CD4+ T cells coordinate critical aspects of adaptive immunity including B cell activation and maturation, CD8+ T cell responses and production of antiviral cytokines. SARS-CoV-2 infection induces robust CD4+ T cell responses1,2 that persist for at least 8–12 months after infection3,4,5,6,7. Such responses are postulated to contribute to the control of SARS-CoV-2 infection through multiple mechanisms. CD4+ and CD8+ T cell responses to spike, membrane and nucleocapsid have been associated with reduced coronavirus disease 2019 disease severity5, suggesting a role in limiting viral pathogenesis. Additionally, spike-specific CD4+ TFH cells can support B cell maturation and neutralizing antibody production following SARS-CoV-2 infection or vaccination8,9,10. Previous studies suggest that TFH cells are useful correlates of neutralizing antibody titers, both after infection3,4,5,6 or following vaccination11.

SARS-CoV-2 affords a unique opportunity to study CD4+ T cell immunity to a novel viral infection directly in humans, in particular for detailed characterization of the establishment, maintenance and recall of both CXCR5 memory T (TM) cells and cTFH cells. Studies suggest both primary infection and vaccination with COVID-19 vaccines predominantly elicit CD4+ central memory T (TCM) cell responses with cTFH, type 1 helper T (TH1)-like and/or type 17 helper T (TH17)-like phenotypes3,11,12,13. Longitudinal follow-up analyses of convalescent cohorts suggest frequencies of spike-specific CD4+ T cells decline in a linear fashion over 8 months11,14, with some evidence that spike-specific cTFH cell frequencies are more stable13. In the context of primary vaccination, CD4+ T cell responses persist for >6 months, with cTFH cell frequencies peaking 1 month after vaccination before declining15.

Due to natural immune decay, vaccine effectiveness has shown signs of decline in many countries, making understanding the dynamics of recall of T cell memory by either booster vaccination or breakthrough infection increasingly important. Vaccine boosters have been reported to augment serological responses relative to two-dose immunization16; however, the impact of boosting on T cell memory is currently poorly defined. Furthermore, the emergence of variants of concern with substantial evasion of neutralizing antibody responses, such as B.1.1.529 (Omicron), have emphasized the likely importance of memory B and T cell recall in preventing severe illness following breakthrough infection. Immunization of COVID-19 convalescent cohorts therefore offers a robust model in which to study immune recall to antigenic challenge and the interplay of infection-induced and vaccine-induced immunity. Despite intense interest in the robust neutralizing antibody responses elicited by vaccination of convalescent individuals17, relatively little is known of the associated CD4+ T cell responses11,14,18.

To date, CD4+ T cell recognition of SARS-CoV-2 has been mostly assessed using stimulation-based assays (activation-induced marker (AIM) or cytokine expression) that quantify bulk responses to the spike, or specific protein sub-domains19. Characterization of the activation state or phenotype of antigen-specific cells directly ex vivo can be challenging due to the requirement for in vitro stimulation. Peptide–MHC (pMHC) tetramers allowing direct tracking of epitope-specific T cells has facilitated a detailed understanding of T cell memory following viral infection, particularly for CD8+ T cells20. However, epitope-level resolution of CD4+ T cell responses during acute viral infections is less common, with most data derived from chronic infections such as human immunodeficiency virus, hepatitis C virus, Epstein–Barr virus or cytomegalovirus21,22. In the context of SARS-CoV-2, pMHC tetramers have been used to demonstrate that two-dose mRNA vaccination drives robust germinal center and TFH cell responses in lymphoid tissues lasting for more than 6 months23. However, spike-specific CD4+ T cells in blood displayed limited expansion or activation following vaccination and were of a predominately TH1-like, not TFH-like, phenotype19. Whether these contrasting observations reflect epitope-specific differences, or the restriction of TFH cells to the draining lymph node, is currently unknown. Additional studies are needed to assess the relative immunodominance of different spike-derived CD4+ T cell epitopes, to compare infection-induced and vaccination-induced T cell antigen receptor (TCR) repertoires, and to understand how CD4+ T cell memory is recalled following booster vaccination or breakthrough infection.

Here, we use a novel HLA-DRB1*15:01 tetramer presenting a SARS-CoV-2 spike epitope (S751–767) to quantitatively and qualitatively characterize memory CD4+ T cells and cTFH cells following infection and vaccination. Our data provide key knowledge of nascent and recalled CD4+ T cell immunity, generating critical insights into the nature of prominent CD4+ T cell epitopes and biomarkers of effective immunity against SARS-CoV-2.


Identification of HLA-DRB1*15/S751-restricted CD4+ T cells

We3,24 and others25,26,27 previously identified an immunogenic CD4+ T cell spike-derived peptide encompassing the sequence NLLLQYGSFCTQLNRAL (S751–767; termed S751 hereafter). Antibody blockade of HLA molecules in AIM assays suggested that S751 peptide presentation occurred through HLA-DR (Extended Data Fig. 1a). HLA-DRB1*15:01 was the only HLA-DR allele shared by the majority of responders to S751 (Supplementary Table 1), and computational analysis of HLA–peptide binding using NetMHCII 2.3 (ref. 28) similarly predicted strong binding (half-maximal inhibitory concentration (IC50) of 12.1 nM) between S751 and HLA-DRB1*15:01. We therefore generated HLA-DRB1*15:01/ S751 tetramers to identify epitope-specific CD4+ T cells.

Tetramer specificity was assessed in HLA-typed individuals following SARS-CoV-2 infection or receiving COVID-19 vaccination. Staining of cryopreserved peripheral blood mononuclear cell (PBMC) samples with S751-PE tetramers identified a clear population of CD4+ tetramer-binding (TET751+) T cells in HLA-DRB1*15 individuals after infection (Fig. 1a) or vaccination (Fig. 1b). In contrast, individuals lacking the HLA-DRB1*15 allele exhibited no or negligible TET751+ cells (Fig. 1a,b). TET751 staining was compared to that of an HLA-DRB1*15:01 tetramer loaded with the previously described S236 epitope19 (Fig. 1c). S751 tetramer co-labeled cells when stained on both APC and PE fluorochromes, but did not co-stain with HLA-DRB1*15:01 tetramers loaded with a control CLIP or S236 peptide, confirming specificity for S751 (Fig. 1d). In vitro culture of post-vaccine PBMCs with S751 peptide confirmed both the proliferative capacity and specificity of the TET751+ cells (Extended Data Fig. 1b).

Fig. 1: Specificity of HLA-DRB1*15:01/S751 tetramer.
figure 1

a, Staining of HLA-DRB1*15:01/S751 tetramer on cryopreserved PBMCs from individuals who recovered from mild COVID-19 infection with or without the HLA-DRB1*15 allele. b, Tetramer staining on PBMCs collected before or after two doses of BNT162b2 vaccine in individuals with or without the HLA-DRB1*15 allele. c, Staining of PBMCs from a BNT162b2-vaccinated individual with HLA-DRB1*15:01 tetramer loaded with S236 peptide. d, Co-staining of PBMCs from a BNT162b2-vaccinated individual with HLA-DRB1*15:01 tetramers loaded with S751, S236 or a control CLIP peptide.

