Neural substrates underlying eyeblink classical conditioning in adults with alcohol use disorders

Abstract

Background:

Excessive alcohol consumption produces changes in the brain that often lead to cognitive impairments. One fundamental form of learning, eyeblink classical conditioning (EBC) has been widely used to study the neurobiology of learning and memory. Participants with alcohol use disorders (AUD) have consistently shown a behavioral deficit in EBC. The present functional magnetic resonance imaging (fMRI) study is the first to examine brain function during conditioning in abstinent AUD participants and healthy participants.

Methods:

AUD participants met DSM-IV criteria for alcohol dependence, had at least a 10 year history of heavy drinking, and were abstinent from alcohol for at least 30 days. During fMRI, participants received auditory tones that predicted the occurrence of corneal airpuffs. Anticipatory eyeblink responses to these tones were monitored during the experiment to assess learning-related changes.

Results:

Behavioral results indicate that AUD participants showed significant conditioning deficits and that their history of lifetime drinks corresponded to these deficits. Despite this learning impairment, AUD participants showed hyperactivation in several key cerebellar structures (including lobule VI) during conditioning. For all participants, history of lifetime drinks corresponded with their lobule VI activity.

Conclusions:

These findings suggest that excessive alcohol consumption is associated with abnormal cerebellar hyperactivation and conditioning impairments.

Alcohol is one of the most widely abused drugs in the world (World Health Organization, 2014) and heavy/hazardous alcohol use often results in a broad range of physical and cognitive deficits. Adults with alcohol use disorders (AUD) (individuals who meet DSM-IV criteria for alcohol dependence) often experience long lasting neurological damage accompanied by learning and memory impairments (Rando et al., 2010, Sullivan et al., 2003, Pitel et al., 2007). A better understanding of the neural substrates that underlie these memory deficits may be achieved by taking advantage of well-understood fundamental, learning paradigms that have clear outcome predictions.

One of these paradigms is eyeblink classical conditioning (EBC), a simple, yet elegant model of learning that has contributed greatly to our knowledge of the neural correlates of learning and memory. This paradigm involves the pairing of a neutral conditioned stimulus (CS; e.g., a tone) and an unconditioned stimulus (US; e.g., a corneal airpuff). The US is often a biologically salient stimulus that is sufficient to elicit an unconditioned response (UR; e.g., a blink). Following multiple CS-US pairings, the organism learns to produce an adaptive, conditioned response (CR), such as a blink, in anticipation of the US presentation, suggesting that an association between the CS and US has been learned. This model system has been extensively studied and consequently, the neural circuitry supporting EBC has been mapped out in considerable detail. Overwhelming evidence from human and laboratory animal studies suggest that the cerebellum is critical for this form of learning (Christian and Thompson, 2003, Woodruff-Pak, 1988).

Both human and laboratory animal studies have reported that chronic, heavy alcohol consumption has detrimental effects on multiple brain regions, including the cerebellum (for reviews see Cheng et al., 2015, Zahr and Pfefferbaum, 2017). Vast functional and structural alterations in the human cerebellum as a consequence of AUD have been documented. For example, adults with AUD have diminished gray matter in the cerebellar hemispheres and vermis (Chanraud et al., 2007, Sullivan et al., 2000) and a reduction in white matter fibers (Chanraud et al., 2009). Histological studies also show significant cerebellar Purkinje cell loss in adults with AUD (Torvik and Torp, 1986, Andersen, 2004, Phillips et al., 1987). Alcohol-related damage is not restricted to structural brain changes as task-based functional magnetic resonance imaging (fMRI) measured significant changes in activation patterns as well. Cerebellar hyperactivation in adults with AUD has been reported in finger tapping, working memory, and auditory learning tasks (Parks et al., 2003, Desmond et al., 2003, Chanraud-Guillermo et al., 2009).

Given these functional and structural changes in the cerebellum and the importance of the cerebellum during EBC, it should not be surprising that adults with AUD consistently show impairments on multiple variants of the eyeblink conditioning procedure. McGlinchey-Berroth and colleagues (1995) showed that abstinent alcohol patients and amnesic Korsakoff patients were impaired at single-cue delay eyeblink conditioning, a procedure in which the CS and US overlap and coterminate. Similarly, chronic heavy alcohol intake by patients also produced deficits in a slightly more complex paradigm, single-cue trace eyeblink conditioning, a procedure that includes a temporal gap and stimulus-free period between CS offset and US onset (McGlinchey et al., 2005). Even when AUD patients were able to learn a discrimination task (i.e., learn that one CS predicts the US and another CS predicts its absence), they were not capable of relearning this task when the contingencies were reversed (Fortier et al., 2008). Finally, adults with AUD had trouble appropriately timing their CRs at longer interstimulus (ISI) intervals (McGlinchey-Berroth et al., 2002) and data from cerebellar patients suggest that the area supporting adaptively timed CRs is the superior cerebellar cortex (Gerwig et al., 2005).

Heavy alcohol consumption produces eyeblink conditioning deficits and functional and structural changes in brain substrates that support EBC. However, a direct functional relationship between these neural mechanisms and poor conditioning performance remains unexplored since only behavioral studies of human EBC have been performed in AUD participants. The present study is the first application of this model learning system and concurrent fMRI in AUD participants to characterize vulnerable neural structures that are uniquely recruited during basic learning processes.

