Investigating the link between serum concentrations of brain-derived neurotrophic factor and behavioral measures in anxious alcohol-dependent individuals

Abstract

Brain-derived neurotrophic factor (BDNF) plays a role in different neurophysiological processes including those involved in alcohol- and anxiety-related behaviors. Preclinical and clinical studies indicate that chronic excessive alcohol use leads to a downregulation of BDNF production in the periphery and in the brain. In addition, a decrease in BDNF concentrations in the blood has been reported to be associated with increased anxiety levels. Non-treatment seeking alcohol-dependent individuals with high trait anxiety were enrolled to assess whether serum BDNF concentrations may be linked to self-reported levels of alcohol drinking, anxiety, and other behavioral measures. Participants had a current diagnosis of alcohol dependence, high trait anxiety score, and were not seeking treatment for alcohol dependence or anxiety. A fasting blood sample was collected from each participant and serum BDNF was measured using an enzyme-linked immunosorbent assay (ELISA). Behavioral data were collected on the same day, including measures of alcohol drinking, craving, dependence severity, and anxiety. Bivariate correlations were run between BDNF levels and behavioral measures. Serum BDNF concentrations were negatively correlated with average drinks per drinking days (r = −0.41, p = 0.02) and positively correlated with obsessive compulsive drinking scale (r = 0.48, p = 0.007) and state-trait anxiety inventory (r = 0.52, p = 0.003) scores. These findings shed light on the possible role of the BDNF system in the neurobiology of alcohol- and anxiety-related behaviors.

1. Introduction

Alcohol use disorder (AUD) is a chronic relapsing brain disease with high global prevalence and is associated with several negative health-related outcomes (Collins, 2016). Individuals with AUD go through patterns of compulsive alcohol use, together with a loss of control over alcohol intake (Burchi, Makris, Lee, Pallanti, & Hollander, 2019). Despite alcohol being widely consumed, only a percentage of drinkers progress to AUD, implying the existence of pathways that may promote or protect from the development of AUD in the majority of moderate drinkers (Ron & Barak, 2016). For example, intracellular signaling cascades and molecular adaptations involved include the cyclic AMP-dependent protein kinase A (PKA) pathway, the extracellular signal-regulated kinase (ERK) 1/2 pathways, glial cell line-derived neurotrophic factor (GDNF), and brain-derived neurotrophic factor (BDNF), the latter being the focus of this study (Ron & Barak, 2016; Ron & Berger, 2018).

BDNF, a widely expressed neurotrophin in the developing and adult brain, signals through its binding to the tropomyosin receptor kinase B (TrkB), also known as the tyrosine kinase receptor B, and participates in a wide range of neurophysiological processes including synaptic plasticity, learning, memory, and neurogenesis (Kowianski et al., 2018). Previous studies suggest that malfunctioning of the BDNF/TrkB signaling may be linked to several neuropsychiatric disorders, including AUD (Andreatta et al., 2019; Autry & Monteggia, 2012; Duman & Li, 2012; Ron & Berger, 2018). In addition to being expressed in the brain, BDNF is also found in non-neuronal tissues such as the heart, liver, kidneys, and blood (Bathina & Das, 2015; Dincheva et al., 2017; Endlich et al., 2018; Kreusser et al., 2008; Serra-Millas, 2016). BDNF concentrations in the blood are widely accepted as a proxy to reflect brain BDNF levels, thus serum BDNF has been investigated as a potential biomarker for a variety of neuropsychiatric disorders (Ornell et al., 2018; Palma-Alvarez et al., 2017; Peng, Li, Lv, Zhang, & Zhan, 2018; Polyakova et al., 2015).

