Abstract

Background

   Research comprehensively examining factors for COVID-19 DNA vaccine
   responses is lacking, particularly in Asian populations. This study
   aims to investigate biomarkers of reactogenic and immunogenic responses
   after DNA-based COVID-19 vaccination in a Taiwanese population.

Methods

   A genome-wide association study (GWAS) of 415 Taiwanese healthcare
   workers was conducted to identify genetic variants associated with
   reactogenic and immunogenic responses to the first and second doses of
   ChAdOx1 nCoV-19 vaccine. Furthermore, gene set enrichment analysis was
   conducted to elucidate the underlying biological pathways. Finally, a
   polygenic score (PGS) was utilized to assess the synergistic host
   effects on neutralizing antibody (NT50).

Results

   We identified 501 suggestive significant genetic associations with
   vaccine responses, enriched in lipid and lipophilic vitamin metabolism,
   interleukin signaling, and neurotransmitter release pathways. Moreover,
   we observed a combined effect of genetics with age and sex on NT50
   after the second dose. Notably, the negative correlation between age
   and NT50 was stronger in lower PGS groups (ρ[lowPGS] = -0.5,
   ρ[mediumPGS] = -0.2, ρ[highPGS] = -0.0072).

Conclusion

   Our study fills a critical gap by addressing the lack of research on
   genetic factors of ChAdOx1 nCoV-19 vaccine responses in Asian
   population, providing valuable insights into the genetic basis of
   DNA-based vaccine responses. The synergic host effect highlights the
   value of integrating genetic information with other host factors as a
   biomarker to predict individual vaccine responses. Our findings can
   contribute to personalized vaccination strategies and future
   vaccination policies.

   Keywords: polygenic effect, host factors, COVID-19 vaccine responses,
   genomics, biomarker

Introduction

   The global pandemic of coronavirus disease 2019 (COVID-19) has led to
   more than 770 million reported cases and 7 million deaths since the
   emergence of severe acute respiratory syndrome coronavirus 2
   (SARS-CoV-2) in 2019 ([68]1). In Taiwan, more than 10 million cases and
   17 thousand deaths of COVID-19 have been reported ([69]2). This public
   health emergency mobilized the entire world to develop novel treatment
   and preventive agents against the disease. The widespread use of safe
   and efficacious vaccines is regarded as the key to prevent disease
   spread. The global endeavor resulted in the development of new vaccines
   in a short period of time. The first vaccines were developed and
   authorized for use around the world at the end of 2020 ([70]3).

   ChAdOx1 nCoV-19 (AZ1222) is one of the earliest COVID-19 vaccines to be
   developed. It is a viral vector-based vaccine that utilizes a
   replication-deficient chimpanzee adenoviral vector to encode the
   SARS-CoV-2 spike protein. Clinical trials showed that ChAdOx1 nCoV-19
   is safe and efficacious to prevent symptomatic and severe COVID-19. The
   estimated efficacy of the vaccine against symptomatic COVID-19 was 74%
   in populations that received the full two-dose regimen ([71]4).
   Furthermore, it has been administered to hundreds of millions of people
   in nearly 150 countries ([72]4, [73]5). However, variations in the
   reactogenicity and immunogenicity were observed ([74]4, [75]6). Side
   effects were milder and less frequent in elderly, while the
   neutralizing antibody responses decreased with age ([76]6, [77]7).
   Furthermore, females were reported with higher odds of experiencing
   adverse events after vaccination than males ([78]8).

   Previous studies showed that host genetics is associated with COVID-19
   vaccine responses ([79]9–[80]15). The HLA-A*03:01 is associated with
   side effects and humoral responses after mRNA vaccinations in the US
   and European populations ([81]9–[82]11). Furthermore, the associations
   of humoral responses with HLA-DRB1*13:02, HLA-DQA1*01:01, and
   HLA-DQB1*06 were reported in UK populations after vaccination with mRNA
   vaccines or ChAdOx1 nCoV-19 vaccine ([83]12, [84]13). In Asian
   populations, loci other than the HLA region were reported. A total of
   14 loci on different chromosomes are associated with side effects after
   mRNA vaccination in the Japanese population ([85]14). Similarly,
   non-HLA variants are associated with antibody responses after
   inactivated COVID-19 vaccines in a Chinese population ([86]15). These
   findings underscore the significance of genetic associations with
   vaccine responses. However, most of these studies focused exclusively
   on either reactogenic or immunogenic responses following mRNA
   vaccination. Information regarding the association between host factors
   and DNA-based vaccine responses in the Asian population remains
   limited. Furthermore, there is a lack of study exploring the combined
   effect of host factors on COVID-19 vaccine responses. Therefore, in
   this study, we aim to investigate the associations of host demographics
   and genetic factors comprehensively with both the reactogenic and
   immunogenic responses after ChAdOx1 nCoV-19 vaccination in a Taiwanese
   population. Furthermore, we extended the exploration for the
   synergistic effects of genetic and demographic factors on
   vaccine-induced immune responses.

