Low birth weight associates with hippocampal gene expression

Neuroscience 318 (2016) 190–205

LOW BIRTH WEIGHT ASSOCIATES WITH HIPPOCAMPAL GENE EXPRESSION J. P. BUSCHDORF, a* M. L. ONG, a S. X. ONG, a J. L. MACISAAC, b K. CHNG, a M. S. KOBOR, b,c M. J. MEANEY a,d AND J. D. HOLBROOK a

Key words: low birth weight, hippocampus, DNA methylation, mental disorders, microarray.

a Singapore Institute for Clinical Sciences, Agency for Science, Technology and Research (A*STAR), 30 Medical Drive, Singapore 117609, Singapore

INTRODUCTION

b Centre for Molecular Medicine and Therapeutics, Child and Family Research Institute, Department of Medical Genetics, University of British Columbia, 950 West 28th Avenue, Vancouver, BC V5Z 4H4, Canada

Fetal growth and development, reflected in birth weight, are influenced by the quality of the prevailing environment and predictive of later metabolic health (Barker et al., 1993a,b; Rich-Edwards et al., 1997). Low birth weight also associates with the risk for multiple brain-based disorders such as attention-deficit hyperactivity disorder (Bhutta et al., 2002), depression (Thompson et al., 2001; Alati et al., 2007), and schizophrenia (Abel et al., 2010) as well as with endophenotypes for mental disorders such as the thickness of regional cortical surface areas, total brain and caudate volumes and cognitive abilities (Raznahan et al., 2012; Walhovd et al., 2012). These effects are present across the birth weight spectrum and linked to tissue- and gene-specific epigenetic modifications (Meaney et al., 2007). Alterations in DNA methylation have been associated with metabolic diseases like type 2 diabetes (Park et al., 2008) as well as mental disorders like schizophrenia and bipolar disorder (Dempster et al., 2011; Xiao et al., 2014). Importantly, DNA methylation can be influenced by internal and external environmental cues such as diet, stress and glucocorticoid secretion, in particular during critical periods of prenatal and early postnatal development (Meaney et al., 2007). Modifications of DNA methylation can be stable and may relate to life-long phenotypic consequences, potentially including increased risk for the development of disease (Gluckman et al., 2009). Whereas changes in gene expression and DNA methylation have been detected in postmortem brain samples of subjects suffering from psychiatric disorders (Grayson and Guidotti, 2013) as well as in response to clinically relevant early environments (McGowan et al., 2009; Provencal et al., 2012), little is known about the adaptations early in development that occur in clinically relevant brain regions in response to the conditions that associate with low birth weight. Such changes would potentially set up a mechanism for a predisposition for an increased risk of disease development established during fetal life. The hippocampus is implicated in anxiety, depression, and schizophrenia (Small et al., 2011; Femenia et al., 2012), all of which associate with birth weight. Low birth weight within the normal range associates with decreased hippocampal volume (Buss et al., 2007) and impaired

c Human Early Learning Partnership, School of Population and Public Health, University of British Columbia, Canada d Canadian Neuroepigenetics Network, Douglas Mental Health University Institute, McGill University, 6875 LaSalle Boulevard, Montreal, Quebec H4H 1R3, Canada

Abstract—Birth weight predicts the lifetime risk for psychopathology suggesting that the quality of fetal development influences the predisposition for mental disorders. The connectivity and synaptic network of the hippocampus are implicated in depression, schizophrenia and anxiety. We thus examined the underlying molecular adaptations in the hippocampus as a function of the fetal conditions associated with low birth weight. We used tissues from the nonhuman primate, Macaca fascicularis, to identify changes in hippocampal gene expression early in postnatal development associated with naturally occurring low compared with normal birth weight. Microarrays were used to analyze gene expression and DNA methylation in the hippocampus of five low- and five normal-birth weight neonates. Real-time PCR was employed to validate differentially expressed genes. Birth weight associated with altered global transcription in the hippocampus. Hierarchical clustering of gene expression profiles from 24,154 probe sets grouped all samples except one by their birth weight status. Differentially expressed genes were enriched in biological processes associated with neuronal projection, positive regulation of transcription and apoptosis. About 4% of the genes with differential expression co-varied with DNA methylation levels. The data suggest that low birth weight is closely associated with hippocampal gene expression with a small epigenetic underpinning by DNA methylation in neonates. The data also provide a potential molecular basis for the developmental origin of an enhanced risk for mental disorders. Ó 2016 IBRO. Published by Elsevier Ltd. All rights reserved.

*Corresponding author. Tel: +65-6407-0568; fax: +65-67766840. E-mail address: [email protected] (J. P. Buschdorf). Abbreviations: DAVID, Database for Annotation, Visualization, and Integrated Discovery; LBW, low birth weight; NBW, normal birth weight; qPCR, quantitative real-time PCR. http://dx.doi.org/10.1016/j.neuroscience.2016.01.013 0306-4522/Ó 2016 IBRO. Published by Elsevier Ltd. All rights reserved. 190

191

J. P. Buschdorf et al. / Neuroscience 318 (2016) 190–205

performance on hippocampal-dependent tasks (Strauss, 2000; Kirkegaard et al., 2006; Gieling et al., 2012). Studies with rodent or non-human primate models show that prenatal exposure to elevated glucocorticoid levels impairs growth, resulting in a lower birth weight, and constrains hippocampal development (Uno et al., 1989, 1994; Coe et al., 2003; Lister et al., 2005). We used the non-human primate, Macaca fascicularis to identify changes in hippocampal gene expression and DNA methylation early in development associated with naturally-occurring instances of low birth weight (LBW) as compared to normal birth weight (NBW). Our data suggest that impaired fetal growth associates with altered global hippocampal transcription which is not broadly accompanied with DNA methylation changes.

EXPERIMENTAL PROCEDURES Animals and collection of brain tissue All animals were healthy. They were bred and sacrificed at the Nafovanny facility (Vietnam), which is a facility accredited by the Association for Assessment and Accreditation of Laboratory Animal Care International (AAALAC International). Dams and offspring were exposed to the local climate. There were no differences in the diet between dams. All neonates in this study were female and from unique matings. They were sedated between 09:20 and 11:15 am with an intramuscular injection of ketamine-HCl (15 mg/kg), and exsanguinated under anesthesia. Whole brains were collected, flash frozen and stored at
80 °C. Tissue processing Coronal sections from the right hippocampus were prepared in a series of two sections of 300 lm and 20 lm respectively by cryosectioning and thaw-mounted on polylysine-coated slides. Slides were stored at minus 80 °C until further processing for Nissl staining and nucleic acid extraction. Nissl staining facilitated the identification of the region of interest. We defined the anterior hippocampus as the hippocampal part that is present in the coordinates A9.6 to A5.6 of the stereotaxic brain atlas by Szabo and Cowan (1984). The lateral geniculate nucleus was used as a landmark (Fig. 1). A standard protocol was used for Nissl staining of 20 lm brain sections. Pictures were taken with an Axio Observer Z1 microscope (Zeiss, Germany) and stitched with TissueFAXS suite version 3 software (TissueGnostics, Austria). RNA and DNA extraction The area used for nucleic acid extraction from the hippocampus is indicated in Fig. 1. Total RNA and DNA was isolated from a single 300 lm section using the All Prep DNA/RNA Micro Kit (Qiagen, Singapore) following the manufacturer’s protocol, including DNase digestion for RNA samples. The nucleic acid samples were subjected to spectrophotometric measurement and the quality of RNA (RIN values) determined using an Agilent

Fig. 1. Representative Nissl staining of female Cynomolgus macaque neonate at dots demarcate the region analyzed in represents 1 mm. GL, dorsal lateral hippocampus.

the hippocampus from a the age of 6.6 days. Black our study; the scale bar geniculate nucleus; Hc,

Bioanalyzer. All RNA samples had an RIN value between 7.7 and 9.1. Expression microarray Five LBW and five NBW samples (plus a technical and a biological replicate) were hybridized to Affymetrix Rhesus Genome Arrays (ORIGIN LABS, Singapore). The RNA samples were processed with the Affymetrix 3’ IVT Express Target Amplification, Labeling and Control Kit according to the manufacturer’s protocol. Briefly, 60 ng of total RNA was reverse transcribed to produce cDNA, which was subsequently used as a template to create biotin-labeled aRNA (amplified RNA). Four unlabeled, polyadenylated RNA spikes in different concentrations were included to assess the target preparation steps. The aRNA was then fragmented and hybridized to Affymetrix Rhesus Genome Arrays for 16 h at 45 °C with rotation at 60 rpm. Four pre-labeled bacterial hybridization controls were included as well. Arrays were washed and stained with the Affymetrix GeneChip Hybridization, Wash and Stain Kit and scanned using an Affymetrix 3000 7G scanner. All controls were identified on all arrays and showed signal intensities corresponding to their initial relative abundance (data not shown). Expression microarray data analysis Raw probe set intensities were normalized using the GCRMA (GC content – Robust Multi-Array Average) algorithm (Wu et al., 2004) in the Affymetrix R package and were log2-transformed. After filtering out probesets for which at least one sample was non-expressing (expression below 3.5), 24,154 probesets remained.

