key: cord-0929714-aiz3y2l5 authors: Park, Sunmin; Kang, Suna title: Association between Polygenetic Risk Scores of Low Immunity and Interactions between These Scores and Moderate Fat Intake in a Large Cohort date: 2021-08-19 journal: Nutrients DOI: 10.3390/nu13082849 sha: fc7626d612c3a253c850766803387feb1e77d839 doc_id: 929714 cord_uid: aiz3y2l5 White blood cell (WBC) counts represent overall immunity. However, a few studies have been conducted to explore the genetic impacts of immunity and their interaction with lifestyles. We aimed to identify genetic variants associated with a low-WBC risk and document interactions between polygenetic risk scores (PRS), lifestyle factors, and nutrient intakes that influence low-WBC risk in a large hospital-based cohort. Single nucleotide polymorphisms (SNPs) were selected by genome-wide association study of participants with a low-WBC count (<4 × 10(9)/L, n = 4176; low-WBC group) or with a normal WBC count (≥4 × 10(9)/L, n = 36,551; control group). The best model for gene-gene interactions was selected by generalized multifactor dimensionality reduction. PRS was generated by summing selected SNP risk alleles of the best genetic model. Adjusted odds ratio (ORs) of the low-WBC group were 1.467 (1.219–1.765) for cancer incidence risk and 0.458 (0.385–0.545) for metabolic syndrome risk. Vitamin D intake, plant-based diet, and regular exercise were positively related to the low-WBC group, but smoking and alcohol intake showed an inverse association. The 7 SNPs included in the best genetic model were PSMD3_rs9898547, LCT_rs80157389, HLA-DRB1_rs532162239 and rs3097649, HLA-C rs2308575, CDKN1A_rs3176337 and THRA_rs7502539. PRS with 7 SNP model were positively associated with the low-WBC risk by 2.123-fold (1.741 to 2.589). PRS interacted with fat intake and regular exercise but not with other nutrient intakes or lifestyles. The proportion with the low WBC in the participants with high-PRS was lower among those with moderate-fat intake and regular exercise than those with low-fat intake and no exercise. In conclusion, adults with high-PRS had a higher risk of a low WBC count, and they needed to be advised to have moderate fat intake (20–25 energy percent) and regular exercise. The immune system is composed of innate and adaptive immunity systems. When foreign materials and pathogens enter the body, innate immunity is activated, and adaptive immunity is subsequently initiated. Immunity dysregulation results in immune deficiency or immune system overactivation [1] . As encountered in lymphopenia, immune exhaustion renders the individual susceptible to infection, certain cancers and sepsis, whereas its overactivation is associated with autoimmune diseases. White blood cells (WBC), also called leukocytes, are responsible for innate and adaptive immunity [2] . WBCs protect the body against bacterial and viral infections and are more strongly linked to the innate immune system [3] . Natural killer cells, mast cells, macrophages, eosinophils, basophils, neutrophils, and dendritic cells are components of the innate immune system, whereas other less numerous WBC, including B and T cells, are components of the adaptive immune system [4] . Therefore, WBC is associated mainly with innate immunity to fight against pathogen attacks, and it is involved partly in adaptive immunity [4] . Immediate response to a noxious challenge is achieved by activating the innate immune system, which manifests as the rapid induction of acute inflammation. However, low immunity fails to fight against pathogens susceptible to severe infection and potential cancer risk, while subclinical immune overactivation induces persistent inflammation, called chronic low-grade inflammation [5, 6] . Persons with either low-or overactive immunity are susceptible to infection severity and mortality during pathogen attack since persons with low immunity cannot efficiently eliminate pathogens and those with overactive immunity have a high chance of inducing cytokine storms [1] . Thus, optimal regulation of immunity is a strategy used to reduce metabolic syndrome (MetS), infection, and cancer risk [3] . No promising biomarkers of immunity are available in clinical settings, but WBC counts are generally used to assess the immune status [7] . WBC counts provide an easily accessed and reliable biomarker of overall immunity in the clinical setting. Under normal circumstances, WBC counts are considered to range from 4 to 11 × 10 9 /L. However, a WBC count of 6.2 × 10 9 /L is determined as the cutoff point for elevated metabolic syndrome risk in the Korean population [2] . Therefore, a WBC count of 4-6.2 × 10 9 /L is considered normal in a narrow definition. High WBC counts are related to cigarette smoking, splenectomy, bacterial infection, inflammatory disease, leukemia, and tissue damage. In