Long-term ethanol exposure: Temporal pattern of microRNA expression and associated mRNA gene networks in mouse brain

Long-term alcohol use can result in lasting changes in brain function, ultimately leading to alcohol dependence. These functional alterations arise from dysregulation of complex gene networks, and growing evidence implicates microRNAs as key regulators of these networks. We examined time- and brain region-dependent changes in microRNA expression after chronic intermittent ethanol (CIE) exposure in C57BL/6J mice. Animals were sacrificed at 0, 8, and 120h following the last exposure to four weekly cycles of CIE vapor and we measured microRNA expression in prefrontal cortex (PFC), nucleus accumbens (NAC), and amygdala (AMY). The number of detected (395–419) and differentially expressed (DE, 42–47) microRNAs was similar within each brain region. However, the DE microRNAs were distinct among brain regions and across time within each brain region. DE microRNAs were linked with their DE mRNA targets across each brain region. In all brain regions, the greatest number of DE mRNA targets occurred at the 0 or 8h time points and these changes were associated with microRNAs DE at 0 or 8h. Two separate approaches (discrete temporal association and hierarchical clustering) were combined with pathway analysis to further characterize the temporal relationships between DE microRNAs and their 120h DE targets. We focused on targets dysregulated at 120h as this time point represents a state of protracted withdrawal known to promote an increase in subsequent ethanol consumption. Discrete temporal association analysis identified networks with highly connected genes including ERK1/2 (mouse equivalent Mapk3, Mapk1), Bcl2 (in AMY networks) and Srf (in PFC networks). Similarly, the cluster-based analysis identified hub genes that include Bcl2 (in AMY networks) and Srf in PFC networks, demonstrating robust microRNA-mRNA network alterations in response to CIE exposure. In contrast, datasets utilizing targets from 0 and 8h microRNAs identified NF-kB-centered networks (in NAC and PFC), and Smad3-centered networks (in AMY). These results demonstrate that CIE exposure results in dynamic and complex temporal changes in microRNA-mRNA gene network structure.

Introduction Excessive, chronic alcohol use can evoke persistent alterations in brain function that result in alcohol dependence [1,2]. Such alterations involve complex gene networks that span multiple brain regions [3], and there is growing evidence that microRNAs may play an important regulatory role in alcohol's effects on brain function [4][5][6][7]. MicroRNAs are short, non-coding RNAs that utilize sequence complementarity to bind RNA transcripts, thus modifying their expression [8]. A single microRNA can simultaneously alter expression of numerous genes, while multiple microRNAs can act coordinately to precisely control the expression of a single gene [7,9].
The chronic intermittent ethanol vapor (CIE) paradigm is known to increase voluntary ethanol consumption in rodents and is considered to be a model of ethanol dependence [22][23][24]. Recently, Smith and colleagues [25] conducted a time-course experiment examining gene expression in five brain regions following CIE vapor and identified a PFC gene co-expression module enriched with predicted binding sites for several microRNAs that may regulate genes within the module, but microRNA levels were not directly assessed. Previously, we used the CIE model to investigate gene expression changes in three different brain regions (amygdala, AMY; nucleus accumbens, NAC; prefrontal cortex, PFC) at 0, 8 and 120h following exposure to 4 cycles of CIE vapor. These time points represent distinct responses to ethanol exposure, including intoxication (0h), withdrawal (8h) and protracted abstinence (120h). In all brain regions, we identified overlapping time-based gene clusters and gene co-expression modules that shared cell-type specific signatures, perhaps suggesting a common regulatory mechanism. We hypothesized that CIE-induced alterations in micro-RNA expression could be linked with some of the gene expression changes, enabling us to define the role of microRNAs in gene network changes produced by CIE exposure and withdrawal.
We report microRNA expression changes in three brain regions (AMY, NAC and PFC) at three time points following 4 cycles of CIE vapor in mice. For each brain region, differentially expressed (DE) microRNAs were paired with their putative mRNA targets DE in our previous study [26]. Importantly, the mRNA and microRNA expression profiles in the current study were obtained from the same samples used in our earlier study (24). Two separate approaches (discrete temporal association, hierarchical clustering) were combined with pathway analysis to further characterize the temporal relationships between DE microRNAs and their 120h DE targets. Robust microRNA-mRNA network alterations in response to CIE exposure were identified demonstrating that CIE exposure results in complex temporal changes in microRNA-mRNA gene network structure.

