Gene expression in the mesocorticolimbic system: implications in alcohol and nicotine dependence

Alcohol is a drug legally consumed in most countries of the world. According to the World Health Organization, alcohol abuse results in 3·2 % of all deaths annually and accounts for 4·0 % of disease burden worldwide⁽¹⁾. A total of 10–18 % of injured patients attending emergency departments are alcohol-related cases⁽¹⁾. In the USA, 8 % of adults fall under the criteria of the Diagnostic and Statistical Manual of Mental Disorders, 4th edition (known as DSM-IV) for alcohol dependence or alcohol abuse⁽²⁾. Globally, alcohol abuse is the second most common mental disorder⁽Reference Robins and Regier³⁾.

Cognitive defects are a core feature of alcoholism. In recent studies, up to 80 % of alcoholics displayed cognitive deficits in abstract problem solving, spatial and verbal learning, memory formation, perceptual motor skills and problem-solving strategies⁽Reference Bates, Bowden and Barry^4–Reference Pishkin, Lovallo and Bourne⁶⁾. The loss of these cognitive abilities may significantly impair quality of life and affects the treatment process and prognosis⁽Reference Bates, Bowden and Barry⁴⁾.

The propensity to abuse alcohol is influenced by both positive (pleasurable effects) and negative (withdrawal, depressed mood states and craving) consequences and overall alcoholism is a complex disease caused by environmental and genetic factors as well as alcohol-induced neuroadaptive changes in distinct brain regions.

Commonality theories of drug abuse propose that dependence on one drug increases the likelihood of dependence on another⁽Reference Levinson, Gernstein and Maloff^7, Reference Blume, Schmaling and Marlatt⁸⁾. These theories are supported by the many reports of alcoholism and smoking co-morbidity. The incidence of smoking in alcoholics is estimated to be higher than 80 %⁽Reference Burling and Ziff^9–Reference Batel, Pessione and Maitre¹¹⁾. Alcoholics become more dependent on nicotine and have difficulty quitting smoking⁽Reference Larsson and Engel¹²⁾.

Understanding the neurobiological mechanism that contributes to alcohol abuse is critical in developing new pharmacotherapies and treatment strategies. Currently approved pharmaceutical treatments have small to medium effect and none address the persistent increased risk of relapse after drinking cessation. Additionally, the robust co-morbidity of alcohol and tobacco abuse needs to be addressed.

The dopaminergic mesocorticolimbic system (MDS) of the brain is crucially implicated in the development of drug dependence, as this neural circuit mediates motivation processes and appears to be a common target for many, if not all, drugs of abuse. Dopaminergic neural projections arise in the ventral tegmental area (VTA) in the midbrain and project to several forebrain regions including the nucleus accumbens (NAC) and the prefrontal cortex (PFC). Addictive drugs increase extracellular dopamine preferentially in the NAC either directly or via action on the VTA⁽Reference Di Chiara and Imperato^13, Reference Di Chiara and Bassareo¹⁴⁾. This is significant because the NAC plays a crucial role in learned reward anticipation⁽Reference Wilson and Bowman¹⁵⁾, drug-related learning⁽Reference Borlikova, Le Merrer and Stephens¹⁶⁾ and craving⁽Reference Dayas, Liu and Simms¹⁷⁾. Repeated abuse results in tolerance to the drug's effects and ultimately results in dependence which is characterised by withdrawal symptoms upon cessation of drug use⁽Reference De Biasi and Salas^18–Reference Vengeliene, Bilbao and Molander²⁰⁾. Craving may persist for months or years after cessation and is likely to be a consequence of changes in cell structure, function or connectivity induced by long-term drug exposure.

Alcohol's action on the MDS is complex. Alcohol does not bind a specific receptor; instead, it interferes with the activity of a number of proteins including ligand-gated ion channels⁽Reference Vengeliene, Bilbao and Molander²⁰⁾. In animals, acute alcohol enhances γ-aminobutyric acid (GABA)ergic and inhibits glutamatergic neurotransmission within the MDS⁽Reference Vengeliene, Bilbao and Molander²⁰⁾. Prolonged drinking has the opposite effect, resulting in region-specific alteration of GABA and glutamate receptor subtype expression⁽Reference Vengeliene, Bilbao and Molander²⁰⁾. Nicotine enhances an excitatory, glutamatergic drive on dopaminergic neurons by binding to specific, pre- and postsynaptic nicotinic acetylcholine receptors in the VTA⁽Reference Nisell, Nomikos and Svensson^21–Reference Schilstrom, Fagerquist and Zhang²³⁾.

The present review focuses on alcohol and nicotine abuse and gives an insight into current approaches that reveal the molecular networks underlying dependence on these drugs. Microarray gene expression profiling simultaneously compares the expression levels of thousands of genes in a tissue. Linear amplification of total RNA allows the use of small amounts of starting material and is a useful tool for exploratory investigation of gene expression in tissue with limited availability. These technologies can provide a snapshot of the transcriptome of specific brain regions relevant to drug addiction. As reviewed below, most work to date has been conducted in animal models. Additionally, several research groups have embarked on elucidating the impact of drugs of abuse on the transcriptome of the human brain. In parallel studies, several groups have characterised the protein changes in selected brain regions of the alcoholic using high-throughput proteomics.

Alcohol-related brain gene expression in animal models

The heritability of alcoholism is estimated to be as high as 50 to 60 %⁽Reference Bierut, Schuckit and Hesselbrock^24, Reference Bevilacqua and Goldman²⁵⁾. Most diagnostic criteria of alcoholism are behavioural, including failure to control drinking quantity and frequency and continuation with knowledge of the social and medical consequences. It has been argued that there is no rodent model that resembles human alcoholism in all important aspects⁽Reference Crabbe²⁶⁾. However, the power of rodent genetics allows a focus on the genetic basis of specific behavioural traits⁽Reference Crabbe²⁶⁾. Several rodent strains have been selected for display of particular alcohol-related behaviours and then used to identify candidate genes associated with these behaviours. To identify genetic components, two inbred strains, which differ in the response of interest, are chosen and intercrossed to produce recombinant inbred strains⁽Reference Crabbe, Phillips and Kosobud²⁷⁾. The recombinant strains are then used for quantitative trait locus (QTL) analysis to define chromosomal regions contributing to the phenotype of interest⁽Reference Gora-Maslak, McClearn and Crabbe²⁸⁾. Such a genomic region may harbour a large number of genes and one or several of these may be associated with the selected phenotype. The display of the phenotype could result from the expression of a polymorphism of a target gene(s) and an altered functionality of the gene product. The altered expression pattern of other genes may be a downstream consequence of this changed function.

