DNA Hypermethylation of Brain-Derived Neurotrophic Factor is Associated with High Levels of Polybrominated Diphenyl Ethers in the Cord Serum: A Laizhou Wan Birth Cohort Study From China

Recent studies have suggested that in utero exposure to environmental organic pollutants, such polybrominated diphenyl ethers (PBDEs) and bisphenol A may induce epigenetic alterations, specically by modifying the patterns of DNA methylation. We investigated the relationship between PBDE exposure and BDNF DNA methylation. We measured the levels of eight PBDE congeners in umbilical cord serum samples using gas chromatography/mass spectrometry. Genomic DNA was extracted from whole cord blood samples. The methylation status of the BDNF gene was determined using methylation-specic PCR with primers specic for either methylated or unmethylated alleles. The unconditional logistic regression model was used to investigate the relationship between PBDE levels and BDNF DNA methylation.


Background
Polybrominated diphenyl ethers (PBDEs) are organic compounds used as ame retardants in a range of products including electronic equipment, vehicles, furniture, computers, textiles, and infant products (Byun et al., 2015). As PBDEs are not chemically bound to these products, they leach from the products with time and use and nally bio-accumulate through the food chain (Shi et al., 2017;Ding et al., 2015).
Previous studies have found that PBDEs are accumulated in human samples such as blood, placenta, breast milk, hair, and tissues (Wang et  Exposure to PBDEs has been linked to a wide range of adverse health effects; they can disrupt endocrine function, affect neurodevelopment, and be teratogenic (Ding et al., 2017;Lyche et al., 2015; 2020; Ishfaq Ahmad Sheikh and Mohd Amin Beg 2020). Recent studies have shown that prenatal exposure to homologues of similar human load PBDEs is associated with irreversible impairment of brain function in offspring, suggesting that PBDEs can cause neurotoxicity and cause learning and memory impairment (Vuong et al., 2017;Ding et al., 2015). Therefore, investigating the effect of PBDE exposure on neurobehavioral development and the underlying mechanism has recently become a focus of research.
Brain-derived neurotropic factor (BDNF) is an important member of the neurotrophin family. It has a central role in neural differentiation, survival of nerve cells, neurite outgrowth, and synaptic plasticity. It is involved in the process of long-term potentiation, which is one of the best known biochemical mechanisms that underlie learning and memory (Leal et al., 2014;Leal et al., 2017). Human BDNF gene is located on chromosome 11p13-14, and consists of 11 exons (I-IX, Vh and VIIIh) and 9 promoters (Pruunsild et al., 2007). BDNF promoter IV is one of the most widely investigated promoters in the contest of DNA methylation changes associated with alterations in BDNF expression (Aarons et al., 2019). For example, McKinney et al (2015) found that BDNFIV methylation in the orbital frontal cortex in the old man was negatively associated with BDNF expression. Xie et al (2017) found that in the periphery blood, BDNFIV methylation was related to mild cognitive impairment. Studies also indicated that BDNF promoter IV appeared to be uniquely sensitive to neuronal stimuli in vitro (Tao et al., 1998;Martinowich et al., 2003) and in vivo (Fuchikami et al., 2008;Sui et al., 2010).
Recent studies have suggested that in utero exposure to environmental organic pollutants, such as . In utero PBDEs exposure has been associated with adverse foetal growth. However, no studies investigate how PBDEs exposure through umbilical cord affects cord blood BDNF DNA methylation. In this study, we investigated BDNF DNA methylation and PBDEs levels in cord blood, and examined the effects of PBDE exposure during pregnancy on BDNF DNA methylation. Our ndings may provide insights to understand the adverse effects of in utero PBDE exposure on foetal growth and development through novel epigenetic mechanisms.

