The term "genetical genomics" is entering the common parlance of researchers, denoting the combined use of genetic marker information and transcriptome analysis. Complex trait phenotyping can be fruitfully combined with genetical genomic analysis to ascertain the candidate genes and gene product interaction pathways which significantly influence the variation in expression of a phenotype of interest. We and others have utilized the genetical genomics and phenomic approaches to identify genes and pathways important in genetically influenced complex traits such as obesity, respiratory function and "addictive" behavior [1-3]. In the area of addictive behavior, our laboratory has focused on certain endophenotypes, including the genetic contributors to acute functional tolerance to ethanol [4,5] and alcohol preference in mice [2]. These studies produced results that implicated the protein products of genes important in learning and memory as contributors to acute functional tolerance to ethanol, and proteins important in orosensory systems and information processing as being important in alcohol preference in mice.

The genus/species Rattus norvegicus has been used extensively in studies of addictive behavior, not only with regard to ethanol, but also to cocaine, opiates, cannabinoids, nicotine, etc. [6,7]. Although several research groups have suggested that similar or identical biochemical systems in particular brain areas mediate self-administration of addictive drugs [6,8-10], such proposals have not been tested by applying the unbiased approach of genetical genomics, combined with phenotyping, to studies with rats. In the present study, we have used the HXB/BXH recombinant inbred rat strains [11], which represent a unique resource for alcohol research. These rats have previously been used for quantitative trait locus (QTL) analysis of various cardiovascular phenotypes and metabolic and behavioral traits [11]. We have generated brain transcriptome data for 28 recombinant inbred strains of HXB/BXH rats [11], and have combined these data with genetic marker data for these animals [12]. We have also tested the rats in a standard procedure for measuring alcohol preference [13] and these data have allowed us to generate insight into a complex of genes and their protein products which have significant association with the trait of alcohol preference/consumption in rats.

We have also taken the opportunity to directly compare the candidate genes and pathways for alcohol consumption that we identified using the rodent model, to candidate genes in human populations. We used the "Addictions Array" [14] to assess the relationships among a panel of genetic markers (SNPs) for "alcoholism and addiction" candidate genes, and the phenotype of alcohol consumption, in two populations of humans that had been characterized in the WHO/ISBRA Study on State and Trait Markers of Alcohol Dependence [15]. The results of this analysis demonstrate a convergence of the human and animal results and provide a differentiation between the genetic polymorphisms associated with predisposition to alcohol drinking and the susceptibility to alcohol dependence.

The reasons for differential intake of ethanol by individuals in a population, be it humans or other animals, have been the subject of an immense amount of research. The current dogma posits that both environmental and genetic factors contribute to individual differences in alcohol consumption [32,33]. What is sometimes confusing about the literature on alcohol consumption is the lack of discrimination between alcohol consumption in non-dependent vs alcohol-dependent individuals [34]. The DSM IV and ICD-10 criteria for alcohol dependence focus intensively on dichotomizing alcohol drinking behavior in dependent and nondependent individuals [35,36] and several recent reviews by authors of this current work [37], and others [9], have detailed evidence and hypotheses regarding the progression of events which transform the non-dependent alcohol intake phenotype to the alcohol-dependent alcohol intake phenotype.

Epidemiologic studies do indicate that the higher the levels of alcohol intake by an individual, the higher is the propensity for an individual to become alcohol dependent [38]. In this regard, high levels of alcohol intake become a risk factor for transition to alcohol dependence, but there is no a priori reason to assume that the genetic factors that may be responsible for modulating an individual's non-dependent levels of alcohol consumption are the same factors that predispose or protect an individual from becoming dependent. In other words, a genetic relationship between the propensity for high alcohol intake and the propensity for becoming alcohol dependent has not been demonstrated. In fact, one can interpret some of the data collected with mice to show a dissociation between a propensity for high alcohol preference and propensity for physical dependence. C57BL/6 mice are a genetically inbred strain that has been demonstrated to have one of the highest levels of alcohol preference, but these mice have a low propensity to develop alcohol dependence (exemplified by signs of CNS withdrawal hyperexcitability) [2,39,40]. It should thus be emphasized that the phenotype we are investigating in this paper is non-dependent alcohol intake.

