Ethanol-BDNF interactions: Still More Questions than Answers

Abstract

Brain Derived Neurotrophic Factor (BDNF) has emerged as a regulator of development, plasticity and, recently, addiction. Decreased neurotrophic activity may be involved in ethanol-induced neurodegeneration in the adult brain and in the etiology of alcohol-related neurodevelopmental disorders. This can occur through decreased expression of BDNF or through inability of the receptor to transduce signals in the presence of ethanol. In contrast, recent studies implicate region-specific up-regulation of BDNF and associated signaling pathways in anxiety, addiction and homeostasis after ethanol exposure. Anxiety and depression are precipitating factors for substance abuse and these disorders also involve region-specific changes in BDNF in both pathogenesis and response to pharmacotherapy. Polymorphisms in the genes coding for BDNF and its receptor TrkB are linked to affective, substance abuse and appetitive disorders and therefore may play a role in the development of alcoholism. This review summarizes historical and pre-clinical data on BDNF and TrkB as it relates to ethanol toxicity and addiction. Many unresolved questions about region-specific changes in BDNF expression and the precise role of BDNF in neuropsychiatric disorders and addiction remain to be elucidated. Resolution of these questions will require significant integration of the literature on addiction and comorbid psychiatric disorders that contribute to the development of alcoholism.

1. Introduction

Alcoholism is a world-wide public health problem and it is estimated that over 14 million Americans suffer from some form of alcohol-related disorder (McGinnis and Foege, 1999). Long-term alcohol exposure in the adult causes neurodegeneration (atrophy of both grey and white matter), Wernicke-Korsakoff syndrome, tremors, alcoholic psychosis, delirium tremens, and withdrawal seizures (reviewed by Harper and Matsumoto, 2005). Alcohol-Related Neurodevelopmental Disorder/Fetal Alcohol Spectrum Disorder (ARND/FASD) is a leading cause of birth defects in the Western world, completely preventable, and irreversible. Exposure to ethanol during brain development induces apoptosis and impairs neuronal migration, resulting in hyperactivity, increased impulsivity, cognitive and motor deficits (Mukherjee et al., 2006).

Alcoholism is difficult to treat, with high relapse rates. Ethanol affects multiple neurotransmitter systems in the brain, including opiates, GABA, glutamate, serotonin, and dopamine (reviewed by Diamond and Gordon, 1997). Prolonged exposure leads to pharmacodynamic changes in these systems resulting in withdrawal-induced anxiety and seizures. Acute withdrawal is generally managed pharmacologically with benzodiazepines, anticonvulsants and dopamine antagonists but craving leading to relapse is more difficult to treat. Several different classes of neuroactive agents are in use therapeutically for preventing relapse. For example, acamprosate probably targets the glutamatergic system, naltrexone and nalmefene block opiate receptors and ondansetron targets 5-HT₃ receptors. In addition, the anticonvulsant topiramate has recently been shown to be effective in increasing abstinence rates among alcoholic patients (Johnson et al., 2007); the pharmacological mechanisms underlying this response are presumably related enhanced activity of GABA receptors and antagonism of glutamate receptors. This dual mechanism of action at both excitatory and inhibitory synapses is believed to mitigate the effects of chronic ethanol exposure on glutamatergic and GABAergic receptors in the mesocorticolimbic circuit (Johnson 2008). Interestingly, topiramate has also been investigated as a treatment for binge eating (Shapira et al., 2000, McElroy et al., 2003) suggesting that this compound may be effective in other appetitive disorders. None of these agents are particularly effective in alcoholism, however, only modestly increasing abstinence rates (reviewed by Mann 2004; Johnson 2008).

Alcoholism represents a heterogeneous disease and different subtypes of alcoholics are likely to respond to different therapeutic interventions. Cloninger and colleagues (1981) attempted to categorize alcoholics in a cross-fostering study of Swedish adoptees and identified 2 subtypes. Type 1 alcoholics had a later age of onset, harm-avoidant personalities and low criminality. Type 2 alcoholics had a family history of alcohol abuse, violent behavior, novelty seeking and an earlier age of onset. Cardoso et al. (2006) subsequently identified 5 subtypes of alcoholics: anxiopathic (anxious), heredopathic (family history), thimopathic (affective symptomology), sociopathic and ad(d)ictopathic (polysubstance abusing). As these classifications suggest, alcoholics frequently suffer from comorbid psychiatric disorders as well, making elucidation of the underlying motivation to drink and selection of appropriate therapeutic interventions difficult. Anxiety, depression, post-traumatic stress disorder, bipolar disorder and schizophrenia all predispose patients to alcohol abuse (reviewed by Goldstein et al., 2006). The genetic and environmental factors that contribute to mental illness therefore contribute to the development of alcoholism and substance abuse in a large percentage of patients.

The neurotrophin BDNF is an attractive candidate molecule for mediating many of the processes mentioned above. Evidence has accumulated that implicates BDNF as a regulator of development (Klein, 1994), modulator of appetite (Lebrun et al., 2006), facilitator of synaptic plasticity (Pang and Lu, 2004), potential therapy for neurodegenerative disease (Pezet and Malcangio, 2004) and mediator of addiction (Bolanos and Nestler, 2004). In addition to numerous earlier studies on BDNF in the context of FASD, recent biochemical and genetic data implicate BDNF and its associated signaling intermediates in ethanol-induced neurotoxicity, acute tolerance and addiction. There are fascinating possibilities for integrating the extensive literature on BDNF in neurodevelopmental and neuropsyciatric disorders with the recent observations of the effects of ethanol on BDNF expression and signaling. However there still remain unresolved questions, particularly in terms of region-specific effects, pharmacology and potential for pharmacotherapy.

2. A Brief History of BDNF

Neurotrophins were originally characterized as target derived substances that support the innervating axon to a level that is proportional to the size of the target tissue (reviewed by Levi-Montalcini, 1987). Mid-twentith century observations, combined with the earlier observations of Ramon y Cajal in the late 1800’s, cemented the neurotrophic hypothesis which prevails, with modifications, to this day in developmental neuroscience. BDNF is a member of a family of cysteine knot, dimeric neurotrophic substances that also includes NGF, Neurotrophin (NT)-3, and NT-4/5 in mammals. NGF has served as a model for neurotrophins since its relatively early discovery by Cohen and Levi-Montalcini in the mid 20^th century (Levi-Montalcini and Cohen, 1960). BDNF was purified over 20 years later (Barde et al., 1982) and this discovery was rapidly followed by homology cloning of a family of neurotrophins with both distinct and overlapping actions (reviewed by Barde 1994; Reichardt 2006). NGF is the survival factor for sympathetic neurons and cholinergic neurons in the CNS. BDNF and NT-4 are more widely expressed, with the highest levels of both neurotrophins in the hippocampus, cerebellum and cortex. BDNF can be produced by astrocytes under pathological conditions and after monoaminergic receptor stimulation (Zafra et al., 1992; Juric et al., 2006). BDNF is also produced by platelets (Karege et al., 2002), lymphocytes (Kruse et al., 2007), and the vascular endothelium (Wang et al., 2006; Kermani and Hempstead, 2007). BDNF is an angiogenic factor, and therefore may be involved in vascular and hemodynamic responses to ethanol as well.

There is considerable complexity within the BDNF gene itself. BDNF mRNA is encoded by up to 9 potential exons in humans (Pruunsild et al., 2007) and 8 in rodents (Aid et al., 2007) but was initially characterized as having 4 exons (Timmusk et al., 1993). Anti-BDNF transcripts are also present in humans (Pruunsild et al., 2007). The first 4 (7–8, according to the recent nomenclature) exons are encoded by unique promoters and are differentially regulated by neuronal activity and during pathological states. The final exon contains the BDNF coding sequence. Exons I–III (the original nomenclature will be used since most studies still use this scheme) are expressed in brain while exon IV is active in lung and heart (Timmusk et al., 1993). BDNF exons I–III are induced by multiple forms of plasticity and during seizures. Transcription of exons I and II is dependent on protein synthesis, while exons III and IV are not. This property gives exon III (in brain) the ability to be readily inducible and, not surprisingly, this promoter is regulated by calcium, mitogen-activated protein kinases and cAMP (Shieh et al., 1998; Tao et al., 1998; Chen et al., 2003a). In addition, Marini and colleagues identified an NK-κB binding site within promoter III that is regulated by NMDA receptors and contributes to NMDA-mediated neuroprotection (Marini et al., 2004). The BDNF promoter is also regulated by epigenetic methylation. Calcium-dependent phosphorylation of Methyl CpG binding protein (MeCP2) leads to derepression of the promoter (Chen et al., 2003b). Interestingly, BDNF mRNA is also targeted to different regions within the cell for translation, depending on the promoter used (reviewed by Tongiorgi et al., 2006).

Neurotrophins signal through high affinity tropomyosin-related kinases (Trks) which were originally named for a translocation of a portion of the TrkA gene to the smooth muscle tropomyosin locus in a colon carcinoma (Mitra et al., 1987). TrkB and the truncated receptors were subsequently characterized (Middlemas et al., 1991). Neurotrophins bind with high affinity to Trk receptors (10–100 pM). TrkA is the preferred receptor for NGF, TrkB binds NT-4 and BDNF, while TrkC is the high affinity receptor for NT-3 (reviewed by Chao and Hempstead, 1995). Trks are not completely selective and will bind other neurotrophins, although with reduced affinity. This is particularly true for NT-3, which can bind TrkA and TrkB with an affinity of 300–500 pM (Barbacid, 1995; Ip et al., 1993a,b). Traditionally neurotrophin concentrations have been expressed as ng/mL. The affinity of BDNF in these units is less than 2.8 ng/mL for high affinity binding and 28 ng/mL for low affinity binding and approximately 10 ng/mL for non-specific activation of TrkB by NT3. Affinity is also regulated by alternative splicing of transcripts that leads to an insertion in the juxtamambrane region of all Trks and relaxes ligand preference (Clary and Reichardt 1994; Shelton et al., 1995). Both TrkB and TrkC can be expressed as truncated receptors that lack the tyrosine kinase domain which can modulate the activity of the full length receptor. In addition, neurotrophins bind a low affinity (~1 nM) receptor as well, p75, which is a member of the tumor necrosis factor receptor superfamily and has been linked to both cell death and plasticity. The p75 receptor complex is beyond the scope of this review but is the subject of several recent reviews (Dechant and Barde, 2002; Woo et al., 2005). BDNF is produced as a pro-peptide and is cleaved by multiple proteolytic pathways for both activity-dependent and constitutive secretion. Recent evidence suggests that p75 is the preferred receptor complex for pro-BDNF while Trks bind the mature neurotrophins (Reviewed by Lu, 2003).

