If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Elevated amygdala responses during de-novo Pavlovian conditioning in alcohol-use disorder are associated with Pavlovian-to-Instrumental transfer and relapse latency
Corresponding author: Claudia Ebrahimi, Charité – Universitätsmedizin Berlin, corporate member of Freie Universität Berlin and Humboldt-Universität zu Berlin, Department of Psychiatry and Neurosciences | CCM, Charitéplatz 1, 10117 Berlin, Germany
Affiliations
Charité – Universitätsmedizin Berlin, corporate member of Freie Universität Berlin and Humboldt-Universität zu Berlin, Department of Psychiatry and Neurosciences | CCM, NeuroCure Clinical Research Center, Berlin, GermanyDepartment of Psychiatry and Psychotherapy, Technische Universität Dresden, Dresden, Germany
Charité – Universitätsmedizin Berlin, corporate member of Freie Universität Berlin and Humboldt-Universität zu Berlin, Department of Psychiatry and Neurosciences | CCM, NeuroCure Clinical Research Center, Berlin, Germany
Charité – Universitätsmedizin Berlin, corporate member of Freie Universität Berlin and Humboldt-Universität zu Berlin, Department of Psychiatry and Neurosciences | CCM, NeuroCure Clinical Research Center, Berlin, GermanyTechnische Hochschule Aschaffenburg, University of Applied Sciences, Aschaffenburg, Germany
Charité – Universitätsmedizin Berlin, corporate member of Freie Universität Berlin and Humboldt-Universität zu Berlin, Department of Psychiatry and Neurosciences | CCM, NeuroCure Clinical Research Center, Berlin, Germany
Department of Psychiatry and Psychotherapy, Technische Universität Dresden, Dresden, GermanyNeuroimaging Center, Technische Universität Dresden, Dresden, Germany
Applied Computational Psychiatry Lab, Mental Health Neuroscience Department, Division of Psychiatry, University College LondonApplied Computational Psychiatry Lab, Max Planck UCL Centre for Computational Psychiatry and Ageing Research, Queen Square Institute of Neurology, University College LondonCamden and Islington NHS Foundation Trust
Department of Psychiatry and Psychotherapy, Technische Universität Dresden, Dresden, GermanyDepartment of Addiction Medicine and Psychotherapy, kbo Isar-Amper Klinikum Region München, Haar, Germany
Charité – Universitätsmedizin Berlin, corporate member of Freie Universität Berlin and Humboldt-Universität zu Berlin, Department of Psychiatry and Neurosciences | CCM, NeuroCure Clinical Research Center, Berlin, Germany
Charité – Universitätsmedizin Berlin, corporate member of Freie Universität Berlin and Humboldt-Universität zu Berlin, Department of Psychiatry and Neurosciences | CCM, NeuroCure Clinical Research Center, Berlin, GermanyCharité – Universitätsmedizin Berlin, corporate member of Freie Universität Berlin and Humboldt-Universität zu Berlin, NeuroCure Cluster of Excellence, Berlin, Germany
Contemporary learning theories of drug addiction ascribe a key role to Pavlovian learning mechanisms in the development, maintenance and relapse of addiction. In fact, cue-reactivity research has demonstrated the power of alcohol-associated cues to activate the brain’s reward system, which has been linked to craving and subsequent relapse. However, whether de-novo Pavlovian conditioning is altered in alcohol use disorder (AUD) has been rarely investigated.
Methods
To characterize de-novo Pavlovian conditioning in AUD, n=62 detoxified patients with AUD and n=63 matched healthy controls completed a Pavlovian learning task as part of a Pavlovian-to-instrumental transfer (PIT) paradigm during an fMRI session. Patients were followed up for 12-months to assess drinking behavior and relapse status.
Results
While patients and controls did not differ in their ability to explicitly acquire the contingencies between conditioned and unconditioned stimuli, patients with AUD displayed significantly stronger amygdala responses towards Pavlovian cues; an effect primarily driven by stronger BOLD differentiation during learning from reward compared to punishment. Moreover, in patients compared to controls, differential amygdala responses during conditioning were positively related to the ability of Pavlovian stimuli to influence ongoing instrumental choice behavior, measured in a subsequent PIT test. Finally, patients who relapsed within the 12-month follow-up period showed an inverse association between amygdala activity during conditioning and relapse latency.
Conclusions
We provide evidence of altered neural correlates of de-novo Pavlovian conditioning in patients with AUD, especially for appetitive stimuli. Thus, heightened processing of Pavlovian cues might constitute a behaviorally relevant mechanism in alcohol addiction.
