Address correspondence and reprint requests to A. K. Kinnunen, Novartis Pharma AG, WSJ-360.616, CH-4002 Basel, Switzerland. E-mail: email@example.com
Exposure of pregnant women to stress during a critical period of fetal brain development is an environmental risk factor for developing schizophrenia in the adult offspring. We have applied a repeated variable stress paradigm to pregnant Sprague–Dawley rats during the last week of gestation coinciding with the second trimester in human brain development. Here we report our findings from a microarray analysis of the frontal pole of the prenatally stressed adult offspring and non-stressed adult controls complemented with measurement of plasma corticosterone levels following exposure to an acute stress. The direction of change of selected genes was confirmed by real time quantitative fluorescence PCR and in situ hybridization. The analysis revealed significant changes in genes associated with the NMDA receptor/postsynaptic density complex and the vesicle exocytosis machinery including NMDA receptor NR1 and NR2A subunits, densin-180, brain enriched guanylate kinase-associated protein, synaptosome-associated protein of 25 kDa, synaphin/complexin and vesicle-associated membrane protein 2/synaptobrevin 2. Interestingly, some of the changes in this animal preparation are analogous to changes observed in schizophrenic and bipolar patients. Our results suggest that application of a repeated variable prenatal stress paradigm during a critical period of fetal brain development reprograms the response of the hypothalamo–pituitary–adrenal axis to acute stress and results in gene expression changes that may have enduring effects on synaptic function in the offspring during adulthood.
In searching for a novel heuristic model for schizophrenia, rats were exposed to a repeated variable prenatal stress paradigm during the last week of pregnancy (Koenig et al. 2002b; Koenig et al. submitted). This period of rat brain development corresponds closely to the second trimester of human gestation (Bayer et al. 1993). The locomotor response to amphetamine and phencyclidine (PCP) was enhanced in young adult prenatally stressed (PNS) offspring. PNS offspring also displayed a disruption of sensorimotor gating as measured by prepulse inhibition (PPI) and sensory gating assessed using the N40 response (Koenig et al. 2002; Koenig et al. in preparation). Enhanced response to psychostimulants, loss of normal PPI and abnormal P50 gating (analogous to N40 gating in rodents) are common deficits in schizophrenic patients (e.g. Light and Braff 1999; Braff et al. 2001). The behavioral data prompted us to use microarray analysis, real time fluorescence based PCR and in situ hybridization histochemistry to examine the molecular changes resulting from exposure to the repeated variable prenatal stress paradigm. Enrichment of pre- and postsynaptic genes among the differentially expressed genes suggests exposure to repeated variable stress during a critical period of brain development may contribute to behavioral, molecular and neurochemical changes similar to those observed in schizophrenic and bipolar patients.
The repeated variable prenatal stress paradigm
Timed pregnant Sprague–Dawley rats (Charles River Laboratories, Wilmington, MA, USA) were subjected to a repeated variable stress paradigm from embryonic day (E) 14 and E22. The stress paradigm consisted of: (i) 1 h restraint in cylindrical plastic restrainers (Harvard Bioscience, Boston, MA, USA); (ii) exposure to a cold environment (+ 4°C, 6 h); (iii) overnight food deprivation; (iv) swim stress in room temperature water (15 min); (v) reversal of the light–dark cycle; and (6) an overnight exposure to social stress induced by overcrowded housing during the dark phase of the cycle. Stressors were applied in a randomized manner to prevent accommodation with up to three stress sessions each day, except for the 6 h cold stress which was the only stress performed that day; from E14–22, all stresses were performed at least twice and all mothers received exactly the same stressors. Control dams remained in the animal room and were not exposed to normal animal room maintenance procedures. Following birth, all dams and pups were left undisturbed in their cages until weaning on postnatal day (P) 22, when the male and female offspring were separated and housed with 1–2 same sex littermates per cage with free access to food and water. The animals used in this study were maintained in facilities fully accredited by the American Association for the Accreditation of Laboratory Animal Care (AAALAC), studies were approved by the University of Maryland School of Medicine IACUC and conducted in accordance with the Guide for Care and Use of Laboratory Animals provided by the NIH.