Ethics statement
All procedures were approved by the Medical University of South Carolina Institutional Animal Care and Use Committee and adhered to NIH Guidelines. The Medical University of South Carolina animal facility is accredited by the Association for Assessment and Accreditation of Laboratory Animal Care.

Animals and chronic ethanol inhalation procedure
Adult male C57BL/6 mice, purchased from Jackson Laboratories (Bar Harbor, ME), were used in this study. Mice were individually housed under a 12-hr light/dark cycle (lights on at 4:00 AM) in a temperature-and humidity controlled-animal facility. The animals had free access to food (Teklad rodent diet) and water throughout the experiment. The study began after an acclimation period (one week) and mice were monitored daily by the animal facilities staff and the research technicians.
Chronic ethanol inhalation procedures, tissue harvest, and brain dissection methods were as previously described [26][27][28][29]. Briefly, chronic intermittent ethanol vapor exposure (or air) was delivered in Plexiglas inhalation chambers to drug-naïve C57BL/6J (B6) male mice (8 treated and 8 controls per group). Ethanol treatments were performed in the laboratory of Dr. H.C. Becker (Medical University of South Carolina, Charleston, SC, USA). Mice were administered alcohol (1.6 g/kg; i.p.) and the alcohol dehydrogenase inhibitor pyrazole (1 mmol/kg; i.p.) prior to vapor ethanol exposure in inhalation chambers. Control subjects received pyrazole injections in saline and received similar daily handling. Chamber ethanol concentrations were monitored daily to induce stable blood ethanol concentrations within the range of 180-200 mg/dl. Ethanol was administered 16 h/day in 4 weekly cycles and alternated with 1 week in between cycles in which the mice were left undisturbed (mimicking drinking weeks). Animals were sacrificed at 3 time points: 0-, 8-and 120-hours following the last ethanol vapor or air treatment. RNA purification, quantification and quality assessment were performed as in [26]. Briefly, total RNAs were purified using the MagMax 96 for Microarrays (Ambion, Austin, TX) kit using a modified protocol that allowed recovery of small and microRNAs, as well as total RNA. RNAs were quantified on a NanoDrop 1000 spectrophotometer (Thermo Fisher Scientific, Inc., Rockland, IL) and RNA quality was assessed on either the Agilent 2100 Bioanalyzer or 2200 TapeStation (Agilent Technologies, Santa Clara, CA). Recovery of low molecular weight RNAs (i.e., <200 nucleotides) was confirmed by presence of a diffuse band/peak around 100 nucleotides. (These are the same RNAs prepared in [26]; the protocol modification was inadvertently omitted from that publication.)

MicroRNA array analysis
Total RNAs from amygdala (AMY), nucleus accumbens (combined core and shell, NAC) and prefrontal cortex (PFC) were shipped to the Molecular Genomics Core Facility at Moffitt Cancer Center (Tampa, FL). Samples were biotin-labeled using the FlashTag Biotin HSR RNA Labeling Kit (Affymetrix, Santa Clara, CA) and hybridized to GeneChip miRNA 3.0 arrays (Affymetrix, Santa Clara, CA) according to manufacturer instructions. This platform used annotations from miRbase version 17 and contained probe sets for over 19,000 mature micro-RNAs from 153 species. Transcript abundance was measured by fluorescent intensity after scanning with the Affymetrix GCS3000 scanner and generation of cel files with Affymetrix AGCC v3 software. MicroRNA array data have been submitted to the NCBI Gene Expression Omnibus (GEO) (http://www.ncbi.nlm.nih.gov/geo/) under accession number GSE90608. Arrays were hybridized with material from a single animal (no pooling); thus, 144 samples (16 animals x 3 tissues x 3 time points) were profiled.
