Critical windows of susceptibility in the association between manganese and neurocognition in Italian adolescents living near ferro-manganese industry

Abstract

Introduction:

Understanding the neurodevelopmental effects of manganese (Mn) is complicated due to its essentiality for growth and development. While evidence exists for the harmful effects of excess Mn, pediatric epidemiologic studies have observed inconsistent associations between Mn and child cognition.

Objective:

We sought to estimate prospective associations between Mn measured in three different early-life time windows with adolescent cognition using deciduous teeth biomarkers.

Methods:

Deciduous teeth were collected from 195 participants (ages 10 – 14 years) of the Public Health Impact of Manganese Exposure (PHIME) study in Brescia, Italy. Measurements of tooth Mn represented prenatal (~14 weeks gestation – birth), early postnatal (birth – 1.5 years) and childhood (~1.5 – 6 years) time windows. Neuropsychologists administered the Wechsler Intelligence Scale for Children, 3rd edition (WISC-III), to obtain composite IQ and subtest scores. Associations between tooth Mn at each time window and adolescent WISC-III scores were estimated using multivariable linear regression. We tested differences in associations between Mn and outcomes across time windows using multiple informant models. Sex-specific associations were explored in stratified models.

Results:

Adjusted associations between tooth Mn and composite IQ scores were positive in the prenatal period and negative in the childhood period. Associations were strongest for subtest scores that reflect working memory, problem solving, visuospatial ability and attention: prenatal Mn was positively associated with Digits backward [SD change in score per interquartile range increase in Mn: β = 0.20 (95% CI: 0.02, 0.38)] and Block design [β = 0.21 (0.01, 0.41)] and early postnatal Mn was positively associated with Digits forward [β = 0.24 (0.09, 0.40)], while childhood Mn was negatively associated with Coding [β = −0.14 (−0.28, −0.001)]. Sex-stratified analyses suggested different Mn-cognition associations for boys and girls and was also dependent on the time window of exposure.

Conclusion:

Our results suggest that exposure timing is critical when evaluating Mn associations between Mn and cognition. Higher prenatal Mn was beneficial for adolescent cognition; however, these beneficial associations shifted towards harmful effects in later time windows. Cognitive domains most sensitive to Mn across time windows included visuospatial ability, working memory, attention and problem-solving.

1. Introduction

Manganese (Mn) is an essential nutrient required for growth and neurodevelopment, but in excess is a potent neurotoxicant (Balachandran et al. 2020), which has created considerable uncertainty in understanding the role of environmental exposures to manganese. The situation is further complicated because exposure timing may influence effects, given that central nervous system nutritional needs for Mn may vary by age. Among its many critical roles in cellular processes, Mn is a cofactor for superoxide dismutase (SOD), an important enzyme for protecting against oxidative stress. On the other hand, in excess, Mn is a transition element that can catalyze redox reactions and may contribute to oxidative stress that can affect central nervous system structures or alter their developmental trajectory (Balachandran et al. 2020; Chen et al. 2015). While the developing infant appears to experience an increased nutritional requirement for Mn during the prenatal and early postnatal period (Mistry and Williams 2011), pregnancy and age specific ideal intake levels (exposure) for Mn have not been identified, and it is unclear at what levels Mn shifts from beneficial to toxic (Aschner and Aschner 2005; Erikson et al. 2007; National Academy of Science 2001).

Environmental Mn overexposure can occur through ingestion of contaminated drinking water or inhalation of air pollution containing Mn. Sources include drinking water with naturally occurring or anthropogenic Mn (Bouchard et al. 2018; Rodrigues et al. 2016). Airborne emissions are another source, from combusted gasoline enhanced with methylcyclopentadienyl manganese tricarbonyl (MMT) (WHO 2011), and from Mn-related industrial emissions including mining (García-Chimalpopoca et al. 2019), ferroalloy and steel processing (Bauer et al. 2020a; Haynes et al. 2015; Rodrigues et al. 2018), and agriculture using Mn-containing fungicides (Mora et al. 2015; van Wendel de Joode et al. 2016). From the in utero period through adolescence, children may be uniquely susceptible to Mn overexposure, because maternal Mn absorption is up-regulated during pregnancy (Erikson et al. 2007), infant homeostatic mechanisms of Mn absorption and excretion are underdeveloped (Aschner and Aschner 2005), and pubertal changes involve Mn-targeted systems (Kern et al. 2010; Wahlstrom et al. 2010b). Mn neurotoxicity in adults is well characterized in occupational studies (Bowler et al. 2006; Mergler et al. 1994; Racette et al. 2012). Yet, while there is heightened potential for Mn overexposure in pediatric populations, evidence for neurodevelopmental toxicity remains inconsistent, likely due to variability in timing of exposure, exposure metric and route, biological sex of child, neurobehavioral outcome assessed, age at assessment and co-exposures to other neurotoxicants (Bauer et al. 2020b).

Prior work suggests that Mn neurodevelopmental toxicity depends on timing of exposure (Bauer et al. 2017; Claus Henn et al. 2018; Gunier et al. 2015; Horton et al. 2018; Mora et al. 2018). Understanding critical windows of susceptibility is particularly important for pediatric neurobehavioral outcomes because brain development is a dynamic and complex process of tightly choreographed events, extending from the in utero through adolescent periods (Blakemore and Choudhury 2006; Sowell et al. 2001). Neurological processes transpire throughout this period, in which neurogenesis, neuronal migration, and specialization occur in utero, the bulk of myelination arises in the first few postnatal years of life, and refinement continues via synaptic pruning, myelination, neuronal transmission and neural circuitry restructuring in later childhood and adolescence (Blakemore and Choudhury 2006; Giedd et al. 1999; Suzuki 2007; Wahlstrom et al. 2010a). The time at which toxicant exposure occurs may interfere with specific neurodevelopmental cascades and could determine the extent of damage and the ensuing phenotype (Rice and Barone Jr. 2000). However, few studies have measured Mn over these critical periods of neurodevelopment.

