If I were to make a movie about the COVID-19 pandemic, I would call it The Triumph of the Vaccines, and antibodies would sweep all of the top acting awards at the Oscars. In the first sequel — The Omicron Phase — new heroes are born, and the strong supporting roles of T cells in the COVID Vaccine Universe begin to get more interesting. In this issue of Nature Immunology, Minervina et al. do a deep character study of a group of very picky assassins known as The Killers1, but in a cliff-hanger ending, it looks as if The Killers may turn out to be heroes too, though we may not know for sure until the script for part 3 is developed.

The Killers, of course, are CD8+ cytotoxic T cells, and they have been extensively studied in SARS-CoV-2-infected people and in vaccine recipients2. CD8+ T cells target peptide fragments (epitopes) derived from proteins (antigens), presented on class I major histocompatibility complex molecules (known in humans as human leukocyte antigens, or HLAs). As of 9 March 2022, the Immune Epitope Database (IEDB3) had curated 1,316 unique class I HLA–epitope pairs, with numbered HLA-encoding genetic alleles mapped to four digits. Minervina et al. focused on characterizing a relatively small number of these pairs — a panel of 18 HLA–epitope pairs assembled from 5 HLA alleles, 6 spike epitopes and 12 non-spike epitopes, each previously confirmed in multiple publications — which allowed them to characterize the T cells in considerable depth. Focusing in this way allowed the group to assemble a high-quality panel of DNA-barcoded HLA–peptide multimers (staining reagents for the identification of antigen-specific T cells) that they then used to determine T cell specificities, antigen receptor sequences and transcriptomes at single-cell resolution. Others have also begun to use barcoded HLA multimer technology to study SARS-CoV-2-specific T cells4.

We are not yet to the point at which a comprehensive panel of barcoded multimers could practically be applied to study the entire spectrum of SARS-CoV-2 T cell responses — for example, the IEDB currently lists 220 epitopes restricted by HLA-A*02:01 — so prudence requires asking very pointed questions with this powerful tool. Minervina et al. focused on both sides of what has become known as hybrid immunity, induced by infection followed by vaccination or vice versa5, and their data shed light on what happens to SARS-CoV-2-specific CD8+ T cells under these conditions. Their experimental approach is outlined in Fig. 1.

Fig. 1: Structure of the study by Minervina et al.1.
figure 1

The schematic highlights the assembled cohorts, the parameters analyzed at the single-cell level and the different groups of T cells analyzed.

Immunologists are natural worrywarts, and it is easy for us to imagine all the bad things that could happen to T cells repeatedly exposed to antigen. The happy news from the cohorts studied by Minervina et al. is that none of these misfortunes occur to any great extent.

First, T cell exhaustion does not appear to be a matter of significant concern. Transcriptome analysis of dextramer-positive T cells using the uniform manifold approximation and projection (UMAP) dimensionality reduction approach revealed a cluster with an exhaustion-associated profile that was detected in about 50% of the donors, but usually at a low frequency compared with the entire dextramer-positive population. When longitudinal data were available, the frequency of cells in this cluster decreased over time following the last antigen exposure.

The next thing that T cell immunologists worry about is that repeated exposure to antigen might cause repertoire narrowing and that this might adversely affect the capacity of memory T cells to participate in recall responses to variants that do not completely evade specific responses. This has been observed previously by Dirk Busch and Eric Pamer in T cell responses of mice subjected to repeated exposure to Listeria monocytogenes6, but Minervina et al. looked for it in their data set and found no evidence for it.

