Uncovering neurodevelopmental windows of susceptibility to manganese exposure using dentine microspatial analyses

Abstract

Background

Associations between manganese (Mn) and neurodevelopment may depend on dose and exposure timing, but most studies cannot measure exposure variability over time well.

Objective

We apply temporally informative tooth-matrix biomarkers to uncover windows of susceptibility in early life when Mn is associated with visual motor ability in childhood. We also explore effect modification by lead (Pb) and child sex.

Methods

Participants were drawn from the ELEMENT (Early Life Exposures in MExico and NeuroToxicology) longitudinal birth cohort studies. We reconstructed dose and timing of prenatal and early postnatal Mn and Pb exposures for 138 children by analyzing deciduous teeth using laser ablation-inductively coupled plasma-mass spectrometry. Neurodevelopment was assessed between 6–16 years of age using the Wide Range Assessment of Visual Motor Abilities (WRAVMA). Mn associations with total WRAVMA scores and subscales were estimated with multivariable generalized additive mixed models. We examined Mn interactions with Pb and child sex in stratified models.

Results

Levels of dentine Mn were highest in the second trimester and declined steeply over the prenatal period, with a slower rate of decline after birth. Mn was positively associated with visual spatial and total WRAVMA scores in the second trimester, among children with lower (<median) tooth Pb levels: one standard deviation (SD) increase in ln-transformed dentine Mn at 150 days before birth was associated with a 0.15 [95% CI: 0.04, 0.26] SD increase in total score. This positive association was not observed at high Pb levels. In contrast to the prenatal period, significant negative associations were found in the postnatal period from ~6 to 12 months of age, among boys only: one SD increase in ln-transformed dentine Mn was associated with a 0.11 [95% CI: −0.001, −0.22] to 0.16 [95% CI: −0.04, −0.28] SD decrease in visual spatial score.

Conclusions

Using tooth-matrix biomarkers with fine scale temporal profiles of exposure, we found discrete developmental windows in which Mn was associated with visual-spatial abilities. Our results suggest that Mn associations are driven in large part by exposure timing, with beneficial effects found for prenatal levels and toxic effects found for postnatal levels.

1. INTRODUCTION

Manganese (Mn) is an essential nutrient necessary for normal brain development (Prohaska 1987; Takeda 2003; Wedler 1993). Maintaining Mn in the correct metabolic balance is important, as excess Mn exposure critically affects brain function (Aschner 2000). In children, both excessive intake and low blood Mn levels have been associated with adverse neurodevelopmental outcomes (Chung et al. 2015; Claus Henn et al. 2010; Menezes-Filho et al. 2011; Rink et al. 2014; Takser et al. 2003; Zoni and Lucchini 2013). Active transport of Mn across the placenta suggests that it plays a critical nutrient role in fetal development, but active transport during pregnancy also occurs in the context of an incompletely formed blood-brain barrier, which may make the fetus more vulnerable to relatively small disturbances in Mn homeostasis in pregnancy (Aschner and Aschner 2005; Ballatori et al. 1987; Yoon et al. 2011). Differences in Mn homeostatic mechanisms in young children, who absorb and retain a larger fraction of ingested Mn than adults, may create additional windows of vulnerability. Interestingly, blood Mn levels in children are consistently higher than blood levels in adults, suggesting that Mn continues to play an important developmental role in the postnatal period. On the other hand, dietary Mn deficiency can adversely affect neuronal activity and energy metabolism in the brain (Prohaska 1987; Takeda 2003; Wedler 1993). The effect of Mn on neurodevelopment may depend as much on the developmental stage when higher exposure is experienced as on dose, as the body’s need for Mn appears to vary over time. This may explain why exposures prenatally have markedly different effects than those observed with postnatal exposure (Ericson et al. 2007; Gunier et al. 2015; Mora et al. 2015). There is also evidence that Mn may interact with other metals, including lead (Pb), to enhance their toxicity (Claus Henn et al. 2012; Wright et al. 2006). These findings are supported by recent animal studies where co-exposures with Pb and Mn produced effects distinct from those observed under single exposure scenarios (Betharia and Maher 2012). Furthermore, in animals, Mn exposure exerts a variable effect on visual, motor and cognitive domains of intellectual function measures, necessitating the study of domain-specific associations in humans rather than only global scores of intellectual performance (Betharia and Maher 2012; Bonilla 1984; Cordova et al. 2013; Kern et al. 2010).

A major barrier to epidemiologic studies of the neurodevelopmental effects of prenatal Mn exposure is the lack of suitable biomarkers to measure fetal uptake. Maternal blood is a strong surrogate when passive transport occurs across the placenta and when there is a high correlation between maternal blood and cord blood levels, as is the case for blood lead. Mn, however, is actively transported, and it is unknown when in pregnancy the active transport mechanism is most active, as there are no methods to safely measure both maternal blood Mn and fetal blood Mn except at birth. Given all these factors, prenatal susceptibility windows when Mn exposure may be most strongly linked to neurodevelopmental outcomes have not been identified. Prior studies have utilized primarily cord blood or serum (Claus Henn et al. 2017; Lin et al. 2013; Yu et al. 2014), maternal blood (Chung et al. 2015; Claus Henn et al. 2017; Takser et al. 2003), and maternal or infant hair (Takser et al. 2003) to characterize Mn levels during the prenatal period. However, these spot measures are prone to measurement error, capturing only a short time frame of exposure, and cannot factor in the role of active transport, which may vary by pregnancy stage, genetics, co-exposures or disease states. A tooth biomarker, however, would overcome many of these barriers because deciduous ‘baby’ teeth are formed in pregnancy and levels in teeth represent fetal exposure downstream from active placental transport. Several studies using tooth-matrix biomarkers have distinguished trimester-specific Mn levels, and reported time-specific associations with neurodevelopmental outcomes (Gunier et al. 2015; Mora et al. 2015). Recently, we have improved upon this method by developing and validating tooth Mn biomarkers that can provide fine-scale temporal profiles of exposure on a nearly weekly basis over the second and third trimesters and during early childhood (Arora et al. 2011; Arora et al. 2012; Arora and Austin 2013). These tooth-matrix methods are based on the incremental nature of tooth development that commences prenatally and proceeds into childhood (akin to growth rings in a tree). Our more recent methodology allows for identification of narrower time frames, during which Mn may influence neurodevelopment (Arora and Austin 2013; Austin et al. 2013). Given the rapid, complex, and dynamic nature of brain development, it is paramount to characterize time-specific associations in this type of detailed manner.

