Risk and Resilience : The Role of Brain-derived Neurotrophic Factor in Alcohol Use Disorder

Brain-derived neurotrophic factor (BDNF) is well known for its role in synaptic plasticity. More recently, BDNF has come to be regarded as a potential resilience factor in a variety of conditions that are characterized by maladaptive neuroplasticity, including alcohol use disorder. Research in animal models suggests that BDNF may serve as a protective gatekeeper in the transition from social drinking to compulsive alcohol consumption. Further, stress-induced modification of BDNF signaling may have an important role in anxiety-modulated alcohol consumption. In this article, we will review recent studies of BDNF and alcohol use in human participants. Studies included in this review have used genetic or epigenetic approaches or have measured peripheral BDNF protein levels in serum or plasma. Importantly, a number of these studies have incorporated neuroimaging methods to provide information about structural and/or functional central nervous system correlates of BDNF measures in relation to alcohol use. While there remains a great deal of variability in research findings from human participants on this issue, a number of studies suggest that behavioral or pharmacological interventions designed to enhance neuroplasticity may be a promising avenue for future research on alcohol use disorder.


Introduction
Like many human conditions, the development of alcohol use disorder (AUD) is influenced by complex interactions of genetic and environmental factors that shape both risk and resilience to the spiral of addiction.Historically, AUD research has tended to focused on risk.Researchers have sought to determine which factors or combinations of factors might increase an individual's chances of transitioning from social drinking to AUD and/or relapsing after periods of abstinence from alcohol.However, as noted in a recent review by Logrip and colleagues [1], alcohol is widely consumed, but only a small percentage of drinkers (i.e., 5% [2]) go on to develop AUD.These statistics have prompted researchers to shift their attention to uncovering factors that might promote resilience in the face of risk for AUD [3].Research findings from animal models and human participants are converging to suggest that brain-derived neurotrophic factor (BDNF) may have a role in this process.BDNF may function as a resilience peptide [4] in an endogenous protective pathway [1] that could potentially be exploited in the prevention and treatment of AUD.In the current review, we will describe contemporary models of BDNF and alcohol interactions, which have derived largely from animal research, and summarize recent studies of such interactions in human populations.Potential avenues for future research will also be discussed.

Brain-derived neurotrophic factor
Brain derived neurotrophic factor is the most abundant neurotrophin in the mammalian central nervous system (CNS) [5].Synthesized in the endoplasmic reticulum as a pre-pro-molecule, BDNF undergoes two cleavage steps from pre-pro via pro-BDNF to its mature form, which is then packaged in secretory vesicles [5][6][7].Following neural activity-dependent release, BDNF binds to tropomyosin-related kinase (TrkB) receptors, activating downstream signaling cascades that are essential for neuronal integrity and survival [5,8,9].Historically, BDNF is perhaps best known as a protein essential for long-term potentiation (LTP) in the hippocampus, a neuronal process that is necessary for learning and episodic memory [10][11][12].In this role and others, BDNF promotes the formation and survival of dendritic spines, which are crucial for proper synaptic plasticity and connectivity in the adult brain [13,14].
The human BDNF gene is located on the short arm of chromosome 11.The most commonly studied BDNF polymorphism is Val 66 Met (rs6265), a single-nucleotide substitution of A for G at nucleotide position 196, producing a non-conservative substitution of a valine with a methionine at codon 66 of this gene [10,[15][16][17].The location of this single-nucleotide polymorphism (SNP) in the
pro-domain of the BDNF gene results in impaired intracellular processing, trafficking, and extracellular secretion of BDNF protein [10].As a result, this SNP may impair neuronal plasticity [18][19][20] and ultimately has the potential to affect higher functions such as memory and cognition [20,21].In a number of studies, the BDNF Val 66 Met polymorphism has also been associated with genetic susceptibility to neuropsychiatric disorders including Alzheimer's disease, autism, depression, eating disorders and schizophrenia [22,23].
Alcohol use disorder (AUD) has also been described as a disease of maladaptive plasticity [24][25][26], though there is some debate as to whether structural plasticity and synaptic changes drive addictive behaviors, or whether they reflect homeostatic compensations to drug use [27].Recently, researchers have begun to appreciate the potential role of BDNF as a protective gatekeeper in the transition from casual drinking to compulsive alcohol consumption [1].One model of direct BDNF and alcohol interactions [1] suggests that BDNF signaling in corticostriatal pathways may act to keep alcohol consumption in moderation through mechanisms related to habit learning and compulsive behaviors, while risk for transition to uncontrolled drinking may occur when BDNF levels in the medial prefrontal cortex are reduced.Another model [28] suggests that BDNF may serve as a mediator in the relationship between stress and alcohol consumption.These models will be reviewed in the next section of this article.