) highlights the motivational power of environmental stimuli to promote craving, drive recurrent drug use and ultimately increase relapse risk. However, the underlying Pavlovian learning process, whereby initially neutral stimuli (conditioned stimuli, CS+) acquire motivational properties through repeated pairings with the hedonic effects of a reinforcer like alcohol (unconditioned stimulus, US) has been rarely investigated in AUD (
Human neuroimaging research elucidated an extended network subserving Pavlovian threat and appetitive conditioning, including the amygdala, hippocampus, ventral striatum (VS) entailing the nucleus accumbens (NAcc), dorsal anterior cingulum (dACC) and orbitofrontal cortex (
). In AUD, surprisingly little is known about the underlying Pavlovian learning process, and we are unaware of any imaging studies investigating de-novo Pavlovian conditioning with drug- or non-drug rewards in this psychiatric condition. This might be partly due to methodological challenges human appetitive conditioning research is facing (
). In contrast, two fMRI studies used a threat conditioning protocol in AUD patients, providing first evidence for attenuated BOLD responses towards threat-predicting cues: Yang and colleagues (
) found attenuated neural differentiation in pregenual ACC, medial prefrontal cortex (PFC) and posterior cingulate cortex (PCC) between a CS predicting a high- vs. low-heat US in alcohol dependent men, while BOLD reactivity in posterior insula towards the high- vs. low-intensity US itself was increased in patients compared to controls. Recently, Munich et al. (
) showed attenuated amygdala involvement during threat conditioning using mild electric stimulation as US in patients with AUD compared to healthy participants. In spite of general blunting, remaining amygdala activation scaled positively with dependence severity, as well as measures of depression, anxiety, and perceived stress (
) did not differ between AUD patients and healthy controls in these imaging studies, two laboratory studies showed blunted differential physiological responses during Pavlovian threat conditioning in high- compared to low-risk AUD populations (
). In line with these findings, reduced amygdala activation has further been observed in response to aversion-inducing, alcohol-related cues in patients with AUD compared to control participants (
On the other hand, generic or idiosyncratically appetitive conditioned cues like the sight or smell of an alcoholic beverage have been shown to bias attention and approach tendencies, induce physiological arousal, and often increase subjective craving in AUD (e.g., 17,18), for review, see 19). BOLD responses elicited by such alcohol-associated cues were predictive of subsequent relapse, most consistently in the VS (
) showed that a novel cue paired with alcohol elicited increased orienting responses that correlated with subjective liking of alcohol in social drinkers. In another study, only participants scoring low on self-reported alcohol sensitivity – a proposed risk phenotype for AUD (
) – demonstrated conditioned neurophysiological responses during second-order conditioning with an alcoholic olfactory cue, suggesting this group might be more susceptible to attribute incentive salience to novel, alcohol-associated cues (
Transfer of incentive salience from a first-order alcohol cue to a novel second-order alcohol cue among individuals at risk for alcohol use disorder: electrophysiological evidence.
). In addition, we previously showed an increased ability of de-novo conditioned Pavlovian cues to bias instrumental choice behavior in recently detoxified alcohol dependent patients compared to healthy participants, measured using a Pavlovian-to-Instrumental transfer (PIT) task (
Altogether, impaired threat conditioning in combination with increased cue reactivity could point towards a unique pattern of associative learning alterations in AUD. On the one hand, reward-associated Pavlovian conditioning might be exaggerated, resulting in elevated reactivity towards drug-associated cues. On the other, a reduction in threat conditioning could make subjects more vulnerable to engage in drug-taking behaviors despite severe negative consequences (
). To test this hypothesis, we here investigate for the first time appetitive and aversive de-novo conditioning as part of a PIT paradigm during fMRI in a large sample of 62 recently detoxified AUD patients and 63 matched control participants. We further explore the behavioral and clinical relevance of these associative learning processes by linking differential BOLD responses during Pavlovian learning to the instrumental choice bias in the subsequent PIT phase and prospective relapse risk during a 12-month follow-up period.
2. Methods and Materials
2.1 Participants
As a part of the LeAD study (Learning and Relapse Risk in Alcohol Dependence; clinical trial preregistration identifier: NCT01679145), 62 recently detoxified alcohol-dependent patients (referred to hereinafter as AUD) and 63 healthy controls matched for age, gender, and smoking status were included at two German study sites in Berlin and Dresden (Table 1; see Supplementary Material for exclusion criteria including Table S1 and Figure S1 for participant flowchart). Only participants showing a significant degree of CS-US contingency knowledge post-learning were included in the final analyses (Figure 1B). After detoxification, patients were followed up for twelve months to assess relapse status (see Supplementary Material for details on follow-up assessments). Follow-up information were available for 44 AUD patients (27 relapsers vs. 17 abstainers).
