Predictive markers of SARS-CoV-2 spike-specific cytotoxic T cell activity following Omicron breakthrough infection

ABSTRACT

At a population level, evaluation of vaccine efficacy and of immunity in general focuses on measuring humoral immunity, especially antibody-mediated neutralization. However, antibody-dependent cell-mediated cytotoxicity (ADCC) and cellular immunity are key processes for pathogen clearance and often provide broader protection against emerging variants. More information on vaccine and infection-related determinants or markers of cell-mediated cytotoxicity is needed to inform design of standardized, high throughput tests assessing these processes. We previously showed that persons infected with SARS-CoV-2 prior to vaccination (hybrid immunity) had polyfunctional CD8⁺ T cells and robust ADCC against SARS-CoV-2 spike (S), while persons infected with Omicron after vaccination (breakthrough infection) had weak ADCC. In the current study, we investigated the impact of Omicron breakthrough infection on S-specific cellular immunity, including polyfunctional CD8⁺ T cell and cytotoxic T lymphocyte (CTL) levels of 28 previously uninfected individuals vaccinated either two or three times against SARS-CoV-2. In the group with three vaccinations, Omicron breakthrough infection significantly increased the frequency of circulating SARS-CoV-2 S-specific T cells, polyfunctional interferon-gamma and interleukin-2-producing SARS-CoV-2 S-specific CD8⁺ T cells and CTL after in vitro stimulation. Matrix testing with overlapping S peptides indicated recognition of one to two CTL epitopes per individual and revealed two previously unreported epitopes. Circulating S-specific cellular immune response magnitude neither correlated with, nor predicted CTL activity against SARS-CoV-2 S, but polyfunctional S-specific CD8⁺ T cell frequency seven days after stimulation and antibody reactivity against determinants indicating robust ADCC were both associated with day 10 CTL activity.

Introduction

Measuring serum or plasma antibody reactivity is a simple, accessible way to assess an anti-microbial immune response, with higher antibody levels generally indicating better responses. Adaptive immunity that prevents microbial infection is largely mediated by neutralizing antibodies that physically block receptor-mediated entry into susceptible host cells [1]. However, when antigenic variation or other immune evasion mechanisms enable microbes to establish chronic infection or to reinfect, neutralizing antibody levels against transient variants have little relevance to ongoing immune protection [2]. Once infection occurs, specific cell-mediated killing of infected cells limits cell-to-cell spread and attempts microbial clearance. Cell-mediated killing of infected host cells can limit the severity of illness and is often maintained against microbial variants that evade antibody neutralization [3–6]. Natural killer (NK) cell triggering by IgG subclass antibodies against microbial antigens expressed on the host cell surface and the coordinate recognition of class I human histocompatibility linked leukocyte antigen (HLA)/microbial peptide complexes by clonotypic T cell receptors (TCR) of cytotoxic T cells (CTL) ensure specificity of cell-mediated killing. Identification of antigen-specific T cells and functional evaluation of mechanisms underlying cytotoxicity against infected cells requires specialized expertise and materials, is labour intensive and is often prohibitively time consuming for population level studies. Therefore, more information on determinants of vaccine and infection-induced antigen-specific cell-mediated killing is required to design efficient, reliable assays indicating induction of effective cell-mediated killing of infected host cells by vaccination or following infection.

Antibody dependent cell-mediated cytotoxicity (ADCC) capacity varies with NK cell status, IgG fine specificity and IgG subclass distribution [7–9]. It is often independent of overall antibody levels or reactivity against neutralizing determinants and, thus, cannot be reliably measured in a simple ELISA format [8,9]. Measurement of antigen-specific cytokine production by peripheral blood mononuclear cells (PBMC) in ELISpot assays and of activation-induced markers by flow cytometry (AIM assay) assesses cellular immunity and distinguishes activated CD4⁺ from CD8⁺ T cells, but provides little information on CTL activity [10–14]. Tetramer-based flow cytometry measures expansion of CD8⁺ T cells with TCR specific for antigens of interest and can also assess their expression of relevant cytotoxins (perforin and granzymes), but does not directly address cytotoxic function and generalization across the greater population is limited by HLA diversity [15,16]. Single cell sequencing can identify expansion of T cell receptors recognizing specific HLA class I/microbial peptide complexes, but this also does not directly address function [17]. Cytotoxicity assays that directly measure ADCC or CTL activity usually require in vitro manipulation of effector cells, individualized target cell production, co-incubation of effectors and targets under specialized conditions and quantitative interpretation of cell death indicators. Despite the importance of cell-mediated cytotoxicity in anti-microbial immunity, its measurement remains restricted to the basic research laboratory and has a limited role informing public health policy.

To rationally and comprehensively assess population level immune protection against SARS-CoV-2 acquired through vaccination or infection, more accessible and informative markers of specific cellular immunity and cytotoxicity against infected cells are needed. Better basic understanding of factors associated with induction of robust microbe-specific CTL activity and antibodies triggering ADCC will help to identify such markers. To this end, we investigated immunological features associated with Omicron breakthrough infection and subsequent enhanced cytotoxicity against SARS-CoV-2 S sensitized target cells. Although not apparent from ELISpot assays with resting PBMC, S-specific polyfunctional CD8⁺ T cell (defined herein as producing IFN-γ and IL-2) frequencies following in vitro expansion may be a useful marker of features underlying effective CTL activity. Further, antibody reactivity against specific determinants of S previously associated with robust ADCC is also associated with S-specific CTL activity and may be indicative of a robust immune response in general.