192

J. P. Buschdorf et al. / Neuroscience 318 (2016) 190–205

Hierarchical clustering was performed on the 24,154 probesets in Array Studio (OmicSoft) using the Pearson correlation dissimilarity measure and Ward’s minimum variance method. We also carried out principal components analysis on the 24,154 probesets in Array Studio (OmicSoft). Before performing the differential expression analysis across birth weight, we further removed probesets without grade ‘‘A” annotation (Affymetrix, 2006). This resulted in 19,186 remaining probesets. We performed one-way ANOVA and regression on these probesets based on either birth weight groups or actual birth weight levels in Array Studio (OmicSoft). We retained only probesets that passed a significance level of P < 0.05 in both the regression and the two-group analyses. We further applied a fold-change cut-off of 1.5 to reduce the number of candidate probesets to 608. P-values were adjusted for multiple testing using the Benjamini-Hochberg method (Benjamini and Hochberg, 1995). We performed Gene Ontology enrichment analysis of the 451 differentially expressed (DE) genes in DAVID (Database for Annotation, Visualization, and Integrated Discovery) (Dennis et al., 2003) using default parameters. Significance was assessed using a hypergeometric distribution test and was corrected for multiple testing using the BenjaminiHochberg method (Benjamini and Hochberg, 1995). Pathway enrichment analyses were performed in GeneGo (Metacore 5.0) against the background list of human genes that mapped to the rhesus chip, and significance was determined in a similar manner as described for Gene Ontology.

Quantitative real-time PCR (qPCR) First strand synthesis of cDNA was carried out with 200 ng total RNA and oligo-dT primers using the Transcriptor First Strand cDNA synthesis Kit (Roche, Singapore). cDNAs were tested for genomic contamination with a genomic contamination primer assay (Qiagen, Singapore). qPCR was performed on an LC480 II Real-Time PCR instrument with Light Cycler 480 SYBR Green I Mastermix (Roche) or QuantiFast SYBR Green Mastermix (Qiagen, Singapore). Primer efficiencies were determined with CAmpER software (www.cebitec.uni-bielefeld.de) using the DART method (Peirson et al., 2003). Quantification was performed with the Advanced Relative Quantification application of the Light Cycler 480 software 1.5 (Roche, Singapore). Samples were analyzed in triplicates and normalized to the expression of b2microglobulin. Two samples (one NBW sample, no 5 and one LBW sample, no 10) were removed from the qPCR analysis because of multiple outliers. Outliers were defined as normalized expression values more than two standard deviations away from the group mean of the remaining values. The removal of samples with outliers resulted in equal average age (4.8 days for both groups). One-tailed Student’s t-tests were used to statistically compare differences in gene expression between samples from low- and normalbirth weight neonates with significance set at a value of P < 0.05.

DNA methylation array analysis 1 lg of genomic DNA was bisulfite converted using EZ-96 DNA MethylationTM Gold Kit (Catalogue No.: D5007, Zymo Research). Successful conversion was confirmed via methylation-specific PCR prior to proceeding with subsequent steps of the Infinium assay protocol. The bisulfite converted genomic DNA was isothermally amplified at 37 °C for 22 h, enzymatically fragmented, purified and hybridized on an InfiniumÒ HumanMethylation 450 BeadChip (Catalogue No.: WD314-1002, Illumina Inc.) at 48 °C for 18 h. The BeadChip was then washed to remove any un-hybridized or nonspecific hybridized DNA. Labeled single-base extension was performed on primers hybridized with DNA and the hybridized DNA was removed. The extended primers were stained with multiple layers of fluorescence, the BeadChip was then coated using a proprietary solution and scanned using the IlluminaÒ iScan system. The image data were processed with the Genome StudioTM Methylation Module software. Further data processing was performed as described previously (Ong et al., 2014) to obtain the beta values. After this step, 338,803 (70%) probes with a P-value detection of <0.05 were retained. Infinium 450K data filtering The probes designed on the Infinium 450K array are designed for the human genome. Therefore, we performed a three-step filter to remove divergent probes that could bias the methylation measurements. The filtering is based on sequence alignment score, uniqueness of the sequence match and the position of the mismatch relative to the CpG being measured (Ong et al., 2014). This resulted in the retention of 205,990 (42.4%) probes for subsequent differential analyses. We also obtained the equivalent macaque ortholog gene name associated with these CpGs from (Ong et al., 2014). Correlation analysis of expression with methylation levels For each of the differentially expressed genes, we determined the available gene-related CpGs on the Infinium 450K array after filtering for divergent probes. We then computed the pairwise Pearson correlation between expression level and each of the corresponding CpG across the samples. And for each gene with at least one CpG typed on the array, we corrected for the total number of CpG:gene expression tests by using a corrected P-value cutoff of 0.05/(number of CpGs present in gene). Cell-type correction analysis We used 5,141 available differentially methylated probes for neuron and glia identified in Guintivano et al. (2013) and applied their method in R on the Infinium 450K Infinium data to estimate the cell type proportion for each sample.

193

J. P. Buschdorf et al. / Neuroscience 318 (2016) 190–205

3.0 20-Oct-07 10 13-May-08 326 6.6

Study approval

RESULTS

3.2 19-Oct-07 5 29-May-08 351 4.6 3.2 19-Oct-07 4 20-May-08 368 7.3 3.1 27-Oct-07 3 16-Apr-08 358 3.6 3.3 20-Oct-07 2 15-Apr-08 371 5.6 299–480 NBW group

33/32 4 63/2 358 (40.7)/351

Original cohort (n = 65)

3.4 22-Oct-07 1 9-Apr-08 358 2.5 361 (8.2) 50-65th 4.7 4.8 Dam’s weight proir to breeding in kg and Date weight was measured Neonate ID Date killed Birth weight in g Age in days Mean birth weight in g and (standard deviation) Birth weight percentile of the original female cohort Mean age in days – microarray Mean age in days – qPCR

We first determined whether hippocampal gene expression patterns differed between the two groups. RNA from the five LBW and five NBW samples as well as a technical and a biological replicate were hybridized to Affymetrix Rhesus Genome Arrays. It was not known whether birth weight-specific signatures could be identified from gene expression data and accordingly whether different birth weight samples could be classified based on gene expression patterns. We therefore performed an unsupervised hierarchical

Number of females/males Number of still births Number of singletons/twins Mean (standard deviation)/median birth weight of females (excl still births and twins) in g Range of birth weights in g

Hippocampal gene expression clustered according to birth weight

Table 1. Cohort and sample details of all animals (Macaca fascicularis)

Sixty-four pregnant Cynomolgus macaque dams (Macaca fascicularis) were monitored prior to delivery at the Vietnam Primate Breeding and Development Corporation. These 64 pregnant dams gave birth to 65 male and female neonates (original cohort of neonates; Table1). The normative birth weight range was assessed from single births and 10 female neonates were selected based on their birth weights to comprise two groups (Table1): (1) LBW group with n = 5 and a birth weight range of 314–326 g (4th to 27th birth weight percentile of the original female cohort) and (2) NBW group with n = 5 and a birth weight range of 351–371 g (50th to 65th birth weight percentile of the original female cohort). The average age of the NBW group was 4.7 days; that of the LBW group was 5.1 days. The respective mean birth weights differed significantly (t8 = 9.77, p < 0.0001; mean = 361 g, SEM = 3.7 g for NBW versus mean = 319 g, SEM = 2.3 g for LBW). The dams’ weight prior to breeding (Table 1) correlated significantly with the neonates’ birth weight (Pearson’s correlation coefficient of 0.695, P-value: 0.026). As hippocampal development is affected by gestational conditions in model organisms (Mallard et al., 2000; Lister et al., 2005; Noorlander et al., 2014) and as we had previously found that birth weight affects hippocampal mineralocortiocid:glucocorticoid receptor ratio in the same set of animals (Ong et al., 2013), we chose to study the anterior hippocampus dissected from the right temporal lobe (Fig. 1).