RMA [31]) and identification of sample outliers using all mouse probesets. Five samples (one AMY 120h control, one PFC 0h control, two NAC 8h controls and one NAC 0h CIE-treated) were identified as outliers using Gene Expression Console and were removed. Data were filtered to include only mature mouse microRNAs with a detection p value < 0.06 on 80% of arrays. Mouse microRNAs were updated to miRbase version 21 annotations using miRbaseTracker [32]. Differential expression analysis for each time point was conducted using empirical Bayes moderated t-statistics from the Bioconductor package limma [33] to compare treated and control mice. MicroRNAs were considered DE at a nominal value 0.05. A nominal, rather than FDR-corrected, p value was utilized to preserve as much DE information as possible for our subsequent systems-level analyses as use of an overly-stringent statistical cutoff would cause us to lose valuable network relationships. Although depleted RNA samples precluded confirmation of our results by RT-qPCR, we have used RT-qPCR to validate this microRNA array platform in a previous publication [7]. In that study, RT-qPCR confirmed microRNA expression changes for six different microRNA families in both mouse and human brain tissue.
Two separate approaches (discrete temporal association, hierarchical clustering) were used to uncover the temporal relationship(s) between microRNAs and their downstream targets in each brain region. Both methods used the "microRNA target filter" utility in Ingenuity Pathway Analysis (IPA, September 2015 release, Qiagen Redwood City, www.qiagen.com/ingenuity) to associate microRNAs with their previously reported DE mRNA targets [26]. It should be noted that microRNAs and mRNAs were isolated from the same samples. Data from mRNA analyses were deposited in the NCBI Gene Expression Omnibus (GEO) under accession number GSE 60676. Only experimentally observed microRNA-target interactions and those predicted with high confidence were used. IPA assigns "high confidence" to interactions involving a conserved or highly conserved microRNA as defined by TargetScan [34][35][36][37] and at least one conserved site on the targeted sequence or having a TargetScan total context score of -0.4 or less.
In the first discrete temporal association analysis, differentially expressed microRNAs from each time point were paired with differentially expressed targets (predicted and validated) at each time point. To maintain consistency with previous analyses, we used linear fold changes and an FDR cutoff of 0.05 for targets DE at 0 or 8h and a nominal p value cutoff of 0.05 for targets DE at 120h. (Previously, these thresholds were chosen in order to maximize the amount of information we could extract from the data. Since the 120h time point did not produce any DE mRNAs after correction for multiple comparisons, implementing an FDR cutoff for all time points would have excluded this entire time point from analysis. Thus, we chose to utilize an FDR threshold where possible and a nominal p value otherwise. Although this is a pitfall in our study, we reasoned that it would allow us to utilize all of our data while maintaining as much rigor in the analysis as possible.) Only microRNA-mRNA associations consistent with changes in microRNA proceeding changes in mRNA were considered. For example, microRNAs DE at 8h were paired with targets from the 8 and 120h time points but not the 0h time point. In all, six paired datasets were created for each brain region (S1 Table). Paired data sets are codified as TimePoint_DEmiR/TimePoint_DEtargets. For example, "0hDEmiR/120hDEtargets" specifies the microRNAs DE at 0h and their targets DE at 120h. We were particularly interested in genes dysregulated at 120h, which represents a state of extended withdrawal, as a result of microRNAs dysregulated at 0h (intoxication) and 8h (withdrawal). Thus, the datasets representing these temporal associations were evaluated with the "Core Analysis" option in IPA using minimum p value to resolve duplicate probes and Illumina MouseRef-8 version 2.0 as the reference set. An alternative approach, hierarchical clustering, used centered and scaled log ratios of DE microRNAs from all time points per brain region with the package "clValid" (version 0.606) [38]. This R package was used to compare multiple algorithms simultaneously to identify the best clustering approach and the optimal number of clusters. Clustered microRNAs were plotted with R package "MmPalateMiRNA" [39] with scaled log-ratios vs time. Clustering was visualized using GraphPad Prism version 6.07 for Windows, GraphPad Software, La Jolla California USA, www. Fig 2A). Three microRNAs were DE in all brain regions (as shown in Fig 2B). Fold changes in bold text indicate significant (p < 0.05) dysregulation at the given time point.

