Prenatal exposure to silver is associated with an elevated risk for neural tube defects: a case–control study

Exposure to copper, silver, and titanium has been reported to be associated with a variety of adverse effects on humans, but it is little focused on the fetus. We investigated the associations between prenatal exposure to the three metals (copper, silver, and titanium) and risk for fetal neural tube defects (NTDs). Placental samples from 408 women with pregnancies affected by NTDs and 593 women with normal pregnancies were collected from 2003 to 2016 in Pingding, Xiyang, Shouyang, Taigu, and Zezhou counties of China. Multilevel mixed-effects logistic regression and Bayesian kernel machine regression (BKMR) were used to evaluate the single and joint effects of the metals on NTDs. Silver was associated with an increased risk for NTDs in a dose–response fashion in single-metal logistic regression, with adjusted odds ratios (95% confidence intervals) of 1.78 (1.04–3.06) and 1.92 (1.11–3.32) in the second and third tertiles, respectively, compared to the lowest tertile. BKMR revealed toxic effects of silver on NTDs and the association appeared to be linear. No interaction of silver with any of the other two metals was observed. Besides, silver concentration was positively correlated with maternal certain dietary intakes. Placental high silver concentrations are associated with an elevated risk for NTDs. Maternal diet may be a source of silver exposure.


Introduction
Neural tube defects (NTDs), including anencephaly and spina bifida, are severe and debilitating congenital malformations of the central nervous system that arise from incomplete closure of the neural tube in early embryonic development (Wallingford et al. 2013). The prevalence of NTDs differs greatly between and within countries, varying from 0.5 to more than 10 per 1000 pregnancies (Greene and Copp 2014;Ren 2015). In China, the birth prevalence of NTDs was 1.4 per 1000 births during 2006 and 2008 with northern region having higher prevalence than southern region (Li et al. 2013). Although intensive researches have been conducted, the etiology of NTDs is not yet fully understood. A variety of causes or risk factors have been proposed, involving genetic or environmental origin or a complex combination of their interactions. However, little is known about the role of environmental factors except for folate and some medications. Identification of environmental risk factors is important for NTD prevention because these factors may be modifiable.
Copper, silver, and titanium are three transition trace elements, the former involved in various biological processes and toxic in excess, while the latter two have no known essential role in biology (Bulcke et al. 2017;Lansdown 2007;Zierden and Valentine 2016). Moreover, the three elements are present in cosmetics, which are commonly used by women all over the world (Borowska and Brzóska 2015;Fage et al. 2016;Panzarini et al. 2018). It was reported that exposure to excessive copper, silver, and titanium has a very wide range of adverse effects on health, such as Alzheimer disease, Parkinson disease, yellow nail syndrome, argyria, leukopenia, and renal dysfunction (Hadrup et al. 2018;Kim et al. 2019;Pohanka 2019). Evidence from animal models, including fish, mice, rats, and zebrafish, indicated that exposure to elevated copper, silver, and titanium could induce growth reduction, morphological malformations in embryos, and neonatal deaths (Ahmad et al. 2022;Cao et al. 2010;Malik et al. 2000;Yoo et al. 2016). Several studies also suggested that excessive copper concentrations in maternal serum or plasma are risk factors for NTDs (Cengiz et al. 2004;Demir et al. 2019;Zeyrek et al. 2009), whereas no such association was found in another maternal serum-based study (Tian et al. 2021). A case-control study from Mexico suggested higher silver in maternal hair may be participating in the development of NTDs (Ramírez-Altamirano Mde et al. 2012); in contrast, another study in China found similar concentrations in hair samples of 191 newborns with NTDs and 261 healthy infants, and it also reported that higher levels of titanium in maternal scalp hair were associated with risk for NTDs (Li et al. 2016).
Therefore, the relationships between copper, silver, and titanium and NTD risk in offspring are still needed to elucidate. Furthermore, blood has a short turnover; as such, blood-based markers may only reflect short-term exposure (Hughes 2006;Laohaudomchok et al. 2011). Maternal hair is prone to external environmental contamination (Karagas et al. 2000). Placenta, a temporary organ during pregnancy, is not only easy to obtain, but also can furnish valuable information on the exposure of fetuses in utero (Baergen 2007;Esteban-Vasallo et al. 2012). Although this tissue acts as a barrier to harmful substances for developing fetuses, some chemicals, including copper, silver, and titanium, can across and accumulate to certain extent in placenta (Iyengar and Rapp 2001;Li et al. 2019). Thus, placental concentrations of the three elements can be considered suitable prenatal exposure biomarkers.
In the present study, using placental concentrations of copper, silver, or titanium, reliable biomarkers of reflecting intrauterine exposure for fetuses, we intended to examine the association between maternal exposure to these elements during pregnancy and risk for NTDs in offspring with a case-control design. Multilevel mixed-effects logistic regression and Bayesian kernel machine regression (BKMR) were used to evaluate the single and joint effects of the metals on NTDs, and the results of the two methods were interpreted jointly forward. Moreover, Spearman correlation was used to investigate whether placental element concentrations related with maternal diet frequencies.

