Ethics Statement

All procedures were approved by Virginia Commonwealth University Institutional Animal Care and Use Committee under protocol number AM10332 and followed the NIH Guide for the Care and Use of Laboratory Animals (NIH Publications No. 80–23, 1996).

Animal microdissection and acute ethanol administration

Animals were treated according to protocols for animal care established by Virginia Commonwealth University and the National Institute for Health. Adult male B6129SF2/J and B6;129S7-Fyn^tm1Sor/J mice obtained from Jackson Laboratories at 12 weeks of age and were housed 4-5 per cage with ad libitum access to water and standard rodent chow (#7912, Harlan Teklad, Madison, WI) on a 12 hr light/dark cycle with Harlan Sani-chips bedding (#7090A). Mice were habituated to the animal facility for 1 week prior to initiating experiments.

Control and Fyn-null mice (n=18 of each genotype) were administered intraperitoneal injections of saline for 3 days to habituate them to the injection process; on day 4 mice received either an injection of saline (n=9 of each genotype) or 3 g/kg (20% v/v) of ethanol (n=9 of each genotype), a sedative-hypnotic dose [23]. Animals were sacrificed by cervical dislocation and decapitation at a 4-hour time-point. Our laboratory has previously found that a 4-hour time point captures a spectrum of early, intermediate, and late gene expression responses to ethanol (Ravindranathan and Miles, unpublished). Microdissection of individual brain regions was conducted exactly as described previously [17]. Brain regions were individually frozen immediately with liquid nitrogen, and subsequently stored at -80°C until isolation of total RNA.

Tissue pooled from three mice of the same genotype/treatment group was homogenized in PureZol Reagent (Bio-Rad Laboratories, Hercules, CA) using a Tekmar homogenizer, and total RNA was isolated with the Aurum Total RNA fatty and Fibrous Tissue Kit according to the manufacturer’s instructions. RNA concentration was determined by absorbance at 260 nm, and RNA quality was analyzed by 260:280 nm absorbance ratios and electrophoretic analysis (Experion; Bio-Rad Laboratories, Hercules, CA). Double-stranded cDNA and biotin-labeled cRNA was synthesized using reagents and protocols from the microarray manufacturer (Affymetrix, Santa Clara, CA).

Microarray Hybridization and Scanning

Biological replicates (n=3) from pooled samples within each treatment group and genotype were hybridized to individual microarrays for prefrontal cortex (PFC), nucleus accumbens (NAc), and ventral midbrain (VMB) areas (n=36 total microarrays). Arrays for a single brain region were processed together in one day, using a supervised randomization of samples in order to minimize potential batch effects. Labeled cRNA samples were analyzed on oligonucleotide arrays (Affymetrix Mouse Genome 430 2.0 arrays) that contain ~36,000 genes and expressed sequence tags. Hybridization, washing, staining and scanning were performed according to manufacturer protocols (Affymetrix).

Microarray Data Analysis

Microarray data were initially processed using GeneChip Operating Software v4.1 (GCOS, Affymetrix). Arrays were normalized to a median total hybridization intensity (target average intensity, 190) and quality was assessed by array scaling factor (<3), 3’-5’ expression ratios for control probes, absent/present calls (%present ≥ 55% for all arrays) and by inspection of pairwise Pearson correlations and scattergrams. Arrays passing quality control were processed for any potential hybridization batch effects [24]. Processed arrays were initially filtered for Absent/Present calls [25] to eliminate probesets consistently called “absent” across all samples, and then subjected to S-score analysis [26]. The S-score algorithm, developed in our laboratory for analyzing Affymetrix arrays was applied to compare hybridization signals between two arrays from differing treatment samples. An S-score of |2| corresponds to a p = 0.0455, uncorrected for biological variability or multiple comparisons. S-scores were generated using all of the pairwise comparisons between ethanol treated controls and Fyn knockout mice, as well as basal differences between genotypes. The average S-scores from all individual subject pairwise comparisons were used to represent a single biological replicate in downstream analyses. Within each brain region and treatment, S-scores were divided by the greater of 1 or the group standard deviation from same/same comparisons to reduce the contribution of biological or technical noise, as described previously [17]. Array data has been deposited in the Gene Expression Omnibus database (www.ncbi.nlm.nih.gov/geo/) with accession number GSE 9028.

