Prenatal Lead Exposure, Genetic Factors, and Cognitive Developmental Delay

Key Points Question Is prenatal lead (Pb) exposure together with genetic factors associated with the risk of cognitive developmental delay (CDD)? Findings In this cohort study of 2361 mother-child pairs followed up prenatally to child age of approximately 2 years, prenatal exposure to even low levels of Pb was associated with an increased risk of CDD in children, especially those with a high genetic risk. Meaning These findings suggest that prenatal Pb exposure and its interaction with genetic factors might jointly contribute to CDD risk, indicating the possibility of an integrated strategy for the assessment of CDD risk.

which was adapted for application in the Chinese population. 2,3As with its precursor, the BSID-CR includes scales for mental and psychomotor factors.The mental (psychological) and psychomotor development indices (MDI and PDI, respectively) were derived from the raw scores of the appropriate scales which were then adjusted based on the child's age as necessary.This was achieved via the formula: MDI or PDI = 100 + (X -Xm) / sd × 16, where Xm and sd are the mean and standard deviation of the abovementioned raw scores, respectively, and X is the raw score of the subject child. 3Specifically, MDI is a representation of a child's cognitive development with respect to language ability, generalization, categorization, memory, and social skills among other aspects, while PDI serves the same purpose for psychomotor development, such as muscle coordination, and gross and fine manipulative skills.Overall, children with better cognitive development tend to have higher MDI scores, and for this study, the BSID-CR grading scale standard 3 was applied, whereupon an MDI ≤ 79 was considered as cognitive developmental delay (CDD).Children with PDI ≤ 79 was considered as motor developmental delay (MDD).Tests were administered by mental healthcare professionals who were not informed of the children's medical records (including family background, genetic factors, Pb exposure levels, etc.).For quality control, the entire evaluation process was monitored and fully documented, and the tests were conducted by similarly blind research assistants at the study hospital in a secluded area where background noise was at a minimum.

eMethods 2. Detailed Procedures for the Measurement of Pb Concentrations
The plasma sample was properly mixed after being defrosted at 4°C, and a 90 μL aliquot was taken at room temperature (20-25°C) into a fresh tube for metal analysis.Samples were first diluted with ultrapure water (1:10) and then shaken to mix.The mixture was then 1:1 acidified with a mixture of Triton 0.04%, butanol 1%, and nitric acid 2% (1:1:1, v/v/v) and digested at room temperature overnight after vortex mixing.Personnel who weren't aware of the grouping information performed the assays.All participants' plasma samples were randomly tested and quantified.The method's accuracy was tested using certified reference materials (ClinChek® human plasma controls for trace elements no.8883 and no.8884, Recipe, Munich, Germany).The reference materials' metrics were all within the recommended range for Pb.Blank samples and quality control samples were analyzed every 30 samples in all batches.Inductively coupled plasma mass spectrometry (ICP-MS, Agilent 7900, Agilent Technologies, USA) was used to analyze the obtained solutions.Furthermore, to assess the whole blood Pb for children, the venous blood samples from 2-year-old children were measured using a Trace Element Analyzer AS-9000C (Wuhan Aoshou Medical Technology Co., Ltd) in accordance with standard operating procedures.For all Pb concentrations, the coefficients of variation (CVs) were less than 10% intra-and inter-day, while the limit of detection (LOD) for plasma Pb in mothers (using ICP-MS) was 0.005 ng/ml, and that of the blood Pb in children (using the trace element analyzer AS-9000C) was 0.001 µg/dL.None of the concentrations recorded in collected samples were lower than the respective LOD.

eMethods 3. DNA Extraction for Human Genome-Wide Arrays
Upon collection at delivery, samples of cord blood were placed in EDTA-coated vacuum tubes, centrifuged, and placed at -80°C until detection.DNA of leukocytes in the samples were extracted and then evaluated (OD260/280 ratio and concentration) using a NanoDrop Spectrophotometers (Thermo Scientific, Wilmington, DE, USA).The DNA with a OD260/280 ratio of 1.7 -2.0, a OD260/230 ratio of 2.0 -2.2, and haven't been degraded was performed subsequent genome-wide arrays.

