Bidirectional Mendelian Randomization Analysis for Vitamin D and Thyroid Peroxidase Antibody

Purpose Whether Vitamin D deficiency or insufficiency is associated with thyroid autoimmunity was debated for long time. This study was still to explore the causal relationship of 25 (OH) D with a thyroid peroxidase antibody (TPOAb). Methods The data were obtained from a cross-sectional study, SPECT-China study, which was performed in 23 sites in East China during 2014 to 2016. 10636 participants were finally included in this study. Genotyped four 25 (OH) D-related and four TPOAb-associated single nucleotide polymorphisms (SNPs) created their genetic risk scores (GRS). Bidirectional mendelian randomization (MR) analysis was used in this study. Results 25 (OH) D GRS was significantly associated with 25 (OH) D (B −0.093, 95% CI −0.111, −0.074) and TPOAb level (B 0.067, 95% CI 0.002 to 0.132). TPOAb GRS was significantly associated with TPOAb concentration (B 0.345, 95% CI 0.135 to 0.556), but not 25 (OH) D (B −0.030, 95% CI −0.091 to 0.030). Using 25 (OH) D_GRS as instrumental variable in the MR analysis, a causal relationship of genetically determined 25 (OH) D with increased TPOAb concentration (B −0.720, 95% CI −1.429 to −0.012). No relation was found between genetically instrumented TPOAb and 25 (OH) D. Conclusion A higher VD_GRS was associated with higher risk of increased TPOAb concentration, which supports a causal association between decreased vitamin D and increased concentration of TPOAb in an eastern Chinese population.


Introduction
yroid dysfunction including overt and subclinical hypothyroidism in the population has considerable consequences for a number of health issues, including insulin resistance, metabolic syndrome, worse lipid profile, central adiposity, and obesity [1][2][3][4].
e most common cause of hypothyroidism is the Hashimoto thyroiditis (HT) and the basic mechanisms in the development of thyroid autoimmunity may be due to a combined TPO-and Tg-specific cytotoxic immune response [5]. It was reported that the prevalence of detectable thyroid antibodies, primarily TPOAb, comprises 10-12% of the healthy population [6][7][8]. Despite the prevalence and adverse outcomes of autoimmune-mediated thyroid disease, its etiology remains incompletely understood [9,10].
Vitamin D deficiency is also a pandemic health problem in both developing and developed countries [11]. Recently, the actions of vitamin D have been shown to go beyond calcium/phosphorus homeostasis via bone formation and resorption to higher susceptibilities of immune-mediated disorders, including chronic infections and autoimmune diseases [12]. Several epidemiological studies showed lower vitamin D levels to the pathogenesis of increasing TPOAb [6,[13][14][15][16][17]. However, conflicting studies were also present reporting that no significant association between the serum vitamin D levels and thyroid autoimmunity [18][19][20]. us, whether low vitamin D levels truly associated with AITD, whether the association is causal, and if so, its causal direction, is still unclear.
e Mendelian randomization (MR) approach was taken widely used for assessing causality in population studies [21], which is the main limitation of a cross-sectional study. Using the genetic variants as the instrumental variable (IV) has become a widely-used approach for causal inference [22]. In this study, if low 25 (OH) D causally induces high TPOAb, genetic variants associated with lower 25 (OH) D should be associated with higher TPOAb concentration, and vice versa. ese genetic variants are inherited independent of potential confounding factors [22]. us, MR could avoid problems in conventional epidemiological studies such as residual confounding and reverse causation [23].
In the present study, on the basis of the large community-based sample of Chinese participants from SPECT-China study (Survey on Prevalence in East China for metabolic diseases and risk factors), we performed a bidirectional MR approach to explore the causal association between increased TPOAb levels and decreased 25 (OH) D levels. TPOAb and Vitamin D genetic risk scores (TPOAb_GRS and VD_GRS) were constructed to represent the genetic susceptibility.

Study Participants.
e data were from the SPECT-China study (ChiCTR1900021356), which is a large crosssectional study. Recruitment and enrollment of the study have been previously described in detail [24][25][26]. From 2014 to 2016, 12666 subjects who were Chinese citizens, ≥18 years old, and had lived in their current area for ≥6 months were recruited for the SPECT-China study from 23 sites in Shanghai, Zhejiang, Jiangsu, Anhui, and Jiangxi Province. Among them, genotype information was available in 10672 participants (84.3%). We excluded the participants who missed information on more than two single nucleotide polymorphism (SNP) genotypes (n � 20), missing the concentration of TPOAb (n � 14) and 25 (OH) D (n � 2). 10636 participants were involved in the final analysis. All participants provided written informed consent before data collection.
e study protocol was approved by the Ethics Committee of Shanghai Ninth People's Hospital, Shanghai JiaoTong University School of Medicine. All procedures followed were in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national) and with the Helsinki Declaration of 1975, as revised in 2008.