Materials and Methods

Participants

A total of 89 adults from the greater Baltimore region were recruited and enrolled in the study. Prospective participants were provided with study information and screened for eligibility via email and phone according to standard MRI safety protocols. MRI exclusionary criteria: a history of any central nervous system disorder, a history of anticonvulsant medication in the past 3 months, treatment with antidepressant medications within the last 6 months, metal implantation in the body (e.g. pacemaker, pumps), or a history of significant closed head trauma (e.g. concussion). 59 were determined to be ineligible for various reasons (positive drug test, current drinking, and MRI incompatibility). Of the remaining 30, participants were classified as adults with AUD (n=18, 12 male) or healthy participants (n=12, 8 male). All participants provided written informed consent on the day of their first visit and were compensated for their time. Study procedures were reviewed and approved by the Johns Hopkins Institutional Review Board.

Screening

Right-handed females and males who were native English speakers between the ages of 18 and 60 with no history of neurological or psychiatric disorders were selected to participate in the study and were invited to visit the lab for further assessment. Blood Alcohol Content (BAC) was assessed via breathalyzer, and a urine sample was assessed for pregnancy and recent drug use. We used the Alco-Sensor IV handheld tester (Intoximeter Inc, St. Louis, MO) to measure breath alcohol levels from recent alcohol consumption including the past 24 hours to rule out recent drinking on the day of all experimental sessions. Biomarkers for long-term excessive alcohol consumption were also used, including blood tests for carbohydrate deficient transferrin (CDT) or Phosphatidylethanol (PEth). We tested the participant’s urine for drug use with the Reditest® Panel-Dip (Redwood Toxicology Laboratory, Santa Rosa, CA) which included tests for the following: Benzodiazepines (BZO), Cocaine Metabolite (COC150 or COC300), Marijuana (THC), Methamphetamine (M-AMP), Opiates (OPI300 or OPI2000), and Oxycodone (OXY). Participants with a positive pregnancy test, drug screen, and/or a non-zero BAC were determined ineligible to complete the study.

Eligible participants completed written assessments including the Alcohol Use Disorders Identification Test (AUDIT) (Saunders et al., 1993), Fagerstrom Test for Nicotine Dependence (Fagerstrom, 1978), Beck’s Depression Inventory-II (Beck, 1996), and Shipley’s Vocabulary and Abstraction Assessments (Zachary, 1986). The AUDIT and Fagerstrom test were used to quantify alcohol and nicotine use, respectively and the BDI-II was used to quantify depressive symptoms. Participants who scored less than 12 on the Shipley’s Vocabulary Assessment were determined ineligible. These tests were further used in combination with semi-structured interview assessments to determine eligibility. Smoking was entered as a covariate in statistical analyses.

Eligible participants then completed the Semi-Structured Assessment for the Genetics of Alcoholism (SSAGA) (Bucholz et al., 1994), Lifetime Drinking History (LDH) (Skinner and Sheu, 1982), and Time-line Follow-back (TLFB) (Sobell, 1992, Sobell et al., 1988) with a Masters-level research interviewer (J.R.M.) to further investigate prior and current drug and alcohol use as well as medical and psychiatric history. The SSAGA yielded diagnoses for alcohol abuse/dependence and other commonly occurring psychiatric disorders under a polydiagnostic system that included Research Diagnostic Criteria (RDC), DSM-III, DSM-IIIR, DSM-IV and Feighner criteria. This interview was used to rule out all current DSM-IV Axis I disorders in study participants. The LDH used structured interview prompts to quantify the following: total lifetime alcohol consumption (number of standard drinks in a lifetime), number of years consuming over 80 grams/day (5.7 standard drinks/day), length of heavy drinking (number of years since drinking level reached a mean of 80 grams/day for at least one month), age of heavy drinking onset (age drinking level reached a mean of 80 grams/daily for at least one month), maximum consumption on a single occasion, total quantity of alcohol consumed during the last six months, and date of last drink. At the time of the initial in person assessment, TLFB interview quantified alcohol and drug use on each day during the past three months. Interim TLFB interviews were conducted on each day that the participant visited the lab to monitor drinking behaviors and eligibility. All interviews were supervised and evaluated by a neuropsychologist with considerable expertise in psychological assessment and diagnosis to confirm that participants did not have any psychological disorders or substance dependencies that were exclusionary.

Eligible participants underwent physical examination and laboratory testing at the Johns Hopkins General Clinical Research Center (GCRC). The team’s physician or other qualified personnel performed a complete physical exam and a brief neurological exam designed to quantify possible cerebellar damage, including assessments for saccade accuracy, pursuit, and gaze holding, as well as the Brief Ataxia Rating Scale (BARS). A blood count, comprehensive metabolic panel, and syphilis screening were performed to detect any physical disorders that may complicate interpretation of functional brain activation measurements. AUD participants were required to be abstinent for at least 30 days prior to participation, confirmed by blood testing. A carbohydrate deficient transferrin (CDT) or Phosphatidylethanol (PEth) test was administered to detect recent heavy alcohol consumption. Prospective alcohol participants were excluded (n=0) if results were non-zero.