Both full length and truncated TrkB receptors are expressed throughout the brain, however the full length receptor is mostly associated with neurons while the truncated receptor is also expressed on glia (Altar, et al., 1994; www.brainatlas.org). The highest levels of TrkB are in hippocampus, cerebellum and cortex but target areas for dopamine (DA) neurons also show diffuse TrkB expression. Diencephalic neurons, hypothalamic and midbrain monoaminergic neurons (raphe, locus coeruleus, Substantia Nigra (SN), Ventral Tegmental area (VTA)) express TrkB and respond to BDNF with increased neurotransmitter synthesis and increased survival (Akbarian et al., 2002; Altar et al., 1992; Altar et al., 1994; Altar et al., 1999; Baquet et al., 2005; Madhav et al., 2001). TrkB is expressed on the majority of mesencephalic DA neurons while a subpopulation also produce BDNF (Numan and Seroogy, 1999). BDNF, NT-4 and TrkB are often expressed in the same cell populations within the CNS but there are subtle differences between mRNA expression and localization by immunohistochemistry (Altar et al., 1997; Conner et al., 1997; Krause et al., 2008). For example, BDNF mRNA is expressed in the Basolateral Amygdala (BLA) and Lateral Amygdala (LA), CA3 and dentate gyrus with less expression in CA1. BDNF mRNA is generally below detection in the Nucleus Accumbens (NAc), an area intimately involved in addiction, but high in cortical areas and in dopaminergic neurons in the VTA and SN (Altar et al., 1997; Conner et al., 1997). Neurotrophins are often synthesized in one area and retrogradely or anterogradely transported. Areas such as the striatum, NAc, Central Amygdala (CeA) and Bed Nucleus of the Stria Terminalis (BNST) are devoid of BDNF mRNA detected by in situ hybridization but rich in BDNF-containing fibers detected by enzymatically amplified immunohistochemistry (Altar et al., 1997; Baquet et al., 2004; Conner et al., 1997; Guillin et al., 2001; Agassandian et al., 2006; Krause et al., 2008). Recent studies on alcoholism using commercially available antibodies coupled to immunogold, in situ PCR and RT-PCR have challenged these initial observations, however (for examples see Pandey et al., 2006 and McGough et al., 2004).

Deletion of TrkB developmentally is lethal shortly after birth, probably because the animals fail to feed (reviewed by Klein, 1994). BDNF deletion produces a more subtle phenotype with bobbing, head turning and spinning by 2 weeks of age (Ernfors et al., 1994; Jones et al., 1994). There is near complete loss of neurons in the vestibular ganglia of homozygotes and partial, progressive loss of these neurons in heterozygotes which likely contributes to this phenotype. There is a loss of motor neurons in TrkB knock outs (Klein, 1994) but no loss with BDNF deletion (Ernfors et al., 1994; Jones et al., 1994). Loss of BDNF also produces a selective depletion of Neuropeptide Y (NPY) and parvalbumin expression in the cortex with no change in the total number of interneurons, implicating BDNF in GABAergic interneuron function and peptide expression (Jones et al., 1994). Interestingly, NT-4 deficient mice show no gross neurological abnormalities (Conover et al., 1995) but, like BDNF deficient mice, show deficiencies in gustatory papillae and taste bud formation (Liebl et al., 1999). Impaired development of gustatory systems may in turn regulate appetitive behaviors in BNDF deficient mice (reviewed by Krimm, 2007). The more subtle phenotype of the BDNF^+/− mice is discussed later as it relates to anxiety, depression and ethanol consumption.

Because of the severe developmental phenotype of the complete knock out, recent work has employed region-specific gene deletion strategies, such as Lox-Cre recombinase under the control of forebrain or cortex-specific promoters (Gorski et al., 2003; Zorner et al., 2003; Baquet et al., 2004; Strand et al., 2007). Similarly, drug-inducible promoters have been constructed that allow the gene to be deleted selectively during developmental windows or in adults (Glorioso et al., 2006). Deletion of BDNF selectively during development, either by using conventional knock-out strategies (heterozygotes), the αCaMKII promoter driving Cre or a drug inducible strategy, leads to impaired hippocampal function, hyperactivity and hyperphagia (Duan et al. 2003; Glorioso et al., 2006; Kernie et al., 2000; Lyons et al., 1999; Rios et al., 2001; Zorner et al., 2003). Cortex-specific deletion of BDNF also causes degeneration of striatal medium spiny neurons (Baquet et al., 2004; Strand et al., 2007). Because no BDNF expression was detected in striatum when measured in reporter mice with LacZ inserted within the BDNF gene, Baquet and colleagues attributed the degeneration of medium spiny neurons to a loss of trophic support from cortex. A subsequent study from this group found striking similarities in gene expression patterns between cortex-specific delection of BDNF in transgenic mice and human and mouse models of Huntington ’s disease (Strand et al., 2007). The severe phenotype of the TrkB^−/− and BDNF^−/− mice indicates an essential role in for this neurotrophin in CNS development. Region-specific and drug-inducible knock out strategies in the maturing mouse reveal an essential role for BDNF in neuronal survival, learning, locomotor activity and appetite regulation.

Signaling events initiated by TrkB activation are well-established. Binding of BDNF to TrkB results in receptor autophosphorylation and recruitment of adaptor proteins which link to Ras family proteins and Extracellular-signal Regulated Kinases (ERK), as well as to activation of phospholipase C and phosphatidyl inositol 3-OH kinase (PI 3-Kinase) pathways (reviewed by Reichardt 2006). Upon ligand binding and phosphorylation, local signaling events occur but the BDNF-TrkB signaling complex is frequently internalized and transported from the cell periphery to the nucleus (reviewed by Ginty and Segal, 2002). ERK is activated following recruitment of the adaptor protein Shc and subsequent recruitment of Grb and Sos (or a similar nucleotide exchange factor) to a phosphotyrosine residue through SH2 domain interactions. Subsequent activation of a small molecular weight GTPase of the Ras family results in Raf activation and sequential phosphorylation of MEK and then ERK. Activation of the ERK pathway links to phosphorylation of numerous other transcription factors (discussed below) thereby changing gene transcription, but TrkB also couples to translation of plasticity-associated proteins such as Arc (Yin et al., 2002). In addition to the initial Ras-activated pathway, ERK can also be activated by Trks in a sustained fashion that involves a different SH2-binding partner to activate B-Raf through Rap-1 (Kao et al., 2001; York et al., 1998).

TrkB activation of PI-3 kinases results in the production of membrane phosphatidyl inositol 3,4,5 triphosphate and recruitment of phospholipid-dependent kinases through membrane association and subsequent phosphorylation of the pro-survival kinase Akt. Akt phosphorylates multiple substrates involved in cell survival and metabolism in an isozyme-specific fashion. Interestingly, activation of the pro-survival pathways appears to be independent of receptor internalization (Kuruvilla et al., 2000; MacInnis and Campenot, 2002).

Phospholipase C activation results in the production of 2 messengers, diacylglycerol and Ca²⁺. Membrane 4, 5 phosphatidyl inositol is cleaved to generate diaclyglycerol and inositol 1, 4, 5 triphosphate, leading to protein kinase C activation and the release of intracellular calcium. In addition to these canonical pathways, other TrkB effectors such as ERK5 have also been characterized (Cavanaugh et al., 2001). The truncated TrkB and TrkC isoforms do not activate these pathways but decrease full-length Trk receptor activation (Eide et al., 1996) and signal through G-proteins to increase intracellular calcium (Rose et al., 2003). TrkB.T1 increases filopodia formation (Hartmann et al., 2004) and truncated TrkC can signal through PSD-95 to increase membrane ruffling (Esteban et al., 2006).

Activation of signal transduction cascades by TrkB leads to expression and/or phosphorylation of multiple transcription factors. Spatial and temporal integration of these signals and induction of immediate early genes regulates gene transcription in response to BDNF in the context of concurrent synaptic activity. The best characterized of these factors are Cyclic AMP Response Element Binding protein (CREB), AP1 (Fos/Jun) and the serum response element binding/ternary complex transcription factor Elk-1. CREB is phosphorylated through multiple pathways, including calcium-dependent kinases, MAP kinase pathways and cAMP-dependent protein kinases (reviewed by Impey and Goodman, 2001). Elk-1 is phosphorylated through JNK in addition to ERKs 1, 2 and 5 in response to growth factor stimulation (reviewed by Turjanski et al., 2007). In addition, TrkB activates many of the same transcription factors as PKC cascades through distinct and overlapping pathways. Since these transcription factors are regulated by multiple overlapping pathways and neuronal activity through ionotropic receptors, the repertoir of transcription factors available to activate a given promoter is complex. Phosphorylation of transcription factors is frequently used in mapping studies but this too is a dynamic process. Regulation may not occur immediately after a stimulus through phosphorylation of a transcription factor but may require multiple pathways to be simultaneously active. Moreover, synthesis of a trans acting factor or second growth factor can modulate the later stages of gene expression.

Receptor signaling is also a dynamic process and this is particularly true for TrkB. NGF exposure increases TrkA in a positive loop (Holtzman et al., 1992) while BDNF activation of TrkB not only leads to desensitization but also reduces levels of TrkB protein but not mRNA (Frank et al., 1996; Frank et al., 1997). The ability of TrkB to signal in heterologous systems and in cerebellar granule cells is maximal at 10–100 pM BDNF concentrations and desensitizes at concentrations above 250 pM (Carter et al., 1995; Ip et al., 1993a,b). This concentration-response relationship is retained at the level of ERK activation by BDNF in cerebellar granule cells (Ohrtman et al., 2006). BDNF-supported survival in cerebellar granule cell and hippocampal neurons also plateaus and declines (Ip et al., 1993; Segal et al., 1992), indicating that both the pro-survival and plasticity signals desensitize. The pattern of neurotrophin signaling can also determine the phenotypic response, with sustained signaling leading to differentiation and intermittent signaling leading to proliferation (Vaudry et al., 2002; York et al., 1998). Therefore, it is necessary to consider not only concentration but the temporal signaling pattern generated by the neurotrophin receptor. This complex pattern of regulation indicates that the effect of the neurotrophin is dependent on concentration, receptor variant expressed, place of expression, duration of exposure, internalization and cross-regulation by neurotransmitters and hormones. The regulation of these pathways is the subject of several excellent reviews (Jeanneteau and Chao, 2006; Lee et al., 2002) and is only introduced here as it relates to studies on ethanol-BDNF interactions.

