SARS-CoV-2-specific T Cell Memory is Sustained in COVID-19 Convalescents for 8 Months with Successful Development of Stem Cell-like Memory T Cells

Study cohort

We recruited 94 individuals with SARS-CoV-2 infection. The peak disease severity was evaluated according to the NIH severity of illness categories¹⁹: asymptomatic (n=6), mild (n=44), moderate (n=23), severe (n=13), and critical (n=8). Whole blood samples were obtained longitudinally (2-4 time points) from 38 patients or at a single time point from 56 patients. Whole blood was collected 1 to 254 days post-symptom onset (DPSO). Finally, a total of 160 PBMC samples were analyzed. Among 160 PBMC samples, 33 samples were obtained in the acute phase when viral RNA was still detected (1 - 33 DPSO), and 127 samples were obtained in the convalescent phase after the negative conversion of viral RNA (31 - 254 DPSO). In the current study, we defined the acute phase as 1 - 30 DPSO, and the convalescent phase as 31 - 254 DPSO. The demographic and clinical characteristics of enrolled patients are presented in Supplementary Table 1.

SARS-CoV-2-specific T cell responses are sustained 8 months after the infection

First, we performed direct ex vivo interferon-γ (IFN-γ) enzyme-linked immunospot (ELISpot) assays following stimulation of PBMCs with overlapping peptide (OLP) pools spanning the spike (S), membrane (M), and nucleocapsid (N) proteins of SARS-CoV-2. S-, M-, and N-specific IFN-γ spot numbers increased during the acute phase. Subsequently, there was a decreasing tendency until 60 - 120 DPSO, and the IFN-γ responses were maintained over 8 months (Fig. 1a). These kinetics were also observed when S-, M-, and N-specific IFN-γ spot numbers were summed (Fig. 1a). In each PBMC sample, all three OLP pools evenly contributed to the IFN-γ responses without antigen dominance (Fig. 1b). We divided convalescent samples into three groups based on the DPSO at sample collection: T1 (31 - 99 DPSO), T2 (100 - 199 DPSO), and T3 (200 - 254 DPSO). We found no significant difference in IFN-γ spot numbers among groups (Fig. 1c), and an even contribution of antigens for the IFN- γ response was maintained among groups (Extended Data Fig. 1). Next, we focused on longitudinally tracked samples from 30 individuals in the convalescent phase. S-, M-, and N- specific and summed IFN-γ spot numbers were stable during the convalescent phase (Fig. 1d). When we compared paired samples from the same patient at two time points (t1, 31 - 100 DPSO; t2, 200 - 254 DPSO), we found no significant difference in IFN-γ spot numbers (Fig. 1e).

• Download figure
• Open in new tab

Fig. 1. SARS-CoV-2-specific IFN-γ responses over 8 months post-infection.

PBMC samples (n=135) from individuals with SARS-CoV-2 infection (n=76) were stimulated with OLPs of S, M, or N (1 μg/mL) for 24 h and the spot-forming units of IFN-γ-secreting cells examined by ELISpot. a, Scatter plots showing the relationship between DPSO and IFN-γ responses. The black line is a LOESS smooth nonparametric function, and the grey shading represents the 95% confidence interval. b, The composition of S-, M-, or N-specific IFN-γ responses among the total IFN-γ responses in each individual. c, IFN-γ responses were compared between T1 (n=46, 31 - 99 DPSO), T2 (n=37, 100 - 199 DPSO), and T3 (n=23, 200 - 254 DPSO). Data are presented as median and interquartile range (IQR). d,e, IFN-γ responses were analyzed in longitudinally tracked samples (n=82) from 30 individuals. d, Scatter plots showing the relationship between DPSO and IFN-γ responses. e, IFN-γ responses was compared between paired samples at two time points (n=14; t1, 31 - 100 DPSO; t2, 200 - 254 DPSO). Statistical analysis was performed using the Kruskal-Wallis test with Dunns’ multiple comparisons test (c) or the Wilcoxon signed-rank test (e). n.s, not significant.

