SARS-CoV-2 spike-specific TFH cells exhibit unique responses in infected and vaccinated individuals

Long-term humoral immunity to SARS-CoV-2 is essential for preventing reinfection. The production of neutralizing antibody (nAb) and B cell differentiation are tightly regulated by T follicular help (TFH) cells. However, the longevity and functional role of TFH cell subsets in COVID-19 convalescents and vaccine recipients remain poorly defined. Here, we show that SARS-CoV-2 infection and inactivated vaccine elicited both spike-specific CXCR3+ TFH cell and CXCR3− TFH cell responses, which showed distinct response patterns. Spike-specific CXCR3+ TFH cells exhibit a dominant and more durable response than CXCR3− TFH cells that positively correlated with antibody responses. A third booster dose preferentially expands the spike-specific CXCR3+ TFH cell subset induced by two doses of inactivated vaccine, contributing to antibody maturation and potency. Functionally, spike-specific CXCR3+ TFH cells have a greater ability to induce spike-specific antibody secreting cells (ASCs) differentiation compared to spike-specific CXCR3− TFH cells. In conclusion, the persistent and functional role of spike-specific CXCR3+ TFH cells following SARS-CoV-2 infection and vaccination may play an important role in antibody maintenance and recall response, thereby conferring long-term protection. The findings from this study will inform the development of SARS-CoV-2 vaccines aiming to induce long-term protective immune memory.


INTRODUCTION
Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is the causative agent of coronavirus disease 2019 (COVID- 19), which poses a serious health threat and has had considerable socioeconomic consequences worldwide. 1,2Effective strategies are urgently needed to establish persistent and immune memory of the appropriate magnitude at the individual and population levels to prevent the continued spread of infection.Thus, understanding how immune memory is successfully established in vaccinated individuals and those who have recovered from COVID-19 is essential for rational vaccine design to elicit long-lasting humoral and cellular immune responses against SARS-CoV-2 and variants of concern (VOCs).
][5][6][7][8][9][10][11][12] The early appearance of neutralizing antibodies (nAbs) associated with less severe disease in acute COVID-19 and the persistence of nAbs in recovered individuals contribute to preventing reinfection by blocking virus entry. 13,14Endemic human coronaviruses (HCoV-229E, HCoV-OC43, HCoV-NL63, and HCoV-HKU1), which usually only infect the upper respiratory tract, frequently cause homologous reinfection, and this may be due to the short lifespan of the immunoglobins that they induce. 15][23] The production and maturation of high-affinity antibodies and memory B cells, as well as long-lived plasma cell differentiation, mainly rely on germinal center (GC) reactions in secondary lymphoid tissues, which are tightly regulated by T follicular helper (T FH ) cells. 24][27][28] Circulating T FH cells may serve as GC T FH cell counterparts, as they express low levels of PD-1, ICOS, and Bcl6 and exhibit a memory phenotype. 29][32][33][34] In the acute phase of SARS-CoV-2 infection, large amounts of low-affinity antibodies are produced rapidly, and in parallel, antigen-specific CD4 + T cells and circulating T FH cells appear, which in turn contribute to antibody production to combat infection. 4,13,35,36In COVID-19 convalescents, we and others have previously demonstrated that CXCR3 + T FH cells directly correlate with anti-spike antibody responses, [37][38][39] and that some epitope (HLA-DRB1*15:01/S751)-specific T FH cells with a long half-life (T 1/2 = 227 days) also predominantly exhibit a CXCR3 + CCR6 − phenotype. 40Several studies have also shown that CCR6 + T FH cells (most of which are CXCR3 − ) predominate and are maintained longer after recovery in the observation period. 6,37,41,42The administration of the mRNA vaccine elicited high nAb levels and circulating T FH cell responses similar to those seen in COVID-19 convalescents. 11,43][46][47] These responses ensure antibody production to prevent reinfection or maintain the levels of circulating antibodies.8][49][50] Thus, addressing the longevity of antibody and T FH cell responses, as well as the functional role of T FH cells, in supporting the antibody response in COVID-19 convalescents and vaccinated subjects is urgently required to guide the development of long-term protective vaccines.
In this study, we performed a longitudinal investigation of the kinetics and longevity of spike-specific antibody and T FH cell subset responses in SARS-CoV-2 infection and vaccination and addressed the functional roles in supporting memory B cell differentiation into antibody-secreting cell (ASCs), as well as antibody production.Our findings provide new insights into the development of interventions and vaccines against SARS-CoV-2 and VOCs.

Persistence of spike-specific circulating T FH cell and subset responses in COVID-19 convalescents
To longitudinally assess circulating T FH cell and antibody responses after recovery from COVID-19, 104 blood samples were collected from 37 convalescents 2, 5, 8, 12, and 24 months after COVID-19 symptom onset (Fig. 1a and Supplementary Table 1).PBMCs were isolated and cultured for 24 h with stimulation in the presence of SARS-CoV-2 spike protein or BSA (5 μg/mL).Negative control PBMCs were collected from 14 healthy individuals before the COVID-19 pandemic and stimulated in the same manner.Circulating T FH (CXCR5 + CD4 + CD3 + T) cells were gated, and antigen-specific T FH cells were identified by CD154 (CD40L) assay (Supplementary Fig. 1).The results showed that the frequencies of CD154 + T FH cells at 2, 5, 8, and 12 months, but not at 24 months, were significantly higher in the spike-stimulated group than in the BSA group; there was no difference in frequencies between the stimulation group and BSA group in healthy PBMCs (Fig. 1b, c, upper panel).Longitudinal analysis revealed that the spike-specific T FH cell responses and the ratio of positive responders (stimulation index >2) declined gradually from months 2 to 24 (Fig. 1d, e, up panel).To explore the kinetics and longevity of T FH cell subset responses, stimulated PBMCs were gated into CXCR3 + and CXCR3 − T FH cell subsets, followed by analysis of CD154 expression (Supplementary Fig. 1).The results showed that the number of CD154 + CXCR3 + T FH cells from months 2 to 24 was significantly enhanced after spike stimulation compared with BSA stimulation, while spike-responsive CXCR3 − T FH cells were only seen from 2 to 8 months, and no significant changes were observed in CD154 + CXCR3 − T FH cell frequency from 12 to 24 months (Fig. 1c, middle and bottom panels).Because most of the COVID-19 patients included in our study had been vaccinated by 24 months, only five COVID-19 convalescents who remained unvaccinated were available for analysis at this time point.Therefore, we confirmed the finding at 24 months using an alternative assay (Supplementary Fig. 2).Longitudinally, the spike-specific CXCR3 + T FH cell responses and the ratio of positive responders declined from the 2nd to the 5th month but remained relatively steady from the 5th to the 24th month, while the number of spike-specific CXCR3 − T FH cells, similar to spike-specific T FH cells, declined slowly from months 2 to 24 (Fig. 1d, e, middle and bottom panel).Together, these findings demonstrate that spike-specific CXCR3 + and CXCR3 − T FH cells exhibit distinct kinetics, and that spike-specific CXCR3 + T FH cell responses may persist for more than 2 years in COVID-19 convalescents.

