Short- and longer-term brain and behavior responses to LSD in healthy people

This study used a randomized, double-blind crossover design. Forty-five healthy participants between the ages of 21 and 55 received one dose of 100 µg LSD and one inactive placebo (Apotheke Dr. Hysek AG, Biel, Switzerland). All participants provided written informed consent and were compensated. The study was conducted in accordance with the Declaration of Helsinki, approved by the Bern Cantonal Ethical Committee (BASEC-No. 2021-01322) and registered on ClinicalTrials.gov (NCT05177419↗). See supplement for details on inclusion criteria and drugs.

Participants completed a screening visit and seven experimental sessions including the two dosing days (Fig. 1; see supplement for details). Dosing sessions were conducted in a quiet, comfortable room with one or two investigators present at all times (Supplementary Fig. S1). A washout period of >4 weeks was required between the LSD and placebo appointments. Baseline measures of EEG and TMS were taken at the baseline visit and repeated six and seven hours after drug administration, respectively, as well as the next day and 6–8 days later (mean: 7.01). Motor learning was completed the day after dosing. Questionnaires were completed at baseline and one week after dosing. Experiments were always conducted in the afternoon in identical order and participants reported on hours of sleep and exercise in the past 24 h.

Auditory ST (Supplementary Fig. S2) was used to assess acute and post-acute changes in auditory ERPs in response to LSD. In addition to information on ERPs, ST was developed as a translational measure of long-term potentiation (LTP) in humans [17]. It thus served as an attempt to index LTP-like changes over frontocentral electrodes following a protocol used in previous studies [17, 18].

EEG recordings were conducted using 32 actively shielded wet electrodes (actiCAP slim EEG system) in 10-20 system locations with an electrode cap and a LiveAmp amplifier (Brain Products GmbH, Germany). Signals were recorded at 1000 Hz and impedances were kept below 25kΩ as recommended by the manufacturer for actively shielded electrodes. Participants were instructed to sit still and attend to a series of identical tones (1000 Hz with Hanning window at edges, 50 ms duration, 60 dB) over in-ear headphones (ER-3C, Etymotic Research, IL, USA) with eyes open. In the first block, single tones were played at intervals of 1800–2600 ms for 4.5 min. The tetanization block was played after a one-minute break and consisted of tones played at 13 Hz for two minutes. Finally, after a two-minute break, the first block was repeated three more times at 15-min intervals.

Pre-processing was conducted offline using EEGLAB (Version 2024.2) [19]. Data were downsampled to 250 Hz and bandpass filtered between 1 and 30 Hz. Epochs began 200 ms before stimulus onset and ended 600 ms after onset. A baseline correction was applied using the 200 ms before stimulus onset. Bad channels were interpolated and large artefacts were removed, then independent component analysis was used to remove eye movement artefacts. Smaller artefacts were flagged using a custom script and approved for removal by manual inspection.

Similar to previous auditory ST studies [18], we analyzed the amplitude of auditory N1 (greatest negative peak 50–150 ms after stimulus) and P2 (greatest positive peak 150–250 ms after stimulus) ERP components averaged over frontocentral electrodes (Cz, Fz, FC1, and FC2), both before and after tetanization. This choice of electrodes is consistent with previous studies [17, 18], and it also reflects the region at which both components showed the strongest signal in our data (Supplementary Fig. S3). Participants who showed any increase in ERP amplitude after ST were classified as “responders” in calculations of responder rates, consistent with previous studies [18]. Finally, we attempted to replicate a previous exploratory analysis from a study using ST [20], which demonstrated a trend-level increase in evoked theta power two weeks after psilocybin treatment (see Supplemental Methods).

The TMS-based PAS method was used to investigate both motor cortical excitability and LTP-like changes in the motor domain during and after LSD exposure [21]. PAS was conducted using a MagPro X100 (Tonica Elektronik A/S, Denmark) coupled to a Dantec Keypoint electrical stimulator (Alpine Biomed ApS, Skovlunde, Denmark) and adhesive electrodes for electromyography of the right abductor pollicis brevis (APB). In preparation for TMS, participants were fitted with electrodes for measuring motor-evoked potentials (MEPs) over the abductor pollicis brevis (APB) of the right hand using the belly-tendon configuration [22]. Participants wore a reusable cap on which the approximate location of the motor hotspot was marked using the C3 point on the 10/20 system. The TMS coil was placed over this hotspot at an angle of 45 degrees to the midline, and the correct hotspot was identified by determining the point at which a TMS pulse of 70% maximum strength evoked the highest amplitude in the right APB. Subsequently, pulse strength was calibrated so that a single pulse produced an MEP with a peak-to-peak amplitude of ~1 mV in the relaxed APB. After calibration at the baseline visit, pulse strength and hotspot location were recorded and kept consistent for each participant throughout the study.