Research addressing associations of Mn measured across multiple time periods and children’s cognition has used several different biomarkers of exposure including maternal blood (Claus Henn et al. 2017; Mora et al. 2018; Takser et al. 2003), cord blood (Claus Henn et al. 2017; Mora et al. 2018; Takser et al. 2003; Zhou et al. 2020), and teeth (Bauer et al. 2017; Claus Henn et al. 2018; Gunier et al. 2015; Mora et al. 2015). Because Mn (II) is somewhat of an analog to calcium, teeth incorporate Mn as they mineralize daily growth rings, allowing for the identification of exposure timing and intensity in validated exposure assessment protocols (Arora et al. 2011, 2012, 2014; Hare et al. 2011). Tooth matrix formation begins during the second trimester of gestation and continues until the tooth is shed (~6-12 years) (Nanci 2018). Thus, researchers can collect naturally shed teeth and perform neurobehavioral assessment in tandem, without concerns for temporality that preclude understanding causal associations in cross-sectional study designs.

The objective of this study was to evaluate associations of tooth Mn measured in three early life periods and cognition among adolescents living near varied ferro-manganese industry in northern Italy who participated in the Public Health Impact of Manganese Exposure (PHIME) study. We measured Mn in dentine of deciduous teeth to represent three distinct time frames, or potential exposure periods, (prenatal, early postnatal, childhood) and estimated associations of each time window with neurocognitive outcomes from the Wechsler Intelligence Scale for Children (Version III, Italian language). Given previous findings from the PHIME cohort of sex differences in Mn associations with motor function (Chiu et al. 2017), visuospatial learning and memory (Bauer et al. 2017) and behavior (Broberg et al. 2019), we also explored sex-specific associations.

2. Methods:

2.1. Study participants

Participants for this analysis were part of the ongoing PHIME study. A detailed description of the study design was previously published (Lucchini et al. 2012). Briefly, 720 children ages 10-14 years were recruited from three demographically similar but geographically distinct sites in the province of Brescia, Italy, a region with varied ferroalloy activity: Bagnolo Mella (BM), an area with currently active ferroalloy industry since 1974; Valcamonica (VC), an area with historical ferroalloy production for over a century that ended in 2001; and Garda Lake (GL), a tourist region with no history of ferroalloy activity (for maps see Butler et al. 2018; Lucchini et al. 2007). Whereas BM is in the plains area of the region, VC is a pre-Alp valley with naturally higher Mn levels in soil. Our data have shown higher Mn soil levels in VC, compared to BM and GL (Borgese et al.; Ferri et al. 2015), yet higher Mn levels in airborne particles (Rosa et al. 2016) and deposited dust in BM vs. VC and GL (Lucas et al. 2015).

To be eligible for the study, all participants must have been (1) born to families living in the designated study area since the 1970s; (2) living in the study area since birth, and (3) ages 10-14 years at enrollment. Participants were excluded if they (1) had any diagnosed metabolic, neurological, hepatic or endocrine diseases, or clinically diagnosed hand/finger motor deficits; (2) took any prescription psychoactive drugs; (3) had clinically diagnosed behavioral manifestations of cognitive impairment, or (4) had inadequately corrected visual deficits.

A total of 720 participants were enrolled into the study, 312 participants in the first phase (2007-2010) and 408 in the second phase (2010-2014). The two phases reflect two waves of funding but were conducted by the same researchers using identical questionnaires and study protocols. The second phase added a third site (BM) in addition to VC and GC, as well as measurement of the Home Observation Measurement of the Environment (HOME) Short Form, a measure of cognitive stimulation and emotional support in the home (National Longitudinal Surveys 1979). In 2013 we obtained supplemental funding to collect and measure manganese in deciduous teeth. Teeth were available from a subset of participants (n=195), which is the final sample for this analysis.

Eligible participants received a detailed description of the study procedures before consenting to participate. The Institutional Review Boards at the Ethical Committee of Brescia, the Icahn School of Medicine at Mount Sinai, and the University of California, Santa Cruz approved all PHIME study protocols.

2.2. Tooth collection and Mn measurement

Manganese levels were measured in one naturally shed tooth (non-carious) per participant. Collected teeth included incisors, canines and premolars. Analytical methods for measuring Mn in teeth have been validated and previously described in detail (Arora et al. 2011, 2012; Gunier et al. 2014). Briefly, teeth were sectioned on a vertical plane to expose the full cross-section of dentine. We used microscopy to identify the neonatal line (NL), a histological feature formed at birth (Sabel et al. 2008). Using the NL as a spatial and temporal reference point, we then measured Mn concentrations in areas of the tooth representing specific developmental windows using laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) (Arora et al. 2011, 2012). Primary dentine formed in the prenatal period (second trimester to birth) was sampled to obtain prenatal Mn exposure information; primary dentine formed between birth and 1.5 years of age was sampled to obtain early postnatal Mn exposure information; and secondary dentine, formed between tooth root completion (~1.5 years) and ~6 years of age, was sampled to obtain childhood Mn levels. Forty measurements were taken from the tooth matrix across exposure periods and used to calculate an area under the curve (AUC) for each period. Mn concentrations were normalized to calcium (Ca) concentrations (⁵⁵Mn:⁴³Ca ratio) to account for within- and between-subject variation in tooth mineral density. The limit of detection was 0.03 AUC Mn⁵⁵:Ca⁴³ x 10⁴. Values below the detection limit were assigned half of the lowest value among the samples above the detection limit (n=2 for early postnatal; assigned a value of 0.015 AUC Mn⁵⁵:Ca⁴³ x 10⁴). Laboratory technicians were blinded to participants’ neurobehavioral and demographic information.

