Time-Course Analysis of Brain Regional Expression Network Responses to Chronic Intermittent Ethanol and Withdrawal: Implications for Mechanisms Underlying Excessive Ethanol Consumption

Abstract

Long lasting abusive consumption, dependence, and withdrawal are characteristic features of alcohol use disorders (AUD). Mechanistically, persistent changes in gene expression are hypothesized to contribute to brain adaptations leading to ethanol toxicity and AUD. We employed repeated chronic intermittent ethanol (CIE) exposure by vapor chamber as a mouse model to simulate the cycles of ethanol exposure and withdrawal commonly seen with AUD. This model has been shown to induce progressive ethanol consumption in rodents. Brain CIE-responsive expression networks were identified by microarray analysis across five regions of the mesolimbic dopamine system and extended amygdala with tissue harvested from 0-hours to 7-days following CIE. Weighted Gene Correlated Network Analysis (WGCNA) was used to identify gene networks over-represented for CIE-induced temporal expression changes across brain regions. Differential gene expression analysis showed that long-lasting gene regulation occurred 7-days after the final cycle of ethanol exposure only in prefrontal cortex (PFC) and hippocampus. Across all brain regions, however, ethanol-responsive expression changes occurred mainly within the first 8-hours after removal from ethanol. Bioinformatics analysis showed that neuroinflammatory responses were seen across multiple brain regions at early time-points, whereas co-expression modules related to neuroplasticity, chromatin remodeling, and neurodevelopment were seen at later time-points and in specific brain regions (PFC or HPC). In PFC a module containing Bdnf was identified as highly CIE responsive in a biphasic manner, with peak changes at 0 hours and 5 days following CIE, suggesting a possible role in mechanisms underlying long-term molecular and behavioral response to CIE. Bioinformatics analysis of this network and several other modules identified Let-7 family microRNAs as potential regulators of gene expression changes induced by CIE. Our results suggest a complex temporal and regional pattern of widespread gene network responses involving neuroinflammatory and neuroplasticity related genes as contributing to physiological and behavioral responses to chronic ethanol.

Fig 1. Schematic representation of experimental design and analysis.

Fig 2. Overlap between CIE- regulated probesets at each time-point across brain regions.

Table documents number of probesets significantly regulated by CIE (FDR ≤ 0.01) at each time point within individual brain regions (shaded cells) and overlap with same timepoints across other brain regions.

Fig 3. Overlap between CIE-regulated probesets and modules identified by WGCNA.

Cell numbers indicate number of overlapping probe-sets, Cell color indicates significant overlap. Significant overlap: p-value ≤ 0.005 and Odds Ratio ≥ 3. Names and number of genes for each module are listed at far left columns within each brain region.

Fig 4. Representative kinetic profiles for module eigengenes.

Module eigengene expression vs. time plots for the black, greenyellow, blue, and magenta modules in PFC, pink module in NAC, salmon module in CEA, darkturquoise module in BNST, and turquoise module in HPC. (detailed discussion of module functions in Results section). Triangle = CIE, Circle = Ctrl.

Fig 5. Bioinformatic analysis of PFC green module containing Bdnf.

A) miRvestigator results of top miRNA motifs with complementary binding sequences in the PFC Green module. B) Network representation of the PFC Green module based on adjacency. Edge transparency indicates Pearson correlation coefficient. Node size reflects within-module connectivity determined by WGCNA. Node color indicates log-ratio of gene expression at 5 days CIE vs. Ctrl. Genes with mmu-let-7c-1 complementary sequences are highlighted. C) Average RMA value (log2 scaled, ±S.E.) expression of Bdnf at each time-point and treatment condition in the prefrontal cortex. (* = LIMMA FDR ≤ 0.05).

Fig 6. Network level analysis of PFC green module.

A) Disruption of co-expression with CIE in genes regulated at 5 days (LIMMA FDR ≤ 0.05). Node size = within module connectivity. Ordered by within module connectivity at 5 days in Ctrl mice. B) Histograms for FDR of genes in PFC Green module at each time point. Dark grey = overlap of genes regulated at 5 days (LIMMA FDR ≤ 0.05). C) Eigengene expression time course for green module genes in control or ethanol (CIE) treated animals.

Fig 7. GeneMANIA analysis of genes from HPC turquoise module related to chromatin modification.

Chromatin modification genes were identified from Gene Ontology analysis of the HPC turquoise module (S10 Table) and submitted to the GeneMANIA resource (www.genemania.org) for identification of network interactions using default criteria and databases.

Fig 8. Expression patterns for representative candidate genes.

A) Average RMA value (log2 scaled) expression of candidate genes at each time-point and treatment condition in the PFC cortex. (* = LIMMA FDR ≤ 0.05, ** = LIMMA FDR ≤ 0.01) B) Average RMA value (log2 scaled) expression of candidate genes at each time-point and treatment condition in the HPC. (* = LIMMA FDR ≤ 0.05, ** = LIMMA FDR ≤ 0.01).