S751-specific T cells are not cross-reactive with human coronaviruses

SARS-CoV-2 spike cross-reactive CD4+ T cells have been identified in uninfected individuals and linked to sequence conservation between SARS-CoV-2 and endemic human coronaviruses (hCoVs), particularly within the S2 domain29,30. Alignment of the spike from SARS-CoV-2 and hCoVs (NL63, 229E, OC43 and HKU1) demonstrated limited conservation of residues within the S751 epitope (Extended Data Fig. 1c). Prediction of epitope recognition by NetMHCII 2.3 suggested an analogous epitope in NL63 and 229E (Extended Data Fig. 1c) could potentially bind HLA-DRB1*15:01 with a similar affinity to S751. We therefore stimulated PBMCs in vitro for 11 d in the presence of interleukin (IL)-2 with either the S751 peptide or analogous hCoV-derived peptides from NL63, OC43 or 229E. Staining with the S751 tetramer demonstrated robust recognition of cells expanded by the SARS-CoV-2 S751 peptide, but minimal expansion by hCoV-derived peptides (Extended Data Fig. 1d). To confirm this result, PBMCs stimulated with the S751 peptide were restimulated with either S751 or corresponding hCoV peptides. Expanded cultures showed strong AIM and CD154 responses to S751, but failed to respond to the analogous hCoV peptides (Extended Data Fig. 1e). Together, these data indicate that S751-specific T cells are not cross-reactive with hCoVs and that CD4+ cells binding DRB1*15:01/S751 tetramers are primed by the SARS-CoV-2 spike.

Establishment of S751-specific T cell memory following COVID-19

We next studied TET751+ T cells in a cohort3,24 of individuals who recovered from mild/moderate COVID-19 with the HLA-DRB1*15 allele (Supplementary Table 2; see gating in Supplementary Fig. 1a,b). In individuals with samples collected 20–60 d after symptom onset (n = 19), the median frequency of TET751+ cells was 0.0136% (interquartile range (IQR) 0.0095–0.0224%; Fig. 2a), approximately 34-fold higher than in HLA-DRB1*15 uninfected or unvaccinated individuals. TET751+ cells were predominately CD45RA, as expected for antigen-experienced cells following infection (Fig. 2b). Within this memory population, TET751+ cells were predominately CCR7+CD27+ (median 82.6%, IQR 70.9–88.4%), and were enriched for this TCM cell phenotype relative to the bulk CD4+ TM population (P = 0.0004; Fig. 2b).

Fig. 2: Establishment of S751-specific CD4+ T cell memory following mild COVID-19.
figure 2

a, Frequency of TET751+ cells (as a percentage of total CD4+) among uninfected (n = 9) or COVID-19 convalescent individuals sampled 20 to 60 d after symptom onset (n = 19). The lines indicate the median, and bars indicate the IQR. b, Representative plots demonstrating expression of CCR7 and CD27 on TET751+ or bulk TM (non-naïve CD4+) cell subsets. Comparison of TCM (CCR7+CD27+) cell phenotype between TET751+ or bulk TM cells assessed using the Wilcoxon test (two-sided, n = 19). c, Representative staining and frequencies of TET751+ cells 1 to 15 months after symptom onset (n = 21). d, Expression of CD38 on TET751+ cells during longitudinal follow-up (n = 19 at 20–60 d and >120 d; n = 15 at 61–120 d). Statistics were assessed by Kruskal–Wallis and Dunn’s multiple-comparisons tests. e,f, Expression of CCR6 and CXCR3 on TET751+ or TM cells during early (e) or late (f) convalescence (n = 19). Lines indicate the median and IQR. Statistics were assessed by Wilcoxon test (two-sided).

Source data

Longitudinal analysis of bulk S-specific CD4+ T cell responses by AIM assay previously estimated a half-life (T1/2) of 94–207 d13,24,31. To refine these kinetics at the level of a single epitope, we tracked the frequency of TET751+ cells in 21 individuals over 23 to 450 d after symptom onset (Fig. 2c). Notably, this allowed quantification of epitope-specific T cells at time points where S751 peptide-specific responses were undetectable by AIM (Extended Data Fig. 2a). TET751+ T cells declined rapidly during early convalescence, with a T1/2 of approximately 20 d (95% confidence interval (CI) 13–30 d; Fig. 2c and Extended Data Fig. 2b), before reaching a level of stable maintenance (T1/2 of ~377 d; 95% CI 283–503 d; Fig. 2c and Extended Data Fig. 2b). The median frequency of TET751+ cells at days 365–450 (n = 17) was 0.0038% (IQR 0.0024–0.0061%), approximately 3.6-fold lower than that during early convalescence but still significantly higher than that in uninfected controls (P < 0.0001; Extended Data Fig. 2c).

Spike-specific CD4+ T cell responses identified by AIM assays exhibit diverse phenotypes, with prominent CCR6+CXCR3 and CCR6CXCR3+ populations3,5. TET751+ T cells showed evidence of activation (by CD38 expression) for >60 d before returning to a resting phenotype (Fig. 2d). During early convalescence, TET751+ cells were either CCR6CXCR3+ (median 46.0%, IQR 36.1–53.9%) or CCR6CXCR3 (median 38.20%, IQR 33.8–50.4%; Fig. 2e). In contrast to the prominent CCR6+ S-specific population previously identified by AIM3,5, TET751+ cells were rarely CCR6+ and, over time, became proportionally enriched for CXCR3 expression, with a significant increase in the frequency of CCR6CXCR3+ cells among the TET751+ population in late versus early convalescence (P = 0.0003; Fig. 2e,f).