Importantly, analytical methods for measuring Mn and other elements in the tooth biomarker have been extensively validated (Arora et al. 2006; Arora et al. 2011; Arora et al. 2012; Arora and Austin 2013; Arora et al. 2014; Hare et al. 2011). Validation studies in animals and humans demonstrate good correlations of tooth levels with levels in the environment and in other biospecimens. For example, in rodents with lifelong Mn exposure, tooth Mn levels were significantly, positively associated with levels in blood, brain, and bone (Austin et al. 2017). In a cohort of children living in an agricultural area where Mn-containing pesticides are used, Mn levels in dentine formed during the second trimester were significantly associated with Mn levels in house dust, and levels in dentine formed close to time of birth were significantly associated with cord blood Mn levels (Arora et al. 2012). In the same cohort, dentine Mn was also significantly associated with other exposure factors such as proximity to pesticide use and having a parent who was employed in pesticide spraying (Gunier et al. 2013). Levels of Pb in teeth measured using the same approach were associated with levels in umbilical cord blood and childhood blood, demonstrating that tooth Pb levels also reflect both timing and intensity of exposure (Arora et al. 2014).

Here, we apply tooth-matrix biomarkers in a prospective pregnancy cohort study in Mexico City to uncover potential prenatal and early postnatal developmental windows of susceptibility to Mn. We estimate time-specific associations between early life Mn levels and visual-motor, visual-spatial and fine-motor ability measured by the Wide Range Assessment of Visual Motor Ability (WRAVMA). Given prior evidence (Claus Henn et al. 2012; Gunier et al. 2015; Kim et al. 2009), we also explore effect modification by sex and Pb.

2. METHODS

2.1. Study participants

Mother–child pairs in this study were drawn from four successively enrolled longitudinal birth cohort studies in Mexico City that comprise the Early Life Exposures in MExico and NeuroToxicology (ELEMENT) project. These cohorts are uniquely poised to examine the effects of Mn because (1) air pollution, of which Mn and other metals are key components, is severe in Mexico City (Calderon-Garciduenas et al. 2013); (2) rich sources of Mn in the diet, such as beans, are commonly consumed in Mexican diets; and (3) higher Mn exposure, measured in biological and environmental samples, has been reported in Mexico than in the U.S. and Canada (Santos-Burgoa et al. 2001). Detailed information on the study design and data collection procedures has been published previously (Braun et al. 2012; Ettinger et al. 2009; Gonzalez-Cossio et al. 1997; Hernandez-Avila et al. 2002). Briefly, mothers were originally recruited during pregnancy or at delivery between 1994 and 2004 to investigate the long-term consequences of prenatal environmental factors on child development (Ettinger et al. 2009; Gonzalez-Cossio et al. 1997; Tellez-Rojo et al. 2004). All four cohorts were recruited at public maternity hospitals, which serve low to moderate income populations in Mexico City. Exposure, outcome, and covariate data from all cohorts were collected in a standard manner by the same study staff, allowing us to pool data across cohorts (see Supplemental Table 1 for comparison of characteristics across cohorts). Exclusion criteria included an Apgar score of ≤6 at 5 minutes, a condition requiring treatment in a neonatal intensive care unit; serious birth defects; and maternal factors including psychiatric illness, seizures, or kidney or cardiac disease; preeclampsia (systolic BP > 140 mmHg or diastolic BP > 90 mmHg); gestational diabetes; daily consumption of alcoholic beverages; addiction to illegal drugs; continuous use of corticosteroids; and factors that could interfere with calcium metabolism. Only one child for each mother was included in this study, regardless of birth order. Of the 1,079 children born into the cohort and followed until 5 years of age, 826 (77%) returned for an additional follow-up assessment between 2008 and 2010 at 6–16 years of age. Tooth collection was incorporated into this visit. Participants were asked to bring or mail in teeth as they were naturally shed. Deciduous teeth were available for metals analysis on 138 of these children.

All participating mothers signed a written consent form and received a detailed explanation of the study intent and research procedures. In addition, at the 6- to 16-year follow-up assessment, the children signed a written assent of minor form and received a detailed explanation of the study. All participants were encouraged to ask questions about the study in order to ensure their understanding. The research protocol was approved by the Research, Ethics, and Biosafety committees of the National Institute of Public Health of Mexico and National Institute of Perinatology, and by the human subjects committees of the Harvard T.H. Chan School of Public Health, Michigan School of Public Health, and the American British Cowdray Medical Center.

2.2. Measurement of manganese in teeth by LA-ICP-MS

We analyzed teeth that were free of obvious defects such as caries and extensive tooth wear. Our approach to measuring Mn and other metals in teeth using laser ablation-inductively coupled plasma mass spectrometry (LA-ICP-MS) and assigning developmental times has been detailed elsewhere (Arora and Austin 2013; Austin et al. 2013; Hare et al. 2011). Briefly, we used the neonatal line (a histological feature formed in enamel and dentine at birth) and daily growth incremental markings to assign temporal information to sampling points. Dentin of incisors, the first teeth to mineralize, begins to form at approximately 3 months gestation and continues until the tooth is shed (Ash and Nelson 2003).

The laser ablation unit used was a New Wave Research NWR-193 system (Kennelec Technologies, Mitcham, Victoria, Australia) equipped with an excimer argon-fluoride laser emitting a nanosecond laser pulse with a wavelength of 193 nm, and a two volume ablation chamber. An approximately 40 cm length of Tygon® tubing (i.d. 3 mm) connected the laser ablation unit to an Agilent Technologies 8800 (Agilent Technologies Australia, Forrest Hill, Victoria, Australia) ICP-MS. The instrument was fitted with a ‘cs’ lens system for enhanced sensitivity. The system was tuned daily for sensitivity using NIST SRM 612 (trace elements in glass). Polyatomic oxide interference was evaluated and minimized by monitoring the Th+/ThO+ (m/z 232/248) ratio. Typical oxide formation was consistently under 0.3%. Using the laser we sampled 50 sampling points of 35 μm diameter in enamel and primary dentine adjacent to the enamel-dentine junction. Depending on tooth type, these 50 sampling points per tooth covered a timeline from the start of the second trimester (14 weeks gestation) through one year of age, thereby representing a sampling frequency of approximately every 10 days. The detection limit was 0.05 μg/g for each metal. Metal intensities were normalized to calcium (Ca) to control for any variations in mineral content within a tooth and between tooth type and individuals, and because there are no matrix-matched certified reference materials for dentine-Mn. Data were analyzed as ⁵⁵Mn:⁴³Ca and ²⁰⁸Pb:⁴³Ca ratios.