What emanates as a conclusion from our genetical genomic/phenomic approach for the search for "candidate genes" that influence non-dependent alcohol intake is that a group of genes/gene products that can be linked to systems which, in the rat (and human), control appetite and satiety, play an important role in variations in the non-dependent alcohol intake phenotype (see 5). It is of interest that our earlier work, which involved a meta-analysis of data from recombinant inbred mice, selectively bred mice and a large panel of inbred mice, indicated that orosensory systems in the mice are critically involved in alcohol selection in a two-bottle choice experimental design [2]. Our two studies, taken together with a number of prior studies [41], indicate that alcohol selection by rodents involves the sensory and caloric information transduction systems dealing with the recognition, and the satiating and rewarding properties of foodstuffs. Ethanol, unlike other psychoactive drugs, has significant caloric value (seven cal/gm) and these calories have been clearly shown to have relevance to an organism's energy status [42].

Figure 3 illustrates that most of the candidate genes that we identified can be related within a neurobiological pathway that was constructed based on analysis of the literature. Many of the candidate genes identified in this study code for products (protein) that have been implicated in regulation of feeding and energy metabolism. For example, the cholecystokinin 2 (CCK2) receptor is expressed in hypothalamic regions such as the paraventricular nucleus (PVN), which contains the circuitry necessary for maintaining energy homeostasis (food intake vs energy expenditure). CCK released from the gut acts as a "satiety hormone" [43], and can activate neurons in various hypothalamic regions [44]. The melanocortin 4 receptor (Mc4r) is also expressed in the hypothalamic regions that regulate energy metabolism, and activation of these receptors by the endogenous agonist, MSH, also inhibits food intake. It has been suggested that these actions involve changes in GABA release from hypothalamic interneurons [45,46]. Both Mc4r and CCK2R are expressed at higher levels in animals that consume higher amounts of alcohol, suggesting that these animals may have an inherent propensity for reduced food intake and lower energy expenditure.

However, animals and humans ingest food not only for nourishment, but also for the rewarding properties of food, and motivational mechanisms are important for generating responses needed for food-seeking and consumption behavior. The neuronal pathways and neurotransmitter substances that regulate food intake for energy homeostasis, and for the rewarding properties of food, are intimately connected but not identical. Particularly relevant to our current studies are the hypothalamic nuclei with connections to the ventral tegmental area (VTA) that generate and carry information on hunger and satiety, and initiate the cascade of events resulting in "wanting/liking" food or other "rewards" [47,48]. Much of the hypothalamic input to the VTA which provides information on the energy status of the mammalian organism is initiated in the arcuate nucleus of the hypothalamus and transmitted via the lateral hypothalamus to the VTA dopaminergic neurons. The transmitter substance that has been associated with stimulatory input from the lateral hypothalamus is orexin, which activates the dopaminergic neurons of the VTA [49]. GABA neurons provide the major direct inhibitory input to the dopaminergic neurons of the VTA, but GABA neurons arising in the arcuate nucleus also provide inhibitory input to the orexin neurons residing in the lateral hypothalamus [49]. Thus, GABAergic activity related to food/energy requirements can generate inhibitory influences on the VTA dopaminergic neuron firing either directly or through inhibition of orexin signaling to the VTA.

The products of many of the candidate genes identified in this study can affect GABAergic neuronal activity. One of the most studied systems in this regard, and one suggested to link the homeostatic and reward pathways associated with food intake [50], is the endocannabinoid system. The endogenous agonists anandamide and 2-AG (2-arachidonoyl glycerol) act primarily through presynaptic cannabinoid 1 (CB1) receptors to inhibit GABA release [51]. The endocannabinoid system in the hypothalamus has been implicated in regulation of both food and ethanol intake [52,53]. 2-AG can act as an anterograde or retrograde messenger to inhibit GABA release [54], and can then be inactivated by hydrolysis by monoacylglycerol lipase or fatty acid amide hydrolase [55]. More recently, however, another group of enzymes has been proposed to inactivate 2-AG, not by hydrolysis, but by conversion of 2-AG to diacylglycerol. The enzyme carrying out this reaction is the product of Mogat2 (also referred to as Mgat2))[56]. Mogat2 mRNA was found to be expressed at lower levels in the higher alcohol-drinking rats, possibly resulting in a reduced rate of metabolism of 2-AG, leading to greater inhibition of GABA release.

