Our study population was similar to the previous report (19). A total of 236 Japanese healthcare workers working at Funairi Citizens Hospital in Japan who received their first dose of the vaccine between March and May 2021 participated in the study. All the participants received two doses of BNT162b2 (Pfizer/Biotech). The vaccinations were administered at the intervals specified in the protocol, that is the second dose was administered three weeks after the first dose. Ultimately, 213 participants with no missing data were included in the analysis, including 35 males (16.4%) and 178 females (83.6%). The age distribution was as follows: 20–29 years: n=29 (13.6%), 30–39 years: n=54 (25.4%), 40–49 years: n=58 (27.2%), 50–59 years: n=45(21.1%), and ≥60 years: n=27 (12.7%).

This study was approved by the Ethics Committee for Human Genome Analysis at Hiroshima University (Hi-258). Written informed consent was obtained from all the participants.

Blood samples were collected three weeks after the first vaccination (just before the second vaccination), three weeks after the second vaccination, and five months after the second vaccination. Total anti-SARS-Cov2 spike IgG antibodies were quantitatively measured using the VITROS SARS-Cov-2 S1 Quant IgG antibody reagent (CLEIA, Ortho Clinical Diagnostics). Quantitative values were determined using the WHO standard binding antibody unit/ml (BAU/ml) (20). The upper limit of quantification was 4000 BAU/ml. Participants with specific N-antibodies detected by the Elecsys^®Anti-SARS-CoV-2 (ECLIA, Roche Diagnostics) were defined as previously infected and excluded from the analysis.

DNA samples from the study subjects were genotyped for 33 functional polymorphisms in 15 candidate genes. The candidate genes were selected on pathophysiological hypotheses based on the best evidence from studies published under the keywords, “vaccine,” “immune response,” and “antibody production.” Single nucleotide polymorphisms (SNPs) in genes encoding functional molecules involved in the immune responses that lead to the establishment of vaccine-induced acquired immunity were investigated. These processes can be categorized into the following four phases: 1) antigen-presenting cell (APC) activation, 2) T cell activation, 3) T cell and B cell (T-B) interaction, and 4) B cell survival ( Figure 1 ). If there were more than four candidate polymorphisms in a gene, selection was based on the number of publications in the single nucleotide polymorphism (SNP) database of the National Center for Biotechnology Information (21, 22). The list of each SNPs and its allele frequency is presented in Table 1 (12–17, 23–51).

Genetic analyses were performed as previously described (28). Briefly, blood samples from the study subjects were collected in EDTA tubes and assigned a blinded unique identification number. Genomic DNA was extracted from whole blood samples using a QIA Cube (QIAGEN, Hilden, Germany). SNP genotyping was performed using TaqMan SNP Genotyping Assays (Thermo Fisher Scientific, MA) according to the manufacturer’s protocol. Two allele specific TaqMan probes containing different fluorescent dyes and polymerase chain reaction (PCR) primer pairs were used to detect specific SNP targets. Quantitative PCR was performed using a Rotor-GeneQ (QIAGEN). We described the zygosity of the A>G SNP as AA or GG and AG in the case of homozygosity and heterozygosity, respectively. Genotypes were analyzed in either a recessive or dominant model, which was selected as the more impactful model based on the results of previous studies or our data set.

We analyzed the association between antibody titers and genetic polymorphisms at 3 weeks after the first vaccination, 3 weeks after the second vaccination, and 5 months after the second vaccination, which was analyzed as initial response, boosting effect, and maintenance ability of antibodies, respectively.

Analyzing the impact of genetic profile for boosting effect, we divided subjects into 2 groups: the individuals with a serum antibody level ≥4000 BAU/ml (upper limit of measurement) defined as high responders and those with <4000 BAU/ml defined as low responders at 3 weeks after the second vaccination.

To account for the level of antibody titer at boosting effect for following maintenance ability, subjects were categorized into three groups by the antibody titers at 3 weeks after the second vaccination (≥4000 BAU/ml, ≥1825 BAU/ml, and <4000 BAU/mL, or <1825 BAU/ml). The median antibody titer at 3 weeks after the second vaccination was 1825 BAU/ml, excluding high responders.

To analyze of the maintenance of protective antibodies, we focused on the neutralizing antibodies. Previously, our collaborators have reported a proportional relationship between the total IgG levels of anti-S protein antibodies and neutralizing activity (19). The predictive formula was as follows: neutralizing inhibitory activity (%) = 48.2 × log (total IgG level of anti-S protein BAU/ml) - 53.5. The negative threshold for neutralizing activity was evaluated using a surrogate virus neutralization test, which showed 30% inhibition compared with negative sera (52). A total IgG level of 54 BAU/ml, which was expected to maintain 30% neutralizing activity, was calculated using a predictive formula. We analyzed the predictive gene profiles below the negative threshold.

All statistical analyses were performed using JMP Pro statistical software package 15.0.0 (SAS Institute, Cary, NC, USA). We used multivariate multiple regression and multivariate logistic regression analyses to identify the factors that were significantly associated with antibody production and maintenance. We used stepwise logistic regression with forward-backward elimination to determine the best combination of variables. The accuracy of the predictive model for high responders and maintenance was assessed using the area under the receiver operating characteristic (ROC) curve.