Dynamics of spike-specific antibody responses in COVID-19 convalescents and correlations with T FH cell frequency
To assess the kinetics of antibody responses in COVID-19 convalescents, we examined spike-specific antibodies in the plasma at 2, 5, 8, 12, and 24 months.The endpoint titers and avidity index of the spike-specific antibodies (immunoglobin A (IgA), IgG, and IgG subclasses (IgG1, IgG2, and IgG3)) were determined.Endpoint titers of spike-specific IgG, IgG1, IgG3, and IgA were detectable at each time point, but all declined significantly from 2 months to 5 months and then remained stable from 5 months to 24 months (Fig. 2a).Of note, spikespecific IgG2 was present at low or undetectable levels at each time point (data not shown).IgA antibody levels have been reported to decline rapidly, and most are short-lived. 51Here, IgA antibodies were found to persist for at least 24 months in symptomatic convalescents (Fig. 2a).In contrast, the avidity of the spike-specific IgG, IgG1, IgG3, and IgA antibodies increased over time with different kinetics from endpoint titers, indicating that the antibodies continued to mature during the convalescent phase (Fig. 2b).
Next, we examined the neutralization activity of a plasma antibody against SARS-CoV-2 spike pseudotyped virus.The neutralization titers decreased significantly from 2 months to 5 months and then gradually declined from 5 months to 24 months (Fig. 2c).Within the first 8 months, all individuals showed nAb titers above the cut-off value (≥30); however, the titers then declined, with 87.5% (21 out of 24) and 100% (5 out of 5) maintained until 12 and 24 months, respectively (Fig. 2c).These data revealed that the nAb level dropped in the early phase of recovery, while maintained at lower levels for a long term, which is consistent with the dynamics of the antibody endpoint titer (Fig. 2a).The neutralization titers were positively associated with the endpoint titers of IgG, IgG1, IgG3, and IgA from 2 months to 12 months; less significant correlations were found at 24 months, because this time point only included five samples (Supplementary Fig. 3).These results suggest that spike-specific IgG and IgA responses both persisted for at least 2 years in the vast majority of convalescents and contributed to the neutralization effect.
The frequencies of spike-specific T FH cells and their subsets, as well as their antibody titers, declined gradually in a similar pattern over the observation period (Figs. 1, 2).Thus, we further analyzed the correlations among spike-specific T FH cells or their subsets and antibody titers at each time point.The endpoint titers for IgG and IgA were positively correlated with the proportions of CD154 + CXCR3 + T FH cells but not CD154 + T FH cells and CD154 + CXCR3 − T FH cells at 5 months, while this correlation was not found at other time points (Fig. 2d, e and Supplementary Fig. 4), which indicated the well balance between T FH cell response and antibody response only observed at 5 months.Together, these results further demonstrate the potential role of spike-specific T FH cells in supporting high-affinity antibody maintenance in COVID-19 convalescents.
Longitudinal analysis of spike-specific T FH cell and antibody responses in inactivated vaccine recipients As described above, SARS-CoV-2 infection elicited persistent and highly correlated antibody and T FH cell responses, and these responses were maintained for up to 2 years.To assess the dynamics of nAb and T FH cell responses following vaccination, we recruited 26 participants who had completed the standard vaccination procedure involving two doses of inactivated vaccine (Sinovac) and collected blood samples at multiple time points to analyze the antibody titers and T FH cell responses (Fig. 3a and Supplementary Table 2).The nAb titer peaked 14 days after the second vaccination dose, and then rapidly decreased until 150 days after the second vaccination dose (Fig. 3b).Unlike the neutralization kinetics, the IgG avidity was low at day 14 and 28 after the first dose and day 14 after second dose and then plateaued from day 60 through day 150 after the second dose (Fig. 3c).Notably, the peak of the vaccine-derived IgG antibody avidity index was lower than that of natural infection with SARS-CoV-2 at 8 months (natural infection peak index versus vaccination index: median, IQR: 82.00%, 72.42-90.21%vs. 55.37%,49.93-61.15%).These findings suggest that two-dose immunization with inactivated vaccine elicited production of a lower level of nAbs that were not fully mature (low avidity) compared with natural infection with SARS-CoV-2.
Next, we assessed spike-specific T FH cell and subset responses to vaccination.The frequencies of spike-specific T FH cells and CXCR3 + T FH cells significantly increased upon spike protein stimulation at day 14, while the number of spike-specific CXCR3 − T FH cells increased at day 28 after the first dose, and the responses lasted throughout the observation period (Fig. 3d, left panel).Longitudinally, the responses of spike-specific T FH cells, CXCR3 + T FH cells, and CXCR3 − T FH cells increased compared with day 0 and peaked at 14 days after the second vaccination dose (Fig. 3e).Interestingly, the ratio of spike-specific CXCR3 + T FH cell positive responders increased dramatically and peaked at 14 days after the second vaccination dose (Fig. 3f).To explore the potential link between T FH cells and antibody responses in inactivated vaccine recipients, we analyzed the relationships between spike-specific T FH cells, CXCR3 + T FH cells, and CXCR3 − T FH cells and nAbs at each time point.We found frequencies of spike-specific CXCR3 + T FH cells before the second dose (28 days after the first dose) were positively correlated with the peak nAb titers 14 days after the second dose, while no such correlations between nAb titers and spike-specific T FH cells or CXCR3 − T FH cells were observed (Fig. 3g).This indicates that early primed T FH cells, especially spike-specific CXCR3 + T FH cells after first dose, contribute to antibody production in the 14 days after the second dose, highlighting a key role for spike-specific CXCR3 + T FH cells in the early activation of CD4 + T cells.These results suggest that two doses of inactivated vaccine can efficiently elicit nAb production and activate T FH cell responses.
A third booster dose augments the T FH cell response and promotes spike-specific antibody potency and affinity maturation Because the nAb produced in response to the standard two-dose vaccination regimen waned significantly after 6 months, a third booster dose was recommended.Previous studies have shown that COVID-19 vaccine boosters elevate antibody responses, but it remains unclear how the T FH cell response helps enhance the antibody response. 52,53To test the antibody and T FH cell responses before and after the third dose, we recruited 24 individuals who had received two vaccination doses at least 6 months later and collected PBMCs before and 14 days after the third booster dose (Sinovac) (Supplementary Table 2).Spike-specific T FH cells, CXCR3 + and CXCR3 − T FH cell subsets all responded to spike stimulation before and after the third dose (Fig. 4a); however, only the responses and responder ratio of spike-specific T FH cells and CXCR3 + T FH cells, but not spike-specific CXCR3 − T FH cells, expanded dramatically in response to the third dose (Fig. 4b, c).The third dose dramatically increased the neutralization titer by an average of more than 12-fold (Fig. 4d) and significantly promoted antibody affinity maturation (Fig. 4e), which reached a level similar to that induced by natural infection (natural infection peak index versus vaccination index: median, IQR: 82.00%, 72.42-90.21%vs. 86.11%,84.99-88.42%).Furthermore, the frequencies of spikespecific CXCR3 + T FH cells, but not spike-specific T FH and CXCR3 − T FH cells, after the third dose were significantly correlated with neutralization titers after the booster dose (Fig. 4f), indicating that the spike-specific T FH cells, especially spike-specific CXCR3 + T FH cells, elicited by the two-dose vaccination regimen further supported antibody production following the third dose.Thus, these findings show that a third vaccine dose augmented the antibody response and induced expansion of spike-specific T FH cells and their subsets, preferentially spike-specific CXCR3 + T FH cells, which may contribute to the enhancement of antibody quality and quantity.Of note, although no significant expansion in spike-specific CXCR3 − T FH cells was observed after the booster dose, these cells still remained at comparable levels with spikespecific CXCR3 + T FH cells, suggesting that they may co-contribute to antibody elevation.