Genetic predisposition to alcohol preference and to the effects of alcohol

One approach to identify specific genes and pathways associated with a tolerance of the acute effects of alcohol or preference for alcohol consumption is to compare the baseline gene expression of rodent strains with different phenotypes (Table 1). Bhave et al. investigated the expression levels of gene variants encoding enzymes concerned with alcohol metabolism⁽Reference Bhave, Hoffman and Lassen²⁹⁾. The expression of alcohol dehydrogenase gene variants was measured in whole-brain extracts of two alcohol-naive mouse strains with different alcohol preference and was shown to be associated with the selected behaviour⁽Reference Bhave, Hoffman and Lassen²⁹⁾. Tabakoff et al. used mice strains bred for differences in tolerance to acute effects of alcohol and compared whole-brain gene expression between representative individual animals⁽Reference Tabakoff, Bhave and Hoffman³⁰⁾. The data were processed using several methods of data filtering followed by statistical analyses. A group of genes, which were concerned with glutamate receptor activity, was revealed. These mapped to previously identified QTL. The candidate genes included those encoding the N-methyl-d-aspartic acid (NMDA) receptor and the glutamate receptor δ2 protein. Carr et al. aimed to identify genes located within one specific, alcohol-related QTL⁽Reference Carr, Kimpel and Liang³¹⁾. The researchers investigated gene expression in the NAC, PFC, amygdala (AMYG), hippocampus and striatum from congenital rat strains differing only in one alcohol preference-related QTL region on chromosome 4 and exhibiting the selected phenotype⁽Reference Carr, Kimpel and Liang³¹⁾. δ2 Glutamate receptor expression was altered in multiple brain regions, while combined analysis, across all regions, implicated protein phosphatase C signalling and neuropeptide Y (NPY) in alcohol preference⁽Reference Carr, Kimpel and Liang³¹⁾. The involvement of glutamatergic neurotransmission, particularly receptors, in the acute effects of alcohol is well established⁽Reference Vengeliene, Bilbao and Molander²⁰⁾. Protein phosphatase C and NPY signalling contribute to several aspects of alcohol-related behaviour including sensitivity to acute effects of alcohol and control of alcohol consumption⁽Reference Vengeliene, Bilbao and Molander^20, Reference Newton and Messing³²⁾.

Table 1 Baseline gene expression in rodent strains

NMDA, N-methyl-d-aspartic acid; NPY, neuropeptide Y; QTL, quantitative trait locus; NAC, nucleus accumbens; PFC, prefrontal cortex; AMYG, amygdala; MAPK, mitogen-activated protein kinase; CREB, cAMP response element binding.

In a larger study, the baseline whole-brain gene expression of three selected lines, and six isogenic strains of mice, which again differed markedly in voluntary alcohol consumption, was compared⁽Reference Mulligan, Ponomarev and Hitzemann³³⁾. Genes associated with the mitogen-activated protein kinase (MAPK) signalling pathways which regulate activity of cytoskeletal elements and mediate the influence of NFκB on transcription had predominantly higher expression in alcohol-preferring animals and were proposed to be linked to the alcohol-related behaviour. Additionally, a group of candidate genes mapped to an alcohol preference QTL on chromosome 9. Inbred Long-Sleep and Short-Sleep mice differentially express a number of genes thought to be implicated in sensitivity to the effects of ethanol⁽Reference Bennett and Johnson³⁴⁾. A promoter analysis of genes differentially expressed in the cerebellum in these strains identified cAMP response element binding (CREB) signalling to be a common feature of many of these genes⁽Reference Acquaah-Mensah, Misra and Biswal³⁵⁾. In a study comparing gene expression levels in the PFC of alcohol-preferring and non-preferring rats, MAPK and CREB signalling was again associated with alcohol-related behaviour⁽Reference Sommer, Hyytia and Kiianmaa³⁶⁾. Additionally genes concerned with mitochondrial function were differentially expressed between the rat strains⁽Reference Sommer, Hyytia and Kiianmaa³⁶⁾. NFκB, Jun and CREB signalling as well as mitochondrial function are involved with cell survival and apoptotic events. The signalling pathways and transcription regulatory elements identified by these studies might modulate the detrimental effects of alcohol on the brain.

Arlinde et al. ⁽Reference Arlinde, Sommer and Bjork³⁷⁾ compared gene expression in the NAC, hippocampus, cingulated cortex and AMYG of non-congenital alcohol-preferring, alcohol non-preferring and non-selected Wistar rat strains using Affimetrix arrays containing probes for 1000 genes. Forty-eight genes were found to be differentially expressed between the alcohol-preferring and alcohol non-preferring rat strains. Notably, genes concerned with MAPK signalling were induced in the NAC of alcohol-preferring rats and the levels of a group of transcripts involved with the regulation of cell structure were altered in the AMYG, pointing to region-specific pathways involved in alcohol preference. With the exception of NPY, which was differentially expressed in the hippocampus, none of these genes is located near known QTL for alcohol preference or consumption in rats. This discordance might have been due to the relatively small number of genes analysed. Alternatively, regulatory elements of the genes identified might be located in such QTL. Using eight inbred mice strains with distinct alcohol-related behaviour, the differential expression of genes involved in MAPK signalling in the ventral striatum was confirmed⁽Reference Letwin, Kafkafi and Benjamini³⁸⁾. Further, differentially expressed genes were associated with learning and memory while genes concerned with cell survival again featured in the cerebellum⁽Reference Letwin, Kafkafi and Benjamini³⁸⁾. The NAC and AMYG are involved in drug-related learning and craving. The differential expression of genes associated with plasticity within these regions of the MDS may therefore reflect a heightened predisposition for the development of dependence and modulate craving in alcohol-preferring animals.

