Waning cellular immune responses and predictive factors in maintaining cellular immunity against SARS-CoV-2 six months after BNT162b2 mRNA vaccination

Several clinical trials have shown that the humoral response produced by anti-spike antibodies elicited by coronavirus disease 2019 (COVID-19) vaccines gradually declines. The kinetics, durability and influence of epidemiological and clinical factors on cellular immunity have not been fully elucidated. We analyzed cellular immune responses elicited by BNT162b2 mRNA vaccines in 321 health care workers using whole blood interferon-gamma (IFN-γ) release assays. IFN-γ, induced by CD4 + and CD8 + T cells stimulated with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) spike epitopes (Ag2), levels were highest at 3 weeks after the second vaccination (6 W) and decreased by 37.4% at 3 months (4 M) and 60.0% at 6 months (7 M), the decline of which seemed slower than that of anti-spike antibody levels. Multiple regression analysis revealed that the levels of IFN-γ induced by Ag2 at 7 M were significantly correlated with age, dyslipidemia, focal adverse reactions to full vaccination, lymphocyte and monocyte counts in whole blood, Ag2 levels before the second vaccination, and Ag2 levels at 6 W. We clarified the dynamics and predictive factors for the long-lasting effects of cellular immune responses. The results emphasize the need for a booster vaccine from the perspective of SARS-CoV-2 vaccine-elicited cellular immunity.

. Characteristics of the study participants who had underwent a series of blood collections. Health care workers were classified as either medical or nonmedical staff. The medical staff included nurses, physicians, pharmacists, clinical laboratory technicians, radiologists, and medical students. Nonmedical staff included administrative workers, medical clerks, and other staff members who did not have direct contact with patients. BMI body mass index. www.nature.com/scientificreports/ the participants reported no underlying medical problems. Only 10 (3.1%) participants reported active smoking, and 28 (8.7%) reported past smoking. A total of 179 (55.7%) participants reported social drinking, and 62 (19.3%) reported daily alcohol consumption. Twelve participants who answered that they had been diagnosed with COVID-19 or showed a high titer of anti-nucleocapsid SARS-CoV-2 IgG antibodies (10.0 AU/mL <) during the observation period were analyzed separately; the other 309 participants were analyzed first.

Cellular immune responses and humoral responses to vaccination. SARS-CoV-2-specific T-cell
responses were measured via IGRA to evaluate cellular immunity. We measured IFN-γ levels using the Ag1 tubes which consist of epitopes of CD4 + T cells derived from the S1 subunit of the spike protein and the Ag2 tubes which consist of epitopes of CD4 + and CD8 + T cells derived from the S1 and S2 subunits. The time course of IFN-γ levels in response to Ag1 and Ag2 is shown in Fig. 1 Ag2 IFN-(IU/ml) Figure 1. Dynamics of the cellular immune response as the IFN-γ level changes. Blood samples were collected and stimulated with Ag1 (upper) and Ag2 (lower) derived from the SARS-CoV-2 spike protein before and after full vaccination. The time-dependent change of IFN-γ levels, induced by the incubation of lymphocytes with SARS-CoV-2-specific epitopes of CD4 + (Ag1) or CD4 + and CD8 + (Ag2) T cells, were demonstrated. Baseline indicates before the first dose of vaccination, and 3 W means before the second dose. Then IFN-γ levels were also measured in blood samples 3 weeks (6 W), 2 months (3 M), 3 months (4 M), and 6 months (7 M) after the second vaccination. Box plots show the median and interquartile range associated with the min and max. The Friedman test was used to calculate the P values (****P < 0.0001). www.nature.com/scientificreports/ Regarding humoral immune responses, all participants showed increased levels of anti-spike IgG antibodies above the cutoff level (50.0 AU/mL) at 6 W, with a median level of 13,200 (IQR,200) AU/mL. However, at 7 M, the median level was 839 AU/mL (IQR, 562-1280 AU/mL). Hence, a 93.6% decrease in the antibody titer was observed after 6 months (Fig. 2).