Omicron subvariants show distinct escape patterns corresponding to vaccine-elicited specific T FH cell responses
Recent SARS-CoV-2 variants, especially the Omicron subvariants, have spread throughout the population, escaping vaccineelicited antibody responses and consequently caused breakthrough infections. 22,23Whether these subvariants also escape vaccine-elicited T FH cell responses is unclear.Here, we tested the vaccine-elicited T FH cell response to two representative Omicron subvariants, BA.1 and BA.4/5.PBMCs from 17 vaccinees who had received two or three doses of inactivated vaccine (14 days after the second or third booster dose) were stimulated with BSA or with spike protein from a prototype SARS-CoV-2 (Wuhan-Hu-1) strain and Omicron subvariants BA.1, or BA.4/5 (Supplementary Table 3).Similar to prototype spike protein stimulation, BA.1 spike protein stimulation also expanded significant specific T FH cell, CXCR3 + T FH cell, and CXCR3 − T FH cell responses compared with BSA stimulation; however, BA.4/5 spike protein stimulation only induced a significant specific CXCR3 + T FH cell response (Fig. 5a).Vaccine-elicited spike-specific T FH cells, CXCR3 + T FH cells, and CXCR3 − T FH cells showed a distinct recall response to prototype spike, BA.1, and BA.4/5 spike protein stimulation.Prototype and BA.1 spike protein stimulation elicited similar spike-specific T FH cell and subset responses, while BA.4/5 spike protein stimulation induced significantly lower vaccine-elicited T FH cell responses (Fig. 5b).Furthermore, the ratio of positive responders decreased slightly in response to BA.1 spike protein stimulation, but dropped dramatically in response to BA.4/5 spike protein stimulation, in comparison to prototype spike protein stimulation (Fig. 5c).These results  Mann-Whitney U test was used to analyze the differences between the indicated time points, *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.P < 0.05 was considered to be a two-tailed significant difference, ns, not significant.d-e Correlations of frequencies of spike-specific T FH cells, CXCR3 + T FH cells, and CXCR3 − T FH cells (data presented with background subtracted) with IgG and IgA endpoint titers at 5 months after illness onset (n = 17).Spearman's rank correlation coefficient was used to describe the association between the frequency of T FH cells and subsets with the IgG or IgA endpoint titers.P < 0.05 was considered to be a two-tailed significant difference suggest that vaccine-elicited T FH cell memory exhibits distinct response patterns to Omicron BA.1 and BA.4/5, even though multiple mutations have accumulated in both variant spikes.2][23] These findings suggest that the evolution of SARS-CoV-2 variants gradually enables them to escape the antibody and T FH cell responses elicited by previous infection or vaccination.
Spike-specific CXCR3 + T FH and CXCR3 − T FH cells exhibit distinct activation statuses and IL-21 secretion capacity We have shown that spike-specific CXCR3 + T FH cells and spikespecific CXCR3 − T FH cells exhibit distinct responsiveness and persistence in both convalescents and vaccinees (Figs. 1, 3, 4).To further assess the functional differences between the spikespecific T FH cell subsets, we compared the activation status of and secretion of IL-21 (a T FH cell signature cytokine) by T FH cells, especially CXCR3 + T FH cells and CXCR3 − T FH cells upon antigen exposure.
To investigate the activation status (HLA-DR, ICOS, and PD-1) of T FH cells, PBMCs from 16 COVID-19 convalescents, 17 vaccinees, and 19 healthy controls were stimulated with BSA or spike protein (Supplementary Table 4).The results showed that the proportions of HLA-DR + CD154 + , ICOS + CD154 + , and PD-1 + CD154 + cells among T FH cells, CXCR3 + T FH cells, and CXCR3 − T FH cells were significantly enhanced upon spike protein stimulation compared with BSA stimulation in both COVID-19 convalescents and vaccinees, except for the proportions of HLA-DR + CD154 + in T FH cells and CXCR3 − T FH cells, as well as ICOS + CD154 + cells in CXCR3 − T FH cells, from convalescents (Fig. 6a-c).There were no changes in the activation status of T FH cells or subsets in healthy controls upon spike protein stimulation (Fig. 6a-c).Furthermore, spike-specific CXCR3 + T FH cells were more highly activated than spike-specific CXCR3 − T FH cells in both COVID-19 convalescents and vaccinees upon spike protein stimulation, while there was no difference in healthy controls (Fig. 6d).These results suggest that SARS-CoV-2 infection and vaccination elicit an expansion of spikespecific T FH cells that can be reactivated upon antigen exposure, and that spike-specific CXCR3 + T FH cells are more highly activated than CXCR3 − T FH cells upon antigen exposure.
IL-21 secreted by T FH cells is critical in supporting B cell differentiation and antibody production. 25,54Therefore, to further explore the functional difference in spike-specific CXCR3 + and CXCR3 − T FH cells, PBMCs from 12 COVID-19 convalescents, 13 vaccinees, and 11 healthy controls (Supplementary Table 4) were stimulated with spike peptide pools.Intracellular IL-21 secretion was significantly increased in T FH cells, CXCR3 + T FH cells, and CXCR3 − T FH cells from both COVID-19 convalescents and vaccinees, but not from healthy controls, upon peptide stimulation (Fig. 6a-c, right panel).Additionally, we found that intracellular IL-21 secretion was prominent in convalescents, while it was relatively low in vaccinees (Fig. 6d, right panel).Intriguingly, the amount of IL-21 secreted by CXCR3 + T FH cells was significantly higher than that secreted by CXCR3 − T FH cells from both COVID-19 convalescents and vaccinees, but not from healthy controls, upon peptide stimulation (Fig. 6d, right panel).The higher activated status of and IL-21 secretion by CXCR3 + T FH cells compared with CXCR3 − T FH cells upon antigen exposure indicates that spikespecific CXCR3 + T FH cells may play a greater role in supporting B cell function than spike-specific CXCR3 − T FH cells in both COVID-19 convalescents and vaccinees.
