Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Placental CpG methylation of infants born extremely preterm predicts cognitive impairment later in life

  • Sloane K. Tilley,

    Roles Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Validation, Visualization, Writing – original draft

    Affiliation Department of Environmental Sciences and Engineering, Gillings School of Global Public Health, University of North Carolina, Chapel Hill, North Carolina, United States of America

  • Elizabeth M. Martin,

    Roles Data curation, Methodology, Writing – review & editing

    Affiliation Department of Environmental Sciences and Engineering, Gillings School of Global Public Health, University of North Carolina, Chapel Hill, North Carolina, United States of America

  • Lisa Smeester,

    Roles Data curation, Investigation, Methodology, Project administration, Resources, Supervision

    Affiliation Department of Environmental Sciences and Engineering, Gillings School of Global Public Health, University of North Carolina, Chapel Hill, North Carolina, United States of America

  • Robert M. Joseph,

    Roles Conceptualization, Funding acquisition, Investigation, Project administration, Supervision, Writing – review & editing

    Affiliation Department of Anatomy and Neurobiology, Boston University School of Medicine, Boston, Massachusetts, United States of America

  • Karl C. K. Kuban,

    Roles Conceptualization, Funding acquisition, Project administration, Supervision, Writing – review & editing

    Affiliation Department of Pediatrics, Boston Medical Center, Boston, Massachusetts, United States of America

  • Tim C. Heeren,

    Roles Conceptualization, Funding acquisition, Methodology, Project administration, Supervision, Writing – review & editing

    Affiliation Department of Biostatistics, Boston University, Boston, Massachusetts, United States of America

  • Olaf U. Dammann,

    Roles Conceptualization, Funding acquisition, Project administration, Supervision, Writing – review & editing

    Affiliation Department of Public Health and Community Medicine, Tufts University School of Medicine, Boston, Massachusetts, United States of America

  • T. Michael O’Shea,

    Roles Conceptualization, Funding acquisition, Methodology, Project administration, Resources, Supervision, Visualization, Writing – review & editing

    Affiliation Department of Pediatrics, School of Medicine, University of North Carolina, Chapel Hill, North Carolina, United States of America

  • Rebecca C. Fry

    Roles Conceptualization, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Visualization, Writing – review & editing

    rfry@unc.edu

    Affiliations Department of Environmental Sciences and Engineering, Gillings School of Global Public Health, University of North Carolina, Chapel Hill, North Carolina, United States of America, Curriculum in Toxicology, School of Medicine, University of North Carolina, Chapel Hill, North Carolina, United States of America

Placental CpG methylation of infants born extremely preterm predicts cognitive impairment later in life

  • Sloane K. Tilley, 
  • Elizabeth M. Martin, 
  • Lisa Smeester, 
  • Robert M. Joseph, 
  • Karl C. K. Kuban, 
  • Tim C. Heeren, 
  • Olaf U. Dammann, 
  • T. Michael O’Shea, 
  • Rebecca C. Fry
PLOS
x

Abstract

Background

The placenta is the central regulator of maternal and fetal interactions. Perturbations of placental structure and function have been associated with adverse neurodevelopmental outcomes later in life. Placental CpG methylation represents an epigenetic modification with the potential to impact placental function, fetal development and child health later in life.

Study design

Genome-wide placental CpG methylation levels were compared between spontaneous versus indicated deliveries from extremely preterm births (EPTBs) (n = 84). The association between the identified differentially methylated CpG sites and neurocognitive outcome at ten years of age was then evaluated.

Results

Spontaneous EPTB was associated with differential CpG methylation levels in 250 CpG sites (217 unique genes) with the majority displaying hypermethylation. The identified genes are known to play a role in neurodevelopment and are enriched for basic helix-loop-helix transcription factor binding sites. The placental CpG methylation levels for 17 of these sites predicted cognitive function at ten years of age.

Conclusion

A hypermethylation signature is present in DNA from placentas in infants with spontaneous EPTB. CpG methylation levels of critical neurodevelopment genes in the placenta predicted later life cognitive function, supporting the developmental origins of health and disease hypothesis (DOHaD).

Introduction

Preterm infants have an increased risk of cognitive impairment later in life [1]. In the Extremely Low for Gestational Age Newborns (ELGAN) cohort, indicators and antecedents of perinatal inflammation are associated with a range of neurodevelopmental outcomes in early and later childhood, including mental and motor impairment, behavioral problems, and cerebral palsy [26]. As the placenta mediates maternal-fetal interactions, it is likely that placental signaling is involved in these adverse developmental outcomes [7].

A potential mediator of gene and subsequent protein expression in the placenta is DNA methylation. DNA methylation of cytosine (CpG methylation) is an epigenetic modification that can influence gene expression levels without changing DNA sequences [710]. Placental CpG methylation has been reported to be a major mechanism by which the placenta dynamically responds to changing conditions throughout pregnancy and thus potentially affects long-term child health outcomes [711]. For example, CpG methylation in the placenta is associated with exposure to environmental factors including inorganic arsenic [12]. Further, placental CpG methylation is associated with newborn neurobehavior [13]. Interestingly, placentas derived from a male or female pregnancy have vastly different CpG methylation signatures [14], which may play a role in the observed differences in neurobehavioral outcome differences between males and females. Taken together, these data support that CpG methylation in the placenta is associated with in utero exposure to toxic substances and neurobehavior at birth.

The placenta serves as an environmental mediator as well as being the environment experienced by the gestating fetus. It plays a critical role in pregnancy, and abnormalities of the placenta have been linked to preterm delivery. Preterm birth is not considered to simply be the early induction of the same biological process as term birth [15]. In fact, two distinct subtypes of preterm birth have been identified and are suspected to have different underlying etiologies [16]. Indicated preterm birth is associated with placental dysfunction, including preeclampsia and intrauterine growth retardation. Spontaneous preterm birth, in contrast, is characterized by an inflammatory placental phenotype and is associated with pre-labor premature rupture of membranes, chorioamnionitis, and placental infections [16].