Materials and methods

Study approval and subjects

This study conformed to recommendations of the Canadian Tri-Council Policy Statement: Ethical Conduct for Research Involving Humans and ethical approval was given by the Health Research Ethics Authority of Newfoundland and Labrador. Peripheral blood was collected between 2021 and 2023 from study subjects at approximately three-month intervals and a questionnaire addressing previous SARS-CoV-2 testing/infection history was administered at study intake after written informed consent in accordance with the Declaration of Helsinki. This study included 22 individuals with no known exposure to SARS-CoV-2 prior to receiving three doses of a COVID-19 vaccine before experiencing mild/moderate Omicron infection and six individuals who experienced breakthrough infection after two vaccinations (Table 1). Participants self-declared symptoms, medical treatments they were receiving and information on any comorbidity. Persons with any known underlying immune compromising condition or on immunosuppressive treatment were excluded.

Table 1.

Relevant features and categorization of breakthrough infection study cohort.

		PV2 Breakthrough n = 6	PV3 Breakthrough n = 22
	Median Age (IQR)	45 (33.5-56.75)	58.5 (40.75-66)
	Mean Age ± SD	44.5 ± 15.8	55.2 ± 13.0
	Female n (%)	3 (50)	14 (63.6)
	Male n (%)	3 (50)	8 (36.4)
Symptoms n (%)	Mild	3 (50)	16 (72.7)
	Moderate	3 (50)	6 (27.3)
Vaccine 1	ChAdOx1-S n (%)	N/A	6 (27.3)
	Moderna n (%)	2 (33.3)	N/A
	Pfizer-BioNTech n (%)	4 (66.7)	16 (72.7)
Vaccine 2	ChAdOx1-S n (%)	N/A	N/A
	Moderna n (%)	3 (50)	1 (4.5)
	Pfizer-BioNTech n (%)	3 (50)	21 (95.5)
	Mean Days Between Vaccination ± SD	88.7 ± 52.0	75.9 ± 19.6
Vaccine 3	Moderna n (%)	N/A	13 (59.1)
	Pfizer-BioNTech n (%)	N/A	9 (40.9)
	Mean Days Between Vaccination ± SD	N/A	180.7 ± 30.1
Samples	Mean Days PV2 ± SD*	55.2 ± 21.3	N/A
	Mean Days PV3 ± SD*	N/A	35.4 ± 6.7
	Mean Days POMI ± SD*	76.7 ± 97.0	45 ± 20.3
	Mean Days Between PV2 and POMI ± SD*	195.8 ± 49.1	N/A
	Mean Days Between PV3 and POMI ± SD*	N/A	119.7 ± 60.4

*Days post second vaccination (PV2); days post third vaccination (PV3); days post Omicron infection (POMI); mild = few symptoms < 7 days; moderate = multiple symptoms > 7 days.

Blood sample processing

Samples were collected and processed as previously described [9]. Briefly, whole blood was collected by venipuncture in acid citrate dextrose vacutainers, after which plasma was collected following 10 min centrifugation at 500 g and stored at −80°C. Peripheral blood mononuclear cells (PBMC) were isolated from anti-coagulated blood by density gradient centrifugation using the Canadian Autoimmunity Standardization Core consensus standard operating procedure (version: March 21, 2019). Freshly isolated PBMC were resuspended in freezing medium consisting of fetal bovine serum (FBS; HyClone™, GE Healthcare Life Sciences, Logan, UT, USA) supplemented to 10% dimethyl sulfoxide (DMSO; Sigma-Aldrich, St. Louis, MO, USA) and cooled at 1°C per minute overnight to −80°C, then maintained in liquid nitrogen until use.

Cell lines and viruses

All cell lines and PBMC were cultured with 5% CO₂ at 37°C using lymphocyte medium (LM) consisting of RPMI-1640 with 10% FBS (HyClone™), 200 IU/mL penicillin/streptomycin, 0.01 M HEPES, 1% L-glutamine (all from Invitrogen, Toronto, ON, Canada) and 2 × 10⁻⁵ M 2-mercaptoethanol (Sigma-Aldrich). Autologous B lymphoblastoid cells (BLCL) were generated from fresh PBMC using Epstein–Barr Virus (EBV)-containing B95-8 supernatant. Cyclosporin A (Sigma-Aldrich) diluted in LM was added 24 h after infection to a final concentration of 1 μg/mL and culture medium was supplemented to 20% FBS. PBMC were recovered overnight in LM and used in functional experiments when viability was > 75% as determined by trypan blue staining.

HLA typing

An aliquot of 5 × 10⁶ BLCL was washed twice with PBS (Sigma-Aldrich) and genomic DNA was isolated by phenol–chloroform extraction followed by ethanol precipitation. Briefly, cell pellets were resuspended in 100 μL PBS and digested in 200 μL lysis buffer (10 mM Tris pH 8.0, 10 mM NaCl, 1 mM EDTA, 0.5% SDS; Sigma-Aldrich) and proteinase K (final concentration 100 μg/mL; Invitrogen) overnight at 42°C. DNA extracted by standard phenol–chloroform method was precipitated with sodium acetate (NaAc)-ethanol (2 volumes 100% ethanol and 1/10 volume 3 M NaAc) overnight at −20°C then washed with 70% ethanol and dissolved in TE buffer [18]. Participant HLA-A, -B and -C genotyping was performed using sequence-specific oligonucleotides by the Nova Scotia Health Authority HLA Typing Laboratory (Halifax, NS, Canada).

