From May 2021 to January 2022, we prospectively enrolled study participants aged ≥18 years with reverse transcription–polymerase chain reaction (RT-PCR)–confirmed SARS-CoV-2 infection who were admitted to the biocontainment units of Seoul National University Hospital or Boramae Medical Center. Uninfected vaccinated individuals were also enrolled. The uninfected vaccinated individuals were enrolled from the non-hospitalized general population. Breakthrough COVID-19 was defined by the presence of COVID-19 symptoms and RT-PCR–confirmed diagnosis of COVID-19 more than 14 days after at least one vaccine dose. Patients with breakthrough COVID-19 were divided into a severe group who required oxygen therapy and a non-severe group who did not require supplemental oxygen therapy during hospitalization (18). The uninfected vaccinated individuals served as a control group.

Within 1 week of onset of symptoms, blood samples were collected from patients hospitalized with COVID-19. Additional serial samples were obtained from a portion of the patients. Blood samples were also collected from unvaccinated patients with severe COVID-19 and uninfected vaccine recipients. Data were collected regarding demographic characteristics, vaccination type and status, days from vaccination to symptom onset, Charlson’s comorbidity index, underlying diseases, and clinical outcomes. Fully vaccinated patients were defined as those with COVID-19 diagnosis more than 14 days after completion of the recommended vaccination regimen.

Anti-S1 (spike subunit) IgG titer was semi-quantitatively measured using an enzyme immunoassay kit (Euroimmun, Lübeck, Germany) according to the manufacturer’s protocol. Optical density (O.D.) ratios were interpreted as follows according to the instructions: <0.8, negative; ≥0.8 to <1.1, borderline; and ≥1.1, positive.

The binding activity of serum antibodies to SARS-CoV-2 receptor-binding domain (RBD) proteins was determined using an ELISA (19, 20). ELISA plates were coated with 100 ng/well of RBD protein, blocked with phosphate-buffered saline (PBS) supplemented with 3% bovine serum albumin, and incubated with diluted serum (1:200) for 2 h. Bound antibodies were detected using horseradish peroxidase–conjugated goat anti-human IgG (Fc) (#ARG23874, 1:12,000, Arigo Biolaboratories, Hsinchu, Taiwan). After washing with PBST three times, 50 μL of 3,3’,5,5’-tetramethyl benzidine was added, followed by the addition of 50 μL of 2 M H₂SO₄ to stop the reaction. Absorbance was measured at 450 nm using an Infinite M200 PRO (TECAN, Zurich, Switzerland).

After whole blood was drawn into heparin vacutainers (Becton Dickinson, NJ, USA), PBMCs were purified using Ficoll-Hypaque (GE Healthcare Life Sciences, Piscataway, NJ, USA). PBMCs were stored in serum-free cryopreservation medium (Cellbanker 2; Zenoaq, Japan) in liquid nitrogen containers until further use.

After thawing, cells were cultured in the presence of 1% penicillin/streptomycin (Thermo Fisher Scientific, Waltham, MA, USA) in complete RPMI-1640 medium containing 10% fetal bovine serum. Thereafter, 1 × 10⁶ PBMCs/mL were stimulated with 0.06 nmol/mL PepTivator^® SARS-CoV-2 Prot_S Complete (i.e., whole-spike Ag), PepTivator^® SARS-CoV-2 Prot_S B.1.617.2 wild-type (WT) reference (i.e., WT Ag), or PepTivator^® SARS-CoV-2 Prot_S B.1.617.2 Mutation Pool (i.e., Delta Ag) (Miltenyi Biotec, Bergisch Gladbach, Germany) for 24 h. Medium alone was used as a negative control. In addition to the antigens, Brilliant Blue 515–anti-human CD4 (clone RPA-T4) antibodies (BD Biosciences, San Jose, CA, USA) for staining CD4 and anti-human CD28/CD49d (clone L293/L25, BD Biosciences) antibodies for co-stimulation were simultaneously added. For the final 4 h of antigen stimulation, cells were treated with BD GolgiPlug^® (brefeldin A) and BD GolgiStop^® (monensin) (all from BD Biosciences).