2.3. Neurobehavioral assessment

Two trained neuropsychologists administered a neurobehavioral assessment when participants were between 10 and 14 years of age, including the Italian language version Wechsler Intelligence Scale for Children - Third Edition (WISC-III) (Wechsler et al. 2006). Twelve subtests from the WISC-III, plus Digit Span, a WISC-III supplemental test, assessed performance in a variety of neurobehavioral domains (Table 1). Subtest scores were combined into three composite IQ measures as follows: 1) Verbal IQ (VIQ), comprising scores from the Arithmetic, Comprehension, Information, Similarities, and Vocabulary subtests, provides an overall assessment of language-based skills; (2) Performance IQ (PIQ), comprising scores from the Block Design, Coding, Object Assembly, Picture Arrangement, and Picture Completion subtests, assesses primarily visual-motor (i.e., non-verbal) performance; and (3) Full-scale IQ (FSIQ), which encompasses all 10 subtests, summarizing VIQ and PIQ. The Digit Span composite score is comprised of Digits Forward and Digits Backward. Age-adjusted standard scores from the Italian normative WISC-III population were used in analyses.

Table 1.

Description of WISC-III^a subtest tasks and associated neurobehavioral domains

WISC-III Subtest	Neurobehavioral Domains	Subtest Task
Arithmeticb	working memory, executive function, attention	Examinee solves arithmetic problems verbally
Comprehension	knowledge, reasoning and executive function	Examinee answers reasoning questions using social knowledge, planning and other abstract skills
Information	verbal knowledge	Examinee answers questions about information learned in school or from life experience
Similarities	executive, verbal, abstract logical thinking and reasoning	Measures ability to relate two words abstractly
Vocabulary	word knowledge	Examinee defines vocabulary words
Block designc	visuospatial organization	Using red and white blocks, examinee replicates designs presented on cards
Coding	motor, strategy, working memory, non-verbal learning	Using a number-symbol code, the examinee draws in symbols that match numbers on a visual grid
Object assembly	visuospatial organization	This task requires assembly of puzzle pieces presented in unorganized form
Picture arrangement	visual sequencing, working memory, executive function	Arrange pictures from beginning to end to tell a meaningful story
Picture completion	visuospatial, attention	Examinee identifies missing parts of pictures
Digits backwardd	working memory, executive function	Examinee hears orally administered sequences of numbers and verbally repeats the sequences backward
Digits forward	attention	Examinee verbally repeats back sequences of numbers presented orally of increasing lengths

^a Wechsler 2006

^bVerbal IQ subtests: Arithmetic, Comprehension, Information, Similarities and Vocabulary

^cPerformance IQ subtests: Block design, Coding, Object assembly, Picture arrangement, Picture completion

^dDigit Span subtests: Digits backward and Digits forward

2.4. Covariates

Information on sociodemographic factors was collected using a standardized questionnaire administered by trained researchers at the time of neurobehavioral assessment. Information included age of participant, birth order (first, second, third or higher), area of residence (BM, VC, GL), self-reported sex of participant, and participant alcohol consumption (yes, no). Socioeconomic status (SES) was categorized into low, medium, or high based on methodology developed in Italy that combines parent education and occupation (Cesana et al. 1995; Chiu et al. 2017; Lucchini et al. 2012b). To obtain a measure of cognitive stimulation and emotional support in the home, we administered the HOME Short Form (National Longitudinal Surveys 1979). Natural tooth wear reduces the most outer layers of dentine and which can impact the accuracy of the prenatal signatures (Nanci 2018). To account for this, we assessed the quality, or wear, of each tooth sample: tooth attrition was rated on a scale of 0 to 3 in which 0 denoted no tooth loss from wear, 1 is less than one-third tooth loss, 2 is more than one-third but less than two-thirds tooth loss, 3 denotes more than two-thirds tooth loss.

Blood lead (Pb) was measured in venous whole blood collected within two weeks of neurobehavioral assessment (when participants were 10-14 years old). Blood samples (4 mL) were collected using a 19-gauge butterfly catheter into a Lithium Heparin Sarstedt Monovette Vacutainer. Measurement of blood Pb concentrations was performed at the University of California, Santa Cruz, using previously described methods (Lucchini et al. 2012b). Hemoglobin (g/dL) was measured in the same blood samples collected for Pb analysis and was considered an indicator of chronic anemia (low iron status), a potential confounder in this analysis.

2.5. Statistical analysis

Univariate and bivariate analyses were performed for each variable. Tooth Mn levels were natural log (ln) transformed to reduce skewness and therefore limit the impact of outliers. Pairwise correlations were calculated using Spearman correlation coefficients for prenatal, postnatal and childhood Mn. We estimated associations between tooth Mn and WISC outcomes including 3 composite scores (FSIQ, VIQ, and PIQ) and 12 subtest scores (Information, Vocabulary, Similarities, Arithmetic, Comprehension, Coding, Block design, Object assembly, Picture arrangement, Picture completion, Digits forward and Digits backward). Each model included prenatal, early postnatal and childhood Mn levels, in order to mutually adjust for one another. Subtest scores were standardized (i.e., z-scored) to compare measures of effect across subtests.