Correlation between the humoral and cellular immune responses after vaccination.
The timedependent attenuation of anti-spike IgG antibody levels was faster than that of IFN-γ levels measured via IGRA, as shown in Figs. 1 and 2. We compared the correlation between anti-spike IgG antibody and IFN-γ levels in response to Ag2 at 6 W and 7 M (Fig. 3). Anti-spike IgG antibody levels were attenuated in all participants, whereas some participants (13.5%) showed elevated IFN-γ levels from 6 W to 7 M. These correlations were very weak at 6 W (R 2 = 0.016, P = 0.025). The regression coefficient at 7 M was increased but still weak compared to that at 6 W (R 2 = 0.033, P = 0.001).  www.nature.com/scientificreports/ analysis are shown in Supplementary Tables S1 and S2. The results of the multiple regression analysis are shown in Table 2. Age (P = 0.047), Ag1 levels at 3 W (P < 0.001), and Ag1 levels at 6 W (P < 0.001) were significantly associated with Ag1 levels at 7 M, while age (P = 0.017), dyslipidemia (P = 0.039), focal adverse reactions only (P = 0.020), lymphocyte (P = 0.012) and monocyte (P = 0.010) counts, Ag2 levels at 3 W (P < 0.001), and Ag2 levels at 6 W (P < 0.001) were significantly associated with Ag2 levels at 7 M. As expected, Ag1 and Ag2 levels were strongly correlated with each other at 7 M. Interestingly, Ag1 and Ag2 levels at 3 W showed a stronger correlation with the level at 7 M than with that at 6 W (partial regression coefficient, 1.366 vs. 1.130 for Ag1 and 1.315 vs. 1.071 for Ag2).
Factors associated with a positive IFN-γ response measured via IGRA 6 months after full vaccination. Multivariate logistic regression analysis was performed to assess the factors associated with a positive response (≥ 0.15 IU/mL IFN-γ) in SARS-CoV-2-specific cellular immunity at 7 M (   Fig. 4), whereas IFN-γ levels and the levels of anti-spike antibodies in response to Ag1 were decreased at 7 M and showed no significant increase compared with those of the other participants.

Dynamics of vaccine-associated immune responses in participants with a positive anti-nucleocapsid antibody titer prevaccination.
We analyzed immune responses in participants with a positive anti-nucleocapsid antibody titer prevaccination. Five participants were selected, and three had a history of COVID-19 infection with no treatment needed. The time course of their immune responses is described in Table 4. One patient had a positive response for both Ag1 and Ag2 at baseline. All patients showed a positive cellular immune response prior to the second vaccination (3 W). Interestingly, four participants (80%) exhibited a significant decline in the cellular immune response at 6 W compared to 3 W. At 7 M, participants still had positive IFN-γ levels, and both anti-spike IgG antibody and IFN-γ levels were significantly elevated in these participants compared with those with no prior elevation of anti-nucleocapsid antibody levels (P < 0.01).