A one-class statistical analysis of microarrays (SAM) was used to within each treatment group/brain region to determine those genes with S-scores significantly different from 0 [27]. Differences in ethanol-regulation of gene expression were determined using 2-class SAM within each brain region (i.e. Fyn KO vs CTL). Within each SAM analysis S-scores were filtered for an average ≥ 1.5 or ≤ -1.5 (composite significance, p < 0.01) to focus on the most biologically robust results. All SAM analyses were filtered for a median false-discovery rate (FDR) ≤ 5%. Significant genes/probesets were subsequently subjected to k-means clustering [28] to determine coordinately-regulated gene expression patterns, and demonstrate brain-region, treatment, and genotype specific differences in gene expression.

Bioinformatics Analysis Of Microarray Data

Toppgene Suite [29] and ErmineJ [30] were used for data exploration of functional classification among gene expression profiles using gene ontology categories, mouse phenotype data, and public pathway databases. ToppFun (functional enrichment analysis within Toppgene Suite) calculations were set to a 5% false discovery rate, with gene limits of n ≥ 3 and ≤ 300 to identify representative a priori ontological categories. ErmineJ gene set analysis was implemented for over-representation analysis (ORA) using the best scoring gene replicate and a gene score threshold of 1.5; reported p-values were corrected for the ErmineJ default of a 10% FDR. For ErmineJ analysis a custom extended myelin-associated gene set was constructed as an additional GO category based upon repeated microarray data cluster associations for myelin genes from our laboratory (Farris and Miles, unpublished data and see 17). Additional pathway analysis of statistically significant genes was done using Ingenuity Pathway Analysis (www.ingenuity.com), a curated bioinformatics resource for discovery of gene-gene interaction networks based on literature association, biological function, and cell-signaling mechanisms.

Fyn-centric correlation networks were constructed using the intersection of Fyn knockout data herein and Fyn expression correlations (Pearson correlation p-value ≤ 0.01) across microrarray datasets previously generated by our laboratory from PFC of the BXD [18] and LXS (Miles and Johnson, unpublished data) recombinant inbred mouse lines (datasets GN135 and GN130, respectively, publically available at www.genenetwork.org). Due to the redundant design of Affymetrix microarrays with multiple probesets representative of a single gene, all probesets representing a single gene were considered for analysis with multiple probesets represented as a single node with the removal of duplicate edges (i.e. correlations) between two genes and self-loops among the same gene. Visualization of gene correlation networks were rendered in Cytoscape (www.cytoscape.org) [31]. Resulting Fyn-centric correlation networks were submitted to GeneMANIA (genemania.org) [32] to independently assess protein and genetic interactions, pathways, co-expression, co-localization and protein domain similarity as well as identify potential candidates not directly identified from microarray analysis.

Gene Expression Pattern Changes In Fyn Null Mice

Genome-wide expression profiling across the mesolimbocortical dopamine pathway of saline or ethanol-treated male Fyn null mice was conducted in order to characterize the molecular mechanisms underlying altered behavioral responses to ethanol in this mutant mouse line. We performed whole-genome expression profiling in the presence and absence of an acute anesthetic dose of ethanol (3 g/kg) since ethanol LORR is the most reproducible ethanol phenotype altered in Fyn null animals. Direct or indirect changes in gene expression resulting from elimination of Fyn were determined through one-class and two-class SAM analysis of S-scores for either basal or ethanol-responsive gene expression. Multivariate analysis using k-means clustering identified region specific patterns of gene expression (Figure 1A; Table S1); basal and ethanol-treated gene expression across each of the three brain regions was combined to increase the statistical power of k-means clustering. As expected from the overall expression distributions (Figure 1B & 1C), most clusters showed basal (KO/CTL) differences that were unique to a single or two brain regions. Several clusters showed possible trends toward differences in ethanol responses, as well as basal expression changes with the Fyn KO, particularly in the VMB (see clusters 2, 3, 4, 6, 9, 11 and 12 for VMB). Generally, differences in basal and ethanol-responsive gene expression were distributed unevenly among the three brain regions with basal alterations ranging NAC > VMB > PFC (Figure 1B) and ethanol- responses varying in the order of VMB > NAC > PFC (Figure 1C).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Basal and ethanol-responsive gene expression differences between Fyn knockout mice and controls.