LBW group

Primate birth weight model

2.8 22-Oct-07 6 9-Apr-08 316 4.5 319 (5.1) 4-27th 5.1 4.8

3.3 20-Oct-07 7 16-Apr-08 323 3.6

2.9 26-Oct-07 8 24-Apr-08 317 6.5

3.0 20-Oct-07 9 9-May-08 314 4.5

Our procedures were approved by the attending veterinary board of Nafovanny (Vietnam) and performed in accordance with the Guidelines on the Care and Use of Animals for Scientific Purposes set by the National Advisory Committee for Laboratory Animal Research (NACLAR) of Singapore which is based on the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes, NHMRC, Australia, as well as the Guide for the Care and Use of Laboratory Animals by National Academy of Sciences, USA, and The Good Practice Guide for the Use of Animals in Research, Testing and Teaching, National Animal Ethics Advisory Committee, New Zealand.

194

J. P. Buschdorf et al. / Neuroscience 318 (2016) 190–205

clustering using the 24,154 expressed probesets across the twelve samples. Importantly, the dendrogram quite accurately grouped the twelve samples by their birth weight status (Fig. 2A). There was only one mismatch,

where a single NBW sample was clustered into the LBW group. We also noted that all the replicates were clustered together and the distances between replicates (for IDs 10 and 8) were shorter than those between

Fig. 2. Comparison of gene expression profiles of low and normal birth weight samples. Hierarchical clustering (a) shows samples as columns and probesets as rows. Logged expression intensities of probesets are shown in the heatmap as green (low) to red (high). Clustering dendogram of samples is above. Sample classification is denoted in the color bar. Gray color indicates LBW samples, white color indicates NBW samples. Principal component analysis (b), component one is on the x-axis and component 2 is on the y-axis, together they explain 38% of variation in the dataset. Gray color: LBW samples; white color: NBW samples.

J. P. Buschdorf et al. / Neuroscience 318 (2016) 190–205

individual samples. Using principal component analysis, we observed a similar pattern, where all but one sample (matching the outlier in the hierarchical clustering method) was clearly grouped according to birth weight by principal component 1 (Fig. 2B). Overall, hippocampal gene expression patterns in neonatal macaques served to successfully cluster the samples by birth weight indicating remarkable sensitivity of the hippocampus transcriptome to the conditions that cause low birth weight. Identification and validation of differentially expressed genes Having established that the hippocampal transcriptome was substantially associated with birth weight, we next asked whether we could identify individual genes for which the expression was statistically significantly different between the birth weight groups. Differentially expressed probesets were identified using one-way ANOVA and regression analyses, both at a significance level of P < 0.05. P-values were adjusted for multiple testing using the Benjamini Hochberg method. Using this very conservative criterion, only one probeset mapping to the TFDP2 gene was significant. Expression at the TFDP2 probe was 1.61 times up-regulated in the hippocampus of LBW animals with an adjusted P-value of 0.047. To more broadly identify candidate genes among the remaining probesets, we used a more relaxed fold-change cut-off of 1.5 with a nominal P-value smaller than 0.05. With this procedure 608 probesets corresponding to 378 unique genes were identified (data not shown), among which 562 were up-regulated and 46 were down regulated in the hippocampus of low-birth weight animals. We then used quantitative real-time PCR (qPCR) to technically validate the results for sixteen genes from the microarray analysis with a preference for histone methyltransferases and transcription factors while other targets were chosen at random. Fourteen of these genes were significantly up-regulated in the hippocampus of LBW neonates (ACTR2, NCOA3, MLL3, SETD2, WHSC1L1, NF1B, MEF2A, MLL, GABRB3, NCAM1, PDE1A, CNTN1, MPPED2, TXNIP) whereas two were unchanged between birth weight groups (PUM1, HS3ST2). Primers were designed that matched the respective probe sequence of the microarray and qPCR was performed with cDNA synthesized from the same RNA samples used for the microarray expression profiling. The results were analyzed with one-tailed t-tests comparing the LBW samples with NBW samples (Fig. 3). Reassuringly, eight genes (ACTR2, NCOA3, MLL3, SETD2, WHSC1L1, NF1B, MEF2A, MLL) were significantly upregulated in our qPCR analysis. The P-value for two additional genes (GABRB3 and NCAM1) approached significance level (i.e., P < .10 > .05). Four genes (PDE1A, CNTN1, MPPED2 and TXNIP) were not significantly up-regulated in our qPCR analysis although the mean expression values were all higher in the LBW samples as compared to the NBW samples. No significant change could be detected for the two

195

genes whose expression had also been found to be unaltered in the microarray expression experiment (PUM1 and HS3ST2). In sum, we found comparable expression changes between array and qPCR assays for twelve out of the sixteen targets (Table 2), and a trend in the same direction as the array results for another two targets.

Age had a negligible effect on gene expression Given the age range of the neonates in our study (2.5–7.3 days) we analyzed whether gene expression was correlated with age. First, we noted that age and birth weight were weakly positively correlated (Fig. 4A). Next, we regressed gene expression measured on the array against age. Contrary to the skewed P-value distribution for birth weight against gene expression indicating an association (Fig. 4B), we found a flat P-value distribution for that of gene expression against age (Fig. 4C), suggesting that the effects of age on gene expression were negligible.

Ten differentially expressed genes were significantly correlated with methylation changes We measured DNA methylation patterns in the same samples using the Infinium 450K array to test whether the observed differences in gene expression as a function of birth weight associated with methylation status. Two hundred and fifty-five of the unique, differentially expressed genes each had at least one CpG typed on the array. The number of CpGs for any differential expressed gene ranged from one to 138, with a median of 7 CpGs across all 255 genes. Of these 255 differential expressed genes, 110 (corresponding to 198 differential expressed gene:CpG pairs) had at least one CpG with a correlation P-value of less than 0.05 After taking into account the number of CpGs for each gene, and correcting for the number of tests, only 10 out of the 255 genes had at least one CpG within the gene with significant correlated expression and methylation. Out of these 10, 9 had a negative correlation, while only 1 had a positive correlation (Table 3).

Cell-type proportion is not significantly correlated with birth weight To assess if cell content of the hippocampal tissue affects birth weight, we used 5141 previously identified Infinium 450K probes differential for neuronal and glial cells to estimate the cell type proportions for each sample (Guintivano et al., 2013). Across the 10 samples, we observed largely consistent neuronal cell proportion ranging from 32% to 37%. In addition, correlation of cell content and birth weight was not significant (P = 0.15), suggesting that the observed changes in gene expression and methylation are unlikely to be due to cell content differences across the samples.

196

J. P. Buschdorf et al. / Neuroscience 318 (2016) 190–205

Fig. 3. qPCR validation of the microarray experiment. Fourteen out of sixteen genes (ACTR2, NCOA3, MLL3, SETD2, WHSC1L1, NF1B, MEF2A, MLL, GABRB3, NCAM1, PDE1A, CNTN1, MPPED2, TXNIP) were significantly up-regulated in the hippocampus of LBW neonates whereas two (PUM1, HS3ST3) were unchanged between birth weight groups as determined by microarray analysis. Primers matched the respective probe sequence of the microarray and cDNA was synthesized from the same RNA samples used for the microarray expression profiling. The results were analyzed with one-tailed t-tests comparing the LBW samples with NBW samples. P-values are shown. Bars represent mean expression values relative to b2-microglobulin expression with SEM. White bar: NBW; gray bar: LBW.

197

J. P. Buschdorf et al. / Neuroscience 318 (2016) 190–205 Table 2. Comparison of the fold change and P-values determined for the microarray and qPCR experiments, respectively. qPCR validation of microarray results for sixteen genes. Fourteen genes were significantly up-regulated in the hippocampus of LBW neonates whereas two were unchanged between birth weight groups as detected in our microarray analysis by one-way ANOVA at a significance level of P < 0.05. qPCR data were analyzed by one-tailed t-tests at a significance level of P < 0.05. P-values below significance level are in bold. Fold change in LBW samples was calculated relative to NBW samples Gene symbol

ACTR2 NCOA3 MLL3 SETD2 WHSC1L1 NF1B MEF2A MLL GABRB3 NCAM1 PDE1A CNTN1 MPPED2 TXNIP PUM1 HS3ST2

Array

qPCR

Fold change

P-Value (one-way ANOVA)

Fold change

P-Value (one tailed t-test)

1.61 1.58 2.03 1.76 1.68 2.28 1.84 2.45 1.51 1.52 1.75 2.03 1.58 2.06 1.00 0.76

0.013 0.020 0.005 0.021 0.006 0.016 0.007 0.012 0.016 0.036 0.028 0.013 0.032 0.016 0.959 0.558