contrast, low WBC counts are associated with bone marrow deficiency or failure, liver or spleen diseases, viral infections, cancers, cancer medication, and severe emotional or physical stress. Therefore, low and high WBC counts are involved in the etiology of different diseases. WBC counts are linked to genetic predisposition and lifestyle factors and their interactions. At the genetic level, WBC counts are associated with human leukocyte antigen (HLA)-C, HLA-G, HLA-DQA1/DRB1, interleukin (IL)-10, and cluster of differentiation 4 (CD4) polymorphisms, which are known to be involved in various immune-related diseases [8] [9] [10] [11] . Genetic variants of these polymorphisms have been mainly studied in the context of immunodeficiency [8, 12, 13] . Individuals with the HLA-C rs9264942 TT genotype demonstrate significantly higher human immunodeficiency virus type 1 (HIV-1) viral loads than those with the CC genotype [8] . The rs1518111 and rs1800872 polymorphisms of IL-10 (a potent anti-inflammatory cytokine) are associated with low CD4 T-cell counts in HIV patients [13] . HLA-B polymorphism has been reported to be associated with penicillin allergy [12] , and the HLA-DQA1/DRB1 polymorphism is found to be significantly associated with hepatocellular carcinoma development (hazard ratio 4.91, 95% CI = 1. 41-17.11 , p = 0.01) [10] . WBC counts also interact with lifestyles, but only a few studies have addressed their interactions. We aimed to identify the genetic variants associated with low-WBC counts (<4 × 10 9 ) and the interaction of their polygenetic risk scores (PRS) with lifestyles, including nutrient intake, smoking, and physical exercise in a large hospital-based cohort. In addition, we examined the relation between PRS and cancers and metabolic syndrome risk. Korean adults >40 years old (n = 58,645) were voluntarily recruited to participate in the hospital-based urban cohort of Korean Genome and Epidemiology Study (KoGES) organized by the Korean Center for Disease and Control during 2004-2013. The replicate study for determining obesity-related genetic variants was conducted in 5493 adults aged 40-79 years to have Korean Chip data in the Ansan/Ansung cohort. The institutional review board of the Korean National Institute of Health approved KoGES (KBP-2015-055), and the protocol of the present study was approved by the institutional review board Hoseo University (1041231-150811-HR-034-01). Written informed consent was obtained from all participants. Information on age, gender, residence area over at least the previous six months, educational level, income, smoking status, physical activities, and daily alcohol consumption were obtained during a health interview [14] . When the participants conducted moderateintensity exercise for 30 min more than three times a week (>150 min/week), they were considered as having regular exercise (physical activity). Moderate-intensity exercise included fast walking, mowing, badminton, swimming, tennis, and jogging. The participants were divided into two groups with and without regular exercise. Smoking status was categorized as current smoker, past smoker, or never-smoker [15] and alcohol consumption as nondrinker (0 g/day), mild drinker (0-20 g/day), and moderate drinker (>20 g/day) [15] . Trained technicians measured body weights, heights, and waist circumferences using standard procedures [16] . Body mass index (BMI) was calculated by dividing weight (kg) by height (m) squared. Blood was drawn after a ≥12 h fast (no food or water), and plasma and serum were separated for biochemical analysis [16] . Fasting plasma glucose concentrations and serum lipid profiles were measured using a Hitachi 7600 Automatic Analyzer (Hitachi, Tokyo, Japan). Serum high-sensitive C-reactive protein (hs-CRP) concentrations were measured by ELISA kit. WBC counts were conducted using EDTA-treated blood. Blood pressures were measured on right arms at heart height in a sitting position after resting for over 10 min. The participants were divided into two groups (<4 × 10 9 /L and ≥4 × 10 9 /L) according to WBC counts for genetic analysis of the low-WBC risk. They were also categorized into three immunity groups, including <4 × 10 9 /L, 4-6.2 × 10 9 /L, and ≥6.2 × 10 9 /L for metabolic analysis. The participants with MetS were categorized according to the 2005 revised National Cholesterol Education Program-Adult Treatment Panel III criteria for Asia as described in the previous studies [17, 18] . The participants answered the history of immunity-related diseases, including allergy, gastritis, asthma, bronchitis, and arthritis, used as confounding variables. Usual food intakes were determined using an SQFFQ developed and validated for KoGES [19] . The questionnaire included 106 food items, and participants selected item frequencies and serving sizes (from