The acute stress procedure
On the morning of P56, after an acclimation period of 5–7 days to the laboratory environment, PNS male offspring and NS male control offspring were placed in cylindrical plastic restrainers for 30 min. At the end of the acute restraint session, the animals were either killed by decapitation or returned to their cages to be killed 120 min or 24 h after removal from the restraint device. Rats belonging to the baseline group were killed without exposure to the acute stress throughout the course of the experiment to control for circadian differences between the animals.
Measurement of the plasma corticosterone concentration
After decapitation, the trunk blood was collected in plastic tubes containing 10% EDTA. Plasma was obtained by centrifugation at 1200 g, and frozen at −80°C. Corticosterone concentrations were resolved using the Corticosterone 125I RIA Kit (ICN Diagnostics, Orangeburg, NY, USA) according to instructions from the manufacturer. Differences across groups were determined by anova and differences between group means were determined by Tukey–Kramer multiple comparisons test. A p-value of < 0.05 was considered statistically significant.
Microarray probe preparation and hybridization
Following killing, the brains were removed from the calvarium and placed on the dorsal side. The frontal pole was quickly dissected using the rostral margin of the olfactory tubercle as the posterior border, and the olfactory bulbs, which were removed, as the anterior border for the dissection. The samples were immediately frozen and stored at −80°C.
Total RNA was isolated using peqGOLD RNAPure™ (PEQLAB Biotechnologie GmbH, Erlangen, Germany), DNAse treated (10 U RNAse free DNAse 2 U/µL and 1 U/µL of the final volume RNAse inhibitor SuperasIn; 20 U/µL; both Ambion, Austin, TX, USA) and repurified using RNeasy (Qiagen Inc., Chatsworth, CA, USA). Samples were labeled and hybridized individually on rat genome RG-U34A microarrays (Affymetrix, Santa Clara, CA, USA) as previously described (Lockhart et al. 1996). Primary image analysis of the arrays was performed using the GeneChip 3.2 software package (Affymetrix) and images were scaled to an average signal intensity (average difference value) of 150.
Microarray analysis was performed using GeneSpring® 4.2.1 (Silicon Genetics, Redwood City, CA, USA) and Novartis Pharmacogenetics Gene Expression Analysis Tools (Novartis proprietary). Microarray quality was first examined and outliers identified based on the percentage of present values, the background noise level and clustering of the individual 48 microarrays using hierarchical clustering of all 8799 genes on the Affymetrix RG U34 A microarray based on Pearson correlation around zero (standard correlation with minimum distance of 0.001 and separation ratio of 0.5). After removal of the outlier microarrays, 25 NS controls and 23 PNS microarrays entered the analysis.
Data normalization and analysis
Three different normalization approaches were used: (i) The expression value for each gene on a chip was divided by the mean of all expression values on that chip assuming that this was at least 10. The bottom tenth percentile was used as a test for correct background subtraction. Thereafter, each gene was normalized to itself by dividing all measurements for that gene by the median of the gene's expression values over all the samples. (ii) The expression value for each gene on a chip was divided by the mean of all expression values on the chip. Thereafter, GAPDH 3′ expression value was divided by the average of GAPDH 3′ expression values across the whole sample set. The result of the GAPDH 3′/Avg GAPDH 3′ was then used to divide the expression values for other genes on that chip. The procedure brings the GAPDH 3′ expression value on every chip to the same value, and thus normalizes the global gene expression with respect to GAPDH 3′. (iii) No further normalization was applied after the average signal intensity on each chip was scaled to 150.
Following the data normalization using approaches (i) and (ii), only genes having raw data expression values = 100 in at least in one of the two conditions (NS and PNS) were included in the analysis. Genes showing a minimum fold difference of 1.5 in the average expression signal intensity between the NS and the PNS with a statistical significance of p = 0.005 (Wilcoxon test) were considered as differentially expressed. In approach (iii), genes were selected based on their logarithmic average difference (AvgLog) in their signal intensity between the PNS and the NS using the Novartis Pharmacogenetics Gene Expression Analysis Tools. Only those genes from the AvgLog list fulfilling the criteria of 1.5-fold and P = 0.005 (Wilcoxon) gene expression difference were included in further analyses.