1.86 1.94 2.62 1.86 1.75 1.74 1.52 1.54 1.59 1.53 1.39 1.23 1.19 1.23 1.23 0.78

0.004 0.029 0.030 0.033 0.038 0.041 0.041 0.047 0.051 0.073 0.113 0.170 0.280 0.309 0.239 0.271

Gene ontology profiling suggested low birth weight associates with hippocampal maturation

Pathway analysis suggested low birth weight associates with apoptosis and neuronal maturation

We performed a gene ontology enrichment analysis on genes mapping to the 608 differential probe sets to identify the biological processes associated with the differentially expressed genes. Table 4 presents the top 15 gene annotation clusters derived from DAVID analysis that showed an enrichment score greater than 1.3 (Huang da et al., 2009), which is equivalent to a non-log scale smaller than 0.05. The cluster with the highest enrichment score (4.8) included the GO terms ‘‘intracellular non-membrane-bounded organelle” and ‘‘cytoskeleton”. The first term includes structures like ribosomes, cytoskeleton and chromosomes. The second highest enrichment score (4.5) was determined for the annotation cluster that contained the terms ‘‘cell projection” and ‘‘neuron projection”. Corresponding to this cluster is the term ‘‘synapse” found in an annotation cluster with an enrichment score of 2.9. The term ‘‘programed cell death” was part of a cluster with an enrichment score of 1.9. These terms suggested altered differentiation and maturation of neurons and glial cells in the hippocampus of LBW animals. Interestingly, the GO terms ‘‘protein methyltransferase activity”, ‘‘histone-lysine Nmethyltransferase activity” and ‘‘chromatin modification” were identified as part of an annotation cluster with an enrichment score of 1.6. Genes identified that fell into these terms were the histone methyltransferases MLL, MLL3, SETD2, PRDM2 and WHSC1L1. Up-regulation of some of these genes in LBW samples was validated by qPCR (Fig. 3). The identification of these GO terms suggested that mechanisms such as histone methylation could imprint the environmental conditions associated with low birth weight onto the epigenome.

We next performed pathway analysis to map the regulated genes onto pathways that might contribute to the biological processes identified by gene ontology enrichment analysis. Table 5 presents the top 20 pathways that passed FDR < 0.05. The most significantly altered pathway was ‘‘Development_ Thrombopoietin-regulated cell process”. Thrombopoietin is expressed in the hippocampus and has a proapoptotic effect on neurons (Diederich et al., 2012). Two additional pathways with an association to hematopoietic growth factors were identified: ‘‘Development_EPOinduced MAPK pathway” and ‘‘Development_G-CSF signaling”. Both EPO and G-CSF support neurogenesis and dendritogenesis in the hippocampus (Diederich et al., 2012). Three Wnt-related pathways were identified: ‘‘Development_WNT signaling pathway Part 2”, ‘‘Development_WNT5A signaling”, and ‘‘Development_WNT signaling pathway. Part 1. Degradation of beta-catenin in the absence WNT signaling”. Wnt5a signaling stimulates synaptic differentiation and function of glutamatergic synapses (Varela-Nallar et al., 2010) and also increases the recycling of GABAA receptors on hippocampal neurons via activation of CaMKII (Cuitino et al., 2010). Among the identified pathways were also three insulin-related pathways and two pathways related to adenosine signaling (Table 5). The results are in agreement with the notion of altered hippocampal development in the LBW group.

DISCUSSION Epidemiological studies frequently link low birth weight to an increased risk for mental disorders (Alati et al., 2007;

198

J. P. Buschdorf et al. / Neuroscience 318 (2016) 190–205

Fig. 4. Effect of neonatal age on gene expression. Correlation analysis between age and birth weight of neonates (a). P-value distribution for association tests relating birth weight to gene expression (b). P-value distribution for association tests relating age to gene expression (c).

199

J. P. Buschdorf et al. / Neuroscience 318 (2016) 190–205 Table 3. Correlation of differentially expressed genes and CpG methylation DE genes

Pearson correlation

P-Value

CpGs that pass Bonferroni

Gene region

CpG island

NLGN4Y SLC4A7 BRWD3 NXPH2 BIRC6 ZNF770 UHMK1 MED13 RALGPS2 RAB30

0.89
0.83
0.81
0.81
0.85
0.86
0.85
0.88
0.86
0.85

5.13E-04 2.82E-03 4.92E-03 4.44E-03 1.88E-03 1.45E-03 1.96E-03 8.65E-04 1.52E-03 2.08E-03

cg00938039 cg02344497 cg15511898 cg15701208 cg26189067 cg05961215 cg06325209 cg04564935 cg07044282 cg06516650

30 UTR;30 UTR Body TSS1500 Body TSS1500 TSS200 TSS200;TSS200 TSS1500 1stExon;50 UTR;Body TSS200

N_Shelf S_Shore N_Shore Island S_Shore

Table 4. Gene ontology analysis of differentially expressed genes from the hippocampus of LBW animals. Each cluster only shows three representative GO terms and its unique genes P-Value CLUSTER 1 Enrichment Score: 4.76 GO:0043232intracellular 5.05E-06 non-membranebounded organelle GO:0043228non5.05E-06 membrane-bounded organelle GO:0005856cytoskeleton 2.05E-04

CLUSTER 2 enrichment score: 4.47 GO:0042995cell 4.79E-06 projection GO:0043005neuron 1.95E-05 projection GO:0044463cell 4.14E-04 projection part

Fold enrichment

FDR

Genes

1.62

0.0069

1.62

0.0069

1.76

Not passed

EIF6, KIF27, UTY, CASK, RPS6KB1, CCT2, IQGAP1, CBX5, ARHGAP21, ACTR2, TOP1, G2E3, KLHL5, RAD21, CSNK2A1, H2AFV, ANK3, KIFAP3, ZNF146, QKI, LRRFIP1, TPR, LRRC7, MYST3, DLG1, TBL1XR1, ANKS1B, ZBTB20, MYO6, KIF5B, RBL2, PCM1, TOX4, BICD2, BICD1, EML4, NAV1, LCA5, PDE4DIP, PARD3, ERBB4, ABI2, NUFIP2, CYLD, GPHN, EZR, HNRNPK, MACF1, EXOC4, SLC4A7, RPL10A, APPBP2, KIF21A, BUB3, APC, DHX9, RNF19A, MCF2, CREB1, TP53BP1, MAP1B, PTPN4, LMNA, PPP1R10, BIRC2, SF3A1, SMC3, FXR1, CSPP1, ATRX, ZFP106, GRIA1, IFT57, WHSC1L1, SUPT16H, JAK1, TMOD3, NFIB

2.48

0.0065

3.15

0.0265

3.23

Not passed

NRP2, SYT1, SH3RF1, PARD3, GABRB3, ABI2, IQGAP1, NRCAM, ACTR2, EZR, ANK3, TTC30B, RASGRP2, EXOC4, SLC4A7, KLHL24, NSF, APC, ANKS1B, MYO6, KIF5B, PLEK, SCN2A, MAP1B, PCM1, NCAM1, CHRM3, LCA5, GSK3B, IFT57, BACE1, RSC1A1, ANKS1B

CLUSTER 3 enrichment score: 3.99 GO:0031981nuclear 1.79E-05 lumen GO:0031974membrane5.25E-05 enclosed lumen GO:0043233organelle 5.91E-05 lumen

1.86

0.0244

1.69

Not passed

1.69

Not passed

CLUSTER 4 enrichment score: 2.92 GO:0030054cell junction 2.42E-04 GO:0045202synapse 3.22E-04 GO:0044456synapse 2.10E-03 part

2.39 2.73 2.85

Not passed Not passed Not passed

SYT1, ANKS1B, PARD3, GABRA2, GABRB3, SCN2A, MPP5, ABI2, RPS6KB1, SSPN, RIMS1, ARHGAP21, GPHN, TMEM47, MACF1, CHRM3, GRIA1, RASGRP2, CDC42BPA, UNC13C, LRRC7, APC, DLG1, ERBB4, MAP1B, CASK, ANK3, EXOC4, ANKS1B, SYT1

CLUSTER 5 enrichment score: 2.83 GO:0005856cytoskeleton 2.05E-04 GO:0015630microtubule 2.95E-03 cytoskeleton

1.76 2.06

Not passed Not passed

EIF6, PARD3, ERBB4, KIF27, CASK, ABI2, IQGAP1, ARHGAP21, CYLD, ACTR2, GPHN, KLHL5, EZR, MACF1, ANK3, KIFAP3, EXOC4, APPBP2,