The cholecystokinin 2 (CCK2) receptor is expressed not only in the hypothalamus, but also in other brain regions, such as the VTA or amygdala, where it is localized presynaptically on GABA neurons [57,58]. The activation of the CCK2 receptor has been shown to promote action potential-induced GABA release in certain brain areas (e.g., hippocampus) [59], but more recent data suggest that the initial increase in GABA release produced by CCK via the CCK2 receptors is followed by a reduction in GABA release [60]. Our findings of higher levels of the mRNA for the CCK2 receptor in high alcohol consuming rats could be construed as another differential means of modulating GABA release in low and high-drinking rats. Another candidate gene product, the P2X₄ purinergic receptor (the product of P2rx4), is also localized presynaptically on GABA neuron terminals [61]. Activation of P2X₄ receptors on GABA terminals may increase GABA release [61]. Since P2X₄ receptor mRNA is expressed at lower levels in the high alcohol-consuming animals, this again would point to diminished levels of GABA release in such rats.

Another identified candidate gene may further regulate the excitability of GABA neurons and GABA release. Ptpre encodes the receptor type protein tyrosine phosphatase epsilon, which has been reported to dephosphorylate and decrease the activity of the voltage-gated potassium channel, Kv2.1 [62]. The Kv2.1 channel is found on somata and proximal dendrites of inhibitory (GABAergic) interneurons [see [62]]. The expression of Ptpre is lower in animals that consume higher amounts of alcohol, and if this results in a lesser level of phosphatase activity and higher levels of the channel activity, one would expect lower excitability of GABAergic interneurons in these animals.

On the other hand, another of the candidate genes, Abat, codes for the enzyme 4-aminobutyrate (GABA) aminotransferase, which degrades GABA. This gene is expressed at lower levels in animals with higher ethanol consumption, and this situation could contribute to enhanced GABA levels in GABAergic synapses in the high alcohol-consuming rats.

The effects on GABAergic activity described above could contribute to a decreased hypothalamic GABAergic inhibitory effect on VTA dopaminergic activity in the high alcohol-consuming rats. Furthermore, the products of a number of the candidate genes can more directly affect the activity of midbrain dopamine neurons. MSH, administered into the VTA, increases dopamine release in the nucleus accumbens via actions on VTA melanocortin 4 receptors [63]. Interestingly, administration of a melanocortin 4 receptor agonist to mice (icv) reduced alcohol drinking [64], and administration of a melanocortin 3/4 receptor agonist to alcohol-preferring rats reduced alcohol intake in association with reduced food intake [65]. The CCK2 receptor is also found in the VTA [57] and likely responds to CCK release from GABAergic neurons synapsing directly with dopaminergic cell bodies in this region. Administration of CCK into the VTA results in increased dopamine release in the nucleus accumbens, mediated by activation of the CCK2 receptors in the VTA [57,66]. Taken together with the inherent differences in expression of genes whose products influence GABA release, the greater expression of the CCK2 and MC4 receptors, if it occurs in the VTA of the high alcohol consuming animals, generates a picture of lower GABAergic inhibitory tone and an increased propensity for direct dopaminergic neuron activation in the VTA by MSH and CCK.

Recent work has shown that the quantities of ethanol consumed by rats in a two-bottle choice paradigm, can, in fact, enhance the electrophysiological activity of VTA dopaminergic neurons [67]. What is unclear about such studies is whether these effects of ethanol are direct or indirect, through modulation of transmitters influencing dopamine neuron firing. For example, acutely, ethanol inhibits the function of the P2X₄ receptor [68], and this action may reduce the presynaptic function of the P2X₄ receptor and reduce GABA release, thus disinhibiting dopaminergic activity. Xiao and Ye [69] have also reported that ethanol can acutely inhibit GABAergic neuron activity in the VTA via other indirect mechanisms, while enhancing such activity in other brain areas. These actions of ethanol in the VTA would tend to disinhibit dopaminergic neuron activity and synergize with the inherently lower GABAergic tone suggested by our studies.