S751-specific and S236-specific T cells following mild COVID-19

HLA-DRB1*15:01 also presents the S236 peptide19, providing an opportunity to contrast T cell responses to two distinct epitopes within the same convalescent cohort. TET751+ and TET236+ cells comprised a similar proportion of the total CD4+ T cell population during early convalescence (Extended Data Fig. 3a). However, the memory phenotype of the two populations was strikingly divergent (Extended Data Fig. 3b). While most TET751+ cells were CD45RA, a significantly greater proportion of the TET236+ population exhibited a naïve phenotype (median 48.2% naïve for S236 versus 16.1% for S751, P = 0.0039; Extended Data Fig. 3b). We therefore compared non-naïve TET236+ T cell frequencies across early convalescent (20–60 d after infection) and uninfected participants. In stark contrast to TET751+ cells (Fig. 2a), frequencies of TET236+ cells were comparable between convalescent and control groups (Extended Data Fig. 3c), suggesting minimal expansion of S236 responses following mild COVID-19. Notably, S236-specific responses remained low even in individuals with the highest frequencies of TET751+ T cells (Extended Data Fig. 3d), with no correlation between S751 and S236 responses (Extended Data Fig. 3e). In contrast to cell frequencies, both the TET236+ and TET751+ populations showed evidence of activation relative to the bulk TM cell population (median 1.7% ICOS+CD38+ for TM, 5.6% for S236 and 23.6% for S751; Extended Data Fig. 3f). Within the total activated CD4+ TM cell compartment, however, TET751+ cells accounted for a significantly greater proportion of cells relative to TET236+ cells (P = 0.004; Extended Data Fig. 3g). Overall, while S751-specific T cells differentiate and proliferate during SARS-CoV-2 infection, the S236 response is subdominant and poorly recruited into the primary response, underscoring the heterogeneity of the CD4+ T cell response to SARS-CoV-2 spike.

TET751 + T cells with a cTFH cell phenotype

Similarly to other viral infections, the frequency and phenotype of spike-specific cTFH cells correlate with neutralizing antibody titers following COVID-19 (refs. 3,4,6,32). Given the relative dominance of the S751 response, we characterized the phenotype of the TET751+ cTFH cell population. During early convalescence (20–60 d after symptom onset), a median of 9.9% (IQR 5.6–18.4%) of TET751+ cells were cTFH cells (CD4+CXCR5+; Fig. 3a), broadly similar to the median frequency of total CD4+ T cells with a cTFH cell phenotype (11.2%, IQR 6.1–13.7%; Fig. 3a). The frequency of TET751+ cTFH cells declined over time in a single-phase pattern of decay with a T1/2 of 227 d (95% CI 179–287 d; Fig. 3b and Extended Data Fig. 2d). Among samples collected 365–450 d after symptom onset, TET751+ cTFH cells were detectable (frequency ≥ 0.003%) in 13 of 17 (76.5%) individuals. In contrast to the bulk TET751+ population at early convalescence (Fig. 2e), TET751+ cTFH cells predominately exhibited a CCR6CXCR3+ phenotype (median 66.7%, IQR 53.6–75.0%; Fig. 3c). Overall, we find that mild SARS-CoV-2 infection establishes long-lived spike-specific CD4+ T cell memory with both cTFH cell and CXCR5 phenotypes.

Fig. 3: Frequency and phenotype of CXCR5+ S751-specific cTFH cells.
figure 3

a, Representative staining and frequency of CXCR5+ cells among CD4+ TM and TET751+ cells 20–60 d after symptom onset (n = 19). Statistics were assessed by Wilcoxon test (two-sided). b, Longitudinal analysis of the frequency of TET751+ cells among the cTFH (CXCR5+ TM) cell population (n = 21). c, Expression of CCR6 and CXCR3 among TET751+ (blue) or bulk (gray) cTFH cells at 20–60 d after symptom onset (n = 19). Graph indicates the median and IQR.

Source data

Dynamics of TET751 + T cells after primary vaccination

S751-specific and S236-specific T cell responses in the context of vaccination were assessed using ten seronegative HLA-DRB1*15:01/02 participants who were immunized with a COVID-19 vaccine (n = 7 BNT162b2, n = 2 ChAdOx nCoV-19, n = 1 NVX-CoV2373; Supplementary Table 3). We note that direct comparisons should not be drawn between vaccine platforms due to the limited number of samples. All vaccinees exhibited expansion of TET751+ T cells after a single dose (median 30.3-fold increase), with 9 of 10 exhibiting a further increase following dose 2 (Fig. 4a). As previously reported20, TET236+ cells became activated after the first and second vaccine doses compared to pre-vaccine samples (P = 0.038 and 0.015, respectively), confirming the recruitment of S236-specific cells into the response (Extended Data Fig. 4a). However, vaccine-induced expansion of TET236+ memory T cells was only evident in 4 of 10 donors after the first dose, with a median 2.4-fold increase, and the S236 response remained subdominant to S751 (Extended Data Fig. 4b).

Fig. 4: Induction of S751-specific CD4+ T cells following primary vaccination.
figure 4

a, Frequencies of S751-specific T cells at baseline, week 3 after dose 1 or week 2 after dose 2 among nine previously uninfected individuals. Statistics were assessed by the Wilcoxon test. b, Representative staining of the S751 tetramer among CD4+ T cells following immunization with one dose of BNT162b2 in a previously uninfected individual. c, Longitudinal S751-specific T cell frequencies at baseline and following BNT162b2 vaccination among nine previously uninfected individuals. Closed circles indicate samples collected after dose 1, and open circles indicate samples collected after dose 2. d, CCR6 and CXCR3 expression on TET751+ cells at week 3 after dose 1 or week 2 after dose 2 among seven individuals vaccinated with BNT162b2. Lines indicate the median, and bars indicate the IQR. e, Longitudinal S751-specific cTFH cell frequencies at baseline and following BNT162b2 vaccination among seven previously uninfected individuals. Closed circles indicate samples collected after dose 1, and open circles indicate samples collected after dose 2. f, Circos plots indicating pairing of TRAV and TRBV genes among sorted TET751+ cells for three individuals after the second vaccine dose.

Source data

We next assessed the precise kinetics of the prominent S751-specific response following BNT162b2 vaccination. Longitudinal samples from vaccinated individuals demonstrated a rapid expansion of TET751+ T cells as early as 7 d after dose 1 (Fig. 4b), with increasing TET751+ cell frequencies over the next 21 d in all participants (Fig. 4b,c). Anti-S IgG titers were not detected above baseline until at least day 11 after dose 1 (Extended Data Fig. 5a), consistent with other reports19,30. TET751+ T cell frequencies peaked 7–14 d after the second dose and were maintained above baseline to 241 d (Fig. 4c). The memory phenotype of TET751+ cells shifted from predominately TCM (CCR7+CD27+; median 82.2%, IQR: 75.6–84.5%) after dose 1 to more heterogeneous TCM, transitional memory T (TTM) and effector memory T (TEM) phenotypes after dose 2 (median 57.0% TCM, IQR 37.8–69.6%; Extended Data Fig. 5b), with the majority of cells lacking CCR6 or CXCR3 expression (Fig. 4d).