It has been shown that different normalization approaches while aiming to decrease intersample variation produce different end results (Hoffmann et al. 2002). To identify a core group of genes which were consistently differentially expressed regardless of the normalization used, the gene lists generated by each normalization approach were compared. Only genes that were differentially expressed in at least two normalization approaches were identified. These sets of consistently differentially expressed genes were then used for further analyses.
Hierarchical clustering of the frontal pole experiment
Hierarchical clustering of the frontal pole experiment was performed with GeneSpring® 4.2.1 (Silicon Genetics, Redwood City, CA, USA) and applied on data normalized using approach 1. Gene expression similarity for all of the 8799 genes was measured by using Pearson correlation around zero (‘standard correlation’), with minimum distance of 0.001 and separation ratio of 0.99. Gene expression was the average expression of the six replicates per time point.
Hierarchical clustering of the genes
The most consistently differentially expressed genes were clustered hierarchically based on individual expression values across the whole sample set. Gene expression similarity was measured by using Pearson correlation around zero (standard correlation), with minimum distance of 0.001 and separation ratio of 0.5.
Self-organizing map algorithm (SOM; Kohonen 1991) was used to explore the variations in expression patterns within the frontal pole experiment. The experiment was first normalized using approach 1. Thereafter, to reduce noise, only genes having raw data expression values = 100 in at least one of eight conditions (where condition equals time point after the acute stress), and a change in gene expression at p = 0.1 (anova) at any of the eight conditions were included in the SOM analysis. For the frontal pole, 566 genes were included in the pattern analysis. SOM was applied in 7 × 9 grid format, the number of iterations was 10000 and the neighborhood radius was 4.
Real-time quantitative fluorescence-based PCR
Twelve differentially expressed genes were chosen for quantitation using a fluorescence-based real time PCR (Taqman, Applied Biosystems, Foster City, CA, USA). Primers and probes (Table S1; Microsynth, Balgach, Switzerland) were designed using ABI PrimerExpress software (Applied Biosystems, Foster City, CA, USA). All the Taqman probes were 5′ FAM (6-carboxyfluorescein)- and 3′ TAMRA (6-carboxytetramethylrhodamine)-labeled (Microsynth GmbH, Balgach, Switzerland). BLAST searches were performed to confirm primer and probe specificity. PNS (n = 16; four per time point) and NS (n = 16; four per time point) frontal pole RNAs were used as templates. The amount of total RNA was determined using the RiboGreen RNA quantitation method (Molecular Probes, Eugene, OR, USA). Four hundred nanograms of the total RNA (15 ng/µL) was treated with DNAse I mix (RNAse free DNAse kit, Qiagen Inc.). Five nanograms of the DNAse-treated RNA was used to control for genomic DNA-contamination using real time PCR and if negative, 250 ng of the RNA was reverse transcribed using random priming and Omniscript RT Kit (Qiagen Inc.) as described by the manufacturer. Five nanograms of cDNA was used as a PCR template. Primers were used at 300 nm and probes at 175 nm in qPCR Mastermix (Eurogentec, Seraing, Belgium) in a 25-µL reaction volume. The thermal cycle conditions used were: 2 min 50°C, 10 min 95°C, followed by 40 two-step cycles at 95°C for 15 s and one min at 60°C. The relative standard curve method (User Bulletin 2, PE Applied Biosystems, 1997) was used to determine the amount of mRNA of the gene of interest relative to an endogenous control gene. Phosphoglycerate kinase I (PGK1) and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) were chosen as control genes. For each gene, the samples were run as triplicates, and the reactions were repeated at least once. PCR for the gene of interest and for the endogenous control were always performed on the same reaction plate. The average and the standard deviation of the relative expression values were calculated and the significance of the change was tested using Student's t-test with unequal variance. A p-value of < 0.05 was considered significant.
NMDA-receptor-subunit in situ hybridization histochemistry
The cDNA template for the NR1 riboprobe was obtained from Dr Stephen Heinemann (Salk Institute, La Jolla, CA, USA; Hollmann et al. 1993) and the NR2A probe from Dr Dennis J. Goebel (Wayne State University, Goebel and Poosch 1999). The in situ hybridization was performed according to published protocols (Bischoff et al. 1997). The sections were dried and exposed to Hyperfilm (Amersham, Piscataway, NJ, USA). The films were imaged through a Sony CCD camera interfaced with a Power Macintosh computer via a Data Translation FrameGrabber digitizer. The relative gray-scale values were measured using NIH Image software (version 1.62) and the level of hybridization in medial prefrontal cortex was expressed as a percentage of the control hybridization. A non-parametric statistical analysis was used to determine the level of significance between the NS and PNS animals.