Effect of alcohol administration on gene expression in the dopaminergic mesocorticolimbic system

To complement the comparison of baseline gene expression in rodent strains with distinct alcohol-related behaviours, the differences in the effect of alcohol administration on gene expression in several regions of the MDS in these strains have been investigated (Table 2). Kerns et al. exploited the differences in alcohol-related behaviour of C57BL/6 (B6) and DBA/2J (D2) mice: B6 mice drink more alcohol than D2 mice and display a larger locomotor response to acute alcohol exposure⁽Reference Kerns, Ravindranathan and Hassan³⁹⁾. Gene expression in the VTA, NAC and PFC before and after acute intraperitoneal alcohol injection was compared. Differentially expressed genes were then mapped to previously established alcohol-related QTL in a comprehensive approach to identify genes and signalling pathways associated with the alcohol response. Expression of genes associated with differentiation was altered in the VTA in both strains. Alcohol also affected transcript levels of several growth factors and developmental genes in the NAC, although the response varied in the different strains. In D2 mice, for example, brain-derived neurotrophic factor and genes associated with neurotransmission were altered in the NAC and genes associated with glucocorticoid signalling, neurogenesis and myelination were influenced in the PFC. Hence, a striking theme common to the brain regions investigated in this study was the altered expression of genes associated with plasticity, which suggests neuroplastic alterations within the MDS in the response to acute alcohol administration. Differentially expressed gene sets associated with neuroplasticity and myelination also mapped to chromosomal regions harbouring several ethanol behavioural QTL⁽Reference Kerns, Ravindranathan and Hassan³⁹⁾. The studies highlight the importance of neuroplasticity-related pathways within several regions of the MDS which may modulate the susceptibility to excess alcohol consumption.

Table 2 Effect of alcohol administration in the animal model

VTA, ventral tegmental area; NAC, nucleus accumbens; PFC, prefrontal cortex; PI, phosphatidylinositol; AMYG, amygdala; EAAT1, excitatory amino acid transporter 1; GluR2, glutamatergic α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor subunit 2; MAPK, mitogen-activated protein kinase; GABA, γ-aminobutyric acid.

As emphasised by Singh et al. the cause and effect between differential gene expression and changed phenotypes remain to be established⁽Reference Singh, Treadwell and Kleiber⁴⁰⁾. Testing the co-segregation of the phenotype of interest and expected gene expression level in independent genetic crosses or using recombinant inbred mice strains could strengthen this link⁽Reference Singh, Treadwell and Kleiber⁴⁰⁾. Towards this aim, Hu et al. utilised thirty BXD recombinant inbred strains of mice, twenty inbred strains of mice, and two replicate lines of mice selectively bred for differences in functional tolerance to acute alcohol to compare baseline gene expression⁽Reference Hu, Saba and Kechris⁴¹⁾. Differentially expressed genes were mapped to known QTL. The level of selected gene expression was correlated to the phenotype and the heritability was monitored. Eight genes were identified as candidate genes for acute alcohol tolerance. They included K voltage-gated channel, Shab-related subfamily, member 1, which dampens excitability of neurons and serine/threonine kinase, which can influence glutamatergic transmission. In combination with previous studies⁽Reference Tabakoff, Bhave and Hoffman^30, Reference Carr, Kimpel and Liang^31, Reference Letwin, Kafkafi and Benjamini^38, Reference Kerns, Ravindranathan and Hassan³⁹⁾, the results again strongly imply neuron excitability and particularly glutamatergic transmission as a target for acute alcohol. The identification of similar genes and signalling pathways associated with alcohol-related behaviours in various rodent models highlights the benefits of combining gene expression studies with genetics to reveal gene networks in the brain that may be involved in a predisposition to alcohol abuse and addiction.

Inbred rodent strains of identical genetic background were utilised to establish the impact of alcohol administration on brain gene expression. The aim of these studies was to mimic neuroadaptations associated with alcohol tolerance and dependence. Rulten et al. investigated the impact of acute alcohol on midbrain gene expression in mice, 2 h following a single intraperitoneal ethanol dose⁽Reference Rulten, Ripley and Hunt⁴²⁾. Genes concerned with transcription factor Sp1 and NFκB signalling pathways were found differentially expressed. Sp1 and NFκBiα expression were also altered in mice fed an escalating ethanol diet for 3 weeks⁽Reference Rulten, Ripley and Hunt⁴²⁾, implicating these pathways in acute as well as chronic alcohol responses.

Wistar rats offered a 12 % alcohol solution as the only source of liquid for 12 weeks were used to elucidate the effect of chronic alcohol exposure on gene expression in the hippocampus⁽Reference Saito, Smiley and Toth⁴³⁾. The expression of sets of genes encoding oxidoreductases and proteins concerned with membrane trafficking was altered. This may reflect oxidative stress and an interference with membrane function. Rimondini et al. investigated the impact of 7 weeks of repeated cycles of alcohol vapour treatment and withdrawal on gene expression in the AMYG and the cingulate frontal cortex of Wistar rats⁽Reference Rimondini, Arlinde and Sommer⁴⁴⁾. This treatment paradigm induces increased voluntary alcohol consumption. In the AMYG, transcript levels of insulin-like growth factor 2 and genes involved in phosphatidylinositol 3-kinase signalling were affected. The role of these genes in responses to alcohol dependence within the AMYG is unknown. Expression of genes associated with MAPK signalling was altered in the PFC. Once more, induction of NFκB confirms adverse effects of alcohol on this brain region. Elevated levels of genes concerned with neurotransmission were revealed in the cortex and included the cannaboid receptor CB1. Accordingly, administration of cannaboid receptor antagonist SR141716A reduces alcohol consumption in strains of alcohol-preferring rats, presenting a pharmaceutical target for the treatment of alcoholism⁽Reference Gessa, Serra and Vacca⁴⁵⁾. The induction of the excitatory glutamate transporter 1 (known as EAAT1 or GLAST) and glutamatergic α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor subunit 2 observed in this study⁽Reference Rimondini, Arlinde and Sommer⁴⁴⁾ may represent an adaptation to altered glutamatergic transmission. Repunte-Canonigo et al. utilised the same treatment and animal model to investigate the impact of intermittent alcohol administration on the transcriptome of the NAC, medial PFC and AMYG⁽Reference Repunte-Canonigo, Lutjens and van der Stap⁴⁶⁾. An observed alteration in expression of genes regulating protein kinase A signalling in all three brain regions could have a downstream modulatory effect on CREB signalling. A number of dopamine-regulated transcripts were differentially expressed in the medial PFC and NAC, which might indicate an alteration in dopaminergic drive. Together with a region-specific change in expression of GABA and AMPA receptor subunits, these results suggest extensive but specific adaptive alteration to neurotransmission in response to chronic alcohol administration.