Vaccine-induced activation of antigen-specific cTFH cells has proven to be an important predictor of the magnitude of the serological response following immunization33. The CD38+ICOS+PD-1+CXCR3+ cTFH cell population that emerges following influenza and yellow fever vaccination contains a high proportion of vaccine-specific cTFH cells, and temporally associates with the emergence of antibody-secreting cells34,35. We therefore assessed both total cTFH cell activation and TET751+ cTFH cell frequencies following BNT162b2 vaccination. There was limited evidence of a coordinated emergence of an activated cTFH cell population following dose 1, with one participant showing an increase in ICOS+CD38+ cTFH cells at week 1 after immunization, and a second participant exhibiting a transient peak at week 3 (Extended Data Fig. 5c). Across all seven participants, there was no further evidence for cTFH cell activation after dose 2 (Extended Data Fig. 5c). Assessment of TET751+ cTFH cells confirmed that, consistent with other reports11,19, vaccine dose 1 did drive expansion of antigen-specific cTFH cells in all participants (Fig. 4e). Frequencies of TET751+ cTFH cells remained relatively stable after vaccine dose 2 (Fig. 4e), with limited evidence of boosting in contrast to the total TET751+CD4+ population (Fig. 4c). Nonetheless, TET751+ cTFH cells exhibited a shift from a CCR6CXCR3 phenotype after dose 1 toward a CCR6CXCR3+ phenotype following dose 2 (Extended Data Fig. 5d). Overall, we find that spike-specific cTFH cell activation or expansion is limited to dose 1.

Restricted TCR repertoire of vaccine-induced TET751 + T cells

Analysis of SARS-CoV-2-specific TCR repertoires has previously facilitated the identification of CDR3 sequence motifs associated with disease severity36, and identification of hCoV cross-reactive TCR clonotypes present in uninfected individuals37,38. We therefore sorted single TET751+ T cells from three uninfected vaccinees and analyzed TRAV/TRBV gene usage and CDR3 sequences. Among the 136 TCRαβ pairs recovered, TRBV expression was highly skewed toward TRBV24-1 (55% of recovered sequences), TRVB20-1 (18%) and TRBV6-1 (9%) genes (Fig. 4f). Given the restricted TRBV repertoire, we sorted TET751+ cells from an additional two donors and compared CDR3 sequences from these three TRBV families. We identified at least five public clonotypes shared between two or more vaccinees (Extended Data Fig. 6), with highly conserved TRBV CDR3 sequence motifs evident in both public and private clonotypes. These data contrast with a public TCR clonotype recently reported for the HLA-DPB1-restricted S167 epitope24, which exhibited a conserved TRAV CDR3 motif but greater TRBV diversity.

Recall of S751-specific T cells after booster vaccination

Administration of third (booster) vaccine doses has been shown to substantially augment the spike-specific serological response17 and reverse both waning neutralizing antibody titers and vaccine effectiveness. However, it is unclear to what extent T cell responses are recalled or boosted. Eight participants in our primary vaccine cohort received a third dose at a median of 215 d later (range 157–241; Supplementary Table 3). We observed a median 7.3-fold drop in the frequency of the TET751+ T cell population in samples collected before dose 3 compared to samples collected after dose 2 (Fig. 5a). Booster vaccination efficiently expanded TET751+ T cell frequencies to levels equivalent, but not superior, to those after dose 2 for all participants (Fig. 5a). This stands in contrast to S-specific IgG, where peak titers after dose 3 were significantly higher than those after dose 2 (P = 0.0078; Fig. 5b).

Fig. 5: S-specific CD4+ T cell and serological response to vaccine dose 3.
figure 5

a, Frequencies of S751-specific T cells 2 weeks after dose 2, before dose 3 and at peak response following dose 3 (n = 8). Colors indicate the primary vaccination platform. Statistics were assessed between dose 2 and dose 3 responses by the Wilcoxon test (two-sided). b, S-specific IgG endpoint titers at matched time points to a. Statistics were assessed by Wilcoxon test (two-sided) between dose 2 and dose 3 (n = 8). c, Representative time course of the S751-specific response to third-dose vaccination, indicating the frequency of TET751+ T cells, and the expression of ICOS/CD38 and CCR6/CXCR3 on the S751-specific population. d, Proportion of TET751+ T cells over time following dose 3, and the frequency of ICOS+CD38+ or CXCR3+CCR6 cells among the TET751+ population at the indicated time points (n = 7). e, Kinetics of S-specific IgG titers following vaccine dose 3 (n = 7). f, Proportion of bulk TM or TET751+ cells expressing CXCR5 5–6 d after dose 3 (n = 7).

Source data

At 3 d after boost, TET751+ cells declined in frequency compared to baseline samples, with robust expansion evident around days 5 and 6 (Fig. 5c,d). In most participants, TET751+ T cell frequencies peaked at days 5–7, which coincided with maximal T cell activation (ICOS+CD38+ phenotype; Fig. 5d). Elevated S751-specific T cell frequencies persisted for at least 20–30 d after vaccination (Fig. 5d), although TET751+ cells had largely returned to a resting phenotype by this time (Fig. 5d). Interestingly, 6 of 7 individuals with longitudinal data showed shifting chemokine receptor expression during recall, with TET751+ cells expressing high levels of CXCR3 only between days 5 and 10 after dose 3 (Fig. 5c,d). S-specific IgG titers following a booster peaked between days 10–15, and were stable for the duration of follow-up (Fig. 5e). Despite the robust serological impact, TET751+ T cells did not show substantial evidence of CXCR5 expression after the booster dose (Fig. 5f). A median of 4.1% of TET751+ cells exhibited a cTFH cell phenotype at the peak of response, which was similar to or lower than the frequency of cTFH cells among bulk TM cells (Fig. 5f).

Given the subdominant S236 response during primary vaccination, we assessed whether the frequency or activation of TET236+ T cells changed after boosting. Booster vaccination did not significantly impact the frequency of S236-specific T cells (Extended Data Fig. 7a), although once again there was moderate evidence of activation (Extended Data Fig. 7b,c). It is pertinent to note that the S236-specific response was highly heterogeneous across the eight vaccinees. While some participants showed no change in TET236+ T cells, one participant exhibited a 4.2-fold increase in TET236+ T cell frequency that was maintained to 15 d after the third dose, with activation of up to 28% of the antigen-specific cells (Extended Data Fig. 7c). Host-specific differences in antigen processing or presentation may contribute to the differential activation and expansion of S236-specific and S751-specific T cells in such individuals.

Recall of infection-induced TET751 + T cells by vaccination

To further understand the impact of immunological priming on the recall of CD4+ T cell memory, we assessed S751-specific responses among convalescent individuals receiving COVID-19 vaccines (often termed ‘hybrid immunity’18,39,40). We obtained longitudinal samples from 12 HLA-DRB1*15 convalescent individuals who then received at least one dose of a COVID-19 vaccine (n = 7 ChAdOx nCoV-19, n = 5 BNT162b2; Supplementary Tables 2 and 4). Pre-vaccine samples were collected no more than 4 months before immunization, and participants were vaccinated a median of 441 d after SARS-CoV-2 symptom onset. We found that neutralization titers in the hybrid immunity cohort were comparable to those after third-dose responses in naïve vaccinees, both of which were significantly higher than titers after primary vaccination (P = 0.0234 versus two doses, P < 0.001 vs one dose; Fig. 6a). Regardless of the vaccine platform, robust expansion of TET751+ T cells was observed 1–2 weeks after immunization of the convalescent cohort (Fig. 6b). Vaccination with the adenoviral vaccine resulted in a median 6.4-fold increase of TET751+ cells (IQR 3.4–8.7), while mRNA vaccination drove significantly greater expansion (median 17-fold, IQR 13.6–26.3; Fig. 6c). Interestingly, mRNA vaccination in convalescent individuals also drove significantly greater expansion of TET751+ T cells than boosting of the primary vaccine cohort (median 6.3-fold change, IQR 5.7–7.3; Fig. 6c).

