Airway-resident T cells from unexposed individuals cross-recognize SARS-CoV-2

To examine for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) cross-reactive T cells in the lower respiratory tract airways, we used cryopreserved BAL samples taken during bronchoscopy of ten healthy donors in 2016–2018, before the onset of the pandemic. Cryopreserved peripheral blood mononuclear cells (PBMCs) from paired blood samples were also analyzed. Donors had been challenged with Streptococcus pneumoniae and influenza vaccine 6–19 weeks before bronchoscopy. Oropharyngeal swabs or nasosorption samples used for PCR screening of multiple respiratory viruses were negative in all subjects, apart from rhinovirus in one (patient cohort details in Methods and Supplementary Table 1). Mononuclear cells from BAL were stimulated with pools of SARS-CoV-2 peptides of specificities previously associated with protective pre-existing T cells1,4: three pools of overlapping peptides spanning the core replication transcription complex (RTC) nonstructural proteins (NSPs) NSP7, NSP12 and NSP13 from ORF1ab and a pool of predicted epitopes from the structural protein spike. After overnight peptide stimulation, antigen-specific T cells with antiviral functional potential were enumerated by intracellular cytokine staining for interferon (IFN)-γ and/or tumor necrosis factor (TNF) (CD8+ and CD4+) and CD40L (CD4+), after subtracting the background effector function seen in the donor-matched unstimulated control well (<0.6% background cytokine production in all cases, gating strategy, example FACS plots and all background values in Extended Data Fig. 1a,b).

SARS-CoV-2 peptide-reactive CD4+ and CD8+ T cells were detected in six of ten of the pre-pandemic BAL samples (Fig. 1a,b and Extended Data Fig. 1c). Responses were detectable against each of the RTC regions, including the highly conserved RNA polymerase (NSP12)5,6 that we found to be most strongly associated with abortive infection, as well as spike. SARS-CoV-2 cross-reactive CD4+ T cells were present in BAL at higher frequencies than CD8+ T cells (Fig. 1c), as noted in previous studies of the periphery1,7,8,9,10. However, CD4+ and CD8+ T cell responses were correlated in frequency within the same donor BALs (Extended Data Fig. 1d), indicating a coordinated response. Lung CD4+ and CD8+ T cells responded to SARS-CoV-2 peptides with induction of TNF and IFN-γ, with TNF being more abundantly produced than IFN-γ (Fig. 1d). Co-staining showed that some responding cells were multifunctional, with most IFN-γ-producing CD4+ and CD8+ T cells also producing TNF and a proportion of IFN-γ- and TNF-producing CD4+ T cells co-expressing CD40L (Extended Data Fig. 1e–h). The proportion of CD4+ T cells producing TNF was particularly high in some donors, but they were proportional to IFN-γ+CD4+ T cell and multifunctional (TNF/IFN-γ or TNF/CD40L CD4+ T cell) responses (Extended Data Fig. 1h and Fig. 1e), supporting antigen-specific triggering, perhaps amplified by a bystander TNF response11.

Fig. 1: Crossreactive SARS-CoV-2-specific T cells are present in pre-pandemic BAL samples.figure 1

a,b, Frequency of TNF- (a) and IFN-γ-producing (b) CD4+ and CD8+ T cells in BAL for each peptide pool. Doughnut plots at the top show the percentage of detectable responses. c, Percentage of TNF- or IFN-γ-producing CD4+ or CD8+ T cells in BAL. d, Frequencies of Sars-CoV-2-specific CD4+ and CD8+ T cells producing TNF or IFN-γ. e, Correlation of TNF+ CD4 T cells versus TNF+IFN-γ+ (left) and CD40L+TNF+ (right) CD4 T cells in BAL (n = 10 biologically independent samples examined over one independent experiment). Bars at median (ad), with gray area representing mean + 2 s.d. of DMSO control; Kruskal–Wallis one-way ANOVA test and Dunn’s multiple comparison (a and b); Wilcoxon’s paired test (c and d); Spearman’s correlation (e).