Nicotine-responsive gene expression in the animal model

Effects of nicotine administration on gene expression in the dopaminergic mesocorticolimbic system

To a lesser extent than alcohol, the effect of acute and chronic nicotine on gene expression in the brain has also been investigated in rodents (Table 3). In mice, acute nicotine altered the expression of genes associated with plasticity in the VTA⁽Reference Chen, Che and Zhang⁵⁰⁾. Ras homologue gene family, member A, a gene involved in structural remodelling, was then selected from this gene set for a case–control gene association study in human subjects and found to be significantly associated with smoking initiation⁽Reference Chen, Che and Zhang⁵⁰⁾. Intermittent nicotine administration followed by withdrawal elicited significant locomotor sensitisation in quasi-congenic RQI and donor BALB/cJ mice but not in C57BL/6ByJ mice. Gene expression was monitored in the VTA of these strains⁽Reference Vadasz, Saito and O'Brien⁵¹⁾. Nicotine-responsive genes were concerned with signal transduction, intracellular protein transport, proteasomal ubiquitin-dependent protein catabolism, and neuropeptide signalling. While some nicotine-responsive genes and pathways overlapped, there were distinct differences between the strains. This suggests common regulatory mechanisms across inbred strains but also, that a small difference in genetic constitution significantly affects transcriptome response to nicotine.

Table 3 Effect of nicotine in the animal model

RhoA, Ras homologue gene family, member A; VTA, ventral tegmental area; MAPK, mitogen-activated protein kinase; PFC, prefrontal cortex; AMYG, amygdala; PI, phosphatidylinositol; NAC, nucleus accumbens; EGFR, epidermal growth factor receptor.

An earlier microarray study established that repeated nicotine treatment affects several signal transduction pathways in the PFC, NAC, VTA and AMYG⁽Reference Konu, Kane and Barrett⁵²⁾ of Holtzman rats. In the PFC, VTA and AMYG, genes involved in MAPK signalling were revealed as nicotine sensitive. mRNA levels of genes involved in phosphatidylinositol signal transduction were affected in the NAC, VTA and AMYG in a region-specific manner. Transcript levels of genes associated with the epidermal growth factor receptor (EGFR) signalling pathway were altered in the AMYG. In a follow-up study using pathway-focused arrays these signalling cascades were investigated further in the PFC, striatum, medial basal hypothalamus, AMYG and VTA of the rat brain in response to subacute and chronic systemic nicotine administration⁽Reference Li, Kane and Wang⁵³⁾. Rats were killed at 3, 7 and 14 d of nicotine administration. The observed changes in gene expression were highly region-specific and dynamic over the time studied. Thus, genes involved in signalling events were up-regulated at day 3 in the AMYG and at day 7 in the NAC and down-regulated over time in the VTA. These changes may be an adaptive process to chronic nicotine treatment. The expression of genes associated with the cytoskeleton was reduced after 14 d in the AMYG. Such genes were altered in the opposite direction in the PFC, which may indicate region-specific restructuring events in response to chronic nicotine. Genes concerned with inhibitory GABAergic neurotransmission were down-regulated while those concerned with glutamatergic transmission were induced, indicating an increased excitatory drive in response to systemic nicotine administration. In the medial basal hypothalamus a reduction of transcripts associated with ubiquitination was evident over time. These alterations were further explored in the PFC and medial basal hypothalamus in a third study by the same research group, again utilising customised pathway-focused microarrays⁽Reference Kane, Konu and Ma⁵⁴⁾. The 14 d nicotine administration applied in this study modulated expression of genes associated with the proteasome–ubiquitin system in a region-specific manner. The authors hypothesise that the observed changes may be caused by oxidative stress or changes to synaptic activity. Together with previous findings⁽Reference Vadasz, Saito and O'Brien⁵¹⁾ this strongly suggests that nicotine has distinct effects on cellular protein homeostasis within the MDS.

Another research group investigated gene expression in response to repeated intermittent nicotine administration on the parietal cortex of mice and identified altered expression of a set of transcription factors⁽Reference Belluardo, Olsson and Mudo⁵⁵⁾. Subsequently they utilised in situ hybridisation to confirm elevated levels of the immediate early genes, Egr-1 and Egr-2, as well as the orphan nuclear receptor, Nr4a1, in the parietal cortex and the hippocampus. The transcription factor activities of the protein products of these nicotine-sensitive genes could induce expression of a large number of other genes in a secondary response.

Gene expression profiling of the human prefrontal cortex of chronic alcoholics

Changes to gene expression and molecular pathways within the MDS associated with long-term alcoholism have been identified by investigating human post-mortem brain tissue of chronic alcoholic and control cases⁽Reference Flatscher-Bader, Zuvela and Landis^63–Reference Iwamoto, Bundo and Yamamoto⁶⁹⁾. The PFC (particularly Brodmann area 9; BA 9) is involved in working memory and is connected to the hippocampus, association areas and parietal cortex, thus providing information on the context of stimuli for higher-order sensory processing. Alcohol abuse has an adverse impact on this brain region. Older alcoholics exhibit a decrease in white matter volume⁽Reference Harper and Kril⁷⁰⁾ and a significant neuronal loss in grey matter⁽Reference Kril, Halliday and Svoboda^71, Reference Kril and Harper⁷²⁾. The PFC was the first brain region in human subjects to be investigated by gene expression profiling by two research groups⁽Reference Flatscher-Bader, Zuvela and Landis^63–Reference Liu, Lewohl and Dodd^67, Reference Iwamoto, Bundo and Yamamoto⁶⁹⁾.