2.3. Measurement of child neurodevelopment and potential confounders

Child neurodevelopment was assessed between 6 and 16 years of age using the Wide Range Assessment of Visual Motor Abilities (WRAVMA). The WRAVMA is an age-normed test of visual motor development that is moderately correlated with IQ (r~0.60) (Adams and Sheslow 1995). Trained study staff (three testers) administered the WRAVMA pegboard (fine motor), matching (visual spatial), and drawing (visual motor) subtests. On the pegboard task, children insert as many round pegs on a wooden board as possible within 90 seconds, using the dominant hand. On the matching task, children choose a picture that “goes best” with a stimulus picture. On the drawing task, children copy a series of line drawings arranged in order of increasing difficulty. Scores on individual subtests are reported and then combined to yield a total score of overall visual motor development. Each subtest and total scores have an expected mean of 100 and standard deviation of 15. The subscale and total scores were used as primary outcomes in our analysis. Study staff administering the WRAVMA were blinded to children’s exposure levels and were overseen by an expert neuropsychologist (LS).

Information on sociodemographic and other characteristics that could confound the relationship between Mn exposure and child neurodevelopment were collected by questionnaire at enrollment, delivery, 1 month postpartum, and/or during subsequent study visits. These characteristics included child sex, maternal characteristics (age at time of delivery, marital status, smoking during pregnancy), maternal and paternal education, and characteristics of the household. Socioeconomic status (SES) index was calculated based on the algorithm AMAI rule 13x6 from 1994, in which families are classified into six levels based on 13 questions about characteristics of the household: (1) education of the head of household; (2) number of rooms; (3) number of bathrooms with showers; (4) type of floor; (5) number of light bulbs; ownership of: (6) car; (7) hot water heater; (8) automatic washing machine; (9) videocassette recorder; (10) toaster; (11) vacuum cleaner; (12) microwave oven; and (13) personal computer (Carrasco 2002). The six levels were then collapsed into low, medium and higher SES. Maternal IQ was assessed with the Information, Comprehension, Similarities, and Block Design subscales of the Wechsler Adult Intelligence Scale, Spanish version (Wechsler 1968). Anthropometric data from the mother and newborn were gathered within 12 hr of delivery. Information on estimated gestational age at birth, based on the date of last menstrual period, and characteristics of the birth and newborn period, including birth weight and head circumference, were abstracted from medical records. Hemoglobin was measured in whole blood samples collected from children at 1- and/or 2-years of age. Following WHO guidance, we adjusted hemoglobin values for altitude by −0.8 g/dl in all participants, because they reside at ~2000 m elevation in Mexico City (WHO 2011). Anemia status (ever anemic during early childhood), was defined as having hemoglobin <11.0 g/dl, after altitude adjustment, at either 1- or 2-years of age (WHO 2011).

2.4. Statistical Analysis

We conducted standard univariate and bivariate explorations of the data and examined distributional plots for all variables. Neurodevelopment (WRAVMA) scores were approximately normally distributed and were modeled as continuous outcomes. Mn and Pb concentrations were right-skewed; therefore the data were natural log (ln)-transformed to satisfy model assumptions.

We used two approaches to uncover discrete developmental periods when tooth Mn levels were most strongly associated with WRAVMA total and subtest (visual motor, visual spatial, fine motor) scores. Initially, we averaged the fine-scale resolution tooth Mn within each subject for each of four time periods: 150 to 90 days before birth (approximately second trimester), 89 days to 1 day before birth (approximately third trimester), birth to 180 days after birth (0 to 6 months postnatal), and 181 to 365 days after birth (>6 to 12 months postnatal). We fit separate multivariable generalized additive models (GAMs) for each of the four time periods and modeled average Mn levels with penalized splines. When associations appeared linear based on visual examination, average Mn levels were also modeled as linear terms. This initial approach using Mn levels averaged over four time periods has two limitations: (1) it treats each time period separately, which fails to account for associations at other time points, and (2) it is subject to exposure misclassification given that tooth metals data are averaged over periods of time when Mn levels are changing. Relatively high correlations of average Mn levels between time windows (e.g., Spearman r = 0.7 for Mn_{second trimester} and Mn_{third trimester}; r = 0.6 for Mn_0–6months and Mn_{>6–12months}) precluded us from fitting models for each time window adjusted for Mn levels from previous time windows (e.g., models for Mn_0–6months including Mn_{second trimester} and Mn_{third trimester}).

As a second approach, to address the limitations of the first approach, we used an adapted distributed lag-type approach that treats X (tooth metals data) as a longitudinal dependent variable and Y (WRAVMA score) as a time-invariant independent variable. The approach is valid because an estimate of association between two variables is sound no matter which is specified as response or predictor, and has the advantage of using the measured exposure concentrations directly with no further assumption about the temporal trend. Specifically, we used the following generalized additive mixed model (GAMM) to characterize associations between the fine-scale resolution tooth Mn data (not averaged) and neurodevelopment: X_i(t) = β₀ + b_0i + f₁(t) + f₂(t)Y_i + f₃(t)Pb_i + β_ZZ_i + ε_it, where X_i(t) is the continuous tooth Mn measurement for subject i at time t, b_0i is the random intercept for subject i, Y_i is the continuous WRAVMA score for subject i, Pb_i is the continuous tooth Pb measurement for subject i, Z_i is the vector of covariates, and f₁(.), f₂(.), and f₃(.) are smooth functions of the data, e.g., f₃(.) is a smooth function of the fine-scale resolution tooth Pb levels (Chen et al. 2015). The model allows for subject-specific random intercepts to account for non-independence of the repeated tooth Mn measurements for each subject, and to allow for subject-specific baseline Mn levels. Additional details are available in previously reported applications for identifying time windows of susceptibility (Chen et al. 2015; Sanchez et al. 2011). For these models, we standardized neurodevelopment scores as well as (natural log-transformed) tooth metal concentrations by dividing by their standard deviations to obtain a common variance. We report results as standardized coefficients; therefore, output of this model can be interpreted as a correlation between tooth Mn levels and neurodevelopment. Smoothly-varying standardized coefficients (f₂(t)) and 95% confidence intervals over time are plotted, and a sensitive window is identified when the estimated confidence intervals on the correlation between exposure and outcome do not include zero.