Peptides and peptide matrix

The peptide pool for Wuhan-Hu-1 S (PepTivator® SARS-CoV-2 Prot_S Complete; Miltenyi Biotec, San Diego, CA, USA) consisting of 15-mers with 11 amino acid (aa) overlap covering functional domains (aa 5–1273) of the S sequence was reconstituted at 50 μg/mL in endotoxin-free ultra-pure water (EF H₂O; Sigma-Aldrich). Individual overlapping peptides (13- or 17-mers, with 10 amino acid overlaps) spanning the canonical Wuhan-Hu-1 S sequence (NR-52402; BEI Resources, Manassas, VA, USA) were reconstituted at 10 mg/mL in DMSO. Sequences for individual 9-mer peptides were determined from published literature and NetMHCpan 4.1 EL predictions (IEDB Analysis Resource) based on cohort HLA typing (Table 2). The peptides were synthesized by Peptide2.0 (Chantilly, VA, USA) at 95% purity and resuspended at 10 mg/mL in DMSO. All peptides were stored at −20°C.

Table 2.

Individual peptides for CTL stimulation.

HLA Allele	Peptide Name	Sequence
A*02	A*02_Wu_S39	YLQPRTFLL
A*02	A*02_Wu_S54	FSTFKCYGV
A*02	A*02_Omi_S54	FFAFKCYGV
A*03	A*03_Wu_S54	KCYGVSPTK
A*03	A*03/B*35_Wu_S54	ASFSTFKCY
A*03	A*03/B*35_Omi_S54	APFFAFKCY
A*03	A*03/11_Wu_S65	RLFRKSNLK
A*11	A*03/11_Wu_S65	RLFRKSNLK
A*24	A*24_Wu_S11&12	RFDNPVLPF
A*24	A*24_Omi_S11&12	RFDNPALPF
A*24	A*24_Wu_S64	NYNYLYRLF
A*24	A*24_OmiR_S64	NYNYRYRLF
A*24	A*24_OmiQ_S64	NYNYQYRLF
A*24	A*24_Wu_S91&92	VYSTGSNVF
A*24	A*24/B*X_Wu_S173	QYIKWPWYI
A*30	A*03/B*35_Wu_S54	ASFSTFKCY
B*07	B*07_Wu_S97&98	SPRRARSVA
B*07	B*07_Omi_S97&98	SHRRARSVA
B*14	B*14_Wu_S34	TRFQTLLAL
B*15	A*03/B*35_Wu_S54	ASFSTFKCY
B*18	B*18_Wu_S74	FELLHAPAT
B*35	A*03/B*35_Wu_S54	ASFSTFKCY
B*35	A*03/B*35_Omi_S54	APFFAFKCY
B*35	A*24/B*X_Wu_S173	QYIKWPWYI
B*44	B*44_Wu_S172	QELGKYEQY
B*44	B*44_Wu_S172&173	YEQYIKWPW
B*44	A*24/B*X_Wu_S173	QYIKWPWYI
B*57	B*57_Wu_S117	RSFIEDLLF
B*57	B*57_Wu_S126a	LAGTITSGW
B*57	B*57_Wu_S126b	GTITSGWTF

HLA alleles and corresponding reactive epitopes in cohort indicated in bold.

To resolve TCR restriction and peptide specificity, we created a peptide matrix using individual overlapping peptides spanning the entire S sequence (180 17-mers and 1 N-terminal 13-mer with 10 amino acid overlaps; NR-52402 BEI Resources). This matrix consisted of 27 pools (13-14 peptides/pool) of equivalent amounts of peptides whereby only two peptide pools contain the same peptide. DMSO was added as a placeholder to pools with only 12 peptides to maintain relatively consistent DMSO concentration across pools (Figure 1).

Figure 1.

Schematic overview of PBMC stimulation and experimental design. Serial PBMC samples were recovered overnight and a fraction applied for 24 h to ELISpot plates with either vehicle (negative control), the full SARS-CoV-2 S peptide pool, anti-CD3 (positive control) or individual peptides to assess circulating S-specific T cell frequency by IFN-γ and/or IL-2 production. Remaining PBMC were stimulated with the full SARS-CoV-2 S peptide pool or individual peptides in the presence of IL-7 with IL-2 added on day 3. On day 7, an aliquot of the cultured cells was restimulated for 5 h with either vehicle, the full SARS-CoV-2 S peptide pool or individual peptides, followed by flow cytometric analysis of individual T cell subsets for IFN-γ and IL-2 production. On day 10, the remaining expanded cells were assessed for CTL activity using a ⁵¹Cr-release assay against autologous BLCL pulsed with vehicle, the full SARS-CoV-2 S peptide pool, or individual peptides selected based on donor HLA typing. A matrix approach was used to deconvolute unknown CTL epitopes using BLCLs pulsed with matrix peptide pools, with confirmation using individual peptides from BEI on days 11-12.

ELISpot

Serial PBMC samples were recovered overnight in LM and acute T cell responses measured using IFN-γ/IL-2 FluoroSpot (ImmunoSpot®, Cleveland, OH, USA) designed for 96-well low autofluorescence PVDF-membrane plates as outlined by the manufacturer. See Figure 1 for sample flow chart. Vehicle controls of EF H₂O and DMSO were included to obtain background signals and anti-CD3 (OKT3, ATCC® CRL-8001) served as a positive control. Peptide pools were used at a final concentration of 1 μg of each peptide/mL to stimulate 2 × 10⁵ PBMC (final volume 200 μL) for 24 h. Cells and peptides or cells and controls were diluted in CTL-Test™ Medium (ImmunoSpot®) and each condition was performed in duplicate. Background (vehicle) spot counts were subtracted from stimulated spot counts and the frequency of polyfunctional, IFN-γ and IL-2 spot forming cells per million PBMC (SFC/10⁶ PBMC) was captured on a CTL ImmunoSpot® S6 Universal Analyzer and illustrated using GraphPad Prism.