39]55 To discriminate between the functional roles that these cells play in supporting ASC differentiation and antibody production, we cocultured T FH (CXCR5 + CD4 + CD3 + T) cells or non-T FH (CXCR5 − CD4 + CD3 + T) cells from healthy controls (n = 6), convalescents (n = 5), and vaccine recipients (n = 9) with autologous memory B cells (5 × 10 4 cells for each cell type) for 6 days in the presence of staphylococcal enterotoxin B (SEB) (Supplementary Table 5).Total ASCs (CD38 hi CD27 hi CD19 + B cells) and spike-specific ASCs were measured by FACS (Fig. 7a, b), and spike-specific IgG in the supernatant was measured by ELISA after coculturing (Fig. 7c-f, right panel).In healthy controls, T FH cells but not non-T FH cells efficiently supported autologous memory B cells differentiation into total ASCs, and, as expected, spike-specific ASCs and IgG were rarely observed (Fig. 7c).T FH cells from COVID-19 convalescents and vaccinees efficiently supported autologous memory B cells differentiation into both total ASCs and spikespecific ASCs and produced spike-specific IgG (Fig. 7d, e).Receiving a third booster dose of the vaccine (n = 7) further enhanced the humoral immune responses, as more spike-specific ASCs and IgG were produced than before receiving the booster dose (Fig. 7f).Moreover, the frequencies of spike-specific ASCs correlated with spike-specific IgG OD 450 (Supplementary Fig. 5).Taken together, these findings suggest that T FH cells, but not non-T FH cells, are the major player in supporting antibody recall response in COVID-19 convalescents and vaccinees upon antigen re-exposure.
As described above, we found that spike-specific CXCR3 + T FH cells were associated with antibody response magnitude and were more responsive than spike-specific CXCR3 − T FH cells upon antigen stimulation in both COVID-19 convalescents and vaccinees (Figs.1-6).However, the roles and functional differences between spike-specific CXCR3 + T FH cells and spikespecific CXCR3 − T FH cells in supporting ASC differentiation and antibody production remained undefined.To address this question, bulk CXCR3 + and CXCR3 − T FH cells were sorted from COVID-19 convalescents (n = 9), vaccinees (n = 7), and healthy ).Data are presented as the mean ± SEM, Mann-Whitney U test was used to analyze the differences between the indicated time points.For (d-e), *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.P < 0.05 was considered to be a two-tailed significant difference, ns, not significant.f Percentage of spike-specific T FH cell, CXCR3 + T FH cell, and CXCR3 − T FH cell responders (stimulation index > 2 was considered to be a positive response) before vaccination (day 0, n = 21) and after vaccination (day 14 after 1st dose, n = 22; day 28 after 1st dose, n = 23; day 14 after 2nd dose, n = 21; day 60 after 2nd dose, n = 22; day 150 after 2nd dose, n = 17).g Correlations among frequencies of spike-specific T FH cells, CXCR3 + T FH cells, and CXCR3 − T FH cells (data presented with background subtracted) on day 28 after 1st dose and nAb titers on day 14 after 2nd dose (n = 22).Spearman's rank correlation coefficient was used to describe the association between the frequency of T FH cells and subsets with the nAb titer.P < 0.05 was considered to be a two-tailed significant difference SARS-CoV-2 spike-specific T FH cells exhibit unique. . .He et al. controls (n = 8) and cocultured with autologous memory B cells (Fig. 7g, h and Supplementary Table 5), respectively.After coculture, B cell differentiation and spike-specific antibody production were assessed.The results showed that CXCR3 − T FH cells were more efficient than CXCR3 + T FH cells in supporting memory B cell differentiation into total ASCs in healthy controls, convalescents, and vaccinees (Fig. 7i-k, left panel), which is consistent with a previous report that circulating human CXCR3 − T FH cells are more efficient than CXCR3 + T FH cells in supporting B cell differentiation. 29However, CXCR3 + T FH cells showed superior  and (d, e), paired t test was used to analyze the differences between the two groups, *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.P < 0.05 was considered to be a two-tailed significant difference, ns, not significant.f Correlation of the frequencies of spikespecific T FH cells, CXCR3 + T FH cells, and CXCR3 − T FH cells (data presented with background subtracted) with neutralization titers after the third booster dose (n = 21).Spearman's rank correlation coefficient was used to describe the association between the frequency of T FH cells and subsets and the nAb titer.P < 0.05 was considered to be a two-tailed significant difference capacity to CXCR3 − T FH cells in supporting spike-specific ASCs differentiation in both COVID-19 convalescents and vaccinees (Fig. 7j, k, middle panel).No spike-specific ASCs differentiation or spike-specific antibody production was observed when CXCR3 +/− T FH cells were cocultured with autologous memory B cells from healthy controls (Fig. 7i).We did not observe higher spike-specific IgG production in CXCR3 + T FH cells compared with CXCR3 − T FH cells from COVID-19 convalescents and vaccinees cocultured with autologous memory B cells, as expected (Fig. 7j, k, right panel).This might be due to the lower ex vivo proportion of CD154 + CXCR3 + T FH cells within bulk CXCR3 + T FH cells compared with CD154 + CXCR3 − T FH cells within bulk CXCR3 − T FH cells (Supplementary Fig. 6), although spike-specific CXCR3 + T FH cells exhibited a more activated status, and greater IL-21 secretion than spike-specific CXCR3 − T FH cells in convalescents and vaccinees (Fig. 6).These findings further confirm that spikespecific T FH cell subsets, especially the spike-specific CXCR3 + T FH cell subset, play a functional role in supporting antibody maintenance and recall response in natural infection and vaccination.Targeting T FH cells, especially CXCR3 + T FH cells, may therefore be a novel approach to enhancing the long-term efficacy of SARS-CoV-2 vaccines. 56