A growing body of evidence suggests a potential link between the BDNF system and anxiety-like behaviors. Substantial evidence comes from genetic association studies, particularly a common single nucleotide polymorphism (SNP) of the bdnf gene that results in a valine (Val) to methionine (Met) substitution (Val66Met), leading to impaired BDNF secretion (Egan et al., 2003; Notaras, Hill, & van den Buuse, 2015). Preclinically, BDNF Val66Met mice are prone to exhibit increased anxiety-like temperaments (Bath et al., 2012; Chen et al., 2006; Dincheva et al., 2017; Soliman et al., 2010; Yu et al., 2012). Clinically, the majority of studies demonstrate an association between the BDNF Val66Met polymorphism and high anxiety levels (Cagni et al., 2017; Gonzalez-Castro et al., 2019; Jiang et al., 2005; Kang et al., 2019; Keyan & Bryant, 2019; Montag, Basten, Stelzel, Fiebach, & Reuter, 2010; Wei et al., 2012; Wei et al., 2017), while a few reports suggest that carriers of the Val/Val allele may be more susceptible to anxious behaviors (Lang et al., 2005; Min et al., 2013; Nestor et al., 2019).

Beyond genetic association studies, reduced BDNF expression has been linked to increased anxiety-like behaviors in rodents (Autry, Adachi, Cheng, & Monteggia, 2009; Kikusui et al., 2019; Uysal et al., 2011), and pre-treatment with BDNF or a BDNF receptor agonist amended anxiety in early-weaned mice (Kikusui et al., 2019). In a recent study, deletion of the bdnf gene in mice led to enhanced anxiety levels, while viral BDNF expression in the amygdala reversed this effect, probably through regulation of GABAergic neurotransmission (Xie et al., 2019). Deletion of TrkB, thereby leading to diminished BDNF signaling, was also found to be associated with higher anxiety levels (Bergami, Berninger, & Canossa, 2009; Bergami et al., 2008). It should be noted that some studies have found different results, i.e., higher BDNF expression in the context of anxiety (Berardino et al., 2019; Monteggia et al., 2007; Rankov Petrovic et al., 2019). A systemic review and meta-analysis of human studies concluded that BDNF levels are lower in individuals afflicted with anxiety disorders, but this notion appears to be influenced by the specific type of anxiety disorder (Suliman, Hemmings, & Seedat, 2013).

Regarding alcohol addiction and other related outcomes, the BDNF system appears to have a complex and variable interaction with alcohol, depending on the quantity of alcohol consumption and severity of alcohol dependence. In rats, moderate (10%) ethanol consumption leads to increased BDNF levels in the dorsal striatum, resulting in the attenuation of alcohol intake by downstream activation of the mitogen-activated protein kinase (MAPK) extracellular signal-regulated kinase 1/2 (ERK1/2) (Jeanblanc et al., 2009; Jeanblanc, Logrip, Janak, & Ron, 2013; Logrip, Barak, Warnault, & Ron, 2015; Logrip, Janak, & Ron, 2008). An upregulation of BDNF signaling pathways was also observed in the hippocampus following conditions of chronic and moderate ethanol intake in C57BL/6J mice (Stragier et al., 2015). By contrast, long-term excessive alcohol intake leads to reduced BDNF mRNA expressions in cortical brain regions, including the medial prefrontal cortex (mPFC) (Darcq et al., 2015; Fernandez, Lew, Vedder, & Savage, 2017; Logrip, Janak, & Ron, 2009; Orru et al., 2016; Tapocik et al., 2014). Moreover, a negative correlation between serum BDNF concentrations and alcohol intake was found in Long-Evans rats after intermittent access to 20% alcohol using a 2-bottle paradigm (Ehinger et al., 2019). While BDNF homozygous knockout (KO) mice do not survive beyond the first postnatal week (Ernfors, Lee, & Jaenisch, 1994), BDNF heterozygous KO mice (expressing approximately 50% of the BDNF gene) consume more alcohol and show enhanced alcohol conditioned place preference, compared to their wildtype counterparts (Hensler, Ladenheim, & Lyons, 2003; McGough et al., 2004). Mice carrying the orthologue of the human Met66BDNF polymorphism (i.e., Met68BDNF) show excessive and compulsive alcohol drinking habits, which are reversed by overexpression of the wildtype Val68BDNF allele in mPFC of Met66BDNF mice or by treating the Met66BDNF mice with a TrkB agonist (Warnault et al., 2016).

