Associations of metal mixtures with thyroid function and potential interactions with iodine status: results from a cross-sectional study in MEWHC

Few studies are available on associations between metal mixture exposures and disrupted thyroid hormone homeostasis; particularly, the role of iodine status was ignored. Here, we aimed to explore the cross-sectional relationship of blood cell metals with thyroid homeostasis and explore the potential modifying effect of iodine status. Among 328 workers from the manganese-exposed workers healthy cohort (MEWHC), we detected thyroid function parameters: thyroid stimulating hormone (TSH), total triiodothyronine (TT3), free triiodothyronine (FT3), total tetraiodothyronine (TT4), free tetraiodothyronine (FT4) as well as calculated sum activity of peripheral deiodinases (GD) and thyroid’s secretory capacity (GT). Inductively coupled plasma mass spectrometry (ICP-MS) was used to measure 22 metal concentrations in blood cells. Based on the consistent results of least absolute shrinkage and selection operator (LASSO) and Bayesian kernel machine regression (BKMR) analyses, there were significant positive associations between copper and TSH (β = 2.016), iron and FT4 (β = 0.403), titanium and GD (β = 0.142), nickel and GD (β = 0.057), and negative associations between copper and FT4 (β =  − 0.226), selenium and GD (β =  − 0.332), among the participants. Interestingly, we observed an inverted-U shape relationship between magnesium and FT4. Furthermore, we found a synergistic effect between arsenic and copper on the TSH level, while antagonistic effects between nickel and copper as well as nickel and selenium on the TSH level. We observed a modified effect of iodine status on association between strontium and GD (Pinteraction = 0.026). It suggests metal mixture exposures can alter thyroid homeostasis among the occupational population, and deiodinase activity had a modified effect on association between strontium and GD. Validation of these associations and elucidation of underlying mechanisms require further researches in the future.


Introduction
Thyroid gland is highly susceptible to environmental pollutants (Gorini et al. 2020), and disrupted thyroid homeostasis has shown associations with various adversely health outcomes, such as thyroid disease, cardiovascular disease, type 2 diabetes, and so on (Jabbar et al. 2017;Rong et al. 2021).
Recently, evidence emerged and showed links between metal exposure and thyroid homeostasis disruption, for instance, lead, mercury cadmium, and arsenic (Chen et al. 2013;Yorita Christensen 2013;Nie et al. 2017;Guo et al. 2018). However, few studies were available on trace metals, for example barium, molybdenum and selenium (Yorita Christensen 2013; Guo et al. 2018). On the other hand, the most were focused on general population or pregnant women in USA (Chen et al. 2013;Kim et al. 2022), Greece (Margetaki et al. 2021), Korea (Kim et al. 2021), and China (Guo et al. 2018;Sun et al. 2019;Wang et al. 2020;Hu et al. 2021;Wu et al. 2021); moreover, they reported inconsistent observations. Limited evidence was available on occupational population exposed to metal though the complex exposure model and relative high level of metal exposure.
In reality, the health effects of mixture exposures and single exposure model are no longer equivalent (Wu et al. 2016;Conley et al. 2021), and interactive effect may play a role. Indeed, Katerina et al. found cadmium may be a synergistic agent of lead (Margetaki et al. 2021). Arsenic has a synergistic effect with cadmium and mercury, and cadmium and mercury can enhance the harmful effect of arsenic on thyroid function during pregnancy (Wang et al. 2020). Metal mixture exposures should be assessed more comprehensively in order to better understand their effects on thyroid function. Disruption in iodine transport or deiodinases might be the potential mechanisms of thyroid dyshomeostasis related to environmental chemicals (Miller et al. 2009). To date, limited evidence is available on those indicators involved in associations between metal exposures and thyroid hormonal effects (Krieg 2016;Nie et al. 2017;Chung et al. 2019). In addition, iodine acts as a component of triiodothyronine and tetraiodothyronine and contributes to thyroid hormone synthesis in an irreplaceable way. But too much iodine inactivates deiodinase, especially if selenium is deficient. It was suggested that thyroid hormone-disrupting toxicant exposure in pregnancy have adverse effects on fetal development, particularly in those whose iodine status was deficiency (Demeneix 2019). However, the studies that assessed potential interactions between metal exposure and iodine status on thyroid hormones were sparse.
In this article, we aimed to explore the relationship of metal mixture exposures with thyroid function parameters as well as explore potential interactions with iodine status among participants from a ferro-manganese refinery.

