Gene expression profiling in the stress control brain region hypothalamic paraventricular nucleus reveals a novel gene network including Amyloid beta Precursor Protein

Basal gene expression differences between C57BL/6J and DBA/2J mice in PVN

To compare expression profiles between the two mouse strains C57BL/6J and DBA/2J and evaluate expression changes at different time points after stress exposure, we used cDNA microarrays. We first evaluated the differences of the two mouse strains in their basal expression profile in the PVN (see Fig. 1A for scheme of all comparisons). The microarray analysis revealed 670 genes with more than 1.4 fold difference in expression.

ACTH and gene expression profiles after forced swimming

To determine transcriptional profiles after stress, mice had been subjected to forced swimming for 5 min at 8:00 h or 12:00 h and then decapitated at 16:00 h in order to isolate the brains for our microarray studies as previously described [21] (experimental plan depicted in Fig. 1B). In this previous study the impact of forced swim stress on stimulation of the HPA axis was evaluated by the determination of corticosterone (CORT) levels in the serum. In order to monitor the activation of the pituitary that responds to the CRF release, we also measured the ACTH levels. A set of animals of both strains was decapitated directly after stress exposure at 08:00 h and at 12:00 h (non stressed control animals were also included). The results show a clear increase for ACTH after stress in both mouse strains and at both time points selected in our study (Fig. 2), like previously also for CORT [21]. Thus, the pituitary as well as the adrenals responded immediately to the stressor. When animals were decapitated 4 h or 8 h after the stress, the levels of ACTH were not different from the non stressed controls, indicative of a functional negative feedback mechanism (Fig. 2). The levels in non-stressed and stressed animals were higher at 16:00 than in the morning though, most likely due to the circadian rhythm of the mice.

To reveal gene expression differences, samples from non stressed mice were compared to samples from mice that had been stressed 4 h or 8 h before decapitation. In C57BL/6J mice, 123 genes were ≥1.4 fold regulated 4 h after stress and 88 genes 8 h after stress. In DBA/2J mice, 185 genes and 96 genes were regulated at the respective time points. (Additional File 2, 3, 4 and 5).

In DBA/2J mice, among the genes regulated 4 h after stress, mitogen activated protein kinase 3 (MAPK3), Guanine nucleotide binding protein, alpha inhibiting 2 (GNAi2), nuclear factor of activated T-cells, cytoplasmic, calcineurin-dependent 1 (NFATC1) were most striking (Table 4), while among the genes regulated 8 h after stress Amyloid beta (A4) precursor protein (APP), cyclin-dependent kinase inhibitor 1B (P27) (CDKN1B), Transcription factor 7-like 2, T-cell specific, HMG-box (TCF7L2) were most prominent (Table 5).

The reaction to stress on the transcriptome level differs between mouse strains and displays phases

Possible signalling pathways elicited after forced swimming in the PVN

The observation that the percentage of receptors and signalling molecules among the regulated genes decreased from 4 h to 8 h after stress (Table 6), together with the phased reaction of the transcriptome to stress (Table 7), led us to hypothesise that genes regulated at 4 h have pathway connections to genes regulated at 8 h. Employing a pathway building program to test this hypothesis, we identified links between genes responding at 4 h and genes responding at 8 h. For example, GNAi2, found to be up-regulated 4 h after stress in DBA/2J mice, is upstream of APP, which is up-regulated 8 h after stress. Additionally, NFATC1, found to be up-regulated 4 h after stress, is upstream of heat shock protein 70, HSPAIA (HSP70), which increases 8 h after stress (Fig. 3A).

In C57BL/6J mice, p21[CDKN1A]-activated kinase 2 (PAK2), down-regulated 4 h-after stress, inhibits the expression of 'expressed in non-metastatic cells 1 protein' (NME1) via regulation of tumor necrosis factor alpha (TNFα, FIG. 3B). NME1, in turn, is upstream of sprouty homolog 2 (Drosophila; SPRY2), which is regulated 8 h after stress. Another connection between genes that are both regulated 8 h after stress is between GNAO1 and RGS2 (Fig 3B).

Validation of GNAi2 and APP expression and regulation in the PVN

To validate and further analyse the expression changes of the genes GNAi2 and APP that are linked by a pathway in the PVN of DBA/2J mice, we used real-time PCR. RNA samples from the original punctures were amplified (one round of amplification) and subjected to RT-PCR without pooling. The results confirmed the up-regulation of GNAi2 4 h after stress (2.2 fold, Fig. 4A) detected by the microarray (2.1 fold). To test whether this regulation is specific for DBA/2J mice, or may also occur in C57BL/6J mice (but was not detected in our microarray screening, because of the selection filtering), we also tested the respective samples from C57BL/6J mice. The results showed a non-significant increase in this mouse strain, implying that the regulation is rather strain-specific (Fig. 4A).