In the present study, the goal was to determine whether there are CpG methylation level differences in placental tissue from spontaneous versus indicated EPTB in the ELGAN cohort. As spontaneous EPTB is characterized by an inflammatory intrauterine environment, it was of interest to test whether genes that are differentially methylated in placentas from spontaneous EPTB also were associated with a neurodevelopmental outcome. The results indicate that methylation of genes in placental tissue predicts children’s cognitive function at ten years of age.

Materials and methods

Study subject recruitment

The recruitment process for the ELGAN study has been described in detail elsewhere [17]. Briefly, between 2002 and 2004, women who gave birth before 28 weeks gestational age at one of the 14 hospitals in five states in the United States were invited to participate in the study. A total of 1,249 mothers and 1,506 infants enrolled in the study. From these infants, 1,365 placentas were collected. Based on currently available placental epigenetic data and demographic information, a sub-cohort of 84 mother/infant pairs were investigated in the present study based on the child having intellectual deficit (n = 18), autism spectrum disorder without intellectual deficit (n = 18), or neither intellectual deficit or autism spectrum disorder (e.g. controls) (n = 48) [18, 19]. The lowest gestational age among infant participants for the current study was 23 weeks.

Sample collection

Women were asked to contribute their placentas for the ELGAN study. After delivery, placentas were placed in a sterile exam basin and were biopsied under sterile conditions in a sampling room. To collect the chorion, the amnion was pulled back using sterile technique to expose the chorion at the midpoint of the longest distance between the cord insertion and the edge of the placental disk. A piece of tissue was removed by cutting at the base of the section after applying traction to the chorion and the underlying trophoblast tissue. The collected specimen was placed in a cryo-vial and immediately immersed in the liquid nitrogen. Placental samples were stored at -80°C until shipment to the University of North Carolina at Chapel Hill for processing as detailed [20].

DNA extraction and Illumina 450K methylation assay

A subsection of placental tissue was cut from each frozen sample on dry ice, washed in sterile 1X PBS to remove residual blood. Subsections were then homogenized in Buffer RLT with β-mercaptoethanol (Qiagen, Valencia CA). The AllPrep DNA/RNA/miRNA Universal Kit (Qiagen, Valncia CA), in accordance with the manufacturer’s instructions, was used to isolate DNA and RNA sequences greater than 18 nucleotides in length. Isolated DNA was first bisulfite-converted using the EZ DNA methylation kit (Zymo Research, Irvine, CA). Converted DNA was then hybridized onto the Illumina HumanMethylation450 BeadChip (Illumina, Inc, San Diego, CA), which assesses the methylation levels of a total of 486,428 individual probes. Each probe measures the methylation level at a single CpG site. BeadChip microarray data were collected at Expression Analysis, Inc (Durham, NC; www.expressionanalysis.com). Methylation levels were calculated and expressed as β values (β = intensity of the methylated allele (M)) / (intensity of the unmethylated allele (U) + intensity of the methylated allele (M) + 100), as previously described elsewhere [21]. Batch effect was evaluated using principle component analysis (PCA) and was not a significant source of variation (p = 0.25). For data filtration, probes with high detection P-values (P > .01) were considered to be unreliable and removed from analysis (n = 24,591), as recommended by the manufacturer. Subsequently, data were normalized using the beta-mixture quantile (BMIQ) normalization methodology [22]. This was performed using the wateRmelon package (version 1.11.0) in R (version 3.2.3; The R Project for Statistical Computing) [23]. Probes that represent known single nucleotide polymorphisms (SNPs) (n = 84,124) and probes that are not associated with an annotated gene were removed, leaving a total of 286,410 probes for further analyses, representing 20,420 genes.

Classification of pregnancy disorders

Disorders leading to EPTB may be classified into two groups, based on pregnancy complications, placental histology, and placental microbiology [16]. Spontaneous EPTB is characterized by preterm labor, prelabor premature rupture of membranes, placental abruption, or cervical insufficiency and is characterized by histologic chorioamnionitis and placental microbe recovery. Indicated EPTB is characterized preeclampsia or fetal indication/intrauterine growth restriction and is characterized by a paucity of organisms and inflammation but the presence of histologic features of abnormal placentation. The sample in the current study include n = 59 from spontaneous EPTB and n = 25 from indicated EPTB [16] selected for analysis based upon subjects with age 10 follow up and also currently available placental CpG methylation data.

Cognitive assessment at ten years of age

When study participants were ten years of age, general cognitive ability (or IQ) was assessed with the School-Age Differential Ability Scales–II (DAS-II) Verbal and Nonverbal Reasoning scales [24]. Two subtests from the DAS-II and five subtests from the NEPSY-II were used to assess executive function [25]. DAS-II Recall of Digits Backward and Recall of Sequential Order measured verbal working memory. NEPSY-II Auditory Attention and Auditory Response Set evaluated auditory attention, set switching and inhibition. NEPSY-II Inhibition and Inhibition Switching assessed simple inhibition and inhibition in the context of set shifting, respectively. NEPSY-II Animal Sorting measured concept generation and mental flexibility. In order to obtain a unitary measure of cognitive function, a latent profile analysis (LPA) was used to identify subsets of study participants with similar profiles on all measures of IQ and executive functioning. Four profiles of cognitive function were identified: normal (34% of ELGAN cohort), low-normal (41%), moderately impaired (17%), and severely impaired (8%) [19, 26, 27].