Materials and methods

Study design

   A schematic overview of the study design is shown in [87]Figure 1 . In
   this study, 415 healthcare personnel were recruited during the initial
   COVID-19 vaccination programs in Taiwan. Clinical responses toward
   vaccination, including reactogenicity (side effects) and immunogenicity
   [antibody responses against the nucleocapsid protein (anti-N), the
   spike protein receptor-binding domain (anti-RBD), and the half-maximal
   neutralizing antibody (NT50)] were collected at pre-specified time
   points after vaccinations. In addition, genomic DNA (gDNA) was
   collected for genotyping and a genome-wide association analysis was
   conducted to investigate the association between genomic variations and
   COVID-19 vaccine responses. Variant annotations and pathway enrichment
   analysis were conducted to further elucidate the mechanisms underlying
   the COVID-19 vaccine responses. Finally, a polygenic score (PGS) was
   calculated to examine the cumulative genetic influence and its
   interaction with other host factors. This study was conducted in
   compliance with The Code of Ethics of the World Medical Association.
   Written informed consents were obtained from the participants before
   enrollment. The study protocols were approved by the Institutional
   Review Board of Taipei Medical University (TMU-JIRB N202103088 and
   N202107009). Detailed methodologies are described in following
   sections.

Figure 1.

   [88]Diagram depicting the process of vaccine administration, tracking,
   and analysis. It shows ChAdOx1 priming and boosting over 8-12 weeks,
   followed by blood and serum collection for genotyping and assays. A
   questionnaire assesses reactogenicity symptoms like pain, fatigue, and
   fever. DNA and genotype are analyzed using TWB2.0 Microarray. Genome
   wide association analysis results are shown in Manhattan plots.
   Polygenic score and functional annotation are highlighted alongside
   gene and pathway enrichment analyses.
   [89]Open in a new tab

   Study design. Anti-N, anti-nucleocapsid antibodies; ECLIA,
   electrochemiluminescence immunoassay; Anti-RBD, anti-spike
   receptor-binding domain antibodies; CMIA, chemiluminescent
   microparticle immunoassay; Ab, antibody; CPE, cytopathic effect; VNT,
   virus neutralization test. Created in BioRender. Chou, W. (2025)
   [90]https://BioRender.com/a50i612.

Study participants and recruitment

   Participants in the study were recruited during the immunization
   programs for the healthcare personnel at Taipei Medical University
   Hospital (TMUH) and Taipei Municipal WanFang Hospital (WFH) between
   March and September of 2021. Individuals with previous COVID-19
   diagnosis or had been quarantined due to contact with
   SARS-CoV-2-positive patients were excluded from the study. All
   participants received two doses, including the priming dose (PD) and
   booster dose (BD), of 0.5 mL ChAdOx1 nCOV-19 (AstraZeneca)
   vaccinations, administered at intervals of eight to twelve weeks.
   Peripheral blood samples were collected from the participants two weeks
   after the first/priming dose (PD14), and four weeks after the
   second/booster dose (BD28). For participants enrolled between their
   priming and booster doses, demographic information of their priming
   doses was retrieved retrospectively from their vaccination records,
   while the responses were available only toward the booster dose.

Definition of phenotypes

   A questionnaire was distributed to collect the experiences of solicited
   side effects after the first and second doses. The participants were
   classified as cases of any side effects (AnySE) if they reported
   experiencing any of the side effects listed in the questionnaire. The
   participants were cases of local side effects (LocSE) if they reported
   any side effects at the injection site. Finally, the participants were
   cases of systemic side effects (SysSE) if they had any side effects
   other than those at the injection site. To monitor the antibody
   responses after vaccination, the SARS-CoV-2 IgG II Quant assay (Abbott
   Park, IL, USA) was used to measure the anti-spike RBD IgG antibodies
   (Anti-RBD) in participants’ serum after the first (PD14) and the second
   (BD28) vaccination. Furthermore, neutralizing antibodies in the
   participants’ serum after the second vaccination (BD28) was measured
   using a cytopathic effect-based serum virus neutralization (SVN) assay.
   The NT50 was defined as the reciprocal of the highest dilution that can
   inhibit 50% of the viral infection based on the cytopathic effect
   detected with crystal violet ([91]7). Finally, to differentiate the
   responses elicited from natural infections and vaccinations,
   anti-Nucleocapsid antibodies (Anti-N) in the participants’ serums after
   the second vaccination (BD28) was measured with the Elecsys
   Anti-SARS-CoV-2 assay (Roche, Mannheim, Germany). The cutoff index
   (COI) ≥ 1.0 was used as the qualitative determination of anti-N
   positivity as manufacturer’s recommendation.