Corticosterone and global gene expression responses to acute stress
Plasma corticosterone concentrations were measured on P56 in young adult PNS and NS rats during basal non-stressed conditions and immediately, 2 and 24 h after an acute 30 min restraint stress. Under baseline conditions, the plasma corticosterone concentrations in the PNS and in the NS animals were not statistically different (Fig. 1a). Acute exposure to restraint stress for 30 min resulted in a dramatic and highly significant elevation in plasma corticosterone concentrations in the NS animals (p < 0.001) which was similar in the PNS group. Two hours after removal from the restraining device, plasma corticosterone concentrations declined to baseline levels in the NS animals. In the PNS animals however, plasma corticosterone concentrations remained significantly elevated at 2 h but returned to baseline levels after 24 h.
Unsupervised hierarchical clustering analysis of all 48 frontal pole microarray datasets based on the average gene expression in each treatment group for all 8799 genes, grouped the NS and PNS datasets in two clearly separate clusters (Fig. 1b). Gene expression 2 h after acute stress differed most from the baseline condition in both the NS and PNS clusters. Whereas the global gene expression pattern almost returned to baseline patterns 24 h after acute stress exposure in the NS group, global gene expression 24 h after the acute stress in the PNS group more closely resembled the 30 min gene expression pattern showing that maternal stress in utero had markedly influenced global gene expression in response to acute stress. The PNS and NS groups divided into two separate clusters in line with the gestational stress status. Interestingly, in the PNS group corticosterone concentrations 2 h after acute stress exposure were significantly higher than in the NS group where concentrations had returned to baseline (Fig. 1a). This is paralleled in the global gene expression pattern at 24 h in the PNS cluster (Fig. 1b).
Genes in the NMDA-receptor/postsynaptic density complex are differentially regulated in the PNS animals
To determine the differentially expressed genes in the frontal pole between the NS (n = 25) and PNS (n = 23) animals, three ways of normalizing the data and three independent analyses of differentially expressed genes were conducted. Thirty-four genes were identified as differentially expressed using approach (i), 43 with approach (ii), and 57 with approach (iii) between the NS and the PNS groups (Fig. 2a). It is recognized, and shown by this study, that different normalization approaches while aiming to decrease intersample variation produce different end results (Hoffmann et al. 2002). To identify a core group of consistently changed genes only genes that were identified in at least two normalization approaches were included. Using this unusually stringent restriction, 35 consistently differentially expressed genes were identified in the frontal pole (Fig. 2a) and subjected to hierarchical clustering (Fig. 2b).
Of the 35 differentially expressed genes, interestingly, three were neurotransmitter receptor subunits or associated proteins: one NMDA-receptor associated protein (NMDA receptor glutamate-binding subunit/grina), and two GABA-receptor subunits (GABAB 1c and GABAAγ−1). Two of the genes are involved in calcium/calmodulin signaling (CaMKII inhibitor α, CaMKII γ), two are associated with neurotransmitters vesicles or vesicle recycling (synapsin 2 and Vamp2/synaptobrevin II) and two are sodium-potassium-ATPases catalyzing the hydrolysis of ATP coupled with the exchange of Na+ and K+ ions across the plasma membrane (Na+,K+-ATPase α1 and Na+,K+-ATPase α2). Seven genes can be classified as part of the postsynaptic density complex (NMDA receptor glutamate-binding subunit/grina, CaMKII γ, CaMKII inhibitor α, densin-180, guanylate-kinase associated protein 1 (begain), rhoC/ras homolog 9 and a ras-related GTPase, ragB).