Table 1Sample characteristics
N
AUD Patients
N
Healthy Controls
p-value
Sociodemographics
Gender
62
♀:13, ♂:49
63
♀:10, ♂:53
.61a
Mean
SD
Mean
SD
Age in years
62
43.98
11.59
63
42.86
11.19
.59
Smokers
75.8 %
73.0 %
.88a
Education (in years)
62
14.37
3.12
61
15.9
3.89
.02
Socio-economic status (SES)
54
-0.41
1.88
42
0.49
1.81
.02
Neurocognitive functioning
Verbal intelligence (MWT-B)
60
104.52
9.43
62
104.66
9.53
.93
TMT-A (seconds)
60
29.42
8.7
62
28.31
9.39
.50
TMT-B (seconds)
60
69.98
26.46
62
60.16
22.54
.03
AUD severity
years with diagnosis (DSM-IV)
57
11.35
10.24
-
number of DSM-IV symptoms
58
5.71
1.24
63
0.51
1.03
<.001
Severity of AUD (ADS)
62
15.31
7.06
63
1.94
2.93
<.001
Lifetime alcohol consumption in kg (pure alcohol)b
62
1717.26
1180.3
63
303.67
988.74
<.001
Craving (OCDS-G total score)
61
12.84
8.33
62
2.87
2.89
<.01
days of abstinence before scanning
62
20.31
11.65
63
88.89
342.16
.12
Personality
Impulsivity (BIS total score)
59
30.47
6.4
62
29.15
5.55
.23
Socio-economic status (SES) was computed as the sum of self-rated z-transformed scores of social status, household income, and inverse personal dept scores (
Wittchen H-U, Pfister H (1997): DIA-X Interview. Instruktionsmanual Zur Durchführung von DIA-X-Interviews (Instruction Manual for the DIA-X-Interview). Frankfurt: Swets & Zeitlinger.
Figure 1A Exemplary appetitive conditioning trial. In each trial, a CS (fractal image combined with one out of five pure tones) was presented either on the right or left side of the screen for 3 s. After a fixed 3-second-trace interval, the associated monetary US (or neutral outcome (0 Cent)) appeared on the opposite site for 3 seconds (100% reinforcement schedule). Trials were separated by a jittered ITI (exponentially distributed; range: 2-6s; mean=3s). The paradigm comprised 5 different conditions (two appetitive, two aversive, and one neutral condition). CS assignment to conditions was counterbalanced across participants. B CS-US contingency knowledge. Mean probability of choosing the higher-valued CS during post-conditioning forced-choice task did not significantly differ between patients with AUD and healthy controls (W=2515.5, p=.77; AUD: n=75, mean(SD) = 85.6(18.1); HC: n=69, mean(SD) = 87.8(16.7)). Only participants performing significantly over chance (teal color-coded participants; i.e. over 50% correct choices, as confirmed by a binomial test) were considered contingency aware (83% of AUD patients, 91% of healthy controls) and included in the final sample (participant characteristics, see Table 1; see Supplementary Table S2 for sample characteristics of aware vs. unaware participants).
The paradigm consists of four parts: instrumental conditioning, Pavlovian conditioning, Pavlovian-to-Instrumental Transfer (PIT), and a forced-choice task to assess CS-US contingency awareness (see Supplementary Material and Figure S2 for task details).
Instrumental conditioning. Participants learned to collect ’good‘ shells and to leave ’bad‘ shells via probabilistic monetary feedback. Shells could be collected via repeated button presses and participants completed up to 120 trials depending on task performance.
Pavlovian conditioning. The task consisted of two appetitive conditions (CS paired with monetary win +2€ or +1€, respectively), two aversive conditions (CS followed by monetary loss -1€ or -2€, respectively), and a neutral control condition without monetary feedback (0€), using five different multimodal cues as CSs (see Figure 1A). Each CS was presented 16 times, resulting in a total of 80 trials. Participants were instructed to attend to the relations between CS and US and to memorize the pairs. They were further informed that they would receive the displayed, cumulated money after the session.