Source data

BAL samples from the lower respiratory tract would be expected to contain some tissue-resident memory T cells (TRM) that are specialized for long-lived sentinel pathogen defence12; to examine this we compared the residency profile (CD69/CD103 expression) of global T cells in BAL with paired PBMCs from the same donors. The t-distributed stochastic neighbor embedding (t-SNE) analysis confirmed differential segregation of cells within PBMCs and BAL expressing residency markers, particularly for CD69+CD103+ co-expression in the global CD8+ pool and single-positive CD69+CD103− in the CD4+ compartment (Fig. 2a,b). Among PBMCs, only 7.6% of global CD4+ T cells expressed CD69 compared with 66.7% in BAL (Fig. 2c,d). As previously reported, only a small population of airway CD69+CD4+ T cells co-expressed CD103 (ref. 13). Circulating CD8 +T cells were predominantly CD69−CD103−, whereas CD8+ T cells with the residency phenotypes CD69+CD103− and CD69+CD103+ dominated in BAL samples (Fig. 2c,d).

Fig. 2: Enrichment of resident phenotype in global and antigen-specific T cells in BAL.figure 2

a, The t-SNE plots of concatenated PBMCs and BAL CD4+ (top) or CD8+ T cells (bottom) highlighting sample tissue origin. b, The t-SNE plots of concatenated PBMCs and BAL CD4+ (top) or CD8+ T cells (bottom) highlighting expression of CD69 and CD103. c, Example plot and percentage of CD69 expression on CD4+ T cells from PBMCs or BAL. d, Example plot and percentage of CD103 and CD69 expression on CD8+ T cells from PBMCs or BAL. e, Example plots of TNF- or IFN-γ-producing CD4+ T cells (top) and CD69 expression on total, TNF+ or IFN-γ+CD4+ T cells (bottom) using NSP12 peptide pool. f, Example plots of TNF- or IFN-γ-producing CD8+ T cells (top) and CD103 versus CD69 expression on total, TNF+ or IFN-γ+ CD8+ T cells (bottom) using NSP12 peptide pool. g, Comparison of CD4+CD69+ and CD69− cells in BAL producing TNF (left) or IFN-γ (right) per peptide pool. h, Comparison of CD8+CD69+CD103+, CD69+CD103− and CD69− cells in BAL producing TNF (left) or IFN-γ (right) per peptide pool. ah, Four biologically independent samples with the highest frequencies of IFN-γ production examined over one independent experiment. Bars at median (c, d, g and h); floating bars indicating the mean, minimum and maximum values within the dataset (e and f); Kruskal–Wallis one-way ANOVA test and Dunn’s multiple comparison (cf and h); Wilcoxon’s paired test (g).

Source data

In four BAL donors we detected virus-specific T cells of sufficient magnitude to assess their residency phenotype; SARS-CoV-2 RTC and spike-specific CD4+ T cells tended to have even higher frequencies of CD69 expression than global BAL T cells, especially among the IFN-γ+ population, where ~90% expressed CD69 (Fig. 2e and Extended Data Fig. 2a). Similarly, a high proportion of SARS-CoV-2-specific CD8+ T cells had a residency signature (mean 70.6% CD69+CD103+, 97% CD69+; Fig. 2f). In this limited sample size, the expression of residency markers was comparable across all antigen specificities examined (Extended Data Fig. 2a,b). Overall, the frequencies of SARS-CoV-2-specific T cells were significantly enriched within the CD4+ and CD8+ TRM cell subsets compared with the non-TRM cell pools in the airways (Fig. 2g,h and Extended Data Fig. 2a,b).

Tissue-residency provides a mechanism to retain an enriched frequency of virus-specific memory T cells at the site of antigen encounter. Having identified SARS-CoV-2 RTC and spike-cross-reactive T cells with a TRM cell phenotype in the lower airways, we postulated that they would be selectively enriched at this site compared with the circulation. To examine whether pre-existing SARS-CoV-2-cross-reactive T cells were selectively enriched at the site of infection, we compared their frequencies after overnight stimulation of paired blood and BAL samples from the same donors. The percentage of CD4+ and CD8+ T cells producing TNF or IFN-γ (Fig. 3a,b) in response to the four SARS-CoV-2 peptide pools tested was substantially higher within BAL than PBMCs. CD4+ and CD8+ T cells tended to have broader specificity (responding to more peptide pools) in BAL compared with PBMCs (Fig. 3c,d). In addition, a higher proportion of donors had a detectable response to each peptide pool in their airway than in their blood sample (Fig. 3e,f). Overnight stimulation did not result in high expression of CD69 among peptide-responsive populations in the periphery (Extended Data Fig. 3a,b). This contrasted with the high CD69 expression of peptide-specific T cells described above in the airway, reinforcing the latter being a feature of TRM cells rather than simply a result of peptide activation.