Discussion
We monitored and analyzed the cellular immune response elicited by the BNT162b2 mRNA vaccine until 6 months after the second vaccination. IFN-γ, induced by the incubation of lymphocytes with SARS-CoV-2-specific epitopes of CD4 + (Ag1) or CD4 + and CD8 + (Ag2) T cells, peaked at 6 W and typically decreased between 6 W, 3 M, and 7 M. The degree of reduction in IFN-γ levels in response to Ag1 and Ag2 from 6 W to 7 M was 54.9% and 60.3%, respectively, whereas that of the anti-spike IgG antibody levels was 93.7%. This observation that the IFN-γ titer tends to persist at 6 months after full vaccination compared with the anti-spike antibody titer is consistent with another similar study using the BNT162b2 mRNA vaccine 23,24 .   www.nature.com/scientificreports/ The attenuation of cellular and humoral immune responses was similar to that reported in previous studies conducted with health care workers 22 . A similar decline in anti-spike IgG antibody levels was also observed in other studies with the BNT162b2 mRNA vaccine 19,25 , except among elderly people and people with immunosuppressive conditions. Between 3 and 4 M, the anti-spike antibody titer decreased by 44%, and the IFN-γ titer in response to Ag2 decreased by 5.2%, while between 4 and 7 M, the antibody titer decreased by 71.9%, and the IFN-γ titer decreased by 27.6%. Our data demonstrated that the decline in both humoral and cellular immune responses was relatively high between 4 and 7 M compared with that between 3 and 4 M. This observation is consistent with evidence that the rate ratio for SARS-CoV-2 infection and severe diseases increased 6 months after full vaccination compared with that after 4 months 26 . These results emphasize the need to administer a booster dose to induce sufficient immunity for disease prevention 27,28 .
Regarding the correlation of the levels of anti-spike IgG antibodies and IFN-γ, we observed only a weak correlation at both 3 M and 7 M. Another study conducted with health care workers also showed a stronger correlation between these levels 6 months after full vaccination compared with that 3 months after vaccination 21 , but correlations at both times seemed stronger than those in our study. Another study in Japan also noted a very weak correlation between these levels, similar to that in our study 29 ; this difference may be attributed to differences in the population. Between 4 and 7 M, six participants were positive for anti-nucleocapsid antibodies (excluded from the main analysis), with suspicion of breakthrough infection. Their humoral and cellular immune responses were not impaired at 4 M compared with those of the other participants, and participants with confirmed COVID-19 infection between 4 and 7 M had mild symptoms and did not require antiviral or steroid therapy (data not shown).
At baseline, IFN-γ levels in response to Ag2 were above the cutoff in 18 (5.8%) participants, without an increase in anti-nucleocapsid antibodies. They still had significantly higher IFN-γ levels in response to Ag2 but the levels response to Ag1 were not significant at 7 M compared with those below the cutoff level. In fact, 6/18 (33.3%) participants had higher IFN-γ levels at 7 M than at 4 M, whereas we found the same tendency in 79/291 (27.1%) participants with levels below the cutoff at baseline, suggesting that circulating activated memory T cells specific for SARS-CoV-2 could be maintained longer in participants with positive responses to Ag2 at baseline. Our results suggest that blood CD8 + T cells in some people already showed reactivity with epitopes from the SARS-CoV-2 spike protein before vaccination, and they had a long-lasting cellular immune response. This early reactivity could presumably be attributed to past exposure to seasonal coronaviruses, which have conserved peptides 30,31 . Other studies have demonstrated that these cells might be important for immune defense against SARS-CoV-2, as blood CD8 + T cells with these conserved epitopes are associated with mild symptoms in patients with COVID-19 infection 30 and long-lasting protective immunity 32 . Considering that cellular immune responses, particularly CD8 + T-cell responses, contribute to protection against severe SARS-CoV-2 infection 16 , people with natural CD8 + T-cell responses to SARS-CoV-2 may be protected against developing severe disease. Further studies are needed to clarify the distinct phenotype of T cells that contributes to evoking effective cellular immunity.
Participants with a positive anti-nucleocapsid antibody titer at baseline showed a positive cellular immune response with the first vaccination, but 80% of them showed a decline in the response to the second vaccine. Similar results regarding vaccination with subjects who recovered from COVID-19 infection have been reported both in cellular and humoral immune responses 33,34 . Regarding humoral responses, memory B-cell responses one week after the first vaccination were boosted in people who had recovered from COVID-19 infection, while naive individuals required two doses to reach comparable memory B-cell levels 34 . Regarding cellular responses, naive individuals exhibited higher SARS-CoV-2-specific CD4 + T-cell activation and proliferation than individuals who had recovered from COVID-19 infection 2 weeks after the second vaccination 33 . Consistent with previous studies, our study implicates that the first vaccination works as a booster and that no further immune response is observed after the second vaccination in participants with prior COVID-19 infection, and the response decreases during the time course. Regulatory T cells may be also involved in this effect 33 . Although it is questionable whether memory T cells are activated sufficiently with one vaccination in individuals with prior COVID-19 infection compared with those in fully vaccinated naive individuals, in our results, both cellular and humoral responses were relatively high in individuals with prior COVID-19 infection compared with those in naive individuals at 7 M.