Similarly to GNAi2, we validated the expression and regulation of APP 8 h after forced swimming by real-time PCR (Fig. 4B), which was found in the microarray analysis.

To visualize the regulation of this expression with spatial resolution in the PVN, in situ hybridization was performed on coronal brain-sections, followed by semi-quantification of the mRNA signal. The results showed a strong signal in the PVN area (Fig. 4C) and the analysis confirmed again the up-regulation to a level of 1.6 fold (data not shown). In situ hybridization showed also specific expression of APP in the PVN (Fig. 4C).

Finally, to test whether stress-regulation of GNAi2 and APP is specific to the PVN, we also analysed the hypothalamic region just anterior and posterior to the PVN. The results show no significant change in the expression levels of GNAi2 and APP for the respective time point (Fig. 4D).

Clustering analysis of GNAi2 - APP

Having corroborated a potential role of the GNAi2 - APP connection in stress response of DBA mice, we used clustering analysis to identify genes that display similar expression changes throughout the conditions we analysed. In setting up the clustering analysis, we considered both up- and down-regulation, because some transcriptional regulators, such as GR, are able to both up- and down-regulate genes. Genes identified in a clustering analysis may be under a common transcriptional control, or influencing each other.

The dendrograms revealed 10 and 15 genes for GNAi2 (two transcripts of this gene were spotted) and 12 for APP (Fig. 5, see also Additional File 6 for the expression profiles). Several genes with known function were among these nearest neighbours.

To test potential, already described connections between these genes, we again used a pathway building program. Interestingly, we found that GNAi2 is located in a signalling cascade, where 'Heart and neural crest derivatives expressed transcript 2' (HAND2) is upstream, APP is downstream and MAP3K2 further downstream. DHDDS and GNAi2 share a common upstream regulator (AGT) as well as a common downstream target (PARK2). GNAi2 and COPS5 have a common target as well. In addition, GNAi2 and APP are found both downstream of another common regulator (RBL2, Fig. 6A). Furthermore, for APP we identified common upstream regulators with PAPOLA (CDC2 and TNF), EN1 as its regulator and HSD17B4 as its target. Interestingly, DGKZ and APP are linked in a feedback loop (Fig. 6B). Thus, the clustering analysis revealed functionally related genes, indeed.

Animal Experiments

Our study was performed with C57BL/6J and DBA/2J male mice with an age of 3-5 months (supplied from Harlan Winkelman, Borchen) and single housed under a 06:00-18:00 light cycle and subjected to forced swimming to induce an emotional stress, as previously described (Tsolakidou et al 2008). Briefly, the animals were placed in a 24 cm high 11 cm diameter cylinder filled with water at 22-25°C for 5 min. Afterwards they were transferred to their own original cage (i.e. single housed again). The animals elicited a mixture of behavioural activity that can be described as climbing, swimming and immobility, with the latter reflecting a passive stress coping behaviour. The experimental procedures were conducted in accordance with the Guide for the Care and Use of Laboratory Animals of the European Union (European Communities Council Directive 86/609/EEC) and approved by the Government of Bavaria, Germany.

The mice were separated into three groups per strain as previously described (Tsolakidou et al 2008) (see Fig. 1B). The first and the second group were subjected to stress at 8:00 or at 12:00, respectively. The third group was not stressed and further on used as reference group for the microarray experiments. All animals were decapitated at 16:00 to avoid possible interference by circadian variations of corticosterone (CORT) levels. Thus, the first and the second group have been actually sacrificed 8 h or 4 h after stress, respectively. Trunk blood was collected for determination of ACTH concentrations (10 animals/condition) and dissected brains from the same animals were frozen on dry ice and stored at -80°C (to be used for expression profiling). To monitor hormone levels acutely after stress, a small set of animals (5/condition) were sacrificed acutely at the respective time points (8:00 and 12:00). Plasma ACTH concentrations were determined in a radioimmune assay (ICN Biomedicals).