Genotyping and quality control

   Peripheral blood mononuclear cells (PBMCs) were isolated from blood
   using Histopaque-1077 (Merck Sigma-Aldrich, MO, USA), and gDNA was
   extracted with the QIAshredder and AllPrep DNA/RNA kit (QIAGEN, Hilden,
   Germany). Genotyping was performed on the Axiom Genome-Wide TWB 2.0
   Array Plate. Quality control filtering of the genotype data was
   conducted using PLINK 2 ([92]16), removing variants with a call rate <
   0.95, minor allele frequency (MAF) < 0.01, non-autosomal variants, and
   those deviating from Hardy-Weinberg equilibrium (pHWE < 1×10^-6).
   Furthermore, subjects with mismatched sex, sample call rate < 0.98, or
   outlying heterozygosity rate (> 3 times of the standard deviation from
   the mean) were excluded. In addition, pairwise kinship coefficients
   were calculated and the subjects with lower genotype call rate within
   each related pair (kinship coefficient > 0.88) were further excluded.
   Moreover, population stratification was assessed using a principal
   component analysis (PCA) with the 1000 genome project (1KGP) data
   (phase 3 v5; release 20130502). Following the above steps, genotype
   imputation was performed using the Michigan Imputation Server ([93]17)
   with data from the east Asian population in 1000 Genome Phase 3 as
   reference panels. After imputation, only common variants (MAF ≥ 0.01)
   with good imputation quality (call rate ≥ 0.9 and R2 ≥ 0.3) were
   included in the genome-wide association analysis.

Genome-wide association study and polygenic score

   To investigate the genetic variants associated with responses to the
   COVID-19 vaccine, we performed a genome-wide association study (GWAS)
   of any side effects (AnySE), local side effects (LocSE), systemic side
   effects (SysSE), and anti-RBD after the priming (PD14) and booster
   (BD28) doses, and NT50 after the booster dose (BD28) of COVID-19
   vaccine. The GWAS was conducted using PLINK2 ([94]16), including age
   and sex as covariates. For responses after the second vaccination,
   anti-N positivity at BD28 was included as an additional covariate. The
   quantitative traits, including anti-RBD and NT50, were log-transformed.
   The genome-wide association with quantitative and binary traits were
   tested using additive genetic models, employing linear regression for
   quantitative traits and logistic regression for binary traits. A
   genome-wide significant (GWS) threshold of p < 5×10^–8 and a suggestive
   significant threshold of p < 10^–5 were applied in this study. The
   association results were visualized using Manhattan plots. Furthermore,
   quantile-quantile plots (Q-Q plots) and genomic inflation factors
   (λ[GC]) were utilized to assess genome inflation and potential
   population stratification in each analysis. Moreover, a polygenic score
   (PGS) for NT50 at BD28 was calculated for each subject by summing up
   the effects of the suggestive significant variants, weighted by their
   dosage of effect allele in each locus. The participants were grouped
   into PGS tercile subgroups and association of NT50 with age and sex
   were conducted in each PGS subgroup to assess the synergistic effects
   of genetic and other host factors on NT50 responses.

Variant annotation

   To assess the biological function and effects of genetic variants
   associated with COVID-19 vaccine responses, the variants achieving
   suggestive significance were annotated with their corresponding
   encoding genes (HGNC gene symbols) using the Ensembl Variant Effect
   Predictor (VEP) v110 with the GRCh38 v110 cache file for Homo sapiens
   ([95]18). The parameter “–pick” was used to pick one consequence block
   for each variant. Furthermore, GWAS catalog was queried to identify
   reported variants associated with COVID-19 vaccine responses using the
   “gwasrapidd” R package (accession date: Feb. 16, 2024) ([96]19). The
   variants were classified as known variants if there is a variant
   reported in relation to the corresponding phenotypes in the GWAS
   catalog within the 500 kb interval.

Gene set enrichment analysis

   To evaluate and compare the mechanisms underlying the variable
   responses to COVID-19 vaccination, a gene set enrichment analysis was
   conducted for the genes encoding any suggestive significant variants.
   The analysis employed over representation analysis (ORA) using the R
   package “clusterProfiler” (v4.7.1.003) ([97]20). To assess the
   biological process and pathways involved, gene ontology biological
   process (GOBP), Kyoto Encyclopedia of Genes and Genomes pathway (KEGG
   PATHWAY), Reactome and WikiPathway were included in the analysis. A
   false discovery rate adjusted p value < 0.05 was used as the
   significance threshold for the enrichment analysis.

Statistical analysis

   In this study, the statistical tests between vaccine responses and
   demographic factors in the overall and the PGS subgroup analysis were
   performed using R version 4.2. The McNemar’s χ2 test was used test the
   differences of reactogenicities and Wilcoxon’s signed rank test was
   used to test the differences of immunogecities measured in different
   timepoints. The Fisher’s exact test was used to test the independence
   between two categorical variables. The Mann-Whitney test was used to
   assess the differences in quantitative variables between two groups.
   Finally, smooth curves between two quantitative variables were fitted
   using the local polynomial regression via the LOESS function, and their
   correlation was evaluated utilizing the Spearman’s correlation method.
   The p value < 0.05 was used as the significance threshold.