Study subjects
Study participants were women enrolled in an ongoing case-control study established in 2003 in five rural counties (Pingding, Xiyang, Shouyang, Taigu, and Zezhou) of Shanxi Province in northern China, as previously described Ren et al. 2011). Briefly, the main purpose of the study is to identify environmental factors for major external structural birth defects, i.e., NTDs, orofacial clefts, and congenital hydrocephalus. Eligible cases are pregnant women whose current pregnancies are affected by the above-mentioned malformations of any gestational age in live births, stillbirths, or elective pregnancy terminations. Following the recruitment of a case, a woman with healthy fetuses, of similar gestational age (± 4 weeks) as the cases, and reside in Shanxi Province during this pregnancy was enrolled as a control. In the present study, we used data collected from January 2003 to December 2016. To obtain a maximum sample size, we concluded controls recruited for other defects as well as NTDs, resulting in a sample consisting of 408 NTD cases and 593 controls. The distribution of samples was pictured on Figure S1. This study was approved by the Ethics Committee of Peking University (Beijing, China). All study participants provided informed consent.
At enrollment, we obtained information on sociodemographic and lifestyle factors with a structural questionnaire administrated by local healthcare staff through a face-to-face interview. The following covariates were assessed: maternal age at pregnancy (years), prepregnancy body mass index (kg/m 2 ), ethnicity (Han/non-Han), parity (one/more), education levels (primary school or lower, junior high school, or high school or higher), occupation (farmer/non-farmer), self-reported history of pregnancy affected by birth defects (yes/no), have had a fever (≥ 38.5 °C) or influenza (yes/no), passive smoking (ever/never), folic acid supplementation (yes/no), frequency of selected food groups during the periconceptional period, and gestational age at placenta collection (weeks). Maternal dietary habits during the periconceptional period were also collected through the questionnaire, including the frequency of eating meat or fish, egg or milk, fresh vegetables or fruits, and bean or its products, as well as drinking tea.

Placenta collection and element quantification
After delivery or elective termination, the placenta was immediately collected and stored at − 20 °C with labeled polyethylene plastic bag. The samples were shipped with dry ice to our laboratory for storage at − 20 °C prior to processing. The details of sample preparation have been described elsewhere (Pi et al. 2019). Shortly after thawing at 4 °C, about 6 g specimens on the fetal side were collected within 3 cm around the insertion point of the umbilical cord, rinsed with deionized water, and freeze-dried 24 h (ALPHA2-4 LD plus, Christ, Germany). About 0.1 g lyophilized placental tissue was successively mixed with nitric acid and deionized water, digested under a high-pressure microwave system (Ultra WAVE, Milestone, Italy), diluted with deionized water, and finally added with internal standard (rhodium and indium), which can be well corrected for matrix effects of the sample, for determination of concentrations of the elements with inductively coupled plasma-mass spectrometers (ICP-MS, 7700x, Agilent, USA). The operating conditions of ICP-MS are listed in Table S1. Concentrations of elements were expressed with nanogram or microgram per gram dry weight. The sensitivity and stability of ICP-MS were calibrated by analyzing the internal reference agent (rhenium, GSB 04-1745(rhenium, GSB 04- -2004. The accuracy and precision of ICP-MS was checked by analyzing the standard reference material (SRM) with known concentrations of three elements (SRM: pig liver for copper, GBW10051; human hair for silver and titanium, GBW09101b). Quality control and assurance of placental element concentration assessment were achieved by using ultra-pure grade chemicals in the whole experiment, 1 blank solution control and 1 replicated measurement in every 30 samples, and masking the operators of the group status of the samples (case or control). The correlation coefficients of the standard curves for copper, silver, and titanium were all over 99.9%. The detection limits of copper, silver, and titanium were 3.2 ng/g, 0.32 ng/g, and 1.07 ng/g, respectively. The detection rates for copper and titanium were 100%, while silver was 79%. The samples with silver levels below the detection limit were imputed as half of the detection limit.