CD8⁺ T cell stimulation

A total of 2 - 5 × 10⁶ PBMC from the same serial samples recovered for ELISpot were stimulated with the S peptide pool (Miltenyi Biotec) as previously described [9] or individual peptides (BEI Resources; Peptide2.0) at 1 μg of each peptide/mL for 1 h in a minimal volume after which time 1 mL of LM containing 25 ng/mL IL-7 (National Cancer Institute, Frederick National Laboratory for Cancer Research, Frederick, MD, USA) was added and cell culture continued. IL-2 (final concentration 10 IU/mL; National Cancer Institute) was added on day 3. Cells were subsequently used on day 7 (intracellular cytokine stain) and day 10 (autologous CTL assay) as described. See Figure 1 for sample analysis flow chart.

Intracellular cytokine staining

On day 6, cultured cells were counted and resuspended at 2 × 10⁶ cells/mL in LM. Effector T cells (0.5 × 10⁶ cells) were restimulated for 5 h on day 7 with the S peptide pool (Miltenyi Biotec) as previously described [9] or individual peptides (BEI Resources and/or Peptide2.0) at 1 μg/mL in the presence of 10 μg/mL Brefeldin A (Sigma-Aldrich) in a final volume of 500 μL. Effector T cells were stained with directly conjugated mAb against human CD3 (VioGreen, REA613, Miltenyi Biotec), CD4 (APC-Vio770, REA623, Miltenyi Biotec), CD8 (AlexaFluor® 700, HIT8a, BioLegend), CD57 (FITC, NK-1, BD Biosciences, Toronto, ON, Canada), and IFN-γ (APC, 4S.B3, Invitrogen) and IL-2 (PE, MQ1-17H12, Invitrogen) using the Inside Stain Kit (Miltenyi Biotec) as per manufacturer’s instructions and as previously described in [9]. Data were acquired on a CytoFLEX flow cytometer and analyzed and illustrated using Kaluza Version 2.1 (Beckman Coulter, Brea, CA, USA) and GraphPad Prism. Intracellular IL-2 and IFN-γ from CD4⁺ or CD8⁺ T cells responding to peptide restimulation was expressed with the signal from non-restimulated cells subtracted as background. Refer to Figure 1 for sample analysis flow chart.

Autologous CTL assay

On day 9, cultured cells were counted and resuspended at 0.5 × 10⁶ cells/mL in LM. On day 10, autologous BLCL targets were labelled for 90 min with 100 μCi Na₂⁵¹CrO₄ (PerkinElmer, Akron, OH, USA) at 37°C, 5% CO₂, then washed four times in PBS containing 1% FBS before resting in 5 mL lymphocyte medium for 1 h at 37°C, 5% CO₂ to minimize spontaneous release and resuspended at 10⁵ cells/mL in fresh LM. Autologous BLCL (5 × 10⁴) target cells were pulsed with S peptide pool (Miltenyi Biotec) or individual peptides (BEI Resources; Peptide2.0) at 1 μg/mL in individual wells of a 96 well U-bottom plate (final volume 80 μL; performed in duplicate) at 37°C, 5% CO₂ for 30 minutes prior to addition of effector cells. Effector cells were introduced at effector to target (E:T) ratios of 5:1 and 20:1 and cytotoxic activity measured by ⁵¹Cr release over 5 h at 37°C, 5% CO₂ (final assay volume 300 μL). ⁵¹Cr release was measured in 125 μL supernatant on a Wallac 1480 Wizard gamma counter and percent specific lysis calculated by (experimental ⁵¹Cr release – spontaneous ⁵¹Cr release) / (maximum ⁵¹Cr release – spontaneous ⁵¹Cr release) x 100.

If resolution of TCR restriction and peptide specificity was not obtained by individual peptides selected on day 10, the ⁵¹Cr assay was repeated on day 11 or 12, as indicated above, using BLCL pulsed with S peptides arranged in a matrix (Figure 1) to determine peptide recognition by effector T cells at E:T 5:1. ⁵¹Cr release was measured as indicated above. See Figure 1 for sample analysis flow chart.

Peptide reactivity ELISA

Chemiluminescence ELISA was performed as previously outlined [8] using individual overlapping peptides spanning the canonical Wuhan-Hu-1 S sequence (NR-52402; BEI Resources). Total IgG binding was measured in a 1 h incubation with 100 μL/well of 1:50,000 HRP-conjugated polyclonal goat anti-human IgG (Jackson ImmunoResearch Labs) and developed with SuperSignal™ ELISA Pico Chemiluminescent Substrate (ThermoFisher, Waltham, MA, USA). Reactions were read within 20 mins on a Fluoroskan Ascent FL luminescence reader (Labsystems, Lima, Peru). Background signal (diluent only) was subtracted from all values and corrected relative light unit (RLU) values obtained. Relevant peptide S^1142-1165C was synthesized by Peptide2.0 at 95% purity, resuspended as above and chemiluminescence ELISA performed.

Statistics

Statistical analyses were performed using GraphPad Prism with two-sided P-values < 0.05 considered significant. Normality of data distributions was assessed using Shapiro–Wilk test. Differences in means with standard deviation (SD) or medians with interquartile range (IQR, calculated as IQR = Q₃ − Q₁) between groups were compared using one-way ANOVA, Student’s t-test, paired t-test, Friedman test or Mann–Whitney U-test as appropriate based on normality of data distribution and paired sampling.

Results

Omicron breakthrough infection expands circulating SARS-CoV-2 S-specific T cells

Samples collected from six individuals who experienced Omicron breakthrough infection after two COVID-19 vaccinations and 22 individuals who experienced breakthrough infection after three vaccinations were tested pre- and post-infection for levels of cellular immunity against SARS-CoV-2 S protein. Their general characteristics, vaccine information, timing of events and symptom severity are shown in Table 1. Previous infection with SARS-CoV-2 was ruled out by longitudinal measurement of antibodies specific for SARS-CoV-2 nucleocapsid (N) protein as previously described [19].