Previous reports on peripheral BDNF concentrations in AUD individuals are not conclusive, with some studies showing no significant findings (Costa, Girard, Dalmay, & Malauzat, 2011; Heberlein et al., 2010), while others noting lower peripheral BDNF concentrations in AUD individuals compared to non-AUD controls (Zanardini et al., 2011; Zhang et al., 2018). In line with the preclinical data, a recent systematic review and meta-analysis found that AUD individuals typically show a global decrease in BDNF levels (Ornell et al., 2018). More recently, Zhou and colleagues reported that peripheral BDNF and TrkB levels were negatively correlated with average daily alcohol consumption (Zhang et al., 2018). In addition to peripheral measurements, human genetic analyses have also found a link between the BDNF system and AUD. The BDNF gene polymorphism Val66Met has been associated with higher risk of relapse, as well as earlier onset and greater severity of AUD (Matsushita et al., 2004; Wojnar et al., 2009). Interestingly, while the majority of studies report that human carriers of the Val66Met polymorphism were found to be more anxious and consume higher amounts of alcohol (Andreatta et al., 2019; Colzato, Van der Does, Kouwenhoven, Elzinga, & Hommel, 2011; Gonzalez-Castro et al., 2019; Zou et al., 2019), it is also notable to impart the presence of studies that found no association between the BDNF Val66Met polymorphism and AUD (Grzywacz, Samochowiec, Ciechanowicz, & Samochowiec, 2010; Nedic et al., 2013). Given that AUD and anxiety are highly comorbid (Gimeno et al., 2017; Smith & Randall, 2012), and BDNF appears to be linked to both conditions, it is plausible to study the potential role of the BDNF system in the shared neurobiology of alcohol- and anxiety-related behaviors. The aim of the present study was to investigate the relationship between serum BDNF concentrations and behavioral measures in a sample of alcohol-dependent individuals with high trait anxiety.

2. Methods

2.1. Participants

This was a secondary study utilizing blood samples and data from a previously reported human laboratory study (ClinicalTrial.gov – NCT01751386) (Farokhnia et al., 2017). Participants were recruited through general flyers and advertisements in the community. Upon completing an initial phone screen, potentially eligible volunteers were invited to the National Institute on Alcohol Abuse and Alcoholism (NIAAA) outpatient clinic at the National Institutes of Health (NIH) Clinical Center in Bethesda, Maryland to complete a screening visit. Based on the information provided at screening, participants who were deemed eligible took part in a randomized, double-blind, placebo-controlled human laboratory study with baclofen, the details of which have been reported in the parent study (Farokhnia et al., 2017). Briefly, participants were of general good health according to medical history, physical examination, and blood/urine lab tests. They were included in the study if they had a current diagnosis of alcohol dependence according to the Structured Clinical Interview for DSM-IV-TR (SCID), a high trait anxiety score (≥40) recorded from the State-Trait Anxiety Inventory (STAI), and were not seeking treatment for anxiety or alcohol dependence, an approach consistent with the NIAAA Council guidelines on administering alcohol in human studies (NIAAA). Individuals were excluded if they had a current diagnosis of substance dependence (other than alcohol and nicotine), lifetime diagnosis of schizophrenia, bipolar disorder, or other psychoses, lifetime history of attempted suicide, a diagnosis of major depressive disorder within the past 6 months, history of epilepsy or alcohol-related seizures, positive urine test for drugs of abuse at any time during the study, clinically significant medical problems, or current use of psychotropic medications that could not be discontinued. The full lists of inclusion and exclusion criteria can be found in the parent study (Farokhnia et al., 2017) and in the supplementary information section (Appendix 1).