Fig. 6: Recall of S751-specific CD4+ T cells following vaccination of COVID-19 convalescent individuals.
figure 6

a, Neutralizing antibody titers against SARS-CoV-2 among the naïve vaccination cohort (3 weeks after dose 1 (n = 9), 2 weeks after dose 2 (n = 9) or 2 weeks after dose 3 (n = 5)) and convalescent individuals (2 weeks after dose 1 (n = 10)). Statistics were assessed by Kruskal–Wallis test with Dunn’s post test comparing groups against the convalescent group (two-sided). Lines indicate the median, and bars indicate the IQR. b, Changes in TET751+ T cell frequency between pre-vaccine and post-vaccine samples (1–2 weeks after dose 1) among convalescent individuals. Statistics were assessed by the Mann–Whitney test. c, Fold change in TET751+ T cell frequencies according to the vaccine platform. Statistics were assessed by the Kruskal–Wallis test with Dunn’s post test. Lines indicate the median, and bars indicate the IQR. d, Time course of S751-specific T cell expansion in a single convalescent individual before and after a single dose of BNT162b2. e, Longitudinal tetramer frequencies among 12 individuals (n = 7 AstraZeneca, red; n = 5 Pfizer/BioNTech, blue). Pre-vaccine samples are set at day −1. Blue or red shading indicates days 3–4 or 4–12 after immunization, respectively. f, Expression of CCR7/CD27 and CCR6/CXCR3 on TET751+ cells at early (5–10 d) or late (>11 d) time points after vaccination. n = 5 ChAdOx nCoV-19, red; n = 5 BNT162b2, blue; statistics were assessed by the Wilcoxon test (two-sided).

Source data

To precisely map the kinetics of antigen-specific T cell recall, we analyzed samples from day 3 to day 65 after vaccination. The frequency of TET751+ cells in the circulation consistently declined at day 3 relative to baseline samples (Fig. 6d,e), likely reflecting activation and/or retention of antigen-specific cells in lymphoid tissues. By day 5, robust CD4+ T cell proliferation was evident (Fig. 6d). Among individuals with frequent sampling, the frequency of TET751+ cells tended to peak between days 9 and 12 after vaccination (Fig. 6d,e), potentially later than observed in the third-dose cohort. While contraction could not be followed in BNT162b2-vaccinated participants due to the 3-week boost schedule, participants vaccinated with ChAdOx nCoV-19 exhibited a gradual decline in TET751+ cells over 40 d after their first dose (Fig. 6e). In contrast to the predominant resting TCM cell phenotype of S751-specific T cells generated after infection, early recall responses (days 5–10 after vaccination) exhibited a notable shift toward TTM (CCR7CD27+) and TEM (CCR7CD27) cell phenotypes (Fig. 6f), before returning to a TCM cell-dominated phenotype (Fig. 6f). The shifts in memory phenotype coincided with transient high levels of CXCR3+ expression (Fig. 6f).

Activation of S751 + cTFH cells in hybrid immunity

In contrast to primary vaccination, activated cTFH cells were rapidly and transiently induced by vaccination of convalescent participants (Fig. 7a and Extended Data Fig. 8a). The appearance of activated cTFH cells occurred as early as day 4 after vaccination and typically waned by day 12 (Fig. 7a). Compared to parental cTFH cells, these activated cells were enriched for CXCR3 expression (Fig. 7b), resembling the cTFH1 cells recalled by influenza vaccination. Previous work has suggested that ~40% of activated cTFH cells exhibit specificity for vaccine antigens35, but pMHC tetramers offer the opportunity to determine the prominence of individual epitopes in this population. Across individuals, the frequency of S751-specific cells within ICOS+CD38+ cTFH cells ranged from less than 1% up to 11.9% (Fig. 7c).

Fig. 7: Vaccine-associated recall of activated and S751-specific cTFH cells in convalescent individuals.
figure 7

a, Representative staining and frequency of ICOS+CD38+ cTFH cells following single-dose vaccination of convalescent participants (n = 12). Gray shading indicates days 4–11 after vaccination. b, Representative staining of CCR6 and CXCR3 on total cTFH (CD4+CXCR5+) cells or ICOS+CD38+ cTFH cells. Data are representative of ten individuals with samples available from days 5–11 after vaccination. c, S751 tetramer binding within the ICOS+CD38+ cTFH cell population in two individuals with low (<1%) or high (>10%) S751-specific frequencies. d, Longitudinal S751-specific cTFH cell frequencies among n = 12 convalescent participants following vaccination. The limit of detection (0.003%) is indicated by the dashed line. e, Expression of ICOS and CD38 on TET751+ cTFH cells (blue) in a single individual over time. f, Proportion of S751-specific cTFH cells with an activated (ICOS+CD38+) phenotype over time (n = 10). g, Expression of PD-1 on S751-specific cTFH cells following vaccination (n = 10).

Source data

The persistence and long-term activation state of cTFH cells is unclear. We tracked the frequency and phenotype of TET751+ cTFH cells, and found that these cells are recalled with similar kinetics to CXCR5 cells, and persist in the circulation for substantially longer than the ICOS+CD38+ cTFH cell population (Fig. 7d). Phenotypic analysis of TET751+ cTFH cells clearly demonstrated that while this population emerges at day 5 after vaccination with an ICOS+CD38+PD-1+ phenotype (Extended Data Fig. 8b), ICOS and CD38 expression are rapidly lost over the subsequent 7 d (Fig. 7e,f). By 4 weeks after vaccination, less than 50% of S751-specific cTFH cells expressed PD-1 (Fig. 7g), indicating that recalled antigen-specific cTFH cells persist in the circulation as a resting, CD38ICOSPD-1+/− pool. Together, these data indicate that enumerating only activated or PD-1+ cTFH cells likely underestimates the total spike-specific population, particularly at later time points.

Finally, we assessed whether S751-specific cTFH cell memory established by primary immunization or infection would be predictive of the neutralizing antibody responses to booster dose or first-dose (hybrid immunity cohort) vaccines. Interestingly, pre-boost/vaccine TET751+ cTFH cell memory frequency positively correlated with post-vaccine neutralizing antibody titers among the combined third-dose/hybrid immunity cohort (P = 0.032, r = 0.563; Extended Data Fig. 8c), suggesting that previous establishment of spike-specific cTFH cell memory may support the serological recall response upon antigen reexposure.