(A) k-Means cluster analysis of differential gene expression for S-scores; green = decreased relative expression, red = increased relative expression, black = no change in expression (Control EtOH = CTL EtOH, KO EtOH = Fyn Knockout EtOH, KO / CTL = Fyn Knockout Saline / Control Saline). (B) Venn diagram of overlapping and non-overlapping genes differentially expressed between saline treated Fyn knockout mice and controls. (C) Venn diagram of overlapping and non-overlapping genes differentially expressed between acute ethanol (3 g/kg) treated Fyn knockout mice and controls. (NAC - Nucleus Accumbens, PFC - prefrontal cortex, and VMB - ventral midbrain).

https://doi.org/10.1371/journal.pone.0082435.g001

Bioinformatic Analysis Of Basal Expression Changes In Fyn Null Mice

In order to assess the potential biological significance of basal expression differences due to a null mutation for Fyn we performed a functional over-representation analysis for each individual brain-region using the ToppGene Suite web-portal (FDR ≤ 0.05, Table 1 Basal Gene Ontology Categories; full-set available as Table S2). Evaluation of functional changes may be important for determining the role of Fyn within different brain-regions and subsequent phenotypic impact of a Fyn null mutation. Although some categories are certainly redundant, the overall number of differentially expressed genes for each brain-region was disproportionate to the number of significant functional groups. PFC had the smallest change in total number of genes among the three brain structures; however, the PFC also showed the largest number of significant functional categories.

Myelin-Associated Gene Expression

As suggested by previously published research with a less extensive expression analysis [39,40], our results detected a significant decrease of myelin-associated gene expression in Fyn knockout mice. The 76 genes with altered basal expression overlapping across all three brain regions (Figure 1B) were over-represented for structural constituent of the myelin sheath (GO: 0019911; P-value = 0.000105 uncorrected for multiple comparisons). This ontological category, as defined by the Gene Ontology Consortium (http://www.geneontology.org/) is limited by the inclusion of only 5 myelin genes (Mal, Mbp, Plp1, Tspan2, Mobp), two of which (Mbp and Plp1) were decreased in Fyn null animals on our microarray analysis. Therefore, we extended this study to include other myelin-associated genes by using ErmineJ [30] for over-representation analysis. We tested the over-representation of a dozen myelin-associated genes (see Methods) based on their absolute expression within each individual brain region (Figure 2A, FDR ≤ 10%). As also shown in Figure 2C (and see Table 1 and Table S2), basal expression for these myelin genes were decreased in Fyn null animals within the NAC and PFC, but not the VMB, were significantly over-represented for myelin-related genes. Ingenuity Pathway Analysis of the NAC dataset (Figure 2B) further showed the functional relationship between Fyn kinase and myelin, as well as other potentially related genes such as myelin basic protein expression factor 2 repressor (Myef2). Our prior genomic studies also identified differential basal expression of myelin genes between C57BL/6J and DBA/2J mice [17], which show divergent behavioral responses to acute ethanol. Together these results suggest that a hypomyelination phenotype for Fyn knockout animals may be important for the interpretation of ethanol behavioral phenotypes in this model system. More generally, these data may suggest that coordinate differential expression of a myelin-related gene cluster and associated signaling mechanisms could contribute to acute ethanol behavioral responses.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Over Representation Analysis (ORA) of Myelin-Associated Gene Expression.

(A) ErmineJ heatmap of NAC gene expression showing a coherent decrease for myelin-related basal gene expression in Fyn knockouts compared to control. (B) Ingenuity Pathway Analysis (IPA) of myelin-related genes and literature associations; numbers shown are mean S-score of three biological replicates. Red indicates increased expression in Fyn knockout vs controls and green fill indicates decreased expression in the Fyn knockout. (C) Table display of ErmineJ corrected p-values (10% FDR) for myelin gene expression in NAC, PFC, and VMB.

https://doi.org/10.1371/journal.pone.0082435.g002

Ethanol-Responsive Gene Expression

In addition to basal differences in gene expression, differences in ethanol-evoked signaling events may contribute to altered ethanol behavioral phenotypes in Fyn kinase knockout animals. Using gene expression as a surrogate measure of signaling mechanisms altered by acute ethanol we performed microarray analysis of NAC, PFC, and VMB from knockout and control animals 4 hours post an acute ethanol exposure (Figure 1A). A 4-hour time-point was chosen due to prior studies from our laboratory showing that this time-point captured a spectrum of early, intermediate, and late gene expression responses to ethanol ([17] and data not shown). We chose a 3 g/kg (i.p.) dose of acute ethanol that has been previously used in regards to behavioral genetics of ethanol’s sedative-hypnotic effects [23].