BDNF and alcohol-a corticostriatal connection
Earlier work in rodent models (i.e., 1990s to 2010) has been reviewed by Ghitza and colleagues [29].
More recently, Logrip and colleagues [1] have provided an extensive review of direct interactions between BDNF and alcohol based on the recent animal research literature.The effects that Logrip and colleagues [1] describe are alcohol-specific, that is, these interactions with BDNF do not occur in response to other appetitive reinforcers such as sucrose [30].They are regionally specific, in that the interactions appear in the dorsal striatum, corticostriatal pathways and the medial prefrontal cortex, rather than in other addiction-relevant areas such as the nucleus accumbens in the ventral striatum [30][31][32][33].
Further, the effects are moderated by alcohol dose.The relationship between BDNF levels and alcohol use appears to change as alcohol administration levels transition from moderate to heavy exposure.For instance, Logrip and colleagues [1] report that, in response to acute low-dose and/or longer-term moderate alcohol administration below the level of intoxication, BDNF mRNA increased in the dorsal striatum [32,33].Further, acute application of alcohol onto striatal neurons increased BDNF expression, translation and secretion of BDNF protein, and TrkB receptor activation [1,34].In contrast, when animals consumed alcohol at levels sufficient to induce intoxication, particularly over longer periods of time, researchers reported long-lasting reductions of cortical BDNF mRNA levels and a decrease in BDNF expression, specifically in the medial prefrontal cortex [1,30,31].Logrip and colleagues [1] describe this breakdown in alcohol's ability to upregulate BDNF expression as a possible loss of a protective mechanism.In other words, the presence of sufficient levels of BDNF in these corticostriatal pathways may promote resilience and serve as a gatekeeper in the transition to more compulsive alcohol consumption.

BDNF as a mediator of stress and alcohol interactions
In addition to the model described by Logrip and colleagues [1], other researchers point to the involvement of BDNF in the relationship between stress and alcohol consumption.Many people view alcohol use as an effective means of coping with stress.This perception is due in part to alcohol's anxiolytic effects on the GABA A receptor [35,36].In particular, acute alcohol consumption increases pre-synaptic GABA release and enhances post-synaptic GABA A receptor function in the central amygdala, ultimately resulting in an acute reduction of anxiety and responsiveness to stress [37][38][39].In fact, the central amygdala has recently been described as a hub for anxiety and AUD [40].These connections may explain, in part, why stressful life events tend to prompt alcohol consumption for some people, and may increase risk of relapse for those with AUD [41][42][43][44].However, it is also important to note that alcohol is perceived as a stressor by the body, and thus activates the neuroendocrine stress response system [44].Chronic alcohol exposure and withdrawal may trigger dysregulation of the hypothalamic-pituitary-adrenal (HPA) axis and the brain's stress-reward circuitry, leading to an escalating spiral of stress, negative affect and alcohol consumption [24,[44][45][46][47].
In addition, a growing body of literature suggests that risk for the development of addictive behaviors is influenced by both genetic and environmental factors [48].For example, environmental influences such as negative early life experiences may cause long-lasting epigenetic changes in expression of stress-related genes that influence brain function and behavior, thus increasing risk for affective disorders and AUD later in life [49,50].Such epigenetic changes do not modify the DNA sequence itself, but rather result in covalent modifications of either DNA (i.e., methylation) or of the histone skeleton (i.e., acetylation) around which it is packaged into chromatin [50,51].Thus, one's adaptive or maladaptive responses to stressors in adulthood may reflect the resilience or vulnerability of one's neuroendocrine profile that was shaped by early gene x environment interactions [5].Further, the use of addictive substances including alcohol may further interact with one's genome and modify responses throughout the lifespan [3,52].Cadet [3] provides an excellent review of how interactions of drugs of abuse with an individual's genome, and the impact of the environment, may shape an individual's response to addictive substances.
BDNF may be a key component in these processes.Moonat and Pandey [28] describe a model, in which stress-induced changes in BDNF signaling lead to altered synaptic plasticity, which ultimately increases risk for both anxiety and alcohol consumption.It is well known that stressful experiences, and the resulting release of glucocorticoids such as cortisol, impact neuroplasticity in the brain through BDNF-related mechanisms [5,53,54], though it is important to recognize that other molecules are also involved in stress-induced neuroplasticity [55].The influence of stress on the brain has been described as an inverted U-shaped curve, in which moderate stress for short periods may have a positive effect on neuronal activities.Conversely, intense stress tends to have a negative effect that intensifies over time.As described by McEwen and colleagues [55], moderate short-term stress results in enhanced synaptic transmission and LTP, while moderate chronic stress facilitates remodeling of neural architecture and successful adaptation.In contrast, intense acute stress may lead to a short-term suppression of these functions, and intense chronic stress may lead to neurochemical distortions, loss of resilience, impaired remodeling and lack of recovery [55].Stress-related changes in neuroplasticity may be particularly pronounced in areas of the brain that underlie behavioral coping responses; however, the direction of the effects may differ by brain region [5,55,56].For example, some studies of human participants suggest that early life stress may induce volumetric decreases in the prefrontal cortex, along with increases in amygdala volume [57], and greater vulnerability to affective disorders [58].
Through epigenetic mechanisms, early life stress may upset the equilibrium of glucocorticoid-and BDNF-signaling pathways, thus tilting the aforementioned inverted U-shaped curved in the direction of increased risk [5].For instance, studies suggest that rats reared in stressful environments may have increased methylation of BDNF DNA, leading to decreased BDNF expression in the prefrontal cortex in adulthood [50,59], a profile that has been associated with excessive alcohol consumption in other animal models [1].It is particularly interesting to note that the opposite may also be true.Branchi and colleagues [60] devised a study in which young mice were provided with environmental enrichment through a communal nesting arrangement (i.e., three mothers with their offspring) that offered enhanced maternal behavior and peer interactions compared to the standard laboratory conditions (i.e., one mother with her offspring).As adults, mice reared in a communal nest had higher levels of histone acetylation in the BDNF gene, a condition that is permissive of BDNF expression, and exhibited a faster increase in BDNF protein levels following environmental challenge [60].