Pavlovian-to-Instrumental Transfer. During the PIT phase the influence of the learned Pavlovian conditioned stimuli on instrumental choice behavior was measured. Participants performed the instrumental task while one of the Pavlovian CSs tiled the background without receiving feedback.
Forced-choice task. Finally, CS-US contingency knowledge of Pavlovian learning was assessed, where participants had to choose the higher-valued CS out of two CSs presented on the left and right site of the screen. Each CS combination was presented 3 times in pseudo-randomized order. Only participants performing significantly over chance (83% of AUD and 91% of control participants) were considered contingency aware and included in the final analyses, as contingency awareness seems necessary for Pavlovian trace conditioning to occur (
R Core Team (2017): R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. Retrieved from http://www.r-project.org
). The alpha level was set at p<.05 for all analyses.
CS-US contingency awareness. Contingency awareness was measured as percentage of higher-valued CS choices during the forced-choice task and group differences were examined via Mann-Whitney U test (see Supplementary Material for more detailed analyses).
Pleasantness and arousal ratings. Subjective ratings of CS pleasantness and arousal, obtained at the end of the PIT paradigm, were analyzed in separate linear mixed-effects models (LMMs) including CS value, group, and study site (see Supplementary Material for details). Aversive and appetitive conditioning were investigated separately, given first evidence of deficits in Pavlovian threat conditioning in high-risk samples (
fMRI. After standardized preprocessing (see Supplementary Material), an event-related analysis was applied using the GLM approach within SPM 12 (Welcome Department of Imaging Neuroscience; www.fil.ion.ucl.ac.uk/spm/) on two levels. For each participant, onset regressors for each CS and US type were modeled as stick functions and convolved with the canonical HRF. Additional nuisance regressors included an eye-tracker recalibration period after half of the trials (mean duration 71.6 s), modeled as box-car function and the 6 movement parameters to account for movement-related variance. Baseline contrasts for each CS were computed and entered into a random-effects flexible factorial model on the second level, together with the group factor (AUD/HC). We investigated main effects across participants as well as group differences for the following three contrasts: Pavlovian learning was probed by contrasting CSs across valence conditions with the control condition (0€), taking into account the grading within appetitive and aversive conditions (i.e. Pavlovian CSs: -2€ -1€ 0€ +1€ +2€, contrast ‚Pavlovian learning’: [+2 +1 -6 +1 +2]). We then separately investigated appetitive and aversive Pavlovian learning (contrast ‚aversive Pavlovian learning‘: [+2 +1 -3 0 0]; contrast ‚appetitive Pavlovian learning‘: [0 0 -3 +1 +2]). Group differences were investigated by testing the group x contrast interaction, followed by post-hoc t-tests in case of a significant effect.
We focused our analyses on three predefined regions-of-interest (ROI): Amygdala and hippocampus, due to their central role in appetitive and aversive Pavlovian (trace) conditioning (
). Bilateral ROIs for amygdala and hippocampus were derived using the WFU PickAtlas (http://www.fmri.wfubmc.edu/download.html) and the VS as a functionally defined mask using the BrainMap database (
). ROI-analyses were performed at p<0.05 FWE-correction, complemented by exploratory whole-brain analyses at p<0.05 FWE correction at the cluster level, using a cluster-forming threshold of p<0.001 uncorrected and cluster extend of ten contiguous voxels. To account for multiple comparisons across ROIs, p-values were additionally adjusted for the number of ROIs using Bonferroni correction.
Brain-behavior associations: Individual PIT effects (see Supplementary Materials) were entered as a covariate within SPM in a separate second-level GLM with the ’Pavlovian learning‘ contrast and the group factor (AUD/HC), allowing for an interaction between group and covariate. We focused our ROI analysis on the amygdala and VS shown to modulate neural PIT effects (
To investigate whether neural signatures during Pavlovian learning were predictive of subsequent relapse, we re-run the flexible-factorial model and informed the group factor by patients’ prospective relapse status (relapsers vs. abstainers vs. HC). We further explored whether neural responses during Pavlovian learning correlated with relapse latency in prospective relapsers using simple regression analysis with the ‘Pavlovian learning‘ contrast and the number of abstinence days till relapse as a covariate.
Study site was included as additional covariate in all analyses.
Results
Explicit learning of CS-US associations: contingency awareness
Contingency awareness was assessed post-learning in a forced-choice task, using data from all participants providing high-quality fMRI data (75 AUD patients vs. 69 HC, Figure 1B, see Supplementary Figure S1 for participant flowchart). Overall performance was at 86.6% correct choices (SD=17.4; range: 16.7-100), with no differences between groups (W=2515.5, p=.77), indicating equal levels of contingency awareness (Figure 1B; see also Supplementary Figure S4). All subsequent analyses are based on participants performing significantly over chance (i.e., ’Pavlovian learner‘, as confirmed by binomial test).