Eligible participants returned for a second and third visit to complete cognitive testing and brain imaging. At the beginning of each visit, participants were screened for recent drug use and alcohol consumption via urinalysis and breathalyzer. If participants reported current nicotine use, they were offered a 14 mg Nicoderm patch two hours prior to the fMRI scan. Active smokers were not expected to quit smoking prior to the EBC testing, but to better control nicotine levels, they were instructed to not smoke on the day of testing and offered a nicotine patch. If they elected to use the patch, they were assessed for contraindications, and a carbon monoxide (CO) test was conducted to determine a baseline CO level on those who requested the patch. Participants completed cognitive testing with the team’s neuropsychologist or other qualified personnel to assess IQ, memory (or possible dementia), working memory, executive function, and motor performance. The following tests were conducted: Trail Making, Initial Letter Fluency, Hopkins Adult Reading Test, RBANS, Digit Symbol (WAIS-IV), Visual Puzzles (WAIS-IV), Digit backwards (WAIS-IV), Brief Test of Attention, and Wisconsin Card Sorting Test (WCST) (modified by Nelson, only 48 cards). Participants were then escorted to the Kirby Center for Functional Brain Imaging to receive fMRI scanning, and a nicotine patch was administered if necessary. If a participant received a CO test earlier in the day, they received a second CO test to insure that covert cigarette smoking had not occurred prior to the nicotine patch administration.

Behavioral Apparatus

The EBC equipment used in the present study was identical to that used in our previous studies (Cheng et al., 2008, Cheng et al., 2010). Stimulus presentation was controlled, and behavioral data were recorded using a laptop computer interfaced to an NI USB-6218 data acquisition module running custom software developed under LabView version 7.1 (National Instruments, Austin, TX). Presentation of auditory stimuli was achieved by using MRI compatible pneumatic headphones (MRA, Inc.). A movie (Charlie Chaplin’s The Gold Rush) was projected onto a back illuminated screen and viewed through an adjustable mirror attached to the head coil. Air puff delivery was gated by a solenoid valve (Asco, Florham Park, NJ), and an MRI-compatible fiber-optic probe (RoMack Inc., Williamsburg, VA) measured the reflectance of infrared light from the left eye (Miller et al., 2005). The fiber-optic probe and air nozzle was housed and connected to the end of a Loc-Line hose system (Lockwood Products, Inc, Lake Oswego, OR), allowing for flexibility in targeting the left eye. This was attached to a headband such that all headgear was anchored to the participant’s head to minimize effects of head motion.

Stimuli

The CS was a 1000 Hz tone (95 dB) delivered binaurally and the US was a 5 psi airpuff (measured at the delivery site). The CS lasted 850 ms and coterminated with a 100 ms corneal airpuff to the left eye. Trials were grouped into blocks: 9 trials/blocks, 2 sec/trial, and a 3-9 (average of 6) sec intertrial interval (ITI), such that each block lasted 50 sec (Figure 1). Blocks were separated by 14 sec rest periods. Participants received four blocks of tone alone and four blocks of airpuff alone interleaved (Unpaired) and 16 blocks of tone-airpuff presentations (Paired). The scanning session lasted 1600 sec or 26.7 minutes. The variable ITI in the current paradigm is similar to the ITI we used previously (Cheng et al., 2014, Cheng et al., 2017) and was selected to ensure a sufficient number of trials to permit analysis of the fMRI data. These temporal parameters were also selected to provide enough trials to permit conditioning within a limited time period and to ensure participant comfort.

Figure 1.

Experimental Design. All participants received presentations of tones (95 dB) and airpuff (5 psi). During the Unpaired Session, tones and airpuff presentations were presented separately. During the Paired Session, tone-airpuff pairings were presented such that the airpuff coterminated with the tone.

MRI Scanner

Scanning was performed on a Philips Achieva 3-Tesla scanner with a 32-channel head coil. Functional images were collected by using a T2*- weighted gradient echo planar imaging (EPI) pulse sequence. Twenty oblique (perpendicular to the axis of the hippocampus) slices (3 × 3 × 6 mm) were collected (TR: 2000 ms, TE: 30.0 ms; FOV, 22 cm; flip angle, 61°) in a series of 800 sequential images (for a total of 1600 sec). Structural images were acquired by using a T1-weighted magnetization prepared rapid acquisition gradient echo (MPRAGE) pulse sequence.

Behavioral Data Analysis

The criterion for a response to be considered a CR is that the difference between the maximum and minimum responses in a 500 ms pre-US time window must exceed four times the standard deviation of the mean of the baseline period (250 ms before CS presentation). A time window of 500 ms pre-US presentation was selected to minimize the inclusion of voluntary and alpha responses as CRs (Spence and Ross, 1959). Performance was expressed as % CR. Peak response latencies relative to CS onset during this 500 ms time window were also calculated as an additional behavioral measure of conditioning, as they indicate at which point in time maximal eyelid closure occurs. A post-experimental questionnaire consisting of seven true/false statements (Manns JR et al. 2000) was given to participants to probe their conscious awareness of the CS-US relationship. Participants were classified as aware if they answered six or more questions correctly and unaware if they answered fewer than six questions correctly. Statistical analyses were performed with SPSS. With smoking as a covariate, a Group (AUD vs. healthy) x Awareness (Aware vs. Unaware) x Block (1-5) repeated-measures analysis of variance (ANOVA) was performed on % CR and the latency of the peak response. Awareness was included as a factor because several reports suggest that awareness may account for some variability in CRs (Knuttinen et al., 2001, Lovibond et al., 2011, Weidemann et al., 2016).