Another intriguing aspect of our candidate gene list is the presence of a number of genes coding for products that can affect postsynaptic aspects of GABA receptor trafficking. These candidate genes and gene products are related to rho GTPases, which affect the cytoskeleton, membrane trafficking and cell adhesion through their regulation of actin dynamics. For example, Cdc42 is a member of the rho GTPase family that can stimulate actin polymerization, affecting specific steps of vesicle trafficking such as those involved in endocytosis [70]. Cdc42 can be activated by a protein called collybistin, which is a gephyrin-binding protein [70]. Gephyrin is a scaffolding protein for GABA_A and glycine receptors [71], and these interactions provide support for a specific role for Cdc42 in clustering and trafficking of GABA_A receptors.

Two of the candidate genes that we identified are Cdc42-associated proteins. Tnk2 codes for a cytosolic non-receptor tyrosine kinase, also known as Ack (activated Cdc42-associated tyrosine kinase). This kinase has been identified as a downstream effector of Cdc42 which is important for regulation of receptor degradation via endocytic mechanisms [72]. It is widely expressed in brain, including various nuclei of the mesencephalon [73]. F2rl1 codes for a protease-activated receptor known as Par2. This G protein-coupled receptor is also widely expressed in brain, including relatively high expression in amygdala, and is found in the soma and dendrites of GABAergic and glutamatergic neurons [74,75]. Par2 agonists activate several cell signaling molecules, including phosphoinositide 3-kinase. This latter activity requires Cdc42 activation by Par2 [76]. The negative association of the expression of Ack with alcohol preference, and the positive association of Par2 expression, could enhance the endocytic recycling process associated with Cdc42-mediated actin polymerization in the animals that consume higher amounts of alcohol, leading to more rapid GABA receptor desensitization.

The products of some other candidate genes are associated with actin organization and microtubule activity, which is also important for the trafficking of receptors and other proteins. Afap1l1 codes for an actin filament associated protein. The expression of this gene is positively correlated with alcohol consumption. Microtubules are involved not only in receptor transport, but also in the movement of gephyrin to and from the synaptic plasma membrane [77]. Dynein is a microtubule motor that participates in axonal transport of neurofilaments [78], and that interacts with beta tubulin [79], the protein product of another of the candidate genes, Tubb6. The expression of all of the above-mentioned candidate genes is associated positively with alcohol consumption in the RI rat strains, supporting the postulate that there may be enhanced GABA_A receptor trafficking (endocytosis), in the brains of animals that have a propensity to consume more alcohol.

Overall, our analysis of the functional pathways defined by the identification of the candidate genes in the HXB/BXH rat strains focuses attention on GABAergic and dopaminergic activity that may set a tone that predisposes to (or against) voluntary alcohol consumption. We are proposing that, in rats, a lower inherent GABAergic tone generated by reduction in presynaptic release and a more responsive GABA receptor desensitization system predisposes to higher alcohol consumption in a free choice experimental paradigm. There are numerous studies showing that administration of agonists or antagonists of GABA or dopamine receptors can alter alcohol consumption or self-administration by rats [e.g., [80-90]].

Although we have focused attention on neuronal systems which mediate the animals' recognition of the energy status of the body and the rewarding properties of caloric substances, it should be clearly stated that the candidate genes we identified, and their products, do play important roles in other anatomically-defined neural systems. A case in point may be that dopaminergic neurons which innervate the dorsal striatum (cell bodies in the substantia nigra) may have as much to do with appetitive behavior as the systems we describe innervating the nucleus accumbens from the A10 nucleus (VTA) [91]. Another area of brain where a number of the products of the candidate genes that we identified interact is the amygdala. For instance, Mc4r is expressed in several nuclei of the amygdala and in the bed nucleus of the stria terminalis [92,93]. Studies with melanocortin 4 receptor null mutant mice showed that restoration of the receptor in a population of amygdala neurons reduced food intake [94], similar to the effect of activation of this receptor in hypothalamus. Cckbr is expressed in the amygdala as well, and CCK increases GABA release in amygdala via the CCK2 receptor [95]. Higher levels of GABA release in the amygdala may set a tone that predisposes to higher alcohol consumption, given the finding that delivery of GABA_A receptor antagonists to the amygdala reduced ethanol self-administration [81]. The endocannabinoid system can also interact with the CCK system in the amygdala, as well as modulating GABA release in this area of brain. CB1 receptor agonists inhibit the release of CCK from GABAergic neurons and Chhatwal et al. [96] have shown that the effects of cannabinoids on the extinction of conditioned fear responses are mediated through CCK2 receptors. Clearly the neuronal systems that affect conditioned fear (anxiety) responses are also involved with appetitive and reward pathways [97,98].