Measurements.
A single assessment protocol of interview and collection of biological specimens at each site was undertaken. Blood samples of each participant were obtained from 7 : 00 Am to 10 : 00 Am after fasting for at least 8 hours. Blood samples were refrigerated immediately after phlebotomy, and after 2∼4 hours they were centrifugation and the serum was aliquoted and frozen in a central laboratory. TPOAb was measured by the chemiluminescence immunoassay (Siemens, immulite 2000, Erlangen, Germany) and the 25 (OH) D was detected using a chemiluminescence assay (Siemens ADVIA Centaur XP, Germany). Body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared.

Statistical
Analysis. Data were analyzed by using IBM SPSS Statistics, Version 22 (IBM Corporation, Armonk, NY, USA). All analyses were two-sided. A P value <0.05 was considered significant. Continuous variables were expressed as the mean (±standard deviation) values, and categorical variables were presented as percentages. 25 (OH) D and TPOAb was logarithmically transformed before analysis. e additive genetic model for each SNP (coded as 0, 1 and 2) was used to construct GRS. For the VD_GRS, we created a weighted score by multiplying each SNP by a weight based on its effect size with 25 (OH) D obtained from a large study containing Asian population [30]. For the TPOAb_GRS, the weights were from meta-analysis of Atherosclerosis Risk In Communities study (ARIC) and Study of Health in Pomerania-TREND (SHIP) [29]. e characteristics of each SNP in the VD_GRS and TPOAb_GRS are summarized in Table 1.
To examine the strength of the allele scores as instruments, the F-statistic was approximated from the proportion of variation in the respective phenotype (R 2 ) explained by the allele score, (F − stat � (R 2 * (n − 2))/(1 − R 2 )) [31].
Linear regression analyses were used to determine the association of the two GRSs and TPOAb with 25 (OH) D, and the association of the two GRSs and 25 (OH) D with TPOAb. Model 1 adjusted for age, sex. Model 2 adjusted for terms in model 1. Model 2 adjusted for terms in model 1 and waist to hip ratio.
Regarding the MR analysis, the weighted VD_GRS and weighted TPOAb_GRS were used as the instrumental variables (IV) estimators to measure the strength of the bidirectional causal relationship between 25 (OH) D and TPOAb concentration. e formal MR analyses to estimate the possible causal effect of 25 (OH) D on TPOAb (and vice versa) were conducted using the IV ratio method [22]. For the causal association of increased risk of TPOAb in relation to lower 25 (OH) D, the computational formula was β (VD-TPOAb) � β (VD_GRS-TPOAb) /βVD_GRS-25 (OH) D). In the opposite direction, the computational formula was e model was the same as the above model 2, adjusting for age, sex, and waist to hip ratio. e standard error (SE) and confidence interval (CI) for the IV estimators was estimated by the delta method. e formulas are shown below: (1) To validate the genetic instruments, we assessed the associations between each individual SNP with 25 (OH) D and TPOAb, respectively. We also measured the potential pleotropic associations of each individual SNP and the GRSs with age, sex, BMI, and waist to hip ratio.

Association of Four 25 (OH) D-Related SNPs with 25 (OH) D and TPOAb.
e associations of each individual 25 (OH) D-related SNP with ln-TPOAb are summarized in Figures 1(a) and 1(b). In these four 25 (OH) D-related SNPs, two SNPs located at the GC (rs2282679) and DHCR7 (rs12785878) loci were significantly associated with the level of 25 (OH) D and none of them were significantly associated with the concentration of TPOAb.

Association of Four TPOAb-Related SNPs with 25 (OH) D and TPOAb.
e associations of each individual TPOAbrelated SNP with ln-25 (OH) D are summarized in Figures 1(c) and 1(d). In these four TPOAb-related SNPs, two SNPs located at the HLA-DPB1 (rs9277555) and TPO (rs11675434) loci were significantly associated with the level of TPOAb and none of them were significantly associated with the level of 25 (OH) D.