Global changes in gene expression occurring similarly across all individual brain regions due to acute ethanol exposure were captured using a one-class SAM analysis and included in k-means clustering for visualization (Fig 1A, e.g. cluster 3). Differences in ethanol-responsive expression were determined using a two-class SAM analysis within each brain structure (Figure 1A & 1C). Unlike our previous studies between B6 and D2 mice, although there were a significant number of differences in ethanol responsive genes between controls and Fyn null animals in each brain region (Figure 1C), region-specific ethanol patterns were difficult to discern visually using k-means clustering. This was likely due to large number of changes in expression from the higher dose of acute ethanol (i.e. one-class SAM), resulting in more quantitative rather than qualitative differences in ethanol responses between the control and null animals, as well as the prominent differences in overall basal abundance due to loss of Fyn (Figure 1B). Subsequently we further analyzed these resulting acute ethanol expression differences for separate brain areas using a multivariate bioinformatics approach described in Materials and Methods. Unlike over-representation analysis of basal differences (Table 1), gene ontology analysis of ethanol-responsive expression at a 5% FDR yielded few significant groups related to gene function (not shown; see Table S3 for total gene ontology results). This suggested that basal variation of system-wide changes in expression (i.e. glutamate receptor function and myelin-associated gene expression) is a stronger predictor of acute ethanol sensitivity in the Fyn kinase knockout mouse. Alternatively, it could be that our use of a 5 % FDR is overly conservative in determining ontological groups related to gene function following a transient event such as with a single ethanol exposure on this specific genetic background. However, literature association analysis using curated resources from Ingenuity Pathway Analysis (IPA) and GeneGo did suggest network-level differences in ethanol-regulated gene expression between knockout and control animals (Figure 3 and Figures S1A and S1B) as discussed below.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. Fyn Knockout Animals Have Altered Ethanol-Responsive Gene Networks.

Ingenuity Pathway Analysis from two-class SAM filtering for network-level differences in ethanol-responsive gene expression for the Nucleus Accumbens in control (left panel) versus Fyn knockout (right panel) mice. Genes labeled in green are down-regulated by acute ethanol; Red labeled genes are up-regulated by acute ethanol. Numbers shown are the mean S-score for 3 biological replicates. Data for additional brain regions are shown in Figure S1.

https://doi.org/10.1371/journal.pone.0082435.g003

Fyn-Related Gene Network Analysis For LORR

Null mutations of Fyn increase duration of the loss of righting reflex (LORR) response to acute ethanol [3-5]. In order to further assess molecular mechanisms for Fyn involvement in ethanol LORR, we compared our microarray analysis of Fyn null mice with gene expression correlates of LORR or Fyn expression across panels of recombinant inbred mice (Figure 4A). PFC microarray data was used for this analysis due to its availability and since the PFC dissections include adjacent motor cortex [17] and we have shown that selective modulation of PFC gene expression can alter ethanol LORR [19]. As described in Methods, Fyn gene expression correlates (Pearson correlation p-value < 0.01) were derived in PFC across the BXD (n = 29) [18] and LXS (n = 42) recombinant inbred mouse lines. Additionally, we identified BXD PFC gene expression correlating with a previously published LORR study across BXD RI mice (GeneNetwork ID: 10589 [51]). Overlapping data from these three microarray correlation gene sets together with the basal expression differences identified here in the Fyn null mice (Figure 1, Table S1) allowed definition of a Fyn-LORR gene network (Figure 4A; Figure S2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 4. Fyn-LORR Correlation Network within PFC.