Unique to AMY Unique to NAC Unique to PFC
https://doi.org/10.1371/journal.pone.0190841.t002 Time-dependent changes in microRNAs and associated mRNA networks in response to ethanol vapor graphpad.com. The number of unique, non-overlapping expression patterns indicates the number of assigned clusters for each brain region. For each identified microRNA cluster, IPA was used to construct a network relating the microRNAs with their DE gene targets reported in [26] using the same statistical significance thresholds as described above. First, microRNAs in each cluster and their 120h DE targets were added to a new pathway in IPA, and then related and connected using the Ingenuity Knowledge Base. The "grow" tool was used sequentially to add the ten 0h DE targets, and then the ten 8h DE targets, with the highest connectivity to the network, connecting all molecules after each addition. Default settings were used for all IPA pathway functions.

MicroRNA descriptive statistics
MicroRNAs were profiled in AMY, NAC and PFC at three different time points (0, 8 and 120h) after ethanol vapor treatment. After preprocessing, both the number of detected (395-419) and DE (42-47) microRNA probesets was similar within each brain region. Despite most of the detected microRNAs being expressed in all brain regions, the DE probes were distinct for each brain region (Fig 2). There was little overlap of DE microRNAs across time within each brain region (S1 Fig). MicroRNAs that were uniquely detected and DE in each brain region are shown in Table 1. Only three microRNAs were DE in all three brain regions ( Table 2). MicroRNAs and mRNAs exhibited different temporal patterns of dysregulation ( Fig  3). AMY and NAC showed very similar temporal profiles with the number of DE mRNAs decreasing over time, while DE microRNAs decreased from 0 to 8 hours and then rebounded at 120h. The temporal pattern of expression was unique in PFC because the number of DE microRNAs decreased as a function of time, and the number of DE mRNAs peaked at 8h, then declined at 120h.

Temporal patterns of DE microRNAs and associated gene targets
DE gene targets of the ethanol-responsive microRNAs were identified for each time point. Both experimentally validated and predicted targets (see methods) were utilized. The micro-RNAs changed at the 0h time point had a number of mRNA targets that were changed at the 0, 8 and 120h time points (Fig 4). In general, microRNAs that were changed at 8 or 120h had fewer targets than those changed at 0h, although NAC showed a number of targets for the microRNAs that were changed at 8h. For each brain region, the proportion of shared 0h microRNA targets was greatest (18-22%) for 0 and 8h (proportion calculated as the number of shared targets divided by the total number of targets in the earlier time point x100; data not shown). Most of the targets of microRNAs were dependent on the time point, however, there were some targets of microRNAs from the 0h point that were also changed at 0, 8 and 120h, or at 0 and 120 h (Table 3). There was little overlap between 8h microRNA targets at 8 and 120h (AMY: 4%, NAC: 7%, PFC: 0%).

Functional annotation of paired DE microRNAs with 120h DE targets
To evaluate the role of microRNAs in the persistent (120h) changes in gene expression, paired DE microRNAs and associated 120h DE targets were assessed with IPA's "Core Analysis" to elucidate the predominant biological functions represented in these data. (See Fig 1 and  Time-dependent changes in microRNAs and associated mRNA networks in response to ethanol vapor Methods section "Statistics and Bioinformatics" for details.) Remarkably, hepatotoxicityrelated annotations ("liver cancer", "liver carcinoma", "liver hyperplasia/proliferation") were the top toxicological functions found in several NAC and PFC datasets. In the NAC 8hDE-miR/120hDEtargets dataset, "CNS cell death" showed a strong negative z score (-2.3) while in the PFC 0hDEmiR/120hDEtargets dataset, several cognition-related annotations showed strong positive z scores ("formation of cellular protrusions", 2.2; "growth of neurites", 2.2; "cognition", 2.1; "learning", 2.6; "memory", 2.0). "Organization of actin cytoskeleton" and "formation of lymphocytes" were the most significant annotations in AMY, occurring in the 0hDEmiR/120hDEtargets dataset with z scores of 1.98 and 2.0, respectively.

Hierarchical clustering of DE microRNAs
For each brain region, DE microRNAs were clustered based on expression log ratios. This resulted in 5 distinct temporal clusters each for AMY ( Fig 5) and NAC ( Fig 6) and 6 clusters for PFC (Fig 7). See also S3 Table.