Statistical analysis of differential placental CpG methylation associated spontaneous versus indicated EPTB

To identify CpG sites that were differentially methylated between indicated EPTB (n = 25) and spontaneous EPTB (n = 59), mixed effect regression analysis was run for all annotated probes. Potential confounders were included in the model if they differed between placentas from indicated EPTB and spontaneous EPTB. To be most inclusive of potential confounders in the final regression models, p was set at < 0.20. Variables tested that did not differ between the two groups included maternal age, maternal race, maternal BMI, maternal education level, the number of participants on public health insurance, and maternal exposure to smoke during pregnancy (active or passive). Covariates included in the model were gestational age, infant sex, and infant birthweight. In order to control for multiple tests, FDR q-values were calculated. Significance was defined as an average beta difference ≥ |0.10|, which corresponds to approximately a 5% false positive rate, and a q-value < 0.05 [28]. As a secondary analysis the reference-free method of adjusting for cellular heterogeneity described in Houseman et al. 2014 was conducted using the RefFreeEWAS package in R [29]. This deconvolution method utilizes a surrogate variable analysis (SVA) that is data driven to identify latent variables as surrogates of cellular composition [29]. As in the prior analysis, the predictor variable tested was EPTB, and covariates included in the model were gestational age, infant sex and infant birth weight.

The distribution of differentially methylated probes (DMPs) was contrasted to the observed distribution of all annotated probes by intragene site: (i) from 200–1500 base pairs upstream of the gene transcription start site (TSS1500), (ii) within 200 base pairs upstream of the gene transcription start site (TSS200), (iii) in the 5’ untranslated region of the gene (5’UTR), (iv) in the first exon of the gene (1st Exon), (v) in the body of the gene (Body), and (vi) in the 3’ untranslated region of the gene (3’UTR) [21]. Permutation testing was used to determine if the observed percentage of differentially methylated intragene sites significantly differed from the distribution of all annotated probes.

In order to examine the higher-level biological functions and processes, functional relationships among the differentially methylated genes were assessed using Ingenuity Network Analysis (IPA) (Ingenuity Systems®, Redwood City, CA, USA) and gene set enrichment analysis (GSEA). As the differentially methylated sites were tested in the context of approximately the entire genome (e.g. 286,410 probes representing 20,420/20,623 genes or 99% of the tested genome), the genome was selected as the background for the enrichment analyses. Canonical pathways enriched among this were analyzed and reported. Significance assessed using IPA was assessed using a right-tailed Fisher’s Exact test p-value < 0.0001. GSEA, which uses a rank-based analysis method to assess biological enrichment, was used as a second method to examine pathway enrichment [30]. GSEA examines discordant differences between two biological states by calculating an enrichment score within a ranked list. CpG methylation data along with relevant covariates have been deposited in the gene expression omnibus (GEO) database (GSE106089).

Overrepresentation of transcription factor binding sites among DMPs

To examine whether the CpG sites identified as differentially methylated in spontaneous versus indicated EPTB had common transcriptional regulators, overrepresentation of transcription factor binding sites was performed using Genomatix (Genomatix Software Inc., Ann Arbor, MI). Overrepresentation of transcription factor binding sites was performed using the forward sequence corresponding to each of the probes identified as differentially methylated in spontaneous versus indicated EPTB. To determine whether a binding site is overrepresented, sequences for each probe compared to the MatBase of sequences. This is a database of transcription factor binding site motifs. The subsequent analysis identifies overrepresentation of individual transcription factor binding sites based upon a comparison to the genomic background using a Z-score calculation [31]. The most significant transcription factors are reported.

Logistic regression of DMPs to later life cognitive function

Logistic regression analysis was performed in SAS (Cary, NC) to test whether the DMPs predicted later life neurocognitive function. Methylation levels were calculated as a proportion between 0 and 1. For the purposes of this analysis, β-values were adjusted to β-value*100 in order to examine the change in the odds ratio (OR) for each percent increase in methylation. This transformation does not change the underlying distribution of the data or the sensitivity of the model. The dependent variable was the binary outcome of either (i) no or low impairment (n = 39) or (ii) moderate or severe impairment (n = 43), derived from LPA, as described above. As confounders had been included in the first step of this analysis, the model was run with DNA methylation beta-values as the predictor variable. Sites of CpG methylation were considered to be significantly associated with LPA scores if the associated p-value was < 0.05. As the tests were run independently for the prediction of LPA score, a test for multiple test correction was not performed. Beta estimates, parameter-likelihood ORs and 95% confidence intervals (C.I.) for ORs are reported.

Sensitivity analysis

In the ELGAN cohort, autism spectrum disorder was associated with intellectual deficit [18]. The subcohort used in the current study displayed a higher prevalence of autism spectrum disorder among study participants with intellectual deficit that is higher than would be typical of extremely preterm children with intellectual deficit. Thus, to assess whether this unusually high prevalence of autism spectrum disorder among children with intellectual deficit was a source of bias, the results were compared from regression analyses that either included or excluded the eighteen study participants with autism spectrum disorder.

Results

Study cohort

Maternal demographic data, pregnancy characteristics, and data on birth and later in life outcomes are presented for the ELGAN subjects used in this analysis (n = 59 spontaneous EPTB subjects, n = 25 indicated EPTB subjects) (Table 1) as well as for the larger ELGAN cohort (n = 1,506). Data are presented as the number (%) of subjects in the cohort unless otherwise noted. There were no significant differences in maternal age, maternal race, maternal body mass index (BMI), maternal education level, maternal status of public health insurance, Apgar and maternal exposure to smoke during pregnancy (active or passive) for spontaneous versus indicated EPTBs. The average gestational age of indicated EPTB infants was 1 week greater than that of spontaneous EPTB infants. However, the average birthweight of spontaneous EPTB infants was higher than that of indicated EPTB infants. Among the spontaneous EPTB group, there was also a higher proportion of male births (n = 44, 74.6%) than in the indicated EPTB group (n = 14, 56%).

thumbnail
Table 1. Subject characteristics.

