Associations between genetic variations of HLA and IGHV, vaccination schedule, and COVID-19 vaccine immunogenicity

Abstract

Interindividual genetic variations significantly influenced SARS-CoV-2-specific immunogenicity after vaccination, particularly for human leukocyte antigen (HLA) alleles. However, the mechanisms of different HLA alleles with varying immunogenicity and additional genetic factors related to immunogenicity remained unexplored. This population-based study utilized data of 180,580 vaccinated individuals from the UK Biobank. A genome-wide association study was conducted to identify significant single nucleotide polymorphisms (SNPs) associated with COVID-19 vaccine immunogenicity, identified by immunoglobulin antibody result. We found that SNPs significantly associated with immunogenicity of COVID-19 vaccines were located near the HLA-DQ and IGHV1-69. We further determined HLA-peptide-binding affinities and found that HLA-DQ alleles were significantly associated with immunogenicity due to variations in HLA-peptide binding affinities. Functional genomics data was applied to assess regulatory mechanisms of significant variants near IGHV1-69. Moreover, we employed Mendelian randomization and found that SNPs near IGHV1-69 were associated with immunogenicity by influencing IGHV1-69 expression. Besides, both genetic factors had interactive effects on immunogenicity. These genetic factors modulated the immune response among recipients with different vaccination interval schedules. Furthermore, we observed an interval of five to six weeks that consistently yielded optimal immunogenicity among population, regardless of genetic factors. This study comprehensively demonstrated how interindividual genetic variations affected immunogenicity of COVID-19 vaccines, suggesting the potential for personalized vaccination and administration strategies.

Introduction

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) caused the global outbreak of coronavirus disease 2019 (COVID-19), leading to a pandemic. Adequate prevention actions, especially vaccination, are crucial in reducing the health, social, economic, and mental costs of COVID-19¹. Vaccination is an important preventive strategy to contain the COVID-19 pandemic, with extensive discussions focusing on its efficacy, effectiveness, waning protectiveness over time², and the response to emerging virus variants. In addition, there are significant interindividual variations in vaccine immunogenicity that must be carefully considered. Understanding the influential factors of these variations is essential for optimizing vaccination strategies and enhancing preparedness for the next outbreak.

Considering the interindividual differences in vaccine responses, genetic factors are recognized as significant contributors to variations in immune responses to vaccination. Previous studies have investigated the genetic determinants of immunogenicity for several vaccines^3–5, focusing on pathways such as antigen presentation^3,4 and antibody production^5,6. For COVID-19 vaccines, several studies have examined genetic factors that influence immune responses, particularly those related to antigen presentation, noting that human leukocyte antigen (HLA) alleles are significant contributors. Most research has highlighted associations between HLA polymorphisms and the immunogenicity of COVID-19 vaccines. Specific HLA alleles, such as DQA1*03, DRB1*13, and DRB1*07, have been linked to differences in antibody responses post-vaccination^7–11. Among these, the DQB1*06 allele—particularly the HLA-DQB1*06:04 subtype—has been suggested as potential causal alleles influencing vaccine-induced immune responses^9,11. Furthermore, these HLA alleles were associated with COVID-19 vaccination immunogenicity at different magnitudes, possibly because of different peptide-binding affinities between HLA and viral peptides¹². However, (1) no comprehensive studies have yet provided conclusive evidence to explain the underlying mechanisms by which different HLA alleles lead to conflicting antibody responses; (2) while a few large-scale population-based studies have explored the genetic determinants of COVID-19 vaccine immunogenicity, most have focused primarily on HLA variations related to antigen presentation. It remains unclear whether other genetic factors, potentially involved in antibody production, also play a significant role in modulating immunogenicity.

Therefore, in this study, we aimed to explore the association between the immunogenicity of COVID-19 vaccination and genetic factors using UK Biobank data, which includes genetics, serological tests, and health records. First, we conducted a genome-wide association study (GWAS) to identify single nucleotide polymorphisms (SNPs) associated with immunogenicity, focusing on the HLA region and immunoglobulin heavy chain variable region (IGHV) genes. Second, we investigated HLA alleles that are relevant to the immunogenicity of the COVID-19 vaccine and examined the underlying mechanisms, particularly regarding peptide-binding affinity. Third, we used functional genomics and Mendelian randomization to investigate the regulatory relationship among SNPs, IGHV1-69 gene expression, and COVID-19 vaccine immunogenicity. Finally, we assessed how the contributions of genetic factors to vaccine immunogenicity may vary depending on the timing of vaccination. This study systematically analyzed and expanded our understanding of the genetic factors associated with the immunogenicity of COVID-19 vaccines, which could inform strategies to enhance vaccine efficacy and contribute to better preparedness for future outbreaks of emerging infectious diseases.

Results

Genetic variants associated with immunogenicity of COVID-19 vaccination

To identify genetic drivers of immunogenicity of COVID-19 vaccines, we performed a GWAS using Plink2 on 10,229,681 genotyped or imputed autosomal SNPs in 180,580 vaccinated subjects from the UK biobank dataset, of whom 70,490 were IgG-positive and 110,090 were IgG-negative. More descriptive analysis of the study population can be found in Supplementary Table 1. Although the genomic inflation factor (λGC) was 1.021, inflation in test statistics due to population structure was negligible. Linkage disequilibrium (LD) score regression yielded an intercept of 0.996 (s.e. = 0.0084) and SNP-based heritability (${\hat{h}}_{\mathit{SNP}}^2$) of 0.0185 (s.e. = 0.0043), indicating that genomic inflation was primarily driven by polygenic effects¹³.

