Psilocybin-induced changes in neural reactivity to alcohol and emotional cues in patients with alcohol use disorder: an fMRI pilot study

This pilot study investigated psilocybin-induced changes in neural reactivity to alcohol and emotional cues in patients with alcohol use disorder (AUD). Participants were recruited from a phase II, randomized, double-blind, placebo-controlled clinical trial investigating psilocybin-assisted therapy (PAT) for the treatment of AUD (NCT02061293). Eleven adult patients completed task-based blood oxygen dependent functional magnetic resonance imaging (fMRI) approximately 3 days before and 2 days after receiving 25 mg of psilocybin (n = 5) or 50 mg of diphenhydramine (n = 6). Visual alcohol and emotionally valanced (positive, negative, or neutral) stimuli were presented in block design. Across both alcohol and emotional cues, psilocybin increased activity in the medial and lateral prefrontal cortex (PFC) and left caudate, and decreased activity in the insular, motor, temporal, parietal, and occipital cortices, and cerebellum. Unique to negative cues, psilocybin increased supramarginal gyrus activity; unique to positive cues, psilocybin increased right hippocampus activity and decreased left hippocampus activity. Greater PFC and caudate engagement and concomitant insula, motor, and cerebellar disengagement suggests enhanced goal-directed action, improved emotional regulation, and diminished craving. The robust changes in brain activity observed in this pilot study warrant larger neuroimaging studies to elucidate neural mechanisms of PAT. Trial registration: NCT02061293.

www.nature.com/scientificreports/ The parent study methods and primary outcomes are described in detail elsewhere 16,43 .Briefly, inclusion criteria were: (1) age 25 to 65 years old, (2) confirmed AUD diagnosis using the Structured Clinical Interview for DSM-IV 44 , and (3) had at least 4 heavy drinking days in the past 30 days.Participants were excluded from the study if they had a major psychiatric or substance use disorder other than AUD; any hallucinogen use in the past year or more than 25 lifetime uses; or contraindicated medical conditions or exclusionary medications.Participants in the main trial were randomly assigned to receive two administrations of psilocybin or active placebo (diphenhydramine) with 12 weekly therapy sessions provided by two therapists.Before the medication session, all participants received 4 psychotherapy sessions featuring motivational interviewing and cognitive behavioral therapy, and educational preparation for managing and making use of the psilocybin session (see Bogenschutz and Forcehimes 43 for further information).
A subsample of fourteen participants from the parent clinical trial consented to participate in the ancillary fMRI study and were randomized to psilocybin (n = 6) or placebo (n = 8).The timeline followback (TLFB) was used to quantify baseline drinking, yielding percent heavy drinking days (PHDD), drinks per day (DPD), and percent drinking days (PDD) 45 .The Penn Alcohol Craving Scale (PACS) was used to quantify baseline craving 46 .Baseline demographic and drinking-and fMRI-related group differences were evaluated with independent sample t-tests and Chi-squared tests.Participants completed task-based functional MRI (fMRI) with a target mean range of 2-3 days before and 1-2 days after receiving their first dose of study blinded medication, consisting of either psilocybin (25 mg/70 kg) or diphenhydramine (50 mg), administered orally during a monitored 8-h drug administration session.

fMRI acquisition and analysis
Structural and functional MRI (fMRI) images were acquired with CBI's Siemens Skyra scanner equipped with a 20-channel radio-frequency coil.A T1 weighted image was acquired using an MPRAGE pulse sequence in the sagittal plane with an isotropic 0.8 mm resolution, TE/TR/TI = 3.1/2400/1000 ms, and 224 slices (7 min.).fMRI images were collected in the AP direction with a multi-band gradient echo EPI sequence.Parameters were axial slices with a FOV = 248 mm, TE/TR = 29/1000 ms, 3 mm isotropic resolution, 56 slices, 955 volumes, multiband factor = 8, BW = 2770/Hz/Px, and echo spacing = 0.52 ms.