3. Is BDNF involved in ethanol-induced neurodegeneration?

BDNF was initially identified as a trophic factor; therefore it is not surprising that the first studies of ethanol-BDNF interactions were initiated to examine the role of BDNF in FASD and neurodegeneration with chronic exposure in adults. The effect of ethanol on BDNF expression in the VTA and NAc will be discussed below in detail in Section 6 as it relates to addiction. The hippocampus has been most extensively examined in terms of neurotoxicity because of the role of this structure in memory and the impaired cognitive function observed clinically in alcoholics (Harper and Matsumoto, 2005). The majority of the studies examining BDNF and/or TrkB in adult preparations have examined mRNA levels and are summarized in Table 1.

Table 1.

Summary of studies examining the effect of ethanol on TrkB or BDNF levels in mature animals and in vitro. Blood ethanol concentrations are noted if reported.

Model	Exposure	Protein/mRNA	Change	Reference
Human Patients	Self administration	Plasma BDNF Protein	↓ in alcoholic patients ↓↓with + family history	Joe et al., 2007
C57BL/6J DBA/J2 Mice	IP injection 2g/kg	BDNF mRNA Array RT-PCR	↑NAc DBA/J2 only ↔ PFCx	Kerns et al., 2005
C57BL/6J Mice	Vapor inhalation 2 × 64 hrs, 1 week apart 150–200 mg/dL	BDNF mRNA Array RT-PCR Western blot	↓PFC mRNA 0 HPE ↓protein 0 and 8 HPE in PFC ↔HP, vStr	Melendez et al., 2006
P and NP rats	No exposure	BDNF ELISA normalized to protein	↓ in NAc of P rats	Yan et al., 2005
PKCγ ko mice	Liquid diet 11 days 213 mg/dL	BDNF array BDNF RT-PCR	↓ in KO by array ↔ by PCR cerebellum	Bowers et al., 2006
Rat	Liquid diet 19 weeks	BDNF mRNA RT-PCR	↔ hippocampus	Okamoto et al., 2006
Rat	Liquid diet 28 weeks CET 24 hr WD 150–175 mg/dL	TrkB mRNA Northern blot	↑ Cortex	Baek et al., 1996
Rat	Liquid diet 15 days 101 mg/dL	BDNF Western blot	↔ on diet ↓ at 24 hr WD Cortex	Pandey et al., 1999
Rat	Liquid diet 28 weeks CET 24 hr WD 150–175 mg/dL	BDNF mRNA Northern blot	↓ in HP	MacLennan et al., 1995
Rat	Liquid diet 3 days/week 6–24 weeks	IHC stereology	↑ number of BDNF+ cells ↑ BDNF/cell Somatosensory cortex (layer V)	Bruns and Miller 2007
Rat	Liquid diet 8–24 weeks	BDNF ELISA Normalized to protein	↑ parietal Cx 24 weeks ↑ basal FB 8 and 24 weeks ↑ septum 24 weeks ↓ HP at 8 weeks only	Miller M. and Mooney 2004
Rat	Liquid diet Chronic Episodic exposure 6–24 weeks 137 mg/dL	BDNF ELISA Normalized to protein	↑ parietal Cx ↔ Enterhinal Cx ↑ septum ↑ HP early	Miller M. 2004
Rat	Liquid diet 28 weeks 150–175 mg/dL	BDNF mRNA TrkB mRNA RT-PCR	↔HP ↔ basal FB	Miller R. et al., 2002
Rat	Liquid diet 6 weeks 142 mg/dL	TrkB mRNA BDNF mRNA	↔	Zhang et al., 2000
Rat	Vapor inhalation 4 weeks 94–103 mg/dL	BDNF mRNA RT-PCR ISHH TrkB mRNA RPA	↓ during exposure HP (CA1 and DG) and SON ↑ CA3, DG, SON 12 hr WD ↑ SON, HP TrkB mRNA	Tapia-Arancibia et al., 2001
C57BL/6 mice	2 bottle choice	BDNF mRNA RT-PCR	↑ DStr ↔ PFC, HP	McGough et al., 2004
C57BL/6 mice	IP injection 2g/kg	BDNF mRNA RT-PCR	↑ DStr, HP Not detected NAc
Hippocampal neurons	10–100 mM	BDNF mRNA RT-PCR	↑ 0.5–2 hr ↓ 24–48 hr
Striatal slices	100 mM acute	BDNF mRNA RT-PCR	↑
SH-SY5Y Cells	100 mM	BDNF ELISA	↓ secretion in differentiated cells	Sakai et al., 2005
Cortical neurons	200 mg/dL 48 hr	TrkB Western blot	↔	Seabold et al., 1998
Hippocampal explant cultures	100 mM 4 hr	BDNF mRNA RT-PCR	↓ EtOH combined with TNF, glutamate and H202	Zou and Crews 2006
Cerebellar granule cells	400–1,600 mg/mL	BDNF ELISA	↓ secretion	Heaton et al., 2004
Cerebellar granule cells	100 mM	BDNF ELISA	↓ secretion	Bhave et al., 1999

Abbreviations: DG, dentate gyrus; Cx, cortex; HP, hippocampus; ISHH, in situ hybridization histochemistry; NAc, nucleus accumbens; PFC, prefrontal cortex; RT-PCR, quantitative reverse transcription polymerase chain reaction; SON, supraoptic nucleus; Str, striatum, TNF, tumor necrosis factor; vStr, ventral striatum; WD, withdrawal.

Some of these studies indicate that ethanol does not change TrkB mRNA levels in the adult (Miller et al., 2002; Zhang et al., 2000) while others report up-regulation in vivo during withdrawal (Baek et al., 1996; Tapia-Arancibia et al., 2001). This variability is not surprising since the receptor is not significantly regulated at the level of mRNA expression but shows significant down-regulation at the protein level with chronic BDNF treatment (Frank et al., 1996; Frank et al., 1997); one would therefore expect reciprocal regulation of BDNF and TrkB. The data on BDNF protein and mRNA levels are also inconsistent, with some studies showing no change in hippocampal BDNF levels (Miller et al., 2002; Okamoto et al., 2006), while others show a decrease in BDNF (MacLennan et al. 1995, Tapia-Arancibia et al., 2001). Surprisingly, two of the studies showing disparate effects in the same brain region are from the same laboratory. The authors attribute this discrepancy to the time of analysis. In the MacLennan study, the animals were withdrawn from ethanol while animals were still on drug/diet when sacrificed in the Miller study. This time discrepancy is bridged by Tapia-Arancibia and colleagues who found a decrease in BDNF mRNA while the animals were on drug/diet and an increase after withdrawal.

Zou and Crews (2006) reported a decrease in hippocampal BDNF mRNA in hippocampal explant cultures exposed acutely to ethanol under conditions of oxidative stress combined with tumor necrosis factor treatment. This was accompanied by an increase in NF-κB and a decrease in CREB DNA binding activity. This is in contrast to the observations of McGough and colleagues (2004) who reported an increase in BDNF mRNA that was mediated through Receptor for Activated C-Kinase 1 (RACK1) in hippocampal pyramidal cells in culture, but noted a decrease in BDNF with long-term treatment. As discussed by McGough and colleagues (2004), this may represent a homeostatic pathway that negatively regulates ethanol consumption. These data are complementary to the observation that intracerebroventricular administration of BDNF maintains ethanol tolerance in C57/BL6J mice (Szabo and Hoffman, 1995). Therefore, BDNF induction may regulate pharmacodynamic adaptations to ethanol exposure.

The disparate observations may reflect differences between homeostatic changes (or pharmacodynamic tolerance) that occurs acutely and the pathological changes that lead to neurodegeneration with chronic exposure. The majority of these studies report a decrease in BDNF mRNA in the hippocampus with long-term exposure, suggesting this decrease may play a role in, or is the consequence of, neurodegeneration. However, there are still temporal and methodological issues to be resolved. Long-term ethanol exposure causes hippocampal atrophy in rats (Walker et al., 1980), therefore observations made after 28 weeks of chronic ethanol exposure may occur as a result of a decrease in dendritic complexity and do not address the mechanisms that produced the degeneration. Short term exposure may cause the system to compensate in an attempt at homeostasis or repair (i.e. McGough et al., 2004; Miller 2004) that eventually fails with longer exposures (McGough et al., 2004; Miller and Mooney, 2004;) whereas withdrawal may cause hyperactivity due to upregulation after signal inhibition (Tapia-Arancibia et al., 2001). These variable changes may occur secondarily to an inability of the receptors to transduce signal in the presence of ethanol (See Section 5). However, these studies primarily examined mRNA, therefore the changes reported may not reflect differences in secreted protein or in its activity in vivo.

Recent advances in genomics, proteomics, in situ mapping and kinase activation screening will be useful in resolving many of these differential effects. One of the most pressing questions to be addressed is the promoters that may be differentially regulated during ethanol exposure and withdrawal. For example, calcium-dependent activation of a promoter that produces BDNF mRNA and targets to dendrites would be expected to have a different functional effect than an mRNA species that is translated in the soma. Knowledge of the promoter being regulated during these various states of intoxication and stress/anxiety may also provide clues to the mechanisms and neurotransmitter systems involved. A detailed time-course of the effects of ethanol on signal transduction pathways in a high-throughput assay combined with protein expression and mRNA could be used to construct a temporal picture of changes in signaling that result in tolerance, compensatory regulation and subsequent neurodegeneration.

4. Is BDNF involved in FASD?

Neurotrophins regulate proliferation, migration and differentiation in the developing brain and therefore are likely to be involved in the toxicity of ethanol during brain development. Rodents exposed to ethanol in utero or during the early postnatal period (a period roughly equivalent to the third trimester of human development) display microcephaly, hyperactivity, reduced cerebellar Purkinje cells, ataxia and learning deficits, which are also hallmarks of the human disorder (reviewed by Berman and Hannigan, 2000; Burn et al., 2003). The striking similarities between FASD models and BDNF haplodeficient mice suggest that the two phenotypes may be developmentally linked. There is considerable evidence that some of the teratogenic consequences of developmental ethanol exposure are mediated by inhibition of neurotrophin expression and function in a region-specific and time-dependent manner. This may contribute to apoptosis and impaired synaptogenesis, which are characteristic of FASD (reviewed by Olney 2004). The majority of the studies examining the effect of ethanol on BDNF and its associated signaling pathways during development have been performed in neonatal animals. This is likely because the regions where BDNF is expressed in the greatest amounts are late embryonic or postnatally developing structures. A variety of exposure paradigms have been employed and are summarized in Table 2.