Statistical analysis
Comparisons of baseline characteristics, which were expressed as number (percentage), between NTD and control groups were made with the Chi-square test. Placental concentrations of the elements between NTDs and controls were compared using the Mann-Whitney U test for medians and Student t test for geometric means, respectively. The heterogeneity of element concentrations between five geographical regions was checked by multivariate analysis of variance.
Firstly, we used multilevel mixed-effects logistic regression to evaluate the association between each element exposure and risk for NTDs. Tertiles of concentrations of elements for all subjects in each geographical region were introduced in the model as categorical variables to compare the second and highest tertiles (exposure) with the lowest tertile (non-exposure). The effect sizes of associations were indicated with odds ratios (ORs) and their 95% confidence intervals (95% CIs). Secondly, we employed BKMR to examine the following: (1) the effect of each element on NTDs while fixing the remaining two elements to 25th, 50th, or 75th percentile; (2) the visualizable dose-response relationship between element exposure and NTD risk without coarsening the continuous concentrations; (3) the interactive effects between three elements, which cannot be properly handled with linear regression ; and (4) the posterior inclusion probability (PIP), which represents the contribution of a particular metal in the mixture and ranges from 0 to 1. A total of 10,000 iterations were applied in BKMR. Both regression models adjusted for maternal occupation, education, gestational age at placenta collection, history of pregnancy affected by birth defects, have had a fever or influenza, passive smoking, and folic acid supplementation during the periconceptional period due to their uneven distributions between the case and the control groups.
Correlations between placental element concentrations, including numerical concentration and tertile concentrations, and maternal diet frequencies were assessed by Spearman's coefficient. To evaluate possible confounding by gestational age, which was different between the case and control groups, on the associations between placental element concentrations and risk for NTDs, we performed subgroup analyses with a subset of cases and controls matched by gestational age. To assess the robustness of the results, we further repeated our analyses by including only NTD cases without other system malformations.
The BKMR was implemented using R (version 4.0.3; R Development Core Team), and the remaining statistical analyses were conducted using the Stata 15.0 software (Stata Corporation, College Station, Texas, USA). A two-sided P-value of < 0.05 was considered statistically significant.

Characteristics of the study participants
Characteristics of participating NTD cases and controls are presented in Table 1. Virtually (99.5%) of all participants were of Han ethnicity. Generally, study subjects were young, with a mean age of 26.9 years in cases and 26.6 years in the controls. Compared to controls, cases were less likely to complete high school education and to report folic acid supplementation during the periconceptional period, more likely to work as a farmer, to report a history of pregnancy affected by birth defects, to have a fever or influenza, and to report passive smoking during the periconceptional period. Over half (56.4%) of the cases had a gestational age at placenta collection of fewer than 28 weeks, compared to only 2.4% in controls, reflecting the impact of elective termination following prenatal diagnosis in the case group. Most of the women had a normal prepregnancy body mass index (approximately 62%), with no difference being observed between the two groups. Approximately half of the participants (49%) were primigravida, and there was no difference in parity between the cases and controls.

Element concentrations in cases and controls
Descriptive data of element concentrations in the placental tissue of the study subjects are presented in Table 2. Median concentrations of copper and silver were significantly higher in NTD cases than in controls, while the median concentration of titanium did not show a significant difference. When presented in geometric parameters, similar profiles were exhibited for the two groups. Anencephaly and spinal bifida, the two major subtypes of NTDs, showed a similar pattern for the three elements with total NTDs (Table S2).
Variations in element concentrations were observed among the five counties where the study subjects were recruited (Table S3). The variations by region were then considered with multilevel mixed-effects logistic regression in the following analyses.