DISCUSSION
In this study, we systemically investigated the longitudinal dynamics of spike-specific T FH cell and antibody responses in individuals up to 2 years after COVID-19 recovery and in individuals who received two or three doses of inactivated vaccine.We discovered that spike-specific CXCR3 + T FH cells are persistent and express higher levels of T FH functional molecules than spike-specific CXCR3 − T FH cells in convalescents and inactivated vaccine recipients.The spike-specific memory B cells generated by natural infection and vaccination was efficiently reactivated by spike-specific CXCR3 + T FH cells and CXCR3 − T FH cells to differentiate into ASCs and produce spike-specific antibodies.These findings suggest that SARS-CoV-2-specific long-lasting humoral immunity can be generated and recalled, thus providing new insight into COVID-19 immunity and how to elicit long-term protection.Previous studies have shown that SARS-CoV-2 infection induces robust spike-specific T H 1 and T FH cell responses in the acute phase that are maintained in the convalescent phase and persist for several months to a year. 41,42,57Here, in a longitudinal analysis of circulating T FH cells over a 2-year period in COVID-19 convalescents, we found that spike-specific CXCR3 + T FH cell response persisted for up to 24 months or more.In most symptomatic COVID-19 cases, robust GC reactions occur and are maintained for at least 6 months. 47,50GC T FH cells promote B cell maturation and differentiation into high-affinity memory B cells and long-lived plasma cells.The long-term persistence of circulating T FH cell responses in COVID-19 convalescents may be attributed to active GC reactions in lymph nodes after recovery.
In line with the T FH cells dynamics observed in this study, spike-specific IgG, IgG1, IgG3, and IgA antibody endpoint titers, as well as neutralization titers, showed a sharp decline from month 2 to month 5 and then were maintained at the same level in the majority of convalescents for at least 24 months, which is longer than the 16-month persistence reported recently. 20The initial sharp decrease was most likely due to the short half-life of serum antibodies and ASCs generated in the acute phase. 41ASCs normally decay within a few weeks, and only a smaller population of long-lived plasma cells can live from several months to years. 58IgA antibody has been shown to play a dominant role in neutralization in early infection, although it is short-lived and declines rapidly after infection. 51Here, we found that systematic spike-specific IgA antibodies were maintained for 24 months, although at a lower level.This is much longer than a previous observation of detectable IgA 6 months after infection, and may contribute to long-term protection. 59Antibody maturation is a relatively slow process that is supported by T FH cells in the GC. 24Our result show that spike-specific IgG antibody maturation peaked at 5 to 8 months after infection, while spike-specific IgG3 and IgA were fully mature at the 12 months.This result is in line with an earlier observation in recovered SARS patients that ****P < 0.0001.P < 0.05 was considered to be a two-tailed significant difference; ns, not significant antibody avidity continued to mature from low avidity in the early phase to high avidity at late phases of disease recovery. 60he maturation of antibody avidity also reflects the persistence of GC reactions after SARS-CoV-2 resolution. 50These findings indicate that long-term T FH cell responses are required in COVID-19 convalescents to gradually support high-quality antibody maturation and maintenance.
Inactivated vaccines have been widely used and proven to elicit short-term humoral immunity when administered in two doses, thereby conferring protection from SARS-CoV-2 and VOCs infection or severe disease. 61,62In this study, we also found that immunization with inactivated vaccine elicited humoral immunity in a similar manner to natural infection, but that the magnitudes of the antibody and T FH cell responses were significantly smaller.T FH cells with autologous memory B cells from healthy controls, convalescents, and vaccinees in the presence of SEB (100 ng/mL) for 6 days.i-k Comparison of frequencies of total ASCs, spike-specific ASCs, and spike-specific IgG in the supernatant between CXCR3 + T FH cells and CXCR3 − T FH cells after coculturing with autologous memory B cells from healthy controls (n = 8), convalescents (n = 9), and vaccinees (n = 7) in the presence of SEB (100 ng/mL) for 6 days.For (c-f) and (i-k), paired t test was used to analyze the differences between the two groups.*P < 0.05; **P < 0.01.P < 0.05 was considered to be a two-tailed significant difference; ns, not significant Although nAb levels peaked at 14 days mature antibodies were observed 2 months after the second dose, spike-specific T FH cells were detected 14 days after the first dose, earlier than the appearance of nAbs.However, the nAb level significantly declined 6 months after immunization.Based on the memory immunity generated by two-dose vaccination, administering a third booster dose significantly magnified the nAb responses and increased antibody avidity and the frequency of responsive T FH cells.These results are consistent with several recent studies of inactivated vaccine showing that a third dose increased the antibodyneutralizing effect for SARS-CoV-2 and some VOCs, proving the utility and necessity of the booster dose. 62,63However, the T FH cell and antibody responses elicited by inactivated vaccines seem to last only a short period of time, unlike natural infections, which elicit a lasting and relatively stable immune memory response. 572][23] Thus, systematically and continuously investigating the efficacy of vaccine-elicited T cell and antibody responses to emerging or pre-emergent variants is critical to inform vaccine improvement.
The long-lived antibody response is mainly supported by T FH cells in natural infection and vaccination. 47,64We found that SARS-CoV-2 infection and vaccination induced a persistent spikespecific T FH cell responses, consistent with other studies, thus laying the foundation for long-term immunity. 61,62Notably, spikespecific T FH cells, especially spike-specific CXCR3 + T FH cells, positively correlated with antibody responses at 5 months in convalescents, and the quantity and quality of the antibodies were well balanced.Spike-specific CXCR3 + T FH cells elicited by the first vaccine dose also positively correlated with peak nAb titers, and this relationship was further strengthened by a third booster dose.Functionally, T FH cells, but non-T FH cells, from COVID-19 convalescents and vaccinees could efficiently recall autologous memory B cells to differentiate into spike-specific ASCs and produce antibodies.Therefore, circulating spike-specific T FH cells are a surrogate of bona fide GC T FH cells that support spikespecific ASC differentiation and antibody production in SARS-CoV-2 natural infection and vaccination. 29Of note, our study did not exclude the possibility that non-T FH cells support short-lived plasmablast differentiation and produce low-affinity antibodies in the very early acute phase to constrain infection rapidly, as large amounts of antibodies are produced by extrafollicular B cells in some severe cases. 65ARS-CoV-2 infection-induced circulating T FH cells exhibited a clear phenotypic bias towards a CCR6 + CXCR3 − phenotype, and this subset comprises the majority of spike-specific T FH cells in natural infection. 6,37,41,428][39] The association of T H 1-like (CXCR3 + ) T FH cells with antibody quantity and quality has also been characterized in influenza vaccine recipients and in other chronic viral infections. 30,31,33,34,66Furthermore, we previously demonstrated that the CXCR3 + and CXCR3 − T FH subsets have distinct phenotypes and functions in HCV infection. 34However, the functional difference of spike-specific CXCR3 + T FH and CXCR3 − T FH cells in SARS-CoV-2 infection and vaccination is largely unknown.In this study, we demonstrated that spike-specific CXCR3 + T FH cells are more persistent than spike-specific CXCR3 − T FH cells in COVID-19 convalescents.And, spike-specific CXCR3 + T FH cells exhibited higher activation status and greater IL-21 secretion than spike-specific CXCR3 − T FH cells in both COVID-19 convalescents and vaccinees upon antigen exposure.Coculture of CXCR3 + T FH or CXCR3 − T FH cells with autologous memory B cells from COVID-19 convalescents and vaccinees showed that CXCR3 + T FH cells were more able than CXCR3 − T FH cells to support spikespecific memory B cell differentiation into ASCs.These findings confirm that both spike-specific CXCR3 + T FH cells and CXCR3 − T FH cells play an efficient, functional role in promoting the antibody response in natural infection and vaccination.4,67 Given the importance of T H 1-like T FH cells in supporting the production of high-quality antibodies and the longevity of ASCs, strategies to promote T H 1-like (CXCR3 + ) T FH cell polarization would benefit SARS-CoV-2 vaccine development.
In conclusion, our study demonstrated that SARS-CoV-2 natural infection and vaccination elicit expansion of spike-specific CXCR3 + and CXCR3 − T FH cell subsets, and that both subsets contribute to spike-specific high-affinity antibody maintenance and recall responses, but that they exhibited clear differences in persistence, functionality, and antibody-supporting ability.Spikespecific CXCR3 + T FH cells are more persistent and superior to CXCR3 − T FH cells in supporting antibody production, and may therefore confer long-term protection.These findings will inform vaccine design to provide long-term protection against SARS-CoV-2 and VOC infection by targeting T FH cells, especially CXCR3 + T FH cells.