Fig. 3: SARS-CoV-2-specific T cells are enriched in the lungs.figure 3

a,b, Frequency of TNF- (a) or IFN-γ-producing (b) CD4+ and CD8+ T cells after overnight stimulation of BAL and PBMCs with Sars-CoV-2 peptide pools shown as paired samples for the different peptide pools separately (left box) or all combined (right box). c,d, Number of viral antigens recognized ranked by response level in TNF- (c) or IFN-γ-producing (d) CD4+ and CD8+ T cells. e,f, Percentage of responders in BAL or PBMC samples for TNF+CD4+ or CD8+ T cells (e) and IFN-γ+CD4+ or CD8+ T cells (f). Responses were considered positive when above mean + 2 s.d. of DMSO control after overnight stimulation with SARS-CoV-2 peptide pools. g,h, Percentage of TNF (g) or IFN-γ (h) production by CD4+ or CD8+ T cells after overnight stimulation of PBMCs distributed according to positive detection of responses in BAL (n = 10 biologically independent samples examined over one independent experiment). Bars at median (a, b, g and h); Wilcoxon’s paired test (a and b); Mann–Whitney U-test (g and h).

Source data

Next we examined whether circulating SARS-COV-2-reactive T cells identified after either overnight stimulation or in vitro expansion correlated with BAL responses. RTC and spike-cross-reactive T cells were detectable in more PBMC samples after short-term (10 d) in vitro expansion than overnight, consistent with these low-magnitude responses being below the threshold of detection rather than completely absent from the circulation (Extended Data Fig. 3c,d). Generally, overnight or in vitro expanded SARS-CoV-2-reactive T cell frequencies did not correlate significantly with those in matched BAL (Extended Data Fig. 3e–h), which may reflect the small sample size, but suggests that the magnitude of circulating T cells is not fully representative of the pool compartmentalized within the lung. Notably, the only responses to significantly correlate between blood and BAL were overnight IFN-γ+CD4+ T cells specific for RNA polymerase (NSP12; Extended Data Fig. 3f), the region most highly conserved across human coronaviruses, supporting our previous association of this specificity in the circulation with protection from overt infection in healthcare workers1. Moreover, overnight peptide stimulation of PBMCs revealed significantly increased CD4+ and a trend to higher frequencies of CD8+ T cell responses in individuals with a detectable BAL response to SARS-CoV-2 (Fig. 3g,h).

Finally, we investigated the postulate that seasonal human coronaviruses (HCoVs) represent one probable stimulus for pre-existing SARS-CoV-2-reactive airway T cells. As insufficient BAL remained, we used residual paired PBMCs and sera to test for pre-existing T cell and humoral immunity to seasonal HCoVs. We have shown that 15-mer peptides spanning the nonstructural RTC region are highly conserved between HCoV and SARS-CoV-2 (ref. 1). To assess the crossreactivity of the pre-existing SARS-CoV-2-reactive T cells to HCoVs, we therefore focused on spike, constructing a mapped epitope pool of peptides from the HCoV spike (OC43, 229E and NL63; BEI Resources, peptide arrays, 17-mers), equivalent to the pool we had used to assess SARS-CoV-2 spike-reactive T cells. Both overnight and in vitro expanded CD4+ and CD8+ T cell responses to the seasonal HCoV spike tended to be higher frequency (significantly so for IFN-γ+CD8+ T cells) in those pre-pandemic donors with SARS-CoV-2-reactive T cells in their BAL (Fig. 4a,b). Immunoglobulin (Ig)G antibody titers against HCoVs were also higher in donors with SARS-CoV-2-reactive BAL T cells (significant for 229E, trends for NL63, OC43, HKU1; Fig. 4c), corroborating the evidence for more recent/stronger HCoV exposure being associated with cross-reactive airway T cells. In contrast to the cross-reactive T cells we had observed, no antibodies able to crossreact with SARS-CoV-2 spike or nucleoprotein were detectable in these pre-pandemic sera (Fig. 4d).