To identify the factors that could predict a strong cellular immune response, we performed a multivariate analysis. Older age and dyslipidemia were significantly associated with a strong cellular immune response but not with positive responses at 7 M. Focal adverse reactions after the second vaccination and the blood lymphocyte and monocyte counts were associated with strong and positive cellular immune responses at 7 M. The younger age is associated with the stronger humoral response at earlier stages after full vaccination. However, it is controversial whether the difference is maintained after 6-9 months 35,36 . Other authors demonstrated that low high-density lipoprotein cholesterol or high triglycerides were not associated with the results of IGRA for tuberculosis 37 . Regarding adverse effects, our previous study demonstrated that participants reporting stronger reactions after the second dose than after the first dose showed significantly higher levels of the anti-spike IgG antibody 15 , implying that relatively strong adverse reactions are possibly associated with both strong humoral and cellular immunity induced by BNT162b2 vaccines. In another study, vaccinated individuals with general fatigue and fever as adverse events had stronger cellular immune responses than those without them 8 weeks after the second vaccination 29 . Our study suggests that systemic adverse reactions after the second dose and the lymphocyte count at baseline were positively associated with strong and positive cellular immune responses at 7 M. Other studies demonstrated that humoral responses correlated with higher B-, NK-and CD4-cell counts in the vaccination period, and a low CD4 + /CD8 + ratio was correlated with cellular response failure in patients after allogenic hematopoietic stem cell transplantations 38,39 . Further studies are needed to determine whether the number of lymphocytes correlates with that of memory CD4 + and CD8 + T cells that can react with SARS-CoV-2 epitopes. As expected, Ag1 and Ag2 levels were strongly associated with each other at 7 M, but Ag1 and www.nature.com/scientificreports/ Ag2 levels at 3 W had a stronger association with those at 7 M than at 6 W. These results suggest that the first dose is a better predictor of long-lasting cellular immunity and memory T-cell activity than the second dose, which could induce strong adaptive immunity but result in temporally strong immune responses, and differences between individuals might be more remarkable after the second dose than those after the first dose. We have to note that other factors we didn't investigate might affect our results. For example, in IGRA, HLA is thought to have an essential role through antigen recognition and interaction between antigen-presenting cells and T cells, which might influence our results. In tuberculosis, HLA-DR polymorphism was associated with IGRA sensitivity 40 . Another study demonstrated that HLA-DR was a marker of recently divided CD4 T cells upon M. tuberculosis antigen exposure 41 . For SARS-CoV-2 vaccination, cellular response failure correlated with higher HLA-DR + T cell levels in recipients of allogeneic hematopoietic cell transplantation 38 . This study had several limitations. First, this was a single-center study using the BNT162b2 vaccine and involving health care workers. It remains uncertain whether our findings can be generalized to other populations and COVID-19 mRNA vaccines. Second, we measured levels of anti-spike antibodies instead of SARS-CoV-2 neutralizing antibodies. Neutralizing antibodies are associated with protection 42 . Anti-spike antibodies are known to be correlated with SARS-CoV-2 neutralizing antibodies postinfection 43 and postvaccination 22 , and we confirmed their correlation in small samples after full vaccination (Supplementary Fig. S1). Third, we selected participants with positive anti-nucleocapsid antibodies or with symptoms of SARS-CoV-2 infection based on the questionnaire and possible asymptomatic infections who had negative levels of anti-nucleocapsid antibodies during the examination period. Between 4 and 7 M, Japan was in the fifth wave of the SARS-CoV-2 epidemic due to the Delta (B.1.617.2) variant, and asymptomatic infection might have been more frequent in this period than in other periods.
In conclusion, we observed a substantial decline in both cellular and humoral immune responses 6 months after the administration of BNT162b2 vaccines, although the dynamics of cellular immune responses were somewhat distinct from those of humoral immune responses. Our study emphasizes the need for booster vaccination, and future studies are needed to clarify the dynamics of cellular immune responses after booster vaccination and the clinical significance of the coordination between humoral and cellular immunity.