We found that SNPs at the genome-wide significance level (p < 5 × 10⁻⁸) were concentrated on Chromosome 6 and Chromosome 14 (Fig. 2). Conditional and joint multiple-SNP analysis identified several independent variants associated with immunogenicity of COVID-19 vaccination: two SNPs—rs35540199 (odds ratio (OR), 0.91; 95% confidence interval [CI], 0.89–0.93; p = 4.42 × 10⁻¹⁷) and rs9273912 (OR, 0.89; 95% CI, 0.87–0.91; p = 3.50 × 10⁻³⁰) —were located in HLA-DQA1 and HLA-DQB1, respectively, both of which encode HLA molecules; another independent variant, rs67625077 (OR, 0.95; 95% CI, 0.94–0.97; p = 8.91 × 10⁻⁹), was located on Chromosome 14 near IGHV1-69, a V gene segment (Supplementary Table 2). Consistently, stratified LD score regression dissected SNP-based heritability by cell types and found that SNPs associated with the immune system accounted for 99.8% of total heritability (Supplementary Table 5). We also employed the REGENIE framework to account for relatedness in our analysis, which allows the inclusion of related individuals without the need for prior exclusion¹⁴. Genome-wide significant SNPs remained concentrated on Chromosomes 6 and 14, confirming the robustness of our findings (Supplementary Fig. 1).

Fig. 2. SNPs associated with COVID-19 vaccine immunogenicity.

Genome-wide associations between SNPs and COVID-19 vaccine immunogenicity exhibited by Manhattan plot (a), and Q-Q plot (b). The blue and red lines in (a) marked the genome-wide significance p value of 5 × 10⁻⁶ and 5 × 10⁻⁸, respectively. Q-Q plot showed expected P values on the x-axis and observed P values on the y-axis. Regional genome plots of GWAS signals within ~400 kb upstream and downstream of the top SNPs on Chromosome 6 (c) and Chromosome 14 (d), respectively. R² indicated the linkage disequilibrium correlation with the lead SNP (purple). The genetic regions were based on GRCh37/hg19 genome assembly.

Specifically, we performed separate GWAS for individuals who received only a first vaccine dose (n = 109,106) and those who received a second (booster) dose (n = 71,474) (Fig. 1). In both the first-dose and second-dose cohort, genome-wide significant associations were detected on Chromosome 6 but not on Chromosome 14 (Supplementary Figs. 2, 3). However, when we conducted a pooled analysis combining all vaccinated individuals, with vaccination dose included as a covariate, significant loci on chromosome 14 still emerged (Supplementary Fig. 4). This result was likely attributable to increased statistical power rather than biological heterogeneity between cohorts. To formally verify this, we conducted a pooled analysis that included an interaction term between vaccination dose and SNP. The interaction term was not statistically significant, while the main SNP effect on Chromosome 14 remained significant (Supplementary Fig. 5).

Fig. 1. Flow chart for study cohort.

IgG immunoglobulin G.

HLA alleles and HLA-peptide binding affinity associated with immunogenicity

The significant SNPs on Chromosome 6 ranged from 32.4 to 33.0 Mb (GRCh37/hg19), a region that largely encodes HLA, the major histocompatibility complex in humans (Fig. 2c). The HLA region exhibits extensive genetic polymorphism, and SNPs within this region give rise to HLA alleles, contributing to disease susceptibility and immune responses to vaccination, including COVID-19⁹. Therefore, we analyzed 359 HLA alleles to identify significant HLA alleles in relation to immunogenicity (Supplementary Table 3). Since associations for other non-HLA-DQ alleles were of lower significance, we focused on HLA-DQ alleles. We found that several HLA-DQ alleles were consistently associated with immunogenicity of COVID-19 vaccination after correcting for multiple testing. An additional copy of HLA-DQB1*05:02 (OR, 0.79; 95% CI, 0.72–0.86), HLA-DQB1*05:03 (OR, 0.87; 95% CI, 0.83–0.91), HLA-DQA1*01:01 (OR, 0.89; 95% CI, 0.87–0.91), and HLA-DQB1*05:01 (OR, 0.90; 95% CI, 0.88–0.92) was associated with reduced immunogenicity. In contrast, an additional copy of HLA-DQB1*06:04 (OR, 1.13; 95% CI, 1.08–1.18), HLA-DQB1*02:01 (OR, 1.07; 95% CI, 1.05–1.09), and HLA-DQA1*01:02 (OR, 1.04; 95% CI, 1.02–1.06) was associated with elevated immunogenicity (Fig. 3a; Supplementary Table 3). To further explore class II contributions, we performed a sensitivity analysis assessing the associations of HLA-DQ alleles with immunogenicity after adjusting for class I effects, and found that the effect sizes for HLA-DQ alleles remained largely unchanged (Supplementary Table 4).