Subjective measures of Pavlovian learning: pleasantness and arousal ratings
Subjective CS pleasantness and arousal ratings, acquired post conditioning, were significantly influenced by the conditioning protocol, evident in a linear effect of CS value on pleasantness ratings (b=0.15, SE=0.06, t=2.78, p=.006) and a linear and quadratic effect on subjective arousal (blinear=0.09, SE=0.04, t=2.22, p=.027; bquadratic=0.08, SE=0.04, t=2.24, p=.026; Supplementary Figure S5). This indicated that participants’ pleasantness and arousal ratings reflected Pavlovian value after conditioning. Arousal ratings were higher in AUD patients compared to controls across cues (b=-0.58, SE=0.26, t=-2.22, p=.028), but we did not observe a group by value interaction, indicating groups did not differ in conditioned responses (pleasantness: p=.358; arousal: p≥.158). Separate investigation of appetitive and aversive conditioning revealed the observed behavioral effects were driven by appetitive CSs (pleasantness: b=0.26, SE=0.12, t=2.21, p=.028; arousal: b=0.22, SE=.09, t=2.46, p=.015) rather than aversive CSs (pleasantness and arousal p>=.386), without significant effects of group or group by CS value interaction in neither analysis (p≥.118).
Neural representation of Pavlovian learning: BOLD signals towards appetitive and aversive Pavlovian cues
Across participants, Pavlovian learning induced marginally increased BOLD responses in right amygdala (pFWE ROI =.099; Table 2). Separate investigation of appetitive and aversive Pavlovian conditioning revealed significantly increased BOLD responses towards reward-predicting cues in the left VS (pFWE ROI=.05; Table 2), while aversive Pavlovian conditioning showed no significant differential BOLD responses. No additional activated clusters survived in the whole-brain analyses.
Table 2Region-of-interest analyses of Pavlovian conditioning
Analysis
Contrast
Region
Side
k
Peak voxel MNI
Zmax
pFWE
x
y
z
all participants
Pavlovian conditioning
Amygdala
R
28
-2
-14
3.08
0.033
[CS-2€ > CS-1€ > CS0€ < CS+1€ < CS+2€]
VS
L
-4
12
-8
3.3
0.037
appetitive conditioning
Amygdala
R
28
-2
-14
3.09
0.032
[CS+2€ > CS+1€ > CS0€]
VS
L
-4
12
-8
3.55
0.017*
R
14
6
-12
3.33
0.034
aversive conditioning
-
[CS-2€ > CS-1€ > CS0€]
Group differences
AUD > HC
Pavlovian conditioning
Amygdala
R
28
-4
-22
3.74
0.004*
[CS-2€ > CS-1€ > CS0€ < CS+1€ < CS+2€]
L
-24
-8
-22
3.35
0.014*
AUD > HC
appetitive conditioning
Amygdala
R
26
-6
-22
4.06
0.001*
[CS+2€ > CS+1€ > CS0€]
L
-24
-8
-22
3.56
0.007*
Hippocampus
R
26
-10
-22
3.76
0.004*
L
-24
-10
-24
3.5
0.011*
pFWE: family-wise error-corrected at p<0.05 for bilateral anatomical region; *denotes significance after Bonferroni correction for number of ROI comparisons; L: left hemisphere, R: right hemisphere; VS: ventral striatum, AUD: alcohol use disorder; HC: healthy controls
Group comparison revealed significant different engagement of right amygdala during Pavlovian conditioning (amygdala right: [x:28, y:-4, z:-22], F1,492=14.65, pFWE ROI=.029). Post-hoc analysis showed that AUD patients exhibited significantly stronger differential BOLD responses in bilateral amygdala towards Pavlovian cues relative to healthy controls (see Table 2; Figure 2; complementary analyses are provided in the Supplementary Material). Investigating differential BOLD responses for appetitive and aversive Pavlovian conditioning separately revealed that the observed group difference was specific for reward-predicting cues, assessed with the appetitive Pavlovian conditioning contrast (amygdala right: [x:26, y:-6, z:-22], F1,492=16.75, pFWE ROI=.006; amygdala left: [x:-24, y:-8, z:-22], F1,492=12.84, pFWE ROI=.045). Here, AUD patients additionally showed stronger recruitment of an anterior cluster within the hippocampus ([x:26, y:-10, z:-22], F1,492=14.38, pFWE ROI=.027; Table 2). In contrast, no group differences emerged during aversive Pavlovian conditioning. Results remained significant when contingency unaware participants were also included (see Supplementary Material).