A schematic illustrating PBMC handling, stimulation and analysis is shown in Figure 1. The frequency of circulating T cells producing IFN-γ and/or IL-2 was measured by ELISpot assay before and after Omicron breakthrough infection and expressed as spot-forming cells/10⁶ PBMC. The median frequency of circulating polyfunctional S-specific T cells and of those producing IFN-γ or IL-2 significantly increased when Omicron breakthrough infection occurred after three vaccinations (Figure 2). In contrast, breakthrough infection after two vaccinations had no significant effect on the mean frequency of circulating polyfunctional S-specific T cells, nor on the mean frequency of those producing IFN-γ or IL-2 (Figure 2). Omicron breakthrough infection following two vaccinations had no significant effect on cellular immune response magnitude against SARS-CoV-2 S as indicated by the frequency of polyfunctional, IFN-γ or IL-2 producing T cells while breakthrough infection following three vaccinations increased the frequency of all three categories of cytokine-producing T cells. This suggests that three vaccinations prime S-specific T cells to respond to SARS-CoV-2 infection better than two vaccinations, although the two vaccination group studied herein is too small to support a definitive statement to that effect. Current ELISpot methodology does not enable discrimination of CD4⁺ and CD8⁺ cytokine-producing cells without depletion of one or the other subset.

Figure 2.

Circulating S-specific SARS-CoV-2 T cell activity. Circulating T cell responses were measured over 24 h in ELISpot assays (in duplicate) using vehicle control, full SARS-CoV-2 S peptide pool, anti-CD3 or individual peptides. Persons with two (PV2; blue, right panel; n = 6) or three (PV3; pink, left panel; n = 22) vaccinations followed by Omicron breakthrough infection (POMI) were tested for (A) IFN-γ, (B) dual IFN-γ/IL-2, and (C) IL-2 SFC/10⁶ PBMC. Data are shown as median with IQR (PV3/POMI) and as mean ± SD (PV2/POMI). Groups compared are spanned by lines above them with significant differences (p value) or non significance (ns) indicated on the line. Wilcoxon matched-pairs signed rank test or Student’s paired t-test was used as appropriate for comparisons based on normality of data distribution.

Omicron breakthrough infection increases in vitro expansion of polyfunctional SARS-CoV-2 S-specific CD8⁺ T cells

In a previous study of individuals infected with SARS-CoV-2 before being vaccinated, we found that in vitro expansion of polyfunctional CD8⁺ T cells was greater in those who had experienced more severe infection and that this corresponded with higher CTL activity [9]. Therefore, to investigate the effect of SARS-CoV-2 infection after vaccination, we measured mono- and polyfunctional S-specific CD4⁺ and CD8⁺ T cell activity by cell surface and intracellular flow cytometry after in vitro stimulation prior to and following Omicron breakthrough infection. The gating strategy is depicted in Figure 3A. As with circulating T cell responses, the percentage of polyfunctional, IFN-γ or IL-2-producing CD8⁺ T cells in response to SARS-CoV-2 S peptide stimulation following in vitro expansion from persons vaccinated twice was not significantly increased following Omicron breakthrough infection (Figure 3B-D). In contrast, the percentage of CD8⁺ T cells from persons vaccinated three times that produced IFN-γ and IL-2 or produced either cytokine in response to SARS-CoV-2 S peptide stimulation following in vitro expansion significantly increased following Omicron breakthrough infection (Figure 3B-D). These data suggest that a third COVID-19 vaccination primes S-specific circulating T cells to respond to Omicron breakthrough infection more robustly than two vaccinations and in combination with Omicron breakthrough infection, a third COVID-19 vaccination activates S-specific CD8⁺ T cells for stronger responses to in vitro stimulation. We saw no significant effect of a third vaccination or of Omicron breakthrough infection on the magnitude of in vitro S-specific CD4⁺ T cell responses (data not shown). Following Omicron infection, the percentage of CD8⁺ T cells producing IL-2 in response to S peptide stimulation seven days post in vitro stimulation correlated significantly with the number of circulating S-specific T cells producing IFN-γ on day 1 (p = 0.0119; Figure 3E). While this result does not demonstrate a direct connection, it does support a relationship between aspects of the circulating S-specific T cell response and in vitro activation of IL-2 producing CD8⁺ T cells and CTL.

Figure 3.

IFN-γ and IL-2 production from expanded SARS-CoV-2 S-specific CD8⁺ T cells. (A) Representative gating strategy for identifying CD3 ⁺ CD8⁺ T cells. On day 7, expanded T cells were restimulated for 5 h with the full SARS-CoV-2 S peptide pool. Cytokine production was assessed for (B) IFN-γ, (C) dual IFN-γ/IL-2, and (D) IL-2 expression. Background signals from vehicle controls were subtracted before plotting. Data are shown as median with IQR (PV3/POMI n = 22; PV2/POMI n = 6). (E) Correlation between IL-2 production by CD8⁺ T cells on day 7 and IFN-γ spot-forming units per 10⁶ total T cells on day 1 was assessed. P values in (B-D) were calculated using Wilcoxon matched-pairs signed rank test and are shown above lines spanning the groups compared when significant or as ns when not significant. Significance of correlation in (E) was assessed using Spearman’s correlation.