Fig. 3. HLA-peptide binding affinity and immunogenicity of COVID-19 vaccine.

a Significant HLA alleles associated with COVID-19 vaccine immunogenicity. Significance level (denoted by *) was determined if p < 0.05/359, and 359 was the total number of HLA alleles assessed. b Schematic diagram of HLA-peptide binding and subsequent T-cell antigen presentation. Created in BioRender (https://BioRender.com). Positive associations between HLA-DQ-peptide binding affinity and immunogenicity at allelic level (c–e) and population level (f–h). We plotted the dose-response curves at the log scale. Binding affinity was measured by affinity in nanomolar IC₅₀, log-scaled predicted binding affinity and eluted ligand prediction score. Lower value in affinity in nanomolar IC₅₀, and higher values in log-scaled predicted binding affinity and eluted ligand prediction score indicated greater binding affinity. *p < 0.05; **p < 0.01; ***p < 0.001.

We next explored the mechanism underlying the conflicting relationships between HLA-DQ alleles and immunogenicity of COVID-19 vaccination, and postulated HLA-peptide binding affinity to be the influential factor. The binding affinity between an antigenic peptide and HLA molecule is crucial for immunogenicity, as it directly influences the subsequent antigen presentation to T cells (Fig. 3b). Stronger binding affinity has been shown to elicit efficient antigen presentation and a greater immune response¹⁵. We predicted HLA-DQ-peptide binding affinity based on sequences of HLA molecules and peptides, and calculated the average binding affinity for each HLA-DQ allele¹⁶. We found that HLA-DQ alleles with greater binding affinity to peptides derived from the BNT-162b2 vaccine, indicated by lower affinity in nanomolar IC₅₀, higher eluted ligand prediction score, and higher log-scaled predicted binding affinity, were consistently associated with increased immunogenicity of COVID-19 vaccination (Fig. 3d, e). To further verify this finding, we conducted epidemiological analysis and found that individuals with stronger HLA-DQ-peptide binding affinity to BNT-162b2 peptides had significantly higher odds ratio of immunogenicity (Fig. 3f–h; Supplementary Table 8). Furthermore, we also investigated the binding affinity between HLA-DQ alleles and different SARS-CoV-2 lineages. We found that HLA-DQ alleles with lower binding affinity to BNT-162b2 peptides also exhibited similarly low affinity for peptides from other SARS-CoV-2 strains, indicating the robustness of our findings (Supplementary Fig. 6). Taken together, the above evidence indicated that HLA-DQ-peptide binding affinity was significantly associated with immunogenicity of COVID-19 vaccination.

IGHV1-69 expression and immunogenicity of COVID-19 vaccination

On the other hand, we identified a set of significant SNPs in LD with genome-wide significance near the IGHV1-69 gene. IGHV1-69, as a member of the immunoglobulin heavy chain variable region genes, has been widely reported to be enriched in the antigen repertoire usage during viral infections, including COVID-19¹⁷. The variable domains of IgG are assembled by V(D)J recombination, which assembles V gene and other gene segments into a functional immunoglobulin gene (Fig. 4b). However, it is difficult to identify causative variants with biological mechanism due to LD. Therefore, we utilized functional genomic analysis to investigate the regulatory mechanism between IGHV1-69 and SNPs in high LD (R² > 0.8) with the lead SNP, rs67625077 (Fig. 4a). Then, we examined the epigenetic landscape at the region of interest and found that our SNPs of interest were located near the open chromatin region, identified by DNase-seq among various cell types and ATAC-seq of B lymphocyte samples. Furthermore, histone modification (H3K4me1, H3K4me3, and H3K27ac) peaks, enhancer annotation, and ChIP-Seq signal for RNA polymerase II subunit A (POLR2A) were found in this genomic region, implying a cis-regulatory mechanism. In summary, these findings suggested that our SNPs of interest have potential regulatory functions on downstream genes.

Fig. 4. IGHV1-69 expression and immunogenicity of COVID-19 vaccination.

a Risk SNPs with genome-wide significance near IGHV1-69 were assessed for multiple mechanisms by tracks analyses. DNase-seq track included ENCODE overlaid DNase I datasets with 95 cell lines, and red lines indicated immune-related cell lines. Histone H3 modification (H3K27ac, H3K4me1, and H3K4me3) are associated with active transcription, enhancers, and promoters, respectively. The genetic region shown is chr14: 106,697,000-106,720,000, GRCh38/hg38 assembly. NHLF Normal Human Lung Fibroblasts, NHEK Normal Human Epidermal Keratinocytes, HUVEC Human Umbilical Vein Endothelial Cells, HSMM Human Skeletal Muscle Myoblast. b Schematic diagram of V(D)J recombination process and Mendelian randomization framework. Created in BioRender (https://BioRender.com). c Scatter plot of Mendelian randomization results presenting SNPs with Z score of GWAS on the x-axis and Z score of eQTL on the y-axis.

Moreover, to further identify the relationship between IGHV1-69 expression and immunogenicity of COVID-19 vaccination, we used Mendelian randomization analysis based on summary-level data from our GWAS and large-scale eQTL data¹⁸. We discovered that our SNPs of interest were associated with decreased expression of IGHV1-69. Sequentially, Mendelian randomization found the positive association between IGHV1-69 expression immunogenicity of COVID-19 vaccination (Fig. 4b, c; Supplementary Table 7). In summary, our results provided evidence that a set of SNPs located near the IGHV1-69 gene affected transcription and contribute to the immune response against COVID-19 vaccines, highlighting the importance of this region in vaccine efficacy and immune response.