Table 2.

Summary of experiments examining the effect of ethanol on BDNF or TrkB levels during development. Blood ethanol concentrations are noted if reported.

Model/region	Age and exposure	Protein/mRNA	Change	Reference
Neonatal rat SCN HP	artificial rearing 4.5g/kg PN4–9 BAC 333 mg/dL	BDNF ELISA @5–6 mos Normalized to protein	↓ Suprachiasmic nucleus ↓ HP	Allen et al., 2004
Prenatal rat Olfactory bulb	Intubation of dam 6g/kg/day 380 mg/dL	BDNF mRNA Southern blot	↓ at birth and PN10 Olfactory bulb	Maier et al., 1999
Neonatal rat	Gavage 5g/kg PN 5–8	BDNF mRNA	↓ Cortex	Fattori et al., 2008
Neonatal rat Cerebellum 0–24 hr	PN4, PN7 Vapor 2 hr 45 min Peak BAC ~290 mg/dL	BDNF ELISA Normalized to wet weight	↑ PN4 0 HPE ↔ 2–12 HPE ↑ PN7 2 HPE ↓ PN7 12 HPE	Heaton et al., 2003a
Neonatal rat Cerebellum	Vapor inhalation 2.5 hr PN4–5, PN7–8 PN4–10 Peak BAC ~300 mg/dL	BDNF ELISA Normalized to wet weight	↔ cerebellum with any exposure paradigm 1.5 HPE	Heaton et al., 1999
Neonatal rat Cortex 0–24 hrs	PN7, PN21 Vapor 2 hr 45 min Peak BAC ~290 mg/dL	BDNF ELISA Normalized to wet weight	↑ 0 and 12 HPE ↓ at 2 HPE ↔ PN21	Heaton et al., 2003c
Neonatal rat Striatum 0–24 hr	PN3 PN14 Vapor 2 hr 45 min ~290 mg/dL	BDNF ELISA Normalized to wet weight	↑ only 24 hr WD PN3 ↔ PN14	Heaton et al., 2003b
Prenatal/neonatal multiple regions	Vapor inhalation dam or PN4–10 2.5 hrs/day 161 mg/dL (dam) ~300 mg/dL (neonatal)	BDNF Normalized to wet weight	↔ HP, Septum, Cx/Str or CB at PN1 (prenatal) ↑ Increase Cx/Str,HP PN10 (neonatal) ↔ at PN21	Heaton et al., 2000
Prenatal Multiple regions	Liquid diet 161 mg/dL	TrkB protein Normalized to cyclophilin	TrkB p95 ↓Septum (female, PN 1, PN 10) ↑Cx (male & female) ↓ HP (male PN1) TrkB p145 ↑Cx, CB (female PN1) ↓Septum (female, PN1)	Moore et al., 2004a
Neonatal Multiple regions	Vapor Inhalation PN4–10 266 mg/dL	TrkB protein Normalized to cyclophilin	TrkB p95 ↓Cx (male & female) TrkB p145 ↓HP (female PN10) ↑Septum (male & female) No effect at PN21 ↔CB	Moore et al., 2004b
Prenatal Multiple regions	Intubation of dam 1 and 3 g/kg/day Examined at PN7	BDNF ELISA, normalized to protein BDNF mRNA TrkB pTrkB	↓ HP, Cx, NS decrease in Str (protein) ↓mRNA ↔cerebellum ↔TrkB ↓ TrkB phosphorylation	Feng et al., 2005
Neonatal rat cortex	Intubation of dam during gestation and lactation ~107 mg/dL	BDNF ELISA normalized to protein TrkB mRNA RPA	↓ in cortex PN5–21 ↑ full length and truncated TrkB	Climent et al., 2002
Neonatal rat	PN2–7, in bins intubation 300 mg/dL	TrkB TrkBT1 IHC	↓ on Purkinje cells at all ages	Light et al., 2002
Neonatal rat cerebellum	PN2–3 Analyzed PN4 intubation 300 mg/dL	TrkB ECD and TK TrkB-T2 mRNA BDNF RT-PCR	↓ TrkB.T2 and Total ECD ↔ TrkB-TK or TrkB.T1 ↓ BDNF	Light et al., 2001
Neonatal rat Cerebellum	Time course PN4, PN9 1.5–6 g/kg intubation 135–547 mg/dL	BDNF mRNA TrkB mRNA	dose-dependent ↓ 1–8 HPE PN4 ↔ PN9	Ge et al., 2004

Abbreviations: CB, cerebellum; Cx, cortex; ECD, extracellular domain; HP, hippocampus; HPE, hours post exposure; IHC, immunohistochemistry; PN, postnatal age; RPA, ribonuclease protection assay; SCN, suprachiasmic nucleus; Str, striatum; WD, withdrawal.

The ontogeny of specific neuronal populations in the cerebellum, their migration, synaptogenesis and the factors modulating these processes have been well-characterized (reviewed by Sotelo et al., 2004) making the developing cerebellum an ideal model for examining alcohol-induced teratogenesis. Cerebellar Purkinje cells are particularly vulnerable to ethanol during the first postnatal week but resistant to ethanol during the second postnatal week (West, 1993). The majority of studies examining BDNF levels, TrkB, or BDNF signaling have suggested that ethanol inhibits the neurotrophic activity of BDNF in the cerebellum (Table 2) and this may contribute to cerebellar Purkinje cell loss. Ge and colleagues (2004) performed a detailed dose-response, time-course analysis and found a rapid decrease in BDNF and TrkB mRNA in the cerebellum after exposure on PN4 but not PN9. In contrast, Heaton and colleagues (1999) found no differences in BDNF protein levels in cerebellum with numerous exposure paradigms that span this age but subsequently reported an increase immediately after exposure on PN4 that normalized by 2 hours (Heaton et al., 2003a). Exposure on PN7 increased BDNF levels at 2 hours and decreased levels at 12 hours in this study.

Light and colleagues used immunohistochemistry to detect TrkB specifically on cerebellar Purkinje cells during the vulnerable first postnatal week. Ethanol decreased TrkB immunoreactivity at all ages examined. Furthermore, they also observed a decrease in mRNA for TrkB and TrkB-T1 (Light et al., 2002). These data are particularly compelling since TrkB was detected and selectively decreased directly on the vulnerable cell population, suggesting that TrkB down-regulation and decreased secretion of BDNF by granule cells (Bhave et al., 1999; Heaton et al., 2004) are a prelude to apoptosis.

Many studies have examined changes in tissue homogenates but this represents total cerebellar protein or mRNA and Purkinje cells represent a small proportion of the total number of cerebellar cells (approximately 5%). In situ hybridization studies do not indicate that Purkinje cells contain more BDNF or TrkB mRNA than other cells and their large arbors are still developing and would therefore contribute less to the total protein or mRNA signal at earlier time-points (i.e. vulnerable vs. resistant periods). In addition, granule cells express both BDNF and TrkB mRNA and comprise the majority of cerebellar neurons thereby complicating the analysis of Purkinje cell proteins or mRNA in total cerebellar homogenates.

BDNF and TrkB expression have also been examined in cortex, striatum and hippocampus in developmental exposure models (Table 2). Several lines of evidence supported a role for decreased BDNF activity in the cortex in the development of FASD. Fetal exposure decreases cortical BDNF and increases TrkB/TrkB-T1 ratios as neonates (Climent et al., 2002). Feng and colleagues (2005) reported similar findings, showing decreased BDNF expression in hippocampus and cortex with gestational exposure, but no change in TrkB. Moore and colleagues (2004) reported a decrease in hippocampal TrkB in males and an increase in cortical TrkB in females at PN1 with gestational exposure. Fattori and colleagues (2008) examined BDNF levels in cortex after neonatal exposure and, as with the majority of studies, also observed a decrease in BDNF mRNA. These data are in contrast with the results of Heaton and colleagues (2003c), who found bidirectional changes in BDNF cortex after exposure on PN7. This group also found no differences in corticostriatal BDNF levels with prenatal exposure but found an increase after neonatal exposure (Heaton et al., 2000b). The discrepancy in these data is not likely due to timing of exposure since conflicting results were also obtained with gestational exposure, but may be due age of the animals at analysis. Heatonet al. (2000) measured levels at PN1, while Feng et al. (2005) examined BDNF levels at PN7 and Climent et al. (2002) measured levels at PN5, PN14, PN21 and PN35 after gestational exposure.

Similar discrepancies also exist for the effect in hippocampus, with Heaton and colleagues (2000) reporting a transient increase in hippocampal BDNF at PN10 with neonatal exposure, while Feng and colleagues (2005) report a decrease in BDNF with gestational exposure. This brings into question whether BDNF levels change as a result of ethanol exposure or change as a result of neuronal atrophy, impaired synaptogenesis and accelerated apoptosis. More experiments examining acute effects on signaling during exposure and compensatory regulation over time after the insult in specific cell types are required (i.e. Ge et al., 2004; Light et al., 2002). It is also unclear whether these changes result in a decrease in activity-dependent BDNF release or represent differences in dendritic complexity. Examination of BDNF signaling (discussed below) suggests that ethanol causes a decrease in the ability of TrkB to transduce signal and desensitize, which may lead to compensatory regulation of both BDNF and TrkB in vivo. The data from Feng et al. (2005) and Climent et al. (2002) suggest that the end result is a decrease in BDNF during post-exposure development that may be secondary to impaired synaptogenesis and neuronal apoptosis.

5. Effects of EtOH on BDNF signaling

The literature on BDNF and TrkB expression levels before, during, and after ethanol exposure is extensive; therefore it is surprising that relatively few studies have focused on the direct effect of ethanol on BDNF signaling. Animals exposed prenatally to ethanol have decreased TrkB phosphorylation at 1 week of age that is probably secondary to a decrease in BDNF (Climent et al., 2002; Feng et al., 2005; Fattori et al., 2008). Chronic ethanol exposure increases ERK activity in cultured cortical neurons and pheochromocytoma cells (Kalluri and Ticku, 2003; Roivainen et al., 1995) while acute ethanol administration decreases activation of ERK (Chandler and Sutton, 2005; Davis et al., 1999; Han et al., 2006; Kalluri and Ticku, 2002; Kalluri and Ticku 2003; Tsuji et al., 2003; Fattori et al., 2008; Ohrtman et al., 2006, however, see Acquaah-Mensah et al., 2001).