EIF6, UTY, CCT2, ZEB1, CNOT7, IQGAP1, CBX5, TOP1, G2E3, CSNK2A1, ZNF146, QKI, TPR, LRRC7, MYST3, ANKS1B, TBL1XR1, ZBTB20, MYO6, RBL2, USP1, MED13, TOX4, HNRNPR, PIAS1, MATR3, SRRT, HNRNPK, ZNF326, TFDP2, TCF4, ERCC3, BUB3, DHX9, MLL, CREB1, TP53BP1, MAML2, LMNA, YTHDC1, SF3A1, SMC3, FXR1, DUSP4, ZFP106, SON, WHSC1L1, SUPT16H, TCEB3, NFIB, PPM1K, EDEM3, THBS1, TXNIP, AK3, HSP90B1, HSPD1, RCN2

(continued on next page)

200

J. P. Buschdorf et al. / Neuroscience 318 (2016) 190–205

Table 4 (continued) P-Value

Fold enrichment

FDR

Genes LRRFIP1, KIF21A, DLG1, APC, ANKS1B, DHX9, TBL1XR1, MYO6, KIF5B, RNF19A, MCF2, MAP1B, PTPN4, LMNA, PCM1, BICD2, SMC3, BICD1, EML4, CSPP1, NAV1, LCA5, GRIA1, IFT57, JAK1, TMOD3, PDE4DIP, TBL1XR1

GO:0044430cytoskeletal part

5.23E-03 P-Value

CLUSTER 6 enrichment score: 2.30 GO:0008104protein localization 3.59E-04 GO:0045184establishment of 7.36E-04 protein localization GO:0015031protein transport 1.34E-03

CLUSTER 7 enrichment score: 1.94 GO:0012501programed cell 5.52E-03 death GO:0006915apoptosis 9.38E-03 GO:0042981regulation of 1.04E-02 apoptosis CLUSTER 8 enrichment score: 1.89 GO:0030695GTPase regulator 3.75E-03 activity GO:0060589nucleoside4.69E-03 triphosphatase regulator activity GO:0008047enzyme activator 1.97E-02 activity CLUSTER 9 enrichment score: 1.81 GO:0000166nucleotide binding 5.85E-03 GO:0017076purine nucleotide 7.16E-03 binding GO:0032555purine 7.46E-03 ribonucleotide binding

CLUSTER 10 enrichment score: 1.71 9.07E-03 GO:0051603proteolysis involved in cellular protein catabolic process GO:0006511ubiquitin-dependent 9.43E-03 protein catabolic process GO:0044257cellular protein 9.51E-03 catabolic process CLUSTER 11 enrichment score: 1.71 GO:0005626insoluble fraction 1.21E-02 GO:0005624membrane fraction 1.46E-02 GO:0000267cell fraction 4.32E-02

1.70

Not passed Fold enrichment

FDR

Unique genes

1.92 1.95

Not passed Not passed

1.90

Not passed

NAPB, CANX, RIMS1, IL10, STXBP5L, EZR, MACF1, ANK3, KIFAP3, ZFYVE16, EXOC4, RAB6B, PPP3CA, APPBP2, VPS13B, EXOC5, TPR, ERCC3, AGAP1, NSF, AP3B1, MYO6, PLEK, PPP1R10, ARFIP1, HSP90B1, RAB30, RABEP1, LCA5, GSK3B, TRPS1, RAB22A, TOMM20, AKAP6

1.88

Not passed

1.82 1.68

Not passed Not passed

2.18

Not passed

2.14

Not passed

2.05

Not passed

1.37 1.40

Not passed Not passed

1.41

Not passed

1.83

Not passed

2.48

Not passed

1.82

Not passed

1.67 1.67 1.44

Not passed Not passed Not passed

IER3, MEF2A, MLL, MCF2, SCN2A, BIRC6, ARHGEF17, VAV2, STK4, BIRC2, FXR1, BAG4, TOP1, G2E3, TNFSF10, RAD21, KRAS, RABEP1, IFT57, SOS2, MDM4, HSPD1, THBS1, SH3RF1, IER3, BCLAF1, IL10, CUL5, ERCC3, APC, TXNIP, CREB1, MALT1, HSP90B1, SON, GSK3B RALGPS2, MCF2, RABGAP1L, ARHGEF17, VAV2, RIMS1, IQGAP1, STARD13, RASAL2, ARHGAP21, ARHGAP22, GAPVD1, ARHGAP5, RABEP1, SOS2, RASGRP2, CDC42BPA, AGAP1, PPP1R12B, MMP16, MALT1, TRIM23

HSP90AB1, ACOX1, KIF27, CASK, RPS6KB1, CCT2, PRKG1, TOP1, ACTR2, HSPH1, DDX17, ARHGAP5, CSNK2A1, AGPS, PAK3, DHX36, RAB6B, TPR, AGAP1, AKT3, NSF, HSP90AA1, MYO6, KIF5B, UBE2J1, EEF2, CDKL3, CSTF2T, STK4, HNRNPR, SRPK1, MFN1, NAV1, BMP2K, MATR3, ERBB4, NEK1, UBA6, UHMK1, GPHN, KRAS, LARS, PABPC1, HNRNPC, KIF21A, ERCC3, DHX9, MLL, AK3, EPRS YTHDC2, ELAVL2, MAPK10, TRIM23, SMC3, ATRX, HSP90B1 RAB30, GSK3B, RAB22A, CDC42BPA, JAK1, HSPD1 USP7, TBL1XR1, C10ORF46,RNF19A,USP1, CBL, UBE2J1, BIRC6, UBA6, MALT1, EDEM3, PSMF1, CYLD, G2E3, HSP90B1, CUL5, PSMD12, BACE1, PIAS1, USP34, BUB3, HECTD1, ERCC3, CD46, LMLN, NSF, MMP16, ADAM18

NRP2, CASK, RPS6KB1, EEA1, SLC16A1, CUL5, KRAS, ANK3, RASGRP2, PPP3CA, PLCB1, PLEK, MPP2, MCF2, SCN2A, LMNA, BIRC6, MCTP1, HSP90B1, ZFP106, SLC16A7, GRIA1, BACE1, CNTN1, ADAM18, RSC1A1, MAP1B, EPRS, TNFSF10

J. P. Buschdorf et al. / Neuroscience 318 (2016) 190–205

201

Table 4 (continued) P-Value

CLUSTER 12 enrichment score: 1.65 GO:0016192vesicle-mediated 1.17E-02 transport GO:0016044membrane 2.10E-02 organization GO:0010324membrane 3.27E-02 invagination CLUSTER 13 enrichment score: 1.61 GO:0008276protein 3.36E-03 methyltransferase activity GO:0018024histone-lysine 3.84E-03 N-methyltransferase activity GO:0016568chromatin 1.83E-01 modification CLUSTER 14 enrichment score: 1.61 GO:0019201nucleotide kinase 1.09E-02 activity GO:0004385guanylate kinase 2.01E-02 activity GO:0016776phosphotransferase 2.89E-02 activity, phosphate group as acceptor CLUSTER 15 enrichment score: 1.60 GO:0045893positive regulation of 3.33E-03 transcription, DNA-dependent GO:0051254positive regulation of 3.71E-03 RNA metabolic process GO:0045941positive regulation of 4.53E-03 transcription

Fold enrichment

FDR

Unique genes

1.82

Not passed

1.97

Not passed

2.27

Not passed

MYO6, PLEK, EEA1, NAPB, M6PR, RIMS1, STXBP5L, GAPVD1, KRAS, RABEP1, GRIA1, ZFYVE16, HMGXB4, RAB22A, EXOC4, VAMP4, RAB6B, EXOC5, UNC13C, THBS1, AP3B1, HSP90AA1, LMNA, NRCAM, MFN1, GAPVD1

5.88

Not passed

7.66

Not passed

1.64

Not passed

8.52

Not passed

13.36

Not passed

5.94

Not passed

2.093

Not passed

2.0756

Not passed

1.9472

Not passed

Abel et al., 2010; Betts et al., 2013), including those such as depression and schizophrenia that involve altered hippocampal function. These findings reflect the potential prenatal origins of predispositions for multiple mental disorders. We tested the hypothesis that low birth weight would associate with hippocampal gene expression patterns in non-human primate neonates. We previously showed that the mineralocorticoid receptor/glucocorticoid receptor ratio is decreased (on mRNA as well as protein level) in the hippocampus of the same set of low birth weight animals used in the current study indicating indeed an effect on gene expression associated with birth weight (Ong et al., 2013). Here, we extended our analysis to examine global differences in hippocampal gene