BDNF as a resilience protein
Taken together, these models and their supporting studies reinforce the notion that BDNF may function as a resilience factor at key risk points in the trajectory toward development of AUD.
Theoretically, environmental enrichment interventions might be expected to combat the negative effects of early life stressors, leading to positive changes in BDNF levels in key brain areas such as the prefrontal cortex.These neuroplastic modifications might facilitate the development of better behavioral coping responses [60], thus, decreasing an individual's vulnerability to self-medication [37] and transition to compulsive alcohol consumption in later life [1].However, it is important to note that the aforementioned models, describing how BDNF might contribute to resilience in AUD, were built largely on animal research.The results of studies with human participants have been more varied.The remainder of this review will focus on that work.

AIMS Neuroscience
Volume 3, Issue 4, 398-432.Theoretically, successful treatment to reduce these negative effects of stress might prompt positive changes in BDNF levels, thus decreasing vulnerability to self-medication and potentially interrupting the transition to compulsive alcohol consumption.

Procedure
The literature review procedure focused on recent studies relevant to BDNF in AUD or the development thereof.Articles were retrieved using search engines such as PubMed, GoogleScholar and PsychInfo, and search terms included combinations of brain-derived neurotrophic factor (BDNF), alcohol, and stress.Original research studies focusing on human participants and published in English between January 2010 and February 2016 were reviewed.For review of articles prior to 2010, Forero and colleagues [61] and Haerian and colleagues [62] summarize genetic association studies of BDNF and AUD, and Meng and colleagues [63] cover previous work on peripheral BDNF protein levels in AUD.
Articles focusing on BDNF in healthy control participants while measuring functions relevant to dopaminergic reward were included in the review, as long as alcohol consumption was measured in the study.Articles focusing exclusively on substance use disorder (SUD) without considering AUD
separately were excluded.Using this procedure, we identified a total of 32 studies.When these were categorized by methodology, 18 studies used genetic or epigenetic approaches to study the relationship between BDNF and AUD or factors affecting risk for AUD.These are summarized in Table 1.The remaining 14 studies measured serum or plasma levels of BDNF protein in relation to alcohol use or risk for AUD.These are summarized in Table 2.In Tables 1 and 2, studies are presented chronologically, and alphabetically within a given year.Information includes authors, population and study design, as well as outcomes relevant to BDNF.Prior to reviewing these studies, methodological considerations for working with human participants will be discussed in the next subsection.

Methodological considerations in human studies
Human participants bring with them a plethora of genetic and environmental factors that add variability to scientific research.Some of these factors can be controlled through careful study design and/or may be addressed using sophisticated multivariate statistical techniques.In other cases, the experimental approaches used in animal research are simply impossible to replicate with human participants.Before embarking on a description of study results in Tables 1 and 2, it is important to briefly consider several issues relevant to investigations of BDNF and alcohol use in human populations.

Peripheral versus central measurement of BDNF
Whereas studies in animal models have focused on CNS levels of BDNF mRNA or concentrations of BDNF protein in specific brain regions, the majority of studies in human populations have been constrained to measurement of peripheral BDNF protein levels, as found in serum or plasma.This is a limitation of human studies, as the mechanisms behind BDNF concentration changes in the peripheral circulation are not as well understood [64] and may have uncertain relevance to changes in the CNS [65].BDNF protein can cross the blood brain-barrier in both directions [64,66,67], and studies in rodent models have reported significant positive correlations between levels of BDNF found in serum or whole blood with levels of BDNF in brain tissue [68][69][70].Further, a study in human participants by Lang and colleagues [71] addressed this issue by using proton magnetic resonance spectroscopy to examine the relationship between serum BDNF levels and in vivo concentrations of cerebral N-acetylaspartate, a marker of neuronal integrity.
Results indicated a positive association between peripheral serum BDNF levels and cerebral N-acetylaspartate concentrations in the anterior cingulate, but not in the hippocampus [71].These results at least partially suggest that BDNF serum concentrations could reflect some aspect of neuronal plasticity in the human brain [71].
Timing of sample collection and participant characteristics may also influence results [72].
Researchers have noted that peripheral BDNF levels in humans oscillate in a circadian fashion with the highest concentrations of peripheral BDNF occurring in the morning and levels decreasing throughout the day [66].This pattern has been noted for plasma BDNF levels in both males and females [66,73,74], but perhaps not for serum BDNF levels [75].Further, in women, BDNF levels may be influenced by ovarian function, resulting in a blunted diurnal rhythm during the luteal phase of the menstrual cycle [66,74].Other studies suggest that BDNF levels decrease with age in both sexes, a phenomenon that may contribute to age-related synaptic loss and cognitive decline [76].Use of over-the-counter or prescribed medications, or illicit substances, may also influence BDNF levels and should be assessed via questionnaire in any study where BDNF samples are obtained.Bus and colleagues [72] note that time of sample storage, food intake before sampling, and participants' urbanicity may be additional factors that affect peripheral BDNF levels.Finally, as mentioned earlier in this review, the BDNF Val 66 Met polymorphism affects activity-dependent release of BDNF [10], which could increase individual variability in studies of peripheral BDNF levels.