Ethanol acutely inhibits TrkB signaling to ERK in cerebellar granule cells (Li et al., 2004; Ohrtman et al., 2006) and this is not due to a decrease in TrkB phosphorylation (Li et al., 2004), suggesting an intracellular cite of action. Hippocampal slices prepared from neonatal rat pups stimulated with BDNF also show reduced nuclear translocation of phospho-ERK in CA1 hippocampal pyramidal cells (Davis et al., 1999). A thorough pharmacological analysis TrkB-ethanol interactions in cerebellar granule cells revealed that inhibition of BDNF-stimulated ERK phosphorylation by ethanol is dependent on BDNF concentration and independent of NMDA receptors (Ohrtman et al., 2006). Low concentrations of ethanol (25 mM) inhibit ERK activation when BDNF is present at concentrations in the linear range of the high affinity concentration-response curve (less than or equal to 5 ng/mL; 185 pM). However, high concentrations of ethanol (100 mM) block the desensitization of the ERK response at high BDNF concentrations (>10 ng/mL). This may occur through inhibition of NMDA receptors by ethanol since APV blocks desensitization of ERK stimulated by TrkB (Davis and Ohrtman, unpublished observations). The mechanism for inhibition of BDNF-stimulated ERK activation by ethanol remains elusive but recent evidence suggests that it may involve the signaling intermediate Raf and Raf kinase inhibitor protein (Hellmann et al., 2006). The decrease in ERK activity also correlates with an increase in protein kinase A activity and a decrease in calcium-sensitive protein kinase C activity (Davis et al., 1999). Both of these pathways can cross-regulate ERK (reviewed by Liebmann 2001), therefore inhibition of TrkB signaling may also ultimately involve pathways previously known to be sensitive to ethanol.

Ethanol decreases Akt phosphorylation in the developing, but not adult, brain (Chandler and Sutton, 2005), suggesting that Akt may be involved in the selective vulnerability of the developing brain to ethanol. BDNF protects neurons from ethanol-induced cell death in vitro (Heaton et al., 2000a; Bonthius et al., 2003). NMDA also increases granule cell survival and protects from ethanol toxicity through up-regulation of BDNF (Bhave et al., 1999; Marini et al., 1998). The biochemical mechanisms and isozymes underlying this neuroprotection are less clear but likely involve PI 3-kinase. To address this question, Bhave and colleagues (1999) used pharmacological inhibitors of PI 3-kinase (100 nM wortmannin or 10 μm LY294002) to block the trophic effects of NMDA and its induction of BDNF. Inhibition of PI 3-kinase was toxic to granule cells in these experiments; however confirmatory western blot analysis further implicated PI 3-kinase in the neuroprotection. In a similar study, Heaton and colleagues (2000a) examined the mechanism of BDNF neuroprotection in cerebellar granule cell cultures and also suggested a role for PI 3-kinase. This study used wortmannin at 10–100 μm to reverse the neuroprotective effects of BDNF. The data for wortmannin alone were not presented but the authors stated that it was without effect. This is in contrast to Bhave and colleagues but may be due to the poor stability of wortmannin since the concentrations used are much higher than what is normally required to inhibit PI 3-kinase (D’Mello et al., 1997; Miller et al., 1997; Bhave et al., 1999). Isozyme-selective PI 3-kinase inhibitors have recently been developed (Redaelli et al., 2006) that have reduced global toxicity and could be used to extend these studies. Given the role of PI 3-kinase in neuronal survival, BDNF is most likely mediating neuroprotection from ethanol toxicity through PI 3-kinase in cultured cerebellar granule cells. The situation in vivo is quite different, however, since the effect in the absence of neurotrophic factor supplementation may be a decrease in the ability of endogenous BDNF to increase survival during development secondary to a decrease in BDNF levels and the reduced ability of BDNF to transduce survival signals.

c-Jun N-terminal Kinase (JNK) is a MAP Kinase family protein that is activated under conditions of cellular stress (reviewed by Weston and Davis, 2007). This is isozyme-specific and JNKs can also participate in neuronal differentiation through phosphorylation of cytoskeletal proteins involved in neuronal migration (Reiner et al., 2004; Hirai et al., 2006) and axon formation (Oliva et al., 2006). p46 JNK is also activated by BDNF (Davis and Hassoun, unpublished observations). Activation of the JNK pathway by ethanol has been suggested to be involved in apoptosis (Han et al., 2006; Heaton et al., 2003). However, subsequent studies did not find a correlation between JNK activation and cell death (J.W. Olney, personal communication). Furthermore, BDNF stimulation of AP1 DNA binding activity is inhibited by ethanol in cerebellar granule cells (Li et al., 2004), suggesting reduced genomic signaling through Fos/Jun. Activation of the NF-κB pathway has also been observed after EtOH exposure (Zou and Crews, 2006), and activation of this pathway correlates with cell death under conditions of cellular stress.

Other pathways such as those involving PLC and PKC that couple to TrkB are also regulated by G protein-coupled receptors, stress, integrins and calcium. Chronic ethanol has previously been shown to specifically decrease PLC activity and PLCβ1 expression in vivo and in neuroblastoma cells (Alling et al., 1993; Katsura et al., 1994; Pandey, 1996) and prenatal ethanol exposure decreases PLCβ1 activity in the hippocampus and cortex (Allan et al., 1997). These changes in PLC may represent changes in G-protein coupled signaling since changes in PKCγ, which also couples to TrkB (Widmer et al., 1993), were not observed (Pandey et al., 1996). However, direct stimulation of PKCγ by epidermal growth factor is inhibited by ethanol (Thurston and Shukla, 1992). Many PKC isozymes are regulated by diacylglycerol that is generated by PLC activity and EtOH is known to modulate PKC activity but this depends on the isozyme of PKC examined and the duration of exposure (reviewed by Slater and Stubbs, 1999; Newton and Ron, 2007).

BDNF and multiple G protein-coupled and ionotropic receptors signal to CREB to modulate gene transcription and late phase synaptic plasticity. CREB is also a target of ethanol and multiple studies implicate CREB in ethanol-mediated changes in gene expression. A detailed discussion of region-specific regulation of CREB as it relates to behavioral phenotypes is presented in Section 6 but is introduced here in the context of cellular signaling. An in silico analysis of promoter elements in ethanol responsive genes has identified CREB as a potential mediator of ethanol effects on gene expression (Uddin and Singh 2007). Using mice expressing LacZ under a CRE, an acute exposure to ethanol increased CRE-mediated gene transcription (Asyyed et al., 2007) in NAc, PFC, Septum, BNST, Lateral Habenula, BLA, PVN, hippocampus, VTA and SN, in addition to hypothalamic nuclei. Pandey and colleagues (2004) identified CREB in the CeA and MeA as a potential mediator of ethanol consumption since reduced pCREB was associated with anxiety during withdrawal, and rats that prefer ethanol have lower levels of pCREB in this structure (Pandey et al., 2005). These data are similar to the decrease in pCREB observed in the CeA and MeA after acute amphetamine exposure (McPherson et al., 2007) and may therefore be related to increased catecholamine levels during withdrawal. However, in contrast to Asyyed and colleagues (2007), Pandey and colleagues (2005) reported an increase in pCREB phosphorylation in CeA and MeA, but not BLA in ethanol preferring rats exposed to ethanol by either IP injection or self administration. Chandler and Sutton (2005) reported a decrease in CREB phosphorylation with acute ethanol exposure that was similar to the decreased observed in ERK activation in hippocampus and cortex. Again, this is in contrast to previous observations reporting a decrease in cortical pCREB and BDNF during withdrawal but not with ethanol exposure (Pandey et al., 1999). The different exposure paradigms (IP injection vs. long-term self administration and withdrawal), rodent strain or cell-type selectivity may explain these differences. Acute ethanol can activate adenylate cyclase, which can activate CREB through PKA (Asher et al., 2002; Constantinescu et al., 2002). Chronic exposure leads to compensatory up-regulation of NMDA receptors, with withdrawal producing Ca²⁺-dependent hyper-excitibility (Hoffman and Tabakoff, 1994). This may lead to increased CREB-mediated gene transcription through CaM kinases in certain regions. Therefore, CREB regulation, as with each of these pathways, may be mediated by multiple ethanol-sensitive pathways at different times during and after ethanol exposure. BDNF is just one factor mediating CREB phosphorylation, therefore any changes in pCREB must be carefully considered in the context of the circuit showing activation, the cell type and other ethanol-mediated changes in cell signaling induced by perturbation of ionotropic and G protein-coupled receptor signaling.

BDNF modulates synaptic plasticity in preparations made from mature brain and this has been extensively examined in the modulation of hippocampal LTP (Korte et al., 1995; reviewed by Nagappan and Lu, 2005; Soule et al., 2006). Ethanol inhibits NMDA receptors and NMDA mediated long-term plasticity (Blitzer et al., 1990; Lovinger et al., 1989; Lovinger et al., 1990). BDNF potentiation of NMDA receptor function is inhibited by EtOH and this occurrs at concentrations as low as 10 mM, which have relatively small direct effects on the current itself (Kolb et al., 2005). These data suggest that BDNF-mediated regulation of NMDA receptor currents may be extrememly sensitive to ethanol.

Regional differences in activation vs. inhibition of BDNF-regulated signaling pathways may depend on the local concentration of BDNF, differential sensitivity of the neuronal poulation to ethanol, developmental state of the cell, duration of ethanol exposure and on compensatory regulation of the pathways (homeostasis and pharmacodynamic tolerance). Disinhibition of a circuit (i.e. inhibition of interneuron function) may also result in seemingly paradoxical activation of BDNF signaling pathways in specific neuronal populations by ethanol in vivo. In addition, the effect of ethanol on pro-survival pathways may depend on the differentiation state of the cells examined. Developing neurons are more sensitive to inhibition of Akt by ethanol (Chandler and Sutton, 2005), certain transformed cells show an increase in ERK activation in response to chronic ethanol exposure (Roivainen et al., 1995; Ku et al., 2007), while terminally differentiated cells, for the most part, show inhibition of ERK after acute exposure (Davis et al., 1999; Chandler and Sutton, 2005; Li et al., 2004; Ohrtman et al., 2006). Detailed temporal, global immunohistochemical mapping studies of activated protein kinases, neuronal activity and the determination of neuronal identities will be essential in resolving temporal and regional differences. In addition, selective BDNF deletion using the Lox-Cre system could be used to elucidate the role of BDNF in ethanol-induced changes in neuronal signaling in discrete neuronal populations.