Imaging Data Analysis

Structural and functional imaging preprocessing and statistical analyses were performed with Statistical Parametric Mapping (SPM8) software (Wellcome Department of Cognitive Neurology, London, UK). Preprocessing included motion correction, slice scan time correction, structural data coregistration, normalization, and spatial smoothing. Normalization was performed with the gray and white matter segmentation-based method implemented in SPM (Ashburner and Friston, 2005). This type of normalization was chosen because it yields cerebellar output that closely corresponds to that of the spatially unbiased infra-tentorial template (SUIT) of the human cerebellum and brainstem (Diedrichsen et al., 2009, Diedrichsen, 2006). EPI functional images were realigned and resliced correcting for minor motion artifacts, and structural images were coregistered to the mean motion-corrected functional image for each participant. Whole brain structural and functional data were transformed into standard stereotaxic space according to the Montreal Neurological Institute (MNI) protocol. A Talairach transformation was also performed and used to report non-cerebellar whole brain coordinates (Talairach and Tournoux, 1988, Lancaster et al., 2007). For the cerebellum, structural and functional data were isolated and normalized into standard stereotaxic space using SUIT (Diedrichsen, 2006). Functional images were spatially smoothed with a Gaussian filter (full-width half-maximum 5 mm) and temporally high-pass filtered at 128 sec. Visual inspection of the EPI functional images was performed to monitor for excessive head movement. Furthermore, motion was modeled using the CONN functional connectivity toolbox (Whitfield-Gabrieli and Nieto-Castanon, 2012), and there were no significant group differences in number of motion artifacts (p = 0.4). In a block design analysis, the general linear model was used to estimate individual subject activations (4 blocks of tone alone, 4 blocks of airpuff alone, and 16 blocks of paired tone-airpuffs), and a random-effects analysis was conducted on all participants in SPM8. Estimated mean beta weights were used as a measure of brain activation.

Contrasts (Paired vs. Unpaired) were designed to examine brain activity changes as a function of associative learning and to account for activations related to motor performance of the unconditioned response within each group. Between-group contrasts were also performed to examine brain activity changes as a function of diagnoses (AUD relative to healthy). In addition, regression analyses on cerebellar activity was performed using the reported number of lifetime drinks, % CR, and smoking (pack years) as covariates. To control for Type I error, Monte Carlo simulations (Forman et al., 1995) were performed, which indicated that activation clusters of at least 25 (whole brain) or 10 (cerebellum) voxels were significant at a p < 0.001 level (Cheng et al., 2017). Activation clusters surviving this threshold are reported in tabular form.

Results

Sample Characteristics

Demographic, drinking and smoking history, and screening assessment data are summarized in Table 1. No significant group differences in age, sex distribution, and education were observed (p’s > 0.05). There was a significantly larger percentage of black participants in the healthy participants (75%) relative to AUD participants (22%). AUD participants drank significantly more than healthy participants as indicated by the total number of years of frequent heavy drinking (t(28) = 4.20, p = 0.0002) and total number of drinks consumed over the lifespan (t(28) = 5.65, p = 0.00002). AUD participants were abstinent for a mean of 94.23 months (range: 1.52 – 312.0, SD = 111.16). Smoking history data showed that AUD participants also smoked more than healthy participants as indicated by packs per day (t(28) = 2.61, p = 0.02) and pack years (t(28) = 2.17, p = 0.04), but not by total packs last year (t(28) = 0.38, p > 0.05). Finally, AUD participants scored significantly higher on the AUDIT (t(28) = 5.08, p = 0.00008) and BDI-II (t(28) = 2.38, p = 0.03).

Table 1.

Sample characteristics

	Healthy (n =12)	AUD (n =18)	t or X2
Demographics
Age at scan (years)	46.25 (6.02)	50.74 (7.09)	1.89
Sex (% male)	75.0	66.7	0.24
Education (years)	13.46 (2.17)	14.89 (2.83)	1.56
Race (% Black)	75.0	22.2	8.17**
Ethnicity (% Hispanic)	0.0	5.6	0.69
Drinking History
Total number of years drinking	22.83 (11.05)	25.58 (11.51)	0.66
Total number of years frequent heavy drinkinga	2.5 (6.29)	14.11 (8.86)	4.20***
Total number of drinks consumed (lifespan)	4687.73 (5265.45)	49573.18 (33101.19)	5.65***
Total number of months abstinent	n/a	94.23 (111.16)	n/a
Smoking History
Smokers (% non-smokers)	50.0	27.8	1.53
Total number of years smoking	11.33 (13.39)	14.72 (13.20)	0.68
Total number of years abstinent	1.17 (4.04)	4.89 (9.20)	1.51
Packs per day	0.19 (0.21)	0.54 (0.51)	2.61*
Pack years	4.40 (5.80)	12.49 (14.15)	2.17*
Total packs last year	64.01 (80.80)	79.01 (132.51)	0.38
Screening Assessments
AUDITa	1.83 (1.47)	17.33 (12.81)	5.08***
BDI-IIb	1.83 (3.38)	9.28 (12.59)	2.38*
Fagerstromc	1.5 (2.11)	2.39 (2.36)	1.08

^*p < .05

^**p < .01

^***p < .001 Values are Mean (SD) or Percent.