Sensitive information obtained via self-report
In all of the studies of human participants that are summarized in Tables 1 and 2, researchers obtained alcohol consumption data from self-report questionnaires or interviews with study participants.This approach is influenced by the interaction of social and contextual factors, respondent characteristics and task attributes [77].In general, people tend to under-report their alcohol use [78,79].Despite these limitations, in the field of alcohol research, self-report remains the standard method of assessing alcohol use among human participants [79].
In addition, some studies in Tables 1 and 2 included information about participants' early life experiences.These data were also gathered by retrospective self-report with or without collateral information from family members or other individuals.As a result, these studies also suffer from a number of limitations, which have been reviewed by Thabrew and colleagues [80].For example, previous research has shown poor agreement between actual childhood events and adolescent retrospective reports of childhood treatment [81], as well as considerable temporal instability in repeated reports on childhood physical punishment and sexual abuse from 18-to 21-year-olds [80,82].
Here again, participants have a tendency to under-report negative events.Studies of this issue suggest a situation in which victims may be more likely to provide false negative reports, rather than non-victims providing false positive reports [80,82].However, researchers in this field suggest that instability over time is likely to have only a minimal impact on statistical models linking childhood maltreatment and adult psychiatric disorders [80,81].

Studies using genetic or epigenetic approaches
All publications included in European participants with schizophrenia, AUD or the co-morbid condition, along with healthy controls.Results of that study suggested that the Val 66 Met polymorphism was associated with AUD and risk-taking behavior after drinking, with the strongest associations noted for the Met/Met allele.
It is interesting to note that these associations in Cheah and colleagues' study [86] were significant only among AUDs with co-morbid schizophrenia and not among AUDs without schizophrenia; further, the relationship appeared to be strongest for males.Finally, a longitudinal study of Caucasian, European adolescents by Nees and colleagues [87] found that Met carriers with high levels of alcohol use had a lower level of reward reactivity in the putamen (notably, part of the dorsal striatum; see [1]) and were more likely to orient toward alcohol and to drink alcohol two years later.cerebellar gray matter volumes compared to HCs.This difference in cerebellar gray matter was associated with an interaction of BDNF and allelic variants of GABRA2 (i.e., the gene encoding the alpha-2 subunit of the GABA A receptor), suggesting that these gene/gene interactions may impact AUD risk and brain morphology via signaling in early development and/or gene expression throughout adolescence and young adulthood [88].Hill and colleagues [88] noted that the potential involvement of cerebellar morphology in risk for AUD is a topic of ongoing research, however, the cerebellum may be involved in addictive processes via its critical roles in attention, memory, facial recognition, theory of mind, emotional monitoring and time perception [89][90][91][92][93][94][95].
Possession of the Met allele may also influence cognitive recovery in abstinent AUDs.In a study of adult AUDs following approximately one month of abstinence from alcohol, Mon and colleagues [96]  Other studies examined reward processing in healthy controls, an issue that is relevant to the development of addictive disorders.Pecina and colleagues [98] studied the potential influence of the BDNF Val 66 Met polymorphism on dopaminergic stress and reward processes in HCs using an fMRI task of monetary incentive delay and a PET task measuring striatal dopamine neurotransmission during two challenges of opposite valence: a pain-stress challenge and a placebo analgesic.Results indicated that during the anticipation of monetary loss (large loss minus null condition), BDNF Met carriers had greater blood-oxygen-level-dependent activation in the ventral tegmental area-nucleus accumbens-medial prefrontal cortex circuit as compared to Val/Val homozygotes [98].No genotype differences were seen during the anticipation of reward (large gain minus null condition).Further, in the PET experiment, Met carriers had greater baseline dopamine receptor (D2/3) availability in the left and right ventral striata as compared to Val/Val homozygotes.Relative to Val/Val homozygotes, Met carriers also exhibited greater bilateral dopamine release in the nucleus accumbens during pain challenge and an overall reduction in dopamine neurotransmission during the placebo condition [98].
Pecina and colleagues [98] suggest that these results support the notion that the BDNF Val 66  For drinkers with more severe symptoms of AUD, a genetic component derived from the CREB-BDNF pathway was associated with an imaging component reflecting hyperactivation in the precuneus, superior parietal lobule, and posterior cingulate areas, regions that have been previously associated with craving and alcohol use severity [99][100][101][102][103][104].This association remained significant after correction for multiple testing and when age, sex and race were statistically controlled.