Twelve genes (aldolase A, Na+,K+-ATPase α2, aiPLA2, grina, vamp2, GABAB 1c, RBCKII, carbonic anhydrase II, densin-180, synapsin-2, NR1, ribin) were selected randomly for real time fluorescence-based RT-PCR assessment (Taqman assay; Table 1). The expression differences for NR1, densin-180 and GABAB 1c were significantly different (p < 0.05) by Taqman assay across the sample set and, these were considered confirmed. Five genes (aldolase A, aiPLA2, grina, synapsin 2, vamp2, Na+,K+-ATPase α2; p < 0.1) showed the same trend of change as observed by microarrays. For three genes, the Taqman method did not confirm the microarray finding (carbonic anhydrase, RBCKII, ribin; p > 0.1). These data may reflect differences in sensitivity of the two methods, normalization or, probe design.
Table 1. Fluorescence based quantitative real time PCR (Taqman) assessment of the microarray finding
Relative amount of mRNA1
% Change in the PNS
Normalized to PGK1. aiPLA2, aldolase A, Na,K-ATPase α2, RBCK2 were normalized to GAPDH. For all, n = 14–16 in NS and PNS across all time points. p < 0.05 (confirmed).
NMDA-receptor subunits NR1 and NR2A are differentially expressed in the PNS animals
Intrigued by the possibility of NMDA hypofunction underlying some of the behavioral changes observed in the PNS animals, the expression patterns of the NMDA receptor subunits NR1, NR2A, NR2C and NR2D as well as the NMDA-receptor glutamate-binding subunit/grina in the frontal pole were studied (Fig. 3).
Analysis of the microarray data revealed that NR1, the glutamate-binding subunit/grina and NR2A were differentially regulated by prenatal stress (Fig. 3a,b). NR1 and grina were significantly down-regulated (p < 0.0005 and p < 0.005, Wilcoxon, respectively) in the frontal pole in the PNS offspring compared to the NS group, whereas NR2A was significantly up-regulated (p < 0.005, Wilcoxon). Of the NMDA-receptor subunits, NR1 and NR2A were selected for further studies using real time quantitative PCR and in situ hybridization. Significant down-regulation of the NR1 and NR2A subunits in the PNS frontal pole and prefrontal cortex was confirmed by quantitative in situ hybridization (Fig. 3e–h) and for NR1 also by quantitative real time PCR (p < 0.05; Fig. 3d).
Functional classification of the differentially expressed genes
While our stringent strategy for identifying differentially expressed genes effectively decreased the number of false positive findings, it also increased the number of false negatives. Therefore, after the limiting first round of analysis of the dataset, the criteria were relaxed to include differentially expressed genes arising from any normalization approach. This identified 66 genes with known function which were divided into 30 functional groups to gain an overall understanding of the functional systems affected by prenatal stress (Table S2). The functional classification was performed based on revised annotation information in the Affymetrix Netaffix™ (http://www.affymetrix.com; for RGU34 index date for annotations 07-05-2002), GenBank Unigene, Swall and Swissprot databases and recent publications. Genes involved in vesicle trafficking, transcriptional regulation, and the pre- and postsynaptic compartments of the synapse constituted the largest functional groups. Taken together the contribution of the calcium/CaMK group with three differentially expressed genes, the glutamatergic system with two NMDA-receptor subunits and the NMDA-receptor associated protein, grina and the group of four postsynaptic density genes (densin-180, BEGAIN, PSD-SAP90-associated protein-3, CRIPT) pointed to a remarkable regulation of genes of the postsynaptic complex. Furthermore, among the differentially expressed genes, there are some small GTPases, kinases, phosphatases and heat shock proteins which have been found in the postsynaptic density (Husi et al. 2000; Sheng and Pak 2000; Sheng and Kim 2002). Also striking was the number of differentially regulated genes belonging to the vesicular exocytosis and endocytosis machineries. Two of the genes (SNAP-25 and VAMP-2) encode proteins forming the SNAP receptor (SNARE) protein complex (Sudhof 2000; Lin and Scheller 2000) which plays a central role in neurotransmitter release. SNAP-25 and VAMP-2 were significantly down-regulated by prenatal stress. Interestingly, synaphin 2/complexin II, that regulates the oligomerization of the SNARE complex (McMahon et al. 1995; Tokumaru et al. 2001), a step essential for synaptic vesicle exocytosis, was also significantly down-regulated in the PNS group using normalization approach 3 (Table S2). Of endocytosis-associated genes, the expression of epsin and clathrin associated protein 17 were affected by prenatal stress (Table S2).
