The research team, led by Frederica Dedo Partey of the Noguchi Memorial Institute for Medical Research at the University of Ghana, recruited healthy adults aged 18 to 60 from the Legon community in Accra who had completed a primary vaccination series at least six months earlier with one of four vaccines: AstraZeneca’s ChAdOx1-S, Moderna’s mRNA-1273, Pfizer-BioNTech’s BNT162b2, or Janssen’s Ad26.COV2.S. Blood was drawn at four time points — before the booster and at three, six, and nine months afterward — and peripheral blood mononuclear cells were isolated by density-gradient centrifugation, cryopreserved in liquid nitrogen, and later thawed for functional testing. A critical feature of the cohort is that every participant was nucleocapsid-seropositive at enrollment, indicating prior natural infection, and antibody levels remained stable throughout follow-up, confirming that no new infections occurred during the study. The findings therefore describe how boosters reshape cellular immunity on a background of hybrid immunity — the combination of vaccination and prior infection that most of the world’s population now carries.

Rather than testing crude pools of viral peptides, the researchers took a precision approach. They used NetMHCpan-4.1, a machine-learning tool trained on peptide binding and eluted-ligand data, to predict 10-mer peptides from the spike and nucleocapsid proteins that bind strongly to seven globally prevalent HLA class I supertypes: A01, A02, A03, A24, B07, B40, and B58. Because participants were not HLA-typed, selecting epitopes across these supertypes maximized population coverage and was particularly relevant for a West African cohort with extensive HLA diversity. The predicted epitopes were matched to pre-synthesized overlapping 15-mer peptides spanning the full spike protein (1,273 amino acids) and nucleocapsid protein (419 amino acids), based on the Wuhan reference strain. Each peptide was then tested individually in an ex vivo interferon-gamma ELISpot assay, with 200,000 peripheral blood mononuclear cells per well stimulated for 24 hours. This single-peptide design meant that every positive response could be assigned to a defined viral sequence — a level of resolution that proved essential to the study’s central discovery.

That discovery concerns what actually drives the decline of T-cell immunity over time. Across the whole cohort, time had a powerful effect on total spike-specific interferon-gamma responses, with an omnibus F-test yielding F(3, 96.6) = 33.16 and a p-value of 8 × 10⁻¹⁵. Yet the median per-peptide response — the strength of each individual epitope-specific reaction — did not change significantly over the same period (F(3, 95.8) = 0.62, p = 0.606). The dissociation is striking: the waning of T-cell immunity after boosting was not caused by individual T-cell responses growing weaker, but by the immune system progressively dropping entire epitope specificities from its repertoire. The same pattern held within both booster groups separately, and the residual response stabilized after month six, with no significant difference between months six and nine.

The two booster platforms produced measurably different immune profiles. Participants boosted with the adenoviral vector Ad26.COV2.S showed significantly broader spike-specific epitope recognition — 70 positive peptide responses at month three compared with 34 in the BNT162b2 group, a difference with a p-value of 4.4 × 10⁻⁵ — and a subset of their responses persisted through month nine. The BNT162b2 recipients, by contrast, mounted higher cumulative interferon-gamma magnitude through month six, an advantage that was already present before the booster was administered (p = 0.012), suggesting it reflected their prior vaccination history rather than the booster itself. By month nine the picture had reversed, with the adenoviral-vector group showing higher — though not statistically significant — interferon-gamma levels. The authors propose a biological explanation rooted in platform mechanics: mRNA is inherently transient and subject to interferon-driven degradation, limiting the duration of antigen presentation, whereas adenoviral vectors support more prolonged antigen expression that may recruit subdominant T-cell clones and sustain a wider memory repertoire.

Vaccination strategy — whether the primary series and booster used the same platform — also left its mark. Within the Ad26.COV2.S group, homologous boosting (viral vector followed by viral vector) was associated with significantly broader epitope recognition at months zero, three, and nine, and this subgroup contributed the highest proportion of persistent responses at months six and nine. Two participants, J028 and J171, recognized at least one spike epitope at every time point, with J171 alone accounting for eight of the ten positive responses recorded at month nine. In the BNT162b2 group, no homologously boosted participant showed sustained responses across all post-booster time points, and no significant differences emerged between homologous and heterologous subgroups — a null result the authors attribute partly to limited statistical power. Notably, the finding that homologous adenoviral boosting outperformed heterologous regimens in breadth runs counter to the common assumption that anti-vector immunity undermines repeat adenoviral dosing.