Study population and data collection
The participants were from manganese-exposed workers healthy cohort (MEWHC) focused on exposure biomarkers, early health effects, and susceptibility; while taking it into consideration that diseases are associated with metal exposure from the occupational settings. The cohort has been described in more detail previously (Lv et al. 2014;Zhou et al. 2018). Briefly, we recruited subjects who admitted occupational health examination in 2021. An interview was conducted face-to-face to gather demographic information as well as lifestyle information, for instance smoking status, drinking status as well as tea consumption. This study involved 708 participants, and we selected 341 individuals from the typical occupation. The inclusion criterion was (1) outliers whose concentrations of metals exceeded their three times 99th (n = 4); (2) had thyroid diseases (n = 6); and (3) lacked data on urinary iodine (n = 3). Finally, analysis was performed among 328 participants.
Participants were asked to fast over night before donating their blood. Serum, plasma, and blood cells were separated and stored frozen (− 80 °C) before we performed metal concentration detection procedure. Prior to the collection of samples and face-to-face interviews, an informed consent form was signed by each participant. The study was approved by the medical ethics committee at Guangxi Medical University (ID:20220017).

Thyroid hormone testings and related endpoints
TSH, TT3, FT3, TT4, and FT4 in serum were detected at the First Affiliated Hospital of Guangxi University of Chinese Medicine with an automatic immune analyzer (Abbott I2000SR, USA). To determine the accuracy of specimen measurements, we established a standard curve and conduct quality control testing in accordance with the kit instruction. Reference ranges for normal values for TSH, TT3, FT3, TT4, and FT4 were 0.35-4.94 mIU/L, 0.98-2.33 nmol/L, 2.43-6.02 pmol/L, 62.68-150.8 nmol/L, and 9.01-19.05 pmol/L, respectively. Peripheral deiodinase activity and thyroid's secretory capacity were calculated according to SPINA-GD (G D ) and SPINA-GT (G T ), respectively (Dietrich et al. 2016).
Furthermore, considering urinary iodine is an important covariate influencing thyroid function, we also determined this parameter. Iodine concentration in urine was measured by inductively coupled plasma mass spectrometry (ICP-MS). Using an alkaline diluter which comprised 0.25% tetramethylammonium hydroxide (25%, Macklin) and 0.02% triton (for Molecular Biology, Sigma), each 150 μL of urine sample was diluted to 2.25 mL. In addition, certified reference materials, such as SeronormTM Trace Elements Urine L-1 RUO (No. 210613, Sero, Iodine concentration in Norway) and SeronormTM Trace Elements Urine L-2 RUO (No. 210713, Sero, Norway) were used as quality controls. The recovery rate of internal standard was 80.00-102.23%. In addition, the metal concentration lowering than limit of detection (LOD) was assigned with LOD value divided by √2.

Metal analysis
We determined 22-metal levels in blood cells in accordance with previous protocols (He et al. 2022). Standards, such as Seronorm™ Trace Elements Whole Blood RUO (Nos. 210105, 210205, and 210305, ALS Scandinavia, Sweden) and standard reference material (SRM1640a, Trace Elements in Natural Water from Natural Institute of Standard Technology, Gaithersburg, MD, USA) were used as quality controls. Among the 22 blood metals, the LOD values ranged from 0.001 to 2.020 μg/L. The metal concentration lowering than the corresponding LOD was imputed by the LOD value divided by √2.

Covariates
Seniority refers to the length of time participants have worked for ferro-manganese refinery. Current drinkers were those who have consumed alcohol at least once in a week over 3 months, and non-drinkers who have never consumed alcohol or have formerly consumed alcohol (less than 5 mL). Current smokers were those who have smoked at least one cigarette in a day over 3 months and current smokers (no), as those who have never smoked or have quit smoking over 3 months).