Element concentrations and the risk for NTDs
The associations between tertile concentrations of each element and risk for NTDs are displayed in Table 3. In singleelement regression models, the highest tertile of copper was associated with a 2.28-fold (1.64-3.17) higher risk for NTDs in univariate analyses. However, this association disappeared when confounding factors were adjusted. For silver, the risk for NTDs increased by 2.02-fold (95% CI 1.45-2.82) and 2.19-fold (95% CI 1.55-3.11) in the second and the highest tertiles compared to the lowest tertile. These associations remained after adjustment for potential confounders, and a dose-response relationship remained with an adjusted OR of 1.78 (95% CI 1.04-3.06) and 1.92 (95% CI 1.11-3.32) for the second and highest tertiles, respectively. No association between titanium levels and NTD risk was observed in either unadjusted or adjusted models. In multiple-element regression models, however, the association between silver concentrations and NTD risk turned to non-significant after adjustment for potential confounders. Similar patterns of associations were presented for anencephaly and spina bifida (Table S4). In BKMR, the PIP value of copper, silver, and titanium was 0.0212, 0.7328, and 0.0178, respectively. Silver was the only element that has shown risk effects on NTDs, as indicated by its three point estimates, which are greater than the null value of zero and their lower boundaries of credible intervals exclude the null. Moreover, the three point estimates of silver are almost identical, suggesting that silver's effect is independent of the other two elements (Fig. 1A). No associations between concentrations of copper and titanium and NTD risks were observed (Fig. 1A). As Fig. 1B indicates, NTD risks increased linearly with silver concentrations, although the curve flatted at the highest silver concentrations. No clear evidence of association was shown for copper and titanium concentrations and risk for NTDs (Fig. 1B). No interaction between elements was observed because the slopes of a specific element were similar when the other element was set at the 25th, 50th, or 75th percentile while the third one being held at its median (Fig. 1C), and none of the interactive effects was significant as the credible intervals for copper, silver, and titanium encompass the zero null value (Fig. 1D). When elements were treated as a mixed exposure mixture, NTD risk increased almost linearly across the whole range of mixture concentrations from the 25th percentile through the 75th percentile ( Figure S2).

Element concentrations and dietary intake
Correlations between frequencies of maternal food groups during the periconceptional period and placental element concentrations are shown in Table 4 and Table S5. No correlation was found between maternal diet and copper Table 3 Associations between placental tertile concentrations of elements (dry weight * ) and risk for neural tube defects in northern China Cu copper; Ag silver; Ti titanium; OR odds ratio; CI confidence interval * Standard reference materials: pig liver for copper, GBW10051; human hair for silver and titanium, GBW09101b # P < 0.05 a Calculated by multilevel mixed-effects logistic regression, geographical region as a random effect b Adjusted for maternal occupation, education, gestational age at sample collection, history of pregnancy affected by birth defects, have had a fever or influenza, passive smoking, and folic acid supplementation during the periconceptional period levels, except for the consumption of meat or fish. Higher consumption frequencies of meat or fish, egg or milk, fresh vegetables and fruits, beans or its products, and tea drinking were positively correlated with concentrations of silver. In addition, increased intake of egg or milk and fresh vegetables and fruits was positively correlated with titanium concentrations. Fig. 1 Effect of elements on neural tube defect (NTD) risk (expressed in β probit ) estimated by Bayesian kernel machine regression. Models were adjusted for maternal occupation, education, gestational age at placenta collection, history of pregnancy affected by birth defects, have had a fever or influenza, passive smoking, and folic acid supplementation during the periconceptional period. A Single effect of element on NTD risk (estimates and 95% credible intervals) by comparing the NTD risk when one element changes from its 25th percentile to 75th percentile while setting the other two at 25th, 50th, or 75th percentile. B Univariate exposure-response function for element and change in NTD risk while the other two elements are held at their median levels. C Bivariate exposure-response functions for each element: the top left panel shows for silver (Ag) when copper (Cu) is fixed at the 25th, 50th, or 75th percentile, and titanium (Ti) is fixed at its median level. The others were similar to the top left panel. D Interactive effect of elements, defined as the change in the single-element association when the remaining two elements are fixed at their 25th percentile as compared to when they are fixed at their 75th percentile Table 4 Correlations between placental element concentrations (dry weight * ) and frequencies of maternal food consumption in northern China NTDs neural tube defects * Standard reference materials: pig liver for copper, GBW10051; human hair for silver and titanium, GBW09101b a Frequency of dietary intake was classified into three levels, i.e., < 1, 1 to 6, and > 6 times per week b Concentrations of elements were treated as numerical variables Dietary intake a Copper (μg/g) b Silver (ng/g) b Titanium (μg/g) b