Robust SARS-CoV-2 S-specific CTL activity after Omicron breakthrough infection

We measured specific killing of autologous BLCL sensitized with overlapping S peptides after 10 days in vitro stimulation of PBMC collected before and after Omicron breakthrough infection in six individuals vaccinated twice and 22 individuals vaccinated three times. We considered specific lysis ≥ 10% (the conventional cutoff to indicate specific killing activity) at an E:T of 5:1 detectable S-specific CTL activity. None of the six individuals vaccinated twice had detectable S-specific CTL activity before or after Omicron breakthrough infection, however, prior to Omicron infection, 6/22 individuals vaccinated three times had detectable S-specific CTL activity (Figure 4A). Following Omicron infection, 11 of these 22 individuals had S-specific CTL activity of ≥ 10% specific lysis, an approximately two-fold increase over vaccination alone with an increase in median CTL activity (p = 0.0029; Figure 4A). HLA-A, -B and -C distribution for the cohort of persons vaccinated three times prior to experiencing Omicron breakthrough infection is depicted in pie graphs in Figure 4B. Omicron breakthrough infection significantly increased S-specific CTL activity resulting from three vaccinations, but infection following two vaccinations did not provide a comparable increase in S-specific CTL activity of the six individuals tested.

Figure 4.

SARS-CoV-2 S-specific CTL activity and HLA distribution. (A) On day 10, CTL activity against autologous BLCL pulsed with vehicle, the full SARS-CoV-2 S peptide pool or individual peptides was measured by ⁵¹Cr-release assay. (B) For participants with three vaccinations then Omicron breakthrough infection (n = 22), peptides were selected based on donor HLA-A, -B and -C typing. (C) A representative S protein matrix deconvolution result (performed in duplicate) is shown for one donor using BLCL pulsed with matrix peptide pools (day 11). (D) Summary of HLA restriction for participants with CTL activity (n = 11). Confirmation of CTL restriction was performed using individual peptides from BEI on day 12. In (A), PV3/POMI data are shown as median with IQR and PV2/POMI as mean ± SD. P values were calculated using Wilcoxon matched-pairs signed rank test and shown above lines spanning the groups compared with ns posted if there was no significant difference.

Confirmation of CTL epitopes within SARS-CoV-2 S

Using the peptide matrix testing method depicted in Figure 1, we confirmed recognition of six distinct CTL epitopes within SARS-CoV-2 S predicted on the basis of HLA class I binding. Two of these were previously uncharacterized epitopes: HLA-B*15:01-restricted GVYYPDKVF; HLA-C*12:03-restricted QSAPHGVVF. An example of cytotoxicity assay results confirming recognition of the QSAPHGVVF epitope in the context of HLA-C*12:03 is shown in Figure 4C. It is not possible to estimate the frequency with which these epitopes are recognized amongst HLA-B*15:01- and HLA-C*12:03-expressing individuals in the general population and within different HLA class I haplotypes from these results.

Among the eleven individuals with detectable SARS-CoV-2 S CTL activity following Omicron breakthrough infection, seven recognized a single dominant epitope in S, while CTL of two individuals recognized two independent epitopes. Due to limited sample availability, CTL fine specificity was not determined for the two remaining individuals. The distribution of HLA alleles and corresponding previously identified epitopes for those demonstrating CTL activity on day 10 are illustrated in Figure 4D and listed in Table 2, respectively.

Distinguishing factors associated with SARS-CoV-2 S-specific CTL activity

Since Omicron breakthrough infection expanded the number of individuals with detectable CTL against SARS-CoV-2 (from 6/22 to 11/22) and increased the median level of specific killing, we compared features impacted by breakthrough infection in groups distinguished by whether or not they had detectable SARS-CoV-2 S-specific CTL activity, using the standard cutoff of 10% specific lysis in ⁵¹Cr release assays (low vs high CTL activity; Figure 5A). Although Omicron breakthrough infection increased the frequency of polyfunctional circulating T cells and of those producing IFN-γ or IL-2 in response to SARS-CoV-2 peptide stimulation on day 1 (Figure 2), there was no significant difference in the magnitude of any of these responses between groups with or without detectable SARS-CoV-2 S-specific CTL activity (Figure 5B-D). In contrast, production of IFN-γ alone or of both IFN-γ and IL-2 by polyfunctional SARS-CoV-2 S-specific CD8⁺ T cells after 7 days in vitro stimulation following Omicron breakthrough infection was significantly higher in the group with detectable CTL activity (Figure 5E-F). There was no difference between the groups in the frequency of SARS-CoV-2 S-specific CD8⁺ T cells producing IL-2 alone (Figure 5G).

Figure 5.

Distinguishing features of participants with SARS-CoV-2 S-specific CTL activity. (A) Participants with three vaccinations then Omicron breakthrough infection (n = 22) were stratified into low CTL (< 10% specific lysis, orange) and high CTL (≥ 10% specific lysis, purple). (B-D) Circulating T cell cytokine responses (IFN-γ, dual IFN-γ/IL-2, IL-2) were assessed by 24 h ELISpot. (E-G) Cytokine production by expanded SARS-CoV-2 S-specific CD8⁺ T cells was measured after 5 h restimulation. P value in (A) was calculated using unpaired t-test and lines bisecting groups represent mean ± SD. Lines bisecting groups in (B-G) represent median with IQR and p values were calculated using Mann-Whitney test. The probability of a significant difference between groups compared is shown above lines spanning groups with ns posted when there was no significant difference.

Accessible soluble marker associates with SARS-CoV-2 S-specific CTL activity

Thorough evaluation of population-level immune protection against SARS-CoV-2 requires the identification of accessible and informative markers of cellular immunity and cytotoxic potential. With this in mind, we examined several immunological relationships, including antibody isotype, avidity, and S domain reactivity that might reflect generalized immune programming to favour cell-mediated killing. Although most of these relationships did not yield meaningful associations (data not shown), we observed that individuals in this study with detectable S-specific CTL activity had significantly higher levels of antibodies against a S protein determinant previously associated with robust ADCC [6] (Figure 6A). This relationship was reinforced when we compared antibody levels between a larger group (n = 50) with or without detectable SARS-CoV-2-specific CTL (Figure 6B-C). This suggests that antibodies against microbial determinants associated with robust ADCC may in some circumstances serve as an accessible marker for robust cell-mediated killing in general with potential relevance to cellular immune protection.