6. BDNF in animal models of anxiety, depression and addiction

Addiction and memory share common biochemical substrates and produce long lasting changes in brain circuitry that lead to compulsive drug seeking (reviewed by Hyman et al., 2006). The dopamine hypothesis of addiction-reward posits that increased dopamine release from VTA neurons onto neurons in the NAc underlies the rewarding properties of most drugs of abuse, including ethanol (reviewed by Koob, 2003). This system has been termed the mesocorticolimbic dopamine pathway and includes prefrontal cortex and the extended amygdala. According to this model, repeated exposure to a drug produces synaptic plasticity (LTP and LTD) in this circuit (and others) and changes the anxiety set point resulting ultimately in craving and addiction. The allostatic model for ethanol addiction posits that drug exposure resets the activity of neurons in these reward pathways. Reward and withdrawal-induced anxiety increases the desire to consume ethanol (Koob, 2003). Consumption leads to an increase in dopamine release from neurons located in the VTA onto neurons in the NAc, leading to “reward”. This is followed by cycles of withdrawal-induced anxiety, decreased CREB activation and decreased NPY production in the amygdala and continued drinking (Pandey, 2003). Rewarding properties are reduced with subsequent exposures and more drug must be consumed to feel “normal”. Theories of ethanol reward and addiction implicate the integration of these circuits in ethanol seeking and relapse.

This reward circuitry has recently been extended to include the lateral habenula, which inhibits DA neurons in VTA to block signaling the absence of reward (Matsumoto and Hikosaka, 2007). Ethanol administered by IP injection acutely increases CREB/CRE mediated gene transcription in the lateral habenula, suggesting that this region may also be involved in ethanol intoxication and acute adaptation/tolerance (Asyyed et al., 2006) but how activity in this region changes with prolonged ethanol exposure and tolerance is unknown. In addition, the mesolimbic dopamine pathway regulates depressive-like behaviors in mice (Nestler and Carlezon, 2006). Projections from the VTA to the NAc are regulated by feedback from inhibitory projection neurons to the VTA, as well as excitatory projections to the NAc from the cortex (Nestler and Carlezon, 2006). These circuits are active in stress, anxiety, depression and addiction, therefore I will consider the effects of ethanol on BDNF pathways in the context of what is known about these brain regions in substance abuse, depressive and anxiety-like behaviors.

Multiple lines of evidence suggest that BDNF is involved in VTA-NAc mediated addiction, craving and withdrawal. Initial observations indicated that BDNF infusion directly into the VTA led to a down-regulation of total ERK with no change in ERK activity. This occluded any further enhancement of the pathway by morphine or cocaine, thereby blocking the rewarding properties of the drugs (Berhow et al., 1996). These data also suggest that the ERK pathway can be dysregulated by prolonged BDNF exposure. Amphetamine exposure increases BDNF in the BLA and within its projections to the CeA, striosomes, medial striatum and NAc (Meredith et al., 2002). BDNF potentiates psychostimulant-induced increases in DA both acutely and during sensitization (Altar et al., 1994; Pu et al., 2006; Horger et al., 1999). Conditioned place preference in response to psychostimulants is also reduced in BDNF haplodeficient mice (Hall et al., 2003). Amphetamine increases TrkB expression (Meredith and Steiner, 2006) and BDNF participates in the rewarding properties of psychostimulants by potentiating pre and post synaptic activity (Horger et al., 1999). BDNF also sensitizes midbrain DA neurons to potentiation during psychostimulant withdrawal (Pu et al., 2006). A single infusion of BDNF into the VTA can potentiate cocaine craving for a month (Lu et al., 2004). Taken together, these observations indicate that BDNF can modulate and potentiate reward pathways, but these become dysregulated with chronic exposure to pyschostimulants or infusion of BDNF itself.

In contrast, chronic stress and depression decrease BDNF levels in the hippocampus (Smith et al., 1995). Antidepressants increase BDNF and TrkB levels in hippocampus and it has been posited that this is a mechanism by which antidepressants exert their actions (Nibuya et al., 1995; reviewed by Duman and Monteggia 2006). These findings were extended with a forebrain-specific doxycycline inducible knock-out of BDNF. These mice show profound hippocampal dysfunction, attenuated antidepressant effects and hyperactivity if the gene is deleted late in embryonic development (Monteggia et al., 2004). There were gender differences as well, with males showing more hyperactivity and females showing more depressive-like behaviors (Monteggia et al., 2006). These data further implicate cortical and hippocampal sources of BDNF in antidepressant efficacy. In a related study (Zorner et al., 2003), TrkB^CaMKII-CRE mice were generated to screen for a role for BDNF in depressive-like behavior. This strategy relies on forebrain-specific expression of the Cre recombinase under the αCaMKII promoter, which is expressed after the major wave of cortical development. This strategy circumvents the developmental lethality of the deletion and allows for selective deletion of TrkB in the forebrain. These mice exhibited an amplified impulsive response to novelty and stereotyped hyperlocomotion, suggesting these mice may be a model for ADHD. It is interesting that both models point to a role for impaired forebrain BDNF signaling in the development of impulsivity and hyperactivity since these are characteristics of FASD with hyperactivity more prevalent in males. BDNF would be retained in the dopaminergic and absent in the corticostriatal terminals, possibly leading to an increase in dopaminergic transmission relative to glutamate from the cortex.

BDNF appears to be a key player in antidepressant efficacy, at least in hippocampus, but what about the systems that regulate anhedonia and depressive states? The VTA and projections from this area (i.e. ventral striatum and amygdala) are not only implicated in reward and anxiety but may also be involved in the eitiology of depression (Nestler and Carlezon, 2006). Surprisingly, direct delivery of BDNF into the VTA produced depressive-like behaviors in mice (Eisch et al., 2003). Furthermore, this was reversed by expression of TrkB.T1 in NAc using stereotaxic viral injections. This observation was recently expanded upon to examine the role of BDNF in another depressive-like behavior, social defeat stress (Berton et al., 2006). Mice exposed to a “bully” mouse (a male mouse of a larger strain) have increased mesolimbic BDNF levels and target activity in NAc as measured by c-Fos induction. Therefore, increased TrkB activation is associated with decreased social interactions and heightened accumbal induction of Fos after exposure to the “bully” mouse. Selective deletion of BDNF in the VTA using viral expression of the Cre-recombinase in floxed-BDNF transgenic mice or antidepressant treatment with fluoxetine reversed these effects. This study was recently extended, further implicating BDNF in the VTA and it’s projections in suceptiblity to social defeat stress (Krishnan et al., 2007).

In a complementary series of experiments, transgenic mice over-expressing BDNF under the αCaMKII promoter were generated to increase BDNF specifically in the forebrain (Govindarajan et al., 2006). These mice exhibited increased anxiety-like behavior and increased dendritic spines in BLA neurons but a decrease in depressive-like behavior in the Porsolt forced-swim test. Therefore, an increase in forebrain expression of BDNF increases anxiety but decreases depressive-like behaviors. These data indicate that precise activation of brain areas by BDNF is required for appropriate responses with induction of BDNF being correlated with antidepressant efficacy in the hippocampus but not in the VTA-NAc pathway. Furthermore, these studies suggest that depression, anxiety and locomotor activity are differentially regulated by BDNF, depending on the source.

Depression and anxiety often result from dysregulation of the serotonergic system (reviewed by Hariri and Holmes, 2006) and the role of BDNF in antidepressant efficacy has recently been reviewed in this journal (Kozisek et al., 2007) It has been proposed that modulation of the 5-HT system might be a viable therapy for alcoholism in a sub-set of patients (Johnson 2004; Wrase et al., 2006) and BDNF supports survival of serotonergic neurons (Altar, 1999; Madhav et al., 2001). Lyons and colleagues (1999) found that BDNF^+/− mice had abnormalities in the serotonergic system, with an age-related decline in 5HT and 5-HIAAA. They observed upregulation of 5-HT_1B, 5-HT_2A and 5-HT_2C in hypothalamus and cortex at 6–9 months of age. Hippocampal levels were slightly decreased, but only significantly for 5-HT_1C. This is accompanied by a reduction in dexfenfluramine induction of c-Fos in the cortex, increased aggression, hyperphagia and degeneration of serotonergic neurons. It was subsequently reported that 5-HT_1A receptor-stimulated GTP-γS binding was modestly attenuated in the medial and dorsal raphe nuclei, anterior cingulate cortex, lateral septum, and significantly decreased in the hippocampus of BDNF haplodeficient female mice (Hensler et al., 2003).

Unlike the studies discussed above, these BDNF^+/− mice did not display hyperphagia as measured by an increased preference for saccharine or quinine, but showed a significant preference for ethanol in a 2-bottle choice paradigm. These studies were extended in the same mouse strain (McGough et al., 2004) to suggest that induction of BDNF in the dorsal striatum and hippocampus negatively regulates ethanol consumption. Also using BDNF globally haplodeficient mice, this study replicated the observations of Hensler and colleagues with respect to increased ethanol consumption. In addition, BDNF^+/− mice show increased conditioned place preference to ethanol (2 g/kg) and increased locomotor activity in response to ethanol in this study. Unlike previous studies with this strain (Duan et al., 2003; Kernie et al., 2000) and with forebrain-specific deletion (Monteggia et al., 2004; Monteggia et al., 2006; Rios et al., 2001; Zorner et al., 2003), these mice did not differ in basal locomotor activity but showed enhanced activity in response to ethanol injection. This is in contrast to cocaine sensitivity, where BDNF^+/− mice show reduced locomotor effects and decreased conditioned place preference (Hall et al., 2003). The increased consumption of ethanol in haplodeficient mice could be reversed by systemic delivery of TAT-RACK1. Ethanol increases RACK1 nuclear translocation (Ron et al., 2000) and the TAT-RACK1 construct increases BDNF expression in striatal slices, in hippocampal cultures, and also after IP injection (McGough et al., 2004). These are intriguing studies because they implicate the dorsal striatum, an area involved in habit learning (reviewed by Gerdeman et al., 2003; Yin and Knowlton, 2006), in ethanol self administration.