^aFrequent heavy drinking is defined by the Lifetime Drinking History Assessment as 15 or more days of drinking per month

^bAlcohol Use Disorders Identification Test (Saunders et al., 1993)

^cBeck Depression Inventory-II (Beck et al., 1996)

^dFagerstrom Test for Nicotine Dependence (Fagerstrom 1978)

Behavioral

% Conditioned Responses (Figure 2A):

Figure 2.

Behavioral Results. A) Healthy participants produced significantly more conditioned responses relative to AUD participants at the end of the experiment (Blocks 13-16) and overall (Blocks 1-16; bar graphs). Healthy participants also demonstrated learning-related changes as evidenced by more conditioned responses at the end of the experiment (Blocks 13-16) relative to tone alone presentations (U : unpaired). B) Healthy participants’ peak responses occurred significantly later (closer to US onset) than AUD participants’ peak responses by the end of the experiment (Blocks 13-16). Overall latencies throughout the experiment (Blocks 1-16; bar graphs) showed a trend for healthy participants’ to produce peak responses significantly later than AUD participants (p = 0.08). Healthy participants also showed learning-related changes by producing peak responses closer to US onset at the end of the experiment (Blocks 13-16) relative to tone alone presentations (U : unpaired). C-D) Number of lifetime drinks was negatively correlated with % CR and latency measures, indicating that participants who drank more also demonstrated poorer conditioning. US: unconditioned stimulus, CR: conditioned response. Error bars represent standard error of the mean.

A 2 (Group) x 2 (Awareness) x 5 (Block) repeated-measures ANOVA was performed on % CRs with smoking as a covariate (# of pack years). There were significant main effects for Group (F(1,25) = 7.69, p = 0.01), Awareness (F(1,25) = 4.51, p = 0.044), and Block (F(4,100) = 3.16, p = 0.017). A significant interaction of Block x Awareness (F(4,100) = 3.13, p = 0.018) and trend of Block x Group (F(4,100) = 2.21, p = 0.073) were also detected. Post hoc comparisons (as indicated with *p < 0.05) showed that relative to AUD participants, healthy participants: 1) produced more conditioned responses overall (t(28) = 1.94, p = 0.031) (Figure 2A bar graphs), 2) showed greater % CRs relative to AUD participants during the last block of paired trials (t(28) = 2.52, p = 0.009), and 3) demonstrated learning-related changes over time as shown by more % CRs at the end of the experiment relative to unpaired tone alone presentations (t(11) = 1.92, p = 0.041), while the AUD participants did not.

Latency of Peak Responses (Figure 2B):

A 2 (Group) x 2 (Awareness) x 5 (Block) repeated-measures ANOVA was performed on Latency of Peak Responses (relative to CS onset) with smoking as a covariate (# of pack years). A significant main effect of Block (F(4,100) = 3.59, p = 0.009) and Block x Awareness interaction (F(4,100) = 3.27, p = 0.015) were observed. Post hoc comparisons (as indicated with *p < 0.05 and ^Ɨp < 0.10) showed that relative to AUD participants, healthy participants: 1) produced maximal eyelid closure responses closer to US onset (t(28) = 1.84, p = 0.038) at the end of the experiment (Blocks 13-16) and 2) showed learning-related changes over time as shown by a significant shift of their peak responses closer to US onset relative to unpaired tone alone presentations (t(11) = 2.08, p = 0.031), and 3) showed a trend in overall peak latencies (t(28) = 1.46, p = 0.077) (Figure 2B bar graphs).

Correlation analyses showed that participants’ number of lifetime drinks correlated negatively with their Latency of Peak Responses (r = −0.42, p = 0.02) and approached significance with % CR (r = −0.31, p = 0.09), suggesting that those who conditioned poorly also consumed the most lifetime drinks (Figures 2C and 2D). A restricted analysis of just the AUD participants showed that there was a significant negative correlation between number of lifetime drinks and Latency of Peak Responses (r = −0.60, p = 0.009) but not with % CR (r = −0.29, p = 0.25).

Nonasssociative Measures:

No significant differences between AUD and healthy participants were found for unconditioned response magnitudes, hearing thresholds, correctly answered movie questions, and awareness levels (all p’s > 0.2), suggesting that the learning deficits exhibited by the AUD participants were not due to impairments in US processing, hearing ability, or general attention.