HLA and IGHV1-69 interactively modifying relationships between vaccination schedule and immunogenicity

We found that individuals with HLA-DQ binding affinity above the 50th percentile had significantly higher immunogenicity compared to those below the 50th percentile (OR, 1.04; 95% CI, 1.02–1.07). For IGHV1-69, individuals homozygous for the reference allele (AA) of the lead SNP near IGHV1-69 (rs67625077) showed a higher likelihood of immunogenicity compared to those homozygous for the risk allele (TT) (OR, 1.07; 95% CI, 1.02–1.12). Interactions of HLA-DQ binding affinity and top SNP of IGHV1-69 were also represented significantly positive relationships with immunogenicity, indicating interactive effects of those two genetic factors (Supplementary Table 9). Compared to individuals with HLA-DQ binding affinity below the 50th percentile and homozygous reference allele (AA) of the lead SNP near IGHV1-69, those with HLA-DQ binding affinity above the 50th percentile and homozygous risk allele (TT) of the lead SNP near IGHV1-69 had an OR of 1.07 (95% CI, 1.00–1.14) (Fig. 5a–c; Supplementary Table 8).

Fig. 5. Odds ratio of immunogenicity of COVID-19 vaccine associated with different genetic factors, as well as vaccination durations between different group stratified by genetic factors.

a–c Associations between different genetic factors and odds ratios of immunogenicity, including HLA-DQ binding affinity (controlling for term of top SNP of IGHV1-69 and other covariates mentioned in the GWAS section) (a), top SNP of IGHV1-69 (controlling for term of HLA-DQ binding affinity and other covariates mentioned in the GWAS section) (b), and interaction term of HLA-DQ binding affinity and top SNP of IGHV1-69 (controlling for other covariates mentioned in the GWAS section) (c). Proportions of positive IgG response in relation to vaccination durations, stratified by different genetic factors, including HLA-DQ binding affinity (d, g, j), top SNP of IGHV1-69 (e, h, k), and interaction term of HLA-DQ binding affinity and Top SNP of IGHV1-69 (f, i, l). Error bars denoted two-sided 95% confidence intervals for IgG seropositivity rates, calculated based on binomial distribution. P value was assessed from Wilcoxon matched-pairs signed rank test. For days after first vaccination, we performed analysis in the study population who only received one dose of COVID-19 vaccine. For days after second vaccination, we performed analysis in the study population who ever received a 2nd booster. For interval days between first and second vaccination, we performed analysis in the study population who ever received a 2nd booster and performed Wilcoxon matched-pairs signed rank test among subgroups from 21 to 56 days. The HLA-DQ binding affinity was a binary variable, classified based on the 50th percentile of nanomolar IC₅₀ value of the binding affinity for each population. Values greater than the 50th percentile are coded as “<Average”, while values lower than the 50th percentile are coded as “≥Average” (Since the association between nanomolar IC50 and binding affinity was negative). The “HLA-DQ *IGHV1-69” was defined as interaction of HLA-DQ binding affinity and Top SNP of IGHV1-69, resulting in six distinct groups: individuals with (1) HLA-DQ binding affinity above average and AA genotype of the top SNP near IGHV1-69 (≥Average and AA), (2) HLA-DQ binding affinity above average and TA genotype of the top SNP near IGHV1-69 (≥Average and TA), (3) HLA-DQ binding affinity above average and TT genotype of the top SNP near IGHV1-69 (≥Average and TT), (4) HLA-DQ binding affinity below average and AA genotype of the top SNP near IGHV1-69 (<Average and AA), (5) HLA-DQ binding affinity below average and TA genotype of the top SNP near IGHV1-69 (<Average and TA), and (6) HLA-DQ binding affinity below average and TT genotype of the top SNP near IGHV1-69 (<Average and TT). 1st first, 2nd second, IgG immunoglobulin. *p < 0.05; **p < 0.01; ***p < 0.001.

Furthermore, we examined the modifying effects of genetic factors on the relationships between days after vaccination and the proportion of positive antibody responses (Fig. 5d–l, Supplementary Figs. 7–9). Inverse U-shaped associations were observed between proportions of positive antibody responses and days after the first vaccination, with peak immunogenicity around three to four weeks. After the second vaccination, positive antibody responses increased rapidly, peaked around one to two weeks, and subsequently plateaued. We further identified a peak in immunogenicity occurring around five to six weeks associated with intervals between the two vaccines, which implied an optimal vaccination interval schedule for COVID-19. Furthermore, genetic factors were found to influence the relationship between antibody responses and the vaccination interval. Individuals with HLA subtypes of higher binding affinity and homozygous reference allele (AA) of the lead SNP near IGHV1-69 demonstrated enhanced immunogenicity. However, the optimal vaccination interval for peak immunogenicity remained consistent across individuals with different genetic variations.