Statistical analysis
For Al, Ni, and Sb, blood cell samples with levels lowering than the corresponding LODs were imputed with LOD/√2. The other metals were detected in all samples. For reducing the skewness of distribution in metal concentrations in blood cells and thyroid hormones in serum, concentrations in metals were log10-transformed, and THs were ln-transformed, representing as median (25th and 75th percentiles) for descriptive purposes. Spearman's correlation analysis was performed for estimating the relationship across metals. Frequency (percent) was used to present demographic information. Age (continuous), gender (men/women), body mass index (BMI) (continuous), seniority (continuous), current smoker (yes/no), current drinker (yes/no), tea consumption (yes/no), work shift (yes/no), and sleep quality (good/ ordinary/bad) were treated as covariates, which were entered simultaneously into the model in the following analyses unless otherwise stated.
In order to select metals in blood cells independently related to thyroid function parameters, we used least absolute shrinkage and selection operator (LASSO) model. In the LASSO model, variables will be penalized more severely accompanied with the increased penalty parameter lambda (λ). As a result, a sparse model remains after more coefficients are closing to zero. All covariates were included in the LASSO model without being penalized. We predicated an optimal model using tenfold cross-validation, according to λ where mean square error (MSE) was the minimum.
Furthermore, we conducted the general liner model (GLM) to assess the associations of metals selected by the LASSO model with thyroid function parameters. Considering iodine status may affect the thyroid function parameters, we divided the participants into two groups, according to WHO as requirements are insufficient and adequate or above requirements. Furthermore, we explored the associations between selected metals and thyroid function parameters stratified by iodine status and introduced interaction terms, thereby assessing the potential modification effect.
In recent years, Bayesian kernel machine regression (BKMR) has been increasingly utilized for estimating the health effects related to environmental mixture exposures (Bobb et al. 2018). In this framework, Bayesian variables are selected non-parametrically, which can also identify the importance of components in mixtures as well as evaluate the non-linear dose-response relationship. First, the importance of specific metal on thyroid function parameters was ranked by posterior inclusion probabilities (PIP) function ranging from 0 to 1. The exposure-response shape of specific metal with each outcome was drawn with the other metals and was kept at the 50th percentile. The change in estimated effect on each thyroid function parameter owing to one metal changing from its 25th to 75th percentiles when the other were kept at their 25th, 50th, or 75th percentile, respectively. In order to estimate the cumulative effect of metal mixtures at specific quantiles, we evaluated the effects of the metal mixtures differing from their 50th percentiles. Also, we estimated the interaction between metals on thyroid function parameters.
We conducted all statistical analyses using R (Version 4.1.3) and SPSS (version 21.0, IBM) with two-sided P values > 0.05 considered significant. Table 1 summarizes the demographic characteristics and thyroid function parameters of 328 participants. Briefly, the participants were of a median age of 46.9 years and seniority of 23.0 years, respectively. The participants were predominantly to have a normal BMI. The proportions of current smoker and current drinker were 50.6 and 51.5%, respectively. The majority of the participants (83.2%) tended to have tea consumption. Moreover, 80.5% of the participants involved in work shift and 50.4% of the participants' sleep quality were ordinary or bad. Median levels of TSH, TT3, FT3, TT4, and FT4 were 1.6 mIU/L, 1.6 nmol/L, 4.8 pmol/L, 93.8 nmol/L, and 12.4 pmol/L, respectively.

Blood cell metal concentrations and their correlations
The detection frequency, quartiles, and LODs of metals in blood cells are summarized in Table S1 Briefly, the lowest LOD was 0.001 μg/L for cobalt or cadmium, while the highest was 5.329 μg/L for calcium, and the concentrations of metals were all higher than LODs, except Al, Tin, and antimony whose proportion of concentrations is lower than the corresponding LOD were 3.1, 7.0, and 18.3%, respectively. Additionally, we conducted Spearman's rank-order correlation analysis to discover correlations between metals in Fig. S1.

LASSO regression model
To select metals, the LASSO model was utilized, putting non-penalization on the covariates. We choose the optimal model where the MSE is the minimum based on the tenfold cross-validation function in Fig. 1. We included all metals selected by LASSO into GLM model for exploring the relationships with thyroid function parameters. Cu was positively associated with TSH (β = 2.016). Ni (β = − 0.204, P = 0.058) and Se (β = − 0.919, P = 0.079) were negatively associated with the TSH level, although with borderline significance. We found a positive association between Fe and FT4 (β = 0.403) while a negative association between Table 1 The characteristics and levels of serum thyroid hormones among the 328 participants MEWHC manganese-exposed workers healthy cohort, BMI body mass index, TSH thyroid stimulating hormone, TT3 total triiodothyronine, FT3 free triiodothyronine, TT4 total tetraiodothyronine, FT4 free tetraiodothyronine, GD sum activity of peripheral deiodinases, GT thyroid's secretory capacity Cu and FT4 (β = − 0.226). We found positive associations between Ti (β = 0.142), Ni (β = 0.057), and G D , respectively. Moreover, we observed negative associations between Se (β = − 0.332), Sr (β = − 0.243) and G D , respectively (Table 2).