Pleiotropic Effects of SNPs and Weighted GRS.
e association of these four 25 (OH) D-related SNPs and four TPOAb-related SNPs with major 25 (OH) D and TPOAb related confounders were calculated. erefore, we measured the potential associations of the SNPs with age, waist to hip ratio, BMI, and sex distribution using an additive model. Unstandardized coefficients (standard error) and odds ratio (95% confidence interval) are summarized in Table 2. None of these eight SNPs had pleiotropic effects (all P > 0.05).
en, the association of VD_GRS and TPOAb-GRS with major 25 (OH) D and TPOAb related confounders were calculated for further analysis ( Table 3). Neither of them has significant association with these confounders (all P > 0.05).

Study Characteristics According to Weighted VD_GRS and
TPOAb_GRS Tertiles. We classified study subjects into three groups according to VD_GRS tertiles (Q1: ≤0.82, Q2: 0.83-1.13, Q3: ≥1.14) and TPOAb_GRS tertiles (Q1: ≤0.136, International Journal of Endocrinology   e general characteristic of study subjects in terms of these two tertiles is shown in Table 4. As expected, with the increasing of VD_GRS, 25 (OH) D concentrations significantly decreased and with the increasing of TPOAb_GRS, TPOAb concentrations significantly increased (P < 0.001 and P � 0.021). TPOAb concentrations increased significantly along with the increased VD_GRS levels (P � 0.019). However, there is no significant difference of 25 (OH) D levels between TPOAb_GRS tertiles (P > 0.05). Table 5, in this cross-sectional study, 25 (OH) D was positively associated with the level of TPOAb after adjusted for age, sex, and waist to hip ratio (B � 0.070, 95% CI 0.003, 0.017). en, the association of VD_GRS and tertiles of VD_GRS with TPOAb concentration was measured. Increased VD_GRS was significantly associated with increased TPOAb levels after adjusting for age and sex (B � 0.066, 95% CI 0.002, 0.129) (model 1). Further adjusting for waist to hip ratio did not change the result (B � 0.067, 95% CI 0.002, 0.132) (model 2). e same trend was also seen in the tertiles of VD_GRS (P for trend � 0.027 in model 1 and 0.031 in model 2). Table 6.

Associations of TPOAb_GRS and TPOAb with 25 (OH) D Concentration. Conversely, the association of the TPOAb_GRS and TPOAb with 25 (OH) D is shown in
Although the TPOAb level was significantly associated with the level of 25 (OH) D, increased TPOAb_GRS was not significantly associated with decreased level of 25 (OH) D in both models. e tertiles of TPOAb_GRS showed similar results (both P for trend >0.05).