Limitations of the study
This study had some limitations: a single-marker (CD154) assay was used to identify antigen-specific T FH cells, only five convalescent samples from the 24-month time point were available for analysis, and a limited number of vaccination cohort samples and Omicron subvariants were tested for immune escape capacity.The findings from the co-culture assay should be confirmed using spike protein or peptide pools.

Study subjects
Two cohorts were included in this study: COVID-19 convalescents and vaccinated subjects.A total of 78 COVID-19 convalescents who had recovered from Wuhan-Hu-1 infection from January to March 2020 were recruited from The Central Hospital of Shaoyang, Hunan Province, China, and a subset of these individuals were followed for up to 24 months after symptom onset (Supplementary Tables 1, 4, and 5).In addition, 95 vaccinated subjects were recruited from The First People's Hospital of Chenzhou, Hunan Province, China.All vaccinated subjects received at least two doses of inactivated vaccine (Sinovac, Beijing, China), while a subset of these subjects also received a third dose (Supplementary Tables 2-5).None of the enrolled COVID-19 convalescents had a history of SARS-CoV-2 vaccination, and none of the convalescents or vaccinees had subsequent SARS-CoV-2 or variant exposure or infection, as determined by frequent PCR testing and questionnaire administration during the observation period.For convalescent individuals, blood samples were collected 2, 5, 8, 12, and 24 months after COVID-19 symptom onset.For vaccine recipients, blood samples were collected before vaccination (day 0), 14 and 28 days after the first dose, and 14, 60, and 150 days after the second dose.For vaccinees who received a booster, blood samples were also taken before and 14 days after the third dose.Blood samples taken from healthy individuals before the COVID-19 pandemic were used as the negative control group.PBMCs and plasma were isolated and stored in liquid nitrogen and a −80 °C freezer, respectively.Each participant signed a written consent form.The study protocol was approved by the Institutional Ethical Review Board of The Central Hospital of Shaoyang (V.1.0,20200301) and the First People's Hospital of Chenzhou (V.3.0,2021001), and the study was compliant with the "Guidance of the Ministry of and Technology (MOST), China".