Results

Study subjects and the responses after the first and the second dose of
COVID-19 vaccine

   In this study, a total of 415 participants were recruited, of which 408
   had gDNA available for whole-genome genotyping. One third (n = 136) of
   this cohort are male, and the average age was 38 years old (37.77 ±
   11.78) at enrollment. All of the participants received two doses of
   ChAdOx-1 COVID-19 vaccine with an average spacing interval of 69 days
   (69.11 ± 11.7). The average time of sampling after the first and the
   second dose were 14 days (14.44 ± 1.03) and 28 days (28.33 ± 1.62),
   respectively. Two participants were tested positive for anti-N antibody
   at BD28, indicating past infection with SARS-CoV-2 before this time
   point. As shown in [98]Table 1 , most responses toward the first and
   the second dose are significantly different. A majority of the
   participants (91.8%) experienced at least one solicited adverse
   reactions following the first COVID-19 vaccine, while only 75.8%
   reported side effects after the second dose (p = 1.32 × 10^-7).
   Injection site pain, fatigue, malaise, and fever were the most common
   reactogenicity after the vaccination. In addition, the frequencies were
   lower after the second dose in most types of reactogenicity. As for
   immunogenicity, the anti-RBD antibody increased by four times after the
   second dose (p = 1.63 × 10^-36). The average of 50% neutralizing titer
   (NT50) at four weeks after the second dose was 307 IU/mL (306.74 ±
   284.48).

Table 1.

   Clinical responses after the first and second dose of COVID-19
   vaccination.
   Outcome After the first dose (PD) After the second dose (BD) P
   value[99] ^c
   Reactogenicity
   Any adverse reactions listed below (n (%)) 223/243 (91.8%) 247/326
   (75.8%) 1.32 × 10^-7
    Local side effects (n (%)) 197/243 (81.1%) 198/326 (60.7%) 1.14 ×
   10^-6
     Injection site pain (n (%)) 191/244 (78.3%) 187/326 (57.4%) 4.78 ×
   10^-7
     Injection site red (n (%)) 80/243 (32.9%) 59/326 (18.1%) 4.73 × 10^-6
     Injection site swollen (n (%)) 121/244 (49.6%) 89/326 (27.3%) 1.01 ×
   10^-9
    Systemic side effects (n (%)) 201/243 (82.7%) 184/326 (56.4%) 1.13 ×
   10^-13
     Fever (n (%)) 131/244 (53.7%) 65/326 (19.9%) 5.02 × 10^-17
     Fatigue (n (%)) 192/244 (78.7%) 156/326 (47.9%) 1.92 × 10^-14
     Malaise (n (%)) 150/244 (61.5%) 79/326 (24.2%) 9.44 × 10^-15
     Chill (n (%)) 126/244 (51.6%) 62/326 (19%) 9.43 × 10^-15
     Headache (n (%)) 106/244 (43.4%) 81/326 (24.8%) 7.24 × 10^-9
     Chest pain (n (%)) 8/244 (3.3%) 11/326 (3.4%) 1
     Stomachache (n (%)) 7/244 (2.9%) 7/326 (2.1%) 0.50
     Dyspnea (n (%)) 8/244 (3.3%) 5/326 (1.5%) 0.13
   Immunogenicity
   anti-RBD[100] ^a (AU/mL) (mean ± SD) 308.39 ± 591.14
   (n = 239) 1369.85 ± 1309.51
   (n = 408) 1.63 × 10^-36
   NT50[101] ^b (IU/mL) (mean ± SD) – 306.74 ± 284.48
   (n = 297) –
   [102]Open in a new tab
   ^a

   anti-RBD: Spike protein receptor binding domain antibody.
   ^b

   NT50: 50% neutralizing titer.
   ^c

   The McNemar’s χ^2 test was used to test the differences in the
   proportion of subjects with reactogenicities after the first and the
   second dose of vaccination. The Wilcoxon’s signed rank test was used to
   test the differences in anti-RBD measured after the first and the
   second dose of vaccination. Tests with p < 0.05 were considered
   statistically significant and shown in bold.

COVID-19 vaccine responses in different sex and age

   As shown in [103]Figures 2A-C , a higher proportion of female vaccinees
   experienced side effects after the first dose of the vaccination. The
   frequency of systemic side effects is significantly higher in females
   (p < 0.05, [104]Figure 2C ). The frequencies of overall reactogenicity
   and systemic side effects are significantly higher in females (p <
   0.05) after the second COVID-19 vaccine ([105] Figures 2D-F ). In
   contrast, there is no difference in the immunogenicity at either time
   point between different sexes ([106] Figures 2G-I ).

Figure 2.

   [107]Nine-panel figure depicting sex differences in side effects or
   antibodies after different dose of COVID-19 vaccines with p-values
   indicating statistical significance. Panels A-F: Bar graphs show
   proportions of male and female participants reporting never or ever
   experiencing these effects. Panels G-I: Violin plots illustrate the
   distribution of anti-RBD and NT50 antibody levels by sex, with p-values
   annotating each panel.
   [108]Open in a new tab

   Association of sex and COVID-19 vaccine responses. (A) Overall
   reactogenicity after the first dose; (B) Local side effect after the
   first dose; (C) Systemic side effect after the first dose; (D) Overall
   reactogenicity after the second dose; (E) Local side effect after the
   second dose; (F) Systemic side effect after the second dose; (G)
   anti-RBD antibody level at 14 days after the first dose; (H) anti-RBD
   antibody level at 28 days after the second dose; (I) NT50 level at 28
   days after the second dose. In panel (A–F), the differences in the
   proportion of participants with (red) and without (blue) the
   corresponding side effects in different sex were examined using
   Fisher’s exact test. In panel (G–I), the differences in the
   corresponding immunogenicity responses between females (red dots) and
   males (blue dots) were examined using the Wilcoxon rank sum test. The
   horizontal grey lines indicate the median value in the group. Tests
   with p < 0.05 were considered statistically significant.