Covariates were chosen a priori based on prior literature (Bauer et al. 2017; Claus Henn et al. 2010) and using a directed acyclic graph (DAG). Covariates included in final models were sex (female vs. male); socioeconomic status (low, medium, vs. high); HOME score (continuous); ln-transformed blood Pb (continuous), and tooth attrition (less than one-third, more than one-third, vs. none). Because Mn is both an essential nutrient and a potential toxicant at elevated exposures, it may be nonlinearly associated with neurobehavioral outcomes. We evaluated potential nonlinearity by visual inspection of Mn modeled as a smoothed term in generalized additive models (GAMs) with penalized splines (knots=4). Based on adjusted GAMs, Mn levels in all three time windows appeared to be linearly associated with WISC outcomes. We subsequently used multivariable linear regression adjusting for confounders to estimate associations between each time window of Mn with WISC outcomes.

Given prior evidence of sex-specific Mn effects (Bauer et al. 2017; Claus Henn et al. 2018; Mora et al. 2015), we explored whether associations between tooth Mn and cognition differed between males and females by 1) stratifying models by sex and 2) including an interaction term for each Mn window (sex*Mn) in regression models, for a total of 3 interaction terms in each model. Interactions were considered significant at p < 0.1.

Additionally, we used multiple informant models to test whether the association for Mn and each WISC outcome differed between time windows (e.g., prenatal Mn vs. early life vs. childhood Mn) using adapted code from Sanchez et al., 2011 (Hu et al. 2018; Sánchez et al. 2011). For each WISC outcome, we fit a generalized estimating equation (GEE) model using empirical standard errors and tested whether the associations between Mn and the WISC outcome differed statistically from one another (see supplemental material for multiple informant code).

Multiple imputation

Some participants had missing data on covariates including HOME score (43.2% missing). Multiple imputation was used to impute missing values using chained equations (Buuren and Groothuis-Oudshoorn 2011). We imputed missing values for all participants (n=709) but restricted our analysis to participants with measured IQ and tooth Mn levels (n=195) (White et al. 2011). We included variables in the imputation process that were thought to be related to the missing covariates, including WISC outcomes, metals concentrations in biological as well as environmental samples (Bauer et al. 2020a; Lucchini et al. 2012a), and potential confounders (see Table S1 for list of variables used in imputation). Data were assumed to be missing at random, i.e., missingness did not depend on unobserved data. We generated 40 imputed datasets with the MICE package in R (Buuren and Groothuis-Oudshoorn 2011). For multivariable linear regression models, pooled estimates and confidence intervals were obtained by combining information from the 40 mean and variance estimates using Rubin’s method (Rubin 1987). For multiple informant models, we used proc mianalyze to combine the 40 mean and variance estimates including a test function to compare whether estimates differed between Mn levels at the three different time windows (see supplemental material for code).

Statistical analyses were conducted using R version 3.5.1 (The R Foundation for Statistical Computing, www.r-project.org) and SAS 9.4 for multiple informant models.

3. Results

A total of 195 participants had available data for adolescent WISC-III outcomes and tooth Mn levels. Among the participants, about half were female (53%) and were of medium socioeconomic status (56%) (Table 2). On average, participants were 12 years of age (SD=0.9) and had a HOME score of 6.6 (SD=1.5; range: 1-9). Median (IQR) levels of tooth dentine Mn measured in prenatal, early postnatal and childhood periods, respectively, were: 0.43 (0.19); 0.13 (0.07) and 0.0007 (0.0004) AUC Mn⁵⁵:Ca⁴³ x 10⁴. Tooth Mn levels did not correlate strongly across time windows (correlation between prenatal and early postnatal Mn: rho= 0.29; prenatal and childhood Mn: rho= −0.01; postnatal and childhood Mn: rho= 0.15), which was expected given that blood Mn has shown to decrease over time as well. Most characteristics did not differ by sex, including tooth Mn levels. However, compared to boys, girls were from families with lower socioeconomic status (e.g., 28.6% of girls were low SES vs. 14.5% of boys) and had less tooth attrition (e.g., 3.8% of girls had level 3 attrition vs. 11.1% of boys). Descriptive characteristics were similar for original and imputed data (Table S2).

Table 2.

Descriptive statistics for PHIME participants included in the analysis^a

	All N=195	Girls N=105	Boys N=90
Demographics	N (percent) or Mean (SD)
Age (years)	12.0 (0.9)	12.1 (0.9)	11.9 (0.9)
Sex
Female	105 (53.8%)	--	--
Male	90 (46.2%)	--	--
Socioeconomic statusb
High	42 (21.5%)	22 (21%)	20 (22.2%)
Low	43 (22.1%)	30 (28.6%)	13 (14.5%)
Medium	110 (56.4%)	53 (50.4%)	57 (63.3%)
Site
Bagnolo Mella	82 (42.1%)	45 (42.8%)	37 (41.1%)
Garda Lake	48 (24.6%)	30 (28.6%)	18 (20.0%)
Valcamonica	65 (33.3%)	30 (28.6%)	35 (38.9%)
HOME scorec	6.6 (1.5)	6.6 (1.6)	6.7 (1.4)
Bio markers
Median (IQR) or N (percent)
Prenatal tooth Mn (AUC Mn55:Ca43)d	0.43 (0.19)	0.43 (0.19)	0.46 (0.24)
Postnatal tooth Mn (AUC Mn55:Ca43)	0.13 (0.07)	0.13 (0.07)	0.13 (0.08)
Childhood tooth Mn (AUC Mn55:Ca43)	0.0007 (0.0004)	0.0007 (0.0004)	0.0007 (0.0004)
Blood Pb (μg/dL)	1.3 (0.9)	1.2 (0.8)	1.4 (0.8)
Level of tooth loss due to attritione
None	109 (55.9%)	66 (62.8%)	43 (47.8%)
Less than one third	72 (36.9%)	35 (33.3%)	37 (41.1%)
More than one third	14 (7.2%)	4 (3.8%)	10 (11.1%)

^aDescriptive statistics averaged across 40 multiple imputed datasets

^bSocioeconomic status (SES) was categorized into low, medium, or high based on methodology developed in Italy that combines parent education and occupation (Cesana et al. 1995; Chiu et al. 2017; Lucchini et al. 2012b).