Neuroimaging

Increased activation was found in a number of cerebellar regions in AUD compared to healthy participants, but not vice versa (see Table 2). The AUD group showed significantly greater activity in the inferior and posterior cerebellum, such as right Crus I (44, −82, −36), right Crus II (20, −90, −39), left lobule VIIb (−34, −53, −41) and right Vermis VIIIa (1, −73, −44). Significantly greater activation in AUD participants was also detected in left lobule VI (−34, −60, −26) (Figure 3A), a key region for eyeblink conditioning. Analyses of non-cerebellar regions revealed increased activation in only one structure, the right parahippocampal gyrus (29, −45, −4; cluster size: 13 voxels; z-value: 3.3) in AUD participants. Similar to cerebellar findings, no greater activation was found in healthy participants in the cerebrum.

Table 2.

Cerebellar differences between AUD and healthy participants

AUD > Healthy
X	Y	Z	SPM {Z}	N Vox	Brain region
44	−82	−36	4.2	374	R. Crus I
1	−73	−44	4.2	220	R. Vermis VIIIa
20	−90	−39	3.4	180	R. Crus II
48	−65	−38	3.5	91	R. Crus I
−34	−60	−26	3.6	70	L. Lobule VI
−34	−53	−41	3.6	43	L. Lobule VIIb
44	−72	−46	3.2	27	R. Crus II
16	−61	−51	3.3	22	R. Lobule VIII
Healthy > AUD
X	Y	Z	SPM {Z}	N Vox	Brain region
none

MNI cerebellar coordinates of activation maxima (Schmahmann et al., 2000) in AUD and Healthy participants during eyeblink conditioning. Regions listed were thresholded at a minimum cluster size of 10 voxels and z-scores of p < 0.001

Figure 3.

ROI analyses of lobule VI. A) Significant differential activation between AUD and healthy participants was detected in this region of cerebellar lobule VI. B-C) This difference was largely driven by greater responding by the AUD participants during Blocks 1 and 4. D-E) Only healthy participants showed a significant positive correlation between conditioned responses and activation in lobule VI. Error bars represent standard error of the mean.

Given the importance of lobule VI in human and laboratory animal studies, a region of interest (ROI) analysis was performed using an ROI that was functionally defined. Figure 3B shows that, when the activation cluster (left lobule VI from Table 2) was used to sample activity (peak betas) from individual participants, this differential responding is largely driven by greater activation in AUD participants (t(28) = 2.97, p = 0.006) and using this ROI to further sample activity across blocks, greater activity was measured during the first (t(28) = 2.76, p = 0.010) and last (t(28) = 2.19 p = 0.037) blocks of paired trials by the AUD participants (Figure 3C).

Laboratory animal and human studies have shown that the interposed nucleus, ipsilateral to the trained eye, is essential for EBC (Christian and Thompson, 2003, Thurling et al., 2015). For this reason, structural ROI analyses using cerebellar deep nuclei ROIs from the SUIT atlas (Diedrichsen et al., 2011) were performed. No significant group differences in peak responses were observed in the left dentate, left fastigial, or left interposed nuclei (all p’s > 0.05).

Brain-Behavior Correlations

Given the overwhelming amount of literature that supports the position that cerebellar lobule VI is important for successful EBC and the significant cluster of activation in lobule VI found in the current study, additional analyses focusing on the relationship between behavioral conditioned responses and activity within this region were performed. Individual participant’s peak activation values within left lobule VI were compared to their % CRs. Healthy participants showed a significant positive correlation between % CRs and lobule VI activity (r = +0.72, p = 0.008) while AUD participants (r = +0.02, p = 0.93) failed to show any relationship between responding in this region and their % CRs (Figures 3D and 3E). A Fisher r-to-z transformation was used to determine that the difference between these two correlation coefficients was statistically significant (p < 0.035).

A significant positive correlation (r = +0.47, p = 0.009) between lobule VI activity (as constrained by the activation cluster reported in Table 2) and the number of lifetime drinks reported by each participant was also observed (Figure 4A). Further confirming this pattern of results, an independent regression analysis on cerebellar activity was performed using the number of lifetime drinks as a predictor of interest and showed that greater lifetime alcohol consumption was associated with greater activations in an almost identical region in lobule VI (Figure 4B). No significant clusters of activation were observed following regression analyses using % CRs and smoking (# of pack years) as covariates.

Figure 4.

Relationship between the number of lifetime drinks and lobule VI activation. A) Number of lifetime drinks was positively correlated with lobule VI activation cluster (AUD > healthy; Table 2). B) An independent regression analysis produced a significant cluster of activity in lobule VI that showed a positive correlation with number of lifetime drinks. Higher number of lifetime drinks predicted greater activation in this region of lobule VI. Color bar indicates magnitude of z-scores.

Discussion

A history of heavy/hazardous alcohol consumption can adversely affect brain function leading to cognitive impairments. Improved diagnosis and treatment of these cognitive deficits may come from a better understanding of the neural circuitry underlying fundamental memory tasks, as they represent a foundation upon which more complex cognition is built. One fundamental form of learning is EBC and this is the first study to directly investigate how long term heavy drinking affects neurological function in regions that support EBC. Relative to healthy participants, AUD participants showed impaired EBC on two indices of learning: % CR and peak response latencies. Correlation analyses suggest that this deficit may be driven by the extent of lifetime alcohol consumption. Despite this learning impairment, AUD participants demonstrated greater activation in several cerebellar regions. Specifically, a higher number of lifetime drinks corresponded with higher lobule VI activation. Activity in this area corresponded with % CRs in healthy but not AUD participants. These findings suggest that excessive alcohol consumption is associated with abnormal cerebellar hyperactivation and conditioning impairments.