We used a prototypical PAS procedure as described in previous studies (Supplementary Fig. S2) [23]. Monophasic TMS pulses were always delivered at a rate of 0.1 Hz. First, baseline MEP amplitude in the right APB was determined using a train of 20 TMS pulses. Next, paired stimulation was applied using 180 TMS pulses paired with peripheral electrical stimulation of the right median nerve at 300% of the perceptual threshold, a strength that is noticeable but painless. Median nerve stimulation was delivered with an interstimulus interval of 25 ms before the TMS pulse so that both arrived at motor cortical synapses simultaneously, which is presumably required for induction of Hebbian LTP [24]. Participants were asked to keep the right hand relaxed and silently count the number of pulses in order to keep their attention on the stimuli. Finally, the baseline train of TMS pulses was repeated three times at intervals of 15 minutes (2, 17 and 32 min after the end of the paired pulses) in order to measure changes in MEP amplitude after paired stimulation.

Prior to analysis, MEP peak and latency detection were performed with custom R scripts and confirmed by manual inspection. MEPs with a peak-to-peak amplitude of <50 µV were rejected [25]. Latency was defined as the earliest timepoint at which an MEP deviated from baseline motor activity by at least 25 µV. Participants who showed any increase in MEP amplitude after PAS were classified as “responders” in calculations of responder rates, consistent with previous studies [26]. When displaying grand average MEP waveforms, positive MEP peaks were aligned to the median peak latency of the given condition and visit. This ensured that grand averages preserved true peak amplitudes without flattening due to latency variability.

To assess possible longer-term effects of LSD on neurotrophin expression [27], BDNF was measured in blood plasma and serum before drug intake, as well as eight hours, one day and one week later. Because previous studies have typically used only one blood fraction [28, 29], we chose to measure BDNF in both plasma and serum in order to compare their sensitivity to LSD. Plasma samples were collected with EDTA tubes and serum samples were collected using rapid serum tubes. Serum samples were allowed to clot for five minutes according to manufacturer instructions. Samples were immediately centrifuged for 10 min at 3000 RPM, after which serum and plasma were extracted and frozen at −80° until analysis. BDNF levels were measured using the Biosensis Mature BDNF Rapid ELISA Kit (Thebarton, Australia).

One day after drug administration, participants completed a sequence typing task developed based on previous studies (Fig. 2A, B) [16]. It assessed both online learning during active practice of motor movements and offline learning improvements after rest [30–32]. Participants were instructed to type a 9-digit number sequence (314211223 or 322112413, counterbalanced across drug order) as quickly and accurately as possible using their left (non-dominant) hand. We ensured comparable difficulty between the two sequences by including the same number of taps for each finger, as well as the same set of finger transitions, in both sequences. The task consisted of 10 practice blocks (online learning) and two test blocks after 10 and 80 min of rest (offline learning). Audiovisual feedback on performance was given after each trial and block, and participants were encouraged to beat their high scores from the previous block. Learning in each block was operationalized as the change in average typing speed on correct trials since the previous practice block.

Acute drug effects were assessed using verbal hourly ratings [1, 33], as well as the 5-Dimensional Altered State of Consciousness Questionnaire (5D-ASC) and the Mystical Experience Questionnaire (MEQ) [34–36]. Side effects were assessed using the Swiss Psychedelic Side Effects Inventory (SPSI) [37]. Cognitive flexibility was assessed with the Cognitive Flexibility Inventory (CFI) and perceived psychological stress was measured using the Perceived Stress Scale (PSS) [38, 39]. See supplement for details.