Discussion
is investigation including 10636 community-dwelling Chinese adults, we examined whether these two GRSs, composed of SNPs significantly associated with genetically determined 25 (OH) D and genetically determined TPOAb concentration, were associated with increased TPOAb concentration and decreased 25 (OH) D level, respectively. International Journal of Endocrinology Using the bidirectional MR study design, we found a causal role of vitamin D in the pathogenesis of increasing TPOAb level, while no causal relationship of higher TPOAb concentration to induce lower vitamin D status was found. To the best of our knowledge, for the first time, the results provided novel evidence for a causal relationship between genetically determined Vitamin D and increased TPOAb concentration by using MR. e identification of a causal relationship between 25 (OH) D and TPOAb concentration may have important clinical implications because vitamin D deficiency is common, and vitamin D supplementation is relatively safe and cost-effective [32].
e MR approach has an important benefit that it helps to overcome problems of confounding and reverse causality, which limits the ability to draw causal inferences in nongenetic observational studies [22]. However, several assumptions should be well met in performing a MR analysis [33][34][35]. First, the GRS as an IV was strongly associated with the exposure of interest. All SNPs used in this study have previously been shown to be significantly associated with vitamin D or TPOAb concentration in large meta-analysis of GWAS [29,30]. In the present study, the associations of these two GRSs with the two corresponding exposures were also very significant. Second, the IVs must not be correlated with any confounders of the exposure-outcome association. In our study, we found the two GRSs were not associated with age, BMI, waist to hip ratio, and sex which were common potential confounders of the vitamin D-TPOAb association. We further tested the pleiotropic effects of each SNP on the above confounders and the results showed no SNP had pleiotropic effects. ird, the IV related to outcome only through the exposure of interest. ere should be no direct effect of genotype on disease or any other mediated effectors other than through the exposure of interest. us, we analyzed the association of each SNP and GRSs with the corresponding outcome (vitamin D or TPOAb). All the SNPs and GRSs were not significantly associated with the outcome.   Recently, in a nationwide population-based study, data were obtained from the Korea National Health and Nutrition Examination Survey VI-1 and 2 (2013 and 2014) showed that the higher TPOAb level was more prevalent in the vitamin D-deficient group and vitamin D deficiency affects thyroid autoimmunity and dysfunction in iodine-replete area [36]. In our present study, we found that higher VD_GRS, presenting a low 25 (OH) D level, had a significantly negative association with higher TPOAb, providing strong evidence in support of a causal role of decreased vitamin D on increased TPOAb. ese findings are consistent with evidence from observational studies that have demonstrated that low vitamin D levels influence risk of an increasing TPOAb level, and also in line with a very current case-control study which enrolled 200 euthyroid subjects: 100 newly diagnosed HT patients and 100 healthy individuals, matched for age, sex, and BMI, which aimed to investigate the association of HT with vitamin D status and SNPs of the vitamin D receptor (VDR). It first suggested that vitamin D deficiency may contribute to HT development and/or progression, acting as an environmental trigger [37].
Further, very limited intervention studies testing the effect of vitamin D supplementation on patients with AITD/HT showed that vitamin D supplements decreased TPOAb concentration, in those with vitamin D deficiency or those with normal vitamin D status [38,39]. However, another randomized, double-blind, controlled trial (RCT) suggested Vitamin D 3 supplementation did not affect the TPOAb level [40]. It should be noted that there exists an important difference between MR studies and RCTs. MR studies assess the association of a lifetime of exposure in the general population, whereas RCTs provide insights from supplementation for shorter periods in individuals at risk [41]. us, long-term RCTs may be needed to assess the role of vitamin D supplementation on the treatment of increased TPOAb adequately.
e results from our study provides rationale to further investigate whether vitamin D supplementation may reduce AITDs susceptibility. e identification of vitamin D as a causal susceptibility factor for TPOAb may have important public health implications since vitamin D insufficiency/deficiency is common [11,42], and vitamin D supplementation is both relatively safe and cost-effective [41]. e strength of our study included a relatively large sample size (more than 10 000 participants), well-defined community setting, and a highly homogeneous population. To our acknowledgment, this is the first report exploring the causal association between low vitamin D and high TPOAb concentration using the bidirectional MR study design, creating VD_GRS and TPOAb_GRS, representing the established common genetic variants of vitamin D and TPOAb level used as the IV.
However, several limitations should be acknowledged. First, it should be noted that the TPOAb_GRS we created in this study only for TPOAb concentration, not representing TPOAb positivity. Further studies were needed to explore the causal association between TPOAb positivity and vitamin D. Second, all participants were of were Han Chinese, a majority ethnic group indigenous within China (constitute about 92% of the population of the People's Republic of China). e findings of this study may not be generalizable other ethnicities. ird, 25 (OH) D was measured only once at baseline. Hence, we were not able to control intraindividual variability. Fourth, we build up our GRSs only based on common variants, which were considered to represent limited TPOAb and vitamin D heritability. We were unable to assess the potential contribution of rare variants.

Conclusion
We found that a higher VD_GRS was associated with higher risk of increased TPOAb concentration. is analysis provides evidence supporting a causal association between decreased vitamin D and increased concentration of TPOAb in an eastern Chinese population. Additional studies are needed to validate our findings and elucidate the mechanisms behind these findings.

Abbreviations
TPOAb: yroid peroxidase antibody SPECT-China: Survey on prevalence in east China for metabolic diseases and risk factors GRS: Genetic risk scores MR: Mendelian randomization HT: Hashimoto thyroiditis 25 (OH) D: 25-hydroxyvitamin D IV: Instrumental variable.

Data Availability
e data used to support the findings of this study are included within the article.

Disclosure
is article was presented as a poster presentation in the 88th Annual Meeting of the American yroid Association by Yi Chen. e funders played no role in the design or conduct of the study, collection, management, analysis, or interpretation of data or in the preparation, review, or approval of the article.

Conflicts of Interest
e authors declare that they have no conflicts of interest.

Authors' Contributions
Yingli Lu designed, performed, and supervised this investigation and had full access to all of the data and took responsibility for the integrity of the data and the accuracy of the data analysis. Yi Chen contributed to the discussion, interpretation of the data, and critical revision of the manuscript for important intellectual content. Yingchao Chen and Bing Han contributed equally to this work. ey did perform this investigation, analyzed the data, contributed to the discussion, performed interpretation of the data, and wrote the manuscript. Chunfang Zhu, Qin Li, Chi Chen, Hualing Zhai provided technical or material support, and contributed to the discussion. All authors read and approved International Journal of Endocrinology the final manuscript. Yingchao Chen and Bing Han contributed equally to this work.