Thirty-nine subjects met eligibility criteria and were enrolled in the parent study (Farokhnia et al., 2017). Of these individuals, 32 participants had stored serum samples available for BDNF measurement. One participant had current depressive disorder NOS (not otherwise specified), which was not among the exclusion criteria and is listed in Table 1 as ‘mood disorder’. We excluded, from the present analyses, one participant for having a positive urine drug test for opiates after the consent (also withdrawn from the parent study) and two other participants for failure to arrive to the clinic overnight-fasted, resulting in a final sample size of n = 29 (Figure 1). Most the participants were male (79.3%) and African-American (75.9%). Further details on sample demographics and other baseline characteristics can be found in Table 1.

Table 1:

Demographic Characteristics and Baseline Data

Sample Size (N)	29
Age, Years, Mean ± SD	43.6 ± 9.7
African-American, n (%)	22 (75.9%)
European-American, n (%)	5 (17.2%)
Multi-Racial, n (%)	2 (6.9%)
Male, n (%)	23 (79.3%)
Female, n (%)	6 (20.7%)
Years of Education, Mean ± SD	12.2 ± 3.7
BMI, kg/m2, Mean ± SD	29.1 ± 4.7
Current Diagnosis of Anxiety Disorders, n (%)	5 (17.2%)
Current Diagnosis of Mood Disorders, n (%)	1 (3.4%)
Current Tobacco Use, n (%)	15 (51.7%)
Baseline OCDS Score, Mean ± SD	16.7 ± 7.0
Baseline QSU-B score, Mean ± SD	41.4 ± 18.8
Baseline ADS Score, Mean ± SD	14.0 ± 6.6
Baseline State STAI Score, Mean ± SD	36.7 ± 8.9
Total Drinks (Past Week), Mean ± SD	37.2 ± 26.5
Drinks/Drinking Days (Past Week), Mean ± SD	7.0 ± 3.8
Drinking Days (Past Week), Mean ± SD	5.0 ± 2.0
BDNF concentration, pg/ml, Mean ± SD	6.6 ×104 ± 2.2 × 104
BDNF concentration corrected for TP, pg/ml, Mean ± SD	8.7 × 10−9 ± 3.0 × 10−9

SD: standard deviation; BMI: body mass index; QSU-B: brief questionnaire of smoking urges; OCDS: obsessive-compulsive drinking scale; ADS: alcohol dependence scale; STAI: state-trait anxiety inventory; BDNF: brain-derived neurotrophic factor; TP: total protein.

Figure 1:

Flow diagram detailing the number of participants screened and final sample of this analysis.

2.2. Study Procedures

Participants were instructed to refrain from eating or drinking caloric beverages overnight (starting at midnight) before coming to NIAAA for this first study visit. Participants had to abstain from alcohol for at least 24 hours prior to the visit and had to pass a breathalyzer test (BrAC = 0.00) before the consent process began. After providing written informed consent, a negative urine drug screen and a Clinical Institute Withdrawal Assessment for Alcohol-revised (CIWA-Ar) (Sullivan, Sykora, Schneiderman, Naranjo, & Sellers, 1989) score of ⩽8 were required to proceed. Vital signs were taken, a fasting blood sample was drawn, and after consuming a light breakfast, participants completed a battery of questionnaires. More specifically, alcohol and nicotine craving were assessed using the Obsessive-Compulsive Drinking Scale (OCDS) and the Brief Questionnaire of Smoking Urges (QSU-B), respectively. Severity of alcohol dependence was measured using the Alcohol Dependence Scale (ADS) and drinking data was collected using Timeline Follow-back (TLFB). State anxiety level was recorded using the state anxiety version of the STAI. Unlike trait anxiety, which looks at trait that is relatively stable over time, state anxiety reflects the intensity of anxious feelings the person is experiencing at the moment (Leal, Goes, da Silva, & Teixeira-Silva, 2017). The complete list of assessments is detailed in the parent study (Farokhnia et al., 2017). For the purpose of this study, we analyzed serum BDNF concentrations and behavioral measures collected on the first visit, therefore, before participants were randomized to receive the study drug.