Recall of infection-induced TCR clonotypes by vaccination

To investigate the clonal relationships between CXCR5 TM cell and cTFH cell populations, we sequenced 187 TET751+ cells from an individual during early convalescence, day 8 after vaccination and day 29 after vaccination (CP24, ChAdOx nCoV-19 vaccine). Following infection, 27 clonotypes represented ~80% of the sequenced TCR repertoire (Fig. 8a). Nine of these clonotypes were identified in subsequent samples, constituting ~20% of the post-vaccine repertoire and showing direct recall of S751-specific memory (Fig. 8a). TRAV and TRBV gene usage was similarly biased to that seen in primary vaccination, with prevalent TRBV20.1, TRBV24.1 and TRBV6.1 usage (Fig. 8b). Two additional convalescent participants vaccinated with BNT162b2 (CP28, day 9 after vaccine; and CP60, day 12 after vaccine) similarly showed post-infection clonotypes making up ~25% of the post-vaccine TCR repertoire (Extended Data Fig. 9a).

Fig. 8: Longitudinal tracking of S751-specific TCR clonotypes.
figure 8

a, Persistence of TRBV clonotypes across three longitudinal samples in a single convalescent individual. Colors identify the 27 most frequent clonotypes constituting 80% of the recovered repertoire at the day 70 convalescent time point. b, Circos plots indicating TRAV and TRBV pairing at each indicated time point. c, Clonotype sharing between cTFH cells (CXCR5+) and CXCR5 TM cells across among cells recovered from any time point.

Strikingly, multiple clonotypes from convalescent individuals were shared with naïve vaccinees (Supplementary Table 5), suggesting comparable recognition of spike-derived epitopes between infection and vaccination. A TRBV 20-1 clonotype identified in 4 of 5 naïve vaccinees (CSARRGTEAFF) was also identified in all three convalescent individuals sequenced, highlighting the existence of highly public clonotypes within S751-specific T cell responses. Finally, we compared TET751+ cTFH and CXCR5 TM cell populations to understand the degree of clonal overlap between these functionally distinct CD4+ T cell subsets. Across post-infection and post-vaccination time points, all cTFH cell-derived clonotypes were also identified among TET751+ CXCR5 cells (Fig. 8c and Extended Data Fig. 9b), suggesting no differential recruitment into TM and cTFH cell compartments.


Intracellular cytokine staining and AIM assays have been instrumental in defining CD4+ T cell responses to SARS-CoV-2. Now, novel HLA class II tetramers restricting immunogenic peptides allow epitope-specific characterization of the ex vivo phenotype, establishment and recall of memory CD4+ T cell and cTFH populations. Our study provides a detailed comparative analysis between two HLA-DRB1*15:01 epitopes, demonstrating differential activation and recruitment into immune responses after infection or vaccination, expanding our understanding of CD4+ T cell immunodominance hierarchies in humans. Given the critical role for CD4+ TFH cells in supporting antibody production, identifying the mechanisms governing CD4+ T cell immunodominance can inform rational vaccine design and support the inclusion of prominent T cell epitopes within novel vaccine immunogens28.

We find that S751-specific T cells were detected in both convalescent and vaccinated participants at frequencies comparable to other reported HLA class I41,42,43 and II19,23-restricted SARS-CoV-2 epitopes, as well as epitopes in influenza infection44, Respiratory syncytial virus infection45 or yellow fever vaccination46. Interestingly, while frequencies of SARS-CoV-2 HLA class I-restricted CD8+ T cells appear stable during convalescence41,43, we observed a rapid decline in TET751+ cells during the first 4 months after symptom onset, before a stable equilibrium in both CXCR5 and cTFH cell populations lasting beyond 1 year.

A pertinent difference in the characterization of spike-specific CD4+ T cells detected by the S751 tetramer versus an AIM assay is the relative frequency of CCR6+ cells. Both we3,24 and others5,13 found a substantial proportion of S-specific CD4+ T cells express CCR6 using AIM, albeit without IL-17 production3. This was also the case for S751-specific cells identified by AIM (via OX-40 and CD25/CD137)24, while the S751-tetramer identified only a negligible frequency of CCR6+ cells. Clarifying whether CCR6 expression relates to pMHC class II tetramer affinity, or represents cells upregulating CCR6 upon stimulation requires clarification in future studies.

In contrast to serological responses39,47, less is known about the recall of CD4+ T cells by SARS-CoV-2 vaccines in either previously uninfected or convalescent individuals. We found that infection or primary vaccination elicits a similar frequency of TET751+ CD4+ T cells, albeit with differences in CXCR3 expression that may reflect distinct cytokine microenvironments during T cell priming. Two vaccine doses elicited S751-specific T cells in previously uninfected participants at frequencies comparable to single-dose vaccinated convalescent participants. A third vaccine dose (booster) only restored waning S751-specific T cell immunity to levels seen immediately after the second dose, in contrast to binding and neutralizing antibody titers, which were boosted to elevated levels. Nonetheless, frequencies of TET751+ cTFH cells in pre-immune individuals were positively correlated with neutralizing antibody responses to subsequent vaccine doses, further supporting a role for cTFH cells as correlates of neutralizing antibodies48 and highlighting the value of establishing robust cTFH cell memory during immune priming.

Robust cTFH cell recall was prominent in convalescent participants being vaccinated, but was more limited in the context of a third dose (booster). Whether this is a feature of hybrid immunity17 or extended duration (~1 year) between the immunological ‘prime’ and vaccine boost is unclear. Nonetheless, identification of S751-specific cTFH cells provides a unique opportunity to study cTFH cell memory and recall in humans. Herati et al. previously provided evidence that resting influenza-specific CXCR5+CD38ICOS cTFH cells may constitute memory recallable by subsequent vaccination49. We now find that both SARS-CoV-2 infection or vaccination induces long-lived spike-specific cTFH cell memory; however, the relative contributions of TM cell and cTFH cell populations into recall responses to subsequent vaccination are difficult to deconvolute. After vaccination, S751-specific cTFH cells reacquire a resting phenotype (CD38ICOS) within 2 weeks, and later substantially downregulate PD-1 expression. Therefore, while the study of CD38+ICOS+, or even PD-1+, cTFH cells captures antigen-specific cells during acute time points, the accurate enumeration and phenotypic characterization of memory cTFH cells established after vaccination or infection likely requires antigenic restimulation or pMHC tetramers.