Micropuncture and RNA preparation

Micropuncturing of the PVN and adjacent region on coronal tissue sections (200 μm thick) was applied under dry ice-cold conditions. To control for the accuracy of the puncture the sections were stained (cresyl violet) afterwards. Total RNA was extracted from the collected tissue (Trizol protocol, GibcoBRL). Samples from six animals were pooled to minimize the impact of biological variance, which is intrinsic to all organisms and can be substantial even in inbred mice [57]. After two rounds of amplification (Amino Allyl MessageAmp aRNA kit, Ambion) the RNA was labelled with Cy3 or Cy5 dyes (Amersham)

Microarray hybridization and Analysis

Spotted cDNA microarray chips (MPIP 24K microarray platform with sequences from 19900 different genes, [58] were used, which allow a greater degree of flexibility in the choice of arrayed elements. Internal controls (sequences from other organisms) were used as contribution to ensure the quality of the data [59]. The microarray experiments were performed by competitive hybridisation of two differentially labelled (Cy3 or Cy5) probes of amplified total RNA samples. 10 arrays were used for each comparison, that is 5 technical replicates and a dye-swap with another 5 technical replicates. 20 μg of each Cy3- or Cy5-labeled sample were denatured at 95°C for 3 min in hybridisation buffer (50% formamide, 50 mM sodium phosphate-buffer pH 7.0, 5× Denhard's solution [Sigma, Taufkirchen, Germany], 6× SSC, 0.5% SDS, 0.4 mg/ml murine COT1-DNA [Invitrogen] and 5 μg poly(dA) [Amersham]). The hybridisation was performed in chambers submerged in a water bath at 42°C for 16 h. The arrays were washed for 15 min with 2× SSC/0.2% SDS at 60°C, in 0.5× SSC for 15 min at 60°C, rinsed in 0.2× SSC for 1 min at room temperature, shaken vigorously in 0.05× SSC at room temperature and finally air dried. All slides were scanned immediately afterwards.

Scanning was performed using a ScanArray 4000 laser scanner and ScanArray 3.1 Software (Perkin Elmer, Boston, USA) with a fixed PMT gain of 80%, and 98% (Cy3) or 70% (Cy5) laser power. The QuantArray software 2.1.0.0 (Perkin Elmer, Germany) and the fixed circle-analysis method were used to perform the quantification. Data were imported into a PostgreSQL relational database for further analysis. Raw data were normalized according to the procedure outlined elsewhere [60] and subjected to a two-sided one sample t-test for significantly differential expression. Candidate genes were screened for using thresholds of |fold regulation| 1.414 and |Z-score| 1.423. In order to exclude any possibility of wrong annotation the majority of the differentially expressed genes, in particular those of major interest were sequenced. All data are deposited at the GEO server (GSE20877).

Ontology Groups Classification and Molecular Pathways Building

To classify and characterize the differentially expressed genes that resulted from the microarray analysis the Gene Ontology data annotation (version 161) (http://www.geneontology.org, http://www.ebi.ac.uk/GOA/) was used. In addition, the text mining program Pathway Studio 5.0 (Ariadne Genomics, Rockville, USA) was used for the identification of pathways linking the regulated genes. Downstream targets of the genes regulated 4 h after stress were looked for in the group of genes regulated 8 h after stress and upstream regulators of the latter in the former group. The main criteria applied for the final selection of the pathways were: 1. the last step of the pathway has to indicate molecular synthesis/expression so as to validate the expression change described on the microarray result, 2. confirmation of the annotation of the selected genes by sequencing and 3. validation of the literature references used by the program.

In situ Hybridisation

18 μm thin cryostat sections prepared from frozen stored brains were thaw-mounted on poly L-Lysine-coated slides, dried and kept at -80°C. [α-³⁵S] UTP-labelled riboprobes for amyloid β (A4) precursor protein (APP) and guanine nucleotide binding protein, alpha inhibiting 2 (GNAi2) genes were prepared by in vitro transcription from corresponding cDNA clones (sense probes were also included for control). The sections were dried, fixed in 4% paraformaldehyde, washed in PBS (3 times) and subjected to acetylation using 0.25% acetic anhydride in 0.1 M triethanolamine-HCL, pH 8.0. After subsequent dehydration in increasing concentrations of ethanol, brain sections were saturated with 100 μl of hybridisation buffer containing ~ 2.5 × 10⁶ cpm ³⁵S-labelled riboprobe. Brain sections were incubated overnight at 62°C or 58°C. Then, the sections were rinsed in 4× SSC (standard saline citrate), treated with RNAse A (20 mg/L) and washed in increasingly stringent SSC solutions at room temperature. Finally sections were washed in 0.1× SSC for 1 h at 67°C or 64°C (respectively) and dehydrated through increasing concentrations of ethanol. Autoradiography was on Biomax MR film (Eastman Kodak Co., NY, USA) for 3-5 days. The autoradiographs were digitised, and relative expression levels were determined by computer-assisted optical densitometry (Image J, Scion Corporation). The average value of 4-6 measurements was calculated from each animal [61].