among 1/2, 1, or 2 serving sizes). Overall consumptions were calculated by multiplying item frequencies by serving sizes, and nutrient intakes were calculated from daily food intakes using a Computer-Aided Nutritional Analysis Program (ver. 3.0) developed by the Korean Nutrition Society [19] . The 106 food items were categorized into 29 food types used as independent variables in the factor analysis to determine dietary patterns. The number of factors retained after principal component analysis (PCA) was determined using eigenvalues of >1.5 and the orthogonal rotation procedure (Varimax) [20] . Food groups with factor loadings ≥0.40 significantly contribute to assigning dietary patterns. Four distinct dietary patterns, the Korean balanced diet (KBD), plant-based diet (PBD), Western-style diet (WSD), and ricemain diet (RMD), were selected for Korean dietary patterns. The factor loadings of food groups in the four dietary patterns were presented in Supplemental Table S1. DII was calculated from individual food and nutrient intakes having the potential for anti-inflammation using their dietary inflammatory weights for certain foods and nutrients (energy, 32 nutrients, four food products, four spices, and caffeine), as previously described [21] . It indicated the anti-inflammatory food intake of the participants. Since the SQFFQ did not include spice intakes, we excluded garlic, ginger, saffron, and turmeric intake from DII calculations. DIIs were calculated by multiplying the dietary inflammatory weights of the 38 food and nutrient components by daily intakes and dividing the sum of these products of 38 items by 100. Individual genetic variants were determined by the Center for Genome Science at the Korea National Institute of Health and provided for further study. Genomic DNA was extracted from whole blood, and genetic variants were determined using a Korean Chip (Affymetrix, Santa Clara, CA, USA) designed for Korean genetic research that included known disease-related SNPs [22] . Genotyping accuracy was confirmed with Bayesian robust linear modeling with the Mahalanobis Distance Genotyping Algorithm [23] . In quality control, genetic variants were selected with dish quality control (>0.82) and call rates (>98%) and excluded low-quality SNPs by Axiom Analysis Suit Guideline from Ther-moFisher (Waltham, MA, USA). All genetic variants satisfied Hardy-Weinberg equilibrium (HWE) inclusion criteria (p > 0.05), and the genotype missing rate was less than 5% [22] . The flow chart used to generate polygenetic risk scores (PRS) for low WBC count (<4 × 10 9 /L) risk is shown in Figure 1 . Participants were divided into low-WBC group (n = 4176) or a control (n = 36,551) group in the hospital-based cohort. GWAS was conducted to explore genetic variants associated with low-WBC risk at p < 0.00001 to have a big pool of SNPs to generate the best model explaining immunity-related pathways using PLINK 2.0 (http://pngu.mgh.harvard.edu/~purcell/plink (accessed on 12 January 2021)). The 602 genetic variants were selected from the GWAS, and their gene names were identified using g:Profiler (https://biit.cs.ut.ee/gprofiler/snpense (accessed on 26 January 2021)). Fifty-three genetic variants without corresponding gene names were excluded. Genes of the 549 SNPs selected were screened for immunity, and 21 genes were corresponding to 549 SNPs. The SNPs were then checked for linkage disequilibrium (LD) in the same chromosomes using LocusZoom (https://genome.sph.umich.edu/wiki/LocusZoom_Standalone (accessed on 3 February 2021), and those with weak LD (r 2 < 0.3) were included. Nineteen SNPs were left from LD analysis, and their genetic characteristics were shown in Supplemental Table S2 . As the replicate study, the adjusted ORs of the selected SNPs for the best model were analyzed for low-WBC risk in the 5493 participants in Ansan/Ansung cohort who determined genetic variants with Korean Chip. The number of participants in the case and control groups was 207 and 5286, respectively. The 19 SNPs were used to find the best model using GMDR, and the final best model included 10 SNPs. The best model was chosen with the interactions of potential genetic variants for the low-WBC count by GMDR [17] . Using GMDR, the best SNP-SNP interaction model was selected using a p-value of <0.05 by the sign rank test with trained balanced accuracy (TRBA) and testing balanced accuracy (TEBA) with adjustment for the covariates of age and gender, living area, body mass, and serum hs-CRP concentration [24] . Ten-fold cross-validation was used to check cross-validation consistency (CVC) since the sample size was larger than 1000 [24] . The risk and non-risk alleles of each SNP were counted as 1 [25] . For example, when the C allele was associated with an increased risk of the low-WBC count, TT, CT, and CC were assigned 