Discussion

In this study, we comprehensively analyzed the genetic determinants of immunogenicity following COVID-19 vaccination and quantified their interactive effects and modifications on vaccine responses. We identified two genetic variants that independently and jointly influenced the immunogenicity of COVID-19 vaccines. These variants were associated with two key biological processes: 1) antigen presentation, involving different HLA-DQ alleles and their peptide binding affinities; and 2) IgG production, linked to the expression of the IGHV1-69 gene. Furthermore, we found that these genetic factors modulated the effect of vaccination intervals on immunogenicity by influencing the peak magnitude, without altering the optimal vaccination interval for peak immunogenicity. Our findings expand the understanding of the genetic influences on the variability in immunogenicity to COVID-19 vaccines and have important implications for future vaccine development and administration strategies.

The HLA region is characterized by extensive genetic polymorphism, with SNPs giving rise to diverse alleles that influence susceptibility to vaccination responses¹⁹, including those against COVID-19. In our analysis, several HLA-DQ alleles demonstrated statistically robust associations with COVID-19 vaccine immunogenicity, replicating and extending findings from previous GWAS conducted in the UK⁹, Japan⁷, and other countries²⁰. Our results replicated previous findings that the HLA-DQB1*06:04 allele was the most significant contributor to immunogenicity, along with other similarly significant alleles, further supporting the reliability of our study findings¹¹. We also replicated the association between HLA-A*03:01 and enhanced vaccine-induced immunogenicity, consistent with previous studies^11,21. Furthermore, Bian et al. identified HLA-DRB1*13:02 allele as associated with vaccine-induced antibody responses⁸, which was corroborated by our analysis (Supplementary Table 3). Notably, in HLA-DRB1*13:02, the substitution of arginine with glutamic acid at position 71 altered the electrostatic potential of the P4 pocket, which is predicted to enhance the stability of peptide–MHC complexes. This structural modification likely accounts for the observed increase in HLA–peptide binding affinity and, consequently, the enhanced vaccine-induced immune response in our study.

The impact of genetic factors on vaccine immune responses may stem from different mechanisms. Genomic variants could determine gene subtypes¹¹, or influence gene expression by altering the epigenetic landscape²². These two regulatory mechanisms influencing vaccination efficacy are captured in our study by the HLA-DQ alleles and IGHV1-69 expression. On the one hand, we observed that different HLA-DQ alleles were associated with varying predicted binding affinities for vaccine-derived peptides. These differences in binding affinity may influence the efficiency of antigen presentation and, in turn, affect the magnitude of vaccine-induced immune responses. We validated this association at both the allelic and individual levels. On the other hand, we identified GWAS signals associated with IGHV1-69, highlighting its relationship with the immunogenicity of the COVID-19 vaccine. Previous studies demonstrated that IGHV1-69 gene usage was highly enriched in mature B cells that produce neutralizing antibodies for COVID-19¹⁷, and its polymorphism affected antibody neutralization²². Using Mendelian randomization, we observed a positive correlation between IGHV1-69 gene expression and antibody response, consistent with previous findings, further suggesting its important role in antiviral immunity.

Notably, our study found that individuals with higher immunogenic potential HLA alleles (defined by HLA-DQ binding affinity) and elevated expression potential of the IGHV1-69 (defined by GWAS-identified top SNP) demonstrated enhanced immunogenicity after vaccination. This result suggested that these two key genetic factors collectively improve immune responses to vaccination. Previous studies have indicated that specific antibodies, upon binding to viral antigen epitopes, can influence T cell recognition of the antigen²³. However, the biological mechanism underlying the HLA-IGHV interaction requires further investigation.

Our study observed that the immunogenicity increased until approximately three to four weeks after the first dose, after which it began to decline. Following the second dose, immunogenicity peaked around one to two weeks, with a higher peak than after the first dose. These findings aligned with previous studies showing that a single dose of COVID-19 vaccine raised spike-specific antibodies by day 28, with further enhancement after the second dose²⁴. We also emphasized the critical role of booster shot timing in optimizing immune responses. The WHO Strategic Advisory Group of Experts on Immunization recommended a four to eight-week interval between doses for individuals aged 12 and above²⁵. Similarly, studies of the COVID-19 vaccines suggested that an interval of more than six weeks between doses increases vaccine efficacy²⁶. Our findings supported that an interval of approximately five to six weeks between the first dose and the booster dose provided optimal protection. Booster shots administered too soon or too late may not achieve the desired immune response. These insights can inform resource allocation for COVID-19 and other large-scale infectious disease prevention. Adhering to the optimal vaccination interval can not only help mitigate vaccine shortages in resource-limited areas, but also enhance preventive efficacy in resource-rich areas.

Studies have shown significant variability in host immunogenicity to COVID-19 vaccines, including differences in antibody production, potency, and duration²⁴. Our study identified genetic factors as modest but noteworthy contributors to the magnitude and persistence of vaccine-induced immunity. While the distribution of immunogenic responses following the first and second doses was similar across individuals with different genetic profiles, we observed variability in the peak magnitude of immunogenicity along with vaccination intervals. This discovery suggests personalized vaccinations strategies in vaccine design. Importantly, the optimal vaccination interval between doses remained consistent regardless of genetic factors, suggesting that appropriate vaccination schedules benefited individuals across all genetic profiles.