Study population and biological sample collection
This study was a prospective birth cohort study (Laizhou Wan birth cohort, LWBC) which was started in 2010 to determine the effects of environmental exposures on the health of pregnant women and their children and was performed in the south coast area of Laizhou Wan (Bay) of the Bohai Sea, Shandong province, China. The detailed methods of this study are published elsewhere . In total, 164 pregnant women participated in the study from September 2010 to March 2011. Among these, 145 women agreed to take part in this study (response rate 88.4%). 23 cases were excluded for their insu cient cord blood sample for PBDEs analysis and BDNF gene methylation detection, and 14 were excluded for missing values for major confounders. Finally, 108 women were regarded as the study population in the present study. No substantial differences were found in sociodemographic characteristics between the baseline population (n = 145) and the study population (n = 108). The study activities were approved by the Medical Ethics Committee of Shanghai Xinhua Hospital, Shanghai Jiao Tong University School of Medicine. Written informed consent was obtained from each participant.
Pregnant women were recruited at the time when they were preparing for labour and delivery in a county hospital (the only hospital with an obstetric ward) located in the southern coastal area of Laizhou Wan. All interviews were conducted by a specially trained nurse using structured questionnaires shortly after foetal delivery. Demographic and socioeconomic information (maternal age, height, pre-pregnancy weight, education level, household income, and home address) and maternal information on lifestyle habits (cigarette smoking and alcohol use during pregnancy, employment) were gathered at the enrolment interview. Information about exposure to PBDEs during pregnancy included the women's occupation, whether there were any brominated ame retardant (BFR) production factories near their home, and, if yes, what types of BFRs were produced in the factories. Information on paternal exposure to PBDEs and occupation were also collected during the interview. Prenatal and delivery medical records were abstracted including data about pregnancy and delivery complications, pregnancy weight gain, and infant birth outcome.
Umbilical cord blood was obtained immediately after delivery by medical professionals. Blood samples were collected from an umbilical vein immediately post-delivery using a syringe. Each sample was divided into two aliquots. One of the aliquots was collected into EDTA-containing collection tubes for genomic DNA extraction and the other was collected into two 10-mL red-topped tubes. The latter was allowed to clot and centrifuged at 1500 rpm for 20 min, following which the supernatant was decanted into pre-cleaned glass vials for residue analysis. All samples were coded, frozen, and stored at -80°C until further analysis.

PBDE measurement in umbilical cord serum
Detailed extraction, gravimetric lipid determination procedures, and instrumental analysis protocols of PBDE measurement are described elsewhere (Hovander et al., 2000;Ding et al., 2015). Brie y, after spiking with surrogate standards ( 13 C-BDE-139), about 2 mL of each sample was extracted twice with a mixture of HPLC-grade hexane and methyl tertiary butyl ether (MTBE; 1:1 in volume). Lipid content was determined gravimetrically using the whole extract. The amount of serum lipid was used to express the concentration of BDEs on lipid weight basis. Each extract was cleaned by 2 mL concentrated H 2 SO 4 and cleaned up by liquid-solid chromatography using a SiO 2 /H 2 SO 4 column (SiO 2 /H 2 SO 4 2:1 by weight; 1 g).
To ensure the reproducibility of PBDE concentration measurement, one sample out of 20 was randomly selected and its concentration was measurement was repeated; the reproducibility of this analysis was good. At the same time, for every 12 samples, a blank control was processed to check for interference or contamination from solvents and glassware. The limit of detection (LOD) of the method was de ned as the mean blank mass plus three standard deviations. The LOD ranged from 0.24 to 0.48 ng/g lipid in serum samples. The blank values were less than LOD for all PBDE congeners in the samples.
Detection of BDNF DNA methylation in the cord blood Genomic DNA was extracted from whole blood samples (approximately 200 µL) using a Flexi Gene DNA Kit (Qiagen, USA). Puri ed DNA was quanti ed using an ND1000 spectrophotometer (Eppendorf, D30, Germany). The extracted genomic DNA (20 µL) was treated with sodium bisulphite modi cation using Fast DNA Bisul te Kit (Qiagen, USA), according to the manufacturer's instructions. Elution was performed with 15 µL of Buffer EB. The methylation status of the BDNFIV gene was determined using methylationspeci c PCR (MSP) with primers speci c for either methylated or unmethylated alleles.
After bisulphite modi cation, a region-speci c PCR with a biotinylated primer was performed for human BDNFIV. The PCR primer sequences were as follows: for the methylated reaction, 5′-GGTCGGCGGGGAGTAGTAT-3′ (forward) and 5′-CCCATTCCCAACGCTTACCT-3′ (reverse) and for the EpiTect Control DNA and Control DNA Set were used as positive controls for the methylated and unmethylated genes, respectively. Negative control samples without DNA were included in each set of PCR reaction. Pyro sequencing veri cation analyses were conducted at Sangon Biotech Co., Ltd.
(Shanghai). PCR products were analysed on 2% agarose gel, stained with ethidium bromide, and visualised under UV light ( Supplementary Fig. 1). Each MSP was repeated at least once to con rm the results.
In the measurements of methylation-speci c polymerase chain reaction PCR gel electrophoresis, we determined that only the MSP band was ampli ed as methylation. If only the USP band was ampli ed, it was determined that no methylation occurred. Bands observed in both MSP and USP were considered partially methylated.