Network analysis of concurrent variation of gene expression amongst PFC Fyn knockout array data, Fyn expression correlates across the BXD RI PFC, Fyn expression correlates across the LXS RI PFC, and LORR – BXD gene expression correlates in PFC. (A) Diagram of general approach used to define a Fyn-dependent gene network for the loss of righting reflex behavior. (B) Basal Fyn-dependent gene network showing correlations among BXD RI RMA saline dataset (node connections indicate correlation p-value ≤ 0.01). (C) Ethanol-responsive Fyn-dependent gene network showing correlations among BXD RI S-score dataset (p-value ≤ 0.01). Red circle indicates position of Fyn.

https://doi.org/10.1371/journal.pone.0082435.g004

Analysis of basal gene expression from the Fyn null mice by the scheme mentioned above identified a network of 50 highly inter-correlated genes (Figure 4B, Table S4); the first principal component of these 50 genes significantly correlated to LORR (r = -0.83, p-value = 5.05 E-06; Figure 5, left lower panel). Functional over-representation analysis of this gene network highlighted ion channel activity, function/localization to the post-synaptic density and dendritic spines, and regulation of action potentials (Table S5) as over-represented groups. Neto1, Kcnd2, Dnm3, Gria3, and Homer1 are all localized to the post-synaptic density and also may play a role in modulating synaptic N-methyl-D-aspartic (NMDA) acid receptor function. Neto1 knockout mice have abnormal long-term potentiation, learning, and memory due to altered NMDAR subtype abundance [52]. Kcnd2, also known as Kv4.2, is a voltage-gated potassium channel that regulates spontaneous NMDAR activation and downstream calcium signaling mechanisms [53]. Dnm3 is a binding partner of Homer1 [54] a scaffolding protein associated with type1 metabotropic glutamate receptors and the NMDAR complex. The ionotropic glutamate receptor Gria3 is a member of the AMPA-receptor family, which facilitate fast excitatory synaptic transmission. Thus, although the literature suggests a dominant role of NR2B NMDA receptor subunit function in Fyn modulation of ethanol behaviors, our genomic analysis results suggest that the Fyn-dependent behavioral phenotypes may also be modulated through several genes related to glutamate function and possibly additional mechanisms as well (see discussion below on myelin).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 5. In silico correlation analysis of Fyn, Fyn gene networks and LORR behavior.

Pearson product moment correlations analysis of basal saline (RMA) gene expression (right panels) and ethanol-responsive (S-score) expression (left panels) versus ethanol LORR behavior (Rodriguez et al., 1995; WebQTL Record ID: 10589). Expression for Fyn (upper panels) or Fyn-dependent networks (lower panels) from Figure 4B and 4C are plotted versus LORR. For Fyn-dependent networks, the first principal component (PC1) for expression of all genes in the network was used for correlations with LORR. PC1 values are relative and do not indicate overall positive or negative correlations with LORR.

https://doi.org/10.1371/journal.pone.0082435.g005

In addition to the glutamate-related ion channels and support proteins mentioned above, Kcna2, Gabrb3, Cacna2d1, Scn1a were present in Figure 4B, suggesting Fyn kinase signaling is interconnected with regulation of gene expression for multiple ion channels. The beta-3 subunit of GABA-A receptor has been previously reported to be functionally altered in Fyn kinase knockout mice, contributing to reduced LORR sensitivity with etomidate, a beta-2/beta-3 selective compound [55]. Scn1a is expressed in the nodes of Ranvier of motor neurons, regulating the propagation of action potentials [56], and resides on Chr 2 within a QTL for alcohol preference [57], although no actual link to ethanol behaviors has been shown for this gene.

The Fyn-dependent basal gene network was also significantly over-represented for ‘abnormal white matter morphology’ (p-value = 0.000003; MP:0008026), and as shown in Figure 2, white matter related genes were significantly decreased in the PFC of Fyn knockout mice. Fyn expression was also correlated to Quaking (Qk) (Figure 2B), which is known to regulate the transport of Mbp mRNA from the nucleus [58]. Fyn specifically regulates the activity of Quaking through phosphorylation of tyrosine residues for binding Mbp mRNA within the C-terminal domain of Quaking [59]. Thus, a Fyn-dependent, myelin-related gene expression network altered in Figure 2B could contribute to Fyn modulation of ethanol LORR.