   Age is highly associated with reactogenicity and immunogenicity after
   both the first and second dose of COVID-19 vaccine ([109] Figure 3 ).
   The median age of participants experiencing side effects is
   significantly younger than those without side effects ([110]
   Figures 3A-C, E, F ). Furthermore, antibody levels after either dose
   were significantly lower in older participants ([111] Figures 3G-I ).

Figure 3.

   [112]Violin plots and scatter plots illustrating the relationship
   between age and side effects or antibodies after different dose of
   COVID-19 vaccines. Panels A-F show reactogenicity, local, and systemic
   side effects, grouped by those who experienced side effects ("Ever")
   and those who did not ("Never"), with P values indicating statistical
   significance. Panels G-I display scatter plots with logarithmic scales
   for anti-RBD and NT50 antibodies, showing negative correlations with
   age, accompanied by histograms on top and side margins.
   [113]Open in a new tab

   Association of age and COVID-19 vaccine responses. (A) Overall
   reactogenicity after the first dose; (B) Local side effect after the
   first dose; (C) Systemic side effect after the first dose; (D) Overall
   reactogenicity after the second dose; (E) Local side effect after the
   second dose; (F) Systemic side effect after the second dose; (G)
   anti-RBD antibody level at 14 days after the first dose; (H) anti-RBD
   antibody level at 28 days after the second dose; (I) NT50 level at 28
   days after the second dose. In panel (A–F), the differences in age of
   participants with (red dots) and without (blue dots) the corresponding
   side effects were examined using the Wilcoxon rank sum test. The
   vertical grey lines indicate the median age in the group. In panel
   (G–I), the correlations between age and the corresponding
   immunogenicity responses were tested using Spearman’s correlation
   method. The smooth curves (green lines) were fitted using the LOESS
   method. Tests with p < 0.05 were considered statistically significant.

GWAS for COVID-19 vaccine responses

   Following quality control of genotype data and imputation, a total of
   401 subjects and 7,780,806 variants were included in the GWAS analysis
   ([114] Supplementary Figure S1 ). As shown in [115]Figure 4A , we
   identified a locus on PURPL gene at 5p14.1 that was significantly
   associated with NT50 level after two doses of ChAdOx1 nCoV-19
   vaccination. Carriers with T alleles at the lead SNP rs10068321 had a
   lower logNT50 at BD28 (β = -0.33, p = 2.01×10^-8). In addition, a total
   of 140 variants on 12 genes passed the suggestive significance
   threshold of association with logNT50 at BD28 ([116] Figure 4A ,
   [117]Supplementary Data Sheet 1 ). The genotype data in our cohort is
   clustered with the data from the east Asian population in 1KGP ([118]
   Figure 4B ). A quantile-quantile plot showed that there was no severe
   genomic inflation of the association results ([119] Figure 4C , λ[GC] =
   1.004).

Figure 4.

   [120]Panel A shows a Manhattan plot with significant gene loci such as
   PURPL, LINC01098, and SCD5. Panel B displays a PCA scatter plot
   distinguishing populations by color. Panel C presents a QQ plot with a
   lambda value of 1.004, showing observed versus expected p-values.
   [121]Open in a new tab

   (A) Manhattan plot for GWAS of NT50 after the second dose of COVID-19
   vaccination; (B) The plot of the first two principal components (PCs)
   derived from the principal components analysis (PCA); (C)
   Quantile-quantile plot (Q-Q plot) for GWAS of NT50 after the second
   dose of COVID-19 vaccination. In panel A, the NT50 was log-transformed,
   and age, sex, and anti-N positivity were included as covariates. A
   genome-wide significance (GWS) threshold of p < 5×10^-8 (red line) and
   a suggestive significance threshold of p < 10^-5 (blue line) were
   applied. Genes are labeled if they contain variants passing the
   suggestive significance threshold. In panel (B), the first two PCs were
   derived from the PCA of genotype data from our cohort (dark blue), as
   well as from populations included in the 1000 Genome Project (1KGP)
   data (phase3 v5; (release 20130502). AFR, African; AMR, American; EAS,
   East Asian; EUR, European; SAS, South Asian). In panel (C), the
   diagonal line represents theoretical distribution.