Maternal demographic data, pregnancy characteristics, and data on birth and later in life outcomes are presented for the ELGAN subjects used in this analysis. Data are presented as the number (%) of subjects in the cohort unless otherwise noted.

https://doi.org/10.1371/journal.pone.0193271.t001

Differences in placental DNA methylation of spontaneous versus indicated EPTB

Mixed effect regression analysis was run where confounders were included in the mixed effect model analysis if they differed between subjects who had spontaneous versus indicated EPTB. Comparison of genome-wide CpG methylation levels in placentas from spontaneous versus indicated EPTB identified 250 differentially methylated probes (DMPs) (e.g. CpG sites) (S1 Table). These 250 DMPs were associated with 217 unique genes (S1 Table). Among these, 249 DMPs (99.6%) were hypermethylated, and one DMP (0.4%) was hypomethylated in association with spontaneous EPTB (Fig 1a). A set of three DMPs remained significant after SVA adjustment for cell type (S1 Table). As spontaneous EPTB is characterized by intrauterine inflammation [16], these hypermethylation marks could be marks associated with an inflammatory signature.

thumbnail
Fig 1.

(a) Heatmap of 250 DMPs between placentas from indicated EPTB (n = 25) and spontaneous EPTB (n = 59). Beta-values were mean standardized and red indicates increased methylation levels, while blue indicates decreased methylation levels. Significance was defined as FDR q-value < 0.05 and an absolute beta difference ≥ |0.10|. (b) Intragene probe site distribution for all annotated probes (n = 286,410) contained on the Illumina HumanMethylation450 BeadChip. (c) Intragene probe site distribution for 250 DMPs between placentas from indicated EPTB (n = 25) and spontaneous EPTB (n = 59). The distribution of DMPs associated with intrauterine inflammation contained more probes located within the body region of genes and less probes located within the TSS200 region of genes than would be expected from a random sample of the total probe distribution, as indicated (*).

https://doi.org/10.1371/journal.pone.0193271.g001

Interestingly, there were significantly more DMPs in the body region of genes and significantly less DMPs in the TSS200 region of genes than would be expected from the distribution of all annotated probes on the methylation array (Fig 1b and 1c). These results suggest that DNA methylation within the placenta is not random, and also that spontaneous EPTB is associated with hypermethylation outside of gene promoter regions.

Hypermethylation in genes associated with neuronal development and function

Pathway-based (e.g. Ingenuity) analysis of the 217 unique genes revealed significant enrichment for several canonical pathways known to play a role in neuronal development (Table 2). Among the most significant were UDP-N-acetyl-D-galactosamine biosynthesis, an essential process in producing sialylated glycosphingolipids that are the primary sialic acid transporters in the central nervous system and plays an important role in mammalian development [32, 33]; CXCR4 signaling, in which CXCR4 and its ligand CXCL12 are both released from developing microglia to aid in axon guidance [3436]; cholecystokinine/gastrin-mediated signaling, pathways implicated GABAergic neuronal function and in prenatal origins of autism-like neurodevelopmental disorders in rate, respectively [37, 38]; and protein kinase A signaling, which has been demonstrated to be necessary for maintenance of neuronal developmental plasticity [39] (Table 2). The enrichment of these pathways was validated using gene set enrichment analysis (GSEA).

thumbnail
Table 2. Enriched canonical pathways among the 217 DMP-associated genes.

https://doi.org/10.1371/journal.pone.0193271.t002

Hypermethylated DMPs are enriched for binding sites for transcription factors associated with neuronal development

Enrichment analysis among the 250 DMPs revealed significant enrichment for binding sites of transcription factors with known roles in neurodevelopment. Interestingly, six of the eight most significant transcription factors were basic helix-loop-helix (bHLH) transcription factors (Table 3). Namely, these were aryl hydrocarbon receptor nuclear translocator (ARNT), basic helix-loop-helix domain containing, class B (BHLHB2), basic helix-loop-helix protein known as Dec1, Stra13, Sharp2 or BHLHE40 (DEC1), Hey-like bHLH-transcriptional repressor (HELT), v-myc avian myelocytomatosis viral oncogene neuroblastoma derived homolog (NMYC), and upstream transcription factor (USF). Specifically, BHLHB2, DEC1, and HELT Basic helix-loop-helix hairy and enhancer of split (Hes) family transcription factors, whose primary role is maintaining the population of self-renewing neural progenitor cells throughout neural development via the Notch signaling pathway [40, 41].

thumbnail
Table 3. Transcription factors with enriched binding sites among 250 DMPs.

The most transcription factors with enriched binding sites among the 250 DMPs associated with spontaneous preterm deliveries due to intrauterine inflammation. The p-value was derived from a Z-score representing the overrepresentation of individual transcription factor binding sites and enrichment for individual transcription factor matrices compared against genomic background.

https://doi.org/10.1371/journal.pone.0193271.t003

Hypermethylation of DMPs predicts cognitive function at ten years of age

Logistic regression analysis revealed that 17 of the 250 DMPs predicted moderate to severe cognitive impairment at ten years of age (Fig 2, S2 Table). All of these sites were hypermethylated in association with spontaneous versus indicated EPTB. In addition, at all sites, higher levels of placental methylation predicted greater cognitive impairment at ten years of age (Fig 2). These 17 sites corresponded to 16 unique genes, many of which are involved in neuronal development and function. Among these were adenylate cyclase 7 (ADCY7), Cdk5 and Abl enzyme substrate 1 (CABLES1), G protein subunit alpha o1 (GNAO1), protein kinase C zeta (PRKCZ), retional dehydrogenase 13 (RDH13), ras homolog family member F, filopodia associated (RHOF), SH3 domain binding protein 5 (SH3BP5), and syntaxin 1A (STX1A). For each of these genes, a one percent increase in methylation at their respective probe sites predicted a 4–7% increase in the odds of moderate or severe cognitive impairment at age ten.

thumbnail
Fig 2. A total of 17 probe sites where increased placental CpG methylation predicted more severe cognitive impairment at ten years of age.

