Our studies here showed that C57BL/6NCrl mice express a striking degree of stable, inter-individual variation in ethanol drinking behavior with greater than 10-fold differences within a drinking session. We suspect these differences were generated by subtle environmental differences such as rearing behaviors 
, intrauterine position, social interactions or stress 
. Individual variation within the C57 substrains has been reported for ethanol drinking behaviors 
as well as in stress responsiveness 
that may be a contributing factor to ethanol preference 
. For example, C57BL/6J mice are known to consume slightly more ethanol in a 2 bottle choice paradigm than C57BL/6NCrl 
and also display variation of individual ethanol intake (unpublished observation), albeit the range is smaller than with C57BL/6NCrl mice. C57BL/6J mice have several chromosomal regions that are duplicated in comparison to C57BL/6NCrl mice 
, and may account for the differences in overall consumption and range of individual variation between the two strains. In the C57BL/10 substrain, Little et al
. reported that within-strain preference variation was not correlated with gender or ethanol metabolism, and could not be altered by simple environmental disturbances 
. Gonzalez et al. have shown that C57BL6/J and C57BL/6NCrl also do not differ in ethanol metabolism rates 
. Regardless of which environmental conditions may have contributed to variation in ethanol drinking behaviors, we hypothesized that the differences could be mediated by individual variation in basal gene expression.
The studies here employed a unique experimental design that allowed long-term measures of ethanol drinking behavior, ensured that such behavior was stable upon reinstatement, and permitted assaying gene expression differences in individual animals off ethanol. This allowed identification of expression patterns presumably “predictive” of drinking behavior rather than simply resulting from such. However, even with the current design, we cannot totally eliminate the possibility that some of our gene expression results reflect, rather than cause, individual variation in drinking behavior. Additionally, even though the behavioral data showed extinction of an ethanol deprivation effect after several cycles of withdrawal and reinstatement (Figure S1
), we cannot exclude that some of our microarray results do indeed reflect expression changes secondary to ethanol deprivation. Perhaps most likely, there could be a complex interaction between basal, ethanol-deprivation and ethanol-responsive gene expression patterns underlying these microarray results and perpetuating the long-term drinking patterns seen in these animals. However, as discussed below, the highly significant overlap of our expression gene sets with data from a meta-analysis on basal gene expression correlating with ethanol consumption 
, strongly suggests that our results largely represent basal individual variation in gene expression that influenced drinking behavior.
The current studies showed a potential role for epigenetic regulation of ethanol preference in B6 mice. In the NAc, genes with chromatin remodeling Gene Ontology function or classified in the HDAC complex had differential expression between high and low ethanol-drinking animals (see ). Intriguingly, genes involved in both histone modifications as well as genes involved in DNA methylation events were significantly altered in the nucleus accumbens but not in other brain regions assayed. Thus, our genomic findings suggest an extensive and complex representation of chromatin modification gene networks as contributing to variation in ethanol intake specifically in the nucleus accumbens.
Inhibiting HDAC activity with TSA injections increased ethanol intake above baseline and vehicle-treated levels, supporting a role for chromatin modifications in the modulation of ethanol preference. The complex changes in chromatin modification gene expression made it difficult to predict how directly altering histone acetylation might affect drinking behavior. We suggest, however, that any TSA-induced change in ethanol drinking across individuals or as a population is supportive of our hypothesis regarding a role of chromatin modification in driving individual variation in ethanol intake. This data is the first to show modulation of drinking behavior by altering chromatin acetylation. However, evidence of ethanol-induced chromatin remodeling has been reported in hepatocytes 
and in mouse brain 
. In cultured cortical neurons, ethanol increases NR2B transcription possibly through epigenetic modifications such as methylation of CpG islands 
. Acute ethanol increases histone H3 and H4 acetylation and decreases HDAC activity in amygdala, while ethanol withdrawal produces the opposite response with decreased histone acetylation 
. Social stress induces histone H3 demethylation at certain Bdnf
promoters, leading to decreased Bdnf
. Together, these studies demonstrate that environmental factors such as social stress or drug taking can modify chromatin and support a role for chromatin remodeling in the formation of stable neuronal adaptations underlying individual differences in drinking behavior.
