ABSTRACT: Microarray profiling was used to investigate gene expression in the hypoxic seizure model of acquired epilepsy in the rat, with the aim of characterizing functional pathways which are persistently activated or repressed during epileptogenesis. Hippocampal and cortical tissues were transcriptionally profiled over a one week period following an initial series of seizures induced by mild hypoxia at post-natal day 10 (P10), and the gene expression data was then analyzed with a focus on gene set enrichment analysis, an approach which emphasizes regulation of entire pathways rather than of individual genes. Animals were subjected to one of three conditions, control condition with no exposure to hypoxia, hypoxic seizures, and hypoxic seizures followed by treatment with the AMPAR antagonist NBQX, a compound tentatively proposed to be a modulator of epileptogenesis. Gene expression in the control samples was found to be consistent with known processes of neuronal maturation in the rat for the given time window. The hypoxic seizure response was found to be saliently enriched for components of the PI3K/mTOR and Wnt signaling pathways, alongside gene sets representative of glutamatergic, synaptic and axonal processes, perhaps regulated as a downstream consequence of activation of these pathways. Wnt signaling components were also found enriched in the more specifically epileptogenic NBQX-responsive gene set. While activation of the mTOR pathway is consistent with its known role in epileptogenesis and strengthens the case for mTOR or PI3K pathway inhibitors as potential anti-epileptogenic drugs, investigation of the role of Wnt signaling and the effect of cognate inhibitors offers a parallel avenue of research toward anti-epileptogenic treatment. Cohorts of ten days old (P10) rat pups were subjected to one of three experimental conditions: 1) a control sham treatment inducing no hypoxic seizures (noHS), 2) a treatment under hypoxia inducing hypoxic seizures (HS), in which each animal underwent a 15 minutes passage at 5% O2, and 3) a treatment with hypoxia inducing hypoxic seizures, followed by four intra-peritoneal injections of NBQX over a 48 h period, starting at 30 minutes post-hypoxia (HS+NBQX). Control vehicle injections of DMSO were performed for the animals under the noHS and HS conditions. Time of occurrence and number of seizures for the animals undergoing hypoxia were recorded and analyzed. Times to the first seizure were found to be exponentially distributed, with mean time to first seizure of 114 seconds, and total count of seizures occurring in the 15 minute hypoxic interval was found to be approximately normally distributed, with mean count 10.6 and standard deviation 2.9. For each treatment condition, cohorts of 4 to 6 rat pups were killed at five sampling times spanning 1 hour to 1 week post-hypoxia (1h, 6h, 12h, 48h, 168h post-hypoxia), for a total of 75 animals sacrificed. Whole cortex and hippocampi were extracted from the sacrificed animals and separately profiled on two Affymetrix high-throughput plate microarrays after RNA quality control, resulting in a total of 144 CEL files encapsulating the gene expression data. The 144 CEL files, generated from scans of the individual microarrays on the two 96-well Affymetrix HT Rat Focus Arrays, were processed according to the MAS5 algorithm, resulting for each input CEL file in output of 24,249 probeset-level intensities and corresponding absent/present calls. For each processed scan, two summary statistics, consisting of the total number of present calls on the chip, and the 75th percentile of present call intensities, were then computed. A quality control step was performed by inspecting for outlier scans in the space of these two statistics (no outliers were found). Additional visual inspection of a heat map of clustered profiles identified a potential tissue assignment swap between two samples, which were conservatively eliminated from further analysis. The resulting 142 quality-controlled profiles were then assembled into a single data matrix, and normalized to each other by linear rescaling of all probeset intensities so as to bring the 75th percentile of present call intensities on all chips to a single common value. These processing steps, and most of the subsequent analyses discussed here, were done using the Gecko gene expression analysis platform (Theilhaber et al., 2004). Additional analyses were performed using the R statistical tools framework (R Development Core Team 2008).