Covariates that are known predictors of neurodevelopment or strong potential confounders of the Mn-neurodevelopment association were included a priori in regression models, based on previous literature (Claus Henn et al. 2010; Mora et al. 2015) and on biologic plausibility. In all models, we included child sex, tooth Pb levels (allowed to vary smoothly over time), maternal IQ, maternal education (years of school), and study cohort. WRAVMA scores are age-normed; therefore, child age at time of WRAVMA assessment was not included as a covariate in the models. We controlled for study cohort because cognitive test examiners and childhood blood Pb levels varied across cohorts (Schnaas et al. 2004). We additionally considered the following potential confounders in sensitivity analyses, but did not include them in main models because they may lie on the causal pathway between prenatal Mn and childhood neurodevelopment (e.g. gestational age, birth weight) and/or because they did not materially change the Mn associations: maternal age at delivery and anemia status (ever anemic during early childhood).

Given previous evidence of Mn-Pb interactions (Betharia and Maher 2012; Claus Henn et al. 2012) and the potential for Pb to modify Mn associations, we stratified analyses by Pb at the median of the distribution of average tooth Pb levels (i.e., median of mean values for each subject over entire prenatal and postnatal assessment periods). Recent literature also suggests possible sex-specific Mn associations with neurodevelopment (Bauer et al. 2017; Chiu et al. 2017; Gunier et al. 2015; Menezes-Filho et al. 2011). Therefore, separately, we explored whether Mn associations differed for girls and boys in our cohort using sex-stratified models.

Some participants were missing data on key potential confounders (e.g., maternal IQ, maternal education). In sensitivity analyses, we therefore multiply imputed missing values using chained equations with the MI procedure in SAS (SAS Institute, Inc., Cary, NC, USA) (van Buuren 2007; White et al. 2011). We assumed data were missing at random and generated 20 imputed datasets. We included all exposure, outcome, and potential confounder data thought to be related to the process causing missing data (see Supplemental Table 2). We combined results from models fit with the multiply imputed datasets averaging the results of the 20 imputations to give the final effect estimates for sensitivity analyses. We calculated standard errors using methods that combine the within- and between-imputation uncertainty (Rubin 2004).

We identified outliers of log-dentine Mn (8 measurements on n=6 participants) using the generalized Extreme Studentized Deviate Many-Outlier procedure (Rosner 1983). Results with and without these participants were compared in sensitivity analyses. We also conducted a sensitivity analysis excluding premature (<37 weeks gestation) or low weight (<2500 g) births (n=10).

Statistical analyses were conducted using SAS version 9.4 (SAS Institute, Inc, Cary, NC, USA) and R version 3.3.2 (The R Foundation for Statistical Computing, www.r-project.org).

3. RESULTS

3.1. Sociodemographic characteristics, neurodevelopment scores, and tooth metals levels

A total of 138 children had available neurodevelopment scores and tooth metals data and were included in analyses. Mothers who participated in this study were mostly under 30 years of age (65.6%), had completed high school (55.4%), and reported not smoking during pregnancy (98.3%) (Table 1). A majority of the children were girls (53.6%), and few were born preterm (less than 37 weeks gestation, 5.9%) or had low birth weight (less than 2500 g, 3.0%). A substantial portion of the children had experienced iron deficiency anemia in the first two years of life (57.5%), based on their hemoglobin levels. The median age at WRAVMA testing was 8 years, with a majority (81.2%) completing the test between 7 and 9 years of age. The mean (±SD) WRAVMA total score was 98.0 (±10.6); visual motor, 98.0 (±11.3); visual spatial, 94.9 (±13.4) and fine motor, 104.0 (±13.7). WRAVMA total scores were significantly correlated with visual motor, visual spatial, and fine motor scores (Spearman r=0.6, p<0.001 for each pair), but subtest scores were not highly correlated with each other (e.g., visual motor and fine motor, r=0.2, p=0.05; visual motor and visual spatial, r=0.01, p=0.9). For most characteristics, children included in analyses were similar to those excluded due to missing tooth metals data (Table 1).

Table 1.

Sociodemographic characteristics of participants included in analysis (n=138)^a and excluded from analysis due to missing tooth metals data (n=688)

	Included in analysis		Excluded from analysis
	N	Mean ± SD or %	N	Mean ± SD or %
Child sex:
Male	64	46.4%	355	51.6%
Female	74	53.6%	333	48.4%
Gestational age at birth (weeks)	135	38.9 ± 1.6	685	38.9 ± 1.5
Preterm birth (<37 weeks gestation)	8	5.9%	42	6.1%
Birth weight (kg)	135	3.2 ± 0.4	687	3.1 ± 0.4
Low birth weight (<2500 g)	4	3.0%	44	6.4%
First child:
Yes	19	17.9%	3	0.5%
No	87	82.1%	639	99.5%
Maternal age at delivery (yrs)	131	28.6 ± 5.8	686	28.9 ± 5.9
<30 yrs old	86	65.6%	372	54.2%
Maternal education (yrs school)	130	11.2 ± 3.0	685	9.0 ± 3.9
≥12 yrs school	72	55.4%	218	31.8%
Maternal IQ score	120	93.6 ± 20.0	628	88.1 ± 22.2
Family SES:
Low	65	55.1%	188	54.0%
Medium	50	42.4%	151	43.4%
High	3	2.5%	9	2.6%
Maternal smoking during pregnancy:
Yes	2	1.7%	15	4.5%
No	115	98.3%	320	95.5%
Hemoglobin at 1 yr of age	95	12.4 ± 1.4	380	12.0 ± 1.3
Hemoglobin at 2 yrs of age	111	13.2 ± 1.1	537	12.7 ± 1.2
Anemic at 1 or 2 years of ageb:
Yes (hemoglobin ever <11.0 g/dl)	73	57.5%	429	73.0%
No (hemoglobin never <11.0 g/dl)	54	42.5%	159	27.0%
Study cohort (years of child birth):
1 (1994–1995)	2	1.4%	202	29.4%
2A (1997–2002)	19	13.8%	243	35.3%
2B (1998–2001)	33	23.9%	111	16.1%
3 (2001–2006)	84	60.9%	132	19.2%
Age at WRAVMA testing (yrs)	138	8.0 ± 1.4	688	10.6 ± 2.6
WRAVMA score: Total	138	98.0 ± 10.6	688	98.1 ± 13.1
WRAVMA score: Visual motor	138	98.0 ± 11.3	688	96.8 ± 12.9
WRAVMA score: Visual spatial	138	94.9 ± 13.4	688	94.2 ± 14.7
WRAVMA score: Fine motor	138	104.0 ± 13.7	688	103.0 ± 13.0

^aMissing data for some covariates: birth weight (n=3), gestational age (n=3), first child (n=32), mother’s age at delivery (n=7), mother’s education (n=8), mother’s IQ (n=18), SES (n=20), mother smoked during pregnancy (n=21), hemoglobin at 1 yr (n=43), hemoglobin at 2 yrs (n=27), anemic (n=11).