2.3. Blood Collection, Processing, and Assays

Blood samples for BDNF measurement were collected under fasting conditions. This was performed at approximately the same time of the day for participants (around 9:30 am), considering that BDNF follows a circadian rhythm (Choi, Bhang, & Ahn, 2011; Pluchino et al., 2009). Blood was drawn into serum tubes containing a gel separator and clot activator. After allowing blood to clot for 30 minutes at room temperature, the tubes were centrifuged for 10 minutes at 1000 g and 25°C. Serum was aliquoted into microtubes and stored at −80°C until they were analyzed. We used serum because, compared to plasma, it appears to produce less variability in peripheral BDNF concentrations (Polacchini et al., 2015). BDNF concentrations were measured using an enzyme-linked immunosorbent assay (ELISA) (Promega; Madison, WI, USA). Briefly, ELISA 96-well plates were coated with BDNF antibodies in carbonate coating buffer. Columns on the ELISA plates were prepared for the BDNF standard curve. Final concentrations within the plate were 0–500 pg/ml BDNF. Serum samples were diluted with Block and Sample buffer 1:500 (Promega; Madison, WI, USA) and added in wells, all in quadruplicate. After incubating the plate for 4 hours at room temperature with shaking (400 ± 100 rpm), the plate was washed and secondary anti-human BDNF antibodies were added, before incubating overnight at 4°C with shaking. The plate was then washed with buffer and the appropriate horseradish peroxidase conjugates were added, before an incubation period of one hour with shaking (400 ± 100 rpm). Next, the plate was washed with buffer, tetramethylbenzine was added to each well, and allowed to incubate briefly. The reaction was then stopped by adding 1 N hydrochloric acid. The absorbance was recorded at 450 nm on a plate reader within 30 min of stopping the reaction. Quadruplicates were averaged to calculate a mean BDNF concentration level.

To control for baseline differences in expression levels, BDNF concentrations were also corrected for total protein (Polacchini et al., 2015). Total protein concentrations were measured using blood samples collected during the screening visit prior to the first study visit. Participants were asked to fast overnight the day before their screening visit, and blood samples were collected in the morning after participants consented to the screening protocol. Blood was drawn into lithium heparin tubes and was processed at the Department of Laboratory Medicine, National Institutes of Health Clinical Center, Bethesda, MD, using a Cobas 6000 Analyzer (Roche Diagnostics International Ltd., Switzerland).

2.4. Statistical Analysis

Descriptive statistics were used to summarize baseline and demographic data. Spearman rank correlation coefficient, r, was calculated between log BDNF concentrations, and each of the following variables: OCDS score, QSU-B score, ADS score, total number of drinks over the past week, number of drinks per drinking day over the past week, number of drinking days over the past week, and state STAI score. The same analyses were conducted with BDNF values corrected for total protein (%TP). BDNF levels were also compared between smoking and non-smoking individuals using the Mann-Whitney U Test. Given that BDNF levels are known to decrease with aging (Lommatzsch et al., 2005) and that platelets are a main source of peripheral BDNF (Bus et al., 2011; Karege et al., 2005), we also explored whether BDNF concentrations in this sample may be correlated with age and/or serum platelet count. Finally, we conducted bivariate correlations between total protein concentrations and behavioral measures to make sure our results are primarily driven by BDNF and not total protein concentrations. P-values <0.05 were considered statistically significant. Analyses were performed in SAS 9.3 (SAS Institute, Cary NC). Graphs were produced using GraphPad Prism 8.0 software (GraphPad, San Diego CA).