In early microarray studies on the PFC of the chronic alcoholic, restrictions in human brain tissue resources required pooling of mRNA samples. These studies revealed changes to the expression of several hundred genes in the chronic alcoholic⁽Reference Mayfield, Lewohl and Dodd^65, Reference Lewohl, Wang and Miles⁶⁶⁾, including genes concerned with myelination. Advances in technology later enabled investigators to profile the transcriptome of the PFC in individual samples⁽Reference Flatscher-Bader, van der Brug and Hwang^64, Reference Liu, Lewohl and Dodd^67, Reference Iwamoto, Bundo and Yamamoto⁶⁹⁾. These studies also identified an alteration in the expression of genes concerned with myelination, which may point to an interference of alcohol with membrane integrity and neurotransmission. Further, in two studies the expression of transcription factors, including the immediate early genes, activator protein 1 (AP-1) and CREB, was sensitive to chronic alcohol abuse⁽Reference Flatscher-Bader, van der Brug and Hwang^64, Reference Iwamoto, Bundo and Yamamoto⁶⁹⁾. These results are in accordance with expression profiling in the animal models⁽Reference Sommer, Hyytia and Kiianmaa³⁶⁾ and may indicate an alteration to cell fate in the PFC of the long-term alcoholic. Certainly, loss of neurons is a characteristic of the PFC of the alcoholic⁽Reference Harper and Kril⁷⁰⁾.

The detrimental effects of alcohol on the brain may be an effect of the direct toxicity of alcohol and its first metabolite acetaldehyde. In addition, changes in the balance of inhibitory and excitatory neurotransmission in the cortex may elevate neuronal sensitivity to glutamate during withdrawal to levels causing excitotoxicity⁽Reference Hoffman and Tabakoff^73–Reference Rossetti and Carboni⁷⁵⁾. Interestingly and similar to the animal model⁽Reference Rimondini, Arlinde and Sommer⁴⁴⁾, one study revealed a robust induction of the glial glutamate transporter, excitatory amino acid transporter 1, in the PFC of the human alcoholic⁽Reference Flatscher-Bader and Wilce⁷⁶⁾. Consequently, the altered expression of this transporter was confirmed at the protein level⁽Reference Flatscher-Bader and Wilce⁷⁷⁾. Taken together these studies point to a protective role of excitatory amino acid transporter 1 against the potentially cytotoxic glutamate level following withdrawal⁽Reference Rossetti and Carboni⁷⁵⁾. Chronic ethanol treatment of rats results in DNA damage and an increased level of heat shock proteins in the cortex⁽Reference Calabrese, Renis and Calderone^78, Reference Renis, Calabrese, Russo and Calderone⁷⁹⁾. Correspondingly, in the PFC of the human alcoholic, genes associated with DNA repair and those encoding a number of heat shock proteins and free radical scavengers were induced⁽Reference Flatscher-Bader, van der Brug and Hwang^64, Reference Mayfield, Lewohl and Dodd^65, Reference Iwamoto, Bundo and Yamamoto⁶⁹⁾. The down-regulation of mitochondrial genes including those associated with electron transport or energy production was revealed in two studies⁽Reference Flatscher-Bader, van der Brug and Hwang^64, Reference Liu, Lewohl and Dodd⁶⁷⁾. Mitochondrial function was again implicated in the animal model of alcohol preference⁽Reference Sommer, Hyytia and Kiianmaa³⁶⁾. These changes indicate that disruption of mitochondrial function may be a possible source of oxidative stress. In summary, microarray gene expression studies on the PFC of human alcoholics identified gene groups associated with a disturbance of membrane function, transcription and oxidative stress and cell death and survival. The alteration in transcription observed in these genes may reflect the neuronal cell loss evident in the long-term alcoholic⁽Reference Kril, Halliday and Svoboda^71, Reference Kril and Harper⁷²⁾.

Comparisons of the various studies on the human PFC reveal few common specific alcohol-sensitive genes and only partial overlap in functional gene groups. The inconsistency between the studies emphasises the need for strictly controlled experimental set ups. As reviewed previously, uncharacterised cases as well as experimental factors may play a role in the divergent results⁽Reference Flatscher-Bader, van der Brug and Landis⁸⁰⁾. Possible confounding factors in human studies include mode of death, tissue pH, white matter contamination, post-mortem interval, onset of drinking, age, intoxication state at death, co-morbidity with other drugs of abuse or with underlying psychiatric disorders. Distinct differences in experimental design, application of various microarray platforms together with platform-specific methods to treat and hybridise input RNA sample and the choice of method of analysis may lead to pronounced differences in the results. Despite these caveats, the studies have revealed alcohol-sensitive gene groups highly relevant to the pathology evident in the PFC of the chronic alcoholic. The accordance of genes and gene networks with those identified in expression profiling using rodent models highlights that the two approaches are complementary in the identification of alcohol-responsive molecular pathways.

Comparative gene expression profiling of regions of the dopaminergic mesocorticolimbic system in chronic alcoholics

In a series of microarray studies our research group profiled the PFC (BA 9), NAC and VTA of long-term alcohol abusers⁽Reference Flatscher-Bader, Zuvela and Landis^63, Reference Flatscher-Bader, van der Brug and Hwang⁶⁴⁾ using a consistent method of RNA extraction, linear amplification, microarray platform, hybridization, as well as data analysis. In the studies of the BA 9 region of the PFC and the NAC a two-colour design was used⁽Reference Flatscher-Bader, van der Brug and Hwang⁶⁴⁾, while for expression screening of the VTA a universal reference was applied⁽Reference Flatscher-Bader, Zuvela and Landis⁶³⁾. The results of each microarray study were verified by real-time PCR on selected genes.