^bAnemic defined by WHO for children 6–59 months of age as hemoglobin <11.0 g/dl, after altitude adjustment. Hemoglobin levels in participants were adjusted for altitude by −0.8 g/dl (participants reside at ~2000 m elevation in Mexico City) to define anemia status.

Ref: http://www.who.int/vmnis/indicators/haemoglobin.pdf, accessed May 2016

The overall median (25^th to 75^th percentile) Mn level (as ⁵⁵Mn:⁴³Ca) across all time points was 1.2E-3 (6.8E-4 to 2.0E-3). Levels of Mn (as ln ⁵⁵Mn:⁴³Ca) were highest in the second trimester and declined steeply over the prenatal period. Levels continued to decline after birth but at a slower rate (Figure 1A). The median Mn level (original scale) at 120 (± 7) days before birth was 3.2E-3 ⁵⁵Mn:⁴³Ca (range: 4.6E-4 to 8.1E-3); at birth (± 3 days) was 1.1E-3 (range: 1.7E-4 to 5.1E-3); at 6 months (i.e., 180 ± 7 days) postnatally was 6.9E-4 (range: 2.0E-4 to 2.0E-3); and at 1 year (i.e., 365 ± 7 days) was 6.4E-4 (range: 1.1E-5 to 1.3E-3). In contrast, tooth-Pb levels (as ²⁰⁸Pb:⁴³Ca) were steadier over time, with slightly increasing levels after birth (Figure 1B). Ln-Mn levels at birth (± 3 days) were approximately 60% and 22% lower among children in cohorts 1 and 2B, respectively, than children in cohort 3 (% change calculated as: (e^β−1)*100; cohort 1 vs. 3: β = −0.92 [95% CI: −1.70, −0.14]; cohort 2B vs. 3: β = −0.25 [95% CI: −0.48, −0.03]) (Table 2), although these estimates are based on limited numbers of observations (n=2, n=33, respectively). Tooth-Mn was not associated with sex, gestational age, birth weight, anemia status, maternal age at delivery, maternal education, or maternal IQ. Tooth-Pb levels at birth (± 3 days) were positively associated with maternal age at delivery (β = 0.03 [95% CI: 0.001, 0.05]), and were approximately 50% higher among children in cohort 2B compared to cohort 3 (β = 0.41 [95% CI: 0.10, 0.73]) (Table 3).

Figure 1.

Tooth metal levels from early second trimester (150 days before birth) to one year of age (365 days after birth) for (A) Mn, as ln ⁵⁵Mn:⁴³Ca ratio, and (B) Pb, as ln ²⁰⁸Pb:⁴³Ca ratio. Dots represent individual tooth metal measurements for 138 participants, with approximately 50 measurements per participant. Line represents Loess smoother.

Table 2.

Unadjusted associations between covariates and tooth Mn (ln ⁵⁵Mn:⁴³Ca) around time of birth^a

	N	Beta (95% CI)
Child sex, female vs. male	138	0.08 (−0.11, 0.27)
Gestational age at birth (wks)	135	−0.01 (−0.07, 0.05)
Preterm birth (<37 wks gestation) vs. normal	135	0.07 (−0.34, 0.47)
Birth weight (g)	135	0.16 (−0.07, 0.38)
Mother’s age at delivery (yrs)	131	0.01 (−0.01, 0.03)
≥30 yrs old vs. <30	131	0.02 (−0.19, 0.23)
Mother’s education, <12 yrs school vs. ≥12 yrs	130	0.09 (−0.10, 0.28)
Mother’s IQ score	120	−0.001 (−0.007, 0.004)
Mother smoked during pregnancy, yes vs. no	117	−0.77 (−1.58, 0.04)
Hemoglobin at 1 yr of age (g/dl)	95	−0.05 (−0.13, 0.03)
Hemoglobin at 2 yrs of age (g/dl)	111	0.001 (−0.09, 0.10)
Anemic at 1 or 2 years of age, yes vs. no	127	0.14 (−0.06, 0.34)
Study cohort (years of child birth)
1 (1994–1995) vs. 3 (2001–2006)	2	−0.92 (−1.70, −0.14)
2A (1997–2002) vs. 3 (2001–2006)	19	−0.17 (−0.45, 0.11)
2B (1998–2001) vs. 3 (2001–2006)	33	−0.25 (−0.48, −0.03)

^aAround time of birth defined as birth ± 3 days

Table 3.

Unadjusted associations between covariates and tooth Pb (ln ²⁰⁸Pb:⁴³Ca) around time of birth^a

	N	Beta (95% CI)
Child sex, female vs. male	138	−0.01 (−0.28, 0.26)
Gestational age at birth (wks)	135	0.04 (−0.04, 0.13)
Preterm birth (<37 wks gestation) vs. normal	135	−0.35 (−0.93, 0.23)
Birth weight (g)	135	0.26 (−0.05, 0.58)
Mother’s age at birth (yrs)	131	0.03 (0.001, 0.05)
≥30 yrs old vs. <30	131	0.18 (−0.12, 0.47)
Mother’s education, <12 yrs school vs. ≥12 yrs	130	0.08 (−0.20, 0.36)
Mother’s IQ score	120	0.0004 (−0.007, 0.008)
Mother smoked during pregnancy, yes vs. no	117	−0.43 (−1.62, 0.75)
Hemoglobin at 1 yr of age (g/dl)	95	0.06 (−0.06, 0.19)
Hemoglobin at 2 yrs of age (g/dl)	111	−0.06 (−0.19, 0.07)
Anemic at 1 or 2 years of age, yes vs. no	127	−0.02 (−0.31, 0.26)
Study cohort (years of child birth)
1 (1994–1995) vs. 3 (2001–2006)	2	0.09 (−1.02, 1.20)
2A (1997–2002) vs. 3 (2001–2006)	19	0.38 (−0.01, 0.77)
2B (1998–2001) vs. 3 (2001–2006)	33	0.41 (0.10, 0.73)

^aAround time of birth defined as birth ± 3 days

3.2. Critical windows of neurodevelopmental susceptibility to Mn exposure

With our first approach, using models with Mn levels averaged over each of four time periods, we observed no significant associations for any time period with the WRAVMA total or subtest scores, after adjusting for child age, sex, tooth Pb levels, maternal IQ, maternal education, and study cohort (Figure 2 for total scores; Supplemental Figure 1 for subtests). Slopes were generally weakly positive or null for Mn measurements averaged across the second and third trimesters, and weakly negative for measurements averaged across the postnatal periods, with the exception of the fine motor subtest (associations with total score: second trimester, β = 3.5 [95% CI: −1.6, 8.7]; third trimester, β = 1.6 [95% CI: −3.2, 6.4]; 0 to 6 months, β = −0.5 [95% CI: −4.9, 3.9]; 6 to 12 months, β = −0.6 [95% CI: −6.7, 5.6]).