   None of genome-wide significant signals was identified in the analysis
   of the reactogenicity or the anti-RBD responses after either vaccine
   doses. Nevertheless, a total of 8, 15, and 59 variants are suggestively
   associated with AnySE, LocSE, and SysSE reactogenicity, respectively,
   after the first dose (PD) ([122] Supplementary Figures S2A-C ,
   [123]Supplementary Data Sheet 1 ). As shown in [124]Supplementary
   Figure S2B and [125]Supplementary Data Sheet 1 , the majority of the
   signals for LocSE after PD are clustered in the FAM153CP gene (5q35.3)
   and the ATXN1 gene (6p22.3). In addition, the signals for SysSE after
   PD are mostly clustered in the PDSS2 and SOBP at 6q21 ([126]
   Supplementary Figure S2C , [127]Supplementary Data Sheet 1 ).
   Furthermore, 116 variants are suggestively associated with the anti-RBD
   level 14 days after PD. The clusters of the signals are found in
   chromosomes 1, 4, 5, 8, 9, 10, 13, and 14 ([128] Supplementary Figure
   S2D , [129]Supplementary Data Sheet 1 ). In the GWAS of responses after
   second dose of ChAdOx1 nCoV-19 (BD) vaccination, a total of 82, 48, and
   2 variants passed the suggestive significance threshold for AnySE,
   LocSE, and SysSE, respectively ([130] Supplementary Figures S3A-C ,
   [131]Supplementary Data Sheet 1 ). The signals for AnySE after BD are
   located predominantly on chromosome 8 ([132] Supplementary Figure S3A ,
   [133]Supplementary Data Sheet 1 ). Furthermore, most signals for LocSE
   are clustered in HDAC4 (2q37.3), VOPP1 (7p11.2), and an intergenic
   region at 8q23.2 ([134] Supplementary Figure S3B , [135]Supplementary
   Data Sheet 1 ). Furthermore, 31 variants are suggestively associated
   with the anti-RBD level 28 days after BD. Importantly, a suggestive
   significant variant, rs1077416 (20q13.33, DIDO1), is in proxy of a
   previously reported COVID-19 vaccine response-associated variant in the
   GWAS catalog ([136] Supplementary Figure S2D , [137]Supplementary Data
   Sheet 1 ). To further investigate the regulatory role of these variants
   on gene expression, the Genotype-Tissue Expression (GTEx) Portal
   (release v10) was interrogated for eQTL analysis of these 31 variants.
   As shown in [138]Supplementary Table S1 , the G allele at rs7518426 is
   associated with increased IL19 (p = 4.3 × 10^-8) and IL20 (p = 2.2 ×
   10^-8) expression in testis. Furthermore, the DIDO1 variant rs1077416
   is a cis-eQTL and the T allele is associated consistently with lower
   expression of DIDO1 across multiple tissues, including spleen (p = 2.6
   × 10^-7).

Gene set enrichment analysis

   To understand pathways and biological functions of COVID-19 vaccine
   response-associated genes, we conducted gene set enrichment analysis.
   The genes containing the variants with suggestive significance were
   used as the input for the analysis. In this study, immune-related
   pathways were consistently enriched with COVID-19 vaccine
   response-associated genes across multiple databases. LocSE-PD14 genes
   were linked to NFAT signaling ([139] Supplementary Figure S4 ), while
   second-dose-associated genes were enriched in KEGG infection pathways
   ([140] Supplementary Figure S5 ), interleukin signaling in Reactome
   ([141] Supplementary Figure S6 ), and virus- and interleukin-related
   pathways in WikiPathways ([142] Supplementary Figure S7 ). Furthermore,
   fat-solubale vitamin and lipid metabolic pathways were enriched with
   the vaccine response-associated genes. NT50 BD28 genes were linked to
   fatty acid metabolism and biosynthesis, while AnySE BD28 genes were
   enriched in the vitamin E metabolic process ([143] Supplementary Figure
   S4 ). Similar associations were observed in KEGG ([144] Supplementary
   Figure S5 ), Reactome ([145] Supplementary Figure S6 ), and
   WikiPathways ([146] Supplementary Figure S7 ). Finally, several
   neurotransmitter-related pathways are also enriched. GO terms related
   to synaptic processes and tetrahydrofolylglutamate metabolism were
   enriched ([147] Supplementary Figure S4 ), along with folate and
   fluoropyrimidine pathways ([148] Supplementary Figures S5 , [149]S7 ).
   Reactome analysis further linked AnySE PD14 genes to neurotransmitter
   release cycles of serotonin, norepinephrine, glutamate, dopamine, and
   acetylcholine ([150] Supplementary Figure S6 ).

Combined effects of genetics, age, and sex on NT50 after the second
vaccination

   To understand the combined effects of genetics and other host factors
   on neutralizing antibody levels after the second COVID-19 vaccination,
   we constructed a PGS using summary statistics from the GWAS of logNT50
   at BD28. As shown in [151]Figure 5A , logNT50 at BD28 is higher in
   females in the low PGS group, while there is no sex difference in the
   medium or high PGS group. On the other hand, while neutralizing
   antibody levels after the second shot decrease with age in the low and
   medium PGS group, the age-related effect is not observed in the high
   PGS group. Furthermore, the correlation coefficient increases with the
   level of PGS (ρ[lowPGS] = -0.5, ρ[mediumPGS] = -0.2, ρ[highPGS] =
   -0.0072) ([152] Figure 5B ).

Figure 5.