Comparing differentially expressed genes between the three studies, the highly region-specific action of alcohol was a striking observation⁽Reference Flatscher-Bader, van der Brug and Landis⁸⁰⁾. No alcohol-sensitive gene was common between the three studies and only 4 % between any two. Functional clusters of alcohol-sensitive genes identified in the NAC and VTA were clearly distinct from the PFC. With the exception of a group of genes concerned with transcription in the VTA, none of the main themes present in the PFC was observed in the VTA or NAC. The most pronounced functional group in the VTA contained genes associated with cell signalling. In the VTA and the NAC a number of genes were associated with synaptic and/or structural plasticity.

In early microarray studies, genes associated with alcohol abuse were grouped manually into functional groups⁽Reference Flatscher-Bader, van der Brug and Hwang^64–Reference Lewohl, Wang and Miles⁶⁶⁾. More recently, functional data mining of large microarray datasets using independent softwares has been applied as an unbiased tool to reveal the biology underlying a dataset⁽Reference Flatscher-Bader, Zuvela and Landis^63, Reference Liu, Lewohl and Dodd⁶⁷⁾. The web-based software Database for Annotation Visualisation and Integrated Analysis (DAVID/EASE; National Cancer Institute at Frederick, Frederick, MA, USA; http://david.abcc.ncifcrf.gov/) groups genes into specific categories considering their multi-functionality⁽Reference Dennis, Sherman and Hosack^81–Reference Huang da, Sherman and Lempicki⁸³⁾. Further, the software identifies over-represented annotation terms and biological themes using κ statistics and maps genes to canonical Kyoto Encyclopedia of Genes and Genomes pathways⁽Reference Hosack, Dennis and Sherman⁸²⁾. DAVID/EASE analysis (National Cancer Institute at Frederick) on the datasets of differentially expressed genes was performed for the VTA, PFC (Table 4) and the NAC (Table 5) to reveal over-represented annotations and functional gene groups within the three brain regions. Groups of genes associated with transcription or apoptosis were enriched in the PFC only. Over-representation of genes coding for proteins involved with modification of the extracellular cellular matrix was unique to the NAC. For example, tissue inhibitor of metalloproteinase (TIMP) 1 and TIMP4 were induced, while TIMP2 was down-regulated. This suggests an alteration to extracellular matrix homeostasis. Additionally, genes concerned with cell adhesion featured in a separate cluster. They included integrin, α 6, T-cell lymphoma invasion and metastasis 1, contactin-associated protein-like 2 and ninjurin 2. Genes associated with tissue development, neuronal cell morphology and neuron morphogenesis and differentiation were also tied into multiple independent clusters. These functional themes strongly suggest structural plasticity in the NAC of the chronic alcoholic. To a lesser extent, genes which are expressed during structural plasticity were also identified as enriched in one functional cluster in the VTA⁽Reference Flatscher-Bader, Zuvela and Landis⁶³⁾. Of interest within these functional clusters were molecular targets common to both the NAC and VTA. In both regions, chronic alcohol abuse changed the expression of genes participating in integrin-mediated alteration of the cytoskeleton. Enduring, structural plasticity within the NAC and VTA in response to repeated drug administration has been observed in animal models⁽Reference Robinson and Kolb⁸⁴⁾. These changes are thought to underlie the persistence of craving and relapse. Interestingly, comparing studies on human and animal models, a meta-analysis identified genes concerned with structural plasticity to feature in the group of top candidate genes for alcoholism⁽Reference Rodd, Bertsch and Strother⁸⁵⁾. Further, in the human VTA, genes concerned with neurotransmitter transport and release were significantly over-represented amid the alcohol-sensitive genes identified⁽Reference Flatscher-Bader, Zuvela and Landis⁶³⁾. In conclusion, comparative expression profiling of distinct regions of the human MDS revealed a region-specific impact of alcohol. Structural plasticity in the NAC and VTA may underlie enduring alterations to connectivity within these two brain regions and open up novel avenues for drug research.

Table 4 Functional enrichment clustering of genes associated with alcohol abuse in the human prefrontal cortex (PFC)*

* Differentially expressed genes were identified by comparing gene expression in the PFC of chronic alcoholics and control cases. This gene list was subjected to functional enrichment clustering using the Database for Annotation, Visualization and Integrated Discovery (DAVID; http://david.abcc.ncifcrf.gov/). Clusters with at least one significant annotation term are presented. Annotation terms are listed with associated Gene Ontology (www.geneontology.org/), InterPro (www.ebi.ac.uk/interpro/) and SwissProt/PIR/Uniprot (www.ebi.ac.uk/swissprot/, http://www.uniprot.or) categories.

Table 5 Functional enrichment clustering of genes associated with alcohol abuse in the human nucleus accumbens (NAC)*

* Differentially expressed genes were identified by comparing gene expression in the NAC of chronic alcoholics and control cases. This gene list was subjected to functional enrichment clustering using the Database for Annotation, Visualization and Integrated Discovery (DAVID; http://david.abcc.ncifcrf.gov/). Clusters with at least one significant annotation term are presented. Annotation terms are listed with associated Gene Ontology (www.geneontology.org/), InterPro (www.ebi.ac.uk/interpro/) and SwissProt/PIR/Uniprot (www.ebi.ac.uk/swissprot/, http://www.uniprot.or) categories.