We also evaluated SARS-CoV-2-specific T cell responses by AIM assays^5,20,21 following stimulation of PBMCs with OLP pools of S, M, and N. CD137⁺OX40⁺ cells were considered SARS-CoV-2-specific cells among CD4⁺ T cells, and CD137⁺CD69⁺ cells were considered SARS- CoV-2-specific cells among CD8⁺ T cells⁵ (Fig. 2a, b). We confirmed a strong positive correlation between the frequency of CD137⁺OX40⁺ cells and the frequencies of alternative AIM⁺ cells (OX40⁺CD154⁺ or CD137⁺CD154⁺ cells) among CD4⁺ T cells (Extended Data Fig. 2). The frequency of S-, M-, and N-specific CD137⁺OX40⁺ cells among CD4⁺ T cells increased during the acute phase, decreased until 60 DPSO, and then was maintained over 8 months (Fig. 2c). When the frequency of S-, M-, and N-specific CD137⁺OX40⁺ cells was summed, similar kinetics were observed (Fig. 2c). The frequency of CD137⁺CD69⁺ cells among CD8⁺ T cells was relatively low compared to the frequency of CD137⁺OX40⁺ cells among CD4⁺ T cells, but exhibited similar kinetics (Fig. 2c). We also observed a positive correlation between the frequency of CD137⁺OX40⁺ cells among CD4⁺ T cells and the frequency of CD137⁺CD69⁺ cells among CD8⁺ T cells (Extended Data Fig. 3).

• Download figure
• Open in new tab

Fig. 2. Kinetics of SARS-CoV-2-specific activation-induced marker (AIM)⁺ T cells.

PBMC samples (n=125) from individuals with SARS-CoV-2 infection (n=69) were stimulated with OLPs of S, M, or N (1 μg/mL) for 24 h. The frequency of AIM⁺ (CD137⁺OX40⁺) cells among CD4⁺ T cells and the frequency of AIM⁺ (CD137⁺CD69⁺) cells among CD8⁺ T cells were analyzed. a,b, Representative flow cytometry plots showing the frequency of AIM⁺ cells among CD4⁺ (a) or CD8⁺ (b) T cells. c, Scatter plots showing the relationship between DPSO and the frequency of AIM⁺ cells among CD4⁺ (left) or CD8⁺ (right) T cells. The black is a LOESS smooth nonparametric function, and the grey shading represents the 95% confidence interval.

Collectively, these results indicate that SARS-CoV-2-specific T cell responses are long- lasting over 8 months in COVID-19 convalescents.

SARS-CoV-2-specific T_SCM cells develop after the infection

Next, we examined the differentiation status of SARS-CoV-2-specific AIM⁺ T cells based on CCR7 and CD45RA expression in CD4⁺ (Fig. 3a) and CD8⁺ (Fig. 3b) T cell populations. Among SARS-CoV-2-specific AIM⁺CD4⁺ T cells, the proportion of CCR7⁺CD45RA^- (T_CM) cells was maintained at approximately 50% on average during the study period, and the proportion of CCR7^-CD45RA^- (T_EM) cells increased up to approximately 35% on average until 60 DPSO and then maintained thereafter (Fig. 3c). CCR7^-CD45RA⁺ (T_EMRA) cells were a minor population (< 5%; Fig. 3c). Notably, the proportion of CCR7⁺CD45RA⁺ cells, which include both naïve and T_SCM cells, was approximately 30% on average in the acute phase, decreasing to approximately 10% on average by 60 DPSO and then maintained thereafter (Fig. 3c).