Figure 6.

Association between S-specific CTL activity and antibody reactivity. (A) Linear full length spike Ab epitope reactivity was measured in POMI samples from participants with three vaccinations (n = 22) using chemiluminescence ELISA-based peptide scanning as previously outlined [6,8]. Reactivity against a known ADCC-associated determinant was represented in groups stratified by CTL activity ≥ or < 10% specific lysis. (B) An expanded cohort (n = 50) was grouped by CTL activity and (C) their antibody reactivity against the same determinant represented by group. P values in (A, B) were calculated using unpaired t-test and lines bisecting groups represent mean ± SD. P value in (C) was calculated using Mann-Whitney test and the line bisecting groups represents median with IQR. The probability of a significant difference between groups compared is shown above lines spanning groups.

Discussion

For ease, convenience and throughput, adaptive immunity acquired through infection or vaccination is generally assessed at the population level by measuring antibodies against relevant microbial antigens. Depending on characteristics of the microbe and the host, antibodies present can be diagnostic of past infection, signify ongoing infection or imply protection against reactivation, future infection and illness. While antibodies against microbial neutralizing determinants are of special interest, for some microbes, their relevance can be challenged by microbial variability and mutational escape [2]. Antibodies and cytotoxic T cells that contribute to microbial clearance through cell-mediated killing of infected host cells tend to exhibit broader reactivity than neutralizing antibodies, but their assessment methods are more complex [3–6]. We previously showed that hybrid immunity and antibody reactivity against certain non-neutralizing determinants of SARS-CoV-2 S protein was associated with robust ADCC [6,8,9]. In the current study, we investigated relationships between factors affected by Omicron breakthrough infection following two or three vaccinations and SARS-CoV-2 S-specific CTL activity to identify predictors that may be useful for broad-based assessment of cell-mediated cytotoxicity against SARS-CoV-2 induced by vaccination and/or infection.

We measured circulating S-specific T cells producing IFN-γ, IL-2 or both cytokines before and after Omicron breakthrough infection. As the Public Health Agency of Canada recommended three SARS-CoV-2 vaccine doses at the time, we identified only six study participants experiencing Omicron breakthrough after just two vaccinations. This small sample size cannot adequately address issues of individual variability to allow for definitive conclusion that three vaccine doses prime CTL responses more effectively than two doses. Therefore, interpretation of data from this group should be considered preliminary in the context of the influence baseline immune status and HLA-diversity have on CTL response variability. Now that the first wave of Omicron has passed and most people have been vaccinated more than twice or been infected with Omicron, it is unlikely the group can be expanded. Although the size of the two-dose group is limited, the data acquired are consistent in suggesting that three doses is superior to prime T cells for durable expansion following infection. In the group with two vaccines, Omicron breakthrough did not increase the frequency of circulating polyfunctional, IFN-γ or IL-2-producing S-specific T cells, nor did it increase the frequency of polyfunctional, IFN-γ or IL-2-producing S-specific CD8⁺ T cells following seven days in vitro stimulation. Robust circulating T cell responses were observed with fewer than three vaccine doses, but in our study, three vaccinations were required to condition the immune system to significantly expand S-specific circulating T cell or in vitro-stimulated CD8⁺ T cell responses upon infection. ELISpot assays estimate the frequency of memory T cells activated by antigenic exposure and provide partial insight into their functional characteristics through cytokine production profiles, however, they do not reveal distinguishing phenotypic characteristics of the cytokine-producing cells. If suitable samples become available, further research should address whether the impact of a third SARS-CoV-2 vaccine on T cell responses is primarily quantitative, boosting memory T cell numbers, or qualitative, modulating memory T cell cytokine response profiles.

While the magnitude of circulating T cell responses against SARS-CoV-2 S did not differ between groups with and without detectable S-specific CTL, in vitro expansion of CD8⁺ T cells stimulated with S peptides was greater in the group that developed detectable S-specific CTL. This and the selective boosting of S-specific CTL activity in the group receiving three vaccines raises the possibility of selective induction of a population of progenitor CD8⁺ T cells primed to fuel robust CTL responses. Detecting this population might require separating CD4⁺ and CD8⁺ T cells before ELISpot assessment of circulating T cell responses or detailed phenotypic analysis of circulating S-specific CD8⁺ T cells using other methods incompatible with large-scale studies. Nonetheless, distinguishing this progenitor phenotype in lab-based research studies and determining requirements for its induction in terms of duration and extent of antigenic exposure could enhance vaccine design and public health vaccine distribution programmes. While there was no significant correlation between S-specific ELISpot responses in fresh PBMC and S-specific CD4⁺ T cell responses measured by flow cytometry after seven days in vitro stimulation, there was a significant correlation between S-specific ELISpot responses in fresh PBMC and in vitro CD8⁺ T cell responses post Omicron infection in the group with three vaccines. This is consistent with the possibility that progenitors of the polyfunctional CD8⁺ T cells and CTL detected following in vitro stimulation are present and may be detectable and measurable in circulating PBMC.s.