3. Results

3.1. Correlations between Serum BDNF Concentrations and Behavioral Measures

The bivariate Spearman Correlations between serum BDNF concentrations and behavioral measures are detailed in Table 2. BDNF concentrations were negatively correlated with average drinks per drinking days during the past week (r = −0.41, p = 0.02) (Figure 2a). Moreover, BDNF concentrations were positively correlated with OCDS scores (r = 0.48, p = 0.007) (Figure 2b), as well as state STAI scores (r = 0.52, p = 0.003) (Figure 2c). Similar results were shown for BDNF concentrations corrected for total protein (Table 2). No other significant correlations were found (p’s ≥ 0.05). BDNF concentrations were not significantly different between smoking and non-smoking individuals (U = 89, p = 0.48).

Table 2:

Bivariate Spearman Correlations Between Serum BDNF Concentrations and Behavioral Measures

	Log BDNF Concentration	Log BDNF (%TP) Concentration
Baseline OCDS Score	r = 0.48 p = 0.007	r = 0.49 p = 0.005
Baseline QSU-B Score	r = −0.12 p = 0.66	r = −0.09 p = 0.74
Baseline ADS Score	r = 0.35 p = 0.05	r = 0.32 p = 0.08
Baseline State STAI Score	r = 0.52 p = 0.003	r = 0.49 p = 0.006
Total Drinks (Past Week)	r = −0.30 p = 0.11	r = −0.28 p = 0.12
Drinks/Drinking Days (Past Week)	r = −0.41 p = 0.02	r = −0.40 p = 0.03
Drinking Days (Past Week)	r = −0.02 p = 0.88	r = −0.005 p = 0.97

ADS: alcohol dependence scale; BDNF: brain derived neurotrophic factor; OCDS: obsessive compulsive drinking scale; QSU-B: questionnaire of smoking urges – brief; STAI: state-trait anxiety inventory; TP: total protein.

Figure 2:

The relationships between serum BDNF concentrations and behavioral measures. BDNF: brain derived neurotrophic factor; OCDS: obsessive compulsive drinking scale; STAI: state-trait anxiety inventory.

BDNF concentrations were not significantly correlated with age in this study (p > 0.05). An expected positive correlation was found between BDNF concentrations and peripheral platelet count; however, the correlation reached statistical significance only for BDNF concentrations corrected for total protein (r = 0.40, p = 0.04) and not for raw BDNF concentrations (r = 0.31, p = 0.12). Total protein concentrations were not correlated with any of the behavioral measures (data not shown).

4. Discussion

The main findings of this study indicate that serum BDNF concentrations correlated negatively with alcohol drinking, and positively with alcohol craving and anxiety, in this sample of highly anxious AUD individuals who were active drinkers and not seeking treatment for their alcohol or anxiety problems.

In the published literature, a notable disparity exists in reports of peripheral BDNF concentrations in AUD individuals, which may partially be due to methodological differences across previous studies (Ornell et al., 2018). Different inclusion criteria, such as length of alcohol abstinence, severity of alcohol withdrawal symptoms, and psychological state of enrolled individuals may considerably affect peripheral BDNF levels in an AUD cohort. Despite heterogenous observations, a recent meta-analysis concluded that peripheral BDNF concentrations in humans are generally decreased following chronic exposure to alcohol (Ornell et al., 2018). The negative correlation between BDNF concentrations and alcohol drinking levels found in this study is consistent with a recent study by Zhou and colleagues, where BDNF concentrations were negatively correlated with daily amount of alcohol consumption in a sample of male patients with alcohol dependence (Zhou et al., 2018). Nevertheless, it is important to note that in the aforementioned study, participants were treated with oral oxazepam to alleviate withdrawal symptoms (Zhou et al., 2018). The sample enrolled in this study were actively drinking individuals with high trait anxiety, mild or no alcohol withdrawal symptoms, and with no pharmacological (e.g., benzodiazepines) treatment under way to ease withdrawal symptoms. These criteria led to a unique set of anxious AUD individuals that separates this study from previous studies looking at peripheral BDNF concentrations in relation to alcohol use. Different neuroadaptive processes are involved during active alcohol drinking, withdrawal, and abstinence; it is, therefore, important to differentiate between these two stages when studying neurotransmitter and neuromodulator systems such as that of BDNF (Becker & Mulholland, 2014). To our knowledge, the majority of studies on peripheral BDNF concentrations have been performed in AUD patients during withdrawal or abstinence, and thus cannot directly be compared to the results of this study (Ornell et al., 2018).