Sensitivity analyses
Ninety-two pairs were obtained when cases and controls were matched by gestational age. Similar patterns of placental concentrations of silver in cases and controls were present with the overall analyses (Table S6). Although the dose-response relationship for silver and NTD risk was not significant, largely because of the reduced sample size, the direction of the association was the same as the total sample (Table S7), suggesting that the observed associations were unlikely to be resulted by confounding of gestation. The results did not change meaningfully when NTDs complicated with other malformations were excluded (Table S8 and Figure S3).

Discussion
Using concentrations of copper, silver, and titanium in placental tissue as prenatal exposure markers, this study revealed that exposure to silver was associated with elevated risk for NTDs in a Chinese population with both the traditional logistic regression and the state-of-the-art BKMR. Silver concentrations were positively correlated with frequencies of several food groups. No interaction between silver and any of the other two elements was observed.

Copper, silver, and titanium and NTD risk
Several previous studies have examined the association between maternal exposure to the three elements during pregnancy and NTDs, and our findings are consistent with some but not all. For example, our finding is comparable to a study from Turkey that found no difference in levels of copper in the amniotic fluid between groups of pregnant women complicated with NTDs and healthy fetuses (Ovayolu et al. 2019), analogous to other two studies from Australia (McMichael et al. 1994) and China (Yan et al. 2017), while contrary to other studies (Cengiz et al. 2004;Demir et al. 2019;Zeyrek et al. 2009). Silver concentrations in scalp hair from mothers of newborns with NTDs were significantly higher than in healthy controls in an Iranian study (Ramírez-Altamirano Mde et al. 2012), which is in line with the present study. Maternal hair concentration of silver was not associated with risk for NTDs in a study conducted in the same province as well as another province in northern China (Li et al. 2016). For titanium, however, the previous case-control study using maternal scalp hair suggested that high levels of titanium are possibly correlated with increased risk for NTDs (Li et al. 2016). We further examined the correlation of the three elements' concentrations in two biological specimens of maternal hair and placenta in a subset of samples, but no significant relationship was found (data not shown). This discrepancy may be resulted from the complex mechanism of elements' absorption, metabolism, or excretion.
The results of BKMR are in line with the results of individual element regression. However, contrary to the multi-element linear model, the relationship between silver and NTDs presented a significantly positive and linear dose-response pattern in BKMR. Generally, conventional statistical approaches, such as logistic regression, include a set of exposures of interest into one model; however, such an approach may contribute to the distortion of results if correlations exist in the targeted exposures (Marill 2004). In this study, there is a correlation between placental silver and titanium ( Figure S4, P < 0.001), with an r value of 0.24, which may explain why silver concentrations are associated with risk for NTDs in BKMR while not in logistic regression.
No interaction of three elements was observed, implying no synergistic effect or antagonistic effect of each two elements (i.e., copper × silver). Therefore, the use of BKMR in combination with conventional linear regression can improve the robustness of the results and get a deeper insight into the complex relations among multiple exposures regarding their effect on NTDs.
The biological mechanism of silver-related embryonic toxicity is not fully understood. The possible mechanisms include aberrant oxidative stress, apoptosis, and necrosis induced by silver, which may be potential factors in the pathogenesis of NTDs (Greene and Copp 2014). Studies have shown that exposure to silver ions can directly increase the production of reactive oxygen species (ROS), including superoxide radicals, a decrease in reduced glutathione and increased susceptibility to hydrogen peroxide-induced cell death and disrupts the balance between ROS and antioxidants (Cortese-Krott et al. 2009;Foldbjerg et al. 2009). Moreover, exposure to silver induces apoptosis and necrosis by generating high levels of ROS (Arora et al. 2008;Foldbjerg et al. 2009). Further studies are warranted to fill in the knowledge gap in the NTD developmental toxicity of silver.