Statistical analysis
All data were analysed using the SPSS 20.0 software (SPSS Inc., Chicago, IL, USA). Descriptive statistics were calculated for demographic characteristics of mothers and infants. In our nal analysis, the partial methylation and complete methylation in the samples BDNFIV were combined, and the BDNFIV methylation status was divided into methylated and unmethylated. We analysed BDNFIV methylation as categorical variables. As four PBDE congeners (BDE-28, -47, -99, and − 100) were detected in more than 85% of the cord serum, we analysed these four congeners to assess the relationship between cord blood PBDEs and cord blood BDNFIV methylation. An aggregate variable (∑ 4 PBDE) was generated using the molar sum of these four congeners. Cord blood PBDE levels were classi ed into two groups: detectable levels below the median detectable level, and detectable levels above the median. Two independent Chisquare tests were used to examine the correlation between BDNFIV methylation and PBDE levels.
We estimated adjusted odds ratios (ORs) and their 95% con dence intervals (CIs) of PBDEs associated with BDNFIV methylation using unconditional logistic regression model, and the effects of potential confounders were assessed using a forward stepwise regression process. Maternal age, education, prepregnancy body mass index (pre-pregnancy BMI), weight gain during pregnancy, parity, household monthly salary, type of delivery, infant sex, and gestational age were initially considered as confounders. Confounders were nally selected for regression models if they were related to PBDE levels in the literature and were associated (P < 0.2) with BDNF DNA methylation. Potential confounders included maternal age, education, and household monthly salary, pre-pregnancy BMI, parity, and infant sex.