fMRI alcohol and affective cue task
To integrate cue-elicited responses to alcohol and emotionally valenced stimuli, we employed a visual cue fMRI paradigm.Following the design of Vollstadt-Klein and colleagues 47 , participants viewed pictures of alcohol-containing beverages, and negative affective, positive affective, and neutral pictures from the International Affective Pictures Series (IAPS) 48 .Alcohol, neutral, negative, and positive pictures were matched for color and complexity and other potentially important confounds (i.e., presence of people) and presented in pseudorandomized order.Forty pictures were presented for each stimulus category (alcohol, neutral, negative, and positive) across 8 blocks, equaling 160 stimuli across the two 12-min runs (24-min total task time).Blocks were 20 s in duration; five pictures were presented for 4 s each.Between blocks, participants were asked to rate their craving on a scale from 1 to 5 (1 = "no craving at all" and 5 = "severe craving") within a 10 s timeframe; 15 s of fixation ensued prior to the next block.Pre-to-post treatment changes in fMRI button box craving data was assessed using two-tailed paired t-tests.

fMRI preprocessing
Preprocessing and analysis of fMRI data was completed in SPM12 (Wellcome Trust Centre for Neuroimaging, https:// www.fil.ion.ucl.ac.uk/ spm) and CONN (https:// www.nitrc.org/ proje cts/ conn).Preprocessing steps included slice time correction, realignment to the mean image, co-registration to the skull-stripped T1 image, normalization to MNI space, and spatial smoothing (8 mm FWHM Gaussian kernel).Scrubbing removed functional volumes exceeding 2 mm displacement using the Artifact Detection Tools toolbox and a 128 s high-pass filter removed low-frequency drift.Whole-brain statistical analyses were performed using a general linear model with task regressors convolved with the canonical hemodynamic response function.For activation analyses, 6 realignment parameters were entered as covariates to account for motion.For functional connectivity analyses, the CONN-fMRI toolbox was used to regress out parameters for white matter (5P), CSF (5P), and realignment (12P) with first-order derivatives.Next, data were band-pass filtered (0.008 0.09) and linearly detrended.After this denoising procedure, all quality control measures were above the 95% normal histogram match, suggesting the absence of associations between quality control and functional connectivity metrics 49 : max global signal change (96.5% match, x = 0.02, SD = 0.30), mean global signal (99.1% match, x = 0.00, SD = 0.31), max motion (95.4% match, x = 0.03, SD = 0.30), and mean motion (97.7% match, x = − 0.02, SD = 0.31).

fMRI modeling and analyses
Treatment-by-time interactions were modeled at the first and second level.At the 1 st level, time (post > pre) and condition (alcohol > neural; negative > neutral; and positive > neutral) were modeled.At the 2nd level, the randomized treatment assignment was modeled (psilocybin > placebo and psilocybin < placebo).For withinpsilocybin group effects of time, the 1st level contrasts for psilocybin participants were entered at the 2nd level with activation specified as [1] and deactivation specified as [− 1] using the post > pre contrast.
Vol:.( 1234567890 www.nature.com/scientificreports/Brain regions showing significant treatment-by-time interactions in the alcohol contrast were entered into a seed-based region of interest (ROI) using a generalized psychophysiological interactions (gPPI) approach to identify functional connectivity alterations specific to alcohol processing after controlling for the positive, negative, and neutral conditions (p-FWE < 0.05).For the functional connectivity analyses, gPPI modeled the entire experimental session by calculating regressor and PPI terms for each condition and generating beta weights for interaction terms (Y = Alc + Neg + Neut + Pos + ROI + Alc*ROI + Neut*ROI + Neg*ROI + Pos*ROI + error) 60 .This method enables the isolation of condition-specific modulation of connectivity.
The rationale for using a whole-brain, uncorrected p < 0.005 threshold-rather than an ROI FWE/FDR corrected approach-was on the basis of the following: (1) the present study's sample size was not adequate for multiple comparison correction; (2) the absence of fMRI studies of psilocybin in alcohol use disorder (and all other substance use disorders) and cue-reactivity tasks precluded justifiable hypotheses; (3) widespread abnormalities in neural co-activation in AUD result in a large number of potential ROIs; (4) there is evidence that psychedelics alter global brain dynamics 50 ; and (5) psychedelics cannot be assumed to have effects similar to traditional pharmacotherapies.