Quantitative (q)PCR

A total of 200 ng of amplified RNA from the first amplification round of the microarray analysis was reverse transcribed with Superscript II (Invitrogen, Karlsruhe, Germany) using random hexamer primers (Ambion) according to the manufacturer's protocol. For quality control, a small aliquot of each cDNA was analyzed on an agarose gel.

cDNA of DBA/2J and C57BL/6J mice from the non stressed, 4 h- and 8 h-after stress groups (for DBA/2J: N = 6 in each group; for C57BL/6J: N = 5 for non stressed and 8 h-after stress samples each, N = 6 for the 4 h stress group), was analyzed by qPCR, using the QuantiFast SYBR Green PCR Kit (Qiagen GmbH, Hilden, Germany) according to the manufacturer's instructions. The respective oligonucleotide primers were for GNAi2 (fwd: 5' CACCTCCATCATCCTCTTCC 3', rev: 5' GCACGTGAAGTGCGTGTAGA 3'), APP (fwd: 5' CGGAAGAGATCTCGGAAGTG 3', rev: 5' TGTTCGAACCCACATCTTCA 3') and the house keeping genes glyceraldehydes-3-phosphate dehydrogenase (GAPDH: fwd: 5' CCATCACCATCTTCCAGGAGCGAG 3', rev: 5' GATGGCATGGACTGTGGTCATGAG 3') hypoxanthine guanine phosphoribosyl transferase 1 (HPRT1; fwd: 5' GTCAAGGGCATA TCCAACAACAAAC 3', rev: 5' CCTGCTGGATTACATTAAAGCACTG 3') polymerase (RNA) II (DNA directed) polypeptide B (POLR2B; fwd: 5 'CAAGACAAGGATCATATCT GATGG 3', rev: 5' AGAGTTTAGACGACGCAGGTG 3') or ribosomal protein L13a (RPL13A; fwd: 5' CACTCTGGAGGAGAAACGGAAGG 3', rev: 5' GCAGGCATGAGGC AAACAGTC 3'). Experiments were performed in duplicates on a Lightcycler^®2.0 instrument (Roche Diagnostics, Mannheim, Germany) under the following PCR conditions: Initial denaturation at 95°C for 10 min, followed by 40 cycles of denaturation (95°C for 10 s) and a combined annealing and extension phase (60°C for 30 s). At the end of each run, a melting curve (50-95°C with 0.1°C/s) was recorded to ensure the quality of the PCR product. Crossing points (Cp) were calculated by the LightCycler^®Software 4.0 (Roche Diagnostics GmbH) using the absolute quantification fit points' method. Threshold and noise band were set to the same level in all runs used in a comparison. Relative gene expression was determined by the 2^-ΔΔCT method [62] using the real PCR efficiency calculated from an external standard curve. Cp were normalized to the housekeeping genes GAPDH, HPRT1, POLR2B or RPL13A). Fold regulation values were calculated relative to the expression mean of basal mice. The calculation Cp (mean (GAPDH, HPRT1, POLR2B or RPL13A)) - Cp (mean (cand. gene)) was used for each animal.

RT-qPCR experiments were analyzed using non-parametric tests (Mann Whitney using SPSS 12.0).

Identifying genes with similar regulation profiles/Clustering Analysis

To identify genes that show similar expression ratios across time points, i.e. genes that might be co-regulated or affecting each other in a common pathway, we used cluster analysis. Cluster analysis allows the grouping of expression profiles with respect to their relative similarity or, in mathematical terms, a "distance". We consider expression profiles (the expression ratios of the different measurements) to be similar--and thus having a small distance--when they fulfil two criteria: 1. show a high absolute (not mean-corrected) correlation, and 2. have either the same or the opposite regulation (up or down) at all corresponding measurements.

Translated into distance between expression profiles this means that expression profiles, that can be scaled (with either a positive or negative factor, but not shifting the zero line) onto each other, have a small distance. This distance measure groups genes with similar regulation patterns as close neighbours in the cluster analysis. For APP and GNAi2 we show the respective neighbourhoods (targeting at least 10 neighbouring genes, but showing all genes in the most distant branch) as depicted by the corresponding dendrograms.