• Download figure
• Open in new tab

Fig. 3. Differentiation status of SARS-CoV-2-specific AIM⁺ T cells. a-d,

PBMC samples (n=124) from individuals with SARS-CoV-2 infection (n=68) were stimulated with OLPs of S, M, or N (1 μg/mL) for 24 h and the expression of CCR7 and CD45RA was analyzed in AIM⁺ (CD137⁺OX40⁺) CD4⁺ (a,c) and AIM⁺ (CD137⁺CD69⁺) CD8⁺ (b,d) T cells. a,b, Gating strategies for identifying each memory subset among AIM⁺CD4⁺ (a) or AIM⁺CD8⁺ (b) T cells. c,d, Scatter plots showing the relationship between DPSO and the proportion of the indicated subsets among AIM⁺CD4⁺ (c) or AIM⁺CD8⁺ (d) T cells. e, PBMC samples (n=47) from individuals with SARS-CoV-2 infection (n=34) were stimulated with OLPs of S, M, or N (1 μg/mL) for 24 h and the frequency of T_SCM (CCR7⁺CD45RA⁺CD95⁺) cells was analyzed in AIM⁺CD4⁺ (upper) and AIM⁺CD8⁺ (lower) T cells. Left, The gating strategy for identifying T_SCM cells. Right, Scatter plots showing the relationship between DPSO and the proportion of T_SCM cells among AIM⁺CD4⁺ or AIM⁺CD8⁺ T cells. The black line is a LOESS smooth nonparametric function, and the grey shading represents the 95% confidence interval (c,d,e).

Among SARS-CoV-2-specific AIM⁺CD8⁺ T cells, the proportion of CCR7⁺CD45RA^- (T_CM) cells was approximately 20% on average in the acute phase and gradually decreased during the study period (Fig. 3d). The proportion of CCR7^-CD45RA^- (T_EM) cells increased up to 50% on average until 60 DPSO and was maintained thereafter (Fig. 3d). The proportion of CCR7^-CD45RA⁺ (T_EMRA) cells and CCR7⁺CD45RA⁺ cells was maintained (∼25% and 25-35% on average, respectively) during the study period (Fig. 3d).

We further investigated whether CCR7⁺CD45RA⁺ cells include T_SCM cells, which have a self-renewal capacity and multipotency for differentiation into diverse T cell subsets, by examining CD95, a marker of T_SCM cells^15,16. In both AIM⁺CD4⁺ and AIM⁺CD8⁺ T cells, CD95⁺ cells were a dominant population among CCR7⁺CD45RA⁺ cells (Fig. 3e), indicating that the CCR7⁺CD45RA⁺ cells among AIM⁺CD4⁺ or AIM⁺CD8⁺ T cells are mainly T_SCM cells. The frequency of T_SCM cells among both AIM⁺CD4⁺ and AIM⁺CD8⁺ T cells gradually increased until 120 DPSO, when it reached a stable plateau (Fig. 3e). We found no significant difference in the frequency of T_SCM cells among AIM⁺CD4⁺ and AIM⁺CD8⁺ T cells between T1 (31 – 99 DPSO), T2 (100 - 199 DPSO), and T3 (200 - 254 DPSO; Extended Data Fig. 4a, b).

• Download figure
• Open in new tab

Fig. 4. Frequency and differentiation status of SARS-CoV-2-specific MHC-I multimer⁺ T cells.

PBMC samples (n=15) from individuals with SARS-CoV-2 infection (n=11) were analyzed by flow cytometry. a, Representative flow cytometry plots showing the ex vivo detection of SARS-CoV-2 S₂₆₉ multimer⁺CD8⁺ T cells in the gate of CD3⁺ T cells. b, Scatter plot showing the relationship between DPSO and the frequency of SARS-CoV-2 S₂₆₉ multimer⁺ cells among total CD8⁺ T cells. Samples from the same patient are connected by solid lines. c,d, The expression of CCR7, CD45RA, and CD95 was analyzed in SARS-CoV-2 S₂₆₉ multimer⁺CD8⁺ T cells. IAV MP₅₈ multimer⁺ (n=5) and CMV pp65₄₉₅ multimer⁺ (n=6) cells from the PBMCs of healthy donors were also analyzed. Representative flow cytometry plots (c) show the proportion of the indicated subsets among multimer⁺ cells, and scatter plots (d) show the relationship between DPSO and the proportion of the indicated subsets among SARS-CoV-2 S₂₆₉ multimer⁺ cells. Samples from the same patient are connected by solid lines. Summary data showing the proportion of the indicated subsets among IAV multimer⁺ and CMV multimer⁺ cells are also presented (d). Horizontal lines represent median. e, A representative flow cytometry plot (upper) and summary data (lower) showing the percentage of PD- 1⁺TIGIT⁺ cells among SARS-CoV-2 S₂₆₉ multimer⁺ cells and total CD8⁺ T cells. Statistical analysis was performed using the Wilcoxon signed-rank test (e). **P < 0.01.