Tobacco co-abuse in chronic alcoholics

There is a strikingly high coincidence between alcoholism and smoking. Alcoholism is more prevalent in smokers than non-smokers and 80–90 % of alcoholics are heavy smokers compared with 33 % of the global population⁽Reference Batel, Pessione and Maitre¹¹⁾. Smoking cessation is more difficult in former and current alcoholics and cessation of smoking improves the chances of a reduction in alcohol consumption⁽Reference Larsson and Engel¹²⁾. In addition, smoking may act as a precipitant of relapse to alcoholism⁽Reference Larsson and Engel¹²⁾. In co-morbid cases, drinking elevated craving for cigarettes and smoking cues increased the desire for alcohol⁽Reference Drobes, Beylotte and Scott⁸⁶⁾. Recent meta-analysis of linkage and association studies have identified candidate genes which may predispose an individual to addiction to more than one drug⁽Reference Li and Burmeister⁸⁷⁾. These included genes involved in dopamine and serotonin neurotransmission and in drug metabolism. Potential predisposition to alcohol and nicotine dependence predominated⁽Reference Li and Burmeister⁸⁷⁾. Despite the potential for extensive interactions, most studies on gene expression within the MDS of the human alcoholics using high-throughput genomics have not taken account of the common co-morbidity of smoking and alcoholism⁽Reference Flatscher-Bader, van der Brug and Hwang^64–Reference Iwamoto, Bundo and Yamamoto⁶⁹⁾. Compelling evidence arising from two recent studies identified that smoking has a profound impact on alcohol-related gene expression in the human. In the PFC and VTA, the expression levels of selected alcohol-sensitive genes were established in an extended case set of alcoholics with or without smoking co-morbidity, non-alcoholic smokers and non-smoking, non-drinking controls⁽Reference Flatscher-Bader, Zuvela and Landis^63, Reference Flatscher-Bader and Wilce⁷⁶⁾. In the PFC, smoking alone altered expression of apolipoprotein D (APOD) and midkine and increased variance in the expression of TIMP3 and excitatory amino acid transporter 1⁽Reference Flatscher-Bader and Wilce⁷⁶⁾. The neurotrophic factor, midkine, is induced after ischaemia⁽Reference Wada, Kamata and Aizu⁸⁸⁾. TIMP3 facilitates doxorubicin-induced apoptosis in neuronal culture⁽Reference Wetzel, Rosenberg and Cunningham⁸⁹⁾. APOD function is yet unknown, but the expression of this gene was altered in schizophrenics and patients with bipolar disorder⁽Reference Thomas, Dean and Pavey⁹⁰⁾. Smoking is associated with impairment of cognitive functions⁽Reference Glass, Adams and Nigg⁹¹⁾ and exacerbates alcohol-induced neuronal injury⁽Reference Durazzo, Cardenas and Studholme⁹²⁾. Taken together, these studies emphasise that smoking alone may have a negative effect on cell viability in the PFC.

In the VTA, smoking and alcohol interacted to alter the expression of excitatory amino acid transporter 2 and smoking alone induced expression of vesicular glutamate transporters 1 and 2⁽Reference Flatscher-Bader, Zuvela and Landis⁶³⁾. These changes indicate a potentiation of glutamatergic neurotransmission in response to smoking⁽Reference Mansvelder and McGehee⁹³⁾. Glutamatergic transmission is vital for the control of the VTA and may also be critical to the weighting of novelty and importance of a stimulus, an essential output of this brain region. Therefore enduring plasticity within the VTA may be a major molecular mechanism for the maintenance of smoking addiction and alcohol, nicotine and co-abuse have distinct impacts on glutamatergic transmission with important implications for the control of this core mesolimbic structure. Interestingly, preliminary clinical studies in subjects dependent on nicotine⁽Reference Harris and Anthenelli⁹⁴⁾ and alcohol⁽Reference Penberthy, Ait-Daoud and Breton⁹⁵⁾ show that modulation of glutamatergic function may be a useful tool for initiation of abstinence and prevention of relapse.

These studies of core regions of the human MDS emphasise the necessity to account for smoking as a confounding variable in studies of alcoholism both with regards to dependence but also as a component of alcohol-related brain damage. Future large-scale gene expression studies carefully delineating the effect of alcohol from that of smoking in the human MDS may further enhance our understanding of the impact of alcohol and smoking on the human brain.

Protein expression studies

Transcription profiling provides a snapshot of the mRNA spectrum of tissue under study and indicates the types of changes that may be taking place. However, mRNA is not the functional endpoint of gene expression and often mRNA abundance does not reflect protein levels⁽Reference Gygi, Rochon and Franza^96, Reference Anderson and Seilhamer⁹⁷⁾. Further, drug exposure may influence mRNA splicing or post-translational modification of a target protein, thereby altering function.

Recently, in parallel with genomic analyses, several groups have reported high-throughput studies of the proteome of MDS regions⁽Reference Lewohl, Van Dyk and Craft⁹⁸⁾. Two-dimensional gel electrophoresis, which separates proteins, firstly on isoelectric charge and then by molecular mass, can resolve over 1000 protein species. This methodology is then coupled with matrix-assisted laser desorption/ionisation time-of-flight (MALDI-TOF) to confirm identity of a limited number of protein species. Currently, these analyses have been restricted to soluble proteins of mid-range molecular weight expressed at a relatively high level.

In an initial study using pooled samples from four individuals, Lewohl et al. reported changes in the levels of 183 proteins in the PFC⁽Reference Lewohl, Van Dyk and Craft⁹⁸⁾. More recently in a detailed comparative study, Matsumoto and colleagues documented changes in the proteome of the BA 9 region of the alcoholic brain⁽Reference Alexander-Kaufman, James and Sheedy⁹⁹⁾. Further, the proteome of the grey matter was compared with that of the underlying white matter⁽Reference Alexander-Kaufman, Cordwell and Harper¹⁰⁰⁾. In these studies, the influence of confounding liver disease was also monitored.

The level of several thiamine-dependent enzymes was changed in both grey and white matter. Further, almost 50 % of the alcohol-sensitive proteins were metabolic enzymes, many of which are involved in energy transduction such as those involved in glycolysis and the tricarboxylic acid cycle. Disruption of these pathways, possibly emanating from a deficiency in thiamine metabolism, has several implications. Deprivation of ATP supply may lead directly to a loss of cellular viability but also disruption of the thiamine-dependent enzymes of the pentose phosphate pathway may reduce the supply of reducing equivalents and therefore the cells' ability to combat oxidative stress. Many of the changes in protein expression in the alcoholic were enhanced in cases with liver disease⁽Reference Matsumoto, Alexander-Kaufman and Iwazaki¹⁰¹⁾.