No Association between BDNF and AUD
In contrast to the studies described above, seven of the 13 studies that examined BDNF Val 66 Met allelic frequencies in AUDs versus HCs, or in HCs with and without a family history of AUD, found no significant associations between Val 66 Met and AUD [105][106][107][108][109][110][111].Thus, despite the compelling evidence of studies suggesting potential mechanisms of enhanced AUD risk among Met carriers (or conversely, endogenous resilience to AUD risk among Val/Val homozygotes), the majority of association studies on this topic have actually yielded negative findings, a result that is reinforced by recent meta-analyses [61,62].

Epigenetic Approaches
We reviewed only one study that used an epigenetic approach.Heberlein and colleagues [112] examined potential alterations in BDNF gene methylation and BDNF protein levels in serum during days 1, 7 and 14 of alcohol withdrawal in AUDs.In particular, this study examined DNA methylation at the BDNF promotor, which would be expected to repress gene transcription [113].Relative to baseline BDNF methylation of HCs, methylation of the BDNF promotor was increased in AUDs on days 1 and 7 of alcohol withdrawal.In terms of within-group differences for AUDs, methylation of the BDNF promotor was successively decreased at each time point during alcohol withdrawal, and decreasing methylation was associated with decreasing symptoms of depression during this time-period [112].Somewhat counterintuitively, AUDs with lower BDNF methylation levels at day 14 of alcohol withdrawal were less likely to remain abstinent from alcohol in the long-term, compared to AUDs with higher methylation levels at that time point.However, the researchers reported that peripheral serum levels of BDNF protein were not statistically associated with methylation of the BDNF promotor in this study and did not differ between AUDs and HCs [112].
Additional studies are needed to clarify the significance of BDNF methylation during alcohol withdrawal in relation to long-term sobriety.Recent studies that examined peripheral BDNF protein levels without a genetic or epigenetic component will be summarized in the next section (see Table 2).

Studies examining peripheral BDNF protein levels and alcohol-related variables
Fourteen studies were identified which examined alcohol-related variables along with peripheral BDNF protein levels in serum and/or plasma.These investigations included a wide variety of study designs.They are summarized chronologically in Table 2 and described by study type, below.

BDNF and Alcohol Use in the General Population
In two studies [114,115], researchers recruited participants from non-treatment-seeking, community populations and categorized them into AUDs or other alcohol-using groups versus HCs based on self-reported alcohol consumption and standardized diagnostic criteria.Lhullier and colleagues [114] grouped participants based on responses to the CAGE questionnaire [116], where participants with CAGE scores ≥ 2 were considered to have moderate to severe AUD, and those with scores < 2 were categorized as HCs.In Zhang and colleagues' study [115], participants were grouped as -alcohol users‖ if they reported more than occasional alcohol use and had a total score of at least 3 on the Michigan Alcohol Screening Test (MAST [117]).The comparison group of -non-alcohol users‖ was comprised of those who had never used alcohol or who occasionally consumed fewer than 5 drinks per episode [115].Neither Lhullier and colleagues [114], nor Zhang and colleagues [115] reported statistically significant differences in serum BDNF levels between alcohol and control groups.Further, Zhang and colleagues detected no significant correlation between BDNF and MAST scores or neuropsychological test scores in their sample.