Successful development of SARS-CoV-2-specific T_SCM cells is confirmed by direct ex vivo MHC-I multimer staining

To validate the results from in vitro stimulation-based AIM assays, we detected SARS-CoV-2-specific CD8⁺ T cells by performing direct ex vivo MHC-I multimer staining and examined the differentiation status of MHC-I multimer⁺ cells. We used an HLA-A*02 multimer loaded with SARS-CoV-2 S₂₆₉ (YLQPRTFLL) peptide that has a low degree of homology to the human common cold coronaviruses (ccCoVs)^22,23. MHC-I multimer⁺ cells were detected during the study period until 234 DPSO (Fig. 4a, b), and we determined the proportions of T_CM (CCR7⁺CD45RA^-), T_EM (CCR7^-CD45RA^-), T_EMRA (CCR7^-CD45RA⁺), and T_SCM(CCR7⁺CD45RA⁺CD95⁺) cells among MHC-I multimer⁺ cells (Fig. 4c). Influenza A virus (IAV)- and cytomegalovirus (CMV)-specific MHC-I multimers were also used to stain PBMCs from healthy donors. Similar to the data from AIM⁺CD8⁺ T cells, T_EM and T_EMRA cells were the dominant populations among SARS-CoV-2-specific MHC-I multimer⁺ cells, whereas T_CM cells were a minor population during the study period (Fig. 4d). Among IAV- and CMV-specific MHC-I multimer⁺ cells from healthy donors, T_EM and T_EMRA cells were dominantly present, whereas T_CM cells were scarcely detected. T_SCM cells were also detected among SARS-CoV-2- specific MHC-I multimer⁺ cells during the study period, but not among IAV- and CMV-specific cells (Fig. 4d). In particular, the frequency of T_SCM cells among SARS-CoV-2-specific MHC-I multimer⁺ cells was high 60 – 125 DPSO.

A recent study has revealed two distinct subsets of CCR7⁺ stem cell-like progenitors: CCR7⁺PD-1^-TIGIT^- cells with stem cell-like features and CCR7⁺PD-1⁺TIGIT⁺ cells with exhausted traits²⁴. Therefore, we examined the expression of PD-1 and TIGIT in SARS-CoV-2-specific MHC-I multimer⁺ T_SCM cells. The frequency of PD-1⁺TIGIT⁺ cells was significantly lower among SARS-CoV-2-specific T_SCM cells than among total CD8⁺ T cells (Fig. 4e), confirming that SARS- CoV-2-specific T_SCM cells are bona fide stem-like memory cells.

Polyfunctionality and proliferation capacity are preserved in long-term SARS-CoV-2- specific T cells

As we observed the generation of SARS-CoV-2-specific T_SCM cells with self-renewal capacity and multipotency, we aimed to examine the kinetics of the polyfunctionality of SARS-CoV-2-specific T cells during the study period. To this end, we performed intracellular cytokine staining of IFN-γ, IL-2, TNF, and CD107a following stimulation with OLP pools of S, M, and N and analyzed the polyfunctionality of CD4⁺ and CD8⁺ T cells (Fig. 5a). We defined polyfunctional cells as cells exhibiting positivity for ≥ 2 effector functions. Among SARS-CoV- 2-specific CD4⁺ and CD8⁺ T cells, the average proportion of polyfunctional cells was 25 - 40% and 30 - 50%, respectively, 60 DPSO, and was maintained until 254 DPSO (Fig. 5b). We found no significant difference in the frequency of polyfunctional cells among SARS-CoV-2-specific CD4⁺ and CD8⁺ T cells between T1 (31 - 99 DPSO), T2 (100 - 199 DPSO), and T3 (200 – 254 DPSO; Fig. 5c). Preserved polyfunctionality among SARS-CoV-2-specific CD4⁺ and CD8⁺ T cells was also observed when polyfunctionality was evaluated according to the number of positive effector functions (Fig. 5d).