The ultimate goal of genetic studies using animal models of alcohol consumption is to identify candidate genes that influence a human's level of alcohol consumption. It should be stressed at this point, that the phenotype we utilized for our genetic association studies with humans was also the quantitative measure of alcohol intake. This phenotype was chosen to allow proper comparison with our studies with the rats. We also performed a separate analysis of the genetic association of the phenotype of alcohol dependence defined by both DSM IV and ICD-10 criteria with the panel of 1,350 SNP marker included on the "Addiction Array" [14]. The polymorphisms that we found to be associated with the quantitative alcohol consumption phenotype in human populations, identified genes whose products are involved in the pathways determined from our studies of differential levels of alcohol intake by the panel of recombinant inbred rats. Particularly evident were components of the GABAergic signaling pathway, including the β2 subunit of the GABA_A receptor in the Sydney population, and the GABA synthetic enzyme, GAD1, as well as the MPDZ protein, which can act as a scaffolding protein for the GABA_B receptor [99], in the Montreal population. MPDZ has previously been linked to alcohol withdrawal seizure susceptibility in mice [100], but not with levels of alcohol intake. The other genes identified in the Sydney population have also been linked to mesolimbic dopaminergic activity. CHRM5, the M5 muscarinic cholinergic receptor subtype, localized on VTA dopamine neurons, is thought to contribute to tonic excitatory regulation of dopamine transmission [101]. As noted earlier, our studies do indicate that the products of the candidate genes predisposing high alcohol drinking by rats can be linked to the modulation of VTA dopamine neuron function. Additionally, muscarinic cholinergic receptors in the VTA have been shown to play an important role in alcohol selection by rats selectively bred (P rats) for high alcohol intake [102]. PPP1R1B, also known as DARPP-32, as well as MAP kinase (the product of MAPK1 gene), are downstream targets of dopamine D1 receptors in the medium spiny neurons of the shell of the nucleus accumbens and other brain areas receiving input from the VTA. Although the "Addiction Array" SNPs did not allow us to assess polymorphisms in a number of the candidate genes identified in our studies with rats (e.g., Mc4r), it has been demonstrated that activation of melanocortin 4 receptors in the nucleus accumbens could enhance dopamine D1 receptor-mediated cyclic AMP production [103], and modulate the enzymatic activity of DARPP-32. The Mc4r mRNA, which we found to be elevated in brains of the alcohol-preferring rats, is reported to be found in the nucleus accumbens, apparently in medium spiny neurons, as well as in the VTA and hypothalamus of humans [92,103].

The overall impression that is generated by cursorily examining the results of our studies with rats and humans, and our prior studies with mice, is that little evidence may have been produced to indicate that identical genes or gene products predispose free choice alcohol intake in rodents or humans. What may be missed is the fact that certain identical neurobiologic pathways have been identified in all of these investigations. In rodents, one can posit that neurobiologic systems that participate in sensing and transducing information about the rewarding or aversive properties of foodstuffs play an important role in oral consumption of ethanol in a free-choice paradigm. Polymorphisms in the loci of genes involved in such an appetitive pathway are also associated with quantitative measures of alcohol intake in humans [104,105]. The proposition that appetitive sensory systems and/or caloric qualities of ethanol contribute to ethanol drinking by rodents is not novel [41,106-108], but the identification of such a relationship by an unbiased genetical genomic/phenomic technique allows one to progress to examination of specific pathways and genetic mechanisms in future studies.

Our results also cast some light on the influence of genetic polymorphisms on levels of alcohol drinking vs alcohol dependence in humans. Epidemiologic studies suggest a strong correlation between levels of ethanol consumed and the diagnosis of alcohol dependence [38], and alcohol intake has been taken by some as a surrogate (endophenotype) for alcohol dependence [109]. In our study, the polymorphisms that we found to be associated with the level of alcohol consumption by humans continued to have an influence in the multivariate model, even when current alcohol dependence and alcohol abuse were accounted for. This result, and the lack of association of the candidate genes for levels of alcohol consumption, with alcohol dependence/abuse in the same individuals, suggest that the genetic factors that we identified as predisposing factors for alcohol consumption may have little direct influence on alcohol dependence or abuse, but would be important for generating the major risk factor for dependence (i.e., high levels of alcohol consumption).

