LSD modulates effective connectivity and neural adaptation mechanisms in an auditory oddball paradigm

This single-blind, placebo-controlled, MEG (magnetoencephalography) study (n=20) used a clever auditory test to show decreased top-down information processing (more entropy, more bottom-up) under the influence of LSD (75μg).

Abstract

“Under the predictive coding framework, perceptual learning and inference are dependent on the interaction between top-down predictions and bottom-up sensory signals both between and within regions in a network. However, how such feedback and feedforward connections are modulated in the state induced by lysergic acid diethylamide (LSD) is poorly understood. In this study, an auditory oddball paradigm was presented to healthy participants (16 males, 4 female) under LSD and placebo, and brain activity was recorded using magnetoencephalography (MEG). Scalp level Event Related Fields (ERF) revealed reduced neural adaptation to familiar stimuli, and a blunted neural ‘surprise’ response to novel stimuli in the LSD condition. Dynamic causal modelling revealed that both the presentation of novel stimuli and LSD modulate backward extrinsic connectivity within a task-activated fronto-temporal network, as well as intrinsic connectivity in the primary auditory cortex. These findings show consistencies with those of previous studies of schizophrenia and ketamine but also studies of reduced consciousness – suggesting that rather than being a marker of conscious level per se, backward connectivity may index modulations of perceptual learning common to a variety of altered states of consciousness, perhaps united by a shared altered sensitivity to environmental stimuli. Since recent evidence suggests that the psychedelic state may correspond to a heightened ‘level’ of consciousness with respect to the normal waking state, our data warrant a re-examination of the top-down hypotheses of conscious level and suggest that several altered states may feature this specific biophysical effector.”

Authors: Christopher Timmermann, Meg J. Spriggs, Mendel Kaelen, Robert Leech, David J. Nutt, Rosalyn J. Moran, Robin L. Carhart-Harris & Suresh D. Muthukumaraswamy

Notes

This study was supported in part by the Beckley Foundation.

The same participants were also included in Carhart-Harris et al. (2016) which also used other brain imaging techniques to draw broader conclusions about the LSD/psychedelic experience.

Summary

LSD modulates effective connectivity and neural adaptation mechanisms in an auditory oddball paradigm

Under the predictive coding framework, perceptual learning and inference depend on the interaction between top-down predictions and bottom-up sensory signals. LSD may modulate the feedback and feedforward connections in the brain, which may be a biophysical effector common to several altered states of consciousness.

1. Introduction

Serotonergic psychedelic drugs, such as LSD, psilocybin, mescaline and DMT, have been shown to alter brain mechanisms subserving normal waking consciousness, and may also present a safe, novel and potentially effective treatment option for certain psychiatric disorders.

The auditory oddball paradigm, which is widely used in conjunction with electroencephalography and magnetoencephalography, provides insight into perceptual learning and has become a ubiquitous probe of conscious perception in healthy participants, patient populations, and during pharmacologically altered levels of consciousness.

Auditory oddball paradigms have also been used to determine potential neural mechanisms associated with psychosis. NMDA antagonists and serotonergic psychedelics have been found to reduce the amplitude of the MMN, but DMT has been found to have an effect on both standards and deviants.

Recent interpretations of the mechanisms underlying the MMN fit within the Bayesian brain and predictive coding models of brain functioning. The MMN is modulated by feedforward and feedback connectivity between frontal and temporal areas as well as intrinsic connectivity in primary auditory areas.

Evidence increasingly suggests that serotonergic psychedelics modulate functional connectivity, but few studies have addressed changes in effective connectivity under these drugs.

We used a pseudo-randomized, placebo controlled design to assess how intrinsic and extrinsic (forward and backwards) connectivity are altered under LSD compared with placebo.

2.1. Experimental design and statistical analysis

Twenty healthy participants were recruited for the study, which consisted of a psychiatric interview, ECG, routine blood tests and urine test for drug use. They received placebo on one day and LSD on another day.

On study days, LSD or placebo was administered intravenously through a cannula inserted in participant’s antecubital fossa. Resting state MRI and MEG scans were performed before the auditory oddball task was completed.

Whole-head MEG recordings were performed using a CTF-275 channel radial gradiometer system while participants lay in supine position. ECG and eye-tracking data were also recorded, and statistical analysis was performed on the recorded data.

The oddball paradigm used was a replication of the ‘Optimum-1’ mismatch negativity paradigm. Five different types of deviants were presented, which differed from standards in frequency, duration, intensity, perceived direction, or by having a gap in the middle.

The auditory oddball paradigm was used to study memory trace evoked by different types of deviants. The stimulus-onset-asynchrony (SOA) used was 300 ms and the total number of trials per condition was 3015, divided in 3 blocks lasting 5 min each.

2.2. Data preprocessing

Data were preprocessed using the FieldTrip toolbox and epoches were performed using a time-window starting 60 ms before stimulus onset and ending 300 ms after. Robust averaging was performed and resulted in ERFs corresponding to each of the 4 conditions.

2.4. Source space analysis

We modelled cortical sources based on MRI scans and performed group source inversion using the Multiple Sparse Priors method. We then selected the cortical sources for subsequent DCM analysis based on the main effect of surprise.

2.5. Dynamic causal modelling

Dynamic Causal Modelling (DCM) was used to explore how the mismatch response is generated, and how this may be modulated by LSD. DCM is a model of effective connectivity, and is a valuable tool for understanding ERFs.

The generative model in DCM for ERFs comprises coupled neural mass models representing activity at individual brain sources, and a forward lead field representing how membrane depolarizations are translated to sensor space.