eMethods 4. Genotyping, Quality Control
Genotyping was carried out and completed, both per the manufacturer's protocols, using the Illumina Infinium Asian Screening Array v1.0 BeadChip (Illumina Inc., San Diego, CA, United States) (which contains ~ 700,000 markers) and Illumina iScan System, respectively.
The prospective birth cohort in Wuhan, which consisted of 2361 children, was genotyped for the PRS application.
For the present study, standard quality control was applied at both the sample and variant levels similar to previously reported research.Genotyping quality control was performed using PLINK 1.9 (https://www.cog-genomics.org/plink/1.9/) 4.We excluded samples with call rate <95% or sex discrepancy, or samples with extreme heterozygosity (> 6 SD from the mean).Identity by descent (IBD) analysis to detect cryptic relatedness and members with low call rates were excluded when PI_HAT > 0.185 for a pair of samples.Population structure was evaluated by principal component analysis (PCA) based on LDpruned (r 2 < 0.8) autosomal SNPs.We further excluded SNPs not in Hardy-Weinberg equilibrium (HWE, P < 10 -6 ).Additional SNPs with MAF < 0.01 and call rate < 95% were excluded.
Quantile-quantile (Q-Q) plots were generated to compare observed P-value distributions to expected distributions under a null hypothesis (no stratification of populations).In this study population, the Q-Q plots did not reveal any evidence of stratification in the population (inflation factor λ = 1.005) (eFigure 3 in Supplement 1).

eMethods 5. The Imputation Process
Genotype imputation is an effective and critical method for estimating unobserved genotypes in genomic data from single nucleotide polymorphism (SNP) genotyping arrays.
Fine mapping of variants in genome-wide association studies (GWAS) can be greatly improved by the imputation of optimal reference panels based on population-specific haplotypes.Previous reference panel datasets with limited Chinese samples have limited the quality of reference panel imputation in current Chinese population studies.In addition, insufficient sequencing depth in the constructed datasets may undermine the ability to detect rare variants in imputation.Therefore, population-specific reference panels constructed from large, in-depth WGS datasets of Chinese populations are essential for accurate and comprehensive imputation of genotyping arrays and sequencing data in Chinese populations.
The China Metabolic Analysis Project (ChinaMAP) has achieved high-quality imputation for a Chinese genomic dataset of 10,155 unlinked Chinese individuals with 59.01 M SNPs.
ChinaMAP's reference samples provide more comprehensive imputation and more novel findings for genetic studies of the Chinese population.ChinaMAP's imputation server can be found at The ChinaMAP imputation server is available on the ChinaMAP website (www.mbiobank.com)for genotype imputation, downloading results, and variant databases.
For the Infinium Asian Screening Array (ASA) imputation, the ChinaMAP Reference Board has demonstrated excellent performance in terms of the number of well-imputed variants and the quality of the shared imputed variants. 5ible samples and genotype data were phased with Eagle2 and imputed with ChinaMAP, using default parameters, and ChinaMAP dataset (Phase 1 Integrated Variant Set Release, spanning 10155 samples) was used as a reference.SNPs with MAF < 0.01 were excluded from the analysis.