Statistical analyses were conducted using R (Version 4.5.0) [40]. Effects of LSD were analyzed using linear mixed effects models with fixed effects for drug, visit, and treatment order and random effects for individual participants (lme4 package, version 1.1.37 [41]). Additionally, models assessing hourly ratings of acute drug effects included drug x hour interactions as fixed effects. Models assessing effects of PAS, ST, BDNF, and motor learning included participants’ reported number of hours of exercise and sleep in the past 24 h as covariates [42]. When applicable, baseline values were included as covariates in the models. The two models assessing N1 and P2 auditory ERP amplitudes included fixed effects for tetanus condition (pre- or post-tetanus), and all post-tetanus timepoints were included individually without averaging or other transformations. These models also included three-way interaction effects between drug, visit and tetanus condition. The two models assessing TMS amplitude and latency included fixed effects for PAS condition (pre- or post-PAS) and interaction effects in the same manner, and they additionally included a fixed effect for number of pulses counted during PAS. Following the lack of tetanus and PAS effects, the models were re-run without effects for tetanus and PAS conditions. Models assessing motor learning included fixed and effects for learning condition (online or offline) with each block included individually, as well as an interaction effect between drug and learning condition. Robust estimators were used when assumptions of linear mixed effects models were violated (robustlmm package, version 3.3-3 [43]). Because there were no significant order effects, all LSD and placebo visits were analyzed together. Post-hoc contrasts at each timepoint (on dosing days, one day later, and one week later) were conducted using estimated marginal means using either pairwise comparisons or linear trends (emmeans package, version 1.11.1 [44]). P values were adjusted using the multivariate t method to control the familywise error rate. Sensitivity analyses were used to test whether results were driven by outliers or influential cases. Correlations were calculated using Spearman’s correlation coefficient with FDR correction for multiple comparisons.

Forty-five volunteers were included in the study, and data from 43 (24 women, mean age = 30.4) were analyzed (CONSORT [45] diagram in Supplementary Fig. S4). Demographics are shown in Supplementary Table S1.

LSD significantly increased hourly ratings of most drug effects, as well as scores on the 5D-ASC and MEQ (Supplementary Fig., Supplementary Table). Most participants reported positive effects of LSD on well-being (Supplementary Table). Regarding blinding, all participants correctly identified the 100 µg LSD dose. Blinding of the placebo as low-dose LSD was successful in 79% of participants (mean guess = 14 µg, SD = 11.3). LSD increased the number of acute and post-acute side effects for up to 24 h, though they were generally mild and transient (Supplementary Tables,, Supplementary Fig.). There was one possibly treatment-related serious adverse event that resolved within the study timeframe (see supplement for a detailed discussion of this and other side effects). S5 S2 S3 S4 S5 S6

We first investigated whether ST caused an LTP-like potentiation of ERPs. Contrary to expectations, there was no main effect of ST on amplitude of N1 (β = 0.05, SE = 0.13, p = 0.70) or P2 (β = 0.15, SE = 0.18, p = 0.41) (Supplementary Table S6, Supplementary Fig. S7). The ST responder rate was 43% for N1 and 53% for P2. Planned contrasts by visit revealed that one week after treatment, P2 amplitude significantly decreased after ST in the LSD condition only (β = −0.47, SE = 0.17, p < 0.05). The corresponding drug x ST interaction was initially significant (β = −0.51, SE = 0.25, unadjusted p = 0.04) but did not survive correction (adjusted p = 0.14). There were no other significant drug x ST or visit x ST interactions. To further investigate the post-tetanus reduction in P2 at one week, we analyzed change in P2 amplitudes over all four recording timepoints within that EEG session (one pre-ST, three post-ST). P2 amplitude significantly decreased in the LSD condition compared to the placebo condition (β = −0.22, SE = 0.10, p < 0.05) with a steady linear trend across all four timepoints (Supplementary Fig. S8); this is not consistent with the expected effects of ST.

Regardless of ST, LSD showed main effects on ERP amplitudes (Fig. 3, Supplementary Fig. S9, Supplementary Table S7). Models analyzing all N1 and P2 ERPs together for each visit showed that LSD acutely and significantly reduced N1 (β = 0.5, SE = 0.08, p < 0.001) and P2 (β = −1.33, SE = 0.11, p < 0.001) amplitudes. P2 amplitude was still significantly reduced one day after LSD (β = −0.32, SE = 0.11, p < 0.01), which was the only significant post-acute effect. There was no effect of LSD on evoked theta power at any timepoint (Supplementary Table S8).

We first analyzed whether PAS induced LTP-like effects. There was neither a main effect of PAS on MEP amplitude (β = 7.4, SE = 43.9, p = 0.87) nor any significant PAS x drug interaction for any visit (Supplementary Table S9). The responder rate for PAS was 53.4%. We therefore analyzed all MEPs for each visit together to assess LSD’s impact on corticospinal excitability only.