   [153]Panel A shows box plots comparing logNT50 levels after the second
   COVID-19 vaccine by sex across low, medium, and high PGS groups, with
   p-values indicating statistical significance. Panel B displays scatter
   plots of logNT50 levels against age for high, medium, and low PGS
   groups, with correlation coefficients and p-values indicating different
   trends.
   [154]Open in a new tab

   Combined effects of genetic variants with sex (A) and age (B) on NT50
   levels after the second dose of COVID-19 vaccination. In panel (A), the
   differences in the logNT50 at BD28 between females (red) and males
   (blue) in each PGS group were assessed using the Wilcoxon rank sum
   test. In panel (B), the correlations between age and logNT50 BD28 in
   the low (blue dots and texts), medium (green dots and texts), and high
   (red dots and texts) PGS groups were examined using Spearman’s
   correlation method. The smooth curves for the low (blue line), medium
   (green line), and high (red line) PGS groups were fitted using the
   LOESS method. The tests with p < 0.05 were considered statistically
   significant.

Discussion

   In this study, we investigated host factors associated with the immune
   responses to the first and second dose of ChAdOx1 nCoV-19 vaccine in a
   Taiwanese population. A stronger antibody response with lower
   reactogenicity was observed after the second dose. Females and younger
   age are associated with higher risk of reactogenicity. Age is
   negatively correlated with immunogenicity, while no significant sex
   differences was found in immunogenic responses. Importantly, genetic
   variants are associated with differences in reactogenicity and
   immunogenicity following vaccination, with the associated genes
   enriched in pathways of lipid metabolism, interleukin signaling, and
   neurotransmitter release. Finally, a PGS analysis highlighted the
   synergistic effect of genetics with sex and age on neutralizing
   antibody levels after the second vaccination. With the investigation of
   the genomic features of responses toward the ChAdOx1 DNA viral vector
   vaccine in an Asian population, our study addresses a critical gap in
   vaccine research by providing genomic insights in an underrepresented
   population for a distinct vaccine platform.

   Nearly all solicited side effects were reported less frequently while
   the anti-RBD level is higher after the second vaccination. These
   findings are consistent with a multi-ethnic phase 3 clinical trial of
   ChAdOx1 nCoV-19 vaccine ([155]4). In our analysis, systemic side
   effects were significantly more prevalent in females after either dose
   of vaccination. Similarly, a European cohort study showed that females
   had higher odds of experiencing side effects after COVID-19 vaccination
   ([156]8). Increased post-vaccination side effects in adult females have
   been reported for the Influenza, Hepatitis B, and Yellow Fever vaccines
   ([157]21). In addition to sex, the negative correlation between age and
   antibody responses observed in this study has also been reported in
   other populations ([158]4) and with mRNA COVID-19 vaccines ([159]22).
   These findings align with the observation that the elderly typically
   exhibit reduced vaccine responses, potentially due to immunosenescence,
   inflammaging, suboptimal nutritional status, underlying health
   conditions, and the complex interplay among these factors ([160]23).

   In the GWAS analysis, a total of 501 genetic associations were
   identified with reactogenic and immunogenic responses, passing the
   suggestive significance threshold. Among the anti-RBD BD28-associated
   variants, the variant rs7518426 is associated with increased expression
   of both IL19 and IL20 in testis with similar effect sizes. IL19
   primarily demonstrates anti-inflammatory effect ([161]24), while IL20
   is a pro-inflammatory cytokine ([162]25), suggesting that the variant
   may result in a balanced effect on immune function. Furthermore, an
   intron variant of the DIDO1 gene, rs1077416, was found in proxy of a
   variant (rs7269650) with reported association with antibody responses
   toward inactivated COVID-19 vaccine in a Chinese population ([163]15).
   T allele of this rs1077416 variant is associated with decreased DIDO1
   expression in multiple tissues, including the secondary lymphoid organ,
   spleen. Previous study shows that increased DIDO1 copy number variation
   is correlated with decreased abundance of NK cells, T cells, and
   macrophages in colorectal cancer patients ([164]26). However, how this
   variant might regulate DIDO1 expression to lead to differential
   antibody production after vaccination remains unclear. Notably, a
   genome-wide significant signal was found with the logNT50 level after
   the second vaccination. This logNT50 BD28-associated variant,
   rs10068321, lies within the intron of the PURPL gene, which encodes a
   p53-activated long non-coding RNA (lncRNA). As reported in GTEx portal
   (v10), this variant is associated with alternative splicing of PURPL.
   Activated by p53, PURPL can in turn suppress p53 and its target genes
   ([165]27), some of which are involved in immune regulation pathways
   like cytokine production and inflammation ([166]28). This suggests that
   rs10068321 may impact immune regulation through alternative splicing of
   the PURPL lncRNA, altering its function of p53 activity regulation, and
   potentially influence antibody production after vaccination. Further
   studies are required to clarify this effect of PURPL variant. Moreover,
   previous GWAS studies have linked HLA variants to COVID-19 vaccine
   responses in other populations ([167]9–[168]13). However, our study
   found no significant associations between genetic variants in the HLA
   region and vaccine reactogenic or immunogenic responses. Instead,
   non-HLA variants across different chromosomes were associated with
   vaccine response. Similarly, studies in Japanese and Chinese
   populations also found non-HLA variants linked to side effects and
   antibody responses to mRNA or inactivated COVID-19 vaccines ([169]14,
   [170]15). The role of non-HLA variants has been reported in a variety
   of childhood vaccines ([171]29). In a twin study of Gambian infants,
   heritability of humoral responses for hepatitis B, oral polio, tetanus,
   and diphtheria was 77%, 60%, 44%, and 49%, respectively, with non-HLA
   genes playing a dominant role ([172]30). Similarly, Höhler et al.
   reported that 60% of the heritability of vaccine-induced antibody
   response to HBV is attributable to non-HLA genes ([173]31). These
   precedents demonstrated the role of non-HLA polymorphisms in vaccine
   responses and our findings suggest that pathways beyond HLA may
   contribute to diverse vaccine responses in Asian populations.