The source was modelled as a single Equivalent Current Dipole (ECD), and the lead-field mapping was parameterized in terms of the location and orientation of each dipole.

The current study employed a network of 5 interconnected regions, including the left and right primarily auditory cortex (A1), the left and right superior temporal gyrus (STG) and the right inferior frontal gyrus (IFG).

To test the effect of standard vs. deviant tones, three different models of extrinsic modulation were compared, with the modulation of intrinsic connectivity in A1 being present or absent.

These generative models were inverted under a Variational Bayesian framework to assess parameter likelihood given the data and the model for each subject individually. The model evidence was then calculated using an Expectation-Maximization scheme.

A separate analysis was conducted that included models that replaced the main effects with the interaction term. However, this analysis had lower log-model evidence than the analysis that included the two main effects.

2.6. DCM statistics

Statistical analysis of the DCMs involved two steps: first, the models that best fit the data were identified using Random Effects (RFX) Bayesian Model Selection (BMS). The protected exceedance probability (pxp) was used as an index of model fit, and second, the Bayesian Omnibus Risk (P0) was calculated.

The second set of models was fitted using a posteriori parameter estimates. The results were then tested for interaction effects.

3.1. Sensor space analysis of event-related fields

Based on previous literature using Dynamic Causal Modelling (DCM) for the mismatch negativity, 1178.07 “standard” trials and 234.57 “deviant” trials were selected for analysis at the scalp-level. An ANOVA was performed using Drug and Surprise as the 2 factors of interest.

Surprise to the interaction effect was addressed using post-hoc paired t-tests, which revealed 2 separate clusters of mismatch responses: one at right temporal sensors at 127 ms and an earlier component at right occipital sensors at 67 ms. The comparison of deviant tones for the LSD versus placebo condition resulted in two large significant clusters, an earlier component at 60e93 ms peaking at 80 ms and a later one at 110e127 ms peaking at 117 ms.

Analysis of interaction effects between Drug and Surprise using gap and duration deviants revealed marginal interaction effects in the right temporal and parietal areas, and no significant interaction effects for the direction and intensity deviants. A mismatch response was present only for the placebo condition during the whole time window of interest (166e223 ms), peaking at 170 ms following stimulus onset. A significant effect was also observed for the contrast between standards in the placebo condition versus deviants in the LSD condition. Results reveal significant effects for the frequency deviant in the LSD condition, including an absent early component at occipital sensors and significantly blunted early and late components at parietal, occipital and temporal sensors.

3.3. DCM of standard vs. deviant tones under placebo

Six DCMs differing in the modulation of intrinsic and extrinsic connectivity by the deviant tone were compared for the placebo condition. The model with modulation of backward and intrinsic connectivity was found to be the best fit.

3.4. DCM for the effect of LSD

A second analysis looked at how both the deviant tone and LSD modulated effective connectivity. The model with the greatest model evidence included backward and intrinsic modulation for Surprise (as per the first analysis) as well as backward and intrinsic modulation for Drug.

The results showed that the intrinsic connectivity in A1 increased bilaterally for deviant tones, and decreased bilaterally for LSD. Additionally, there was a marginal decrease in backward connectivity from right STG to left A1 for the LSD effect over the group.

4. Discussion

LSD reduced the negativity to unexpected stimuli, while increasing the negativity to the standard stimulus. This was best modelled by modulations of backward extrinsic connections and intrinsic connectivity in the primary auditory cortex.

Under the predictive coding framework, A1 intrinsic connectivity increases with the divergence between predictions for standard and deviant tones, but is reduced under LSD, potentially resulting in a reduction in the estimated precision of prediction errors and a recursive break down of perceptual learning.

The data relating to the mismatch negativity (MMN) and intrinsic connectivity in primary auditory areas are consistent with empirical and theoretical work on psychosis, and indicate decreased adaptation mechanisms and less divergence between standard and deviant stimuli under LSD.

The findings suggest that psychedelics may share a neural mechanism with schizophrenia, accounting for broad modifications of sensitivity towards external stimuli, while still leaving room for further enquiry on other mechanisms explaining the phenomenological differences between both states.

Changes in connectivity alter the glutamatergic mechanisms underlying associative learning, which can lead to abnormal perceptual phenomena and delusions. This is supported by previous studies demonstrating that acute ketamine administration disrupts bottom-up connectivity and that LSD alters top-down connectivity.

Reduced frontal to temporal connectivity has been identified in patients with reduced levels of consciousness, which leads us to question the specificity of the hypothesis that reduced top-down connectivity is a marker of conscious level.

The MMN was most consistently identified as the best fit for the data in the placebo condition, however, the paradigm employed in this study was different to previous studies, in that it used a short stimulus-onset asynchrony (300 ms) and a relatively high proportion of deviant presentations. We used source analysis to identify sources at the superior temporal gyrus and inferior frontal gyrus and used random effects modelling to identify the winning model.

The current study’s IFG source is posterior compared to previous MMN studies, and may have been affected by the short inter-stimulus interval or the limited sensitivity of magnetoencephalography to frontal radial dipoles.

5. Conclusions

LSD administration reduces the magnitude of the mismatch response, alters neural adaptation mechanisms at the scalp, and alters backward extrinsic connectivity in the fronto-temporal generative network. This results suggest that LSD administration attenuates top-down suppression of prediction error in response to auditory stimuli. Future studies may be conceived to specifically inform on and potentially resolve this matter, by testing specific aspects of psychosis and then testing concrete hypotheses about their underlying mechanism.