Figure 2.

Smoothed associations between tooth Mn levels, averaged over each of four time periods, and child WRAVMA total score. Generalized additive models with penalized splines (constrained to 4 knots), adjusted for child sex, tooth Pb levels (centered at mean of distribution), maternal IQ, maternal education, and study cohort. Models use complete case method. (A) Second trimester, n=104, p_GAM=0.18; (B) Third trimester, n=114, p_GAM=0.50; (C) Birth to 6 months, n=114, p_GAM=0.82; (D) Six to 12 months, n=56, p_GAM=0.85.

With our second approach, using the adapted distributed lag-type method described above, results for Mn associations were consistent with results from our initial approach (i.e., using Mn levels averaged over each of four time periods): positive associations, indicated by standardized coefficients above zero, were estimated during the prenatal period, and negative correlations, suggesting an adverse Mn effect, were estimated mostly during the postnatal period (Figure 3). However, no statistically significant developmental window of susceptibility was identified, as confidence intervals all included zero.

Figure 3.

Associations between fine-scale resolution tooth Mn levels over time and WRAVMA scores. GAMMs using adapted distributed lag-type approach, adjusted for child sex, tooth Pb levels, maternal IQ, maternal education, and study cohort. Models use complete case method. Dark blue line represents association over time; light blue shading represents 95% confidence intervals. Thin black line indicates correlation = 0. (A) Total score; (B) Fine motor subtest score; (C) Visual spatial subtest score; (D) Visual motor subtest score.

When we stratified models by tooth Pb, however, we observed significant associations of tooth Mn levels with WRAVMA total and visual spatial subtest scores, among children with low tooth Pb levels (Figures 4A, 4C). At low Pb, higher tooth Mn in the second trimester was associated with better performance on visual motor ability. Significant positive associations were observed with total WRAVMA scores during the exposure window from 150 to 79 days before birth: at 150 days before birth, a one unit increase in SD of ln-transformed dentine Mn was associated with a 0.15 (95% CI: 0.04, 0.26) increase in SD of WRAVMA total score; at 79 days before birth, a one unit increase in SD of ln-transformed dentine Mn was associated with a 0.10 (95% CI: 0.0002, 0.21) increase in SD of WRAVMA total score. For visual spatial scores, we also found positive associations during the second trimester, with significant associations from 150 to 92 days before birth (150 days before birth: β=0.17 [95% CI: 0.07, 0.28]; 92 days before birth: β=0.10 [95% CI: 0.0003, 0.19]). A similar pattern was observed with fine motor scores, where increasing prenatal Mn levels were associated with better performance in children with low Pb levels, although associations did not reach statistical significance (Figure 4B). In contrast, the pattern with visual motor scores was different from the other three outcomes, with suggestive positive associations in the postnatal period (Figure 4D). Among children with high tooth Pb levels, there were no significant Mn associations with any WRAVMA outcomes.

Figure 4.

Associations between fine-scale resolution tooth Mn levels and WRAVMA scores, stratified by tooth Pb levels at median. GAMMs using adapted distributed lag-type approach, adjusted for child sex, tooth Pb levels, maternal IQ, maternal education, and study cohort. Models use complete case method. Dark blue line represents association over time; light blue shading represents 95% confidence intervals. Thin black line indicates correlation = 0. (A) Total score; (B) Fine motor subtest score; (C) Visual spatial subtest score; (D) Visual motor subtest score.

When we stratified models by child sex, we estimated weak positive correlations in the prenatal period, followed by significant negative correlations during the postnatal period among boys for both WRAVMA total and visual spatial subtest scores. A critical developmental window was identified from 187 to 365 postnatal days, corresponding to ~6 months to 1 year of age, for the association between Mn and visual spatial scores among boys (187 days after birth: β=−0.11 [95% CI: −0.0001, −0.22]; 365 days after birth: β=−0.16 [95% CI: −0.04, −0.28]). Similarly, for total scores, a critical window was identified from ~6 to 10 months of age (i.e., 176 to 309 postnatal days) (Figure 5). A unit increase in SD of ln-transformed dentine Mn was associated with a 0.13 [95% CI: −0.25, −0.01] decrease in SD of total WRAVMA score at 274 days after birth. No significant association was observed for fine motor or visual motor subtests among boys, and no significant association was observed among girls for any outcome.

Figure 5.

Associations between fine-scale resolution tooth Mn levels and WRAVMA scores, stratified by child sex. GAMMs using adapted distributed lag-type approach, adjusted for tooth Pb levels, maternal IQ, maternal education, and study cohort. Models use complete case method. Dark blue line represents association over time; light blue shading represents 95% confidence intervals. Thin black line indicates correlation = 0. (A) Total score; (B) Fine motor subtest score; (C) Visual spatial subtest score; (D) Visual motor subtest score.

In sensitivity analyses, we ran models with multiply imputed covariate data, and results were similar to models using complete data (Supplemental Figure 2). Models excluding outliers of dentine Mn (n=6) were similar to models with all data, although Mn correlations in the postnatal period among children with low Pb were more negative for all outcomes (Supplemental Figure 3). Similarly, in models excluding premature or low weight births (n=10), stronger negative correlations for Mn were estimated in the postnatal period among low Pb children, in particular for total and visual motor scores (Supplemental Figure 4).

4. DISCUSSION

Manganese is an essential nutrient that has clear neurotoxic properties with excess exposure. Increasing evidence supports that both deficiency and excess Mn levels are associated with adverse neurodevelopmental outcomes (Chung et al. 2015; Claus Henn et al. 2010; Gunier et al. 2015; Kordas et al. 2015; Lin et al. 2013; Mora et al. 2015; Ode et al. 2015; Rodriguez-Barranco et al. 2013; Yu et al. 2014). In particular, both higher and insufficiently low blood Mn levels, suggestive of an imbalance in Mn homeostasis, are being associated with poorer developmental test scores (Chung et al. 2015; Claus Henn et al. 2010; Lin et al. 2013; Ode et al. 2015; Yu et al. 2014). The weight of the evidence suggests that the biological role of Mn in brain development is highly complex and that maintaining homeostasis is crucial.