eMethods 6. Polygenic Risk Score (PRS) Calculation
In the present study, we generated the PRS by multiplying the genotype dosage of each risk allele for each variant by its respective weight (β), summing all variants together into a PRS.
According to the number of risk alleles, each SNP was recorded as 0, 1, or 2. The PRS was calculated using the equation: PRS , where M is the total number of SNPs, and βj is per allele for cognitive ability associated with SNPj.The β were taken from a meta-analysis of genome-wide association studies in the CHARGE consortium on cognitive function. 6eMethods 7. Covariates The following data was extracted from participants' medical records: mother's age; gestational age; mode of delivery; gestational diabetes; gestational hypertension; gender of the baby; and birth weight of the baby.The heights and weights of mothers were also collected, from which their body mass index (BMI) before pregnancy were carefully calculated.Standardized in-person interviews with appropriately trained staff at the hospital provided additional information (in particular, the mother's education level and extent of passive smoking, the father's education level, and annual household income).Passive smoking was defined as prenatal exposure to cigarette smoke during pregnancy (whether from the father or other people who smoked at home or work).The 10 principal components (PCs) to be chosen to adjust for potential confounding effects of population genetic group stratification according to scree plot analysis (eFigure 4 in Supplement 1). 7Additionally, the blood Pb level of children involved in the study, as measured using a trace element analyzer (AS-9000C, Wuhan Aoshou Medical Technology Co., Ltd.), was also included as a covariate in the models.Genetic risk and prenatal Pb exposure were also included as covariates in the models when analysing additive interaction effects.

eMethods 1 . 2 . 5 . 2 . 3 . 4 . 5 . 6 .eTable 7 .
Chinese Revision of Bayley Scale of Infant Development for Cognitive Assessment eMethods Detailed Procedures for the Measurement of Pb Concentrations eMethods 3. DNA Extraction for Human Genome-Wide Arrays eMethods 4. Genotyping, Quality Control eMethods The Imputation Process eMethods 6. Polygenic Risk Score (PRS) Calculation eMethods 7. Covariates eTable 1. Descriptive Characteristics of All Mother-Child Pairs eTable Concentrations of Pb Among Study Participants eTable Association Between CDD Risk and Prenatal Pb Exposure in Children With Different Genders eTable Association Between CDD Risk and Pb Exposure in 2-Year-Old Children eTable Sensitivity Analyses Restricted to Children With a Normal Range of Birth Weight eTable Sensitivity Analysis Restricted to Pregnant Women Without Gestational Hypertension or Gestational Diabetes Associations Between MDD Risk and Prenatal Pb Exposure eTable 8. Summary Results of 58 SNPs Used for Polygenic Risk Score eTable 9. General Linear Model Evaluating Association Between PRS and MDI eTable 10.Full Regression Analysis Results for Table 2

eTable 11 . 2 eFigure 1 . 2 . 3 .eFigure 4 .
Full Regression Analysis Results for Figure Flow Diagram of Mother-Child Pairs Exclusion eFigure Association Between Baseline Status and Risk of CDD eFigure Quantile-Quantile Plot and Genomic Inflation Factor Lambda for Associations With CDD Risk Scree Plot of Eigenvalues for the Principal Component Analysis (PCA) eReferences This supplemental material has been provided by the authors to give readers additional information about their work.

eMethods 1 .
Chinese Revision of Bayley Scale of Infant Development for Cognitive Assessment Chinese revision of Baylcy Scale of Infant Development (BSID-CR) is a variant of the Bayley Scales of Infant Development (BSID-1)

eFigure 1 .
Flow Diagram of Mother-Child Pairs Exclusion.This cohort study initially included 2416 mother-child pairs recruited from the birth cohort who had available maternal blood samples and qualified cognitive assessment for children.Among them, 41 participants whose blood samples failed in the SNP chip detection were excluded.Furthermore, participants with missing information on pre-pregnancy BMI (n = 1), annual family income (n = 8), passive smoking status (n = 3), paternal education status (n = 1), and parity (n = 1) were excluded.For these reasons, our final study population consisted of 2361 mother-child pairs.Furthermore, during the follow-up, 1372 children provided venous blood for a Pb exposure measurement at the study hospital at approximately 2 years old.eFigure 2. Association Between Baseline Status and Risk of CDD.The OR (95% CI) for CDD risk at baseline status by child gender, mode of delivery, annual family income, maternal education, and paternal education.OR, odds ratio; CI, confidence interval.eFigure 3. Quantile-Quantile Plot and Genomic Inflation Factor Lambda for Associations With CDD Risk.The results were based on 2361 children in present study.The red dot represents the distribution of P values for the association.The observed versus expected  2 test statistics show no evidence for inflation of  2 tests (inflation factor λ = 1.005).No evidence of population stratification was observed in 2361 children.Adjusted for characteristics of the child (gender, gestational age at birth), mother (age, education status, pre-pregnancy body mass index, passive smoking status, number of deliveries, mode of delivery), father (education status), family (annual family income), and the top 10 principal components of ancestry and genotyping batch.