BDNF in Alcohol Use Disorder
Other studies directly compared BDNF levels in clinical samples of participants with alcohol use disorder (AUDs) versus healthy controls (HCs) under various conditions of alcohol withdrawal and abstinence.
One strategy involved monitoring BDNF protein levels in AUDs undergoing withdrawal from alcohol and then comparing these values to typical BDNF levels in HCs.In two studies, there were no significant differences in BDNF protein levels of HCs versus AUDs entering treatment for alcohol withdrawal.Heberlein and colleagues [118] monitored serum BDNF in AUDs on days 1, 7 and 14 of alcohol withdrawal and found no significant differences in AUDs versus HCs at baseline or over time.
However, BDNF levels were negatively associated with alcohol withdrawal severity in AUDs on day 1; AUDs with higher BDNF levels had less severe alcohol withdrawal symptoms [118].Costa and
colleagues [119] studied serum BDNF in AUDs at the time of initial alcohol withdrawal and at a 6-month follow-up session.Similar to Heberlein and colleagues [118], Costa and colleagues [119] found no significant differences in serum BDNF between HCs and AUDs at treatment intake.Further, when AUDs were categorized based on abstinence status at the 6-month follow-up, it was noted that abstinent and non-abstinent AUDs did not differ on BDNF at intake.However, at the 6-month follow-up session, BDNF levels were higher in abstinent AUDs relative to both HCs and non-abstinent AUDs [119].
In contrast, other studies have noted significant differences in HCs relative to AUDs at treatment intake.A cross-sectional study by Zanardini and colleagues [120] measured both serum and plasma BDNF.While there were no significant differences in plasma BDNF between groups, serum BDNF levels were lower in AUDs at treatment intake, as compared to HCs.The researchers noted that this finding was not related to blood alcohol concentration at intake or treatment with prescription medications [120].Huang and colleagues [121] reported a similar result (i.e., BDNF of AUDs at intake < HCs) and further noted that the lowest levels of serum BDNF for AUDs at intake occurred among those who had experienced delirium tremens (DTs) during alcohol withdrawal.
Patients were then followed over the initial withdrawal period.By day 7 after alcohol withdrawal, serum BDNF had increased significantly in AUDs, such that BDNF of non-DT AUDs was comparable to HCs [121].A more recent study by Kohler and colleagues [122] also found reduced serum BDNF levels in AUDs admitted for alcohol withdrawal, as compared to HCs.After an initial decrease, BDNF levels in AUDs showed a tendency to increase between days 3 and 14 of alcohol withdrawal, but these changes remained below the level of statistical significance [122].Finally, Reynolds, Mueller and MacLaren [123] examined the effects of dexmedetomidine administration on BDNF levels during alcohol withdrawal in AUDs (without comparison to HCs) and found no significant differences in plasma BDNF at baseline or over time (up to 120 hours after starting drug therapy) in low and high dose dexmedetomidine groups versus placebo.
Another research strategy involved monitoring peripheral BDNF protein levels in detoxified, abstinent AUDs relative to HCs.Two studies examined BDNF levels in AUDs who had attained a significant level of abstinence prior to testing.In a cross-sectional study, Meng and colleagues [63] found no significant differences in serum BDNF between AUDs who were 4 to 6 weeks abstinent at the time of the study compared to HCs.This study [63] also found that BDNF levels increased to a similar extent in both groups following a 15-minute session of the Trier Social Stressor Task, which is composed of public speaking and mathematics challenges [124].D'Sa and colleagues [68] also examined serum and plasma BDNF levels in AUDs with one month of abstinence compared to HCs.
Using average values from morning blood samples taken over three consecutive days, D'Sa and colleagues [68] found higher serum BDNF levels in AUDs compared to HCs, but no significant differences in plasma BDNF levels.
Finally, four of the studies in Table 2 focused on comorbid diagnoses in AUDs or other drinking groups.For example, Kim and colleagues [125] studied plasma BDNF levels in male AUDs with AIMS Neuroscience Volume 3, Issue 4, 398-432.
normal glucose tolerance, pre-diabetes mellitus, or diabetes mellitus diagnoses at intake and after 30 days of AUD treatment.Only the pre-diabetes mellitus group showed a significant increase in BDNF after 30 days of AUD treatment.Further, increased BDNF in the pre-diabetes mellitus group was significantly greater than that of the normal glucose tolerance group [125].
Miguez-Burbano and colleagues published two studies [126,127] focusing on plasma BDNF levels in participants positive for the human immunodeficiency virus (HIV), as well as HIV seronegative controls.To examine alcohol consumption, they used participants' self-reported alcohol use and hazardous drinking definitions from the National Institute on Alcohol Abuse and Alcoholism (NIAAA [128]).In the first report [126], the researchers found that plasma BDNF levels were negatively correlated with the number of alcoholic beverages consumed per day, but this relationship appeared only after a threshold level of 3 or more drinks per day had been reached.Further, among hazardous drinkers, these findings were largely due to consumption of beer and liquor rather than wine, a finding that the researchers linked to the antioxidant content of wine versus other alcoholic beverages [126].In multivariate analyses, both hazardous alcohol use and plasma BDNF alterations were independent predictors of HIV-associated neurocognitive disorders.As predicted by long-standing literature (e.g., [129]), hazardous alcohol use was associated with decrements in neurocognitive function.However, Miguez-Burbano and colleagues [126] are among the first to note an inverted U-or J-shaped relationship between plasma BDNF levels and neurocognitive performance in human participants, in which only intermediate levels of BDNF seemed to produce a protective effect.
In a second analysis of this study population by Miguez-Burbano and colleagues [127], the researchers focused on participants' drinking trajectories, plasma BDNF levels, and mood states over a 6-month period.At baseline, there was a negative correlation between BDNF levels and participants' endorsement of feeling nervous and stressed.Next, three drinking trajectories were identified based on participants who (1) increased drinking; (2) drank at constant levels; or (3) decreased drinking over the 6-month study period.Results indicated that women in trajectories 1 and 2 reported greater stress and had lower plasma BDNF than those in trajectory 3. Men in trajectory 1 had lower plasma BDNF and marginally greater stress compared to men in trajectories 2 and 3.The researchers noted that dual diagnoses of anxiety plus stress, or depression plus stress, were present in 26% of the study population and in this case all affected participants reported increased drinking during the study period.This reinforces the notion that stress is a significant factor in BDNF and alcohol relationships [127].
Finally, Neupane and colleagues [130] examined potential associations between serum BDNF and lifetime diagnosis of major depression, recent depression symptoms, or AUD severity.Their participants were abstinent male AUDs with an average of 34 days of sobriety at the time of testing.
While the researchers found no association between serum BDNF and lifetime diagnosis of major depression, BDNF was negatively correlated with participants' number of recent depressive symptoms.Further, participants with more severe AUD symptoms or a previous binge-drinking pattern tended to have higher levels of BDNF at the time of testing, although duration of abstinence
from alcohol was not correlated with BDNF level [130].
Together, the results from studies of peripheral BDNF protein levels in AUD and other drinking populations suggest that BDNF levels may differ between drinking groups, and may fluctuate during alcohol withdrawal and continued abstinence.Changes in peripheral BDNF levels may parallel neuroplastic modifications [71] that occur with chronic alcohol consumption and recovery [96].
Additional studies are needed to determine the nature of the relationship between peripheral BDNF levels and neuronal changes during the critical period of early abstinence and AUD treatment.