In this study, BDNF levels were positively correlated with OCDS, a measure of alcohol craving. The observed opposing directions in which BDNF levels correlated with alcohol drinking and craving may be surprising, yet not unexpected, given that craving is a complex phenomenon and self-reported measures of craving do not necessarily correlate with the amount of alcohol drinking (Haass-Koffler, Leggio, & Kenna, 2014; Swift, 1999). A recent systematic review reported that, in the majority of previous studies, craving was found to be a significant predictor of relapse (Sliedrecht, de Waart, Witkiewitz, & Roozen, 2019); however, the concept of “relapse” does not apply to our study, given that the participants were non-treatment seekers who were drinking actively and heavily before, during, and after the study. BDNF appears to play a role in incubation of craving, which is plausible since BDNF is known to act on the neurocircuitries involved in drug craving (Geoffroy & Noble, 2017). Hilburn and colleagues did not observe any significant correlation between BDNF levels and alcohol craving (Hilburn et al., 2011), whereas the results obtained by Heberlein and colleagues are in agreement with our findings, i.e., BDNF concentrations were positively associated with alcohol craving (Heberlein et al., 2016). Most of the participants in the present study had heavy drinking levels, according to TLFB, but were required to come to the clinic with a BrAC of 0.00 and, therefore, had to be abstinent for several hours. It is conceivable that the participants’ alcohol craving was higher than usual as a result of the short-term abstinence from alcohol and the potential presence of initial, albeit subclinical, symptoms of alcohol withdrawal. Given the protective effect of BDNF against alcohol-seeking behaviors and relapse (Ron & Berger, 2018), it can be speculated that endogenous BDNF concentrations are increased in response to the rebound increase in alcohol craving, hence a positive correlation between the two was observed. On the other hand, it is possible that the processes leading to enhanced alcohol craving and changes in BDNF concentrations are independent from one another. Further investigations are warranted to understand whether and how the BDNF system may regulate and/or be influenced by alcohol craving.

Although the etiological role is complex, anxiety has been associated with elevated levels of alcohol craving and consumption, together with incidence of developing and maintaining AUD (Breese, Sinha, & Heilig, 2011; Haass-Koffler et al., 2014; McCaul, Hutton, Stephens, Xu, & Wand, 2017; Sinha, 2013; Spanagel, Noori, & Heilig, 2014). Moreover, AUD patients are 2–3 times more likely to be diagnosed with an anxiety disorder than non-AUD individuals (Birrell, Newton, Teesson, Tonks, & Slade, 2015; Lai, Cleary, Sitharthan, & Hunt, 2015; Swendsen et al., 1998). While alcohol can initially serve as an anxiolytic agent, chronic drinking may enhance the risk of anxiety (SAMHSA, 2005). In this study, similar to the relationship between BDNF and alcohol craving, we also found a positive correlation between peripheral BDNF concentrations and state anxiety levels measured via STAI (Julian, 2011). While the majority of published studies indicate that disrupted BDNF signaling may be linked to enhanced anxiety levels (Bath et al., 2012; Bergami et al., 2008; Castren & Kojima, 2017; Chen et al., 2006; Dincheva et al., 2017; Kikusui et al., 2019; Soliman et al., 2010; Suliman et al., 2013; Wei et al., 2017; Xie et al., 2019; Yu et al., 2012), it is important to note the presence of other studies that have found the contrary (Berardino et al., 2019; Monteggia et al., 2007; Rankov Petrovic et al., 2019) or inconsistent results (Notaras et al., 2015). Our observation in the present study that high BDNF levels were associated with high anxiety levels is not in agreement with the majority of the published data, which may be, at least in part, due to the specific sample enrolled, i.e., alcohol-dependent individuals with high trait anxiety. Considering that a positive correlation between craving and serum BDNF levels was also observed, it could be hypothesized that state anxiety may be a proxy of (or secondary to) alcohol craving, at least within the context of the present study, resulting in similar relationships with serum BDNF concentrations. As detailed in the Methods section, we used the STAI to assess for trait and state anxiety. The trait STAI score was used as part of the eligibility criteria for the study, and participants had to have a trait STAI score of 40 or more to be deemed eligible. On the other hand, the state STAI scale is used as a variable in the analyses to assess the relationship between serum BDNF concentrations and anxiety levels. Therefore, as shown in Figure 2, it is possible to have participants with a trait STAI score of ≥ 40 (rendering them eligible for this study) presenting with a state STAI score of < 40.