Array analysis has identified BDNF as an ethanol-responsive gene in PFC and NAc, but there are also discrepancies in these observations. Melendez and colleagues (2006) exposed mice by vapor inhalation for 2 “binges” of 64 hours separated by 2 weeks. PFC, NAc and hippocampus were examined but differences in BDNF were only observed in the PFC. This group observed a decrease in BDNF mRNA on the array, which was confirmed by RT-PCR. Protein levels were also examined and found to be reduced in the PFC immediately after the second exposure and 8 hours later. In a similar study, Kerns and colleagues identified NAc BDNF as an ethanol responsive gene in DBA/2J mice using mRNA from animals exposed to acute EtOH by IP injection. They compared low ethanol drinking DBA/2J mice to high ethanol drinking C57BL/6J mice using microarrays and found an increase in BDNF mRNA in the NAc in DBA/2J mice only. These data suggest that BDNF induction in the NAc may be associated with ethanol aversion or dysphoria, similar to the depressive-like effect of intra-VTA administration of BDNF (Eisch et al., 2003) or induction by social defeat stress (Berton et al., 2006). The source of the BDNF mRNA is also a mystery since it is not generally detected in the NAc or at significant levels in striatum under normal conditions (Altar et al., 1997; Baquet et al., 2004; Conner et al., 1997; Zuccato et al., 2001), but may indicate strain differences that mediate preference and aversion. Given that these studies used extremely sensitive techniques, this mRNA may represent increased expression in a small population of neurons or mRNA present in glia, stem cells, platelets or blood cell progenitors. Axonal mRNA from cortex seems unlikely since BDNF protein is transported to striatum from cortex and midbrain structures (Altar et al., 1997; Conner et al., 1997; Gauthier et al., 2004; Strand et al., 2007) but remains a possible explanation. Another intriguing possibility is that the source of BDNF mRNA is the dopaminergic terminals, which are the source of the BDNF that increases D3 receptor expression during cocaine sensitization (Guillin et al., 2001). Alternatively, BDNF induction may signal a significant pathological reaction to a toxic insult, as has been shown to occur with cocaine exposure (Liu et al., 2006) and excitotoxic lesions (Rite et al., 2005).

The role of the extended amygdala in fear, anxiety, hypothalamic-pituitary adrenal axis regulation, appetite and sensation has been studied extensively in rodent models of anxiety-like behaviors and fear conditioning. The amygdala is divided into at least 13 nuclei with multiple subdivisions in rodents based on location and afferent-efferent projections (reviewed by Sah et al., 2003). These nuclei are grouped anatomically into cortical, basolateral and centromedial but extensively subdivided, with signals generally flowing laterally to medially through the nuclei with numerous intranuclear connections and collaterals. The amygdala integrates information from multiple sensory systems, the cortex and the hippocampal formation. Monoaminergic brain stem innervation, particularly in the CeA and BNST, modulates the activity of the circuit. The CeA is considered the primary output, with projections to the hypothalamus, ventral striatum, PAG and BNST which mediate arousal. The basolateral complex also sends reciprocal projections to prefrontal cortex, BNST, VTA, mediotemporal lobe memory structures and the ventral striatum. The precise function of these connections is not well understood but the output of this structure regulates fear, startle, anxiety, pain perception and autonomic function through hypothalamic projections.

One of many self-medication hypotheses for the development of alcoholism exploits the anxiolytic properties of ethanol in the amygdala. Ethanol increases GABAergic transmission through enhanced release of GABA in the CeA and BLA (Roberto et al., 2003a; Zhu and Lovinger, 2006) and can potentiate GABA_A receptor function, although this is controversial (reviewed by Lovinger and Homanics, 2007). In addition, a recent study indicates that there are 2 ethanol-sensitive populations in the BLA (Silberman et al., 2007). The first population is represented by local interneurons and responds to ethanol with enhanced release through a GABA_B-dependent mechanism. The second population, paracapsular neurons, responds to ethanol through norepinepherine receptors. These data suggest that synaptic transmission in BLA target areas would be decreased by ethanol.

Chronic ethanol also increases the sensitivity of CeA neurons to inhibition of NMDA receptors by ethanol (Roberto et al., 2004). However, the mechanism for this effect is still being debated, with Roberto and colleagues suggesting that this is due to changes in receptor expression. Others have shown no differences in receptor levels (Läck et al., 2005), instead implicating post-translational regulation of NMDA receptors in the CeA in neuroadaptation to chronic ethanol and possibly withdrawal-induced anxiety. The BLA has been extensively examined for roles in fear conditioning, addiction and anxiety (reviewed by Davis, 2006; Shekhar et al., 2005). BDNF in the BLA plays a role in Pavlovian fear conditioning (Rattiner et al., 2005) and consolidation of fear extinction (Chhatwal et al., 2006). Other studies have implicated the BNST in anxiety-like behaviors and afferent modulation of the acoustic startle reflex, with the CeA modulating fear conditioning (reviewed by Davis and Shi, 1999; Davis 2006).

Surprisingly, administration of a BDNFantisense oligodeoxynucleotides into the CeA or MeA, but not the BLA, increased anxiety and increased ethanol consumption in rats (Pandey et al., 2005). This was also reflected in reduced activation of ERK and CREB with oligodeoxynucleotide infusion. The increased ethanol consumption is recapitulated in CREB deficient mice (Pandey et al., 2004b), suggesting that reduced CREB activity and decreased BDNF and NPY levels in the CeA and MeA contribute to ethanol drinking. This is in contrast to observations in the hippocampus and cortex, reporting a decrease in pCREB and NPY with ethanol exposure and an increase in both pCREB and NPY in the dentate gyrus/hilus during withdrawal (Bison and Crews, 2003). Opiate withdrawal also induces anxiety, but this involves A1/A2 noradrenergic projections and increased CREB phosphorylation in the CeA and BNST (reviewed by Aston-Jones and Harris, 2004). The identity of the cells expressing BDNF is also of interest. The striatum and CeA, but not the MeA, share a common embryonic origin in the lateral ganglionic eminence (García-López et al., 2008). The majority of the cells in the CeA are of the medium spiny type and presumably GABAergic (McDonald, 1982). Therefore, increased activity in these neurons would presumably lead to decreased activity in target structures such as the brain stem, PVN and VTA but this has not been established.

As discussed above, these data are in contrast to anatomical studies showing no BDNF mRNA in CeA, but heavy staining of fibers and terminals in this area (Conner et al., 1997; Krause et al., 2008). Ethanol induces expression of BDNF mRNA in structures not identified in classical expression studies in normal animals, such as the CeA (Pandey et al., 2005) and dorsolateral striatum (McGough et al., 2004). McGough and colleagues used PCR amplification for detection but did not examine protein levels in these structures. Pandey and colleagues (2005) used in situ RT-PCR and immunogold to detect both mRNA and protein. The speckled distribution of the immunogold particles suggests that BDNF might be in terminals but this study did not specifically address this possibility and in situ PCR for BDNF clearly showed somatic localization but no immunoreactive fibers from the BLA projections, as have previously been described (for images of these projections, see Conner et al., 1997). In addition, ethanol-induced increases in glutamatergic function in the BLA do not affect BDNF expression in the BLA (B.A. McCool, personal communication), suggesting that the BLA is not the source of this change in BDNF. Recent high resolution immunodetection of BDNF protein in the CeA identified immunoreactivity in pericellular baskets surrounding the somata of CeA neurons (Krause et al., 2008). This study used a sheep anti-BDNF antibody that was extensively characterized by Western blot and antigen preabsorption. It is possible that the antibodies employed in the two studies recognize different epitopes present on the pro and mature forms of BDNF or have cross reactivity with other neurotrophins (as has been shown for the Santa Cruz anti-BDNF antibody SC 564, www.scbt.com). Pathological conditions such as excitotoxic lesions (Rite et al., 2005) and cocaine (Liu et al., 2006) can induce striatal BDNF mRNA, suggesting that BDNF induction in these areas may have significant biological consequences, but it is uncertain what this means in relation to addiction or alcoholism. It is also worth noting that NT-3 binding and TrkC expression are high in the striatum (Altar et al., 1994), yet there are few functional studies with this neurotrophin. Immuno-identification of specific ethanol-sensitive populations within the extended amygdala and striatum combined with anterograde tracing could resolve these neuroanatomical interactions and identify the active circuits before, during and after ethanol consumption. The use of colchicine (Altar et al., 1997) to allow BDNF to accumulate in the cell body of origin might also aid in mapping the circuit.

These disparate data between depression/anxiety and ethanol consumption might be explained by the regional and cellular differences in expression (global vs. forebrain principle neurons vs. amygdalar nuclei), disinhibition, or by regional differences in sensitivity of TrkB to activation and desensitization. Most ethanol studies to date have focused on the BLA, CeA and MeA. As discussed above, ethanol enhances GABAergic transmission in the BLA through both paracapsular interneurons and local interneurons (Zhu and Lovinger, 2006; Silberman et al., 2007). Neurons in the BLA are also hyper-excitible during withdrawal (Floyd et al., 2003; Läck et al., 2007) and CRF is increased with chronic ethanol treatment in the CeA (Läck et al., 2005). The BLA projects to the CeA and to the paracapsular intercalated cell masses, which are GABAergic and located at the medial border of the BLA (Royer et al., 1999). The intercalated cell masses and paracapsular neurons express high levels of D1 receptors and are also regulated by inputs from the prefrontal cortex and by catacholamines (Fuxe et al., 2003; Marowski et al., 2005). Ultimately, this provides feed forward inhibition of neurons in the CeA that is modulated by cortical inputs and catacholamines. Therefore, the effects of ethanol exposure on BLA projections to the CeA would be expected to be inhibitory acutely, while chronic exposure produces hyperexcitibility. However, acute ethanol can increase BDNF and pCREB in the CeA, suggesting excitation in this nucleus. Inhibition of the feed forward GABAergic neurons in the intercalated cell masses might be a homeostatic pathway that regulates CeA function and would be modulated by inputs from PFC, similar to the pathway described by Nestler and Carlezon (2006) that regulates excitability of the mesolimbic dopamine pathway in depression. Recent observations suggest that withdrawal causes a decrease in GABA release probability from the lateral intercalated cells with no effect on the local, feed-back type neurons (B.A. McCool, personal communication), suggesting a decrease in feed forward inhibition to the BLA during withdrawal that contributes to hyperexcitibility in the BLA. Interestingly, neurons in the CeA show a decrease in pCREB during ethanol withdrawal (Pandey et al., 1999) and acute amphetamine exposure also decreases pCREB in this region (McPherson et al., 2007). Therefore, these data might be explained by the relative activity of ethanol at each cell type within this local circuit or may reflect withdrawal-induced catecholamine release. These studies also suggest dissociation between BDNF and pCREB that can be modulated by catacholamines in the extended amygdala.