Discussion and Conclusions
There is no doubt that BDNF is one of the key factors in neuroplasticity [55].In recent years, the alcohol research field has gained an increasing appreciation for the involvement of neuroplasticity in risk or resilience to addiction, and AUD has been characterized as a disease of maladaptive plasticity [24][25][26].Animal research suggests that BDNF may directly influence the transition from light to compulsive drinking through actions in the dorsal striatum and prefrontal cortex [1].It is also likely that BDNF plays a key role in stress and alcohol interactions [28].There has been a recent increase in the number of studies examining these issues in human participants.
However, the results of these human studies have been more contentious than those derived from animal models.For example, although meta-analyses of genetic association studies suggest that the BDNF Val 66 Met polymorphism may not be linked to the development of AUD [61,62], other researchers have found greater AUD vulnerability and poorer long-term prognoses among abstinent AUDs who are Met carriers (See Table 1).Among studies examining peripheral levels of BDNF protein in serum or plasma of human participants, the majority of these have noted some relationship between BDNF levels and drinking histories or trajectories, often in relation to other factors such as stress, anxiety or depression.In the majority of these studies, higher peripheral BDNF levels appear to confer resilience, although Miguez-Burbano and colleagues [126] describe an inverted U-or J-shaped relationship between BDNF levels and neuropsychological performance (See Table 2).As in the studies of genetic association, some counterintuitive findings have been noted, and additional research is needed to clarify the relationship between peripheral BDNF and CNS neuroplasticity in human participants with AUD.
The differential findings across studies of human participants are undoubtedly related to the greater variability that is inherent in this type of research and influential factors that may be overlooked in study designs or statistical models.Human beings are extremely complex.The studies summarized in Tables 1 and 2 suggest that race, ethnicity, sex and gender, co-morbid medical history, and use of other prescribed or abused substances are all factors that should be addressed in the design or analyses of projects examining BDNF and addiction.Further, studies that examine SNPs in isolation or use peripheral protein levels from blood samples collected at a single time point may fail to account for the incredible complexity of gene × gene, gene × environment and environment × environment interactions that exist in human populations [50].These are issues of ongoing concern that have been addressed in numerous recent publications, in which researchers have called for additional thoughtful discussion about which strategies to pursue in order to have the greatest impact on the field of alcohol research [48,51,131,132].
In particular, the characterization of the environmental component in studies of human gene × environment interactions has often been surprisingly limited [131].Heath and colleagues [131] suggest that future studies could employ -big-data‖ approaches by including appropriate questions in aggregate databases to capture information on neighborhood environments that could be used to increase consistency across studies.Big-data approaches may also address the important issue of sufficient sample size in studies of gene × environment interactions [133].The possibility of false discoveries due to type I error is a recognized consequence of low power, and may account for some of the inconsistency of outcomes in these types of studies [133].Finally, epigenetic approaches have been described as a key missing link in the etiology of addiction [51].Cecil and colleagues [51] note that longitudinal modeling of environmental and epigenetic data could be used to identify windows of vulnerability when at-risk individuals might benefit most from preventive interventions.

Interventions to Promote Resilience
In a recent review of the emerging neuroscience of resilience, Srinivasan and colleagues [134] observed that individuals have the capacity to enhance their resilience to stress and addiction through behavioral and pharmacological interventions.Given the proposed role for BDNF in resilience to addiction, it is logical to suggest that strategies that enhance BDNF and neuroplasticity in relevant brain areas might have a positive effect on individuals at risk for development of AUD or those in recovery from AUD.As noted recently by Davidson and McEwen [57], exercise [135], cognitive therapy [136,137] and meditation [138] have all been shown to induce plasticity-related alterations in the brain and to support positive behavioral outcomes for a variety of human conditions.For instance, long-duration, moderate exercise and heavy exercise appear to have a positive effect on signaling and trafficking of the BDNF receptor TrkB, which facilitates neuroplasticity and long-term potentiation in the hippocampus [139], an area of the brain that has been implicated in cognitive recovery during the first seven months of abstinence in AUDs ( [97], see Table 1).Other activities which decrease stress, and promote pro-social behavior and well-being may lead to measurable improvements in brain structure and function, such as increases in prefrontal activation and decreases in amygdala activation with corresponding changes in brain volume in these areas [57].This is a particularly exciting result, as animal studies have directly implicated both of these brain areas in BDNF and alcohol interactions that influence the development of addiction [1].A more holistic approach to the study of factors influencing resilience to AUD, encompassing families, communities, and other sources of enrichment (e.g., [3]) may be promising avenue for future research.