Cluster analysis of the frontal pole differentially expressed genes
Using the SOM-algorithm (Kohonen 1991) the changes in the gene expression patterns within the frontal pole samples were correlated. The algorithm places genes with similar expression profiles into nodes and arranges different nodes with similar average expression patterns next to each other on a 7 × 9 rectangular grid. This neighborhood function of the SOM allows visual recognition of larger clusters of genes with similar expression patterns. After prefiltering the microarray data to reduce noise ratio, 566 genes were included in the pattern analysis of 7 × 9 SOM resulting in 63 nodes. Eleven of the total 63 nodes formed two separate clusters with opposite expression profiles and contained the 35 most consistently changed genes (cluster 1 and cluster 2; Table 2). Cluster 1 contained 54 and cluster 2 28 genes, 28 and seven of which, respectively, belonged to the most consistently changed genes.
Table 2. SOM clustering result for the 35 most consistently differentially expressed genes (in bold italics) in the frontal pole
Na,K-ATPase α2 subunit
Na/K ATPase α1 subunit
kinase II gamma
ESTs, 18S, 5.3S and 28S rRNAs
CaM-kinase II inhibitor α
ESTs, by synteny CGI-31
GTP-binding protein (G-α0)
ESTs, similar to SK2
ESTs, 45S rRNA
Cox IV b
ESTs, Highly similar to hypothetical protein MGC4175
NMDA receptor glutamate-binding
phospholipase C form-I (PI-PLC I)
ras-related GTPase, ragB
MAP/microtubule affinity-regulating kinase 3
ATPase isoform 2, Na+ K+ transporting, β polypeptide 2 ESTs, highly similar to hypothetical protein FLJ11046
DNA-binding protein C1D
Smooth muscle myosin
Glucose-regulated protein GRP78
Repetitive DNA sequence LINE3
Kruppel-like factor 13
2.4 kb repeat DNA
Potassium channel TWIK
Intercellular adhesion molecule-1
Carbonic anhydrase II
2.4 kb repeat DNA
GABA(A) receptor α1 subunit
estrogen receptor beta
Casein kinase I alpha L (CKIaL)
L1Rn B6 repetitive DNA
Long interspersed repetitive DNA
The concept of prenatal programming has been a long-standing topic in neuroendocrinology research and there is substantial evidence that prenatal stress changes or reprograms brain function, particularly the HPA axis (for reviews, see Weinstock 2001; Weldberg and Seckl 2001; Seckl 2001). This study addresses for the first time the differences caused by a variable prenatal stress paradigm at the level of global gene expression in the frontal pole using the microarray approach. In addition to cognitive functions, prefrontal cortex as part of the frontal pole, plays an important role in the negative feedback regulation of the HPA axis (Diorio et al. 1993; Akana et al. 2001; Moghaddam 2002; Vermetten and Bremner 2002). The results of this study demonstrate how exposure of pregnant female rats to a repeatedly applied, variable stress paradigm during a critical phase of fetal brain development changes the response of the HPA axis to the acute stress in the adult male offspring. The change in the HPA axis is coupled to changes in gene expression in the frontal pole (Figs 1b and 3). Behaviorally, PNS animals display deficits in sensorimotor and sensory gating and stress reactivity (Koenig et al. 2002b; Koenig et al. submitted). At the molecular level, PNS offspring exhibit gene expression differences in essential functional systems including genes encoding proteins for neurotransmitter release, receptor function and signaling (Figs 2 and 3, Table S2).
Prenatal stress increases maternal corticotropin and corticosterone concentrations. Corticosterone can readily penetrate the fetal brain (Zarrow et al. 1970) and interact with specific glucocorticoid receptors that are present during the last week of gestation in the rat (Meaney et al. 1985; Cintra et al. 1993). Glucocorticoid receptors are nuclear hormone receptors which function as ligand-activated transcription factors directly mediating transactivation of target genes by binding sequence specific recognition elements (glucocorticoid response elements; Whitfield et al. 1999). Glucocorticoid receptors are also known to interact with multiple transcription factors, such as c-jun, nuclear factor-kappa B, the TFIID complex, STAT5, and coactivators known to modulate the function of these signaling molecules (Jenkins et al. 2001; Yudt and Cidlowski 2002). Prenatal stress also increases maternal and fetal catecholamine release (Morishima et al. 1978; Roehde et al. 1989), maternal oxytocin and opioid peptides of which β-endorphin is able to cross the placenta (Sandman and Kastin 1981; Neumann et al. 1998). Molecular pathways leading from prenatal stress to the neuroendocrinological, behavioral, molecular and neurochemical changes in the adult remain poorly understood. In addition to complex glucocorticoid regulated mechanisms, it is likely that the plasticity of the developing brain monoaminergic system participates in these changes (Weinstock 2001; Seckl 2001; Welberg and Seckl 2001).