Dead cells were stained with LIVE/DEAD (Thermo Fisher Scientific) after stimulation. Cells were permeabilized before incubation with peridinin chlorophyll protein complex–anti-human CD8 (clone SK1), Brilliant Violet (BV) 510–anti-human CD3 (clone CHT1), BV605–anti-human CD69 (clone FN50), BUV395–anti-human CD137 (clone 4B4-1), phycoerythrin–indotricarbocyanine (Cy7)–anti-human IFN-γ (clone B27), allophycocyanin–anti-human interleukin(IL)-2 (clone 5344.111), phycoerythrin–anti-human tumor necrosis factor-α (clone Mab11), and BV421–anti-human IL-4 (clone MP4-25D2) antibodies (all from BD Biosciences). Each sample was treated with BD Horizon Brilliant Stain Buffer (BD Biosciences). In each experiment, compensation beads (UltraComp eBeads, Thermo Fisher Scientific) and unstimulated cells were used for compensation. Flow cytometric data were acquired on a FACSymphony system (BD Biosciences) and analyzed using FlowJo software (version 10.7.1; TreeStar, Ashland, OR, USA).

The frequencies of SARS-CoV-2–specific activation-induced marker⁺ (AIM⁺, CD69⁺CD137⁺) CD4⁺ T cells or CD8⁺ T cells and cytokine-producing CD137⁺CD4⁺ or CD137⁺CD8⁺ T cells were assessed (21). To analyze only the SARS-CoV-2–specific response, the percentages of target populations in specimens without antigen stimulation were subtracted from those in stimulated specimens (22). The flow cytometry gating strategy for SARS-CoV-2–reactive T cells and cytokine-producing T cells is shown in Supplementary Figure 1.

The chi-squared test or Fisher’s exact test was used to compare categorical variables, and the Mann-Whitney U test was used to compare continuous variables. To identify risk factors for severe breakthrough COVID-19, variables with a P value of <0.10 in the univariate analysis were included in the multivariable logistic regression analysis. Statistical analyses were performed using SPSS Statistics, version 26.0 (IBM Corp., Armonk, NY, USA). P values <0.05 were considered statistically significant. Data are presented as median with interquartile range (IQR) and as dot plots. All graphs were generated using GraphPad Prism 9 (GraphPad Software, La Jolla, CA, USA).

A total of 108 breakthrough COVID-19 cases admitted within 1 week of the onset of symptoms were enrolled. The non-severe group included 79 (73.1%) patients, and the severe group included 29 (26.9%) patients. In addition, the control group included 22 uninfected individuals who completed the standard doses of SARS-CoV-2 vaccine (Table 1).

The severe group was older (median [IQR], 71 [63–78] years vs. 61 [47–68] years, P<0.001) and had higher Charlson’s comorbidity index (median [IQR], 4.0 [3.0–5.0] vs. 2.0 [0.0–4.0], P<0.001) and longer duration of admission (median [IQR], 9 [8–12] vs. 9 [7–10], P = 0.035) than the non-severe group. Vaccination type and status did not significantly differ between the two groups. The control group was younger than the severe group (median [IQR], 62 [47–73] years vs. 71 [63–78] years, P = 0.006), and the interval between vaccination and sample collection was shorter in the control group than in the non-severe and severe groups.

Anti-S1 IgG titer within 1 week after symptom onset in breakthrough COVID-19 patients was significantly lower in the severe group than in the non-severe group (median O.D. ratio [IQR], 4.99 [1.38–9.02] vs. 8.28 [5.17–10.25], P = 0.007) (Figure 1A). Compared with the uninfected vaccinated control group, the anti-S1 IgG titer was significantly higher in the non-severe group (median [IQR], 4.22 [2.46–6.13] vs. 8.28 [5.17–10.25], P<0.001) but similar in the severe group (median [IQR], 4.22 [2.46–6.13] vs. 4.99 [1.38–9.02], P = 0.591). Anti-S1 IgG titer according to the number of days after symptom onset was significantly higher in the non-severe group than in the severe group on days 5–7 after symptom onset (median [IQR], 8.50 [7.15–9.39] vs. 5.80 [0.55–8.98], P = 0.014) but did not differ significantly between the two groups on days 1–4 after symptom onset (median [IQR], 7.31 [2.77–10.33] vs. 4.81 [2.64–10.73], P = 0.285) (Figure 1B).