MLL, WHSC1L1, PCMTD1, PRDM2, SETD2, MLL3, ATRX, TBL1XR1, RSF1, RBL2, UTY, MYST3

MPP2, AK3, CASK, DLG1

TBL1XR1, MLL, MYO6, KLF12 CREB1, TP53BP1, MAML2, CASK, RORA, SOX6, ZEB1, MED13, CNOT7, IL10, ZBTB38, NCOA3, PIAS1, TCF4, ERCC3, NFIB, RSF1, MYST3, HSP90AB1, TBL1XR1, HSP90AA1, THBS1, ANKS1B, RBL2, YTHDC1, CBX5, SON, CSNK2A1, TFDP2, TCEB3, ZNF292, MEF2A, GPBP1, ZNF518A, ARHGAP22 LRRFIP1, MLL3, NFKBIZ, ZBTB20, ZNF92, GTF2IRD2, ZNF37A, PRDM2, MYNN, ZNF516, BCLAF1, ZNF800, NR3C2, ZNF326, MLLT3, TXNIP, ZNF567, ZNF770, ZNF667, AFF4, PPP1R10, ZFHX4, PHF3, IFT57, TRPS1, WHSC1L1, SUPT16H, RFX3, SETD2, RSC1A1, ZNF146, PDE8A, MDM4, KRAS, HMGXB4, RFX7, ELAVL2, MALT1, ATRX, TOP1, H2AFV, TOX4, ABI2, GIN1, HNRNPK, SOS2, DHX9, SETBP1, CBL, HSPD1,RSF1, PSMF1, APC, PSMD12, MEF2A, PABPC1, SRPK1, NMRAL1, RNF19A, STK4, GSK3B

transcription in naturally-occurring low compared with normal birth weight samples. We found that gene expression patterns for 24,154 probesets clustered the samples almost entirely according to birth weight suggesting that there was a gene expression profile specific for low birth weight. Moreover, gene ontology and pathway analysis suggested that the differentially expressed genes in the low birth weight group, most of which were up regulated, associated with neuronal maturation and apoptosis. These findings suggested an up-regulation of genes implicated in neuronal maturation in hippocampal samples from the low birth weight subjects. While perhaps counter-intuitive, there is emerging evidence for an effect of environmental adversity on the

202

J. P. Buschdorf et al. / Neuroscience 318 (2016) 190–205

Table 5. Metacore Pathway enrichment analysis of differentially expressed genes from the hippocampus of LBW animals. Unique genes are presented. All terms passed FDR < 0.05 Rank

Maps

P-Value

Unique genes

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

Development_Thrombopoietin-regulated cell processes Development_A2B receptor: action via G-protein alpha s Development_WNT signaling pathway. Part 2 Translation_Insulin regulation of translation Development_Activation of Erk by ACM1, ACM3 and ACM5 Development_EPO-induced MAPK pathway Signal transduction_Calcium signaling Development_PIP3 signaling in cardiac myocytes Development_WNT5A signaling Development_A3 receptor signaling Development_G-CSF signaling Development_EDNRB signaling Development_EGFR signaling via small GTPases Development_IGF-1 receptor signaling Signal transduction_Erk Interactions: Inhibition of Erk Development_NOTCH-induced EMT Development_WNT signaling pathway Part 1. Degradation of beta-catenin in the absence WNT signaling Development_Role of HDAC and calcium/ calmodulin-dependent kinase (CaMK) in control of skeletal myogenesis Immune response_BCR pathway G-protein signaling_Regulation of RAC1 activity

8.46E-06 1.74E-05 2.57E-05 6.32E-05 8.27E-05 9.41E-05 9.41E-05 1.21E-04 1.21E-04 1.53E-04 1.53E-04 1.71E-04 1.99E-04 2.14E-04 2.30E-04 2.40E-04 2.40E-04

AKT3, CREB1, GSK3B, SOS2, CBL, RPS6KB1 AKT3, PAK3, CREB1, GSK3B, PLCB1, SOS2, TCF4 APC, CSNK2A1, GSK3B, TCF4, NRCAM AKT3, GSK3B, SOS2, EEF2, EIF4G1, RPS6KB1 CHRM3, ADAM17, RASGRP2, CREB1, PLCB1, SOS2 MAPK10, KRAS, PAK3, STK4, SOS2, CBL AKAP6, CREB1, PPP3CA, MEF2A, PLCB1, EZR AKT3, CREB1, GSK3B, PARD3, SOS2, RPS6KB1 PPP3CA, GSK3B, MAPK10, PLCB1, PRKG1, TCF4 AKT3, CREB1, GSK3B, PLCB1, SOS2, TCF4 AKT3, JAK1, MAPK10, SOS2, CBL, BIRC2 AKT3, RASGRP2, CREB1, PLCB1, PRKG1, SOS2 MAPK10, KRAS, SOS2, VAV2, CBL AKT3, CREB1, GSK3B, M6PR, SOS2, RPS6KB1 AKT3, PPP3CA, MAPK10, DUSP4, DUSP9 ADAM17, NOTCH1, PSEN2 APC, GSK3B, TCF4

2.64E-04

AKT3, PPP3CA, IGF2, RPS6KB1

2.64E-04 3.04E-04

AKT3, PPP3CA, GSK3B, SOS2, VAV2, RPS6KB1 ARHGAP22, MCF2, KIFAP3, VAV2, EZR

18

19 20

pace of neural development in mammals. Prolonged and repeated periods of maternal deprivation in the infant rat, are sufficient to reduce circulating levels of growth hormone (Kuhn and Schanberg, 1998), accelerate amygdala function and amygdala-dependent fear learning (Moriceau et al., 2006; Callaghan and Richardson, 2011). Maternal separation is also associated with increased development of neurons in the medial prefrontal cortex (Muhammad et al., 2012). In humans, Gee et al. (Gee et al., 2013) revealed the early emergence of amygdala - prefrontal connectivity among institutionalized children who lack parental care. Rao et al. (2010) showed a negative correlation between the quality of postnatal parental care and hippocampal volume. Perhaps most compelling is the study of DeBellis and colleagues (Tupler and De Bellis, 2006) who reported that maltreatment was associated with a larger hippocampal volume in young children. Interestingly, in the examples cited above, the accelerated maturation of amygdala function and connectivity with the prefrontal cortex were also associated with increased glucocorticoid signaling (Moriceau et al., 2006; Callaghan and Richardson, 2011; Gee et al., 2013). Stress as well as protein or nutrient restriction, all of which constrain fetal growth, increase circulating glucocorticoids in the mother and her fetus (Meaney et al., 2007; Cottrell et al., 2012). Glucocorticoid administration to pregnant female rodents reduces birth weight (Nyirenda et al., 2001) and maternal adrenalectomy blocks the effects of protein deprivation on fetal growth (Cottrell et al., 2012). Increased glucocorticoid levels are implicated in fetal growth restraint in humans (Goland et al., 1993; Seckl, 2004; Meaney et al., 2007).

Glucocorticoid treatment during late pregnancy also correlates with poorer cognitive performance in humans (Damsted et al., 2011). Accordingly, primates exposed to exogenous glucocorticoids before birth show altered hippocampal volume (Uno et al., 1994). While the potential role of fetal glucocorticoids is currently a matter of speculation, our previous and current findings are consistent with those of the effects of postnatal conditions and suggest that adverse environments can operate during either fetal or postnatal development to accelerate neuronal maturation. Data derived from animal as well as human studies suggest that environmental conditions that influence the quality of fetal growth can have lasting programing effects that are mediated by epigenetic mechanisms (Gluckman et al., 2009). In our study only 10 out of 255 unique, differentially expressed genes (about 4%) had at least one CpG within the gene with significantly correlated expression and methylation. The weak correlation is not necessarily inconsistent with a programing effect via DNA methylation. Studies in rodents have demonstrated that established models of developmental programing during the perinatal or neonatal period induce stable modifications in DNA methylation of specific genes only in adult or young adult animals long after corresponding changes in gene expression are detected (Park et al., 2008; Murgatroyd et al., 2009). For example, in the study of Park et al. intrauterine growth restriction strongly reduced gene expression of the transcription factor Pdx1 in fetal and early postnatal tissues which was not accompanied by an increase in DNA methylation at the proximal Pdx1 promoter. Only in young adult tissues