Mapping the responses onto the spike protein revealed a dynamic immunodominance hierarchy. At baseline, recognition clustered in the N-terminal domain and the fusion peptide region — structurally exposed domains previously shown to harbor conserved T-cell epitopes. After boosting, recognition shifted toward the receptor-binding domain and subdomains 1/2, with peptides such as KLPDDFTGCV (amino acids 421–435) and IHADQLTPTW (amino acids 621–635) among the most frequently recognized at month three. By month nine, only nine spike peptides remained immunoreactive: three each from the N-terminal domain and fusion peptide region, and one each from the receptor-binding domain, subdomain 1/2, and the buried heptad repeat 2. The persistence of N-terminal and fusion-derived epitopes underscores their structural stability, while the rare recognition of a heptad repeat 2 peptide — a region largely hidden in the prefusion spike — hints at unusual HLA binding or exposure to post-fusion conformations.

The nucleocapsid arm of the study offered a natural experiment in memory maintenance without antigen restimulation. Because all administered vaccines encode only the spike protein, any nucleocapsid-specific T-cell responses must have been acquired through prior infection. Despite stable anti-nucleocapsid antibodies and no incident infection, these responses declined from 57 positive peptide reactions at baseline to just three at month nine — evidence that spike-based boosting does nothing to sustain cellular memory against non-spike antigens. Yet two nucleocapsid regions, YGANKDGIIW (amino acids 121–135) and VTPSGTWLTY (amino acids 321–335), were recognized at every one of the four time points. Their repeated immunogenicity, combined with the nucleocapsid protein’s high sequence conservation across coronaviruses, makes them leading candidates for inclusion in multi-epitope T-cell vaccines.

The authors are candid about the study’s limitations. The timing and severity of participants’ pre-enrollment infections could not be established; the interval between primary vaccination and booster ranged from six to 24 months; HLA typing was not performed, so genotype-specific responses could not be assessed; and months six and nine testing was restricted to peptides positive at month three, meaning delayed or newly emerging responses would have been missed. The observed contraction in breadth should therefore be read as a narrowing among the month-three-positive epitopes rather than definitive proof of global epitope loss. Subgroup comparisons were also modestly powered. Even so, the study’s central message stands: the platform and strategy chosen for boosting shape not just how strong T-cell immunity is, but how wide and how long-lasting it remains. For next-generation vaccine design, the implication is clear — durable cellular protection may depend less on maximizing peak response magnitude than on preserving epitope breadth, and conserved targets in the spike N-terminal domain, the fusion peptide, and the nucleocapsid protein deserve priority in the vaccines meant to outpace viral evolution.

Subject of Research: Longitudinal durability and epitope breadth of SARS-CoV-2-specific T-cell responses after mRNA versus adenoviral-vector booster vaccination

Article Title: T-cell recognition of SARS-CoV-2 antigen peptides following booster vaccination with mRNA and viral-vectored COVID-19 vaccines

Article References: Maru, A. K., Kumi, P. K., Pobee, A. N. A., Appiah, N. Y. A., Koranteng, N. A. O., Brenko, T., Osei, F., Teye-Adjei, D., Agyekum, G., Akuffo, L., Dzabeng, F., Kyei-Baafour, E., Obiri-Yeboah, D., Aniweh, Y., Ofori, M. F., Kusi, K. A., & Partey, F. D. (2026). T-cell recognition of SARS-CoV-2 antigen peptides following booster vaccination with mRNA and viral-vectored COVID-19 vaccines. Frontiers in Cellular and Infection Microbiology, 16, Article 1947651. https://doi.org/10.3389/fcimb.2026.1947651

Image Credits: AI Generated

DOI: 10.3389/fcimb.2026.1947651

Keywords: SARS-CoV-2, T-cell immunity, booster vaccination, BNT162b2, Ad26.COV2.S, ELISpot, epitope breadth, hybrid immunity, nucleocapsid protein, spike protein, HLA class I, vaccine design