Serum antibody responses to SARS-CoV-2 RBD_WT and RBD_Delta were compared between groups (Figures 1C, D), but two cases from the non-severe group were excluded due to shortage of samples. Titers of both anti-RBD_WT IgG and anti-RBD_Delta IgG were significantly lower in the severe group than in the non-severe group (median [IQR], 0.26 [0.13–0.60] vs. 0.48 [0.24–0.99], P = 0.017; 0.22 [0.13–0.64] vs. 0.40 [0.22–1.12], P = 0.007). Compared with the control group, the non-severe group showed significantly higher antibody responses to both RBD_WT and RBD_Delta (median [IQR], 0.24 [0.18–0.52] vs. 0.48 [0.24–0.99], P = 0.004; 0.19 [0.14–0.42] vs. 0.40 [0.22–1.12], P<0.001), but the severe group showed similar antibody responses to both RBD_WT and RBD_Delta.

The kinetics of antibody responses over time beginning at symptom onset were evaluated in 17 severe breakthrough cases, 10 non-severe breakthrough cases, and 7 unvaccinated severe cases (Figure 2). In non-severe breakthrough cases, the anti-S1 IgG titer was high at the time of symptom onset or elevated during the early phase of infection. However, the anti-S1 IgG titer increased later in some of severe breakthrough cases. In two severe breakthrough cases, the antibody titer began to increase 2 weeks after symptom onset, similar to unvaccinated severe cases. One of these two severe cases had received an immunosuppressant.

T cell–mediated immune responses against the SARS-CoV-2 WT and Delta variant were assessed using flow cytometry in 10 non-severe breakthrough cases, 5 severe breakthrough cases, and 10 uninfected vaccinated individuals. Clinical data for the study participants whose cellular immune responses were analyzed are provided in Supplementary Table 1. Among severe cases, all PBMCs used in the analysis of cellular immune responses were isolated from blood samples collected before oxygen therapy.

Compared with uninfected vaccinated controls, the proportion of AIM⁺ CD4⁺ T cells specific for whole-spike Ag was significantly lower in both the non-severe breakthrough cases (median [IQR], 0.27% [0.12–0.36%] vs. 0.06% [0.03–0.19%], P = 0.036) and severe breakthrough cases (median [IQR], 0.27% [0.12–0.36%] vs. 0.02% [0.02–0.18%], P = 0.019) (Figure 3A). The proportion of AIM⁺ CD4⁺ T cells specific for the WT and Delta Ags tended to be lower in the severe group than in the non-severe group, although the difference was not statistically significant. In addition, the proportions of cytokine-producing CD4⁺ T cells tended to be lower in the severe group than in the non-severe group (Supplementary Figure 2). Although the differences were not statistically significant, the proportions of AIM⁺ CD8⁺ T cells specific for the WT and Delta Ags tended to be lower in the severe group than in the non-severe group (Figure 3B).

The univariate analysis identified older age, higher Charlson’s comorbidity index, and lower anti-S1 IgG titer in the early phase (≤1 week after symptom onset) as risk factors for severe breakthrough COVID-19 (Table 2). The multivariate analysis identified older age (adjusted odds ratio [aOR] 1.07, 95% confidence interval [CI] 1.01–1.12, P = 0.014) and low anti-S1 IgG titer in the early phase (aOR 0.88, 95% CI 0.77–0.99, P = 0.037) as independent risk factors for severe breakthrough COVID-19.

The results of the multivariate linear regression model are shown in Supplementary Table 2. Among breakthrough COVID-19 patients, anti-S1 IgG titer was negatively correlated with severe COVID-19 (standardized β, −0.227, P = 0.025) and male sex (standardized β, −0.211, P = 0.029).