Demographics
Two participant did not complete both study visits and fMRI malfunctioning resulted in incomplete data collection for one participant at the pre-intervention visit, resulting in the exclusion of three participants from the analysis.Thus, the final sample comprised eleven participants (psilocybin n = 5; placebo n = 6).No group differences were detected in biological sex, age, weight, baseline craving, baseline percent heavy drinking days, baseline drinks per day, or pre/post fMRI framewise displacement (Table 1).However, the psilocybin group scored significantly higher in percent drinking days at baseline relative to the placebo group (Table 1).fMRI scans were collected on average 2.55 days before psilocybin treatment (SD = 1.75; range 1-6) and 1.45 days after treatment (SD = 0.68, range 1-3), falling within the mean target range of 2-3 days before and 1-2 days after.No group differences were detected in the number of days between the first fMRI and treatment (t[9] = -0.77,p = 0.462), between treatment and the second fMRI (t[9] = 0.229, p = 0.82), or between the first fMRI and the second fMRI (t[9] = -0.571,p = 0.582).

Negative emotional cue neural reactivity
In the negative affective cue task, treatment-by-time interactions detected increased activation in 5 clusters (Fig. 2A, Table 3).Of these, 3 clusters showed within-psilocybin treatment effects, mirroring areas from the alcohol cue reactivity task, including the left caudate, left mPFC, and left dlPFC, and uniquely, the right supramarginal gyrus (SMG; Table 3).Deactivation treatment-by-time interactions were detected in 13 clusters (Fig. 2B, Table 3).Of which, 6 clusters showed within-psilocybin treatment effects, including the right insula, left MTG, bilateral lingual gyri, and cerebellum (left 4/5 and right 9; Table 3).

Functional connectivity
Based on activation findings showing (i) significant treatment-by-time interactions and (ii) significant withinpsilocybin effects of time for the alcohol contrast, the following regions were run in a seed-based gPPI functional connectivity analysis: (1)