First, we analyzed the demographic and genetic variables associated with specific antibody production in the initial response phase after the first COVID-19 vaccination. Ten of the 33 candidate SNPs were defined as potential factors by stepwise regression, and seven of the 10 SNPs and young age (<60 years) were significantly associated with the absolute titer of anti-SARS-CoV-2 spike IgG in the initial response phase ( Table 2 ). The statistically significant seven SNPs were within genes encoding immune function molecules belonging to multiple phases, i.e., rs4612666 (NLRP3), rs10925027 (NLRP3), and rs1131454 (OAS1) upon APC activation phase; rs3212227 (IL12B), rs2546893 (IL12B), and rs3087243 (CTLA4) on T cell activation phase, and rs2227284 (IL4) on T-B interaction phase. Thus, SNPs in genes encoding molecules responsible for functions in immunological processes from the priming to B cell activation phases were important predictors in favor of antibodies in response to the initial vaccine, whereas SNPs encoding molecules belonging to the B cell survival phase showed no significant association.

Three weeks after the second vaccination, we observed 53 high responders with anti-SARS-CoV-2 spike IgG at more than 4000 BAU/ml. Ten SNPs in the 33 candidate genes were defined as potential factors using stepwise regression, and six SNPs and young age were significantly associated with high responders ( Table 3 ). Four of the six genes related to high responders overlapped with those related to initial response, i.e., rs4612666 (NLRP3) and rs1131454 (OAS1) on APC activation phase, and rs3212227 and rs2546893 (IL12B) on T cell activation phase.

We proposed a model for predicting high responders using polygenetic factors and age with a receiver operating characteristic (ROC) curve analysis (area under the curve (AUC) =0.74, Figure 2 ). The model equation was as follows; prediction score (P) = 1/(1+e-x) x = -1.09 + 0.71×(age: -1 if 60 years, 1 if under 60)+ 0.58×(sex: -1 if male, 1 if female)+ 0.48×(NLRP3 rs4612666: -1 if CC, 1 if T carrier)+ 0.35×(NLRP3 rs10925027: -1 if T carrier, 1 if CC)+ 0.37×(TNF rs1799724: -1 if CC, 1 if T carrier)+ 0.80×(TNF rs1799964: -1 if T carrier, 1 if CC)+ 0.70×(OAS1 rs1131454: -1 if A carrier, 1 if GG)+ 0.35×(IL1B rs1143627: -1 if A carrier, 1 if GG)+ 0.88×(IL12B rs3212227: -1 if G carrier, 1 if TT)+ 0.77×(IL12B rs2546893: -1 if AA, 1 if G carrier)+ 0.48×(CTLA4 rs231775: -1 if GG, 1 if A carrier)+ 0.44×(CTLA4 rs3087243: -1 if AA, 1 if G carrier).

By analyzing the effect of gene profiles on the ability to maintain antibody titers 5 months after the second vaccination, seven SNPs were defined as potential factors by stepwise regression. Three of the 7 SNPs, rs3212227 (IL12B) in the T cell activation phase, rs1494555 (IL7R) in the T-B interaction phase, and rs1007888 (MIF) in the B cell survival phase, were significantly associated with the maintenance of high titers of anti-SARS-CoV-2 spike IgG ( Table 4 ).

One of the current topics in COVID-19 prevention is the breakthrough infection after vaccination. It has been reported that the neutralizing potency, which is proportional to anti-SARS-CoV-2 spike receptor-binding domain IgG levels, is key to preventing COVID-19 infection (53, 54). We evaluated predictive factors for the individuals whose anti-SARS-CoV-2 spike IgG titer dropped below the negative threshold (details in Material and method) and found that rs12583006 (BAFF) was exclusively identified as a genetic risk factor for rapid attenuation of anti-SARS-CoV-2 IgG (aOR 0.23, 95%CI 0.07-0.80) ( Table 5 ). We proposed a predictive model to identify individuals with rapid anti-SARS-CoV-2 IgG attenuation using ROC curve analysis with or without antibody titer information (AUC=0.86, and 0.76, respectively, Figures 3A, B ). Although the prediction model with anti-SARS-CoV-2 IgG titer information was more accurate, the model consisting of the gene profile alone without antibody titer remained informative. The model equation was as follows; predicted score (P)=1/(1+e-x) x=-1.16 + 0.50×(sex: 1 if male, -1 if female)+ 0.42×(age: 1 if 60 years, -1 if under 60 years)+ 0.38×(NLRP3 rs4612666: -1 if T carrier, 1 if CC)+ 0.91×(NLRP3 rs10925027: -1 if CC, 1 if T carrier)+ 0.40×(IL12B rs6887695: -1 if G carrier, 1 if CC)+ 0.47×(STAT4 rs7572482: -1 if A carrier, 1 if GG)+ 0.51×(IL4R rs1805010: -1 if A carrier, 1 if GG)+ 0.40×(IL7R rs1494555: -1 if A carrier, 1 if GG)+ 0.50×(BAFF rs12583006: -1 if T carrier, 1 if AA), (AUC=0.76, Figure 3B ).