^cHome observation measure of the environment, HOME

^dArea under the curve, AUC

^eTooth attrition is a measure of tooth loss from wear

Mn-neurobehavior associations for each time window were generally small in magnitude and many 95% confidence intervals contained the null (Figure 1 and Table S3). However, effect estimates in the prenatal period were consistently positive, suggesting beneficial effects of prenatal Mn. An interquartile range (IQR) increase in prenatal Mn was associated with increases of ~1-2 points on composite IQ scores [FSIQ: β = 1.65 (95% CI: −0.78, 4.07); VIQ: β = 0.89 (−1.59, 3.36); PIQ: β = 2.01 (−0.56, 4.59)] (Figure 1; z-scored IQ estimates in figure only). The strongest associations for the prenatal period were estimated for Digits backward and Block design subtests, where an IQR increase in prenatal Mn was associated with a 0.2 SD increase in subtest score [Digits backward: β = 0.20 (95% CI: 0.02, 0.38); Block design: β = 0.21 (0.01, 0.41)] (Figure 1). Using multiple informant models, associations were different across Mn time windows for Digits backward (p=0.01; Table S5). Based on exploratory sex-stratified analyses, prenatal Mn was generally more beneficial among boys than girls [e.g., Picture completion, boys: β = 0.39 (0.09, 0.69) vs. girls: β = −0.29 (−0.58, 0.001); p-value for interaction=0.007] (Figure 2 and Table S4).

Figure 1. Adjusted associations¹ between prenatal, postnatal, and childhood tooth manganese levels (⁵⁵Mn:⁴³Ca ratio) and Wechsler Intelligence Scale for Children (WISC-III) outcomes among 195 early adolescents (ages 10-14) enrolled in the PHIME cohort in Northern Italy.

¹ All models use natural log transformed tooth Mn and z-scored age adjusted WISC-III outcomes. All models were adjusted for sex, log-transformed blood lead, socioeconomic (SES) index, HOME score and tooth attrition, and mutually adjusted for log transformed Mn in other time windows. Dots represent effect estimates for interquartile range increase in Mn; dashed lines represent 95% confidence intervals. Shading indicates (from top to bottom): (1) composite scores, (2) subtests that comprise VIQ composite score, (3) subtests that comprise PIQ composite score, (4) Digits Span subtests.

Figure 2. Sex-stratified adjusted associations¹ of interquartile range increase in prenatal, postnatal, and childhood tooth manganese levels (⁵⁵Mn:⁴³Ca ratio) and Wechsler Intelligence Scale for Children outcomes.

¹ All models use ln-transformed Mn and z-scored outcomes. All models were adjusted for ln-transformed blood lead, SES index, HOME score and tooth attrition. All models mutually adjusted for all other Mn time windows. Points represent effect estimates; dotted lines represent 95% confidence intervals. Males (purple) and females (green).

*p-value for sex-Mn interaction term <0.10: Picture completion (p=0.007); Arithmetic (p=0.03); Vocabulary (p=0.08); Digits backward (p=0.08).

Associations between early postnatal Mn (birth to 1.5 years of age) and adolescent WISC-III outcomes were generally null and most estimates were less positive than in the prenatal period. An IQR increase in early postnatal Mn was associated with slight increases in composite IQ scores [FSIQ: β = 0.98 (95% CI: −1.47, 3.42); VIQ: β = 0.78 (−1.71, 3.27); PIQ: β = 0.93 (−1.66, 3.52)]. The association between early postnatal Mn and Digits forward was more beneficial than the association with prenatal Mn [early postnatal: β = 0.24 (0.09, 0.40) vs. prenatal: β = 0.04 (−0.11, 0.19); Figure 1 and Table S3] and associations between Mn time windows were statistically different from one another using multiple informant models (p=0.01; Table S5). There were differences in estimates of effect between boys and girls, where, in contrast to prenatal Mn, postnatal Mn was more beneficial among girls than boys [e.g., Arithmetic, girls: β = 0.32 (0.02, 0.62) vs. boys: β = −0.14 (−0.39, 0.10); p-value for interaction=0.03; Vocabulary, girls: β = 0.23 (−0.07, 0.54) vs. boys: β = −0.19 (−0.46, 0.08); p-int=0.08) (Figure 2 and Table S4)].

For the childhood period (~1.5 to 6 years of age), associations between tooth Mn levels and WISC-III outcomes measured in adolescence were mostly null, although many estimates were negative. Childhood Mn was not associated with composite IQ scores [per IQR increase, FSIQ: β = −0.64 (95% CI: −2.69, 1.42); VIQ: β = −0.13 (−2.23, 1.97); PIQ: β = −1.07 (−3.25, 1.10)]. Subtests with the largest effect estimates were Coding [β = −0.14 (−0.28, −0.001)] and Block design [β = −0.14 (−0.31, 0.03)]. In sex-stratified analyses, most associations were similar for boys and girls, except Digits backward, [girls: β = −0.12 (−0.34, 0.09) vs. boys: β = 0.26 (0.04, 0.48); p-int=0.008] (Table 2), where childhood Mn appeared more harmful among girls.