MEP waveforms are shown in Fig. 4A. There was a significant LSD effect on MEP amplitude (β = 64.4, SE = 43.9, p < 0.01). Post-hoc tests revealed that MEP amplitude was significantly increased in the LSD condition on dosing days (β = 64.4, SE = 26.6, p < 0.05) and significantly decreased in the LSD condition one day after dosing (β = −80.8, SE = 26.1, p < 0.01), compared to placebo (Fig. 4B; Supplementary Fig. S10, Supplementary Table S10). Additionally, there was a main effect of LSD on MEP latency (β = −775.2, SE = 119.8, p < 0.001). Post-hoc tests showed that this effect was driven by dosing days, on which the LSD condition showed significantly decreased MEP latency (β = −775.2, SE = 119.8, p < 0.001) (Fig. 4C, Supplementary Fig. S10). During PAS on dosing days, participants in the LSD condition counted a mean of 163 TMS pulses compared to 177 (out of 180) in the placebo group; this effect approached significance (β = 4.69, SE = 2.67, p = 0.08). The number of pulses counted did not differ between groups one day (β = −0.71, SE = 2.63, p = 0.79) or one week after dosing (β = −1.58, SE = 2.58, p = 54).

There was no effect of LSD on BDNF in either serum (β = 0.49, t = 0.41, p = 0.69) or plasma (β = 0.26, t = 1.6, p = 0.11) at any timepoint (Supplementary Table S11).

On the motor learning task, mixed effects models showed a significant main effect of condition (β = −216.2, SE = 7.4, p < 0.001), indicating that participants’ typing speed improved in the test blocks (offline learning) relative to practice blocks (online learning). There was also a significant drug x condition interaction (β = 24.4, SE = 10.5, p < 0.05). Post-hoc tests showed that the LSD condition showed greater offline speed improvements than the placebo condition (β = −22.32, SE = 9.7, p < 0.05; Fig. 2C, Supplementary Table S12). This was primarily driven by the second test block completed 80 minutes after the end of practice (β = −37.0, SE = 12.6, p < 0.01). In the second test block, the LSD condition was 348 ms faster and the placebo condition was 262 ms faster than in the final practice block, corresponding to a 32.8% greater improvement in the LSD condition. LSD did not affect speed improvement during practice (β = −2.1, SE = 4.8, p = 0.66), initial speed in the first practice block (β = 38.2, SE = 81.0, p = 0.64) or accuracy during online (β = 0.8, SE = 0.6, p = 0.16) or offline learning (β = 1.1, SE = 1.2, p = 0.36), nor were there block-level effects on accuracy. Finally, neither motor learning nor questionnaire scores correlated with neurophysiological outcomes (Supplementary Table S13).

One week after dosing, LSD significantly increased ratings on the Alternatives subscale of the CFI (β = 1.92, SE = 0.63, p < 0.01), but not Control (β = −0.53, SE = 0.67, p = 0.42; Fig. 5, Supplementary Fig. S11, Supplementary Table S14). LSD also significantly reduced PSS scores (β = −1.94, SE = 0.91, p < 0.05; Fig. 5, Supplementary Fig. S11, Supplementary Table S14).

This study had several important limitations. The sample size of 43 may have been underpowered for detecting some interaction effects or complex order effects. Because ST and PAS did not potential neural signals, we could not evaluate LSD’s effects on LTP. The effect of LSD on P2 ERPs one week after dosing should be considered exploratory. Additionally, the baseline covariates included in statistical models were not re-assessed at each visit and we did not include plasma drug concentrations. The motor learning task was only given one day after dosing due to time constraints, but future studies would benefit from investigating the effects of LSD on learning at later timepoints, as well as whether learning gains are stable over time. Several findings were difficult to interpret mechanistically without quantification of possible confounders. Specifically, we did not measure possible confounding factors of motivation, attention or arousal during motor learning or EEG acquisition, which could have impacted the results, nor did we quantify peripheral muscle activity prior to TMS pulses. Finally, we attempted to reduce placebo effects by presenting the placebo as a low dose of LSD, but all participants still identified the real LSD dose due to its strong psychoactive effects. Because many participants in trials of psychedelics have positive expectations of how the drug will affect them, this failure of blinding could have strongly influenced self-report and behavioral outcomes [101].

The data supporting the conclusions in this article are available from the corresponding author upon reasonable request.