Chronic alcohol exposure differentially modulates structural and functional properties of amygdala: A cross‐sectional study

Animal models have shown that chronic alcohol exposure is associated with persistent neuroadaptations in amygdala synaptic function, whereas human studies have consistently reported amygdala grey‐matter volume (GMV) reductions in alcohol dependent patients (ADP). We hypothesised that chronic alcohol use associated with neuroadaptations may entail a reconfiguration of the amygdala's functional interactions and that these mechanisms may be affected by structural atrophy. We compared amygdala resting state functional connectivity (RSFC) using a whole brain seed‐based approach and amygdala GMV in abstinent ADP (n = 20) and healthy controls (HC; n = 39), balanced for age, gender and levels of head motion. The potential moderating influence of age, cumulative alcohol exposure, abstinence length and head motion was further examined in the two groups separately using correlational analyses. We found increased amygdala RSFC with substantia nigra/ventral tegmental area (SN/VTA) in ADP compared with HC. As expected, amygdala GMV was lower in ADP. Multiple regression analyses of the ADP group showed that amygdala‐SN/VTA RSFC increases were primarily associated with cumulative alcohol exposure rather than age, whereas amygdala GMV reductions were primarily associated with age rather than cumulative alcohol exposure. The same association between age and amygdala GMV was not observed amongst HC. Importantly, amygdala GMV and amygdala‐SN/VTA RSFC were uncorrelated in ADP, and neither measure was correlated with abstinence length. These results suggest that chronic alcohol exposure is associated with persistent elevations in amygdala‐SN/VTA RSFC and accelerated age‐related grey‐matter atrophy through potentially distinct mechanisms.

ADP group showed that amygdala-SN/VTA RSFC increases were primarily associated with cumulative alcohol exposure rather than age, whereas amygdala GMV reductions were primarily associated with age rather than cumulative alcohol exposure. The same association between age and amygdala GMV was not observed amongst HC. Importantly, amygdala GMV and amygdala-SN/VTA RSFC were uncorrelated in ADP, and neither measure was correlated with abstinence length. These results suggest that chronic alcohol exposure is associated with persistent elevations in amygdala-SN/VTA RSFC and accelerated age-related grey-matter atrophy through potentially distinct mechanisms. Chronic, excessive alcohol consumption is associated with widespread homeostatic neuroadaptations, whose effects become unmasked during withdrawal as intense autonomic and emotional disturbances, which in turn negatively reinforce alcohol consumption. 1 Relapse is common within the initial months of abstinence, 2 whereas in those who achieve longer-term abstinence, symptoms of anxiety and dysphoria often persist. 3 An influential model proposes that cumulative alcohol exposure elicits an allostatic load, a persistent shift in the baseline function of motivational and stress circuits, thereby predisposing individuals to relapse to their original state of alcohol consumption. 4 Animal models have highlighted the amygdala as a critical site for allostatic changes which drive escalation of alcohol intake and stressinduced reinstatement. 5 The amygdala is an evolutionarily wellconserved complex of subcortical nuclei involved in fear conditioning, acquisition of stimulus-reward associations and bodily responses to stressors. 6 Chronic alcohol exposure has been shown to augment GABA release in the central amygdala 7 via upregulation of 'pro-stress' neuropeptides. 8 The resulting increase in local inhibition in the central amygdala has been hypothesised to drive the net disinhibition of downstream effectors, for example, hypothalamic, dopaminergic or noradrenergic brainstem nuclei, thereby providing a possible feedforward mechanism for maintaining positive and negative reinforcement of alcohol consumption. 5,9,10 Although animal studies have made great strides in elucidating local neuroadaptations within circumscribed brain regions, a more comprehensive neurobiological model of alcohol dependence requires a systems-level mapping of interactions between multiple brain regions in human studies. Resting state functional connectivity (RSFC) has been increasingly applied to characterise network-level differences in substance dependent populations. Studies that have looked at amygdala RSFC in alcohol, 11 cocaine 12 and opioid-dependent patients 13 have reported lower amygdala RSFC relative to healthy participants. However, to our knowledge, no study has specifically explored chronic alterations in amygdala RSFC with reference to cumulative alcohol exposure in abstinent alcohol-dependent patients. Additionally, there is evidence that alterations in RSFC in substance-dependent populations may be related to structural atrophy. 13,14 Smaller amygdala grey-matter volume (GMV) is widely documented in alcohol dependent populations (ADP) and [15][16][17][18] is associated with a greater likelihood of relapse, 16 yet its relationship to amygdala RSFC has not been specifically investigated.
We present here data from the ICCAM Experimental Medicine Platform 19 comparing amygdala RSFC and amygdala GMV in abstinent ADP with age-matched healthy controls (HC). Because animal models implicate multiple effector regions which may be chronically modulated by local neuroadaptations in the amygdala, 7 we applied a whole brain approach to our RSFC analyses. We hypothesised that longer cumulative alcohol exposure in ADP would be associated with greater alterations in amygdala RSFC and reduced amygdala GMV.