4. Discussion

Although manganese is crucial for healthy neurodevelopment, excess levels are neurotoxic and many determinants of age appropriate maintenance dose, homeostatic regulation, deficiency and neurotoxicity remain unclear (Balachandran et al. 2020). Pediatric epidemiologic studies have reported beneficial, null and adverse associations between Mn and neurobehavior, with varying exposure metrics, exposure routes, levels, timing and co-exposures (Bauer et al. 2020b; Coetzee et al. 2016; Lucchini et al. 2017). In this study of Italian adolescents living near ferro-manganese industry, we observed patterns of associations between tooth Mn and WISC-III subtest performance during adolescence: our results suggest a subtle shift over time from the beneficial role Mn plays as an essential nutrient required for brain development during the prenatal period to a more detrimental role as a neurotoxicant when exposure occurs in childhood. Overall, the WISC-III subtests more strongly associated with Mn across time windows measure visuospatial ability, working memory, problem solving and attention.

Findings of Mn-sensitivity in these cognitive domains support the notion that Mn impacts dopaminergic pathways that subserve these cognitive functions. Dopaminergic neurons in the central nervous system including the frontal cortex and extrapyramidal motor regions are known to be sensitive to Mn homeostasis (Aschner et al. 2007; Dorman 2006). Dopamine-rich regions in the extrapyramidal motor region are highly involved in executive function, memory, learning, mood and visuospatial tasks (Sotomayor-Zarate et al. 2014). Congruently, animal neuroimaging studies report Mn-associated changes in the frontal cortex and basal ganglia, and to brain structures related to working memory including the striatum (caudate nucleus), and parietal and frontal cortices (Guilarte 2013). Pediatric imaging research suggests dopaminergic system involvement as well: basal ganglia morphometric changes were found in children exposed to high Mn through drinking water, compared to children with low exposure (Lao et al. 2017). In a neuroimaging study of 6- to 7-year-old Mexican children, maternal blood Mn was associated with reduced intrinsic functional connectivity (iFC) between the basal ganglia and prefrontal cortex (de Water et al. 2018). Additionally, postnatal tooth Mn was associated with increased iFC within cognitive control brain areas in a subset of adolescent study participants (de Water et al. 2018). However, damage to dopaminergic environments is likely not the only mechanism at play involving neurotransmission; other areas of developing interest for Mn neurotoxicity include glutamatergic and GABAergic synapse interference (Burton and Guilarte 2009).

Our results of positive associations of prenatal Mn and neurobehavioral outcomes support higher levels of Mn serving as a beneficial nutrient during this developmental window. In particular, Mn was beneficially associated with subtests that measure working memory and visuospatial abilities (Digits backward, Block design, Object assembly). The neurophysiologic requirements for Mn may be greatest during the prenatal period, in which Mn is actively transported across the placenta to aid in fetal growth and development (Peres et al. 2016). The beneficial trend of Mn in the prenatal period has been observed in other longitudinal studies that used deciduous teeth as a Mn biomarker (Claus Henn et al. 2018; Horton et al. 2018; Mora et al. 2015). In 6- to 16-year-old children residing in Mexico City, prenatal Mn was beneficial for total visual motor abilities and visuospatial ability (Claus Henn et al. 2018) as well as overall behavioral problems and internalizing and externalizing symptoms assessed using the Behavioral Assessment System for Children, second edition (BASC-2) (Horton et al. 2018). In Mexican-American children living near areas of Mn-containing fungicide application in California, prenatal tooth Mn was beneficially associated with cognition, memory and motor abilities at 7 years among participants with low Pb exposure (Mora et al. 2015). Consistent with our findings, pediatric epidemiologic studies using Mn biomarkers of prenatal exposure other than teeth have also reported beneficial (Chung et al. 2015; Freire et al. 2018) and null associations (Kupsco et al. 2020; Zhou et al. 2020) with neurobehavioral outcomes. However, unlike our findings, several prospective studies have reported inverse associations between fetal and maternal measures of prenatal Mn and children’s neurodevelopment, with variation in the direction and shape of the association depending on modifying factors (Claus Henn et al. 2017; Lin et al. 2013; Muñoz-Rocha et al. 2018; Takser et al. 2003; Valeri et al. 2017). Potential differences could be related to overall exposure levels to Mn or differences in the Mn biomarker used. We propose that teeth may be a better reflection of body burden compared to blood or urine because Mn in blood is under regulatory control into a relatively small range of values, and urine is not the main route of Mn excretion (ATSDR 2012). Teeth in contrast are not remodeled after the mineral matrix is laid and likely best reflect the variability in dose over time (Arora et al. 2012). Given the variability in exposure biomarkers, neurobehavioral outcomes, and modifying factors across studies, further investigation of the association of prenatal Mn on children’s neurobehavioral function is warranted.

Our results for Mn and neurobehavioral scores were mostly null for the early postnatal period. The early postnatal period is a unique and highly variable time frame for exposure opportunities, as physiologic characteristics and behavioral patterns vary across early infancy (Cohen Hubal et al. 2000; Ferguson et al. 2017). During this first year of life, dietary patterns, hand to mouth behaviors (e.g., exploration of objects or alleviation of teething discomfort) as well as mobility (crawling or walking) vary between infants as well as within infants across time and can influence Mn exposure (Cohen Hubal et al. 2000; Zota et al. 2016). A possible explanation for our null findings in the postnatal period may be that the exposure window we measured (birth-1.5 years) is too wide to capture important changes in Mn levels at specific, and possibly sensitive, neurodevelopmental time windows. Recent research examining more discrete time frames during the early postnatal period reported that postnatal Mn measured during 6 to 12 months was associated with reduced visuospatial performance, but reported null associations between averaged early postnatal Mn (0-1 year) and total WRAVMA score (Claus Henn et al. 2018). Another study using data from the same cohort found a discrete time window during the fourth postnatal month where Mn was associated with higher internalizing and anxiety scores (Horton et al. 2018).