Figure 1 .
Figure 1.BDNF as a Resilience Protein.A) BDNF may serve as a resilience factor to impede the transition from casual alcohol use to compulsive alcohol consumption.B) Further, various types of stress may have a negative impact on BDNF levels in key brain areas such as the prefrontal cortex, leading to a breakdown of BDNF's protective influence, and C) a potential escalation of anxiety-modulated alcohol misuse.

Volume 3 ,
Issue 4, Met polymorphism influences a) dopamine-mediated responses to stress, b) the cognitive regulation of such responses by positive expectations, and c) anticipatory responses to risk/reward conditions via the ventral tegmental area-nucleus accumbens pathway.In a more comprehensive genetics study, Chen and colleagues[99] used parallel independent component analysis with multiple references (pICA-MR) to examine potential relationships between genome-wide SNPs and alcohol cue-induced brain activations in heavy drinkers.The alcohol cue condition consisted of exposure to and brief tasting of the participants' preferred alcoholic beverage type (versus juice, in a control condition) during fMRI scans.Results of pICA-MR analyses on fMRI (54,937 voxels) and SNP (717,129 loci) data indicated 15 fMRI components and 11 SNP components.

Table 1
[84]ylation.Beyond these aspects, the study designs encompassed a wide range of methods from genetic association studies and drug trials to laboratory stress manipulations, neuropsychological assessments and neuroimaging.3.1.1.Positive associations between BDNF and AUDGenetic association studies have focused predominantly on the relationship between AUD and the BDNF Val66Met polymorphism (rs6265).In our review of the literature, 13 of the 18 studies summarized in Table1included a comparison of BDNF Val 66 Met allelic frequency in participants with alcohol use disorder (AUDs) versus healthy controls (HCs) or in HCs with and without a family history of AUD.Of these 13 studies, five reported differential allelic frequencies across AUD or family history groups.For example, in a study of older Korean men with and without AUD, Shin and colleagues[83]found a higher frequency of Met/Met and Val/Met alleles among AUDs compared to HCs.Colzato and colleagues[84]studied male social drinkers from The Netherlands and found that featured either BDNF genotyping, or assessment of BDNF AIMS Neuroscience Volume 3, Issue 4, 398-432.gene Met carriers (i.e., Met/Met and Val/Met genotypes combined for statistical analyses) reported higher levels of alcohol consumption, and had higher anxiety levels and anticipatory stress responses, compared to those with the Val/Val genotype.Su and colleagues [85] studied Chinese men with and without AUD and found that the Met allele appeared more frequently among AUDs with co-morbid depression compared to AUDs without depression and HCs.Cheah and colleagues [86] studied Taken together, these five studies support the notion that individuals with the BDNF Val 66 Met polymorphism, particularly those homozygous for the Met allele, may have greater risk for developing AUD.It is important to note that participants' sex and psychiatric co-morbidity were recognized in these studies [88]r studies in Table1suggest potential mechanisms that may underlie an enhanced risk for AUD among Met allele carriers.For instance, Hill and colleagues[88]studied adolescent/young adult high-risk offspring of multiplex AUD families and found that offspring of AUDs had greaterAIMS NeuroscienceVolume 3, Issue 4, 398-432.
reported increases in frontal and parietal gray matter volume over time in Val/Val homozygotes but not in Met carriers.Increases in gray matter volume were associated with [97]ovements on neuropsychological tests over time, suggesting that BDNF genotype may influence functionally significant brain tissue volume recovery in abstinent AUDs[96].In another article by the same research team[97], AUDs were tested at approximately 7, 33 and 213 days of abstinence.At all time-points, AUDs had smaller hippocampi than HCs, and at one month of abstinence, lower hippocampal volume was associated with neuropsychological decrements in AUDs.In terms of recovery with continued abstinence, hippocampal volume was increased at a 7-month follow-up for the Val/Val homozygotes, but not for Met carriers.Hoefer and colleagues[97]interpreted these results to suggest that BDNF genotype may regulate functional hippocampal recovery in AUDs.If this is the case, then Val/Val homozygotes may possess a potential advantage over Met carriers in terms of cognitive recovery and resilience in AUD.

Table 1 . Studies using genetic or epigenetic approaches to study BDNF and alcohol in human participants (2010-2016).
Participants were 35 female HACAs (M age = 36 years) and 63 HCs (M age = 35 years).All were White and recruited in France.Participants were BDNF genotyped (rs6265) and completed personality questionnaires.Val/Met and Met/Met were combined for comparison to Val/Val.HACAs and HCs did not differ on genotype distributions.Overall, Val/Val participants had lower levels of the reward dependence personality trait compared to Met carriers.There was also an allele by group interaction.In the Val/Val group, HACAs had lower reward dependence compared to HCs.This difference was not significant for Met carriers, whose reward dependence scores were more similar to Val/Val Chen et al., 2015 [99] Participants were 315 (220 male) heavy drinkers (M age = 32 years; 45% Caucasian) recruited from the Southwest United States.This study investigated relationships between genome-wide SNPs and alcohol cue-elicited brain activations.For drinkers with more severe AUD symptoms, the genetic component derived from the CREB-BDNF pathway was associated with an imaging component reflecting hyperactivation in the precuneus, superior parietal lobule, and posterior cingulate.