| Study description
All data presented here were collected during the first visit of the ICCAM platform study: a three-center (Imperial College London, University of Cambridge and University of Manchester), multisession neuroimaging study to test the effectiveness of compounds in modulating putative brain pathways in addiction processes. 19 During the study visit, 68 HC and 28 ADP underwent a 5-min resting state functional magnetic resonance imaging (fMRI) scan with eyes closed. This visit did not include drug administration. Participants provided their written informed consent in accordance with the Declaration of Helsinki. Ethical approval was obtained from West London and GTAC NRES committee (11/H0707/9), and R&D approval was obtained from relevant Research Governance and PIC (Participant Identification Centre) sites.

| Inclusion and exclusion criteria
Inclusion criteria required ADP to meet DSM-IV criteria for prior alcohol dependence, without current or prior diagnoses of dependence on other drugs, except nicotine. ADP were also required to have been abstinent for at least 1 month prior to scanning, and none were currently in treatment or receiving medication for alcohol dependence. HC were not permitted to any history of drug or alcohol dependence, except for nicotine.
The following exclusion criteria were applied for both ADP and HC: (i) current primary axis I diagnosis; (ii) current or past history of enduring severe mental illness, including psychosis; however, secondary or lifetime history of depression or anxiety was permitted due to high comorbidity; (iii) history of significant neurological diagnosis; (iv) prescription medication that could interfere with study integrity or safety; (v) any MRI contraindication; and (vi) positive breath alcohol or positive urine drug test on scan day (amphetamines, barbiturates, cocaine, opiates and benzodiazepines). On study days, smokers were permitted to smoke cigarettes up to 1 h prior to scanning in order to avoid withdrawal.
In the analyses presented here, seven participants (3 HC and 4 ADP) were excluded due to containing less than 5 min of resting state data after censoring of volumes with excessive head motion (>0.5-mm framewise displacement). Four additional ADP were excluded: one due to corrupted acquisition and three due to failing to pass inclusion criterion for abstinence at the time of baseline scan.
From the remaining pool of 65 HC, 39 were selected to maximise matching to the ADP group on age, gender, smoking status and head motion in scanner (mean framewise displacement). A cumulative alcohol exposure measure was calculated for all ADP taking into account the quantity, frequency and pattern of their drinking during their lifetime. For each year to be counted as 'exposure', male participants had to exceed 400 g of alcohol per week, or have more than three drinking episodes of at least 70 g per week for at least 6 months of that year, and female participants had to exceed 280 g per week or more than three drinking episodes of at least 56 g per week for at least 6 months of that year. For each ADP, a cumulative alcohol exposure score was calculated that was equivalent to the number of years exceeding these criteria.

| Materials
All psychiatric and substance dependence histories were reviewed by two addiction psychiatrists (RSAF and FP) to ensure uniformity of diagnostic thresholds across sites and any discrepancies arbitrated by a third addiction psychiatrist (ARLH). Abstinence length was measured from last date of alcohol consumption. Cigarette package years (pack years) were defined as number of packs of cigarettes (i.e., 20 cigarettes) smoked per day multiplied by the number of years smoked. Symptoms of anxiety were characterised using Spielberger Trait Anxiety Scale (STAI 22 ) and Spielberger State Anxiety Scale (SSAI 22 ). IQ and handedness were assessed using Weschler's Adult Intelligence Scale 23 and The Edinburgh Inventory, 24 respectively.
Resting heart rate (HR), systolic and diastolic blood pressure were measured in a sitting position on the same day prior to the scan. Oblique acquisition enabled greater coverage of inferior regions at the expense of signal dropout affecting the most superior 9 mm aspect of the brain. There was minimal variation in the time of day of scans across all three sites (earliest: noon, latest: 3:30 pm), thus minimising potential effects of diurnal variation in resting state activity. 25 For more detail on the acquisition parameters, see McGonigle et al. 26

| Preprocessing
Preprocessing of the MRI data was performed using a combination of AFNI, 27 FreeSurfer, 28 ANTs, 29 FSL 30 and in-house Python code.
Structural T1 images were intensity normalised, brain extracted and nonlinearly registered to an MNI152 2-mm brain template. GMV maps were computed by deriving partial volume estimates of greymatter from T1 images, registering these to MNI152 space (using the previously computed transformation matrices), applying Jacobian modulation and spatial smoothing at 6-mm FWHM.
EPI images initially underwent despiking, slicing timing correction, rigid-body motion correction, brain extraction and boundary-based registration to their corresponding native T1 image. EPI images were spatially smoothed at 6-mm FWHM. After registration of EPI images,  Figure S1). Amygdala timeseries were extracted from the fully preprocessed resting state EPI of each participant and modelled within the framework of a GLM. The resulting statistical maps provided a voxel-wise estimate of amygdala RSFC for each subject. This same amygdala region of interest was used both for the resting state fMRI and amygdala GMV analyses ( Figure S1).
As an additional control, we repeated the GMV analysis with a structurally defined segmentation of the amygdala that more closely estimated total amygdala volume than our spherical region of interest.
Here, amygdala GMV (total amygdala GMV) was computed in native space by adding up voxel-wise grey-matter partial volume estimate values within individual-level bilateral masks of the amygdala derived using FreeSurfer. Again, whole brain volume and site effects were regressed out before further analyses.