Finally, Minervina et al. went one step further and asked whether mutations of the selected epitopes present in variants of concern (but curiously, and perhaps randomly, not in the Omicron variant) actually evaded recognition by a small subset of T cell clonotypes found in their dataset. Armed with TCR α and β sequences from their single-cell approach, the authors prepared transgenic T cell lines and tested them for reactivity to peptide-pulsed artificial antigen-presenting cells expressing relevant HLA alleles. There are all sorts of caveats with such experiments — the number of unique TCRs studied is small, the transgenic lines may be more sensitive than T cells in the host, the use of synthetic peptides circumvents the natural processing and presentation of epitopes in cells infected with a genuine SARS-CoV-2 virus — and there is no reasonable expectation that all of the results will turn out the same way. This was indeed the observation: some mutations abolished recognition by some T cell clones and not others, whereas other mutations abolished recognition by each of the small number of relevant T cell clones. The authors conclude that their data “suggest that the T cell memory repertoire established by SARSCoV-2 infection or vaccination has great cross-reactivity potential against novel viral variants.” Although this is a fair point and probably about as comforting as we should expect, in my mind, their approach to studying cross-reactivity may be too focused to make a significant contribution to the larger story that has been studied by others using more comprehensive methods, such as T cell stimulation assays with peptide pools covering the entire spike antigens of the ancestral and Omicron strains7.

With vaccines that contain only the spike antigen, there is an opportunity to study two separate groups of T cell: those specific for epitopes in the shared spike antigens, and those that target the non-spike antigens. When vaccination follows infection, the only antigen available during the second exposure is the spike antigen and, as expected, Minervina et al. found that the fraction of the measured response that is specific for spike increases with vaccination. This suggests that boosting has occurred, although this was less obvious from more conventional frequency-based measures of T cell responses that show little average change and a high degree of variability between cohort members. It is likely that a lot of this heterogeneity is due to the difficulties of obtaining samples at intervals that would be more informative and consistent. Limiting analysis to T cells in peripheral blood mononuclear cells and not from the site of infection probably also contributes to the variability.

As the pandemic rolls on, it has become possible to study T cell responses in individuals who were infected after vaccination, in so-called breakthrough infections. Given the extraordinary efficacy of the vaccines, it would be predicted that spike responses would dominate even in this scenario, but this is not what has been seen. Instead, there is evidence that the breakthrough infections drive primary responses to non-spike antigens that are equivalent to those seen in unvaccinated subjects. In infected–unvaccinated subjects, CD8+ cell responses are skewed toward epitopes from the nucleocapsid (N) protein8, perhaps due to more efficient processing and presentation of epitopes from the N protein than of those from the spike. In the breakthrough subjects, this advantage seems to hold, and is apparently not outweighed by the advantages that recall of spike-specific memory cells is presumed to enjoy. The broadening of targeted CD8+ cell specificities in the breakthrough subjects echoes what happens with antibody responses, wherein IgG antibodies to the SARS-CoV-2 N protein are induced. In other words, vaccination with a single antigen does not suppress immune responses to the full range of SARS-CoV-2 antigens upon breakthrough infections, and this is welcome news.

As the first Omicron wave wanes in much of the world, it is a good time to place what we now know about SARS-CoV-2-specific CD8+ T cells into context. Most immunologists agree that Omicron successfully evaded whatever protection the vaccines provided against infection with pre-Omicron strains. At the same time, the vaccine-induced protection against severe disease largely held, especially after a third vaccine dose. Protection against disease is often ascribed to adaptive cellular immune responses, but proof that vaccine-induced cellular responses are (or are not) responsible for protection against COVID-19 disease in humans has not been achieved so far. In the case of COVID-19, even proving correlates of protection against disease in humans is challenging, in part as a result of high vaccine efficacy.

So what are the candidates for future movies in the COVID-19 Vaccine Universe? Fans of T cells are probably hoping to see a story focusing on the introduction of additional antigens into current vaccine formulations, such as constructs encoding the nucleocapsid antigen9. Waning immunity and the emergence of variants of concern may even provide fertile ground for testing the efficacy of this path, and the ability to detect an increase in efficacy following a third dose of the current vaccines provides a useful basis for comparison. As a testable scientific hypothesis, its merit is undebatable. In contrast, what remains highly debatable is where that path should lie on our priority list. Strong observational studies of T cell responses in SARS-CoV-2-infected and vaccinated people may be insufficient to push this path upward on the priority list.