Circulating white blood cells and lung function impairment: the observational studies and Mendelian randomization analysis

Abstract Background Circulating white blood cell (WBC) counts have been related to lung function impairment, but causal relationship was not established. We aimed to evaluate independent effects and causal relationships of WBC subtypes with lung function. Methods The 19,159 participants from NHANES 2011–2012 (n = 3570), coke-oven workers (COW, n = 1762) and Dongfeng-Tongji (DFTJ, n = 13,827) cohorts were included in the observational studies. The associations between circulating counts of WBC subtypes and prebronchodilator lung function were evaluated by linear regression models and LASSO regression was used to select effective WBC subtypes. Summary statistics for WBC-associated SNPs were extracted from literature, and Mendelian randomization (MR) analysis with inverse-variance weighted (IVW) method was applied to estimate the causal effects of total WBC and subtypes on lung function among 4012 subjects from COW (n = 1126) and DFTJ cohorts (n = 2886). Results Total WBC counts were negatively associated with lung function among three populations and their pooled analysis indicated that per 1 × 109 cells/L increase in total WBC was associated with 36.13 (95% CI: 30.35, 41.91) mL and 25.23 (95% CI: 19.97, 30.50) mL decrease in FVC and FEV1, respectively. Independent associations with lung function were found for neutrophils, monocytes, eosinophils and basophils (all p < .05), except lymphocytes. Besides, IVW MR analysis showed that genetically predicted total WBC and neutrophil counts were associated with reduced FVC (p = .017 and .021, respectively) and FEV1 (p = .048 and .043, respectively). Conclusions WBC subtypes were independently associated with lower lung function except lymphocytes. Our findings suggest that circulating neutrophils may be causal factors in lung function impairment. KEY MESSAGES White blood cell (WBC) subtypes were negatively associated with lung function level except lymphocytes in the observational studies. Associations of WBC subtypes with lung function may be modified by sex and smoking. Mendelian randomization analysis shows that neutrophils may be causal factors in lung function impairment.


Supplementary materials
Pearson's correlation coefficients between different WBC subtypes.

Table S2
Relationships of total and differential WBC counts with lung function among subjects without cancers (single-marker model).

Table S3
Relationships of total and differential WBC counts with lung function among subjects without using anti-infectious drugs (single-marker model).

Table S4
Characteristics of SNPs associated with total and differential WBC counts in published GWAS.

Table S5
The associations of total and differential WBC counts related SNPs with confounders in the COW and DFTJ studies.

Table S6
Between-instrument heterogeneity test for the Mendelian randomization analysis of total and differential WBC counts with lung function.

Definitions of common lifestyles
In the NHANES 2011-2012, subjects with a positive answer to the question "Have you smoked at least 100 cigarettes in your entire life" were classified as smokers; otherwise, they were classified as non-smokers [1]. Subjects with at least 12 alcohol drinks per year or in lifetime were classified as alcohol drinkers; otherwise, they were classified as non-alcohol drinkers [2]. Subjects with vigorous work activity ≥75 min/week or moderate work activity ≥150 min/week or a combination ≥150 min/week were classified as exercisers; otherwise, they were classified as non-exercisers [3].
In the COW and DFTJ studies, participants who have smoked >1 cigarette/day for >6 months were classified as smokers; otherwise, they were classified as non-smokers.
Participants who have drunk alcohol >1 time/week for >6 months were classified as alcohol drinkers; otherwise, they were classified as non-alcohol drinkers [4]. During the last 6 months, participants who have spent ≥30 min at a time and ≥5 times/week exercising in leisure time were classified as exercisers; otherwise, they were classified as non-exercisers.

SUPPLEMENTARY NOTES
Single-marker regression, multiple-marker regression, and interaction analyses were performed with SAS 9.4, and LASSO regression and MR analyses were conducted using R 3.5.3 software. SAS and R codes were listed as follows.   .001 †Eosinophil and basophil counts were transformed by common logarithm (log10) to approximate normal distribution.

SAS codes for single-marker regression analysis
Total and differential WBC counts Table S3 Relationships of total and differential WBC counts with lung function among subjects without using anti-infectious drugs (single-marker model). Notes: *Total and differential WBC counts were included in the multiple linear regression model separately, and the model was adjusted for age, sex, race (only in NHANES 2011-2012 population), height, smoking, alcohol use, and exercise. 192.43 Abbreviation: WBC, white blood cell; GWAS, genome-wide association study. *These SNPs were not genotyped or imputed in the participants of present research.
Notes: Summary statistics were derived from the largest GWAS of total and differential WBC in the Asian population [5]. Total WBC counts were standardized by Z-score, and the study sample size was 107,964. Neutrophil, monocyte, eosinophil, and basophil counts were standardized by rank-based inverse normal transformation and the study sample size for the genome-wide analysis was 62,076.            Notes: Effect estimates were obtained from linear regression model for height and logistic regression model for smoking, drinking and exercise, with adjustment for age, gender, and chip types (only in DFTJ study). Fixed-effect (heterogeneity P ≥0.05) or random-effect (heterogeneity P <0.05) meta-analysis was used to combine results from COW and DFTJ studies (n=4,012).
*These SNPs were significantly associated with the covariates at the Bonferroni-corrected level and they were then excluded as IVs in the following MR analysis. †These SNPs were identified as potential outliers by MR-PRESSO method and they were further excluded as IVs in the following MR analysis.