| Statistical modelling
Group comparison of amygdala RSFC in ADP versus HC was implemented in a mixed-effects model, controlling for site (FSL's FLAME1). Results were thresholded using clusters determined by Z > 2.3 and a corrected cluster significance threshold of p < 0.05. Group comparison of amygdala GMV in ADP versus HC was carried out using a two-sample t test, after controlling for site and intracranial volume (ICV) via regression. Further analyses of amygdala RSFC or GMV in regions of interest were undertaken using subject-level parameter estimates, again, after controlling for effects of site and ICV (for GMV only). Linear relationships between clinical variables and brain measures were examined using Pearson correlations and multiple regression.

| RESULTS
This study examined amygdala RSFC and GMV in 20 ADP and 39 HC.
Demographic and clinical characteristics of the two groups are presented in Table 1. No significant group differences were observed for age, body weight, handedness, IQ score, resting HR, systolic and

| No association between amygdala-SN/VTA RSFC and amygdala GMV
The relationships between amygdala-SN/VTA RSFC, amygdala GMV, age and cumulative alcohol exposure are summarised in Figure 3.  Our spherical amygdala region of interest was motivated by our RSFC analyses and kept the same for the amygdala GMV analyses to enable direct comparison between the modalities. To control for potential biases arising from our choice of amygdala definition, we repeated our amygdala GMV analyses using structurally defined total amygdala volumes (see Section 2). Broadly consistent with our original analysis, we found a significant negative correlation between age and total amygdala GMV (r = −0.56, p = 0.01) in ADP, whereas correlations between total amygdala GMV and cumulative alcohol exposure or abstinence length were not significant. The group difference between ADP and HC for total amygdala GMV also remained significant (p = 0.008). However, the functional significance of amygdala-SN/VTA connectivity could be context dependent, showing distinct effects at rest, as in the current study, and during exposure to alcohol cues, as in the study of Beck et al. 18 To our knowledge, this is the first study to show a linear relationship between cumulative alcohol exposure and an amygdala resting state circuit in abstinent ADP. Previously, we have observed a reduction of amygdala-insula RSFC in another abstinent ADP cohort, an effect that was not found to be associated with cumulative alcohol exposure, and which was not replicated in the current cohort. 11 Other studies have also reported correlations between duration of substance use and RSFC in other circuits, in recently-abstinent ADP, 39 currently using cocaine 12 and prescription opioid dependent populations. 13 In the current study, abstinence length was not associ- When our analyses accounted for, both age and cumulative alcohol exposure, amygdala GMV reductions in ADP were only significantly associated with age, and not with cumulative alcohol exposure.

| Control analyses
One possible interpretation is that some features of alcohol dependence, other than duration of exposure, might interact with age, thereby producing accelerated age-related amygdala atrophy.
According to this account, the same duration of alcohol dependence might be associated with different degrees of amygdala GMV reduction depending on an individual's age. This interpretation is consistent with a recent longitudinal study which found accelerated atrophy of the frontal cortex in recently abstinent ADP that was unrelated to the number of years of heavy drinking. 40 Other studies have also reported accelerated cortical atrophy in recently abstinent ADP 41 and cocainedependent population 42 as well as age-related atrophy of the hippocampus in heavy nondependent drinkers. 43 Our findings build on this work by presenting evidence suggestive of accelerated age-related atrophy specifically in the amygdala and in longer-term abstinent ADP. Nonetheless, the exact set of conditions that are necessary to observe age-associated atrophy in ADP remains unknown, for example, minimum duration of diagnosis, severity of addiction or minimum amount of alcohol exposure that is necessary.
Another possibility is that smaller amygdala GMV might be a preexisting vulnerability to alcohol dependence rather than a consequence of chronic alcohol exposure. Indeed, reduced amygdala volumes have been observed in offspring or unaffected first-degree relatives of ADP, [44][45][46] and they have also been associated with a greater likelihood of relapse during early abstinence. 16 Although this does not provide an explanation for the negative correlation between age and amygdala GMV, the accounts of preexisting differences and accelerated-atrophy are by no means mutually exclusive.
Amygdala GMV did not significantly vary with abstinence length across ADP. This is consistent with previous studies finding amygdala GMV  Last, we did not measure respiratory activity, a confound whose importance is becoming increasingly recognised in resting-state fMRI studies. 50 In summary, we showed that ADP exhibit differences in amygdala structure and function that persist in long-term abstinence and which seem to differentially interact with age and cumulative alcohol exposure, respectively. These findings will require replication in a larger dataset, whereas longitudinal studies will be necessary to characterise the timecourse of these effects and to delineate their potential roles in mechanisms of abstinence and relapse. ARL-H has received honoraria from Lundbeck and research support from GlaxoSmithKline for a PhD studentship.

DISCLOSURE/CONFLICT OF INTEREST
All other authors declared no conflict of interest.