Results
The study included 108 participants. Details regarding their demographic characteristics are presented in Table 1. The average maternal age was 28.51 years (SD = 5.12); two-thirds (66.67%) were primiparous, and half (52.80%) had graduated from high school or above. The majority of the women (93.52%) lived in households with a monthly income less than 5000 (¥). More than half (64.81%) of the women had a normal weight before pregnancy. Of the 108 births, 50 (46.3%) were boys and 58 were (53.7%) girls. The mean birth weight and gestational age were 3.40 kg (SD = 442.13) and 39.75 weeks, respectively. Table 2 shows the distribution (25 th , 50 th , 75 th , and 95 th percentile) of the eight individual lipid-adjusted PBDE congeners. The mean (range) level of all PBDE congeners in cord blood was 23.17 (4.17-2459.46) ng/g lipid. The detection rate of PBDE-28, -47, -99, and -100 was over 85% in all samples.
Unconditional logistic regression models were used to analyse the associations between cord blood BDNFIV methylation and PBDE exposure. Methylation level was modelled as a function of PBDE levels, adjusting for maternal age, education, household monthly salary, pre-pregnancy BMI, parity, and infant sex (Table 4). After adjusting for potential confounders, the incidence of BDNFIV methylation status increased when the levels of PBDE congeners increased. The statistical parameters of this association are as follows. BDE-28: OR = 3.32, 95% CI:  Previous studies have observed that infants and children who are prenatally exposed to high levels of PBDEs are more likely to suffer from various adverse health consequences, such as adverse infant birth outcomes Chen et al., 2018). The exact mechanism underlying the adverse effects associated with prenatal exposure to PBDEs remains unclear; however, epigenetic mechanisms have been suggested as a possible pathway linking in utero PBDE exposure to adverse health effects.
In the present study, we evaluated the association between prenatal exposure to PBDEs and BDNF DNA methylation levels measured in cord blood samples. We found that BDE-28, -47, -99, and − 100 were positively associated with BDNF DNA (exon IV) methylation. As DNA methylation is involved in the normal processes of development and genomic imprinting, any changes in the normal regulation of this system might have a high negative impact on the human body (Bernstein et al., 2007). DNA methylation changes have been found in psychiatric post-mortem brain samples (Keller et al., 2010) as well as in peripheral blood of the psychiatric patients (Kang et al., 2013;Ikegame et al., 2013). Meanwhile, a cohort study conducted at the Columbia Center for Children's Environmental Health (CCCEH) found that BDNF DNA methylation in the human blood may be used as a predictor of brain BDNF DNA methylation and gene expression; moreover, it can predict behavioural vulnerability induced by early-life environmental exposure (Kundakovic et al., 2015). Thus, our present study showed that increased levels of PBDEs were related to BDNFIV methylation in the cord blood. Our results contribute to the growing body of evidence that PBDEs may act as developmental neurotoxicants.
This study is the rst to experimentally con rm that PBDE exposure is linked to BDNF DNA hypermethylation in the cord blood. Similarly, associations between DNA methylation levels and prenatal exposure to other compounds, including BPA (Kundakovic et al., 2015), phthalate (Zhao et al., 2015), and air pollutants (Janssen et al., 2013), have been reported. However, the in uence of PBDEs on cord blood BDNF DNA hypermethylation levels has not been described. Most previous studies on the relationship between PBDE exposure and epigenetic changes have reported DNA methylation in different tissues such as placenta and cord blood (Dao et al., 2015;Kappil et al., 2016;Zhao et al., 2016). A positive association was observed between placental global DNA methylation and total PBDE levels in a study by Kappil et al. (2016). Placental DNA hypomethylation was associated with high levels of PBDE exposure in cord blood (Zhao et al., 2016). On the contrary, Dao et al. (2015) found that high maternal BDE-47 exposure was associated with decreased cord blood TNFα DNA methylation.
The biological mechanism for the interference of PBDEs with BDNF DNA hypermethylation is still unclear. Previous studies have shown that prenatal exposure to PBDEs was associated with increased odds of low thyroid hormones (Herbstman et al., 2008;Lin et al., 2011) and disrupted thyroid homeostasis by binding of thyroid hormones with their receptors and transporter proteins (i.e., transthyretin) (Yu et al., 2010;Ernest et al., 2012). Sui et al. (2010) suggested that perinatal hypothyroidism induces epigenetic modulations of DNA methylation and the cross talks involved in chromatin reorganisation or remodelling, thus in uencing BDNF expression. We speculate that PBDEs disrupt thyroid homeostasis and reduce the delity of the epigenetic machinery, resulting in BDNF DNA hypermethylation of cytosine residues.
Our study is one of the rst to show a relationship between PBDE exposure and DNA methylation in cord blood samples. However, there are several limitations to our study. Firstly, we did not measure all PBDE congeners, including BDE-209, which is the main constituent of the deca-BDE mixture. Therefore, our study cannot provide any information on the relationship between BDE-209 and BDNF DNA methylation. Secondly, epigenetic modi cations are tissue speci c. In our study, we acknowledged that cord blood contained a variable number of nucleated newborn red blood cells that are globally more hypomethylated than white blood cell types, which may confound our methylation measurements. It indicates that distinctive cell counting and / or classi cation of cells from whole blood is required to adjust the ndings from the methylation patterns by regression model or access the methylation patterns in different types of blood cells (Dao et al., 2015). However, in our present study we cannot obtain the cell distribution of the immune cells in the cord blood samples we utilized. Given the fact that cord blood may contain stem cells that can populate the brain in later life (Singh et al., 2015) and also provide abundant source of immune cells which are important producers of cytokines (Stock et al., 2000), the cord blood may provide a reasonable surrogate for our target organ/tissues (brain/immune cells) (Dao et al., 2015). Thirdly, as partial methylation was treated as methylated samples in the nal analysis, information bias could not be ruled out. However, after restricting the samples to complete methylation and no methylation samples, BDE-28 and BDE-99 are still related to the methylation status of BDNFIV, suggesting the robustness of the results reported in the present study to some extent. Fourthly, a lack of negative control using unmethylated promoter linked highly expressed genes in this study may con ne our conclusions. More speci ed studies, for instance, conservative gene sequences as negative controls, pyrosequencing, or genome-wide analyses, are required to elucidate the potential mechanisms that how PBDEs induce BDNFIV methylation. Lastly, as people are simultaneously exposed to a variety of environmental pollutants, it is possible that unmeasured confounders have a role to play, which may affect both the exposure levels and BDNF DNA methylation status. In addition, our small sample size may limit our power to detect signi cant relationship in logistic regression models, which calls for further steps to demonstrate the potential in uence of prenatal PBDE exposure on the foetal epigenome, by performing replicated studies in a larger sample pool and exploring the relationship between prenatal PBDEs exposure and epigenetic changes.

Conclusions
In conclusion, our present study found that the levels of PBDEs in the cord serum were much higher compared with those measured in other studies in the general population in China. A number of PBDE congeners, including BDE-28, BDE-47, BDE-99, and BDE-100, were positively associated with BDNF DNA hypermethylation in the cord blood. Our ndings may provide insights to understand the adverse effects of in utero PBDE exposure on foetal growth and development through novel epigenetic mechanisms.

Declarations
(YG2019ZDA29), Three-year Action Project of Shanghai Municipal Public Health System Construction (GWV-10.1-XK11). The authors declare that they have no actual or potential competing nancial interests.

Ethical Approval and Consent to participate
The study activities were approved by the Medical Ethics Committee of Shanghai Xinhua Hospital, Shanghai Jiao Tong University School of Medicine.

Consent for publication
Written informed consent was obtained from each participant.

Availability of supporting data
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no actual or potential competing nancial interests.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. Supplementarymaterial.docx