Our HLA-DRB1*15:01-S751 tetramer enabled characterization of the TCR repertoire, revealing the recruitment of TCR clonotypes established by previous infection into the immune response after subsequent vaccination, clearly demonstrating recall of CD4+ T cell memory. Future studies of the diversity of the total spike-specific T cell repertoire will be informative in understanding whether repeated vaccination selectively expands high-avidity T cell clones. Further, the high degree of clonal overlap observed between S751-specific CXCR5 TM cells and cTFH cells suggests TCR clonotypes do not segregate into distinct CD4+ helper T lineages. Finally, the restricted TCR repertoire and the abundance of highly public TCR clonotypes suggests structural convergence toward paratope hotspots50 and warrants further investigation to understand its molecular basis and potential implications.

Limitations of the current work include the restricted cohort size due to the requirement of individuals with specific HLA alleles. Additionally, the dynamics of S751-specific CD4+ T cells may not be representative of other spike or non-spike CD4+ T cell epitopes. In particular, we note that the description of S751-specific responses in non-mRNA-vaccinated individuals is limited, and we lacked the statistical power to perform a comparative analysis of different vaccine platforms. A more detailed understanding of tetramer-specific responses across vaccine platforms will be important in future studies. Overall, we find that SARS-CoV-2 infection or vaccination generates spike-specific cTFH cells and memory CD4+ T cells capable of being recalled upon antigen reexposure; however, individual epitopes are recruited with varying efficiency. Defining the drivers of CD4+ T cell immunodominance remains critical for understanding vaccine immunogenicity and design.


Participant recruitment and sample collection

The study protocols were approved by the University of Melbourne Human Research Ethics Committee (2056689 and 21198153983), and all associated procedures were carried out in accordance with the approved guidelines. All participants provided written informed consent in accordance with the Declaration of Helsinki. Participants were not compensated for their participation. Sex and age of participants is shown in Supplementary Tables 14.

A longitudinal cohort of individuals who recovered from COVID-19 (previously described in Juno et al.3 and Wheatley et al.24) were recruited to provide additional blood samples following vaccination against SARS-CoV-2. All cohort participants had either a previous positive nasal PCR during early infection for SARS-CoV-2 or clear exposure to SARS-CoV-2 as well as a positive ELISA for SARS-CoV-2 spike and receptor-binding domain protein as previously reported3. Contemporaneous controls who did not experience any symptoms of COVID-19 and who were confirmed to be seronegative were also recruited to provide blood samples before and following vaccination for SARS-CoV-2. No statistical methods were used to predetermine sample sizes. All participants in existing cohorts with the HLA-DRB1*15 allele were included, and our sample sizes are similar to those reported in previous publications19,23. For all participants, whole blood was collected with sodium heparin anticoagulant. Plasma was collected and stored at −80 °C, and PBMCs were isolated via Ficoll-Paque separation, cryopreserved in 90% fetal calf serum (FCS)/10% dimethylsulfoxide (DMSO) and stored in liquid nitrogen. HLA typing by the Victorian Transplantation and Immunogenetics Service was available on all participants.

Generation of MHC II tetramers

Human DRB1*15:01 NLLLQYGSFCTQLNRAL (SARS-CoV-2), DRB1*15:01 TRFQTRFQTLLALHRSYLT and DRA1*01:01/DRB1*15:01 PVSKMRMATPLLMQA (CLIP) biotinylated monomers were generated by ProImmune. Biotinylated monomers were tetramerized by sequential addition of streptavidin-PE (BD Biosciences) or streptavidin-APC (BioLegend).

Tetramer staining

Cryopreserved PBMC samples were thawed in RPMI-1640 with 10% FCS and penicillin–streptomycin (RF10), washed and counted. Up to 10 × 106 PBMCs were washed in 2% FCS/PBS before incubation in 50 nM dasatinib (Sigma) for 30 min at 37 °C. Tetramer conjugated to PE or APC was then added at 4 μg ml−1 for 60 min at 37 °C. Cells were washed in PBS, stained with Live/Dead fixable green dead cell stain (Life Technologies), and incubated for 30 min at 4 °C with a surface stain antibody cocktail. Surface stain antibodies included: CD45RA PerCP-Cy5.5 (HI100), CCR7 Alexa Fluor 647 (G043H7), CD69 APC Fire-750 (FN50), CD27 BV510 (MT-T271), CD4 BV605 (RPA-T4), PD-1 BV650 (EH12.2H7), CCR6 BV785 (G034E3) and CXCR3 PE Dazzle594 (G02H57; BioLegend), CD38 Alexa Fluor 700 (HIT2), ICOS BV421 (C398.4), CD3 BUV395 (SK7) and CD20 BUV805 (2H7; BD Biosciences), and CXCR5 PE-Cy7 (MU5UBEE; Thermo Fisher). Cells were then washed with 2% FCS/PBS and fixed with Cytofix (BD Biosciences), before acquisition on an LSR Fortessa (BD Biosciences). Data were analyzed using FlowJo v10.2 (TreeStar). Data collection and analysis were not performed blind to the conditions of the experiments.

Activation-induced marker assay

Cryopreserved PBMC samples were thawed, seeded at 1–2 × 106 cells per well of a 96-well plate, and rested for 4 h at 37 °C. Cells were then stimulated with 1 μg ml−1 of peptide or an equivalent volume of DMSO for 20 h. In some experiments, CD154 APC-Cy7 (TRAP-1, BD Biosciences) antibody was included in the culture medium for the duration of the stimulation. Cells were then washed in PBS and stained with Live/Dead green, and surface stained with the following antibodies: OX-40 PerCP-Cy5.5 (Ber-ACT35), CD25 APC (BC96), CD137 BV421 (4-B41), CD27 BV510 (MT-T271), CD4 BV605 (RPA-T4), CCR6 BV785 (G034E3) and CXCR3 PE Dazzle594 (G02H57; BioLegend), CD45RA PE-Cy7 (HI100) and CD3 BUV395 (SK7; BD Biosciences) and CXCR5 PE (MU5UBEE, Thermo Fisher). For HLA blocking experiments, PBMCs were preincubated with 8 μg ml−1 of purified HLA-DR (L243, BioLegend), or mouse IgG κ isotype control (MOPC-21, BioLegend) for 1 h before peptide stimulation.

In vitro S751 proliferation assay

To expand S751-specific T cells in vitro, 3–5 × 106 freshly isolated or thawed cryopreserved PBMC samples were seeded in 96-well plates and stimulated with 1 μg ml−1 of SARS-CoV-2 S751, NL63 S801, 229E S618, OC43 S833 or an equivalent volume of DMSO for 9–10 d. At days 3/4 and 6/7, the culture medium was replenished and supplemented with 10 U ml−1 recombinant human IL-2 (PeproTech). On day 9 or 10, cells were stained for S751 tetramer or antigen-specific responses measured via AIM assay. In some experiments, cells were stained with 2.5 μM CellTrace Violet Proliferation Dye (Thermo Fisher) before stimulation with peptide S751. In such cases, PBMCs were cultured for 6 d and supplemented with 10 U ml−1 IL-2 at day 3.