A recent genome-wide association study of alcohol dependence in humans used genes identified by differential mRNA expression in alcohol-consuming rats as a means of attempting to generate more credence in the candidate genes with modest statistical support for association with alcohol dependence in the analysis of the human data [110]. Our studies would caution against such an approach because genetic determinants for alcohol consumption by animals or humans may not be identical with genetic determinants of ethanol dependence. This contention is also supported by QTL analysis of free-choice alcohol consumption [111] and alcohol dependence/withdrawal [112] in animals, and alcohol consumption/alcohol dependence in humans [113-115]. The most recent and instructive study in this regard is a study with humans performed by Hansell et al. [105]. This study utilized quantity/frequency measures of alcohol consumption as a phenotype for a linkage study and found two types of QTLs. The LOD scores for certain QTLs were diminished if the primarily high level alcohol consumers were utilized for analysis, and other QTLs were enhanced if primarily high level consumers were used. What was interesting is that the QTLs which were enhanced when individuals with lower levels of alcohol consumption were dropped from the analysis were those which, in other studies, were related to "alcoholism" or alcohol dependence. QTLs identified as significant and having maximal LOD scores using the full range of alcohol consumption in this population, were not in areas of the human genome previously associated with alcohol dependence. However, at least one such QTL was previously found by de Andrade et al. to be associated with the number of drinks (on average) consumed by the subjects collected in the COGA study [109].

The genetical genomics approach, in combination with phenomics, is a powerful method for determination of candidate genes that contribute to the predisposition to alcohol consumption in rats. Informatics-based analysis of the function of candidate gene products led to the consideration that GABAergic function, and particularly GABA release modulated by peptidergic, purinergic and endogenous cannabinoid systems, as well as GABA receptor trafficking, are important components of the genetic/biochemical pathways that contribute to alcohol drinking. Our results highlight the need to identify neuronal signaling networks based on candidate genes, rather than focusing on individual genes and gene products, when attempting to understand the genetic basis of complex behavioral traits. The comparison of rat and human genetic contributors to the trait of alcohol consumption suggested that one can extrapolate from pathways - not necessarily specific genes - found in animals to begin to elucidate cross-species similarities in the genetic basis of behavior. Our results also emphasize the importance of carefully defining phenotypes for genetical genomic approaches; in this case, although high levels of alcohol drinking are phenotypically correlated with alcohol dependence, the genetic factors that contribute to the full range of alcohol consumption versus alcohol dependence in humans are distinct.

BT planned and supervised all experiments, analyzed data and drafted the manuscript. LS carried out eQTL and bQTL analyses, correlation and all statistical analyses. MPrintz provided data and tissue from the HXB/BXH rats, participated in alcohol consumption studies, and analyzed alcohol consumption data. PF participated in HXB/BXH rat alcohol consumption experiments and analyzed data. CH performed studies using the addiction array. DG supervised addiction array studies and analyzed data. GK conceived, organized and supervised the HXB/BXH alcohol consumption studies. HNR organized and supervised the HXB/BXH rat alcohol consumption studies and analyzed data. KK performed transcription factor and transcription factor binding site analysis. RLB provided data and tissue from P and NP rats. NH performed genotyping of HXB/BXH RI strains. MH performed genotyping of HXB/BXH RI strains. MPravenec provided tissue from HXB/BXH rats. JM provided genotype data from HXB/BXH RI rats. LL recruited subjects from Montreal and provided data and DNA. MD supervised studies and recruited subjects in Montreal. KMC recruited subjects in Australia and provided data and DNA. JW supervised human studies and recruited subjects in Australia. JS recruited subjects in Australia and analyzed data. BG created questionnaire for WHO/ISBRA study and generated algorithms for data analysis. PLH analyzed microarray data and drafted manuscript. WHO/ISBRA Investigators are included as authors on all studies using data from the WHO/ISBRA study, by formal agreement.