Limited in vitro expansion of S-specific CD8⁺ T cells after two vaccine doses and Omicron infection aligned with unchanged circulating T cell levels, again indicating that the third vaccine dose enhanced potential for memory T cell differentiation in vivo and in vitro. Only those participants with three vaccinations consistently showed detectable in vitro activation of S-specific CTL and a measurable impact of breakthrough infection on CTL levels. We confirmed recognition of four previously characterized CTL epitopes and identified two unreported SARS-CoV-2 S-derived CTL epitopes: GVYYPDKVF (HLA-B*15:01) and QSAPHGVVF (HLA-C*12:03). These findings expand the known repertoire of immunogenic regions within the SARS-CoV-2 S protein and highlight the utility of peptide matrix deconvolution in uncovering subdominant or previously unrecognized CTL targets. This is especially pertinent in the context of breakthrough infections where immune recall may be shaped by both vaccination and variant exposure. Identification of novel epitopes is also particularly relevant given the global heterogeneity in HLA and the ongoing evolution of SARS-CoV-2 variants. These epitopes may serve as valuable candidates for inclusion in next-generation vaccines aimed at eliciting broad and durable CD8⁺ T cell responses depending on how commonly they are recognized by other HLA-B*15:01 and C*12:03 individuals. Enlarging the library of confirmed CTL epitopes is helpful to design broadly immunogenic vaccines, especially when overlap between epitopes presented by different HLA types and highly immunogenic sequences is identified.

Among those with detectable CTL responses, we observed significantly greater in vitro expansion of polyfunctional and IFN-γ-producing CD8⁺ T cells after seven days in vitro stimulation and higher antibody reactivity to a S protein determinant associated with ADCC. Among features compared between groups with and without S-specific CTL activity, the level of antibody reactivity against this determinant associated with robust ADCC was the only feature we found significantly different. This relationship was reinforced by S-specific CTL data from additional participants. There is no obvious mechanism connecting induction of antibodies that elicit robust ADCC to induction of robust CTL activity and in cases where immunity stems from infection, the dominant CTL target is seldom the same antigen that triggers ADCC [20,21]. Cumulative antigenic load may be a determining factor as severe infection and more antigenic exposures both favour development of cell-mediated cytotoxicity [9]. Skewing of the response towards cell-mediated cytotoxicity may reflect immune adaptation to address extended microbial spread or persistence with the most effective immune functions for counteracting this. In cases where T cell responses against a microbe occur in the absence of seroconversion, these are generally CD4⁺ T_H1-type responses, supporting a requirement for exposure to more antigen over a longer period to stimulate CD8⁺ T cell responses [22–25]. Given that the role of immune responses is to return the system to equilibrium, positive feedback for both ADCC and CTL activity to eliminate infected cells may coalesce during a protracted immune response. The strong CTL and ADCC responses may be mechanistically unrelated and simply represent two distinct aspects of a robust antiviral immune response. Even without a mechanistic connection, establishing a link between antibody reactivity against determinants associated with robust ADCC and CTL activity in other infections could validate the level of these antibodies as a useful predictor of cell-mediated cytotoxicity.

As previously reported, we found that T cell responses induced by ancestral SARS-CoV-2 S protein were generally conserved against newer variants and expanded by Omicron breakthrough infection [3,4,26–28]. Overall, T cell responses induced by ancestral antigens were approximately 80% conserved against newer SARS-CoV-2 variants [3,4,26,28], however, approximately 20% of individuals in one study lost over half of their T cell reactivity against newer variants, primarily within CD8⁺ responses [27]. Several previous studies reported that detecting SARS-CoV-2 cytokine release is less sensitive, but more specific than AIM or proliferation assays, which yielded positive results with apparently unexposed persons [20,29]. We focused on responses defined by functional cytotoxicity assays against peptide-pulsed autologous target cells. Although this approach is less sensitive, it is differentially discerning compared to broad peptide pool-based assays measuring ex vivo activation-induced marker expression. The greater sensitivity of AIM assays compared to the cytotoxicity assay we employed to define CD8⁺ T cell epitopes underscores our identification of one to two epitopes within S recognized compared to an average of 1.6 epitopes/protein/HLA allele recognized by CD8 ⁺ T cells as estimated by AIM expression [20]. While 15% of a previously studied breakthrough cohort experiencing different histories of infection and vaccination demonstrated increased CD8⁺ T cell responses [26], our study offers preliminary evidence that three vaccines favours enhanced T cell responses following breakthrough infection more so than two vaccines. Vaccination guidelines and the confounding factor of undiagnosed previous infection make evaluation of the impact of two versus three vaccines on responses to breakthrough infection problematic. Although a different vaccine formulation, Chen et al. reported that a third dose of CoronVac was required to elicit broadly reactive immunity against SARS-CoV-2, including CD4⁺ and CD8⁺ T cell responses [30].

In line with other reports, our results suggest that a threshold level of antigenic exposure through vaccination and/or infection underlies the capacity to generate a robust CTL response potentially foretold by selective induction of circulating progenitors. Detection of these cells may require methodology incompatible with population-based studies, but their clear identification would benefit vaccination strategies. If the relationship we observed between SARS-CoV-2 S-specific CTL activity and the level of ADCC eliciting antibodies is mirrored in other viral infections, it could offer a readily accessible and practical method to gauge the potential for a robust CTL response. With increasing application of mRNA vaccine technology, strategies to capitalize on endogenous antigen presentation to CD8⁺ T cells through class I HLA antigens and generate CTL responses will be increasingly important.

Funding Statement

This work was supported by Canadian Institutes of Health Research: [Grant Number VR1 – 173202]; Public Health Agency of Canada: [Grant Number 2223-HQ-000284].

Author contributions

KAH conceived, designed and performed experiments, analyzed data, created figures and illustrations and co-wrote/edited the manuscript with MDG. MDG supported experimental design. DPI assisted with experiments. KMH, KEF, DAB and KAB processed samples. DH coordinated patient consent, questionnaires and collected blood samples. MDG and KAH obtained funding.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The data that support the findings of this study and preserve anonymity of study participants are available from the corresponding author, MDG, upon reasonable request.