Fig. 4: HCoV spike-specific T cell and antibody responses are higher in individuals with detectable Sars-CoV-2-reactive T cell responses in BAL.figure 4

a,b, Frequency of TNF- (a) or IFN-γ-producing (b) CD4+ and CD8+ T cells after overnight and 10-d stimulation of PBMCs with spike peptide pool from HCoVs 229E, NL63 and OC43 combined. c, Serum IgG to spike from HCoVs 229E, NL63, OC43 and HKU1 in arbitrary units (n = 10 (a and b) or n = 9 (c and d) biologically independent samples examined over one independent experiment). d, Serum IgG to Sars-CoV-2 spike and nucleoprotein in arbitrary units. a,b, Results are represented grouped as individuals who tested negative (BAL nonresponders) or positive (BAL responders) for Sars-CoV-2-specific T cells in BAL. Bars at median (a and b), floating bars indicating the mean, minimum and maximum values within the dataset (c and d); Mann–Whitney U-test (a and d).

Source data

A potential confounder for this cohort is the influenza vaccination and pneumococcal challenge that donors had received before BAL sampling, although these are ‘real-world’ physiological exposures, with >10% of the general population being pneumococcal carriers14 and many receiving annual influenza vaccines. In most cases these had been administered >4 months before BAL; the two cases with a 6-week interval were not outliers in their SARS-CoV-2 reactivity (Extended Data Fig. 4a). We have previously noted that pneumococcal and influenza-specific memory responses account for ~1% of BAL T cells15 (Extended Data Fig. 4b), whereas spontaneous T cell TNF release decays over time after pneumococcal challenge to become minimal >1 month post-challenge (Extended Data Fig. 4c), in line with the low levels of background cytokine production seen without peptide stimulation in this cohort. However, it remains possible that previous inflammatory exposures such as these increased the recruitment of SARS-CoV-2-reactive lung TRM cell specificities16. Previous challenge could also have activated global lung T cells, but BAL from a donor without pneumococcal challenge or influenza vaccine still had high expression of CD69 on CD4+ T cells, many of which co-expressed the additional TRM cell marker CD49a (Extended Data Fig. 4d).

We demonstrate that SARS-CoV-2-cross-reactive T cells can reside in the human airways, a key site for frontline protection against inhaled pathogens such as coronaviruses. SARS-CoV-2-reactive T cells were more frequent in the CD4+ than the CD8+ T cell fraction of pre-pandemic lower respiratory mucosa, but both were capable of co-producing IFN-γ and TNF. Although this observational study could not examine their protective role, we speculate that airway memory T cells capable of such rapid, robust cytokine production would provide vital sentinel function, based on in vivo depletion studies in animal models. CD4+ T cell responses localized within the airway, rather than those in the lung parenchyma or vasculature, have been shown to be critical for protection against SARS-CoV; depletion of the airway fraction specifically abrogated the efficacy of mucosal vaccination3. Airway coronavirus-specific CD4+ T cells have been shown to mediate rapid IFN-γ-dependent induction of antiviral pathways such as STAT-1 (ref. 3), consistent with our previous observation of healthcare workers aborting infection before PCR or antibody positivity1. The response by airway CD4+ T cells could include a cytotoxic component and might also initiate the production of CXCR3 chemokines to recruit migratory dendritic cells, able to amplify the CD8+ T cell response for final clearance of infected cells3,17; in line with this we found that frequencies of SARS-CoV-2-specific CD4+ and CD8+ T cells were correlated in donor BALs.