• Download figure
• Open in new tab

Fig. 5. Polyfunctionality and proliferation capacity of SARS-CoV-2-specific T cells. a-d,

PBMC samples (n=90) from individuals with SARS-CoV-2 infection (n=39) were stimulated with OLPs of S, M, or N (1 μg/mL) for 6 h. Intracellular cytokine staining was performed to examine the frequency of polyfunctional cells exhibiting positivity for ≥ 2 effector functions among SARS-CoV-2-specific CD4⁺ and CD8⁺ T cells. a, Representative flow cytometry plots showing the frequency of polyfunctional cells among CD4⁺ (left) and CD8⁺ (right) T cells. b, Scatter plots showing the relationship between DPSO and the frequency of polyfunctional cells among SARS-CoV-2-specific CD4⁺ (upper) or CD8⁺ (lower) T cells. The black line is a LOESS smooth nonparametric function, and the grey shading represents the 95% confidence interval. c, The fraction of polyfunctional cells among SARS-CoV-2-specific CD4⁺ (left) or CD8⁺ (right) T cells was compared between T1 (n=17, 31 - 99 DPSO), T2 (n=39, 100 - 199 DPSO), and T3 (n=25, 200 - 254 DPSO). Data are presented as median and IQR. d, Pie charts showing the fraction of cells positive for a given number of functions among SARS-CoV-2-specific CD4⁺ (left) or CD8⁺ (right) T cells. Each arc in the pie chart represents the indicated function. e, CTV-labeled PBMCs (n=11) obtained after 200 DPSO were stimulated with S OLP pool (1 μg/mL) for 120 h and the frequency of CTV^low and Ki-67⁺ cells among CD4⁺ T cells was analyzed. Representative plots (upper) and summary data (lower) are presented. Statistical analysis was performed using the Kruskal-Wallis test with Dunns’ multiple comparisons test (c) or the Wilcoxon signed-rank test (e). n.s, not significant, ***P < 0.001.

Finally, we examined the antigen-induced proliferation capacity of long-term SARS- CoV-2-specific memory T cells. We performed CellTrace Violet (CTV) dilution assays and Ki- 67 staining using PBMCs obtained after 200 DPSO. CD4⁺ T cells exhibited a robust proliferative response following in vitro stimulation with the S OLP pool (Fig. 5e), indicating that SARS-CoV-2-specific memory T cells elicit rapid recall responses upon viral re-exposure.

Considering preserved polyfunctionality and proliferation capacity of SARS-CoV-2- specific memory T cells, the current data indicate that memory T cells contribute to protective immunity against re-infection even 8 months after the primary infection.

Patients and specimens

In this study, 94 patients with PCR-confirmed SARS-CoV-2 infection were enrolled from Ansan Hospital and Chungbuk National University Hospital, Republic of Korea. Peripheral blood was obtained from all patients with SARS-CoV-2 infection. We also obtained peripheral blood from seven healthy donors. In six asymptomatic patients, the date of the first admission was regarded as 7 DPSO because the date of the first admission among symptomatic patients was an average 7 DPSO. This study was reviewed and approved by the institutional review board of all participating institutions and conducted according to the principles of the Declaration of Helsinki. Informed consent was obtained from all donors and patients.

PBMCs were isolated by density gradient centrifugation using Lymphocyte Separation Medium (Corning). After isolation, the cells were cryopreserved in fetal bovine serum (FBS; Corning) with 10% dimethyl sulfoxide (DMSO; Sigma-Aldrich) until use.

Ex vivo IFN-γ enzyme-linked immunospot assay

Plates with hydrophobic polyvinylidene difluoride membrane (Millipore, Billerica, MA) were coated with 2 μg/mL anti-human monoclonal IFN-γ coating antibody (clone 1-D1K, Mabtech) overnight at 4°C. The plates were washed with sterile phosphate-buffered saline (PBS) and blocked with 1% bovine serum albumin (Bovogen) for 1 hour at room temperature (RT). 700,000 PBMCs were seeded per well and stimulated with 1 μg/mL OLP pools spanning SARS-CoV-2 spike, membrane, and nucleocapsid proteins (Miltenyi Biotec) for 24 hours at 37°C. We used 10 μg/mL phytohemagglutinin as a positive control and an equimolar amount of DMSO as a negative control. Plates were washed with 0.05% Tween-PBS (Junsei Chemical) and incubated with 0.25 μg/mL biotinylated anti-human monoclonal IFN-γ antibody (clone 7-B6-1, Mabtech) for 2 hours at RT. After washing, streptavidin-alkaline phosphatase (Invitrogen) was added sequentially. Precipitates were detected with AP color reagent (Bio- Rad) and the reaction stopped by rinsing with distilled water. Spot-forming units were quantified using an automated ELISpot reader (AID). To quantify SARS-CoV-2-specific responses, spots in the negative control wells were subtracted from the OLP-stimulated wells.