Effect modification by iodine status
A modification effect of the association was observed in Table 3. Stratified by iodine status, the positive relationship of Sr with G D persisted only in participants whose iodine status were insufficient (β = 0.047), while among individuals with adequate or above requirements iodine status (β = − 0.050), the association was in the opposite direction (P interaction = 0.026). However, we did not observe any modification effect of iodine status on the relationship of other metals with thyroid function parameter (all P interaction > 0.05).

Bayesian kernel machine regression analysis
The BKMR model further provided measurement for the overall effects of the metal mixtures on thyroid function parameters and interactions between metals. Figure 2d   remaining metals were fixed at their 50th percentiles. And other metals showed linear relationships with thyroid function parameters which were in line with the results in the GLM model in (Fig. 2a, Figs. S2a and S3 a. Among the expected PIPs, Mg ranked the highest for FT4 (PIP = 0.358), and Cu ranked the highest for TSH (PIP = 0.628).
In addition, we explored the interactions between metals by plotting specific metal effect differing when the other one holding at the 10th, 50th, and 90th percentiles. We observed a synergistic effect between As and Cu; evidence showed that the slope of As and TSH was steeper when Cu was at a high quintile. Interestingly, we found signs of antagonistic effects between Ni and Cu or Se; evidence showed that the slope of Ni and TSH was smoother when Cu or Se was at a high quintile. No indication for an interaction is detected for Se and Cu in Fig. 2c.