Time point-based network analysis
As part of the "Core Analysis", IPA uses "network eligible molecules" (those molecules in the dataset that interact with other molecules in the IPA knowledgebase) as "seeds" to generate networks with a high degree of connectivity relative to all molecules in the knowledgebase. These networks provide insight to molecule connections and relationships that may not be detected by standard functional annotation methods. Using this approach, we compared the top two networks derived from each dataset containing DE microRNAs and their associated DE 120h targets ( Time-dependent changes in microRNAs and associated mRNA networks in response to ethanol vapor from 0 and 8h microRNAs produced NF-kB-centered networks, while the same datasets in AMY generated networks sharing Smad3. We further employed a process of network merging to uncover the most important relationships between 120h DE genes and their associated microRNAs dysregulated during 0 and 8h time points. For each brain region, the top 3 networks obtained from the "0hDEmiR/ 120hDEtargets" and "8hDEmiR/120hDEtargets" datasets were combined, expanding component gene families or complexes to include all members present in the dataset (S3 Fig). Genes DE at 120h and common to these 2 merged networks are given in Table 5. Using the time point-based microRNA-mRNA relationships, we identified potential dysregulated microRNA regulators of these critical network genes (Fig 8).
Cluster-based network analysis. Networks were created from each cluster of microRNAs and their associated targets DE at any time point, emphasizing those targets dysregulated at 120h (S4 Fig). Notably, genes identified as hubs in the cluster-based networks ( Table 6) were also identified as hubs in the network analysis of paired microRNAs and 120h DE targets. For example, Bcl2 showed high connectivity in AMY networks created from both approaches; similarly, for Srf in PFC networks. Additionally, hub genes from the cluster-based networks were also essential members of the combined networks described above. These included Bcl2 in AMY, Mapt in NAC and Fndc3b in PFC (Tables 5 and 6). In general, cluster-based networks exhibited varying degrees of overlap with critical network genes identified in the time pointbased analysis ( Table 5). The greatest overlap occurred between Table 5 genes and AMY cluster 4 (13/21 genes), NAC cluster 3 (9/14 genes) and PFC cluster 4 (10/13 genes) (S5 Fig). MicroRNAs and mRNAs identified by both cluster-and time point-based approaches are highlighted in Fig 8. Hub genes, affiliated microRNAs and expression data for both (direction and time of change, nominal p value) are provided for each cluster. MicroRNAs preceded by a "+" superscript were not DE in our dataset but were added and connected by the IPA algorithm. Italics denote critical network genes (see Table 5). The last column of microRNA data lists several hub gene-connected microRNAs that were also dysregulated in prefrontal cortex of human alcoholics [10] and/or alcohol-dependent rats [9]. These microRNAs are provided as probeset IDs or microRNA names, depending on their presentation in the reference. The last column of mRNA data reports publications that demonstrate a link between ethanol exposure and expression changes for the given hub gene in each miRNA cluster. References are provided as a hyperlink and are prefixed with the abbreviation for the studied species.

Discussion
Gene expression profiling studies demonstrate clearly that alcohol consumption changes brain region-specific transcriptional profiles in human alcoholics and animal models of voluntary consumption [40]. We recently reported time-dependent changes in mRNA expression in mice exposed to chronic intermittent ethanol (CIE) exposure (a model of alcohol dependence). It is well established that miRNAs can alter the expression of many target genes and a number of studies have shown that miRNA expression is altered in response to alcohol abuse in humans [10][11][12][13] and in animal models [7,9,[14][15][16][17]. Rodent drinking models have been important in identifying alcohol-responsive miRNAs and their functional relevance based on responses to expression manipulation. Alcohol-induced changes in microRNAs are associated with cellular Time-dependent changes in microRNAs and associated mRNA networks in response to ethanol vapor tolerance to alcohol [41], antianxiety effects [42], cellular reward mechanisms [20], regulation of alcohol consumption and preference [16,17,20,21], episodes of binge drinking [7,14,43], dependence/withdrawal [9,15,17] and alcohol-induced conditioned-place preference [21]. However, these studies report expression changes that are occurring at a single point in time after alcohol treatment and thus, do not capture dynamic transcriptional regulation. The temporal relationship between alcohol-responsive miRNAs and mRNAs resulting from alcohol challenge has not been investigated directly. The goal of this study was to combine new micro-RNA data with previously identified mRNA [26] changes in brain regions of mice subjected to CIE vapor, a paradigm known to escalate voluntary ethanol consumption in rodents. Micro-RNA expression changes were measured in three brain regions (AMY, NAC, PFC) at three time points (0-, 8-and 120h) following repeated exposures to ethanol vapor. These three time points represent distinct actions of ethanol as they correspond to intoxication, withdrawal and protracted abstinence, respectively. These very different actions of alcohol would be expected to have brain-region and gene network specific actions on the transcriptome as we found in our earlier study (24). Because we propose that at least some of the transcriptome changes are related to changes in microRNAs, we would also expect that the DE microRNAs would vary considerably between time points and brain regions.