Six of ten of the healthy donors tested had T cells capable of crossrecognizing SARS-CoV-2 proteins within their airways before the SARS-CoV-2 pandemic. Although extrapolations cannot be made from such a small cohort to the wider population, enhanced protective immunity in this proportion is compatible with SARS-CoV-2 challenge and household exposure studies where around half of those exposed resisted infection10,18,19,20,21. T cells were reactive against all four regions of SARS-CoV-2 tested; critically, these included T cells specific for the RNA-dependent polymerase (NSP12) and other components of the core RTC (polymerase cofactor NSP7, helicase NSP13) of SARS-CoV-2, which we recently found associated with abortive seronegative infection. Thus, the responses we previously identified in the periphery were probably a representative subset of their more protective lung mucosa-resident counterparts, which either had not yet acquired tissue residence or were ‘ex-TRM cells’22,23. Consistent with this, we noted that circulating polymerase (NSP12)-specific T cell frequencies correlated significantly with lung airway responses identified after overnight peptide re-stimulation. Pre-existing SARS-CoV-2-specific T cell frequencies were enriched in the BAL compared with the circulation. BAL cell yields did not allow a more comprehensive analysis of T cell specificities, but it is likely that responses targeting other regions of SARS-CoV-2 would be similarly enriched in the airways compared with blood. The frequency of lung TRM cells with cross-reactive potential against SARS-CoV-2 may also be underestimated because lavage samples only airway (epithelial) and not interstitial TRM cells, with the latter typically containing more T cells capable of cytotoxicity24,25. Although proliferating interstitial lung TRM cells can function as a reservoir of responses to seed and replenish the airway TRM cells26, it is the airway TRM cells that have been found to form a critical component of protection against animal coronaviruses3 and other respiratory viruses24. Maintaining a large memory pool of pathogen-specific T cells at the site of infection entry is a key protective advantage of the efficient frontline immunosurveillance provided by TRM cells in many tissues12,27,28.

Lung TRM cells may not be as long-lived as other tissue TRM cells, although stable frequencies persisted for more than 1 year in some human lung transplant recipients12,27,29,30. We could not determine longevity because the timing and nature of the original infection priming the cross-reactive lung T cells that we observed are unknown. Despite the small size of the cohort, we were able to detect differences in the strength of T cell and humoral immunity to seasonal HCoVs in those with or without detectable SARS-CoV-2-cross-reactive responses in the airways. This suggests that the timing and/or dose of previous HCoV exposure may shape the effectiveness of airway T cell crossprotection against SARS-CoV-2. Although ‘common cold viruses’ do not typically cause lower respiratory symptoms, respiratory syncytial virus (RSV) has also recently been shown to induce T cells in the lower airways in healthy adults with only upper respiratory tract symptoms13. Pre-existing SARS-COV-2 TRM cells have been identified by activation assays in tonsillar lymphoid tissue extracted before the pandemic31, suggesting that functional responses with crossprotective potential may also develop in the upper airways. However, it is likely that some airborne virus can bypass immune defenses in the upper respiratory tract, underscoring the importance of the lower airway T cells, which we have identified, in preventing the lung pathology that causes severe infection outcomes. Future studies will need to investigate whether, in addition to the pre-pandemic cross-reactive responses observed, airway-resident functional memory T cells can be formed after SARS-CoV-2 infection. T cell receptor sequencing has provided evidence for SARS-CoV-2-specific CD8+ T cells in nasal samples from four donors after severe COVID-19 (ref. 32). Lung TRM cells have been shown to form either after SARS-CoV-2 infection in selected cases where tissue was obtained from surgical resection or post mortem33,34, which did not allow airway responses to be distinguished from parenchymal/vascular TRM cells. It is not yet known whether airway or interstitial lung-resident T cell responses are induced by the current peripherally administered vaccines or whether mucosal delivery is required to achieve this.

Next-generation vaccines are being developed with greater focus on regions beyond the highly variable spike and on inducing mucosal immunity able to provide durable protection at the site of infection35. Our recent study highlighted the potential for RTC-specific T cells to abort early infection before PCR positivity and antibody seroconversion, potentially reducing transmission as well as disease and targeting early expressed viral replication proteins conserved across SARS-CoV-2 variants and other animal and HCoVs1. The data presented in the present study provide an explanation for the observed association between cross-reactive T cells and rapidly aborted infection, representing an immediate response at the site of infection after previous priming by, for example, an HCoV. Our findings underscore the rationale for vaccines delivered by the mucosal route, to expand tissue-resident T cells in the airways, including broadly cross-reactive SARS-CoV-2 RTC and spike specificities, aiming to provide frontline immunosurveillance against future variants.

留言 (0)

沒有登入
gif