Another potential explanation for our null findings during the early postnatal period could be that the specific outcomes measured or the age at time of neurobehavioral assessment is not sensitive to Mn homeostasis in the early postnatal period. Two studies using different outcomes and ages of participants but a similar tooth exposure marker reported significant associations between early postnatal Mn and neurodevelopment and had comparable cumulative average postnatal Mn levels to our study [our study: GM (GSD) = 0.13 (1.56) Mn⁵⁵:Ca⁴³ AUC X 10⁴ vs. Mora et al. 2015: GM (GSD) = 0.14 (2.47)] (Gunier et al. 2015; Mora et al. 2015). Mora et al. 2015 reported adverse associations between early postnatal Mn measured in teeth and outcomes from the BASC-2 assessed at 7 and 10.5 years of age (internalizing and externalizing behaviors, symptoms of hyperactivity), but beneficial associations between early postnatal Mn and memory assessed at 9 and 10.5 years of age (Mora et al. 2015). Gunier et al. 2015 reported early postnatal Mn was associated with reductions in the Mental Developmental Index (MDI) scores of the Bayley Scales of Infant Development second edition (BSID-II) measured at 6 and 12 months, but not for scales assessed at 24 months (Gunier et al. 2015). Similarly, a nonlinear association was observed between postnatal blood Mn measured at 12 months and MDI assessed at 12 months in Mexican infants (Claus Henn et al. 2010). These findings were specific to the exposure period measured (association between 12-month Mn and MDI but null associations for 24-month Mn) and time when the outcome was assessed (association with 12-month MDI but null associations with 18- and 24-month MDI) (Claus Henn et al. 2010). In summary, our null findings in the early postnatal period could be due in part to either the lack of sensitivity to early postnatal Mn homeostasis for these specific outcomes or for the time period of outcome assessment, or that this exposure window (birth to 1 year) is too wide to detect an association.

Associations between childhood tooth Mn and WISC scores were mostly negative, including associations with IQ composite scores, although confidence intervals contained the null. In particular, Mn was adversely associated with subtests that measure visual-motor and visuospatial ability (Coding and Block design, respectively), suggesting that the childhood exposure window may be susceptible to Mn neurotoxicity. Other recent studies investigating tooth Mn and child cognition only examined the prenatal and early postnatal periods. However, several prior cross-sectional studies have reported inverse associations between childhood Mn levels and cognition in children living near industrial Mn emissions. In 7- to 12-year-old children, hair Mn was inversely associated with IQ (Haynes et al. 2015, 2018; Menezes-Filho et al. 2011; Riojas-Rodríguez et al. 2010), verbal memory and learning (Carvalho et al. 2018; García-Chimalpopoca et al. 2019; Torres-Agustín et al. 2013), aspects of visuospatial abilities (Carvalho et al. 2014; Hernández-Bonilla et al. 2016), and working memory (Carvalho et al. 2014; Haynes et al. 2018). The childhood period may mark a time frame when increased Mn may be more harmful than beneficial. As infants age, Mn blood concentrations decrease (Claus Henn et al. 2012; Yoon et al. 2011), likely due to reduced Mn requirements, changes in diet and Mn homeostatic mechanisms that regulate Mn absorption and excretion (Yoon et al. 2011). This research begins to fill a gap in research investigating prenatal, early life and childhood Mn together in association with adolescent neurocognition.

Our finding that associations for WISC-III subtests were stronger than for WISC-III composite scores is consistent with some prior work. In Taiwanese children, inverse associations between cord blood Mn and 2-year scores on the Comprehensive Developmental Inventory for Infants and Toddlers (CDIIT) were strong for cognitive and language scores separately, but not for combined scores (Lin et al. 2013). Inverse associations were reported between cord blood Mn and subscales of attention and non-verbal memory, but not for composite scores of the McCarthy Scales in 3-year-old French children (Takser et al. 2003). Not all studies, however, report associations with subtest scores in addition to composite scores. Taken together with our results, these findings support the need for future studies to evaluate both composite scores and scores from domain-specific tasks to better understand the context of Mn developmental neurotoxicity.

In exploratory sex-stratified analyses, some associations varied by timing of exposure and cognitive domain assessed. Prenatal Mn was beneficially associated with visuospatial abilities (Picture completion) and working memory (Digits backward) among boys, but weakly inversely associated among girls. Consistent with this finding, prenatal tooth Mn was associated with worse performance on the virtual radial arm maze, a complex task of visuospatial learning and memory among girls only in a subset of the current study (Bauer et al. 2017). Along the same lines, among boys, but not girls, higher prenatal Mn was associated with improved visual memory performance (NEPSY-II Memory design) in a study of Californian children living near areas of Mn fungicide application (Mora et al. 2015). Sex differences were also weakly observed for the postnatal period of Mn exposure in our study. Postnatal Mn was associated with improved scores on working memory and executive function (Arithmetic) and verbal knowledge (Vocabulary) among girls, but weakly inversely associated among boys. Conversely, Mora et. al reported postnatal tooth Mn was associated with improved verbal comprehension among Mexican-American boys only (Mora et al. 2015). For Mn measured during childhood (age 1-6 years), associations with working memory (Digits backward) were beneficial among boys and weakly harmful among girls. Additionally, marginal adverse associations were found between childhood tooth Mn and visuospatial abilities (Block design) among girls only. One study of 7- to 11-year-old Mexican children living near Mn mining industry reported that hair Mn was more strongly inversely associated with visuoperception and visual memory in girls (Hernández-Bonilla et al. 2016).