Figure 2Stronger differential BOLD responses in bilateral amygdala during Pavlovian conditioning in AUD patients compared to control participants (amygdala right: Z=3.74, pFWE_ROI=.012; amygdala left: Z=3.35, pFWE_ROI=.041). Group differences were driven by both increased BOLD responses toward Pavlovian CSs in AUD patients compared to healthy controls (pFWE_ROI<.001), as well as increased BOLD responses towards the neutral cue in healthy participants compared to AUD patients (pFWE_ROI≤.012; see Supplementary Material).Visualization threshold of T-map at T≥3.
Association of Pavlovian conditioning with instrumental PIT behavior and prospective relapse
We further investigated whether neural responses during Pavlovian learning were related to the ability of Pavlovian cues to bias subsequent choice behavior (i.e., PIT effect; see Supplementary Table S3 and Figure S3), and to prospective relapse risk.
Across groups, this analysis revealed that increased conditioning-related BOLD activity in right VS was associated with a stronger instrumental choice bias during the subsequent PIT phase ([x:4, y:-14, z:-8], Z=3.48, pFWE ROI=.05; Supplementary Figure S6). Group comparisons showed that BOLD activity in left amygdala was predominantly predictive of patients’ subsequent choice bias, in contrast to healthy controls (left: [x:-26, y:-2, z:-24], Z=3.35, pFWE ROI=.048; Figure 3).
Figure 3BOLD responses in left amygdala during Pavlovian conditioning were positively associated with subsequent PIT behavior in AUD patients compared to control participants (amygdala left: Z=3.35, pFWE_ROI=.048). Visualization threshold of T-map at T≥3.
Finally, we assessed whether neural signals during de-novo conditioning were associated with prospective relapse at 1-year follow-up. Contrasting prospective relapsers with abstainers as well as healthy controls revealed a main effect of group in the right amygdala during Pavlovian learning ([x:26, y:-6, z:-20], F2,416=8.58, pFWE ROI=.033; Figure 4). Post-hoc analyses confirmed that both patient groups showed increased amygdala activity relative to healthy controls (relapser > HC: [x:26, y:-6, z:-20], Z=3.47, pFWE ROI=.033; abstainer > HC: [x:24, y:-4, z:-20], Z=3.40, pFWE ROI=.042). Although amygdala activation did not differ between prospective relapsers and abstainers (pFWE ROI=.22), within patients who relapsed, increased right amygdala activation during Pavlovian learning was associated with reduced relapse latency ([x:20, y:0, z:-20], Z=2.94, pFWE ROI=.047; Figure 4).
Figure 4A Significant group difference between prospective relapsers, abstainers, and controls during Pavlovian conditioning in the right amygdala (F2,416=8.58, pFWE ROI=.033). Both prospective relapsers (Z=3.47, pFWE ROI=.033) and abstainers (Z=3.40, pFWE ROI=.042) showed increased BOLD responses compared to control participants, while patient groups did not significantly differ. B Within patients who relapsed, differential amygdala responses during Pavlovian learning were inversely related to relapse latency (Z=2.94, pFWE ROI=.047). Visualization threshold of F-/T-map at F≥6/T≥3.
In this study, we investigated de-novo Pavlovian conditioning of both appetitive and aversive associations in recently detoxified AUD patients and healthy participants during functional magnetic resonance imaging. While both patients and healthy participants were equally likely to acquire the different CS–US associations in terms of explicit contingency knowledge, Pavlovian CSs elicited significantly stronger BOLD responses in bilateral amygdala in patients compared to controls. This difference was most pronounced for reward-predicting cues. We further related BOLD responses during Pavlovian conditioning to the behavioral choice bias induced by these cues in a subsequent PIT test, as well as to relapse during a 12-month follow-up period. In contrast to healthy participants, left amygdala activation during Pavlovian conditioning was positively associated with the subsequent behavioral PIT effect in patients with AUD, and among patients who relapsed, right amygdala activation was predictive of relapse latency in an exploratory analysis.
Patients with AUD showed elevated amygdala activation towards Pavlovian cues during de-novo conditioning
We observed significant group differences in the amygdala during de-novo Pavlovian conditioning, with stronger differential BOLD responses in patients compared to healthy participants.