Gene expression studies on the human dopaminergic mesocorticolimbic system of heroin and cocaine addicts

The impact of other drugs of abuse, in particular heroin and cocaine, on gene expression in the human NAC has been studied by Albertson et al. using the Affimetrix platform⁽Reference Albertson, Schmidt and Kapatos^102, Reference Albertson, Pruetz and Schmidt¹⁰³⁾. Seven heroin abusers were compared with matched controls and approximately 1000 genes were identified as differentially expressed⁽Reference Albertson, Schmidt and Kapatos¹⁰²⁾. Of particular interest was the down-regulation of genes encoding pre-synaptic proteins. These proteins are involved in aspects of neurotransmitter release, including vesicle storage, release and recycling. The results may indicate decrements in neurotransmission within the NAC of the chronic heroin abuser. In a second study, Albertson et al. compared gene expression of ten chronic cocaine abusers with matched controls⁽Reference Albertson, Pruetz and Schmidt¹⁰³⁾. Only forty-nine genes were identified to be differentially expressed in the cocaine abusers. Therefore and in contrast to the heroin study and to our studies on the NAC of chronic alcoholics, cocaine exposure might affect only a select group of genes in the human NAC. In both studies, differentially expressed genes were manually assigned to diverse functional groups. In the cocaine abusers, a number of genes were glia-specific and included genes involved in myelination. Particularly, a decrease in expression of myelin basic protein, myelin-associated oligodendrocyte basic protein and proteolipid protein was revealed. Myelin basic protein and proteolipid protein constitute about 80 % of myelin in the central nervous system and a dysregulation of myelin in the cocaine abuser is consistent with white matter pathology in these addicts⁽Reference Bartzokis, Beckson and Lu^104–Reference Chang, Ernst and Strickland¹⁰⁶⁾. These studies highlight the specific and diverse actions of different drugs of abuse on the human NAC.

The proteome of the NAC has also been explored in cocaine abusers⁽Reference Tannu, Mash and Hemby¹⁰⁷⁾ and in self-administering Rhesus monkeys⁽Reference Tannu, Howell and Hemby¹⁰⁸⁾. Consistent with the concept of drug-induced plasticity in this region of the human brain, several proteins associated with synaptic plasticity were found to have altered levels in cocaine abusers. In addition, altered metabolic flux was suggested by changes in the level of proteins involved in energy metabolism whilst other changes were suggested to reflect increased oxidative stress. In the subsequent study of the primate model, protein changes indicated a dysregulation of proteins related to cell structure, signalling and metabolism including energy generation in the mitochondria.

Future studies

Rodent models of drug-seeking behaviours and drug dependence continue to be developed. Studies of these models coupled with advances in rodent genetics will provide important information on the mechanism of action at the cellular and molecular level. It will also be important to continue with exploration of the human brain in parallel with these animal studies. Currently, a major difficulty is matching the physiological conditions of each human case. In addition to the variation in lifestyle pre-mortem, differences in post-mortem treatment of tissues are a confounding factor. However, continued improvement in public awareness, documentation and tissue treatment in the various Brain Banks worldwide should result in increased numbers of well-controlled cases for future studies.

Expression profiling experiments, by nature, generate a large amount of data. Application of analysis softwares such as DAVID (National Cancer Institute at Frederick; http://david.abcc.ncifcrf.gov/), Pathway Assist by Stratagene (Agilent Technologies, Santa Clara, CA, USA) and others can indicate underlying biological themes. Recent advances in tools to group genes by variation of expression across individual samples have made possible the exploration of gene networks⁽Reference Horvath, Zhang and Carlson¹⁰⁹⁾. This type of analysis has been directed at diverse biological problems, for example, evolution of the human cortex⁽Reference Oldham, Horvath and Geschwind¹¹⁰⁾ and genes differentially expressed during the progression of Alzheimer's disease⁽Reference Miller, Oldham and Geschwind¹¹¹⁾. Similar approaches to group differentially expressed genes in the drug-affected MDS may reveal expression-related gene groups and genes central in these networks (hub genes) which may have key roles. This will give an insight into the organisation of drug-sensitive genes and possible new pharmacological targets.

Recent proteomic analyses have significantly contributed to understanding the pathophysiology of the complicated disease which is alcoholism. The data from these studies illustrate both the power and limitations of current technologies for proteomic analyses. Clearly these approaches identified important biologically relevant changes at the end point of gene expression; however, available technologies have limited ability to detect high- or low-molecular-weight proteins, and hydrophobic or insoluble proteins associated with membranes. These proteins are often important in understanding the mechanisms of central nervous system diseases. Advances in protein separation and identification have the potential to analyse expression levels of proteins, including neurotransmitter receptors and transporters that have not been examined using current techniques.

Changes in the transcriptome of the alcoholic brain may be a result of many influences. Some changes in expression may be a result of drug toxicity, some may reflect neuroadaptations associated with the development of tolerance and others may result from genetic variation. The latter may have importance in predisposition to drug dependence. Animal studies relating changes in gene expression to behaviours and to QTL maps have been invaluable in identifying potential inheritable genetic factors. Recent genome-wide association studies of alcohol⁽Reference Rice and Saccone^112–Reference Ducci and Goldman¹¹⁵⁾ and nicotine dependence⁽Reference Uhl, Liu and Drgon^116, Reference Bierut, Madden and Breslau¹¹⁷⁾ have identified several potential markers. Further meta-analyses comparing human gene expression and gene association studies will greatly aid the identification of genes and gene networks crucially involved in the predisposition to alcohol, tobacco and co-abuse in humans⁽Reference Li and Burmeister⁸⁷⁾. One recent study resulted in a systems biology network model derived by ontology-based computational genotype–phenotype association analysis of the pharmacogenetics of nicotine addiction and treatment⁽Reference Thomas, Mi and Swan¹¹⁸⁾. The model encoded a disease risk model capturing the underlying biological systems and interactions with the environment. Such combined studies and analyses may reveal novel candidates to improve current treatment regimens to cease drug abuse and prevent relapse.