Age and gender are known to influence circulating BDNF levels (Lommatzsch et al., 2005). Peripheral BDNF levels are known to decrease with age in both sexes, possibly due to age-related neurodegeneration (Oh, Lewis, & Sibille, 2016), whereas ovarian function in women may influence peripheral BDNF levels, depending on the menstrual cycle phase (Giese et al., 2014; Pluchino et al., 2009). Participants of this study ranged from young to middle-aged adults and, as expected, we did not observe a significant relationship between serum BDNF levels and age. With regards to gender, our study was underpowered to investigate possible gender differences because the sample was predominantly male (79.3 %). This limitation appears to be present in other similar studies investigating the role of BDNF in relation to alcohol use (Ornell et al., 2018; Silva-Pena et al., 2018; Zhou et al., 2018) and should be addressed in future studies.

The present study has several strengths and limitations that need to be acknowledged when interpreting the results. Key factors, such as fasting condition and time of blood collection, were standardized across participants. Our strict inclusion and exclusion criteria allowed us to select a homogeneous sample of participants, therefore reducing the risk of random variability in BDNF concentrations. Participants had no clinically-relevant problems that could have impacted peripheral BDNF levels (Pius-Sadowska & Machalinski, 2017; Serra-Millas, 2016). However, we do acknowledge that the homogeneity of the study sample may also represent a limitation, as it limits the generalizability of our finding. The sample size was small and the study did not have a healthy control group (without alcohol dependence) and/or an active comparator group of alcohol-dependent individuals with no or low anxiety levels, limiting our ability to run a case-control analysis. Another limitation was single, and not repeated, measurement of BDNF concentrations in this study. Due to BDNF’s diurnal/circadian fluctuations, especially in women, repeated BDNF measurements over a 24-hour period are recommended over single measurements (Cain et al., 2017). Lastly, this study presents a set of correlational results which reflect the possibility of associations between BDNF and the selected behavioral measures but do not establish causality. An ideal future study would be an a priori study with a larger sample size, more females, a non-AUD control group, and repeated BDNF measurements.

In conclusion, serum BDNF concentrations correlated negatively with alcohol drinking and positively with both alcohol craving and anxiety levels in actively drinking, alcohol-dependent individuals with high trait anxiety. Mechanistic studies are needed to elucidate the possible role of the BDNF system in relation to the neurobiology and consequences of alcohol- and anxiety-related behaviors.

Highlights.

• Serum BDNF concentrations were measured in anxious, alcohol-dependent individuals

• Bivariate correlations were run between serum BDNF levels and behavioral measures

• Serum BDNF levels correlated negatively with average drinks per drinking days

• Serum BDNF levels correlated positively with craving and anxiety scores