It was surprising, that many of the observed gene expression changes in the PNS group were analogous to observations made in microarray analyses of brain samples from BA9 and BA46 prefrontal cortex regions of schizophrenic patients. Of the 35 differentially expressed genes in the frontal pole, aiPLA2, vamp2 and synapsin 2 have been previously indicated as differentially expressed or associated with schizophrenia or bipolar disorder (Ross et al. 1997; Ross et al. 1999; Mirnics et al. 2000; Sokolov et al. 2000; Vawter et al. 2002). Some changes in the PNS animals were opposite to findings in humans and it is likely that some of these differences could be explained by medication. Interestingly, in a recent microarray analysis of acute clozapine treatment in rats (Kontkanen et al. 2002), the functional group of presynaptic genes with vesicle exocytosis function was found to be differentially regulated by clozapine in the prefrontal cortex. It is tempting to speculate that antipsychotic treatment may target some of the same functional groups that are affected by prenatal stress. In light of the possible role of the NMDA-receptor complex in schizophrenia and its central function in cognition (Goff and Coyle 2001; Greene 2001; Wittenberg and Tsien 2002), it is remarkable that the microarray analysis identified changes in the gene expression of components of the NMDA-receptor/PSD complex as one of the largest functional entities affected by prenatal stress (Table S2). The postsynaptic membrane in the brain's excitatory synapses is specialized for responding rapidly to glutamate released from the presynaptic terminal. Associated with the postsynaptic membrane is the postsynaptic density (PSD) which contains a high concentration of glutamate receptors and an ordered array of regulatory, scaffolding and cytoskeletal proteins, including the PSD95-family of membrane-associated guanylate kinases, that in the postsynaptic neuron, initiate and modulate signal transduction (Kennedy 1998; Walikonis et al. 2000; Sheng and Kim 2002). Of the PSD95-family, only PSD-95 and chapsyn110/PSD-93 are present on the Affymetrix RG-U34A microarray. Although the level of expression of the PSD-95/SAP90 was not significantly changed by gestational stress, expression of many of the PSD-95-interacting partners such as BEGAIN, densin-180, NMDA receptor subunit NR2A, CamK II γ, small GTPases and G-proteins were affected (Table S2) rendering it possible that gestational stress could affect the expression of other members of PSD-95 family. Using a correlation based approach, SOM clustering revealed some interesting candidate genes which may be coregulated with PSD-associated proteins. In Node 2.7 (Table 2), the NMDA-receptor subunit NR2A and BEGAIN which are known to interact with the PSD95/SAP90 (Deguchi et al. 1998) and synaptic scaffolding molecule S-SCAM (Hirao et al. 2000) clustered together with another postsynaptic density protein, densin-180 (Apperson et al. 1997). Interestingly, the BEGAIN interacting proteins PSD95/SAP90 and S-SCAM both interact with the C-terminus of the NR2A subunit via their PDZ-domains and likely participate in the assembly of the NMDA receptors (Lin et al. 2001). Also densin-180 has been recently shown to bind PSD95 via its direct interaction with MAGUIN-1 (Ohtakara et al. 2002) and is thus indirectly interacting with NR2A. Furthermore, densin-180 has been shown to interact with calcium/calmodulin kinase II (CaMKII) and could play an important role in localizing or translocating CaMKII at the postsynaptic density after its synthesis (Walikonis et al. 2001). Altogether, three members the calcium/calmodulin signaling were down-regulated in the frontal pole of the PNS animals, including CaMK inhibitor α, calcium–calmodulin phosphodiesterase (CaM-PDE), and CaMKII γ-subunit (Table S2). CaMKII is enriched at the postsynaptic density and it has been demonstrated to be essential for synaptic plasticity and learning (reviewed in Lisman et al. 2002). Down-regulation of the CaMK-related genes in the frontal pole of the PNS is particularly interesting, since cognitive impairments have been repeatedly reported in offspring after maternal exposure to stress. The gene expression differences described in the frontal pole in our study could contribute to the cognitive impairments observed after exposure to gestational stress.