Indeed, DE probesets were distinct for the brain regions (see Fig 1) and within each brain region overlap of DE probesets across time was limited (see S1 Fig). Within each brain region the mRNA response to CIE also produced a fairly unique profile of DE genes at each time point. This emphasizes the unique time-and brain-region specific alterations that occur following CIE vapor. Notably, the temporal patterns of microRNA and mRNA expression patterns differed (see Fig 2). All brain regions exhibited a decrease in DE microRNAs from 0 to 8h and a concomitant reduction of DE mRNAs between 8 and 120h, likely reflecting the time required for the microRNAs to exert their effects on gene targets. These findings are consistent with those identified in previous studies in mouse [7] and human alcoholics [10] which demonstrate over-represented directional changes based on time of alcohol exposure. In contrast to AMY and NAC, the number of DE microRNAs at 120h continued to decline in the PFC, perhaps indicating that microRNA regulation in the PFC is under greater homeostatic regulatory control given the complexity of the transcriptome networks responding to alcohol challenge [7,44]. In PFC, several DE microRNAs are members of the same family, especially at the 0h time point. (S2 Table). The greater redundancy of DE microRNA family members in PFC compared with AMY and NAC could suggest a higher degree of transcriptional fine tuning or regulatory prioritization in this brain region. In addition, temporal patterns of DE microRNA gene targets also varied significantly among brain regions. AMY and PFC microRNAs were most responsive to the direct effects of ethanol (0h), whereas NAC displayed a greater response to ethanol withdrawal (8h). This is supported by the finding that the greatest number of DE gene targets in AMY and PFC were associated with microRNAs DE at 0h while the greatest number of DE gene targets in NAC were associated with microRNAs DE at 8h (Fig 4).
We reasoned that integrating differential expression profiles from both microRNAs and mRNAs would provide greater insight into the perturbed gene networks associated with withdrawal/protracted withdrawal. Our microRNA-target association procedure allowed for both perfect and imperfect sequence matching between microRNA-mRNA pairs. This undoubtedly influenced our results since a microRNA's function is determined in part by the extent of Time-dependent changes in microRNAs and associated mRNA networks in response to ethanol vapor sequence complementarity with its target. In general, a perfect sequence match between microRNA and target is believed to cause mRNA degradation while an imperfect match is believed to result in translational repression [45]. Thus, utilizing only perfect-match micro-RNA-mRNA pairs would have omitted non-degraded mRNA targets from our analysis, potentially resulting in more precise associations. However, the relationship between microRNA-mRNA sequence complementarity and microRNA function is not absolute; and limiting our analysis exclusively to perfectly matched pairs would likely result in a significant loss of information. Thus, we opted to include both perfect and imperfect sequence matches, relying on dual analyses to ensure reliable results. Two different strategies were used to reveal relationships between DE microRNAs and their targets: discrete time-based analysis and a hierarchical cluster-based analysis.