Antibody endpoint titers
The endpoint titers of spike-specific antibodies were determined by measuring the binding activity of serially diluted plasma to the SARS-CoV-2 spike protein by ELISA.In brief, 96-well plates (Corning, NY, USA) were coated with SARS-CoV-2 spike protein (SARS-CoV-2 S1 + S2_ECD, 200 ng/well) (Sino Biological, Beijing, China) in PBS and incubated at 4 °C overnight.The plates were washed five times with PBS-T (0.05% Tween-20 in PBS) and then blocked with blocking buffer (2% FBS and 2% BSA in PBS-T) for 30 min.Two-fold serial dilutions of plasma, starting from a 1:20 dilution, were added to the 96-well plates in triplicate (100 µL/ well) and incubated for 1 h at room temperature.Spike-specific antibodies were detected using horseradish peroxidase (HRP)conjugated anti-human IgG (Jackson ImmunoResearch, PA, USA), IgG1, IgG3, and IgA (BaiaoTong Experiment Centre, Luoyang, China).Plasma samples collected from healthy subjects before the COVID-19 pandemic were used as the negative control group, and SARS-CoV-2 spike RBD-specific monoclonal antibody was generated in the laboratory and used as a positive control.Optical density at 450 nm (OD 450 ) was measured for each reaction, and an OD 450 value three-fold greater than the cut-off value (healthy control group) was considered a positive readout.The highest dilution showing a positive readout was defined as the endpoint titer of the antibody, and the data were logarithmically transformed.

Antibody avidity assay
The avidity of spike-specific antibodies (IgG, IgG1, IgG3, and IgA) was measured using a modified two-step approach that we described previously. 39In the first step, plasma dilutions were optimized to obtain an OD 450 value within the range of 0.5-1.5 to ensure linear measurement of antibody avidity.The second step was an ELISA that included elution with 1 M NaSCN.These measurements were performed in triplicate.The avidity index of an antibody was calculated as OD NaSCN 1M /OD NaSCN 0M × 100%.

Antibody neutralization assay
The neutralization activity of plasma was determined by reduction in luciferase expression after infecting Huh7 cells with pseudotyped virus, as described previously. 39In brief, SARS-CoV-2 pseudotyped virus was incubated in duplicate with serial dilutions of plasma samples (six dilutions: 1:30; 1:90; 1:270; 1:810; 1:2430; 1:7290) at 37 °C for 1 h.Then, freshly trypsinized cells were added, the mixture was incubated at 37 °C with 5% CO 2 for 24 h, and the luminescence was measured.In parallel, control wells with virus only or cells only were included (six replicates).The background value in relative light units (RLUs) (wells with cells only) was subtracted from each value.Plasma from healthy controls was used as a negative control.Plasma from guinea pigs immunized with the SARS-CoV-2 spike protein was used as a positive control.The 50% inhibitory dilution (ID 50 ) was defined as the plasma dilution that reduced the value in RLUs by 50% compared with the control wells (virus + cells).The cut-off value was defined as ID 50 = 30, and samples with an ID 50 > 30 were considered to have a neutralizing effect.

Fig. 2
Fig. 2 Kinetics of spike-specific antibody responses in COVID-19 convalescents.a Endpoint titers and (b) avidity index of plasma spike-specific IgG, IgG1, IgG3, and IgA antibodies among COVID-19 convalescents at 2 months (n = 25), 5 months (n = 25), 8 months (n = 25), 12 months (n = 24), and 24 months (n = 5) after illness onset.The endpoint titer data were logarithmically transformed.c Neutralization titers of COVID-19 convalescent plasma against SARS-CoV-2 pseudotyped virus at the indicated time points.The cut-off value was defined as ID 50 = 30, and samples with an ID 50 > 30 were considered to have a neutralizing effect.For (a-c), data are presented as median ± IQR (25-75%).Mann-Whitney U test was used to analyze the differences between the indicated time points, *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.P < 0.05 was considered to be a two-tailed significant difference, ns, not significant.d-e Correlations of frequencies of spike-specific T FH cells, CXCR3 + T FH cells, and CXCR3 − T FH cells (data presented with background subtracted) with IgG and IgA endpoint titers at 5 months after illness onset (n = 17).Spearman's rank correlation coefficient was used to describe the association between the frequency of T FH cells and subsets with the IgG or IgA endpoint titers.P < 0.05 was considered to be a two-tailed significant difference