Several limitations exist for this study. The sample size was relatively small and therefore has limited statistical power to detect associations in sex-stratified analyses. For this reason, we presented these analyses as exploratory. We have limited information on the prenatal period; there may be unmeasured confounding or effect modification by maternal characteristics such as iron status, as low maternal iron status has been associated with reduced neurodevelopment in offspring and also with increased absorption of Mn (Finley 1999; Gunier et al. 2015; Kupsco et al. 2020). However, hemoglobin was used as a proxy for current iron status and considered as a potential confounder, although it was not included in final models. We also lack information on maternal depression, which has been reported to modify the association between prenatal Mn and cognition (Muñoz-Rocha et al. 2018). Additionally, we cannot rule out unmeasured confounding by earlier exposure to zinc, Pb or other potential co-exposures, which may be associated with levels of early life Mn as well as with neurodevelopment (Adamo and Oteiza 2010; Mahaffey 1990). However, we did adjust for adolescent blood Pb in final models. Effect measure modification by other metals may also be possible, although interactions between adolescent blood Pb and Mn did not improve model fit and were thus excluded. Additionally, because we enrolled participants from three geographic areas with varying ferroalloy activity, study site could act as a potential confounder. However, prenatal and postnatal dentine Mn levels were similar across study sites (Bauer et al. 2017) and neurobehavioral performance did not vary across site (unpublished data). We considered study site as a confounder in sensitivity analyses of prior studies and found no difference in estimates after the addition of study site to models (Bauer et al. 2020a). Finally, we do not expect strong confounding by tooth type, as it is not associated with neurobehavior and was not found to be a confounder in prior similar analyses (Bauer et al. 2017).

Despite these limitations, our study has several strengths. Using a biomarker of Mn exposure instead of environmental media (e.g., dust or soil) allows our study to integrate exposure sources from multiple pathways, including Mn absorbed from the diet and airborne Mn (Butler et al. 2019; Ferri et al. 2012, 2015; Lucas et al. 2015; Pavilonis et al. 2015). This is particularly important in our study because relative contributions of each of these sources may differ by age of the child (e.g., toddlers may be exposed more through dust, take-home exposures, vs. older children may be more exposed via diet, air). Several previous publications in the PHIME cohort have examined exposure sources. For example, we have investigated associations of ferroalloy activity with environmental media including house dust, soil, air and water (Lucas et al. 2015; Pavilonis et al. 2015); uptake of metals in home gardens within these study sites (Ferri et al. 2015); and contributions of these sources to biomarkers (excluding teeth) (Butler et al. 2019). The tooth biomarker captures exposure information integrated over time, beginning at the 2^nd trimester, rather than a spot measurement such as blood or urine, likely reducing exposure misclassification. The tooth biomarker also allows for non-invasive measurement of fetal Mn, which is important given the challenges of determining levels of an actively transported element downstream of placental transfer (Claus Henn et al. 2017; Krachler et al. 1999). Tooth Mn has been correlated with environmental Mn, including proximity to agricultural areas with Mn fungicide application and with parents who work as Mn-containing fungicide sprayers (Gunier et al. 2013). In biological media, tooth Mn levels have been correlated with cord blood Mn as well as with brain, hair, bone and to a lesser degree, with blood, in rodents (Arora et al. 2012; Austin et al. 2017; Gunier et al. 2014; Liang et al. 2016). In our study, we averaged the area under the curve over multiple time points (40 total spot measurements) to derive a cumulative measurement for three distinct exposure windows of prenatal, early postnatal and childhood periods. Although newer methodology for the tooth biomarker exists in which hundreds of measurements for Mn tooth concentrations are taken to capture more discrete time windows of exposure (Claus Henn et al. 2018; Modabbernia et al. 2016), our study nonetheless offers insight into multiple sensitive time windows of Mn exposure.

Another strength of our study is that neurobehavioral assessment was performed during a sensitive neurodevelopmental period (adolescence) that has not been sufficiently studied (Wasserman et al. 2018). Adolescence is a critical time in neurodevelopment in which the dopaminergic system and the prefrontal cortex in general are undergoing profound changes and white matter microstructure is growing and remodeling (Blakemore and Choudhury 2006; Giedd et al. 1999; Lebel and Deoni 2018). Moreover, the adolescent period is likely vulnerable to Mn insult because Mn affects the dopaminergic system (Kern et al. 2010; Wahlstrom et al. 2010b) including white matter connectivity (de Water et al. 2018). Additionally, neurobehavioral assessment in adolescence allows for measurement of outcomes with greater sensitivity and is likely a better determinant of life-long function (White et al. 2009).

Evaluating subtest performance in addition to composite IQ scores allowed us to glean more information about how Mn levels are associated with specific brain function. However, these subtests do not purely measure one domain and do not cover all of the neurobehavior domains that are of interest. Other neurobehavioral functions in children have been reported to be sensitive to Mn exposure, including non-cognitive behavior (Al-Saleh et al. 2019; Carvalho et al. 2018; Horton et al. 2018; Menezes-Filho et al. 2014; Mora et al. 2015, 2018; Oulhote et al. 2014; Rodrigues et al. 2018), motor abilities (Chiu et al. 2017; Hernández-Bonilla et al. 2011; Lucchini et al. 2012a) and olfactory function (Iannilli et al. 2016; Lucchini et al. 2012a).

Conclusions

Our results show that exposure timing is important when evaluating Mn in association with cognition in adolescence. At the levels measured in deciduous teeth in our study, higher prenatal Mn was beneficial for cognition in adolescence. However, these beneficial associations shifted to more harmful associations at later time windows. Cognitive domains most sensitive to Mn across time windows included visuospatial ability, working memory, attention and problem-solving. This work supports the use of neurobehavioral subtests as well as composite scores to help generate hypotheses about Mn-affected brain function.

Highlights.

• Manganese is an essential nutrient but in excess is a neurotoxicant.

• We measured Mn in teeth to estimate exposure during critical developmental windows.

• Adolescent neurobehavior was assessed using the Wechsler Intelligence Scale for Children.

• Mn-neurobehavior associations shifted from beneficial to harmful over time.

• Sex-differences were explored and varied by exposure timing and outcome.