Interestingly, divergent analytical approaches identified a number of common genes that were highly connected within networks (connectivity and hubs). These analyses have identified a number of neuroimmune-and cell death/survival-related pathways including ERK1/2 (mouse equivalent Mapk3, Mapk1; time-point based paring only), Bcl2 (in amygdala networks) and Srf (in prefrontal cortex networks). These identified hub genes demonstrated robust microRNA-mRNA network alterations in response to alcohol exposure. In addition, temporal analyses identified NF-kB-and Smad3-centered networks in NAC and PFC. Mitogen-activated protein kinases (Mapk3 and Mapk1) are critical components of signal transduction pathways that are involved in cell growth, adhesion, survival and differentiation [46] as well as myelination [47]. Rit1, a microRNA target dysregulated at multiple time points in each brain region, was also identified in the current study (see Table 3). This gene is dependent upon MAP kinase signaling pathways which are involved in neuritogenesis [48]. These biological functions are associated with regulation of transcription, translation, and cytoskeletal rearrangements [49]. In addition, these pathways are activated by a variety of signals including cytokines and  Time-dependent changes in microRNAs and associated mRNA networks in response to ethanol vapor heterotrimeric G protein coupled receptors [50]. Both Mapk3 and Mapk1 are related to alcohol actions in mouse [51,52] and human [53] brain. They have also been implicated in addiction to other drugs of abuse, such as cocaine [54]. Bcl2 (B-cell lymphoma 2) family proteins regulate cell death by either pro-or anti-apoptotic mechanisms and Bcl2 is considered a "prosurvival" protein that is regulated by the Erk1/2 signaling pathway [55]. It was reported recently that ethanol exposure results in pre-mRNA mis-splicing of myeloid cell leukemia-1 (Mcl-1) which is an anti-apoptotic member of the Bcl-2 family of proteins, suggesting a potential role for Bcl2 in ethanol-mediated neuronal cell death [56]. The regulation of Bcl-related family members is clearly important to understand in the alcohol field given that alcohol in a variety of settings is known to be involved in cell death [57]. In addition, Bcl2 is upregulated in alcohol preferring mice [51] and protects against ethanol neurotoxicity in neonatal mouse cerebellum [58]. NF-kB-and Smad3 networks were identified as highly connected in NAC. NF-kB is a ubiquitously expressed transcription factor family that controls the transcription of hundreds of genes that are involved in many processes, including inflammation, immunity, cell proliferation, and cell death [59,60] and is known to be an important regulator of neuroinflammation [61]. A number of studies have identified genes that are related to inflammatory/immune responses and mediate their effects through NF-kB [40]. Smad3 (SMAD family member 3) mediates signaling from transforming growth factor beta (TGF-β) that is a regulator of cell proliferation, differentiation and death [62]. Smad3 knockout mice have impaired immune function, suggesting SMAD signaling is involved in regulating the immune response [63,64]. Time-dependent changes in microRNAs and associated mRNA networks in response to ethanol vapor Recently, alcohol-induced microglial changes were found to be over-represented in genes attributed to TGF-β/Smad3 receptor signaling and inflammatory response [65]. Given that TGF β is a cytokine expressed in brain that is capable of controlling microglial activation, this signaling pathway may have the potential to regulate alcohol consumption. MicroRNAs that targeted these hubs included miR-34a-5p, miR-17-5p, miR-181a-5p, miR-16-5p for Bcl2 in the AMY (except for miR-16-5p, these microRNAs are involved with immune response [66]), and miR-335-5p for Bcl2 in the PFC (identified in the time-point and cluster based pairings). In Time-dependent changes in microRNAs and associated mRNA networks in response to ethanol vapor NAC, miR-34a-5p and miR-146a-5p are potentially important regulators of Erk1/2 and NF-kB signaling pathways (the time-point based pairings only), respectively. These results emphasize the importance of investigating the complex temporal relationships that exist between miRNA and gene expression changes in response to alcohol challenge. Time-dependent changes in microRNAs and associated mRNA networks in response to ethanol vapor Many studies have focused on "snapshots" in time to establish the relationship between expression changes of miRNA and mRNA; however, this temporal relationship is complex and difficult to study. The current studies are unique because they address critical questions about the temporal relationship between miRNA regulation and mRNA function. These results demonstrate that alcohol exposure results in complex temporal changes in microRNA-mRNA gene network structure and that manipulation of microRNAs may rescue the aberrant synaptic plasticity associated with alcohol consumption.