Behavior

Since the cerebellum is particularly vulnerable to chronic alcohol exposure and also critically involved in EBC, it was not surprising to find behavioral learning deficits associated with heavy alcohol consumption across species. Using a range of conditioning paradigms and laboratory animal models, behavioral EBC deficits resulting from pre and neonatal alcohol exposure have been extensively reported (Green et al., 2006, Brown et al., 2007, Murawski et al., 2013). Human eyeblink conditioning deficits stemming from alcohol exposure have also been well documented in investigations of both children (Jacobson et al., 2008, Jacobson et al., 2011) and adults (Fortier et al., 2009, McGlinchey-Berroth et al., 2002). For example, children exposed to alcohol prenatally did not condition as well as unexposed children in both delay (similar to the present study design) and trace conditioning, a more difficult procedure in which a temporal gap and stimulus-free period separate the CS and US (Jacobson et al., 2011, Jacobson et al., 2008). This effect does not seem to be restricted to the developing brain as abstinent adults with a history of heavy alcohol consumption also showed behavioral conditioning deficits (Fortier et al., 2009, McGlinchey-Berroth et al., 2002). In addition to standard delay conditioning deficits, adults with AUD were also impaired at more complex conditioning protocols such as trace conditioning (McGlinchey et al., 2005), differential conditioning (i.e., learning to discriminate between two different CSs), and reversal learning (Fortier et al., 2008).

It is interesting to note that more lifetime drinks corresponded to poorer conditioning (Figures 2C and 2D: % CR and peak latencies), suggesting that a more severe drinking history predicted worse learning. This relationship between degree of drinking and conditioning is similar to the finding that abstinent AUD patients with Korsakoff’s syndrome, a disease commonly brought on by severe alcohol misuse, showed worse conditioning relative to abstinent non-Korsakoff’s AUD patients (McGlinchey-Berroth et al., 1995). Additionally, dose-dependent impairments in EBC have also been demonstrated in rats exposed to alcohol neonatally, illustrating the idea that the heaviest exposed groups were most impaired (Brown et al., 2008, Lindquist et al., 2013). These data suggest that conditioning deficits related to AUD reside on a spectrum and that heavier alcohol consumption may exacerbate alcohol-related learning deficits.

No significant group differences in non-associative factors such as hearing thresholds, UR amplitudes, and performance on post-experimental questionnaires (questions about the movie and CS-US relationship) were observed. This lack of group differences suggests that the behavioral differences seen here are deficits specific to conditioning and not performance-related.

Neuroimaging

Given the critical involvement of the cerebellum during this form of learning, it was not surprising to see cerebellar activations as most prominent. This finding is in line with human EBC investigations that highlight the role of the cerebellum using multiple approaches, including neuroimaging (Molchan et al., 1994, Kirsch et al., 2003, Ernst et al., 2017), patient/lesion (Steiner et al., 2019, Gerwig et al., 2003), and neuromodulation techniques (Beyer et al., 2017, Zuchowski et al., 2014, van der Vliet et al., 2018). The present neuroimaging data showed that, despite poorer conditioning, AUD participants recruited more regions and demonstrated higher cerebellar activation than healthy participants. AUD participants showed greater cerebellar responding in eight clusters of activation during conditioning while healthy participants did not show greater activity in any cerebellar region (see Table 2). This pattern and direction of fMRI activity is consistent with other studies that report clear cerebellar hyperactivation associated with deficits in eyeblink conditioning (Cheng et al., 2014, Cheng et al., 2017). In an fMRI study that examined the role of development on EBC, healthy children were able to demonstrate successful conditioning but not to the same degree as young adults. But despite this less than optimal behavioral performance, children produced greater learning-related cerebellar activity than adults (Cheng et al., 2014). Even more relevant to the present data, another fMRI study showed that children with prenatal alcohol exposure did not condition as well as healthy children but again, demonstrated significantly greater cerebellar activations (Cheng et al., 2017).

This cerebellar hyperactivation was not unique to EBC as greater task-based fMRI cerebellar activation related to excessive alcohol consumption has also been previously reported. Desmond and colleagues (2003) investigated abstinent adults with AUD and healthy adults performing a Sternberg verbal working memory task during fMRI. Behaviorally (accuracy and reaction time), the two groups performed at the same level but adults with AUD exhibited greater activation in cerebellar lobule VI (Desmond et al., 2003). Similarly in another study, AUD participants showed comparable performance to controls in a language task but demonstrated greater fMRI activation in their left middle frontal gyrus, right superior frontal gyrus, and cerebellar vermis (Chanraud-Guillermo et al., 2009). Finally, even low levels of alcohol use elicited significantly greater cerebellar activity despite no performance differences in a Counting Stroop task (Hatchard et al., 2015). These findings of greater fMRI cerebellar responding in AUD patients may reflect a compensatory mechanism, whereby greater functional activation is needed for normal performance. However, because the AUD participants in the present study did not show intact conditioning, it is difficult to explain this hyperactivation entirely with a compensatory hypothesis. More work is needed to completely rule out the compensatory explanation as it is possible that the hyperactivation reflects effort by the adults with AUD and that given enough trials, they would achieve stronger conditioning levels. Another possible interpretation is that the increased cerebellar activation may represent excitotoxicity, a process in which cells are damaged or killed due to overstimulation of receptors. It is possible that the cerebellar hyperactivation by the AUD participants represented a neuronal excitotoxic process, resulting in conditioning deficits. This finding of conditioning impairments associated with increased cerebellar activation was consistent with previous fMRI studies of eyeblink conditioning that compared healthy children to adults and healthy children to children exposed to alcohol prenatally (Cheng et al., 2014, Cheng et al., 2017). However, it is important to note that neuronal excitotoxicity as a consequence of AUD has been questioned (Collins and Neafsey, 2016, Crews et al., 2015). Finally, it is also possible that hyperactivation does not represent neurotoxic or compensatory mechanisms but rather supports CS and US processing in the cerebellum in AUD participants.