eTable 1. Descriptive Characteristics of All Mother-Child Pairs
Abbreviations: SD, standard deviation; CDD, cognitive developmental delay.aPvalue for the difference according to the chi-square test and unpaired two-samples t-test.eTable2.Concentrations of Pb Among Study Participants a Children's Group Mothers, GM(GSD) P Value † Children, GM(GSD) P Value bAbbreviations: Pb, Lead; GM, Geometric mean; GSD, geometric standard deviation.aThe unit is ng/mL (to convert ng/mL to g/L, multiply by 1 × 10-6) b P value for differences between various genders or genetic risks using non-parametric tests.eTable3.

Association Between CDD Risk and Prenatal Pb Exposure in Children With Different Genders
a Adjusted for characteristics of the child (gender, gestational age at birth), mother (age, education status, pre-pregnancy body mass index, passive smoking status, number of deliveries, mode of delivery), father (education status), and family (annual family income).b Test for trend based on the variable containing the median value for each tertile.eTable 4.

Association Between CDD Risk and Pb Exposure in 2-Year-Old Children
Abbreviations: CDD, cognitive developmental delay; Pb, Lead; OR, odds ratio; CI, confidence interval.Model 1: Unadjusted model.Model 2: Adjusted for characteristics of the child (gender, gestational age at birth), mother (age, education status, pre-pregnancy body mass index, passive smoking status, number of deliveries, mode of delivery), father (education status), and family (annual family income).Model 3: Model 2 + Prenatal Pb exposure (Ln transformed).a Test for trend based on the variable containing the median value for each tertile.eTable 5.

Sensitivity Analyses Restricted to Children With a Normal Range of Birth Weight
aTest for trend based on the variable containing the median value for each tertile.eTable 6.

Sensitivity Analysis Restricted to Pregnant Women Without Gestational Hypertension or Gestational Diabetes
Abbreviations: CDD, cognitive developmental delay; Pb, Lead; OR=odds ratio; CI, confidence interval.Model 1: Unadjusted model.Model 2: Adjusted for characteristics of the child (gender, gestational age at birth), mother (age, education status, pre-pregnancy body mass index, passive smoking status, number of deliveries, mode of delivery), father (education status), and family (annual family income).Model 3: Model 2 + BLLs in children.aTest for trend based on the variable containing the median value for each tertile.eTable7.

Associations Between MDD Risk and Prenatal Pb Exposure
a Test for trend based on the variable containing the median value for each tertile.eTable 8.

General Linear Model Evaluating Association between PRS and MDI
© 2023 Jia Z et al.JAMA Network Open.eTable 10.

Full Regression Analysis Results for Table 2
© 2023 Jia Z et al.JAMA Network Open.eTable 10.

Full Regression Analysis Results for Table 2 (continued)
Adjusted for characteristics of the child (gender, gestational age at birth), mother (age, education status, pre-pregnancy body mass index, passive smoking status, number of deliveries, mode of delivery), father (education status), and family (annual family income).Model 3: Model 2 + blood Pb levels in children.

Full Regression Analysis Results for Figure 2 (continued)
© 2023 Jia Z et al.JAMA Network Open.eTable 11.

Full Regression Analysis Results for Figure 2 (continued)
© 2023 Jia Z et al.JAMA Network Open.eTable 11.

Full Regression Analysis Results for Figure 2 (continued)
Pb, Lead; OR, odds ratio; CI, confidence interval.Model 1: Unadjusted model.Model 2: Adjusted for characteristics of the child (gender, gestational age at birth), mother (age, education status, pre-pregnancy body mass index, passive smoking status, number of deliveries, mode of delivery), father (education status), and family (annual family income).Model 3: Model 2 + blood Pb levels in children.