Discussion
In this study, we identified metals associated with thyroid function parameters among the participants from MEHWC. LASSO and BKMR models showed consistent findings and indicated positive associations between Cu and TSH, Fe and FT4, Ti and G D , Ni and G D , while negative associations between Cu and FT4 and Se and G D . Interestingly, we observed an inverted-U shape relationship between Mg and FT4. Furthermore, we found a synergistic effect between As and Cu on the TSH level, while antagonistic effects between Ni and Cu as well as Ni and Se on the TSH level. And we observed the modified effect of iodine status on association between Sr and G D .
In comparison to metal concentrations in previous studies, Mn concentration here (median, 29.30 μg/L) was higher than Table 2 Differences in thyroid function parameter for tenfold increase in metals based on LASSO model a Log-transformed metals selected by LASSO-penalized regression models were included in multivariate linear regression models simultaneously (multiple-metal models), with adjustment for gender (men/ women), age (continuous), seniority (continuous), BMI (continuous), current smoker (yes/no), current drinker (yes/no), tea consumption (yes/no), work shift (yes/no), and sleep quality (good, ordinary, and bad) TSH thyroid stimulating hormone, FT4 free tetraiodothyronine, GD sum activity of peripheral deiodinases  (Lin et al. 2021) and Germany (Heitland and Koster 2021) (n = 102, mean, 17.4 μg/L); but it was lower than that in other places among smelters or welders (mean or median, 79 ~ 1300 μg/L) (Jiang et al. 2007;Long et al. 2014;Ou et al. 2018). Interestingly, Mg (median, 56,746.65 μg/L), Fe (median, 1,087,676.66 μg/L), Cu (median, 731.8 μg/L), Cd (median, 4.40 μg/L), Ba (median, 5.91 μg/L), and Pb (median, 61.23 μg/L) were all higher than the general population from human biomonitoring project in Germany (Heitland and Koster 2021). In accordance with airborne metals, metal elements were also detectable in blood cells, including Mg, Mn, Fe, Cu, Cd, and Pb. External exhibits provide an interpretation of metal resources. The higher level of Ba might be the interaction between metals or from diet.
In our research, Fe was positively associated with FT4. Fe is the most abundant transition metal in the human body, and the role of Fe in the thyroid is less defined Biologically, hemecontaining thyroperoxidases (TPO) plays an important role in thyroid hormone production. TPO, whose activity is dependent on Fe, catalyzed iodide incorporation into tyrosine residues of thyroglobulin and covalent bridging of the residues (Ye et al. 2023). On one hand, Fe deficiency decreased plasma concentrations of T4, which interestingly was reversible with iron repletion (Smith et al. 1993). On the other hand, in this study, the level of Fe was relatively higher than the normal population, owing to Fe being one of the main components of the product. To some extent, it can provide a basis for the positive correlation between Fe and FT4. The mechanism of the positive association between Fe and FT4 needs further exploration. Fig. 2 Joint effect of metal mixture exposures on TSH among the participants. a Dose-response relationship between specific metal and TSH when the remaining metals kept at their 50th percentile. b Difference in TSH for increase (25th to 75th) in specific metal when the remaining metals kept at their 25th, 50th and 75th percentile, respec-tively. c Bivariate exposure-response functions for: exposure 1 when exposure 2 was fixed at either the 10th, 50th, or 90th percentile, and other elements are fixed at their 50th percentile. d Overall effect of the metal mixtures on TSH 1 3 More than 300 enzymes are dependent on Mg, including those involved in thyroid hormone synthesis (Reinehr et al. 2008). Mg-rich food, for example, leafy greens, nuts, whole grains, and seeds, are the main source for people to intake Mg. It has suggested links between Mg imbalance and thyroid diseases (Jones et al. 1966;Wang et al. 2018). Interestingly, we observed an inverted U-shaped relationship between Mg and FT4, consistent with previous publications. Previous studies have found increased Mg was associated with hypothyroidism, and decreased Mg was associated with hyperthyroidism (Jones et al. 1966), which provides an explanation or possibility for the inverted U-shaped relationship in this study. However, a meta-analysis showed no significant difference in the Mg level between hypothyroid patients and healthy controls (Talebi et al. 2020).
Strontium (Sr), an element known as calcium-sensing receptor (CaSR) agonists, and Sr 2+ ion seem to be able to replace Ca 2+ ion. Sr-stimulated calcitonin secretion was stronger when CaSR signaling was biased toward extracellular signal-regulated kinase 1 and 2 (ERK1/2) signaling (Thomsen et al. 2012). In the secretion of calcitonin, thyroid hormones T3, T4, and tyrosine combine to stimulate the process of regulating body calcium balance. And yet it is unclear whether Sr in turn affects thyroid function in producing T3 and T4. While we observed a negative association between Sr and FT4, further investigation of this association and its possible mechanism is needed.
Deiodinase is also implicated in disrupted thyroid function due to various environmental contaminants (Boas et al. 2009). Homeostasis of thyroid hormone secretion and synthesis are maintained by sensitive feedback mechanisms. As a result of TSH stimulation, the thyroid produces thyroid hormone T4, which is subsequently deiodinated into thyroid hormone T3, as a more biologically active form. Circulating T3 and T4 control the rate of TSH release. When circulating T3, T4, and TSH levels are reduced, the pituitary will release TSH, which results in an increase in thyroid synthesis of T3 and T4 (Stoker et al. 2004). A mechanism for inhibiting THs might be by inhibiting T4's deiodination or by competing with the hormone's binding to its carrier proteins (Rana 2014).
In this study, we performed two kinds of multi-pollutant analysis approaches (LASSO and BKMR models) to overcome the shortcomings in traditional strategies; our employment of BKMR allowed us to account for non-linearity and the potential for high-dimensional interactions between metals in association with thyroid function parameters. Furthermore, we assessed the urinary iodine concentration among the study participants. Iodine is vital in the synthesis of thyroid hormones. Available information on iodine status allowed us to explore whether iodine status can modify the associations of metal mixture exposures with thyroid hormones, thereby underlying the potential mechanism. However, there were several limitations in our study. First, the reverse causation is an inherent limitation in this cross-sectional study. Second, there was a relatively small sample size, which limited the ability to interpret results. Third, residual confounding could not be dismissed, such as dietary habit.

Conclusions
The current study suggests that metal mixture exposures can alter thyroid homeostasis among the occupational populations, possibly through an alteration of deiodinase activity. Validation of these associations and elucidation of underlying mechanisms require further researches in the future.