These sites corresponded to 16 unique genes. The model represents the increase in the odds of moderate or severe cognitive impairment at age ten for every one percent increase in methylation at the probe site. Significance was defined as a p-value < 0.05 in a logistic regression model.

https://doi.org/10.1371/journal.pone.0193271.g002

In a sensitivity analysis that excluded children with a diagnosis of autism spectrum disorder, 12 of the 17 sites (71%) found to be significant in the original model were confirmed to predict more severe neurocognitive impairment at ten years of age (S2 Table). It is likely that the other five sites failed to reach significance in the sensitivity analysis due to the lowered power of the analysis caused by the restricted sample size.

Discussion

Infants born prematurely have a higher risk of later-life cognitive impairments than those born at term, and the severity of these adverse outcomes is inversely associated with gestational age [19]. Intrauterine inflammation and its antecedents are associated with both premature birth and preterm infants’ risk of later-life cognitive impairments [6]. Although the exact mechanism by which inflammation alters neurodevelopment is unknown, the current literature postulates that placental epigenetics may be an important determinant of changes in early-life neurodevelopmental programming in response to inflammation [7, 11]. For example, DNA methylation levels of the glucocorticoid receptor (NR3C1) and 11β-hydroxysteroid dehydrogenase type 2 (HSD11B1), have been associated with adverse newborn neurobehavior [810].

In the present study, a differential CpG methylation signature was observed between placentas from spontaneous EPTB versus indicated EPTB. Specifically, 250 DMPs representing 217 genes were identified in placentas from spontaneous EPTB. Interestingly, these genes are involved in biological pathways crucial to many aspects of neural development, including axon-glia interactions, neuron migration, and developmental neuronal plasticity [32, 36, 39]. Furthermore, CpG methylation levels of critical neurodevelopment-related genes in the placenta predicted later life cognitive function. Specifically, these were ADCY7, which is involved in the regulation of neural networks [42]; GNAO1, which is the predominant guanine nucleotide-binding protein found in the brain and mutations of which have been associated with severe developmental delays [43]; PRKCZ, which promotes axon differentiation and is involved in neuronal survival, differentiation, outgrowth, and synaptic plasticity [44, 45]; and RHOF, which is essential in the beginning of neuronal dendritic spine formation and also promotes neurite retraction [46]. The results from this study indicate that epigenetic changes in the placentas from spontaneous EPTB are associated with cognitive function later in life.

Prior evidence suggests that gene-specific patterns in differential DNA methylation likely result from altered transcription factor binding, known as the “transcription factor occupancy theory” [47, 48]. To examine the possibility that DMPs observed in spontaneous versus indicated EPTB resulted from the binding of common transcriptional regulators, the 250 DMPs were analyzed and an enrichment for basic helix-loop-helix (bHLH) transcription factors was identified. These transcription factors are known to play a critical role in neuronal development [41]. Previous research has reported an association between perinatal methylation of the bHLH transcription factor HES1 in the umbilical cord and childhood cognitive function and behavior at four years of age [49]. Furthermore, genes whose methylation levels were predictive of later cognitive impairment were predicted to be regulated by these transcription factors. For example, these genes included CABLES1, whose expression is necessary for neurite outgrowth and which is a substrate of c-Abl tyrosine kinase, an important kinase in neurulation [50, 51] and STX1A, a nervous-system specific protein that contributes to neurite outgrowth and has been associated with both autism and attention deficit/hyperactivity disorder [52, 53]. The data suggest that bHLH transcription factors may play a critical role in epigenetic programming in the placenta that affects later life cognitive development.

When interpreting the results of this study, several factors should be considered. First, the present study focuses on CpG methylation in the placenta and not brain tissue. It is established that there are tissue-specific patterns of CpG methylation [54] as well as CpG sites that display conserved methylation patterns across tissues [55]. The results of this study do not suggest that CpG methylation in the placenta would be similar to CpG methylation in the fetal brain. Instead, placental CpG methylation represents a type of “biological recording” of placental signaling pathways that are critical for fetal growth and development. This work contributes to a growing body of literature showing altered placental CpG methylation in relation to infant behavior and potential deficits in cognitive function [8, 9, 5658]. Second, it is recently documented that there are significant sex-based differences in the placenta [14]. There are also established sex-based differences in neurodevelopmental outcomes in male and female adolescents [27]. To address the impact of sex as an influencing factor related to differential CpG methylation, sex was included as a covariate in the analysis. While underpowered to do so here, further research could further explore the relationship between placental DNA methylation and neurocognitive outcomes separately in male and female children. Lastly, while we have used in silico methods for placenta cell deconvolution as described previously [29], future research should aim to catalog DNA methylation patterns by placental cell type.

In summary, a CpG hypermethylation signature was identified in placental tissue from spontaneous EPTB that could mediate, at least in part, the relationship seen between preterm birth and adverse neurodevelopmental outcomes later in life. As survival rates of extremely premature infants have significantly improved with recent medical advances, the population of young children with an increased risk of cognitive impairment has grown and likely will continue to grow in the future [59]. These impairments represent a significant burden to public health, translating into billions of healthcare dollars, a lower quality of life of affected individuals and their families, and a reduction in societal productivity [59]. A total of 17 DMPs were identified between spontaneous and indicated EPTB that could be investigated for use as perinatal clinical epigenomic biomarkers of these later-life impairments. While current standard practice relies on childhood cognitive tests to identify cognitive impairment, identification of neonates at highest risk for adverse neurodevelopmental outcomes could present opportunities for earlier interventions to improve the outcomes of these children [60].

Supporting information

S1 Table. DMPs associated with EPTB.

A total of 250 probes displayed differential methylation (DM) in spontaneous versus indicated EPTB. Gene and intragene region annotations are included. Significance was defined as FDR q-value < 0.05 and an absolute median beta difference ≥ |0.10|.

https://doi.org/10.1371/journal.pone.0193271.s001

(XLSX)

S2 Table. Logistic regression and sensitivity analysis.

