti-TCR Vβ13.2 (H132), APC-Vio770 anti-TCR Vβ2 (REA654), APC-Vio770 anti-TCR Vβ11 (REA559), APC-Vio770 anti-TCR Vβ16 (REA553), APC-Vio770 anti-TCR Vβ17 (REA915), APC-Vio770 anti-TCR Vβ21.3 (REA894). Cells were washed and resuspended in PBS following staining and data were acquired on a BD LSR Fortessa ×20 flow cytometer (BD Biosciences) using FACSDiva (v.9.0). All data was processed using FlowJo analysis software (v.10.6.1).

Serological analysis

Quantitative IgG antibody titers against SARS-CoV-2 S and N and β-HCoV family S proteins were measured using a multiplex serology assay (V-PLEX COVID-19 Coronavirus Panel 2 (IgG) kit, catalog no. K15369U), according to the manufacturer’s instructions. In brief, 96-well plates were blocked and washed. Samples were prediluted 1:4,000 in provided sample diluent and added to the wells in duplicate alongside the reference standard and assay kit controls. Following incubation, washing and addition of anti-IgG detection antibodies, read buffer was added to all wells and plates were measured immediately using a MESO Quickplex SQ 120 System (Meso Scale Discovery). Data were generated by Methodological Mind software (v.1.0.36), adjusted for sample dilutions and analyzed using MSD Discovery Workbench (v.4.0).

Bioinformatic analysis

Spike amino acid alignments were performed using the MUSCLE algorithm54 with default settings. Protein sequence identity was calculated using Expasy55. Correlation plots were prepared in R v.4.0.3 using corrplot (v.0.84).

Statistics and reproducibility

No statistical methods were used to predetermine cohort sizes and researchers were not blinded to the serostatus of donors before Elispot and serological assays. All statistical tests were performed in GraphPad Prism (v.9.3.1).

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.