Discussion
The present study sought to characterize psilocybin-induced alterations in neural activity to alcohol and emotional cues which may account for therapeutic effects in patients with alcohol use disorder (AUD).Psilocybin treatment was associated with engagement of various prefrontal cortical areas (lateral and medial PFC) and the caudate, and disengagement of the insula, motor and cerebellar areas, and temporal, parietal, and occipital cortices.These post-acute effects (i.e.occurring in the days following psilocybin administration) largely implicate brain areas previously reported to be acutely affected by psilocybin 51 .Importantly, group-by-time interactions were mostly driven by changes in the psilocybin group, suggesting that psilocybin-assisted therapy alters neural activity across the cortex and within multiple limbic structures.The high prevalence of overlapping regions across conditions suggests treatment effects were largely non-specific to stimulus type (alcohol, negative, and positive cues), and possibly reflects alterations to the saliency of visual stimuli, affective processing, or a general mood stabilizing effect.Psilocybin-treated patients displayed increased caudate, mPFC, vlPFC, and dlPFC engagement across multiple cue types, suggesting functional reorganization of structures involved in emotional regulation, response inhibition, goal-directed action 47 , and executive functioning 5 .However, the directionality of some of the effects are-at initial pass-inconsistent with normalization of AUD-related dysfunction as meta-analyses indicate hyperactive frontostriatal circuits in AUD.Specifically, studies have reported hyperactivity of the mPFC and dorsal striatum in response to alcohol cues, relative to healthy controls, and treatment-induced downregulation of this pathway within AUD samples 51,52 .While this warrants caution when interpreting the present study findings, a few lines of evidence offer potential explanations.
First, hyperactivity to alcohol cues in these regions are frequently reported in the context of hypoactive responses to other stimulus categories (i.e., negative/stress, neutral, positive stimuli) 53,54 .Such alcohol-specific hyperactivity supports the notions of pathological incentive salience toward alcohol cues, and concomitant devaluation of non-drug stimuli in AUD 7 .Therefore, it is plausible that increased activity in these brain regions across alcohol and affective stimuli reflects a broadening of incentive salience and changes in general affective processing.Such a widening of the attentional scope may be critical to belief updating in predictive coding and Bayesian models of addiction 55,56 , as has been posited to be a mechanism of action of psychedelics 57 .
Secondly, directionality has been mixed as studies have also reported hypoactivity within frontostriatal regions in AUD.For example, hypoactive mPFC and striatum responses to alcohol and negative/stress images, in contrast to hyperactive responses in these regions to neutral/relaxing images, have been reported in AUD compared to healthy controls 58,59 .Since both hyper-and hypoactivity in the mPFC predicted drinking behavior and relapse in these studies, valence-dependent responses in the mPFC may be clinically relevant.Notably, we observed decreases in orbitofrontal cortex (OFC), a subregion of the vmPFC, and increases in the dmPFC, areas responsible for emotional and cognitive aspects of self-referential processing, respectively 60 .In line with our findings, successful inhibition of cue-induced cocaine craving has been negatively associated with OFC activity and positively associated with vlPFC activity in the right hemisphere 61 .Thus, we speculate psilocybin might dampen the emotional and enhance the cognitive self-relevancy of emotionally charged stimuli.It is also important to consider that mPFC and caudate were activated in concert with ventral and dorsal divisions of the lateral PFC, matching what is observed in healthy controls who show greater lateral PFC recruitment compared to AUD patients 62 .Additionally, greater medial and lateral PFC activity during the regulation of alcohol craving and negative emotions has been observed in patients with AUD 63 .Thus, while psilocybin-induced increases in medial PFC is inconsistent with normalization of alcohol cue sensitization in AUD 52 , patterns match neural signatures of cognitive regulation, suggesting that psilocybin may enhance top-down executive control, rather than blunt the saliency of alcohol-related cues 64 .Future studies should consider the complex and potentially opposing roles of ventral, dorsal, and orbital divisions of the medial PFC, and contemporaneous lateral PFC co-activation, when evaluating psilocybin modulating effects on cue-reactivity.www.nature.com/scientificreports/Further support for psilocybin's putative effects on cognitive regulation can be drawn from the neurobiological underpinnings of attentional and inhibitory control in AUD.For example, IFG response is negatively associated with attentional biases to drug cues 65 ; heightened dlPFC and vmPFC is observed during alcohol interference in a Go-NoGo task 66 ; diminished dlPFC recruitment is observed when making reward-related decisions and processing negative prediction errors 67 ; and dlPFC stimulation reduces alcohol craving 68 .In the context of psilocybin treatment, one study found increased dlPFC, vlPFC, and mPFC response in an emotional conflict Stroop task 41 , and another found mPFC functional connectivity changes during a focused attention meditation practice 37 .Considering this research in the context of AUD suggests that psilocybin might diminish preference for alcohol cues and engage hubs of inhibitory control.However, follow-up studies using executive functioning tasks are needed to directly test this proposition.
While comparisons with other studies of psilocybin's action are difficult due to heterogeneities in clinical samples, assessment time points (acute versus post-acute), and task designs, there has been some consistency in reported brain regions, including: the mPFC, a hub of the DMN (see Gattuso et al. 69 for a review of psychedelic effects on the DMN), the ACC and insula, nodes of the SN, and lateral PFC, a hub of the executive control network.Focusing strictly on post-acute effects, psilocybin has been shown to induce connectivity changes in the cingulum, striatum, and mPFC, with decreased mPFC-PCC connectivity predictive of positive mood 4 months later among health controls 37 .In treatment-resistant depression, psilocybin altered mPFC, ACC, and PCC connectivity one day post-treatment, with decreases in mPFC connectivity predicting depressive symptoms 5 weeks later 39 .In a negative affective task similar to the one employed in the present study, dlPFC and mPFC decoupling with the amygdala one day post-psilocybin has been shown to predict reductions in rumination 5 weeks posttreatment 40 .Moreover, Barrett and colleagues found psilocybin increased positive affect and increased PFC response to emotionally conflicting stimuli 41 .