0, 1, and 2. PRS was obtained by summing the number of risk alleles in the best model. PRS of the best model containing 2 or 7 SNPs were categorized as (0-1, 2, and ≥3) and (0-5, 6-7, and ≥8) by PRS, referred to as low-, medium-, and high-PRS groups. A high-PRS indicated a higher number of risk alleles in the best genetic variant-genetic variant interaction model. , medium-, and high-PRS groups. A high-PRS indicated a higher number of risk alleles in the best genetic variant-genetic variant interaction model. Statistical analysis was conducted using SAS version 9.3 (SAS Institute, Cary, NC, USA). Descriptive statistics for categorical variables (e.g., gender and lifestyle) were calculated based on frequency distributions according to WBC and PRS groups. Frequency distributions of categorical variables were analyzed using the chi-squared test. WBC counts were classified as <4.0 × 10 9 , 4-6.2 × 10 9 , and ≥6.2 × 10 9 /L to determine the effects of WBC counts on metabolic syndrome and its components. Adjusted means and standard errors were determined for continuous variables of the control and low-WBC group. The significant differences between the low-WBC and control groups were determined by analyzing covariance (ANCOVA) with covariate adjustment. After covariate adjustment, adjusted odds ratios (ORs) and 95% confidence intervals (CI) of metabolic syndrome and its components for low-WBC risk were calculated by multiple logistic regression analysis. Adjusted ORs and 95% confidence intervals (CIs) of PRS for low-WBC risk or MetS and its components were analyzed after adjusting for covariates. According to the different covariates, two models were included for PRS for low-WBC risk: model 1 included age, gender, residence area, survey year, income, and education level as covariate and model 2 contained the variables in model 1 plus energy intake, smoking status, physical activity, alcohol intake, autoimmune diseases, and serum hs-CRP concentrations as variates. In other logistic regression analyses, covariates in model 2 were used. Participants were categorized into high and low dietary intake groups to examine the interactions between PRS and dietary intake parameters. Two-way ANCOVA with main effects and an Statistical analysis was conducted using SAS version 9.3 (SAS Institute, Cary, NC, USA). Descriptive statistics for categorical variables (e.g., gender and lifestyle) were calculated based on frequency distributions according to WBC and PRS groups. Frequency distributions of categorical variables were analyzed using the chi-squared test. WBC counts were classified as <4.0 × 10 9 , 4-6.2 × 10 9 , and ≥6.2 × 10 9 /L to determine the effects of WBC counts on metabolic syndrome and its components. Adjusted means and standard errors were determined for continuous variables of the control and low-WBC group. The significant differences between the low-WBC and control groups were determined by analyzing covariance (ANCOVA) with covariate adjustment. After covariate adjustment, adjusted odds ratios (ORs) and 95% confidence intervals (CI) of metabolic syndrome and its components for low-WBC risk were calculated by multiple logistic regression analysis. Adjusted ORs and 95% confidence intervals (CIs) of PRS for low-WBC risk or MetS and its components were analyzed after adjusting for covariates. According to the different covariates, two models were included for PRS for low-WBC risk: model 1 included age, gender, residence area, survey year, income, and education level as covariate and model 2 contained the variables in model 1 plus energy intake, smoking status, physical activity, alcohol intake, autoimmune diseases, and serum hs-CRP concentrations as variates. In other logistic regression analyses, covariates in model 2 were used. Participants were categorized into high and low dietary intake groups to examine the interactions between PRS and dietary intake parameters. Two-way ANCOVA with main effects and an interaction term were used to investigate interactions between PRS and lifestyle parameters that affect low-WBC risk after adjusting for covariates. Statistical significance was accepted for p values < 0.05. The participants in the low-WBC group were older than those in the control group. In the low-WBC group, men were much lower than women. Adjusted ORs for genders were inversely associated with WBC counts after adjusting for MetS-related parameters, indicating men were inversely associated with the low-WBC risk (Table 1) . Mean serum hs-CRP concentration was higher in the middle-WBC and high-WBC group than in the low-WBC group, and serum hs-CRP concentration was inversely associated with WBC counts by 0.542-fold. Cancer incidence was higher in the low-WBC group than in the other groups, and adjusted ORs were positively associated with WBC count by 1.467-fold (cutoff: <4.0) ( Table 1 ). The prevalence of MetS was much higher in the low-WBC group than in the other groups, and the components of MetS, including waist circumferences, plasma glucose, total cholesterol, LDL, and TG, concentrations, SBP, and DBP, showed the same trends. MetS was inversely associated with a low-WBC count (<4.0 × 10 9 /L) by 0.458-fold. Mean waist circumference was higher in the control group than in the low-WBC group, and hip circumference was inversely associated with WBC by 0.86-fold (Table 1) . The cutoff points were as following: 1 <55 years old; 2 <0.5 mg/dL for high-sensitive C-reactive protein (hs-CRP), 3 <25 mg/kg 2 for body mass index (BMI), 4 <25% for men and <30% for women; 5 <90 cm for men <85 cm for women; 6 <126 mg/dL fasting serum glucose and 7 <6.5% HbA1c or taking hypoglycemic medication; 8 <230 mg/dL serum total cholesterol, 9 ≤40 mg/dL for men and ≤50 mg/dL serum HDL; 10 <160 mg/dL serum LDL, 11 <150 mg/dL serum triglyceride; 12 <130 mmHg SBP and 13 <90 mmHg DBP or taking hypotensive medication. 14 Values represent adjusted means and 95% confidence intervals (CI) after adjusting for covariates or 15 the number of the subjects and percentage. Covariates used were age, sex, body mass index (BMI), energy intake, income, education, residence area, survey year, and autoimmunity-related diseases, including atopic dermatitis, asthma, allergy, and inflammation-related diseases, alcohol intake, smoking status, and physical activity. 16 Adjusted odds ratio (ORs) and 95% confidence intervals of each parameter for the hypo-WBC (<4.0 × 10 9 /L) risk after adjusting for covariates in logistic regression analysis. a,b,c Different letters indicate significant differences among the groups in the Tukey test at p < 0.05. *** Significantly different for WBC count groups by one-way ANCOVA in continuous variables at p < 0.001. +++ Significantly different WBC count groups by χ 2 test at p < 0.001. hs-CRP, high-sensitive C-reactive protein; HbA1c, blood hemoglobin A1c; LDL, low-density lipoprotein; HDL, high-density lipoprotein; TG, triglyceride; SBP, systolic blood pressure; DBP, diastolic blood pressure. Proportions of smokers and former smokers were much lower in the low-WBC group than in the other groups, and smokers and former-smokers were inversely related to low-WBC risk by 0.298-and 0.352-folds on the reference of the non-smokers (Table 2) . However, the proportion of individuals that exercised regularly was higher in the low-WBC group than in the other groups, and regular exercise was positively associated with low-WBC risk by 1.262-fold. There was a higher proportion in the low-WBC group in low alcohol and coffee intake than the other groups (Table 2 ). Alcohol and coffee intakes were inversely associated with low-WBC risk by 0.849-and 0.856-fold, respectively. Participants in the Nutrients 2021, 13, 2849 7 of 15 low-WBC group had lower energy intake and fat intakes than those in the high-WBC group. However, protein intakes were non-significantly different. Interestingly, vitamin D intakes were significantly higher in the low-WBC group than in the high-WBC group (Table 2) , and inflammation indices were not significantly different. 1 Values represent the number (%) and 2 adjusted means ± standard deviations after adjusting for covariates or the number of the subjects and percentage. Covariates used were age, sex, body mass index (BMI), energy intake, income, education, residence area, survey year, having autoimmunity-related diseases including atopic dermatitis, asthma, allergy, having inflammation-related diseases including gastritis, alcohol intake, smoking, and physical activity. 3 Adjusted odds ratio (ORs) and 95% confidence intervals of each parameter for the hypo-WBC (<4.0 × 10 9 /L) risk the after adjusting for covariates using logistic regression analysis. The cutoff points were as follows: 4 < moderate intensity activity for 150 min/week; 5 <20 g alcohol/day, 6 <3 cups/week, 7 0.05) and MAF (>0.05) criteria (Table 3 ). In Ansan/Ansung cohort, the selected 10 SNPs exhibited similar OR values to those in a hospital-based cohort, but the significance levels were higher in Ansan/Ansung cohort than those in the hospital-based cohort (Table 3 ) since the number of cases (n = 204), and control (n = 5286) was much smaller in Ansan/Ansung cohort. Gene-gene interaction models with 2, 7, 8, 9, or 10 SNPs met the best model criteria. Among them, adjusted ORs of the 2 and 7 SNP models increased low-WBC risk (<4.0 × 10 9 ) by 1.844 (1.165-2.918) and 2.123 (1.741-2.589) folds in participants with high-PRS than those with low-PRS ( Figure 2 ). The best model with 2 SNPs included proteasome 26S Subunit, non-ATPase 3 (PSMD3)_rs9898547 and lactase (LCT)_rs80157389, while that with 7 SNPs contained the SNPs in the 2 SNP model, HLA-DRB1_rs532162239 and