This study has several limitations. First, we utilized only one lead significant SNP near IGHV1-69 to evaluate the genetic effects on Chromosome 14 with respect to immunogenicity. However, other significant SNPs in strong LD with the lead SNP rs67625077 indicate that rs67625077 is a suitable proxy for capturing the immunogenicity-related genetic effects at this locus. Second, compared with continuous antibody metrics, using binary IgG antibody responses as predictors may reduce the statistical power of the GWAS. Nevertheless, the large sample size of almost 200,000 participants in our study helps to mitigate this limitation. Third, the study population was generally older, which may introduce bias when generalizing the findings to the broader population. Further research should include younger individuals to provide a more comprehensive understanding. Fourth, the use of nucleocapsid-specific IgG seropositivity as a proxy for prior SARS-CoV-2 infection may underestimate true infection rates due to waning antibody levels or variability in individual immune responses. Future genetic studies could consider better infection history data. Finally, the inclusion criterion of a minimum one-day interval between vaccination and antibody testing may not fully capture the time required for IgG responses to develop. However, given the large sample size and inter-individual variability in immune kinetics, the overall impact on the GWAS findings is likely modest. Our study also has several strengths. This large-scale and comprehensive analysis provided novel insights into the genetic mechanisms underlying IgG production in the immunogenicity of COVID-19 vaccines, as well as the cumulative influences of antigen presentation and IgG production. Furthermore, we integrated epidemiological analyses to quantify the genetic-vaccination effects on immunogenicity, offering evidence to inform proper and prompt public health responses in the future.

In conclusion, this study validated HLA-DQ alleles as the major genetic determinants influencing the IgG antibody response to COVID-19 vaccination, and further highlighted HLA-DQ-peptide binding affinities as possible influential factors. Additionally, we discovered a novel susceptibility locus on chromosome 14 and confirmed the involvement of IGHV1-69 expression in vaccine immunogenicity. Our findings suggested that both genetic factors independently and jointly influenced the magnitude and persistence of vaccine-induced immunity. This research advances our understanding of the genetic mechanisms underlying vaccine immunogenicity and clinical effectiveness, providing valuable insights for the development of more personalized vaccination strategies.

Methods

Study population

We used data from UK biobank and restricted the sample to participants who had received at least one dose of a COVID-19 vaccine. The outcome of COVID-19 vaccine immunogenicity was assessed using immunoglobulin (IgG) result by self-administered lateral flow test, specifically the Fortress Fast COVID-19 Test or AbC-19^TM kit. We used IgG test positive as the indicator of immunogenicity. We also extracted information on whether participants received a second booster dose, the interval between the first or second vaccination and antibody testing, and the type of antibody test kit. Because lateral flow tests cannot distinguish between vaccine-induced and infection-induced antibodies, an additional Thriva coronavirus antibody test was further used to collect capillary blood from participants. Laboratory analysis was then performed to detect IgG antibodies against the nucleocapsid protein of SARS-CoV-2, in order to determine prior infection status. We excluded all participants who were ever infected with SARS-CoV-2 (Fig. 1). We extracted individual-level data from multiple databases in the UK Biobank, including COVID-19-related data, Genomics data, Hospital Inpatient data, Primary Care records, and Death Register data. The research has been approved by the UK Biobank Ethics and Governance Council. All participants in the UK Biobank provided written informed consent at the time of recruitment, including consent for their data to be used in future health-related research. Therefore, no additional informed consent was required for this study.

Genome-wide association study (GWAS)

We performed linear association analyses between each genetic variant and immunogenicity of COVID-19 vaccines. We performed quality control filter on the variants according to: information score ≥0.4²⁷, minor allele frequency ≥0.001, Hardy-Weinberg equilibrium -log₁₀(P) ≤ 7, and inclusion of only autosomal and biallelic variants. These inclusion criteria were consistent with previous studies on the same study population²⁸. For the GWAS analysis, IgG positivity after COVID-19 vaccination was used as the primary phenotype and dependent variable; independent variables included SNP term, genotyping array, age, genetic sex, and 15 principal components to control for population stratification. We applied a genome-wide significance threshold of p-value < 5 × 10⁻⁸ to account for multiple testing. Related individuals with high kinship coefficients (>0.3) were removed to reduce the impact of cryptic relatedness. Above genome-wide association analyses were implemented using Plink2²⁹. Moreover, we performed conditional and joint analysis to identify multiple independent signals. Specially, we conditioned on the primary associated SNP, selected other significant SNPs in a stepwise manner with p-value threshold of 5 × 10⁻⁸. We set the window size of selection 10 Mb, assuming that SNPs 10 Mb away or on other chromosomes are in linkage equilibrium. Among all SNPs reached genome-wide significance, we identified three multiple independent SNPs associated with immunogenicity of COVID-19 vaccination, located near HLA-DQA1 and HLA-DQB1 on chromosome 6, and IGHV1-69 on chromosome 14 (Supplementary Table 2). Analysis was performed using GCTA-COJO package³⁰.

Stratified linkage disequilibrium score regression

LD score regression assumes that heritability uniformly distributes across the genome. For complex traits, however, associated heritability may cluster in regions. To deal with this drawback, stratified LD score regression was proposed to dissect SNP-based heritability into different categories. We applied a previously published method to attribute SNP-based heritability from GWAS into different functional categories and cell types³¹. Only cell type group related to immune system was significantly associated with vaccination immunogenicity after Bonferroni correction (p = 0.006 ≤ 0.05/9) (Supplementary Table 5). For SNPs related to immune system, they accounted for 23.34% of total SNPs explained but explained almost 99.8% of ${\hat{h}}_{\mathit{SNP}}^2$, with fold change of 4.28. This functional relevance was supported by existing knowledge that immunogenicity was related to immune system.