While we did not observe functional connectivity changes in the mPFC as has been reported in other samples, we found increases in ACC-caudate and vlPFC-precentral gyrus connectivity, suggesting psilocybin may modulate frontostriatal and motor circuits, respectively.Whether these changes reflect top-down or bottom-up modulation deserves attention in future studies using effective connectivity approaches.Our findings of increased PFC activity and functional connectivity with striatal and motor areas add to this growing body of literature, and together, independent research groups are beginning to converge on putative therapeutic substrates of psychedelics 17,37,41 .
Augmented striatal activity to alcohol cues has been most widely reported in the ventral striatum (nucleus accumbens 53 ) and putamen, responsible for reward/motivation and motor control/habitual behavior, respectively, whereas the caudate appears to contribute more to goal-directed action and cognitive control 70 .Given this functional distinction (and concomitant PFC activation), heightened caudate response and caudate-ACC connectivity post-treatment might reflect top-down cognitive control and diminished emotional perturbation.Relatedly, diminished functional connectivity between the striatum and ACC has been associated with AUD severity in a response inhibition task 71 , and abstainers display stronger striatal-ACC connectivity than non-abstainers 72 .Intriguingly, we did not observe decreases in the nucleus accumbens or amygdala as expected.Decreases in the left putamen were evident in the interaction but nonsignificant for within-psilocybin comparisons.Acute reductions in left putamen have been reported following psilocybin administration 51 .In light of these considerations, we speculate that the effects observed in the present study reflect a state of improved self-regulatory control in relation to long-term goal pursuit (sobriety or reduced drinking) and emotional equipoise irrespective of changing environmental stimuli 63 .
Psilocybin-treated patients also displayed broad reductions in insular, motor, temporal, occipital, and cerebellar activity relative to placebo controls.These findings are in line with an activation likelihood estimation metaanalysis in AUD that found hyperactivity and treatment-induced reductions in these brain regions, including after cue-exposure therapy 52,72,73 .Overall, the patterns of deactivation observed after psilocybin point toward normalization.For example, greater activation in insular, temporal, parietal, and occipital cortices have generally been found during alcohol cues exposure in AUD versus health controls 61,68,71 (with some inconsistencies 52 ).A role for the cerebellum in addiction and craving has also emerged 74 , with activity positively correlating with AUD severity 6 .Our findings of attenuated cerebellar response support a growing consensus of its contributions to higher-order cognitive functions such as negative emotionality, salience detection, executive control, memory, and self-reflection 75 .Acutely, psilocybin has also been shown to decrease activity in the insula, hippocampus, motor cortex, and temporal areas, although directionality might be dependent on relative versus absolute measurement 51 .Psychedelics modulate areas rich in 5-HT1A receptor expression, such as the insula, raising the possibility that psilocybin may exert inhibitory effects on the insula via agonism at 5-HT1A receptors 76 .In relation to AUD, decreases in insular activity are in line with previous work showing insular hyperactivity and treatment-induced reductions in AUD 52 .The insula has long been associated with interoceptive components of craving and negative affect 77 .Psilocybin-specific decreases in insular activity were robust for alcohol and negative affective contrasts, but not for positive affective cues, suggesting that attenuation of interoceptive processing is specific to craving and negative affect states.
Unique to positive affective cues, psilocybin reduced left and increased right hippocampus engagement.Interestingly, hemispheric asymmetries have been established for emotional processing, with left hippocampal lateralization occurring when viewing negative versus neutral pictures 78 .Others have observed increases in relative cerebral blood flow in the right hippocampus acutely after psilocybin administration 51 , raising the question whether these changes persist or undergo temporal reconfiguration that ultimately results in durable clinical effects.We speculate that these lateralized, affect-specific responses might reflect the facilitation of natural, nondrug rewards regaining reinforcing properties and a resetting of the hedonic set point as has been qualitatively reported in the parent study 79  www.nature.com/scientificreports/Recent developments in establishing a neural signature of craving have included temporal, parietal, occipital, and cerebellar regions, expanding the neurobiology of addictions beyond the confines of the mesocorticolimbic circuitry which has dominated the field's focus 80 .Koban and colleagues posit that co-activation of visual and posterior attentional areas may be critical to ascribe personal meaning to rudimentary percepts 80 , as has been established for complex emotional states-such as fear and sadness-which are highly embedded in the visual system 81 .From this perspective, it is possible that personal associations with alcohol and emotional contexts are attenuated though PFC engagement and contemporaneous posterior disengagement, giving rise to a decentered, nonjudgmental, and nonreactive perspective as has been reported in the early stages of mindfulness meditation interventions 82,83 .However, in the absence of brain-behavior analyses and relevant fMRI paradigms, extreme caution is warranted when inferring the cognitive and psychological processes underlying these brain findings.Well-powered studies are needed to examine the relationships between these neural correlates and the proposed cognitive constructs.