J. P. Buschdorf et al. / Neuroscience 318 (2016) 190–205

DNA methylation was detected at the proximal promoter. Interestingly, early repression of transcription was accompanied by changes in transcription factor binding, histone acetylation as well as histone methylation pointing to these epigenetic modifications mediating the immediate effects of intrauterine growth restriction (Park et al., 2008). Low birth weight increases the risk for mood disorders and schizophrenia (Alati et al., 2007; Abel et al., 2010; Betts et al., 2013) probably in combination with other environmental factors like childhood maltreatment (Costello et al., 2007; Nomura and Chemtob, 2007). Recently, a genome-wide association study of SNPs affecting the development of bipolar disorder, major depressive disorder and schizophrenia surveying data from over 60,000 participants identified ‘‘histone H3-K4 methylation” and ‘‘histone methylation” as the top two shared pathways across these adult psychiatric disorders, pointing to a common mechanism for these diseases (O’Dushlaine, 2015). Interestingly, by interrogating gene expression and birth weight, which is a risk factor for such neurological diseases, we also found enrichment for histone modification genes. Differential expression of some of these genes was validated by qPCR (Table 2 and Fig. 3). Such epigenetic mechanisms could indirectly associate with processes identified alongside in our gene ontology analysis, for example ‘‘neuron projection”, ‘‘cell junction” and ‘‘synapse”. Again identification of these processes is consistent with the study by O’Dushlaine (2015) who identified ‘‘cell–cell junction” and ‘‘postsynaptic density” among the top shared pathways affected in mental disorders. Together with the epidemiological data indicating low birth weight as a risk factor for psychopathology, our results raise the possibility that histone modifications play a role in regulating the global transcriptomic changes observed in low birth weight infants early on in development, and these changes may lead to greater susceptibility to neurological illnesses later on in life. The use of unfractionated tissue for the analysis of gene expression and DNA methylation will mean that cellular heterogeneity associated with the hippocampal development shortly after birth and/or with birth weight may drive some of the detected molecular changes. To overcome cellular heterogeneity possibly associated with birth weight, we used the bioinformatics model introduced by Guintivano et al. (2013) to quantify the proportion of neurons to glia based on DNA methylation. Due to the removal of 57.6% of all probes of the Infinium 450K array we used about half of the epigenotype-specific markers as compared to the study by Guintivano et al. (2013). We did not detect a significant difference in the neuron to glia ratio between the NBW and LBW groups suggesting no difference in cell composition between these two groups. This result is in agreement with a Western blot for hippocampal glial and neuronal marker protein expression of the same set of animals presented in our previous study (Ong et al., 2013). These results also suggest that the changes in gene expression cannot be exclusively ascribed to neurons or non-neuronal cells. A clear limitation is our small sample size, which limits rigorous statistical correction for multiple testing although

203

notably gene ontology and pathway enrichment analyses passed multiple testing corrections. In addition, we are unable to specify the cause of low birth weight in our animals. This is similar to many epidemiological studies where birth weight is also used as a proxy for fetal growth conditions. However, we noted that the weight of the dams prior to breeding correlated significantly with the birth weight of the neonates. This correlation is consistent with studies in humans which identified pre-pregnancy weight and weight gain during pregnancy as predictors for fetal growth (Ay et al., 2009). The correlation suggests a maternal effect on birth weight in our study. Our data suggest that the hippocampi of low birth weight neonates of Cynomolgus macaques display a distinct gene expression profile as compared to the control group. Differentially expressed genes indicate alterations in the level of neuronal maturation, transcription and apoptosis associated with low birth weight. Our data further suggest that the association between low birth weight and hippocampal gene expression provides a potential molecular basis for the developmental origin of an enhanced risk for mental disorders.

COMPETING INTERESTS The authors have declared that no conflict of interest exists. Acknowledgments—M.S.K. and M.J.M. are Senior Fellows of the Canadian Institute for Advanced Research and members of the Brain Canada – funded Canadian Neuroepigenetics network. M.S.K. is the Canada Research Chair in Social Epigenetics. This study was supported by the Agency of Science, Technology and Research (A*STAR), Singapore.

REFERENCES Abel KM, Wicks S, Susser ES, Dalman C, Pedersen MG, Mortensen PB, Webb RT (2010) Birth weight, schizophrenia, and adult mental disorder: is risk confined to the smallest babies? Arch Gen Psychiatry 67:923–930. Affymetrix (2006) Affymetrix Transcript Assignment for NetAffxTM Annotations Revision 2.3. Alati R, Lawlor DA, Mamun AA, Williams GM, Najman JM, O’Callaghan M, Bor W (2007) Is there a fetal origin of depression? Evidence from the Mater University Study of Pregnancy and its outcomes. Am J Epidemiol 165:575–582. Ay L, Kruithof CJ, Bakker R, Steegers EA, Witteman JC, Moll HA, Hofman A, Mackenbach JP, Hokken-Koelega AC, Jaddoe VW (2009) Maternal anthropometrics are associated with fetal size in different periods of pregnancy and at birth. The Generation R Study. BJOG 116:953–963. Barker DJ, Gluckman PD, Godfrey KM, Harding JE, Owens JA, Robinson JS (1993a) Fetal nutrition and cardiovascular disease in adult life. Lancet 341:938–941. Barker DJ, Hales CN, Fall CH, Osmond C, Phipps K, Clark PM (1993b) Type 2 (non-insulin-dependent) diabetes mellitus, hypertension and hyperlipidaemia (syndrome X): relation to reduced fetal growth. Diabetologia 36:62–67. Benjamini Y, Hochberg Y (1995) Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc Ser B 57:289–300.

204

J. P. Buschdorf et al. / Neuroscience 318 (2016) 190–205

Betts KS, Williams GM, Najman JM, Scott J, Alati R (2013) The association between lower birth weight and comorbid generalised anxiety and major depressive disorder. J Affect Disord 146:231–237. Bhutta AT, Cleves MA, Casey PH, Cradock MM, Anand KJ (2002) Cognitive and behavioral outcomes of school-aged children who were born preterm: a meta-analysis. JAMA 288:728–737. Buss C, Lord C, Wadiwalla M, Hellhammer DH, Lupien SJ, Meaney MJ, Pruessner JC (2007) Maternal care modulates the relationship between prenatal risk and hippocampal volume in women but not in men. J Neurosci 27:2592–2595. Callaghan BL, Richardson R (2011) Maternal separation results in early emergence of adult-like fear and extinction learning in infant rats. Behav Neurosci 125:20–28. Coe CL, Kramer M, Czeh B, Gould E, Reeves AJ, Kirschbaum C, Fuchs E (2003) Prenatal stress diminishes neurogenesis in the dentate gyrus of juvenile rhesus monkeys. Biol Psychiatry 54:1025–1034. Costello EJ, Worthman C, Erkanli A, Angold A (2007) Prediction from low birth weight to female adolescent depression: a test of competing hypotheses. Arch Gen Psychiatry 64:338–344. Cottrell EC, Holmes MC, Livingstone DE, Kenyon CJ, Seckl JR (2012) Reconciling the nutritional and glucocorticoid hypotheses of fetal programming. FASEB J 26:1866–1874. Cuitino L, Godoy JA, Farias GG, Couve A, Bonansco C, Fuenzalida M, Inestrosa NC (2010) Wnt-5a modulates recycling of functional GABAA receptors on hippocampal neurons. J Neurosci 30:8411–8420. Damsted SK, Born AP, Paulson OB, Uldall P (2011) Exogenous glucocorticoids and adverse cerebral effects in children. Eur J Paediatr Neurol 15:465–477. Dempster EL, Pidsley R, Schalkwyk LC, Owens S, Georgiades A, Kane F, Kalidindi S, Picchioni M, Kravariti E, Toulopoulou T, Murray RM, Mill J (2011) Disease-associated epigenetic changes in monozygotic twins discordant for schizophrenia and bipolar disorder. Hum Mol Genet 20:4786–4796. Dennis Jr G, Sherman BT, Hosack DA, Yang J, Gao W, Lane HC, Lempicki RA (2003) DAVID: Database for Annotation, Visualization, and Integrated Discovery. Genome Biol 4:P3. Diederich K, Schabitz WR, Minnerup J (2012) Seeing old friends from a different angle: novel properties of hematopoietic growth factors in the healthy and diseased brain. Hippocampus 22:1051–1057. Femenia T, Gomez-Galan M, Lindskog M, Magara S (2012) Dysfunctional hippocampal activity affects emotion and cognition in mood disorders. Brain Res 1476:58–70. Gee DG, Gabard-Durnam LJ, Flannery J, Goff B, Humphreys KL, Telzer EH, Hare TA, Bookheimer SY, Tottenham N (2013) Early developmental emergence of human amygdala-prefrontal connectivity after maternal deprivation. Proc Natl Acad Sci U S A 110:15638–15643. Gieling ET, Park SY, Nordquist RE, Van der Staay FJ (2012) Cognitive performance of low- and normal-birth-weight piglets in a spatial hole-board discrimination task. Pediatr Res 71:71–76. Gluckman PD, Hanson MA, Buklijas T, Low FM, Beedle AS (2009) Epigenetic mechanisms that underpin metabolic and cardiovascular diseases. Nat Rev Endocrinol 5:401–408. Goland RS, Jozak S, Warren WB, Conwell IM, Stark RI, Tropper PJ (1993) Elevated levels of umbilical cord plasma corticotropinreleasing hormone in growth-retarded fetuses. J Clin Endocrinol Metab 77:1174–1179. Grayson DR, Guidotti A (2013) The dynamics of DNA methylation in schizophrenia and related psychiatric disorders. Neuropsychopharmacology 38:138–166. Guintivano J, Aryee MJ, Kaminsky ZA (2013) A cell epigenotype specific model for the correction of brain cellular heterogeneity bias and its application to age, brain region and major depression. Epigenetics 8:290–302. Huang da W, Sherman BT, Lempicki RA (2009) Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nat Protoc 4:44–57.