Emerging evidence suggests that effects of Mn are dependent on the developmental age at which exposures occur (Claus Henn et al. 2010; Gunier et al. 2015; Lin et al. 2013; Mora et al. 2015). Proper homeostatic regulation can be as time dependent as environmental exposures, as blood Mn has well-known life stage dependencies, with higher levels occurring in infancy/childhood than in adulthood. Because of the complex interaction of Mn with brain function, it has been challenging to disentangle the dose and time periods when Mn is beneficial for neurodevelopment versus when Mn exposure may lead to neurodevelopmental decrements. Furthermore, the effects of Mn may be modified by co-exposure to other metals, with both human and animal studies showing effect modification between concurrent Pb and Mn levels on neurobehavioral outcomes (Betharia and Maher 2012; Claus Henn et al. 2012). Our tooth biomarker enables us to examine time-dependent associations of prenatal Mn with health outcomes, and has the advantage of being a direct measure of fetal levels, rather than a surrogate like maternal blood. Furthermore, because the tooth biomarker measures Mn levels repeatedly over time, the shifts in homeostasis (i.e. decreases or increases in levels due to developmental life stage) are captured, allowing us to determine the role of Mn levels in predicting subsequent neurodevelopment. Here, we combined tooth-matrix biomarkers, which provide fine scale temporal profiles of Mn levels over the prenatal and early childhood periods, with recently developed statistical methods (adapted distributed lag-type approach) to study developmental stages when Mn levels are significantly associated with fine-motor, visual-motor and visual-spatial ability (as measured by the WRAVMA test) in Mexican children.

We observed a complex relationship between Mn exposure in early life and neurodevelopment, which was dependent on developmental stage, level of Pb exposure, and child sex. Small but significant associations were estimated for narrow time windows, and the findings overall suggest a potential shift over time in the direction of the Mn association with visual motor ability scores, from positive associations during the second and third trimesters of pregnancy to weak negative associations during the early postnatal period. The positive association of prenatal tooth Mn levels with visual motor ability scores was most evident in the visual spatial subtest among low Pb children, over a narrow developmental time window from 150 days before birth (approximately gestational week 17) to birth. This significant positive correlation in the prenatal period was not observed among children with high Pb, suggesting that in the presence of higher Pb levels, the benefit of prenatal Mn is dampened. Finally, negative associations estimated in the postnatal period were significant in boys only, suggesting a sexually dimorphic Mn association with visual motor ability.

Our finding of a positive Mn association in the prenatal period with visual motor ability, particularly visual spatial ability, is consistent with Mn being an essential element critical for healthy brain development. Mn may be in greatest demand during the prenatal period when developmental processes of neuronal proliferation and migration are at their peak (Tau and Peterson 2010). The positive Mn correlation we observed in the prenatal period among only low Pb exposed children suggests that higher Pb levels diminish the beneficial Mn effects, thereby modifying the Mn association. Interactions between Mn and Pb have been reported previously (Betharia and Maher 2012; Claus Henn et al. 2012; Kim et al. 2009) and may be related to competitive binding to receptors (e.g., divalent metal transporter [DMT1], transferrin receptor [TfR]) or altered distribution of metals in the body including in target organs such as the brain (Betharia and Maher 2012; Shukla and Chandra 1987). While prenatal Mn levels appear to be positively associated with visual motor ability, Mn benefits could, in the presence of neurotoxic co-exposures such as Pb, be overwhelmed or masked.

In contrast to the in utero period, early postnatal Mn levels were associated with poorer visual spatial ability, particularly among boys. In the postnatal period, the placenta is no longer controlling or regulating Mn levels that are transported to the fetus. Instead, direct exposure to the young child can occur, and with potentially underdeveloped homeostatic mechanisms and a lower need for Mn than during the prenatal period, the developing central nervous system in early postnatal life may be more sensitive to excess Mn levels. Evidence from our study suggests that visual spatial ability, in particular, may be more sensitive to increasing Mn levels than fine motor function, among boys. Mn appears to target the basal ganglia, a brain region whose primary functions are to control voluntary and fine motor movements as well as cognition and emotion, but basal ganglia dysfunction has also been associated with altered visual perception (Middleton and Strick 1996).

Sex-specific Mn associations have also been reported previously in epidemiologic studies, but results are inconsistent. Sexual dimorphism in the association between Mn and neurodevelopment may be related to biological differences in neurochemistry and hormone activity (Ngun et al. 2011), such as testosterone, which has been shown to improve visuospatial performance in animals (Lund and Lephart 2001) and humans (Mueller et al. 2008). Animal studies have demonstrated sex differences in the accumulation of Mn in body tissues (Dorman et al. 2004; Yamagata et al. 2017), neuron morphology (Madison et al. 2011), and performance on tests of behavior and motor function (Yamagata et al. 2017).

Our findings have some consistencies with prior studies. A similar shift in the direction of the Mn association with neurodevelopmental outcomes between the prenatal and postnatal periods has been reported previously (Gunier et al. 2015). Among 197 children residing in an agricultural area of California, significant negative associations were reported between postnatal, but not prenatal, Mn levels in tooth dentine with mental and psychomotor development scores on the Bayley Scales of Infant Development-Second Edition (BSID) at 6 and 12 months of age, controlling for prenatal Mn (Gunier et al. 2015). Although this study did not observe significant beneficial associations of prenatal Mn, as we did in our study, findings are generally consistent with ours: results suggest that Mn in the postnatal period adversely impacts brain development, but that in utero Mn may play a different, potentially beneficial, role during fetal development. This is congruous with prior evidence of up-regulated maternal Mn absorption as a response to high fetal demand for Mn to support healthy and rapid growth and development in the prenatal period (Mistry and Williams 2011; NAS 2001).