Converging lines of evidence identified the amygdala as a core region subserving appetitive and aversive Pavlovian conditioning: The amygdala is critically involved in encoding the state value of motivational salient stimuli, forming CS-US associations, and expression of conditioned responses (
). Amygdala responsivity has been shown to capture individual differences in human threat conditioning, as BOLD signals in this region correlate with physiological conditioning indices (
Long-term expression of human contextual fear and extinction memories involves amygdala, hippocampus and ventromedial prefrontal cortex: A reinstatement study in two independent samples.
), highlighting the importance of this structure for acquisition, recall and expression of conditioned responses.
Our observation of elevated differential amygdala activation during Pavlovian conditioning in AUD patients compared to healthy participants therefore likely reflects enhanced neural encoding of Pavlovian associations – especially rewarding ones – and could reflect greater susceptibility to assign incentive salience to novel, reward-related cues in AUD (
). To our knowledge, this is the first study investigating appetitive Pavlovian de-novo conditioning in AUD patients. Further evidence for enhanced drug-related Pavlovian learning in at-risk participants comes from Fleming and colleagues (
Transfer of incentive salience from a first-order alcohol cue to a novel second-order alcohol cue among individuals at risk for alcohol use disorder: electrophysiological evidence.
), where Pavlovian de-novo conditioning using an alcoholic olfactory cue only induced subjective craving and conditioned event-related potentials in low but not high alcohol sensitive participants, a phenotype associated with risk to develop AUD (
). During instructed threat conditioning, where CS-US contingencies are known in advance, alcohol-dependent men showed attenuated differential BOLD responses towards a high- vs. low-heat predicting cue in cortical regions associated with negative affect regulation, including the pregenual ACC and medial PFC, together with increased posterior insula activation towards the high- vs. low-intensity US itself (
) found that men with a high family history of AUD compared to men without such family history failed to acquire differential SCRs towards threat compared to neutral cues due to reduced CS+ responsiveness. Attenuated differential SCRs and startle responses towards aversive vs. neutral CSs were also observed in young binge drinkers compared to non-binge drinkers (
In our study, investigating aversive and appetitive Pavlovian conditioning separately showed that the group difference in amygdala was primarily driven by enhanced BOLD responses towards reward- and not loss-predicting cues. However, we would refrain from drawing specific conclusions about aversive conditioning in AUD, as the aversive contrast in our paradigm did not elicit significant differential BOLD activation across participants (see also (
)). Therefore, cues signaling threat like electric shock or loud noise might be better suited to study aversive associative learning in future studies.
Conditioned amygdala responses are related to Pavlovian-to-Instrumental transfer and relapse latency in patients with AUD
The PIT paradigm enables investigation of the influence of Pavlovian cues on instrumental behavior - an effect called Pavlovian-to-Instrumental transfer (
By relating neural activity during Pavlovian conditioning to the subsequent behavioral PIT effect, we showed that ventral striatal BOLD responses were positively correlated with the strength of the PIT effect in both patients and control participants. Amygdala activation during Pavlovian conditioning significantly correlated with instrumental choices during PIT in patients with AUD compared to healthy controls. This observation suggests that amygdala engagement during Pavlovian conditioning contributes to instrumental choices towards these Pavlovian cues, and that this association is pronounced in patients with AUD, underlining the behavioral relevance of our neural finding.