Fig. 3
Fig.3Kinetics of spike-specific antibody and circulating T FH cell responses in inactivated vaccine recipients.a Timeline of vaccination and blood sampling.b nAb titers against pseudotyped SARS-CoV-2 spike virus at the indicated time points (day 0, n = 26; day 14 after 1st dose, n = 26; day 28 after 1st dose, n = 26; day 14 after 2nd dose, n = 26; day 60 after 2nd dose, n = 26; day 150 after 2nd dose, n = 26).Data are presented as the mean ± SEM. c The kinetics of spike-specific IgG avidity at the indicated time points (day 14 after 1st dose, n = 26; day 28 after 1st dose, n = 26; day 14 after 2nd dose, n = 26; day 60 after 2nd dose, n = 25; day 150 after 2nd dose, n = 25).Data are presented as the mean ± SEM. d Frequency of spike-specific T FH cells, CXCR3 + T FH cells, and CXCR3 − T FH cells before vaccination (day 0, n = 21) and after vaccination (day 14 after 1st dose, n = 22; day 28 after 1st dose, n = 23; day 14 after 2nd dose, n = 21; day 60 after 2nd dose, n = 22; day 150 after 2nd dose, n = 17), upon stimulation.Paired t test was used to analyze the differences in T FH cell and subset responses between BSA or spike protein stimulation at the indicated time points.e Kinetics of spike-specific T FH cell, CXCR3 + T FH cell, and CXCR3 − T FH cell responses (data presented with background subtracted; responses below background are shown as 0.001%) before vaccination (day 0, n = 21) and after vaccination (day 14 after 1st dose, n = 22; day 28 after 1st dose, n = 23; day 14 after 2nd dose, n = 21; day 60 after 2nd dose, n = 22; day 150 after 2nd dose, n = 17).Data are presented as the mean ± SEM, Mann-Whitney U test was used to analyze the differences between the indicated time points.For (d-e), *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.P < 0.05 was considered to be a two-tailed significant difference, ns, not significant.f Percentage of spike-specific T FH cell, CXCR3 + T FH cell, and CXCR3 − T FH cell responders (stimulation index > 2 was considered to be a positive response) before vaccination (day 0, n = 21) and after vaccination (day 14 after 1st dose, n = 22; day 28 after 1st dose, n = 23; day 14 after 2nd dose, n = 21; day 60 after 2nd dose, n = 22; day 150 after 2nd dose, n = 17).g Correlations among frequencies of spike-specific T FH cells, CXCR3 + T FH cells, and CXCR3 − T FH cells (data presented with background subtracted) on day 28 after 1st dose and nAb titers on day 14 after 2nd dose (n = 22).Spearman's rank correlation coefficient was used to describe the association between the frequency of T FH cells and subsets with the nAb titer.P < 0.05 was considered to be a two-tailed significant difference

Fig. 4 A
Fig. 4 A third vaccine dose promoted spike-specific antibody potency and maturation.a Frequencies of spike-specific T FH cells, CXCR3 + T FH cells, and CXCR3 − T FH cells before and after the third booster dose upon BSA or spike protein stimulation (n = 21).b Comparison of frequencies of spike-specific T FH cells, CXCR3 + T FH cells, and CXCR3 − T FH cells (data presented with background subtracted; responses below background are shown as 0.001%) before and after the third booster dose (n = 21), upon spike protein stimulation.c Percentage of spikespecific T FH cell, CXCR3 + T FH cell, and CXCR3 − T FH cell responders (stimulation index > 2 was considered to be a positive response) before and after the third booster dose (n = 21).d Neutralization titers before and after the third booster dose (n = 24).The cut-off value was defined as ID 50 = 30, and ID 50 > 30 was considered to indicate a neutralizing effect.e Spike-specific IgG avidity index before and after the third booster dose (n = 24).For (a, b),and (d, e), paired t test was used to analyze the differences between the two groups, *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.P < 0.05 was considered to be a two-tailed significant difference, ns, not significant.f Correlation of the frequencies of spikespecific T FH cells, CXCR3 + T FH cells, and CXCR3 − T FH cells (data presented with background subtracted) with neutralization titers after the third booster dose (n = 21).Spearman's rank correlation coefficient was used to describe the association between the frequency of T FH cells and subsets and the nAb titer.P < 0.05 was considered to be a two-tailed significant difference

Fig. 7
Fig. 7 Spike-specific CXCR3 + T FH cells were more able than spike-specific CXCR3 − T FH cells to support ASCs differentiation in vitro.a Diagram of T FH cells or non-T FH cells coculturing with autologous memory B cells.b Representative flow plots of spike-specific ASCs after coculturing T FH cells or non-T FH cells with autologous memory B cells from healthy controls, convalescents, and vaccinees in the presence of SEB (100 ng/ mL) for 6 days.c-e Comparison of total ASCs, spike-specific ASCs, and spike-specific IgG in the supernatant between T FH and non-T FH cells after coculture with autologous memory B cells from healthy controls (n = 6), convalescents (n = 5), and vaccinees (n = 9).f Comparison of the total ASCs, spike-specific ASCs, and spike-specific IgG from before (n = 7) and after (n = 7) the third booster dose.g Diagram of CXCR3 + T FH cells or CXCR3 − T FH cells coculturing with autologous memory B cells.h Representative flow plots of spike-specific ASCs after coculturing of CXCR3 + or CXCR3 − T FH cells with autologous memory B cells from healthy controls, convalescents, and vaccinees in the presence of SEB (100 ng/mL) for 6 days.i-k Comparison of frequencies of total ASCs, spike-specific ASCs, and spike-specific IgG in the supernatant between CXCR3 + T FH cells and CXCR3 − T FH cells after coculturing with autologous memory B cells from healthy controls (n = 8), convalescents (n = 9), and vaccinees (n = 7) in the presence of SEB (100 ng/mL) for 6 days.For(c-f) and (i-k), paired t test was used to analyze the differences between the two groups.*P < 0.05; **P < 0.01.P < 0.05 was considered to be a two-tailed significant difference; ns, not significant