   Our pathway analysis revealed the association of lipid or lipophilic
   vitamin pathways with humoral responses after both the first and second
   doses and the reactogenicity after the second dose of ChAdOx1 nCOV-19
   vaccination. Vitamin D is a lipophilic vitamin that can modulates
   adaptive immunity by suppressing the function of T helper type-1 (Th1)
   cells, promoting the production of anti-inflammatory cytokines by Th2
   cells, and stimulating regulatory T cells ([174]32). Vitamin E also
   exhibits immunomodulatory functions through the anti-oxidant pathways
   ([175]32). Similarly, omega-3 fatty acids have favorable effects on
   immunity and were reported to improve oxygenation in COVID-19 patients
   ([176]32, [177]33). Additionally, several neurotransmitter related
   pathways are enriched with the reactogenicity-associated genes
   following the first dose of ChAdOx1 nCOV-19 vaccine in our study.
   Neurotransmitters, including acetylcholine, norepinephrine, serotonin,
   dopamine, and glutamate, have been reported to possess immunomodulatory
   functions ([178]34, [179]35). The findings suggest that variations in
   genes associated with these pathways may contribute to diverse
   anti-viral antibody production and reactogenicity following the
   administration of the COVID-19 vaccine. Future studies with direct
   measurement of these molecules in vaccine recipients are required to
   validate our findings and to assess if these pathways could be
   modifiable for optimized vaccine responses.

   In this study, we demonstrate the combined effect of host factors on
   neutralizing antibody levels following the administration of two doses
   of ChAdOx1 nCOV-19 vaccine, utilizing PGS analysis. Although no
   significant sex difference in humoral responses was observed, males in
   the low PGS group had lower NT50 levels compared to their female
   counterparts. This is supported by previous studies showing that
   sex-based differences in immune-related genes are associated with
   humoral response in the context of measles and other childhood
   vaccinations ([180]36). Furthermore, adult females were reported with
   higher humoral responses following vaccinations against influenza,
   hepatitis B, yellow fever, rabies, herpes, and smallpox viruses
   ([181]21). This suggests that, in populations with an unfavorable
   genetic background, males may have a reduced ability to produce
   antiviral antibodies after COVID-19 vaccination compared to females.
   Furthermore, our analysis indicated that the negative correlation
   between age and neutralizing antibody levels is less pronounced in the
   high PGS group, suggesting that individuals with favorable genetic
   characteristics may maintain their capacity to produce antibodies
   despite the effects of aging. Our study highlights the importance of
   considering genetic information with other host effects when predicting
   the vaccine responses. This may have clinical implications for
   identifying vaccine recipients who are at risk of suboptimal responses,
   thereby facilitating the optimization of public health vaccination
   strategies.

   There are some limitations in the study. First, despite the effort to
   collect post-vaccination side effects in a timely manner, these side
   effects were self-reported through questionnaires without biochemical
   measurement. Additionally, while the study investigated the genetic
   association with short-term immunogenicity after both doses of the
   vaccine, data on longitudinal follow up was unavailable. Furthermore,
   the present study focuses on common genetic variants, given that the
   limited sample size may not identify the associations with rare
   variants. Future studies are required to validate our findings on the
   genetic features of short-term responses to the viral-vector based
   COVID-19 vaccine in the Asian population. Larger sample sizes will be
   necessary to estimate the effects of the rare variants. Finally,
   investigations of the genetic features related to long-term antibody
   waning or susceptibility to breakthrough infections will complement our
   study and provide deeper insights into the genetic background of
   diverse responses to the ChAdOx1 nCOV-19 vaccine.

Conclusion

   We investigated the host factors associated with ChAdOx1 nCOV-19
   vaccine responses following the first and second doses in a Taiwanese
   population. Our findings indicate that sex, age, and genetic factors
   are associated with the variations in vaccine responses. Additionally,
   we found that pathways related to lipid, interleukins, and
   neurotransmitters were enriched with genes associated with vaccine
   responses. Importantly, our results suggest that the genetic effects
   interact synergistically with the age and sex factors in the
   neutralizing antibody level after two doses of vaccination. This study
   provides valuable insights into the genetic characteristics underlying
   diverse short-term responses to ChAdOx1 nCOV-19 vaccine in the Asian
   population. The implications of our findings may be instrumental in the
   development of personalized vaccination strategies and the optimization
   of future vaccination policies.

Acknowledgments