Multi-color flow cytometry

Cells were stained with fluorochrome-conjugated antibodies for specific surface markers for 10 minutes at RT. Dead cells were excluded using LIVE/DEAD red fluorescent reactive dye (Invitrogen). In intracellular staining experiments, cells were fixed and permeabilized using the FoxP3 staining buffer kit (Invitrogen), and then stained for intracellular markers for 30 minutes at 4°C. Multi-color flow cytometry was performed using an LSR II instrument (BD Biosciences) and the data analyzed in FlowJo software (FlowJo LLC). The fluorochrome-conjugated MHC-I multimers and monoclonal antibodies used in this study are listed in Supplementary Table 2.

Activation-induced marker assay

PBMCs were blocked with 0.5 μg/mL anti-human CD40 mAb (clone HB14, Miltenyi Biotec) in RPMI 1640 supplemented with 10% FBS and 1% penicillin and streptomycin for 15 minutes at 37°C. The cells were then cultured in the presence of 1 μg/mL SARS-CoV-2 OLP pools and 1 μg/mL anti-human CD28 and CD49d mAbs (clone L293 and L25, respectively, BD Biosciences) for 24 hours. Stimulation with an equal concentration of DMSO in PBS was performed as a negative control.

MHC-I multimer staining

PBMCs were stained with APC-conjugated MHC-I SARS-CoV-2 S₂₆₉ pentamer (Proimmune), IAV MP₅₈ dextramer (Immudex), or CMV pp65₄₉₅ dextramer (Immudex) for 15 minutes at RT and washed twice. Additional surface markers were stained using the protocols described above.

Stimulation for intracellular cytokine staining

PBMCs were cultured in the presence of SARS-SoV-2 OLP pools and 1 μg/mL anti- human CD28 and CD49d mAbs for 6 hours at 37°C. Brefeldin A (GolgiPlug, BD Biosciences) and monensin (GolgiStop, BD Biosciences) were added 1 hour after the initial stimulation.

Proliferation assays

PBMCs were labeled using the CellTrace™ Violet Cell Proliferation Kit (Invitrogen) at a concentration of 1.0 × 10⁶ cells/mL for 20 minutes at 37°C. We then added 1% FBS-PBS to the cells for 5 minutes at RT to quench unbound dyes. Cells were washed and cultured in RPMI 1640 (Corning) containing 10% FBS (Corning) and 1% penicillin-streptomycin (Welgene) at a concentration of 500,000 cells per well in the presence of 1 μg/mL SARS-CoV- 2 spike peptide pools and 1 μg/mL anti-human CD28 and CD49d mAbs (clone L293 and L25 respectively, BD Biosciences) for 120 hours. Stimulation with an equal concentration of DMSO in PBS was performed as a negative control. After incubation, cells were harvested and stained with antibodies for analysis by flow cytometry.

Data quantification and statistical analysis

Data were calculated as background-subtracted data. Background-subtracted data were derived by subtracting the value after DMSO stimulation. When three stimuli were combined, the value after each stimulation was combined and we subtracted triple the value derived from DMSO stimulation.

Statistical analyses were performed using GraphPad Prism version 9 for Windows (GraphPad Software). Significance was set at p<0.05. The Wilcoxon signed-rank test was used to compare data between two paired groups. In addition, the Kruskal-Wallis test with Dunns’ multiple comparisons test was used to compare non-parametric data between multiple unpaired groups. To assess the significance of correlation, the Spearman correlation test was used. In cross-sectional analyses, a non-parametric local regression (LOESS) function was employed with 95% confidence interval.