Logistic regression modeling identified 17 DMPs that predicted a significantly greater risk of neurocognitive impairment (LPA score = 3 or 4 versus LPA score = 1 or 2) at 10 year of age. In a sensitivity analysis excluding 18 autistic individuals, 12 (71%) DMPs remained significant.

https://doi.org/10.1371/journal.pone.0193271.s002

(XLSX)

References

  1. 1. Chan E, Leong P, Malouf R, Quigley MA. Long-term cognitive and school outcomes of late-preterm and early-term births: a systematic review. Child Care Health Dev. 2016;42(3):297–312. Epub 2016/02/11. pmid:26860873.
  2. 2. Kuban KC, O'Shea TM, Allred EN, Paneth N, Hirtz D, Fichorova RN, et al. Systemic inflammation and cerebral palsy risk in extremely preterm infants. J Child Neurol. 2014;29(12):1692–8. Epub 2014/03/22. pmid:24646503
  3. 3. Leviton A, Allred EN, Fichorova RN, Kuban KC, Michael O'Shea T, Dammann O, et al. Systemic inflammation on postnatal days 21 and 28 and indicators of brain dysfunction 2years later among children born before the 28th week of gestation. Early Hum Dev. 2016;93:25–32. Epub 2016/01/07. pmid:26735345
  4. 4. O'Shea TM, Allred EN, Kuban KC, Dammann O, Paneth N, Fichorova R, et al. Elevated concentrations of inflammation-related proteins in postnatal blood predict severe developmental delay at 2 years of age in extremely preterm infants. J Pediatr. 2012;160(3):395–401 e4. Epub 2011/10/18. pmid:22000304
  5. 5. O'Shea TM, Joseph RM, Kuban KC, Allred EN, Ware J, Coster T, et al. Elevated blood levels of inflammation-related proteins are associated with an attention problem at age 24 mo in extremely preterm infants. Pediatr Res. 2014;75(6):781–7. Epub 2014/03/13. pmid:24614800
  6. 6. O'Shea TM, Shah B, Allred EN, Fichorova RN, Kuban KC, Dammann O, et al. Inflammation-initiating illnesses, inflammation-related proteins, and cognitive impairment in extremely preterm infants. Brain Behav Immun. 2013;29:104–12. Epub 2013/01/09. pmid:23295265
  7. 7. Nugent BM, Bale TL. The omniscient placenta: Metabolic and epigenetic regulation of fetal programming. Front Neuroendocrinol. 2015;39:28–37. Epub 2015/09/15. pmid:26368654
  8. 8. Bromer C, Marsit CJ, Armstrong DA, Padbury JF, Lester B. Genetic and epigenetic variation of the glucocorticoid receptor (NR3C1) in placenta and infant neurobehavior. Dev Psychobiol. 2013;55(7):673–83. Epub 2012/06/21. pmid:22714792
  9. 9. Conradt E, Lester BM, Appleton AA, Armstrong DA, Marsit CJ. The roles of DNA methylation of NR3C1 and 11beta-HSD2 and exposure to maternal mood disorder in utero on newborn neurobehavior. Epigenetics. 2013;8(12):1321–9. Epub 2013/10/19. pmid:24135662
  10. 10. Marsit CJ, Maccani MA, Padbury JF, Lester BM. Placental 11-beta hydroxysteroid dehydrogenase methylation is associated with newborn growth and a measure of neurobehavioral outcome. PLoS One. 2012;7(3):e33794. Epub 2012/03/21. pmid:22432047
  11. 11. Lesseur C, Paquette AG, Marsit CJ. Epigenetic Regulation of Infant Neurobehavioral Outcomes. Med Epigenet. 2014;2(2):71–9. Epub 2014/08/05. pmid:25089125
  12. 12. Cardenas A, Houseman EA, Baccarelli AA, Quamruzzaman Q, Rahman M, Mostofa G, et al. In utero arsenic exposure and epigenome-wide associations in placenta, umbilical artery, and human umbilical vein endothelial cells. Epigenetics. 2015;10(11):1054–63. Epub 2015/12/10. pmid:26646901
  13. 13. Paquette AG, Houseman EA, Green BB, Lesseur C, Armstrong DA, Lester B, et al. Regions of variable DNA methylation in human placenta associated with newborn neurobehavior. Epigenetics. 2016;11(8):603–13. Epub 2016/07/02. pmid:27366929
  14. 14. Martin E, Smeester L, Bommarito PA, Grace MR, Boggess K, Kuban K, et al. Sexual epigenetic dimorphism in the human placenta: implications for susceptibility during the prenatal period. Epigenomics. 2017;9(3):267–78. Epub 2017/02/25. pmid:28234023
  15. 15. Romero R, Espinoza J, Kusanovic JP, Gotsch F, Hassan S, Erez O, et al. The preterm parturition syndrome. BJOG. 2006;113 Suppl 3:17–42. Epub 2007/01/09. pmid:17206962.
  16. 16. McElrath TF, Hecht JL, Dammann O, Boggess K, Onderdonk A, Markenson G, et al. Pregnancy disorders that lead to delivery before the 28th week of gestation: an epidemiologic approach to classification. Am J Epidemiol. 2008;168(9):980–9. Epub 2008/08/30. pmid:18756014
  17. 17. O'Shea TM, Allred EN, Dammann O, Hirtz D, Kuban KC, Paneth N, et al. The ELGAN study of the brain and related disorders in extremely low gestational age newborns. Early Hum Dev. 2009;85(11):719–25. Epub 2009/09/22. pmid:19765918
  18. 18. Joseph RM, O'Shea TM, Allred EN, Heeren T, Hirtz D, Paneth N, et al. Prevalence and associated features of autism spectrum disorder in extremely low gestational age newborns at age 10 years. Autism Res. 2017;10(2):224–32. Epub 2016/05/26. pmid:27220677