An intriguing target area of the amygdalar projections from the BLA and, to a lesser degree the CeA, neurons is the BNST, which is hyper-excitable during opiate withdrawal through norephinepherine (Aston-Jones and Harris, 2004). Output of this nucleus is also regulated presynaptically by NPY and postsynaptically by CRF/Urocortin (Kash and Winder, 2006). Given the partial efficacy of opiate and CRF antagonists in reducing ethanol consumption and the efficacy of benzodiazepines in treating withdrawal-induced anxiety and seizures, this circuit may be mediating withdrawal. Projections from the CeA and BNST also regulate stress hormone production, which could be correlated with synaptic activity as an in vivo surrogate for activity in this circuit. A complete understanding of the role of BDNF in the amygdala and stress circuitry will require integration of the electrophysiological data with fine mapping of the time course of transcription factor activation and gene transcription both during exposure and after withdrawal. Complex, multi-electrode electrophysiological recordings have been used to elucidate the circuitry of the amygdala (Woodruff et al., 2006) and the effect of ethanol on this structure is a question well suited for this technique. Combined immunodetection would not be possible in real time but the use of a GFP reporter construct (e.g. synaptically driven translation or immediate early gene promoters driving somatic GFP expression) could be used to monitor activity in combination with electrophysiological studies, which may aid in mapping this circuit.

7. BDNF-TrkB polymorphisms in humans: depression, anxiety and addiction

There have been reports of isolated cases of deletions or inactivating mutations in the human TrkB-BDNF system that lead to severe hyperphagia, obesity, hyperactivity and impaired cognitive function (Yeo et al., 2004; Gray et al., 2006). BDNF polymorphisms have been linked to eating disorders in humans (Koizumi et al., 2004; Monteleone et al., 2006; Mercader et al., 2007) and the human phenotype of FASD is remarkably similar to that described for BDNF haplodeficient mice. Therefore, BDNF is involved in neuronal systems that regulate multiple aspects of alcoholism and alcohol use disorders.

Genetic variation at the TrkB gene is associated with alcohol dependence and antisocial personality disorder in a Finnish population (Xu et al., 2007). In particular, a haplotype was detected at a higher frequency in non-alcohol dependent subjects compared with alcohol dependent individuals, suggesting that some genotypes may be “protective.” Whether variation at the TrkB gene alters receptor expression or function is unknown. The BDNF coding sequence is also polymorphic in human populations and several studies have linked these polymorphisms with substance abuse and psychiatric disorders. A dinucleotide repeat in the BDNF gene was recently shown to be associated with drug abuse vulnerability (Uhl et al., 2001). This repeat is located in an intron 5′ to the first coding exon. This study did not specifically examine alcoholics and excluded subjects who did not abuse at least 1 illegal substance, however.

A polymorphism in the coding sequence in the pro region of the protein was also recently identified. This single nucleotide polymorphism (G196A) results in either a valine or a methionine residue at codon 66. BDNF Met66Met shows impaired stimulus-induced secretion and decreased localization in secretory granules (Egan et al., 2003; Chen et al., 2004). Humans with the Met66Met genotype have impaired hippocampal function, with poorer episodic memory and less activation in the hippocampus as measured by functional magnetic resonance imaging (Egan et al., 2003; Hariri et al., 2003). The Met66 allele is also linked to reduced hippocampal volume and depression (Bueller et al., 2006; Pezawas et al., 2004; see Dumas and Monteggia 2006 for review).

Met66Met was recently shown to be associated with alcoholism in violent alcoholics (Matsushita et al., 2004), but this was not replicated in a small sample of Chinese alcoholics with violent tendencies (Tsai et al., 2005). In a follow-up study, the Met66Met genotype was associated with harm-avoidant personality types and depression/anxiety but this may be modulated by the presence of a polymorphism in the promoter with lower DNA binding and reporter activity (Jiang et al., 2005). Val66 was shown to be positively associated with methamphetamine and heroin abuse (Cheng et al., 2005). Interestingly, this allele has also been associated with bipolar disorder (Sklar et al., 2002; Lohoff et al., 2005) and ADHD (Kent et al., 2005). Like humans, Met66Met mice exhibit increased anxiety (Chen et al., 2006).

BDNF is present in plasma and has been used as a surrogate marker for central function (Karege et al., 2002). Joe and colleagues (2007) measured BDNF levels in alcoholic Korean patients and found a decrease in plasma BDNF. This difference was further magnified when examined in terms of family history, with family history positive individuals having consistently lower levels and less variability between subjects. There was significantly more variability in the BDNF levels measured in patients with no family history of alcoholism. Changes in BDNF did not correlate with anxiety or depression in this patient population, which is in contrast to animal studies. Unfortunately this study did not report the genotype of these patients with respect to the Met/Val BDNF polymorphism.

Patients with the Val66Val genotype would be expected to have globally higher BDNF levels. Higher BDNF levels may protect against depression in the hippocampus and through increased survival of serotonergic neurons, but higher levels of BDNF are associated with addiction and stress in the mesolimbic dopamine system in rodent models. In addition, BDNF haplodeficient mice show decreased psychostimulant-induced conditioned place preference, suggesting that patients with lower BDNF secretion might me more resistant to stimulant abuse. BDNF is trophic for GABAergic neurons (Carrasco et al., 2007; Baquet et al., 2004; Strand et al., 2007); therefore, the Met66Met allele might lead to reduced interneuron function, thereby contributing to the anxious phenotype. The Val66Met polymorphism may predispose to addiction in a substance-selective manner but this seems unlikely since the Val66 allele is over-represented in bipolar patients and they show little selectivity in the drugs they chose to abuse (Levin and Hennessy, 2004). Alcoholism and substance abuse has historically been prevalent in writers and artists with comorbid psychiatric disorders who are unlikely to have impaired hippocampal function. A more likely scenario invokes the archaic self-medication theories. Several genes predisposing to alcoholism have been described and progress is slowly being made towards integrating these polymorphisms with behavioral phenotypes. Knowledge of the BDNF genotype and underlying psychopathology may ultimately help determine the motivation to seek alcohol and may eventually be used to tailor therapeutics.

8. BDNF as a therapeutic target for alcoholism?

Specific strategies targeting BDNF therapeutically have a long history of partial successes and failures (reviewed by Pezet and Malcangio, 2004). Therapeutic BDNF was first explored by pharmaceutical companies for the treatment of neurodegenerative disorders. BDNF is trophic for dopaminergic neurons and early preclinical studies suggested that BDNF infusion or transduction of the gene into the striatum could halt degeneration of dopaminergic neurons in the substantia nigra. These types of studies have been tedious due to problems administering a protein, whether systemically, virally, microencapsulated or through a cannula. Since BDNF has a narrow window of activity and TrkB desensitizes (Carter et al., 1995; Segal et al., 1992; Ohrtman et al., 2006), regulated and well controlled levels in selected brain regions are also essential for this type of therapy.

Ethanol inhibits BDNF signaling and BDNF is neuroprotective in several in vitro models of developmental ethanol exposure. BDNF is a stimulus for migration for multiple neuronal populations. Dysregulation of BDNF, TrkB, or the associated signaling pathways during critical migratory periods would presumably be irreversible. Therefore, the ultimate intervention would be aimed at the mother. Despite intervention and close monitoring, a significant number of newborns are still diagnosed with FASDs. Fortunately the developing brain is remarkably plastic and postnatal intervention with environmental enrichment (Hannigan et al., 2006), appropriate nutritional supplementation (Thomas et al., 2004) and motor training (Klintsova et al., 2000) can reverse some of the deficits caused by gestational ethanol exposure. In addition, the characterization of factors governing stem cell production during intoxication and repair may lead to regererative therapies (Nixon 2006). These interventions ultimately involve BDNF and may be augmented by BDNF peptidomimentics. O’Leary and Hughes (2003) recently developed a peptidomimetic partial agonist of TrkB that may prove useful therapeutically. A partial agonist would presumably not show the same desensitization kinetics as the full agonist, which could expand the therapeutic window. Interestingly, exercise is sufficient to induce BDNF (Cotman and Berchtold, 2002) and may also augment other behavioral and nutritional interventions.

A feasible approach to be investigated in alcoholism is pharmacological modulation of BDNF levels. As discussed earlier, antidepressants increase BDNF levels in the hippocampus, however, it was recently shown that high doses of venlafaxine actually decrease hippocampal BDNF (Xu et al., 2003). This decrease is a property shared with the structurally unrelated antidepressant bupropion, which has been shown to be effective in the treatment of cocaine abusers with and without comorbid ADHD (Levin et al., 2002; Margolin et al., 1991). Polymorphisms in the BDNF gene have also been linked to antidepressant efficacy. Patients with the Met66Met genotype have a better response to fluvoxamine, milnacipran and citalopram (Choi et al., 2006; Yoshida et al., 2006). However, therapeutics that globally target TrkB may not be universally efficacious in depression or anxiety where individual variation and regional differences exist in the effect of BDNF.

Pharmacological agents useful in a particular subset of individual patients might be deduced a priori based on serum BDNF levels and genotype. Alcoholic patients with underlying anxiety, a Met66Met genotype and low serum BDNF might be more likely to benefit from SSRI antidepressant therapy for alcoholism. A polysubstance abusing patient (addictopathic) or sociopathic with a Val66Val genotype might benefit from bupropion. Alcoholism, depression and anxiety are polygenic disorders, therefore there are likely to be substantial interactions between BDNF the many systems it regulates, such as catecholamines, CRF and NPY that influence both mood and alcohol consumption (reviewed by Oroszi and Goldman, 2004; Thorsell et al., 2006).

Abbreviations

ADHD

Attention Deficit Hyperactivity Disorder

BDNF

Brain-Derived Neurotrophic Factor

BLA

Basolateral Amygdala Complex

BNST

Bed Nucleus of the Stria Terminalis

CaMK

Calmodulin-dependent Kinase

CeA

Central Amygdala

CNS

Central Nervous System

CREB

Cyclic 3′5′ adenosine monophosphate Response Element Binding protein

DA

Dopamine

ERK

Extracellular-signal-Regulated Kinase

FAS

Fetal Alcohol Syndrome

FASD

Fetal Alcohol Spectrum Disorder

LA

Lateral Amygdala

MeA

Medial Amygdala

NAc

Nucleus Accumbens

NGF

Nerve Growth Factor

NPY

Neuropeptide Y

PI 3-kinase

Phosphatidyl inositol 3-OH kinase

PN

Postnatal

SN

Substantia Nigra

Trk

Tropomyosin-related kinase

VTA

Ventral Tegmental Area