Kirkegaard I, Obel C, Hedegaard M, Henriksen TB (2006) Gestational age and birth weight in relation to school performance of 10-yearold children: a follow-up study of children born after 32 completed weeks. Pediatrics 118:1600–1606. Kuhn CM, Schanberg SM (1998) Responses to maternal separation: mechanisms and mediators. Int J Dev Neurosci 16:261–270. Lister JP, Blatt GJ, DeBassio WA, Kemper TL, Tonkiss J, Galler JR, Rosene DL (2005) Effect of prenatal protein malnutrition on numbers of neurons in the principal cell layers of the adult rat hippocampal formation. Hippocampus 15:393–403. Mallard C, Loeliger M, Copolov D, Rees S (2000) Reduced number of neurons in the hippocampus and the cerebellum in the postnatal guinea-pig following intrauterine growth-restriction. Neuroscience 100:327–333. McGowan PO, Sasaki A, D’Alessio AC, Dymov S, Labonte B, Szyf M, Turecki G, Meaney MJ (2009) Epigenetic regulation of the glucocorticoid receptor in human brain associates with childhood abuse. Nat Neurosci 12:342–348. Meaney MJ, Szyf M, Seckl JR (2007) Epigenetic mechanisms of perinatal programming of hypothalamic-pituitary-adrenal function and health. Trends Mol Med 13:269–277. Moriceau S, Wilson DA, Levine S, Sullivan RM (2006) Dual circuitry for odor-shock conditioning during infancy: corticosterone switches between fear and attraction via amygdala. J Neurosci 26:6737–6748. Muhammad A, Carroll C, Kolb B (2012) Stress during development alters dendritic morphology in the nucleus accumbens and prefrontal cortex. Neuroscience 216:103–109. Murgatroyd C, Patchev AV, Wu Y, Micale V, Bockmuhl Y, Fischer D, Holsboer F, Wotjak CT, Almeida OF, Spengler D (2009) Dynamic DNA methylation programs persistent adverse effects of early-life stress. Nat Neurosci 12:1559–1566. Nomura Y, Chemtob CM (2007) Conjoined effects of low birth weight and childhood abuse on adaptation and well-being in adolescence and adulthood. Arch Pediatr Adolesc Med 161:186–192. Noorlander CW, Tijsseling D, Hessel EV, de Vries WB, Derks JB, Visser GH, De Graan PN (2014) Antenatal glucocorticoid treatment affects hippocampal development in mice. PLoS One 9:e85671. Nyirenda MJ, Welberg LA, Seckl JR (2001) Programming hyperglycaemia in the rat through prenatal exposure to glucocorticoids-fetal effect or maternal influence? J Endocrinol 170:653–660. O’Dushlaine C (2015) Psychiatric genome-wide association study analyses implicate neuronal, immune and histone pathways. Nat Neurosci 18:199–209. Ong SX, Chng K, Meaney MJ, Buschdorf JP (2013) Decreased hippocampal mineralocorticoid:glucocorticoid receptor ratio is associated with low birth weight in female cynomolgus macaque neonates. J Mol Endocrinol 51:59–67. Ong ML, Tan PY, Macisaac JL, Mah SM, Buschdorf JP, Cheong CY, Stunkel W, Chan L, Gluckman PD, Chng K, Kobor MS, Meaney MJ, Holbrook JD (2014) Infinium monkeys: infinium 450K Array for the Cynomolgus macaque (Macaca fascicularis). G3. Park JH, Stoffers DA, Nicholls RD, Simmons RA (2008) Development of type 2 diabetes following intrauterine growth retardation in rats is associated with progressive epigenetic silencing of Pdx1. J Clin Invest 118:2316–2324. Peirson SN, Butler JN, Foster RG (2003) Experimental validation of novel and conventional approaches to quantitative real-time PCR data analysis. Nucleic Acids Res 31:e73. Provencal N, Suderman MJ, Guillemin C, Massart R, Ruggiero A, Wang D, Bennett AJ, Pierre PJ, Friedman DP, Cote SM, Hallett M, Tremblay RE, Suomi SJ, Szyf M (2012) The signature of maternal rearing in the methylome in rhesus macaque prefrontal cortex and T cells. J Neurosci 32:15626–15642. Rao H, Betancourt L, Giannetta JM, Brodsky NL, Korczykowski M, Avants BB, Gee JC, Wang J, Hurt H, Detre JA, Farah MJ (2010) Early parental care is important for hippocampal maturation: Evidence from brain morphology in humans. Neuroimage 49:1144–1150.

J. P. Buschdorf et al. / Neuroscience 318 (2016) 190–205 Raznahan A, Greenstein D, Lee NR, Clasen LS, Giedd JN (2012) Prenatal growth in humans and postnatal brain maturation into late adolescence. Proc Natl Acad Sci U S A 109: 11366–11371. Rich-Edwards JW, Stampfer MJ, Manson JE, Rosner B, Hankinson SE, Colditz GA, Willett WC, Hennekens CH (1997) Birth weight and risk of cardiovascular disease in a cohort of women followed up since 1976. BMJ 315:396–400. Seckl JR (2004) Prenatal glucocorticoids and long-term programming. Eur J Endocrinol 151(Suppl 3):U49–U62. Small SA, Schobel SA, Buxton RB, Witter MP, Barnes CA (2011) A pathophysiological framework of hippocampal dysfunction in ageing and disease. Nat Rev Neurosci 12:585–601. Strauss RS (2000) Adult functional outcome of those born small for gestational age: twenty-six-year follow-up of the 1970 British Birth Cohort. JAMA 283:625–632. Szabo J, Cowan WM (1984) A stereotaxic atlas of the brain of the cynomolgus monkey (Macaca fascicularis). J Comp Neurol 222:265–300. Thompson C, Syddall H, Rodin I, Osmond C, Barker DJ (2001) Birth weight and the risk of depressive disorder in late life. Br J Psychiatry 179:450–455. Tupler LA, De Bellis MD (2006) Segmented hippocampal volume in children and adolescents with posttraumatic stress disorder. Biol Psychiatry 59:523–529.

205

Uno H, Tarara R, Else JG, Suleman MA, Sapolsky RM (1989) Hippocampal damage associated with prolonged and fatal stress in primates. J Neurosci 9:1705–1711. Uno H, Eisele S, Sakai A, Shelton S, Baker E, DeJesus O, Holden J (1994) Neurotoxicity of glucocorticoids in the primate brain. Horm Behav 28:336–348. Varela-Nallar L, Alfaro IE, Serrano FG, Parodi J, Inestrosa NC (2010) Wingless-type family member 5A (Wnt-5a) stimulates synaptic differentiation and function of glutamatergic synapses. Proc Natl Acad Sci U S A 107:21164–21169. Walhovd KB, Fjell AM, Brown TT, Kuperman JM, Chung Y, Hagler Jr DJ, Roddey JC, Erhart M, McCabe C, Akshoomoff N, Amaral DG, Bloss CS, Libiger O, Schork NJ, Darst BF, Casey BJ, Chang L, Ernst TM, Frazier J, Gruen JR, Kaufmann WE, Murray SS, van Zijl P, Mostofsky S, Dale AM (2012) Long-term influence of normal variation in neonatal characteristics on human brain development. Proc Natl Acad Sci U S A 109:20089–20094. Wu Z, Irizarry RA, Gentleman R, Martinez-Murillo F, Spencer F (2004) A model-based background adjustment for oligonucleotide expression arrays. J Am Stat Assoc 99:909–917. Xiao Y, Camarillo C, Ping Y, Arana TB, Zhao H, Thompson PM, Xu C, Su BB, Fan H, Ordonez J, Wang L, Mao C, Zhang Y, Cruz D, Escamilla MA, Li X (2014) The DNA methylome and transcriptome of different brain regions in schizophrenia and bipolar disorder. PLoS One 9:e95875.

(Accepted 5 January 2016) (Available online 12 January 2016)