  19. 19. Joseph RM, O'Shea TM, Allred EN, Heeren T, Hirtz D, Jara H, et al. Neurocognitive and Academic Outcomes at Age 10 Years of Extremely Preterm Newborns. Pediatrics. 2016;137(4). Epub 2016/03/24. pmid:27006473
  20. 20. Onderdonk AB, Delaney ML, DuBois AM, Allred EN, Leviton A, Extremely Low Gestational Age Newborns Study I. Detection of bacteria in placental tissues obtained from extremely low gestational age neonates. Am J Obstet Gynecol. 2008;198(1):110 e1–7. Epub 2008/01/02. pmid:18166321.
  21. 21. Bibikova M, Barnes B, Tsan C, Ho V, Klotzle B, Le JM, et al. High density DNA methylation array with single CpG site resolution. Genomics. 2011;98(4):288–95. Epub 2011/08/16. pmid:21839163.
  22. 22. Teschendorff AE, Marabita F, Lechner M, Bartlett T, Tegner J, Gomez-Cabrero D, et al. A beta-mixture quantile normalization method for correcting probe design bias in Illumina Infinium 450 k DNA methylation data. Bioinformatics. 2013;29(2):189–96. Epub 2012/11/24. pmid:23175756
  23. 23. Pidsley R, YW CC, Volta M, Lunnon K, Mill J, Schalkwyk LC. A data-driven approach to preprocessing Illumina 450K methylation array data. BMC Genomics. 2013;14:293. Epub 2013/05/02. pmid:23631413
  24. 24. Beran TN; Elliott CD. Differential Ability Scales (2nd ed.). Can J Sch Psychol. 2007; (22):128–32.
  25. 25. KMKUK S.. NEPSY-II: Clinical and interpretive manual. San Antonio, TX: The Pyschological Corporation. 2007.
  26. 26. Heeren T, Joseph RM, Allred EN, O'Shea TM, Leviton A, Kuban KCK. Cognitive functioning at the age of 10 years among children born extremely preterm: a latent profile approach. Pediatr Res. 2017;82(4):614–9. Epub 2017/06/06. pmid:28582386.
  27. 27. Kuban KC, Joseph RM, O'Shea TM, Allred EN, Heeren T, Douglass L, et al. Girls and Boys Born before 28 Weeks Gestation: Risks of Cognitive, Behavioral, and Neurologic Outcomes at Age 10 Years. J Pediatr. 2016;173:69–75 e1. Epub 2016/03/24. pmid:27004675
  28. 28. Du P, Zhang X, Huang CC, Jafari N, Kibbe WA, Hou L, et al. Comparison of Beta-value and M-value methods for quantifying methylation levels by microarray analysis. BMC Bioinformatics. 2010;11:587. Epub 2010/12/02. pmid:21118553
  29. 29. Houseman EA, Kile ML, Christiani DC, Ince TA, Kelsey KT, Marsit CJ. Reference-free deconvolution of DNA methylation data and mediation by cell composition effects. BMC Bioinformatics. 2016;17:259. Epub 2016/07/01. pmid:27358049
  30. 30. Subramanian A, Kuehn H, Gould J, Tamayo P, Mesirov JP. GSEA-P: a desktop application for Gene Set Enrichment Analysis. Bioinformatics. 2007;23(23):3251–3. Epub 2007/07/24. pmid:17644558.
  31. 31. Ho Sui SJ, Mortimer JR, Arenillas DJ, Brumm J, Walsh CJ, Kennedy BP, et al. oPOSSUM: identification of over-represented transcription factor binding sites in co-expressed genes. Nucleic Acids Res. 2005;33(10):3154–64. Epub 2005/06/04. pmid:15933209
  32. 32. Yamashita T, Wu YP, Sandhoff R, Werth N, Mizukami H, Ellis JM, et al. Interruption of ganglioside synthesis produces central nervous system degeneration and altered axon-glial interactions. Proc Natl Acad Sci U S A. 2005;102(8):2725–30. Epub 2005/02/16. pmid:15710896
  33. 33. Tian E, Ten Hagen KG. Recent insights into the biological roles of mucin-type O-glycosylation. Glycoconj J. 2009;26(3):325–34. Epub 2008/08/13. pmid:18695988
  34. 34. Bilbo SD, Smith SH, Schwarz JM. A lifespan approach to neuroinflammatory and cognitive disorders: a critical role for glia. J Neuroimmune Pharmacol. 2012;7(1):24–41. Epub 2011/08/09. pmid:21822589
  35. 35. Hagihara K, Zhang EE, Ke YH, Liu G, Liu JJ, Rao Y, et al. Shp2 acts downstream of SDF-1alpha/CXCR4 in guiding granule cell migration during cerebellar development. Dev Biol. 2009;334(1):276–84. Epub 2009/07/29. pmid:19635473
  36. 36. Tiveron MC, Cremer H. CXCL12/CXCR4 signalling in neuronal cell migration. Curr Opin Neurobiol. 2008;18(3):237–44. Epub 2008/07/23. pmid:18644448.
  37. 37. Merali Z, Presti-Torres J, Mackay JC, Johnstone J, Du L, St-Jean A, et al. Long-term behavioral effects of neonatal blockade of gastrin-releasing peptide receptors in rats: similarities to autism spectrum disorders. Behav Brain Res. 2014;263:60–9. Epub 2014/01/28. pmid:24462726.
  38. 38. Ramamoorthi K, Lin Y. The contribution of GABAergic dysfunction to neurodevelopmental disorders. Trends Mol Med. 2011;17(8):452–62. Epub 2011/04/26. pmid:21514225
  39. 39. Cai D, Qiu J, Cao Z, McAtee M, Bregman BS, Filbin MT. Neuronal cyclic AMP controls the developmental loss in ability of axons to regenerate. J Neurosci. 2001;21(13):4731–9. Epub 2001/06/27. pmid:11425900.
  40. 40. Imayoshi I, Kageyama R. Oscillatory control of bHLH factors in neural progenitors. Trends Neurosci. 2014;37(10):531–8. Epub 2014/08/26. pmid:25149265.