Copper, silver, or titanium and dietary intake
Dietary intake is the primary source of human copper exposure, with approximately 75% from solid food and 25% from drinking water (Brewer 2015). In this study, consumption frequencies of meat or fish were correlated with copper concentrations in placental tissue. As a naturally occurring element, silver has a wide variety of applications, allowing exposure through various routes of entry to the body, such as ingestion and skin contact. The WHO has reported that the daily intakes of silver are about 7 μg/person (World Health Organization 2008). A study from the UK estimated 27 μg/day for human ingestion of silver from diets (Dolara 2014); high concentrations of silver in edible mussels (0.02-4.4 μg/g), oysters (1.1-65 μg/g), and mushrooms (3.1 μg/g) were also reported from France (Chiffoleau et al. 2005) and Poland (Falandysz et al. 2008). Additionally, the workplace also is a major silver exposure source for humans. However, participants in the current study reported no occupational exposure, and most of the study subjects were recruited from rural areas. And our results showed that concentrations of silver in placental tissue are significantly correlated with frequencies of all food groups included in the questionnaire. Thus, daily dietary silver intake may be one of the sources of exposure in this population. As a common food additive, titanium ubiquitously exists in foods in the form of titanium oxide, and the content in some dairy products with white colors, such as milk, ranges from 0.10 to 0.26 μg/mL (Yang and Westerhoff 2014). This may partly explain why placental titanium concentrations were positively correlated with the frequency of the consumption of egg or milk in our population.

Element concentrations in different populations
Reported exposure levels (dry weight) vary among populations and biological samples. For instance, compared to the data from women in Poland, women in our population had a similar concentration of copper (4.04 μg/g vs. 5.64 μg/g) but a lower placental concentration of silver (3.0 ng/g vs. 0.84 ng/g) (Kot et al. 2019). Median placental copper concentrations were also similar to those found in previous studies conducted in Bangladeshi (5.3 μg/g) (Kippler et al. 2010) and Japan (3.91 μg/g) (Sakamoto et al. 2013). The mean value for copper in placental tissue in this study (4.23 μg/g) was remarkably lower than the concentration reported in a study from India (70.0 μg/g) (Reddy et al. 2014), similar to those from women in Germany (4.4 μg/g) (Schramel et al. 1988), but higher than those reported from healthy pregnant Spanish women (0.97 μg/g) (Cerrillos et al. 2019). However, due to differences in element assessment methodology, it may not be appropriate to make a direct comparison of specific element concentrations across studies.

Strengths and limitations
Strengths of the present study include homogeneous ethnicity of study participants, relatively large sample size, and considerable information on characteristics to adjust as potential confounders in statistical models. Besides, element concentrations in placental tissue were used as an indicator of environmental exposure in utero, which is believed to be able to reflect longer-term measures compared to maternal or cord blood-based markers (Baergen 2007;Iyengar and Rapp 2001). Finally, BKMR was used to explore the effect of each element while taking others into consideration and interactions between any two of the elements.
Inevitably, there also exist some limitations to our analyses. First, it would be ideal to be able to measure the levels of these elements during the 5th to 7th weeks of gestation, the developmental window of NTDs (Botto et al. 1999). However, it is infeasible due to ethical or technical reasons. Second, most cases but fewer controls had gestational age at sample collection less than 28 weeks, which may confound the examined associations. However, sensitivity analyses showed that our results could not be completely explained by the confounding of gestation. Third, we were unable to isolate a causal relationship between placental element levels and risk for NTDs, as this is a case-control study. Fourth, other co-exposure pollutants including heavy metals were not tested in this study, which may also have effect on the development of NTDs. Finally, other potential exposure sources, such as cosmetics and personal care product use, were not investigated.

Conclusions
Prenatal exposure to high levels of silver in placental tissue was associated with an increased risk for NTDs in offspring in a dose-response manner, and dietary intake may be one of the major sources of maternal silver exposure. Since maternal exposure to heavy metals, including silver, is also common in other populations around the globe, further studies are needed to replicate the findings of this study, and animal studies are warranted to elucidate biological mechanisms underlying the observed statistical association.