In addition to differences in basal gene expression that may contribute to variation in acute ethanol sensitivity of Fyn null mice, the LORR behavior may also be influenced by differences in acute ethanol-evoked signaling events. Again using S-score analysis as a measurement of ethanol-responsive gene expression (i.e. EtOH/Saline), we used the experimental design outlined in Figure 4A, with ethanol-responsive gene expression values from the various datasets for correlation to Fyn or LORR, to identify a set of 32 distinct genes across the four different datasets (Figure 4C, Table S6). The first principal component of the ethanol-responsive network was significantly correlated to the LORR (r = 0.87, p-value = 2.30 e-07), greater in magnitude than for Fyn kinase ethanol-response alone (Figure 5, right panels), suggesting that signaling events modulating these ethanol-responsive genes may also be involved in ethanol LORR behavior.

Of the 32 genes identified by analysis of Fyn/LORR correlates of ethanol-responsive gene expression across Fyn null and recombinant inbred lines (Figure 4C), only 2 genes (Cacna2d1 and Ptprb) were in common with the basal network (Figure 4B) other than Fyn kinase itself, suggesting the ethanol-responsive network is largely distinct from basal differences in expression. However, although the individual genes differed, this Fyn centric ethanol-responsive network was over-represented for functional categories related to ion channels (Table S7), as seen with basal Fyn network; notably including Kcnj9, Cacnb4, Kcnma1, and Gabra2. Several of these genes have been characterized for their association with ethanol-related phenotypes in mice and humans, although their association with Fyn kinase has not previously been identified. In particular, Kcnj9 has been identified as a potential quantitative trait gene (QTG) for sedative-hypnotic withdrawal from ethanol [60], residing on mouse chromosome 1 syntenic to region of human chromosome 1 that has been identified for alcohol dependence [61-63]. Single nucleotide polymorphisms in Gabra2 are associated with alcohol-elicited cues in the medial frontal cortical area [64], alcohol dependence [65], and acute effects of ethanol in humans [66]. Genetic knock-in of an ethanol insensitive mutant for Gabra2 in mice causes increased acute ethanol-induced hypnosis, loss of motor stimulation, and altered ethanol-drinking behavior [67]. In respect to Kcnma1, acute ethanol is also known to modulate the voltage and Ca⁺⁺ sensitivity of the BK potassium channel [68] causing acute intoxication in the model organism Caenorhabditis elegans [69].

The resulting Fyn-related basal and ethanol-responsive gene networks were submitted to GeneMANIA (www.genemania.org) [32] to both validate our networks across multiple external datasets related to gene co-expression, genetic interaction, and protein interaction, and perhaps identify network members missed in our analysis. The networks from GeneMANIA are qualitatively similar to our networks within PFC, suggesting these genes have tightly inter-related functional and regulatory mechanisms. Approximately 65% of all links between two genes within the basal network were due to protein-protein interactions (i.e. physical interactions) (Figure 6A). Although not a direct proof of causality, this GeneMANIA analysis thus provided strong in silico evidence supporting our microarray-derived Fyn-centric networks correlating with LORR behavior. The ethanol-responsive network (Figure 6B) was slightly different with approximately 81% of all connections derived from co-expression of genes; however, a subset of genes may physically interact with Fyn kinase. GeneMANIA datasets do not necessarily address gene-gene interactions following ethanol exposure and thus may underestimate the degree of physical interactions under our particular experimental conditions.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 6. Independent assessment of Fyn kinase correlation networks.

GeneMANIA web-tool (http://genemania.org/) analysis using public resources for Fyn-related gene expression networks: (A) Reconstruction of basal network from Figure 4B. (C) Reconstruction of ethanol-responsive network from Figure 4C. Grey nodes are query genes and white nodes are genes predicted form the GeneMANIA algorithm. Red arrow indicates the location of Fyn kinase.

https://doi.org/10.1371/journal.pone.0082435.g006

GeneMANIA predicted the membership of multiple other candidate genes to both the basal and ethanol-responsive networks (Figure 6, nodes labeled in white). Notably, the analysis predicted the neurotrophic receptor Ntrk2 and the phosphatase Pten within the basal network, and a subunit for the serotonin (Htr2a) and GABA-A (Gabra1) receptor within the ethanol-response network. All of these genes have been previously reported to have associations with the molecular and behavioral responses to ethanol. Thus, our analysis has defined a set of Fyn-related correlation networks from the measurement of steady state mRNA following either an acute dose of saline (basal) or ethanol. Overlapping these results with behavioral correlates of the loss of righting reflex across the BXD RI lines has led to the identification of Fyn-related gene networks that may have a functional role in acute sensitivity to ethanol, as seen with LORR.