NMDA receptor function is under complex regulation during physiological and pharmacological states. Recent data suggest that NMDA receptors on the cell surface may be regulated via a process involving vesicular trafficking and intracellular vesicular pools (Carroll and Zukin 2002). Our finding of a down-regulation in the expression of the key members of the vesicle exocytosis machinery, vamp-2, SNAP-25 and synaphin 2/complexin II and a PKC-binding protein (Figs 2 and 3; Table S2) is therefore interesting from several points. Dysregulation of the vesicle exocytosis machinery in the PNS animals may reflect a defect in neurotransmitter release, that could contribute to the observed behavioral impairments and some of the receptor changes observed. It may also reflect either a slower or a faster vesicle recycling in the PNS animals without having an effect on the gross vesicle exocytosis (Sudhof 2000). On the other hand, PKC-regulated NMDA-receptor trafficking has been shown to occur via SNARE complex-dependent exocytosis (Lan et al. 2001; Scott et al. 2001) and down-regulation in this function could thus result in the decrease in the delivery of NMDA receptors to the cell surface. Exposure to acute stress in normal animals increases the glutamate levels in the prefrontal cortex and up-regulates the expression of NR1 and NR2 subunits (Moghaddam 2002). We report here a significant down-regulation of NR1 in the frontal pole of adult PNS animals across the whole sample set (Fig. 3a,d,e). The paradoxical down-regulation in NR1 expression implies that the PNS animals have lost some of the normal plasticity in the central glutamatergic system. Furthermore, down-regulation of the common subunit of the NMDA-receptor is likely to result in a hypoglutamatergic state in the frontal pole of the PNS animals. We also observed a significant up-regulation of the overall NR2A expression in the frontal pole of the PNS animals (Fig. 3b,f,h). In light of recent evidence, an up-regulation of NR2A in the absence of concomitant up-regulation of the NR1-subunit may lead to an increase in the number of extrasynaptic NMDA receptors and consequent activation of excitotoxic cell death pathways (Hardingham et al. 2002).
This study reconfirms observations that exposure to stress during pregnancy reprograms the offspring's HPA axis as adults, resulting in greater and prolonged elevation of plasma ACTH and/or corticosterone after acute stress (Fig. 1a; McCormick et al. 1995; Barbazanges et al. 1996; Koehl et al. 1999). Using microarray technology, we show in PNS adult male offspring that the global gene expression pattern follows a similar stress response pattern (Fig. 1b). Importantly, the effect of repeated variable stress in utero over-rides the effect of acute stress on gene expression since the NS and PNS treatment groups cluster separately according to the gestational stress status (Fig. 1b). Furthermore, prenatal stress results in differential expression of a considerable number of genes between the two groups (Figs 2 and 3; Table S2). Many of these genes are central to function of the glutamatergic synapse (Table S2).
The results imply that maternal exposure to repeated variable stress paradigm during a critical period of fetal brain development could alter the normal trajectory of brain development disrupting not only the normal stress response and the HPA axis, but also altering global gene expression in different brain regions. It is intriguing and, analogous to most schizophrenic patients, that the behavioral deficits in the PNS animals appear postpubertally (Koenig et al. 2002b; Koenig et al. submitted). The onset of schizophrenia is frequently precipitated by a stressful event and psychological stress is known to exacerbate psychotic symptoms in humans (Duncan et al. 1999). The repeated variable prenatal stress paradigm described here, offers a testable preclinical and clinical model that may be helpful in defining some stress provoked molecular aspects of human disorders displaying deficits in sensorimotor and sensory gating and cognition.
We thank Rainer Maier and Christine Sturchler for valuable discussions in the course of the data analysis; Jose Luis Crespo, Doris Rueegg and Dana Brady for excellent technical assistance with RNA preparation and the animal model; Stephanie Schultz, Lisa M. Barnes, Christian Lavedan and Mihael Polymeropoulos at the Novartis Pharmacogenomics for the microarray hybridization.