Methods
Study design and participants. This prospective observational study was conducted at the University of Tokyo Hospital, Tokyo, Japan. In Japan, SARS-CoV-2 vaccination with the BNT162b2 mRNA vaccine for health care workers was made available in February 2021. Health care workers who were requested to receive the first dose of the BNT162b2 mRNA vaccine at the University of Tokyo Hospital from March 12 to March 31, 2021, were invited to participate in this study. Participants were included if they indicated their willingness to participate prior to the first vaccination. There were no exclusion criteria regarding the health conditions of the participants. This study was conducted in accordance with the Declaration of Helsinki and approved by the ethics committee of the University of Tokyo Hospital (approval number: 2020353NI). Written informed consent was obtained from each participant at the start of the study.

Data collection and SARS-CoV-2 antibody immunoassays. Blood samples and clinical information
of the study participants based on the questionnaire were collected prior to the first (baseline) and second (3 W) doses, 3 weeks after the second dose (6 W), 2 months after the second dose (3 M), 3 months after the second dose (4 M), and 6 months after the second dose (7 M), as previously described in detail 15 . Briefly, the following characteristics were collected via an online questionnaire at the first sample collection: participant age, race, sex, height, weight, occupation type, comorbidities, smoking and drinking status, and exposure to outpatients or inpatients. Self-reported information on the history of COVID-19 infection or close contact with confirmed COVID-19 patients was collected at each visit for blood sample collection. In addition, information on adverse reactions after the first and second doses of vaccination was collected. We measured anti-spike IgG antibody levels (SARS-CoV-2 IgG II Quant assay, Abbott Architect, U.S.) using a CLIA, and the positive cutoff antibody level was defined as 50.0 AU/mL according to the manufacturer's instructions. We also measured anti-nucleocapsid SARS-CoV-2 IgG antibody levels using an automated CLIA (iFlash3000, YHLO, Shenzhen, China) to assess COVID-19 infection status. A positive cutoff index was defined as 10.0 AU/mL, according to the manufacturer's instructions. In addition, we measured the counts and levels of the following before the first and second doses: complete blood cell counts, renal function, liver function, electrolytes, lipids, C-reactive protein, D-dimer, prothrombin time, and activated partial thromboplastin time. Data on epidemiological characteristics and adverse reactions to vaccination were collected using an online questionnaire.

Measurement of the SARS-CoV-2-specific T-cell response.
The T-cell responses in the peripheral blood were evaluated via IGRA. We used QuantiFERON SARS-CoV-2 Research Use Only, including blood collection tubes from the SARS-CoV-2 starter and control sets (Qiagen, Hilden, Germany). The kit consisted of Ag1, Ag2, negative control, and positive control tubes. The Ag1 tubes consisted of epitopes of CD4 + T cells derived from the S1 subunit of the spike protein. The Ag2 tubes consisted of epitopes of CD4 + and CD8 + T cells derived from the S1 and S2 subunits. Whole blood was incubated with SARS-CoV-2 epitopes in tubes at 37 °C for 16-24 h, and the IFN-γ concentration in plasma was measured by enzyme-linked immunosorbent assay according to the manufacturer's instructions. The cutoff for this assay was determined using data from 20 subjects who had tested as nonreactive for SARS-CoV-2 with an authorized RT-PCR test or FDA-authorized serology test and 20 donors who were fully vaccinated (between 2 and 16 weeks after full vaccination) with an FDA EUA-authorized vaccine 44 . With a cutoff value of 0.15 IU/mL, the test sensitivity and specificity were deter-