In the same cohort as Gunier et al. (2015) but including 248 school-age children, positive associations were estimated for prenatal dentine Mn with scores on tests of cognition, motor function, and visuospatial memory (Mora et al. 2015). These results are consistent with our findings, although associations in the California study were observed among boys only. In contrast to our postnatal results, however, Mora et al. reported that higher postnatal dentine Mn was associated with better WRAVMA fine motor scores, the only WRAVMA subtest administered, among boys, although this was only for the non-dominant hand. No significant associations were observed for the dominant hand, which is the hand used in our study. In the California study, results from Pb-stratified analyses varied by outcome, but some Mn associations (e.g., with behavior) were only apparent in children exposed to lower prenatal Pb levels. The primary source of Mn exposure in the California studies was Mn-containing fungicides applied to agriculture, while in our Mexico City cohort, primary sources have not been identified; exposures likely occur from a combination of dietary and airborne sources. The only other study of neurodevelopment that measured Mn in the prenatal and postnatal periods used enamel to obtain an exposure estimate at approximately gestational week 20 and at 7 months postnatal (Ericson et al. 2007). At both time points, higher Mn levels were associated with more behavioral problems (impulsivity, externalizing and attention problems) among 27 U.S. school-age children. Of note, the aforementioned studies employed different approaches for estimating Mn levels in teeth: Ericson et al. (2007) used enamel, capturing only snapshots at two distinct times; Gunier et al. (2015) and Mora et al. (2015), similar to our study, sampled dentine, but used an earlier technology to measure Mn in the tooth-matrix biomarker that was integrated over longer time periods to derive a single prenatal or postnatal measurement, and thus may not have captured time dependent effects with the high resolution that we used here.

Associations of Mn with neurobehavior have also been reported in longitudinal studies using umbilical cord blood and childhood blood Mn measures (Claus Henn et al. 2012; Lin et al. 2013). As previously noted, it is important to consider that associations of maternal pregnancy or cord blood Mn with children’s neurodevelopment may not be directly comparable to associations observed with prenatal dentine Mn biomarkers, as Mn is actively transported across the placenta. The degree of active transport may vary by pregnancy stage, disease state, genetics, or co-exposures. Thus, maternal blood Mn likely has considerably more error than teeth when used as a surrogate for fetal Mn levels. The decline in dentine Mn from the second trimester to birth, as we have shown here, is in contrast to the rise in maternal blood Mn during pregnancy (Takser et al. 2003). While the physiological mechanisms that result in this discrepancy between maternal and fetal Mn uptake during pregnancy are not completely understood, it is possible that mothers with high levels of Mn during pregnancy do not necessarily reflect higher fetal exposure; in fact, the converse might be true. For example, in a study that collected umbilical cord and maternal blood at delivery, Mn levels in cord blood increased linearly with increases in maternal blood to a level of 40 μg/L Mn in maternal blood, but then leveled off or even declined with increasing maternal Mn levels (Claus Henn et al. 2017).

Studies that have dosed animals with Mn alone have shown that higher exposure to Mn during early life, including the postnatal period, results in motor coordination dysfunction (Cordova et al. 2013), hyperactivity and spatial learning deficits (Kern et al. 2010). However, the picture emerging from experiments where animals are co-exposed to both Mn and Pb is more complex, revealing that mixture-exposed rats perform better on some tasks (e.g. learning and memory) but worse on others (e.g., swimming velocity) (Betharia and Maher 2012). As the epidemiologic literature on Mn and neurobehavior expands, and more nuanced analyses are conducted, more of this potential complexity in the Mn associations becomes apparent.

There are several limitations to our study. The relatively small sample size limited the statistical power to detect associations, particularly in the stratified analyses, and precluded us from examining analyses stratified jointly by Pb and sex. Our findings should be confirmed in a larger sample. The distributed lag-type approach we used models the correlation between visual motor ability score and Mn levels while assuming a linear relationship of the outcome with Mn at each time point. Prior evidence from studies using blood Mn as a biomarker suggests that Mn may follow a nonlinear or inverted U-shaped dose-response with neurodevelopment (Chung et al. 2015; Claus Henn et al. 2010; Yu et al. 2014). Therefore, there may be model misspecification if tooth Mn associations are nonlinear at some time points. Our initial analyses using GAMs with average Mn levels in four time windows suggest that most estimated relationships visually appear to be approximately linear. We cannot rule out the possibility of unmeasured confounding by secondhand smoke exposure or by other co-occurring neurotoxicants, such as arsenic. In secondary analyses, we explored models adjusted for secondhand smoke exposure (i.e., time spent in room with smoker on typical day during pregnancy) and overall patterns remained (data not shown). However, only limited data were available on this variable (40% of participants with available data). Effect modification by other unmeasured exposures is also possible, which could enhance or reduce the estimated Mn associations. Higher-order interactions were not assessed in this study but are plausible and will be examined in future analyses with a larger sample size. Last, because deciduous teeth begin developing only in the second trimester, we were unable to examine the effects of first trimester metal exposure on neurodevelopmental outcomes.

Despite these limitations, there are also several unique strengths to our study. This is among the first studies to characterize internal levels of prenatal and early postnatal metals over time in a nearly continuous manner. By linking levels of metals in deciduous teeth to developmental timing of exposure, we were able to capture distinct developmental windows that are more sensitive to the beneficial and/or harmful effects of metals. This biomarker has been validated against other Mn and Pb biomarkers (Arora et al. 2012; Arora et al. 2014; Austin et al. 2017). In addition to this highly innovative exposure measurement, we implemented a novel statistical technique to incorporate the high-resolution exposure data. We adjusted for many important confounders and imputed missing data to maximize our sample size in sensitivity analyses. Finally, the prospective study design combined with a retrospective exposure biomarker allowed us to examine neurodevelopmental outcomes in mid-childhood when assessments are more sensitive and reliably related to cognitive abilities than outcomes measured at younger ages (White et al. 2009).

5. CONCLUSIONS

Overall, in this study of Mexican children, we found discrete developmental windows when exposure to Mn was associated with visual-spatial abilities in childhood. The associations we observed were modest, but suggest that they may be dependent on the developmental age, sex, and level of concurrent exposure to Pb. This is the first study of metals and children’s neurodevelopment to characterize prenatal and early postnatal Mn levels in a nearly continuous manner. These results add to the growing evidence that the relationship of Mn exposure with neurodevelopment is complex and multifactorial.

Highlights.

• Tooth-matrix biomarkers were used to identify windows of susceptibility to Mn

• Dentine Mn levels were highest in the second trimester, then declined

• Prenatal tooth Mn was associated with better visual-spatial scores at low Pb levels

• Postnatal tooth Mn was associated with worse visual-spatial scores among boys

Abbreviations

Ca

Calcium

Mn

Manganese

Pb

Lead

WRAVMA

Wide Range Assessment of Visual Motor Abilities