Single-cell sorting and TCR sequencing

Up to 10 × 106 thawed PBMCs were stained with tetramer, followed by viability staining with Live/Dead green. Cells were then surface stained for 30 min at 4 °C with: CD45RA PerCP-Cy5.5 (HI100), CCR7 Alexa Fluor 647 (G043H7), CD4 BV605 (RPA-T4), CCR6 BV785 (G034E3) and CXCR3 PE Dazzle594 (G02H57, BioLegend), CD3 APC-H7 (SK7) and CD20 BV510 (2H7, BD Biosciences) and CXCR5 PE-Cy7 (MU5UBEE, Thermo Fisher). Cells were sorted into 96-well plates using a BD FACS Aria III sorter and frozen until cDNA synthesis. cDNA was synthesized by reverse transcription using 450 ng of random hexamer primers, 2 µl of 10 mM dNTP, 0.1 M dithiothreitol, 0.25% vol/vol Igepal, RNAsin (Promega) and 120 U Superscript III reverse transcriptase (Invitrogen). PCR was performed at 42 °C for 10 min, 25 °C for 10 min, 50 °C for 60 min and 94 °C for 5 min, and cDNA was stored at −20 °C. TRAV and TRBV genes were amplified by nested PCR. First-round PCR reactions were prepared using 10 µl of cDNA template, 10 mM dNTP, HotStar Taq Plus Polymerase and the following primers (Supplementary Table 6)51: TRAC-EXT, TRAV-EXT (cocktail), TRBC-EXT and TRBV-EXT (cocktail). Secondary PCR reactions were carried out independently for TRAV or TRBV transcripts using 2.5 µl of unpurified primary PCR product and either TRAC-INT/TRAV-INT primers or TRBC-INT/TRBV-INT primer cocktails. All nested PCR reactions were performed for 40 cycles at 95 °C for 20 s, 52 °C for 30 s and 72 °C for 45 s. Recovered PCR products were subjected to Sanger sequencing and productive TCR sequences were aligned using IMGT52. Analysis of clonotype sharing between participants or time points was performed using the Immunarch package (Immunomind) in R 3.6.2. Visualization of alpha and beta V gene pairing was performed using Circlize53.


MaxiSorp plates (96-well; Thermo Fisher) were coated overnight at 4 °C with 2 μg ml−1 recombinant SARS-CoV-2 S proteins (HexaPro). After blocking for 1 h, 25 °C with PBS + 1% FCS, plasma samples were serially diluted in PBS + 1% FCS before incubation for 2 h at 25 °C. Plates were then washed using PBST (PBS with 0.05% Tween-20), before incubation with a 1:20,000 dilution of horseradish peroxidase-conjugated anti-human IgG (Sigma) for 1 h. Plates were washed and developed using tetramethylbenzidine substrate (Sigma), stopped using 0.16 M sulfuric acid and read at 450 nm. Endpoint dilutions were calculated using a fitted curve (four-parameter log regression) and Prism 9.0 software (GraphPad).

Microneutralization assay with ELISA-based readout

Wild-type SARS-CoV-2 (CoV/Australia/VIC/01/2020) isolate was passaged in Vero cells (obtained from the Victorian Infectious Diseases Reference Laboratory) and stored at −80 °C. Next, 96-well flat-bottom plates were seeded with Vero cells (20,000 cells per well in 100 µl). The next day, Vero cells were washed once with 200 µl serum-free DMEM and added with 150 µl of infection medium (serum-free DMEM with 1.33 µg ml−1 TPCK trypsin). Then, 2.5-fold serial dilutions of heat-inactivated plasma (from 1:20 to 1:12,207) were incubated with SARS-CoV-2 virus at 2000 TCID50 (50% tissue culture infectious dose) per ml at 37 °C for 1 h. Next, plasma–virus mixtures (50 µl) were added to Vero cells in duplicate and incubated at 37 °C for 48 h. ‘Cells only’ and ‘virus + cells’ controls were included to represent 0% and 100% infectivity, respectively. After 48 h, all cell culture media were carefully removed from wells and 200 µl of 4% formaldehyde was added to fix the cells for 30 min at 25 °C. The plates were then dunked in a 1% formaldehyde bath for 30 min to inactivate any residual virus before removal from the BSL3 facility. Cells were washed once in PBS and then permeabilized with 150 µl of 0.1% Triton-X for 15 min. Following one wash in PBS, wells were blocked with 200 µl of blocking solution (4% BSA with 0.1% Tween-20) for 1 h. After three washes in PBST, wells were added with 100 µl of rabbit polyclonal anti-SARS-CoV N antibody (Rockland, 200-401-A50) at a 1:8,000 dilution in dilution buffer (PBS with 0.2% Tween-20, 0.1% BSA and 0.5% NP-40) for 1 h. Plates were then washed six times in PBST and added with 100 µl of goat anti-rabbit IgG (Abcam, ab6721) at a 1:8,000 dilution for 1 h. After six washes in PBST, plates were developed with tetramethylbenzidine and stopped with 0.15 M sulfuric acid. Optical density values read at 450 nm were then used to calculate the percentage neutralization with the following formula: (‘virus + cells’ − ‘sample’) ÷ (‘virus + cells’ − ‘cells only’) × 100. IC50 values were determined using four-parameter nonlinear regression in GraphPad Prism with curve fits constrained to have a minimum of 0% and a maximum of 100% neutralization.


To compare the decay phase of CD4+ and TFH cells after the peak, we modeled a fraction, f, of cells at the peak as short-lived cells, and the remainder (1 − f) as long-lived cells, which decay independently. The model can be written as:

$$Y(t) = Y_0\left[ {fe^{ - \delta _1t} + (1 - f)e^{ - \delta _2t}} \right]$$

in which

Y0 = peak levels of cells,

f = fraction of short-lived cells,

δ1 = death rate of short-lived cells, and

δ2 = death rate of long-lived cells.

A censored nonlinear mixed-effect model was used to fit the model to the longitudinal T cell data. The limit of detection was fixed to 0.0001% (for total CD4+ cells) and 0.003% (for TFH cell population). We also tested if the data could be fitted with just a single decay (that is, setting f = 1 and δ2 = 0 in the equation above) or using the non-constrained equation (a biphasic model with both f and δ2 as free parameters). Model comparison was performed based on the likelihood ratio test by comparing the likelihood value of the nested models and the difference in the number of parameters. These analyses were carried out in Monolix R2019B.

Flow cytometry data were analyzed with FlowJo v10.2. Statistical analysis was performed in GraphPad Prism 9 (TreeStar) using non-parametric statistical tests as indicated (making no assumptions about data normality). P < 0.05 was considered significant. No data were excluded from the study.

Reporting Summary

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