Our bioinformatics analysis of gene expression correlating with ethanol drinking also identified gene networks involved in synaptic vesicle biogenesis and recycling (). Many of these genes have previously been implicated as playing a role in ethanol drinking or acute response to ethanol. For example, syntaxin-binding protein, STXBP1, anchors synaptic vesicles to the plasma membrane and was positively correlated to ethanol drinking in our studies. Stxbp1
was previously identified as a candidate gene for a mouse Chr2 ethanol preference locus 
. RAB3A, a small GTPase associated with synaptic vesicle trafficking and neurotransmitter release 
, was positively correlated to ethanol intake and protein expression was 1.7 fold higher in heavy drinking mice. This gene may play a role in sensitivity to the acute ataxic and sedative effects of ethanol in C. elegans
and mice 
We also identified an inverse correlation between Bdnf
mRNA levels and individual ethanol consumption. BDNF regulates multiple synaptic vesicle-related proteins, including several listed in , such as synaptotagmin, synaptophysin 
, AP2 complexes 
, and STXBP1 
. BDNF has been implicated in neuroplasticity from multiple drugs of abuse 
. In clinical studies, peripheral BDNF is lower in dependent alcoholics and patients with a positive family history of dependence as compared to normal controls and dependent patients with a negative family history 
. McGough et al. 
also showed that Bdnf
under-expression in Bdnf
+/- mice caused increased ethanol consumption, consistent with Bdnf
mRNA expression observed in the current study, where Bdnf
is lowest in mice with the highest ethanol intake. We do not believe, therefore, that Bdnf
expression levels seen in our studies were secondary to ethanol exposure itself. In support of this, we and other investigators have shown that acute ethanol injection (2 g/kg i.p.) in B6 or D2 mice increases Bdnf
expression and that after 4 weeks of 2-bottle choice ethanol drinking, Bdnf
is increased in the dorsal striatum versus non-ethanol controls 
. Thus, we suggest that lower Bdnf
expression in the low drinking mice was possibly a causal factor in individual drinking behavior variance, rather than secondary to drinking behavior itself. We cannot currently exclude the possibility that ethanol drinking followed by withdrawal (4 days) caused the correlated changes in Bdnf
expression. Together, these findings on Bdnf
and synaptic vesicle-related gene expression are strong evidence supporting an important link between regulation of synaptic vesicles and individual variation in ethanol intake.
In the present study, many of the most robust gene expression changes were found in the nucleus accumbens and prefrontal cortex. Brain regional differential gene expression is not surprising considering the proposed different roles of each region in ethanol responses 
. Furthermore, we have seen such inter-region diversity with our prior studies on acute ethanol– such as Bdnf
expression only being regulated in nucleus accumbens 
. The regional differences are potentially functionally significant. For example, glutamate signaling, the major excitatory feedback to the ventral tegmental area, was altered in prefrontal cortex. The finding of expression differences related to potential epigenetic regulation events only in nucleus accumbens, is particularly intriguing given the known role of that region in drug reward.
Importantly, we found significant overlap between our gene lists and a previously published meta-analysis of basal brain gene expression across mouse strains with differing ethanol preference 
. Several functional categories potentially involved in drinking phenotypes were also over-represented in both studies, including PI3K/Akt and PTEN signaling, protein ubiquitination and mitochondrial dysfunction. These functional categories together suggest a role for cell survival pathways, altered energy metabolism or potential neuronal toxicity due to ethanol consumption. However, animals from the meta-analysis never consumed ethanol. Therefore it is possible that animals with a proclivity to drink ethanol may have altered signaling in these pathways prior to drinking.
In conclusion, the current experiments have described persistent inter-individual variation of ethanol drinking behaviors in B6 mice and, more importantly, they define gene expression networks that may underlie these individual differences. This study utilizes variation within an inbred strain to minimize genetic influences, isolating changes in gene expression due specifically to environmental factors. These experiments have identified several gene networks previously implicated in responses to ethanol in the NAc and PFC: glutamate signaling, BDNF and genes involved in synaptic vesicle function. Perhaps most importantly, our expression studies and behavioral analysis following histone deacetylase inhibition implicate epigenetic factors involving chromatin acetylation and/or methylation as contributing to environmental modulation of ethanol intake. Defining specific gene networks targeted by these epigenetic modifications is an important goal of ongoing studies. The novel findings presented here could contribute to understanding mechanisms involved in individual risk for alcohol abuse and alcoholism in humans. Future work will focus on characterizing the genesis and implications of gene network alterations and epigenetic modifications associated with variation in ethanol drinking.