rs3097649, HLA-C_rs2308575, cyclin-dependent kinase inhibitor 1A (CDKN1A)_rs3176337, and thyroid hormone receptor alpha (THRA)_rs7502539 (Table 4) . Although the 2 SNPs model sufficiently showed increases in the risk of a low-WBC count, we considered that 2 SNPs might be too small to show associations with metabolic syndrome. Additionally, variation was much greater in the 2 SNP model than in the 7 SNP model. Therefore, we used the 7 SNP model to investigate interactions with lifestyles. Table 3 . The characteristics of the ten genetic variants of genes related to immunity in the risk of low WBC count (<4.0 × 10 9 /L) and used for the generalized multifactor dimensionality reduction analysis. PRS calculated using the 2-and 7-SNPs models were divided into three categories (0-1, 2, and ≥3) or (0-5, 6-7, and ≥8), respectively. Adjusted ORs were obtained by logistic regression after adjusting for various covariates. Two models were composed of different covariates: Model 1 included age, gender, residence area, survey year, income, and education level as covariates, and model 2 contained the variables in model 1 plus energy intake, smoking status, physical activity, alcohol intake, autoimmune diseases, and serum high-sensitive C-reactive protein concentrations as variates. The low-PRS group was used as a reference for logistic regression. Based on the covariates, red and blue boxes indicated adjusted ORs for models 1 and 2, respectively, and lines indicated 95% CIs. There was no significant association between PRS and MetS after adjusting for covariates (Supplementary Table S3 ). BMI and body fat mass also did not have an association with PRS (Supplementary Table S3 ). MetS components including waist circumferences, plasma glucose, HDL, triglyceride concentrations were not associated with PRS after adjusting for covariates (Supplementary Table S3 ). Serum hs-CRP concentrations did not have any relation with PRS (Supplementary Table S3 ). Adjusted odds ratios and 95% confidence intervals of the PRS of 2-and 7-SNP models generated from assessing gene-gene interactions associated with a low white blood cell (WBC) count. PRS of the 2-and 7-SNPs were calculated by summing the number of risk alleles of SNPs. PRS calculated using the 2-and 7-SNPs models were divided into three categories (0-1, 2, and ≥3) or (0-5, 6-7, and ≥8), respectively. Adjusted ORs were obtained by logistic regression after adjusting for various covariates. Two models were composed of different covariates: Model 1 included age, gender, residence area, survey year, income, and education level as covariates, and model 2 contained the variables in model 1 plus energy intake, smoking status, physical activity, alcohol intake, autoimmune diseases, and serum high-sensitive C-reactive protein concentrations as variates. The low-PRS group was used as a reference for logistic regression. Based on the covariates, red and blue boxes indicated adjusted ORs for models 1 and 2, respectively, and lines indicated 95% CIs. Table 4 . The characteristics of the ten genetic variants of genes in the risk of low white blood cell count applied for the generalized multifactor dimensionality reduction analysis (GMDR). Adjusted There was no significant association between PRS and MetS after adjusting for covariates (Supplementary Table S3 ). BMI and body fat mass also did not have an association with PRS (Supplementary Table S3 ). MetS components including waist circumferences, plasma glucose, HDL, triglyceride concentrations were not associated with PRS after adjusting for covariates (Supplementary Table S3 ). Serum hs-CRP concentrations did not have any relation with PRS (Supplementary Table S3 ). No interactions were observed between PRS and age, gender, BMI, or metabolic syndrome that affected low-WBC count risk (p > 0.05). There was no interaction between lifestyles (except fat intake), PRS, and low-WBC risk (p = 0.008; Table 5 ). In participants with a high fat intake, those with a low-PRS had much higher WBC counts than those with a medium or high-PRS ( Figure 3 ). This trend was similar in the low-fat intake. The association between PRS and WBC counts was greater for participants with high-fat intakes than low-fat intakes (Table 5) . Table 5 . Adjusted odds ratios of polygenetic risk scores of the best model (PRS) for the hypo-WBC risk after covariate adjustments according to lifestyles patterns and the interaction of PRS with lifestyles for the hypo-WBC risk. High-PRS (n = 26,899) According to the low and high intake groups, values represent adjusted odds ratio (OR) and 95% confidence intervals. Covariates were age, sex, body mass index (BMI), energy intake, income, education, residence area, survey year, taking immune-related medicine, alcohol intake, smoking status, and physical activity. The cutoff points were as follows: 1