Limitations
This study has major limitations worth noting.The single most limiting factor is the small sample size which restricts generalizations and limits statistical power.Rather than approaching this small dataset with ROI hypotheses, we chose to report whole-brain level changes to serve as a foundation for other work to replicate or disconfirm.We utilized a balanced statistical thresholding approach and sought to isolate psilocybin-specific effects by focusing on treatment-by-time interactions that demonstrated within-psilocybin effects of time.While this approach provides an unbiased method to explore these data in the absence of previous studies, it results in an elevated type 1 error rate due to multiple comparisons.Another limitation related to the small sample size is the lack of control for variables which may contribute to BOLD response in cue-reactivity designs (due to the need to conserve degrees of freedom).For example, biological sex, smoking status, and age may influence AUD responses to cues.The within-subject design of the study, inclusion of motion parameters as covariates, and absence of baseline between-group differences, partially mitigates this concern, but these factors should be accounted for in better powered studies.Other limitations include a nondiverse, homogeneous population which was primarily Caucasian, of young adult age, and of middle-to-high socioeconomic status.Studies with diverse samples are critical to determining for whom psilocybin treatment is (most) beneficial and if shared neural mechanisms underpin therapeutic improvements across populations.All of the findings of this pilot study require replication in larger and more diverse samples before they can be accepted as generalizable.

Conclusion
In summary, this randomized, controlled pilot study provides the first data on neurobiological changes occasioned by psilocybin-assisted therapy in patients with AUD.Key findings are: (1) increased engagement of frontal circuits; (2) widespread disengagement of temporal, parietal, occipital, and cerebellar brain regions; and (3) consistently overlapping neurobiological circuits across stimulus categories, suggestive of alterations to affective processing.While caution is urged due to sample size and lack of stringent multiple comparison correction, the findings are encouraging, suggest large effect sizes, and reveal potential therapeutic neural changes attributable to psilocybin in AUD.
Promisingly, if fMRI metrics prove to be strong proxies of the purported rapid, robust and enduring salutary effects of psilocybin, future investigation in this area holds potential to (i) elucidate the etiology of AUD (ii) identify novel neural targets seeking to optimize and sustain treatment gains (i.e. using neurostimulation technologies or non-psychedelic 5-HT 2A agonists), (iii) reveal transdiagnostic mechanisms of psychiatric conditions, and (iii) facilitate precision-based medicine for AUD and other disorders of addiction.