  41. 41. Imayoshi I, Kageyama R. bHLH factors in self-renewal, multipotency, and fate choice of neural progenitor cells. Neuron. 2014;82(1):9–23. Epub 2014/04/05. pmid:24698265.
  42. 42. Joeyen-Waldorf J, Nikolova YS, Edgar N, Walsh C, Kota R, Lewis DA, et al. Adenylate cyclase 7 is implicated in the biology of depression and modulation of affective neural circuitry. Biol Psychiatry. 2012;71(7):627–32. Epub 2012/01/24. pmid:22264442
  43. 43. Saitsu H, Fukai R, Ben-Zeev B, Sakai Y, Mimaki M, Okamoto N, et al. Phenotypic spectrum of GNAO1 variants: epileptic encephalopathy to involuntary movements with severe developmental delay. Eur J Hum Genet. 2016;24(1):129–34. Epub 2015/05/15. pmid:25966631
  44. 44. Joung I, Kim HJ, Kwon YK. p62 modulates Akt activity via association with PKCzeta in neuronal survival and differentiation. Biochem Biophys Res Commun. 2005;334(2):654–60. Epub 2005/07/14. pmid:16011831.
  45. 45. Zhang X, Zhu J, Yang GY, Wang QJ, Qian L, Chen YM, et al. Dishevelled promotes axon differentiation by regulating atypical protein kinase C. Nat Cell Biol. 2007;9(7):743–54. Epub 2007/06/15. pmid:17558396.
  46. 46. Fan L, Yan H, Pellegrin S, Morigen , Mellor H. The Rif GTPase regulates cytoskeletal signaling from plexinA4 to promote neurite retraction. Neurosci Lett. 2015;590:178–83. Epub 2015/02/11. pmid:25668492.
  47. 47. Martin EM, Fry RC. A cross-study analysis of prenatal exposures to environmental contaminants and the epigenome: support for stress-responsive transcription factor occupancy as a mediator of gene-specific CpG methylation patterning. Environ Epigenet. 2016;2(1). Epub 2016/04/12. pmid:27066266
  48. 48. Sanders AP, Smeester L, Rojas D, DeBussycher T, Wu MC, Wright FA, et al. Cadmium exposure and the epigenome: Exposure-associated patterns of DNA methylation in leukocytes from mother-baby pairs. Epigenetics. 2014;9(2):212–21. Epub 2013/10/31. pmid:24169490
  49. 49. Lillycrop KA, Costello PM, Teh AL, Murray RJ, Clarke-Harris R, Barton SJ, et al. Association between perinatal methylation of the neuronal differentiation regulator HES1 and later childhood neurocognitive function and behaviour. Int J Epidemiol. 2015;44(4):1263–76. Epub 2015/04/25. pmid:25906782
  50. 50. Koleske AJ, Gifford AM, Scott ML, Nee M, Bronson RT, Miczek KA, et al. Essential roles for the Abl and Arg tyrosine kinases in neurulation. Neuron. 1998;21(6):1259–72. Epub 1999/01/12. pmid:9883720.
  51. 51. Zukerberg LR, Patrick GN, Nikolic M, Humbert S, Wu CL, Lanier LM, et al. Cables links Cdk5 and c-Abl and facilitates Cdk5 tyrosine phosphorylation, kinase upregulation, and neurite outgrowth. Neuron. 2000;26(3):633–46. Epub 2000/07/15. pmid:10896159.
  52. 52. Ghiani CA, Starcevic M, Rodriguez-Fernandez IA, Nazarian R, Cheli VT, Chan LN, et al. The dysbindin-containing complex (BLOC-1) in brain: developmental regulation, interaction with SNARE proteins and role in neurite outgrowth. Mol Psychiatry. 2010;15(2):115, 204–15. Epub 2009/06/24. pmid:19546860
  53. 53. Nakamura K, Iwata Y, Anitha A, Miyachi T, Toyota T, Yamada S, et al. Replication study of Japanese cohorts supports the role of STX1A in autism susceptibility. Prog Neuropsychopharmacol Biol Psychiatry. 2011;35(2):454–8. Epub 2010/12/02. pmid:21118708.
  54. 54. Ghosh S, Yates AJ, Fruhwald MC, Miecznikowski JC, Plass C, Smiraglia D. Tissue specific DNA methylation of CpG islands in normal human adult somatic tissues distinguishes neural from non-neural tissues. Epigenetics. 2010;5(6):527–38. pmid:20505344
  55. 55. Woodfine K, Huddleston JE, Murrell A. Quantitative analysis of DNA methylation at all human imprinted regions reveals preservation of epigenetic stability in adult somatic tissue. Epigenetics & chromatin. 2011;4(1):1. pmid:21281512
  56. 56. Paquette AG, Lester BM, Lesseur C, Armstrong DA, Guerin DJ, Appleton AA, et al. Placental epigenetic patterning of glucocorticoid response genes is associated with infant neurodevelopment. Epigenomics. 2015;7(5):767–79. pmid:26343289
  57. 57. Paquette AG, Lester BM, Koestler DC, Lesseur C, Armstrong DA, Marsit CJ. Placental FKBP5 genetic and epigenetic variation is associated with infant neurobehavioral outcomes in the RICHS cohort. PloS one. 2014;9(8):e104913. pmid:25115650
  58. 58. Monk C, Feng T, Lee S, Krupska I, Champagne FA, Tycko B. Distress During Pregnancy: Epigenetic Regulation of Placenta Glucocorticoid-Related Genes and Fetal Neurobehavior. Am J Psychiatry. 2016;173(7):705–13. pmid:27013342
  59. 59. Behrman R. and Butler A. Preterm Birth: Causes, Consequences, and Prevention. National Academies Press (US); 2007.
  60. 60. Guralnick MJ. Early Intervention for Children with Intellectual Disabilities: An Update. J Appl Res Intellect Disabil. 2016. Epub 2016/01/15. pmid:26764896.