GWAS using REGENIE framework

To account for relatedness in our analysis, we utilized the REGENIE framework, which allows the inclusion of related individuals without the need for prior exclusion¹⁴. For the generation of the bed, bim, and fam genotype files used in step 1, we applied stringent quality control filters: SNPs with minor allele frequency (MAF) less than 1%, minor allele count (MAC) less than 100, or missing genotype call rate greater than 1% were excluded. We also removed variants showing significant deviation from Hardy–Weinberg equilibrium (HWE) in the overall sample (p < 1 × 10⁻¹⁵), as such deviations may indicate genotyping artifacts or underlying population structure. Individuals with a genotype missingness rate greater than 2% were also excluded. Furthermore, our analysis was restricted to biallelic SNPs only, since multiallelic variants may complicate modeling and are generally less well-supported by GWAS tools. We also retained only SNPs with unambiguous A, C, G, or T alleles, excluding indels, structural variants, and ambiguous base calls to ensure consistency in downstream analyses. To select representative SNPs and corresponding samples for step 1 modeling, we additionally applied filters based on imputation quality (INFO score > 0.8) and individual missingness (--mind 0.02). Based on available computational resources, we set the block size (--bsize) to 4000. For step 2 association testing, we included the following covariates in the model: GenotypeBatch, Age, GeneticSex, and the first 15 principal components (PC1–PC15). Firth logistic regression with approximate inference was applied to account for potential bias in unbalanced binary traits. To reduce false positives while maintaining sensitivity, we adopted a conservative significance threshold of --pThresh 0.01.

HLA type imputation and HLA-immunogenicity association analysis

Classical HLA types with four-digit resolution at the class II regions were imputed from the SNP genotyping in the UK Biobank population. Imputation of four-digit HLA alleles was carried out using HLA*IMP:02³². For each HLA allele, we performed logistic regression to analyze its association with immunogenicity, adjusting for covariates including the influential factors mentioned in the GWAS section and the type of antibody test kit. Such logistic regression produced a beta estimate: positive value of beta coefficient indicated additional copy of this HLA allele associated with elevated immunogenicity, while negative value indicated reduced immunogenicity. The significance of HLA alleles in the logistic regression model was tested while accounting for multiple testing.

HLA-peptide binding affinity prediction and binding-affinity-immunogenicity association analysis

To assess peptide binding affinity between HLA alleles with peptide from COVID-19 vaccine, we obtained BNT-162b2 sequence in fasta format from GenBank (Accession No. OR134577.1). To predict peptide binding affinity, we used NetMHCpan-4.1 for MHC class I peptide affinity prediction³³, and NetMHCIIPan-4.1 and 4.3 for MHC class II peptide affinity prediction³⁴. This machine learning models were trained on existing binding affinity and eluted ligand data as outputs, using peptide sequence and HLA allele sequences as inputs. We derived the BNT-162b2 sequence into 12- to 18-mers and predicted peptide affinity in nanomolar IC₅₀, eluted ligand prediction score, and log-scaled predicted binding affinity using netMHCPan. Specifically, for each HLA-DQ allele, we calculated the average peptide binding affinity by averaging the binding affinity across all 12- to 18-mers, and assigned this value to the corresponding individuals. For individuals that were heterozygote (two different HLA-DQA1 alleles or two different HLA-DQB1 alleles) or double heterozygote (two different HLA-DQA1 alleles and two different HLA-DQB1 alleles), we calculated binding affinity across all alleles, and assigned the averaged binding affinity to that person.

To explore the association between binding affinity and immunogenicity, we first analyzed associations between immunogenicity and binding affinity at the allelic level. For all HLA alleles, we performed linear regression to examine the association between the beta coefficients representing the relationship between HLA alleles and immunogenicity, and the averaged HLA-peptide binding affinity. Second, for all individuals, we analyzed associations between immunogenicity and binding affinity at the population level. We used logistic regression to analyze the association between individual-level binding affinity and immunogenicity, and adjusting for other covariates. To assess nonlinear dose-response relationships, we repeated the above regression model but replaced the linear term of binding affinity with splines.

Functional genomic analysis of causative variants near IGHV1-69

We used LocusZoom³⁵ to plot SNP-gene relationship in a specific genomic region. We performed tracks analysis to assess potential regulatory functions of the risk SNPs. The conversion of SNPs in different genome assemblies was performed using liftover³⁶. We collected functional genomics data from ENCODE (https://www.encodeproject.org/)³⁷, including chromatin accessibility (DNase-seq, ATAC-seq), histone H3 modification (H3K27ac, H3K4me1, H3K4me3), and ChIP-seq of critical factors (Supplementary Table 6). H3K27ac is a mark of active promoters and enhancers, while H3K4me3 and H3K4me1 usually serve as marks of promoters and enhancers, respectively. Besides 7 cell lines in ENCODE histone modification collection, DND41 and DOHH2 were added in our analysis for their high expression of IGHV1-69, revealed by Harmonizome³⁸. We incorporated ChIP-seq peaks of CCCTC-Binding factor (CTCF) and RNA polymerase II subunit A (POLR2A) in B cell samples in our analysis. We applied annotations of cis-regulatory elements using resources from UCSC genome browser³⁹.