Since extensive evidence from the animal literature suggests that the cerebellar deep nuclei are crucial for eyeblink conditioning (Christian and Thompson, 2003), apriori ROI analyses (Diedrichsen et al., 2011) designed to characterize activity in these regions were performed. Significant group differences related to learning were not observed in these structures, which is consistent with other fMRI/EBC studies (Ramnani et al., 2000, Knuttinen et al., 2002, Cheng et al., 2008). This lack of group differences in activation should be interpreted with caution because imaging the cerebellar deep nuclei can be challenging (Habas, 2010), due to its small size and iron accumulation in this region associated with normal development. Also contributing to the difficulty of imaging these structures is that variability of signal intensities in the deep nuclei increased as one ages (Maschke et al., 2004). However, recent ultra-high-field 7T fMRI studies have been successful at resolving a signal in the cerebellar cortex and deep nuclei (Ernst et al., 2017, Thurling et al., 2015).

Brain-behavior relationship

Correlation and regression analyses showed that participant’s behavioral measures (% CRs and lifetime drinking) were associated with activations in the cerebellum. However in the case of % CRs and healthy participants (Figure 3E), this appeared to be driven by one outlier as once this individual was eliminated from the analysis, the correlation was no longer significant.

Previous human EBC neuroimaging (Logan and Grafton, 1995, Cheng et al., 2014, Cheng et al., 2017) and laboratory animal (Yeo et al., 1984) studies implicate the cerebellar cortex, particularly lobule VI, in the acquisition and expression of behavioral CRs. This region has been identified as one important site of plasticity for EBC as Purkinje cell damage in rats (Nolan and Freeman, 2006) and aspirations of lobule VI in rabbits (Yeo et al., 1984) produced behavioral conditioning deficits. Similarly in humans, patients with damage to this area showed a significant reduction in conditioning levels (Gerwig et al., 2003).

Furthermore, two independent analyses showed that increased lobule VI activity also corresponded to a greater number of lifetime drinks consumed (Figure 4), suggesting that the severity of drinking affects how this structure processes conditioning. Greater cerebellar fMRI activation was also found to be positively associated with severity of AUD during a Stroop task (Wilcox et al., 2019). Interestingly, those participants who showed a reduction in drinking behavior also showed a reduction in cerebellar activation, suggesting that drinking and activity in the cerebellum were linked. The present findings extend our understanding of the role that lobule VI plays during EBC in AUD and healthy populations. They indicate that higher lifetime alcohol consumption predicts higher cerebellar activation and that consideration of drinking severity on a spectrum may be more explanatory over a dichotomous categorization of AUD vs. healthy participants.

Limitations and Future Directions

One consideration of the present study is the number of participants and the issue of replicability. A larger sample size would certainly increase statistical power, however, AUD participants meeting our strict eligibility criteria were rare. These high standards were designed to rule out comorbid factors and other possible non-alcohol related explanations of our effects. Despite these efforts, group differences not directly related to alcohol (e.g. depression, smoking, and race) were still evident in our sample, making interpretation of the divergent findings in learning and brain activations more difficult. Finally, it is important to bear in mind that classical conditioning has benefitted from decades of replicability and research, making this model of learning highly predictable even in smaller sample sizes.

Future studies using brain stimulation techniques (e.g., cathodal transcranial direct current stimulation; tDCS) could further test the prediction that cerebellar hyperactivation is deleterious to successful eyeblink conditioning. Anodal and cathodal tDCS of the cerebellum have also been shown to be effective in modulating conditioned eyeblink responses in healthy individuals (van der Vliet et al., 2018, Beyer et al., 2017, Zuchowski et al., 2014). To date, tDCS of the cerebellum as a way to improve conditioning in an AUD population has not been performed.

Conclusions

This study is an important step towards identifying and characterizing cerebellar functioning in AUD. Our results indicate that AUD is associated with deficits in EBC, that cerebellar activation in AUD may be detrimental to successful EBC, and that the nature of this activation is closely associated with an individual’s drinking history. In sum, classical conditioning is a fundamental task on which more complex cognition is built so understanding the etiology of learning impairments related to alcohol exposure is important for diagnoses, intervention, and treatment of cognitive disorders related to AUD.