Previous research showed that BOLD responses in the NAcc during the PIT phase were predictive of subsequent relapse during follow-up in AUD patients (
). Therefore, we assessed whether the neural signatures of Pavlovian conditioning represent a potential marker for prospective relapse within a 12-month follow-up period. Both prospective relapsers and abstainers showed elevated amygdala responses during Pavlovian conditioning compared to healthy participants, and patient groups did not differ significantly in overall amygdala reactivity. However, we found a significant inverse correlation between right amygdala activation in response to Pavlovian CSs and relapse latency in prospectively relapsing patients. This study, if replicated, may suggest that increased amygdala reactivity towards Pavlovian cues is not a general risk factor of AUD, but could decrease relapse latency in vulnerable persons. Patients who abstained might have additional protective factors, helping them to stay abstinent despite increased amygdala reactivity during Pavlovian learning. Cue-reactivity research revealed that abstinent compared to non-abstinent AUD patients showed increased functional connectivity between limbic regions and prefrontal areas in a cue-reactivity paradigm, potentially helping them to stay abstinent in the presence of craving-inducing alcohol-cues (
). However, increased cue-induced limbic brain activation may not simply promote relapse, but could contribute to salience attribution also required for inhibitory control (
). Complex top-down and bottom-up mechanisms might constitute an important moderating factor also shown to critically interact with cue-reactivity in AUD (
Several limitations need to be considered: First, prospective studies in participants at risk are needed to elucidate whether the observed alterations in Pavlovian conditioning represent a predisposing factor for AUD, or rather develop throughout the disease. Furthermore, our paradigm might not be optimal to disentangle appetitive and aversive conditioning, as the aversive contrast failed to significantly engage relevant brain structures. Third, we acquired no additional psychophysiological measure of conditioned responding, e.g. skin conductance, limiting comparability between studies. Fourth, the observed group difference in amygdala activation during Pavlovian conditioning was due to both increased BOLD responses towards Pavlovian cues in AUD, as well as towards the neutral cue in healthy participants. Our control condition might be affectively more ambiguous compared to a neutral cue not paired with any outcome, as often used in (fear) conditioning paradigms, highlighting the need for careful consideration of adequate baseline conditions in future studies. Moreover, we did not investigate conditioning of drug-related cues, which might tap into more disease specific mechanisms within largely overlapping neural circuits (
To conclude, we provide evidence for altered Pavlovian learning processes in patients with AUD, reflected in increased amygdala recruitment that was especially pronounced during reward-associative learning. Increased amygdala reactivity was related to subsequent PIT behavior as well as to relapse latency during a 12-month follow-up period. These findings may reflect greater susceptibility to assign incentive salience to novel, reward-related cues in AUD (
), a process that might contribute to bias patients’ behavioral choices in the presence of these Pavlovian stimuli.
Our findings extend evidence in AUD and related high-risk populations on Pavlovian conditioning, and point towards patterns of associative learning alterations, whereby conditioned responses towards reward- or drug-associated Pavlovian cues are increased (
Transfer of incentive salience from a first-order alcohol cue to a novel second-order alcohol cue among individuals at risk for alcohol use disorder: electrophysiological evidence.
). This might promote elevated reactivity towards reward-associated cues including drug cues on the one hand, while subjects engage in conditioned behaviors despite severe negative consequences on the other.
Interestingly, the reported conditioning alterations in AUD are different from that seen in other patient populations including post-traumatic stress disorder and anxiety, whereby both threat and safety cues elicit increased physiological responses and neural activation of the amygdala, suggesting abnormal fear generalization (
). Given the outlined evidence, investigating both reward and threat conditioning processes in mental disorders could represent a fruitful avenue for future research, as it enables to dissociate learning alterations in different value domains (
). Moreover, investigating individual differences in these learning mechanisms might provide valuable insights about the role of Pavlovian conditioning in addiction maintenance (e.g., (
Ultimately, characterizing alterations in neural structures subserving Pavlovian learning processes, that is, mechanisms at the center of influential theories of addiction (
), could foster our understanding of AUD as a disorder driven by maladaptive learning and provide targets for future therapeutic interventions aim to counteract the motivational power of alcohol-related cues (
We thank the LeAD study teams in Dresden and Berlin for data acquisition. This work was supported by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) FOR 1617 – grants HE2597/14-1, HE2597/14-2, SCHL1969/2-2, ZI1119/3‐1, ZI1119/3‐2, ZI1119/4‐1 – and in part by TRR 265, Project ID 402170461, subprojects B01 and B03 (
Transfer of incentive salience from a first-order alcohol cue to a novel second-order alcohol cue among individuals at risk for alcohol use disorder: electrophysiological evidence.
R Core Team (2017): R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. Retrieved from http://www.r-project.org
Long-term expression of human contextual fear and extinction memories involves amygdala, hippocampus and ventromedial prefrontal cortex: A reinstatement study in two independent samples.
Wittchen H-U, Pfister H (1997): DIA-X Interview. Instruktionsmanual Zur Durchführung von DIA-X-Interviews (Instruction Manual for the DIA-X-Interview). Frankfurt: Swets & Zeitlinger.
QJMH has obtained fees and options for consultancies for Aya Technologies and Alto Neuroscience. AH, CE, FS, MG, MNS, MS, KC, and USZ reported no biomedical financial interests or potential conflicts of interest.
00011-3&id=gr1.jpg){kind=link}
00011-3&id=gr2.jpg){kind=link}
00011-3&id=gr3.jpg){kind=link}
00011-3&id=gr4.jpg){kind=link}