Mendelian randomization analysis of causative variants near IGHV1-69

In addition, we used Mendelian randomization analysis to determine the causal association between IGHV1-69 expression level and the immunogenicity of COVID-19 vaccination, based on summary data from GWAS and eQTL. In the framework of Mendelian randomization analysis, the outcome of interest (Y) is the immunogenicity of COVID-19 vaccination. The gene expression (X) stands as the exposure of interest (i.e., IGHV1-69), while the single nucleotide polymorphism (SNP) or genetic variant (Z) acts as the instrumental variable. To denote the effect size of gene expression on the outcome of interest, we use b_xy. Similarly, b_zy represents the effect size of the SNP on the outcome of interest, and b_zx is the effect size of the SNP on gene expression. In the Mendelian randomization framework, our primary objective is to ascertain the effect size of gene expression on immunogenicity, denoted by b_xy, which can be calculated as b_xy = b_zy/b_zx. Based on previous research, b_xy can be estimated from summary-level data, despite the outcome of interest, gene expression, and SNP being obtained from different data samples; the statistical power of b_xy estimation can be significantly increased if b_zy and b_zx are derived from two independent datasets with large sample sizes⁴⁰. Our GWAS analysis was based on nearly 200,000 individuals, while eQTL data were procured from a separate large-scale sample, eQTLGen consortium¹⁸. Therefore, we calculated the test statistic of b_xy by: $T\approx \frac{Z_{\mathit{zy}}^2{Z}_{\mathit{zx}}^2}{Z_{\mathit{zy}}^2+Z_{\mathit{zx}}^2}$, where z_zy and z_zx were z-test statistics obtained from GWAS and eQTL data set. The positive association between gene expression and outcome was visualized by scatter plot of Z-scores derived from the GWAS and eQTL summary data. Among all 10 SNPs located near IGHV1-69 and reached genome-wide significance (i.e., 5 × 10⁻⁸), 7 of them had available records in eQTL. The expression level of IGHV1-69 was positively associated with immunogenicity, since SNPs of interest were (1) all negatively associated with immunogenicity, according to GWAS results, (2) all negatively associated with reduced expression of IGHV1-69, as indicated by negative z-value in Supplementary Table 7.

Interactive effects of genetic factors on the immunogenicity of COVID-19 vaccine

To evaluate possible interaction between HLA alleles and lead SNP near IGHV1-69 on COVID-19 vaccine immunogenicity, we first calculated an HLA-DQ binding affinity by treating continuous nanomolar IC₅₀ of binding affinity at individual level into a binary variable. Individuals were categorized into two groups according to whether their binding affinity was above or below the average based on the nanomolar IC₅₀ of binding affinity. Separate logistic regression models were performed to evaluate associations between COVID-19 vaccine immunogenicity and genetic factors: (1) the HLA-DQ binding affinity, (2) the lead SNP near IGHV1-69, and (3) the interaction term of HLA-DQ binding affinity * lead SNP near IGHV1-69, adjusting for other covariates. The interaction model additionally included the main effects of both genetic variables.

Modification of genetic factors on relationships between date of vaccinations and immunogenicity

To further investigate whether genetic factors contribute to the heterogeneity of associations and how they modify the relationships between date of vaccinations and immunogenicity, we calculated the proportion of vaccinated individuals with IgG test positive (i.e., immunogenicity) by days after vaccination among groups with: (1) HLA-DQ binding affinity above average versus HLA-DQ binding affinity below average, (2) homozygous or heterozygous reference allele (AA or AT) of the lead SNP near IGHV1-69 versus homozygous risk allele (TT) of the lead SNP near IGHV1-69, and (3) HLA-DQ binding affinity above average and AA of the lead SNP near IGHV1-69 versus HLA-DQ binding affinity below average and TT of the lead SNP near IGHV1-69 (comparisons of other groups were presented in Supplementary Table 9). For days after the first vaccination, the analysis was conducted in individuals who received only one dose of the COVID-19 vaccine. For days after the second vaccination and interval days between the first and second vaccination, the analysis focused on those who received a second booster dose. We also excluded participants from cohorts if they received the first or the second vaccine on the day of the IgG antibody test.

Author contributions

Limei Ke and Guoqing Feng equally contributed to this manuscript. Limei Ke: Conceptualization, Methodology, Investigation, Visualization, Writing—original draft, Writing—review & editing; Guoqing Feng: Conceptualization, Methodology, Investigation, Visualization, Writing—original draft, Writing—review & editing; Bo Yin: Methodology, Investigation; Keyi Lyu: Investigation; Wen Fang: Investigation; Qian Di: Supervision, Conceptualization, Methodology, Investigation, Visualization, Writing—original draft, Writing—review & editing.

Data availability

This study uses data from the UK Biobank. All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. GWAS results are available at the GWAS Catalogue (Study GCP ID: GCP001302, submission ID: 683071c82fffc10001b6a360).

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41541-025-01284-w.