Modulation of long-term potentiation following microdoses of LSD captured by thalamo-cortical modelling in a randomised, controlled trial

Microdosing psychedelics — repeatedly taking sub-hallucinogenic doses — is reported anecdotally to benefit mood and cognition, but controlled clinical evidence for mechanisms is sparse. Preclinical work suggests serotonergic psychedelics such as LSD are psychoplastogens that can enhance structural and functional markers of neuroplasticity; however, direct measures of synaptic plasticity (for example long-term potentiation, LTP) are not feasible in human trials. Non-invasive EEG paradigms that induce visual LTP-like responses (the Teyler protocol) provide an indirect index of Hebbian plasticity by measuring changes in visual evoked potentials (VEPs), notably the early N1b and the later P2 components, which have been related to short-term potentiation and early LTP respectively. Murphy and colleagues set out to test whether repeated microdoses of LSD modulate neural plasticity as indexed by a visual LTP EEG paradigm. They combined conventional scalp-level ERP analyses with source-localised dynamic causal modelling using a biologically informed thalamo-cortical model (TCM) to infer laminar and thalamic microcircuit changes. The primary hypothesis was that LSD microdosing would increase potentiation of LTP-associated ERPs (particularly late P2, possibly early N1b) after six weeks of dosing, and that the TCM would reveal parameter changes consistent with enhanced plasticity; acute (2.5 h) effects were treated as exploratory.

This was a randomised, placebo-controlled trial conducted at the University of Auckland. Eighty healthy male volunteers were randomised 1:1 to receive either 10 µg LSD base sublingually or an inactive placebo. Only males were recruited because menstrual-cycle variation affects LTP measures and phase-locking was not feasible in the available sample. EEG assessments occurred at three timepoints: a drug-free Baseline, a Treatment session one week later (EEG recorded 2.5 h after the single administered dose), and a drug-free Final session six weeks later conducted two days after participants had completed 14 repeated microdoses given every third day. Participants were instructed to avoid alcohol for 24 h before EEG and to maintain usual caffeine; recording times were kept consistent within participants where possible. Exclusions from specific analyses were applied if data were missing/corrupted or if protocol timing/dosing criteria were not met (specific counts are reported in the Results). Sixty-four channel EEG was acquired (Brain Products system); full acquisition and pre-processing details are provided in the supplementary material. The visual LTP (Teyler) protocol presented horizontal and vertical high-contrast sine gratings across four phases: Baseline low-frequency presentation (1 Hz, 120 trials per orientation), a photic tetanus (one orientation at 9 Hz for 1000 presentations over 2 min, orientation counterbalanced), an early post-tetanus 1 Hz block (2 min after tetanus) to capture short-term potentiation (STP), and a late post-tetanus 1 Hz block 40 min later to capture early LTP (e-LTP). ERP analysis used FieldTrip and SPM12 to compute difference waves (Early: 2 min post-tetanus minus pre-tetanus; Late: 40 min post-tetanus minus pre-tetanus) within an occipito-parietal region of interest (a predefined list of electrodes). Time windows for N1b (early) and P2 (late) were determined from parameter-finding analyses. Main planned tests were four separate 2 × 2 repeated-measures ANOVAs of Group (Placebo vs. LSD) × Session (Baseline vs. Treatment or Baseline vs. Final) for Early and Late contrasts, with familywise error-corrected statistics (FWE-c p < 0.05) for main/post-hoc effects; interaction effects were considered at an uncorrected p < 0.001 as per prior literature. For mechanistic inference, source localisation identified a peak voxel of LTP effects in visual cortex and extracted a 5 mm virtual local field potential (LFP) per participant. A biologically grounded thalamo-cortical model (TCM) parameterising interlaminar cortical cell types (superficial pyramidal, superficial interneurons, spiny stellates, deep pyramidal, deep interneurons) and thalamic populations, plus receptor/channel decay constants (AMPA, NMDA, GABA-A/B, M- and H-channels), was fitted to the virtual LFPs using Dynamic Causal Modelling (DCM) in SPM12. Two contrasts were encoded to capture non-linear STP and linear e-LTP changes across blocks (contrast vectors [-1 1 0] and [-1 0 1]), and group- and session-level parameter differences were analysed using parametric empirical Bayes (PEB) in a hierarchical 'PEB of PEBs' approach. Baseline vs. Treatment modelling included 73 participants and Baseline vs. Final modelling included 51 participants after exclusions.

Sample sizes after exclusions were: Baseline vs. Treatment N = 73 (placebo n = 36, LSD n = 37) and Baseline vs. Final N = 51 (placebo n = 26, LSD n = 25). Reasons for exclusion included missing/corrupted data, inter-visit interval < 1 week for acute contrasts, Final session not two days after last dose, or participants not having exactly 14 doses. ERP (scalp-level) results — Acute (Baseline vs. Treatment): Early (N1b window). A Group × Session interaction was detected in a right-lateralised cluster peaking at 198 ms (F(1,142) = 13.05, FWE-c p = 0.0201). There was also a lateralised main effect of Group with peaks at ~175 ms (left) and ~170 ms (right), where the LSD group showed less potentiation than placebo (left peak t(142) = 4.30, FWE-c p = 0.0011; right peak t(142) = 3.97, FWE-c p = 0.0033). Post-hoc testing indicated the interaction was driven by a within-group change in the placebo group (more negative component at Treatment versus Baseline peaking ~200 ms), whereas the LSD group showed no within-group change in this region. Inspection of the raw ERPs suggested that the interaction cluster fell just prior to the P2 peak rather than neatly within a canonical N1b window. ERP — Acute Late (P2 window). No Group × Session interaction or main effect of Group was found; there was a main effect of Session with a central cluster peaking at 192 ms (F(1,142) = 15.6, FWE-c p = 0.0153), indicating that P2 post-tetanus amplitude was reduced on the Treatment day relative to Baseline across groups (t(142) = 3.95, FWE-c p = 0.0077). ERP — Sustained (Baseline vs. Final): Early condition showed no significant effects (no Group × Session interaction, no main effects). Late condition produced a Group × Session interaction with a left-lateralised peak at 203 ms (F(1,98) = 12.16, reported FWE-c p = 0.0596 and uncorrected p = 0.0007). There was a strong main effect of Session (peak 222 ms, F(1,98) = 34.07, FWE-c p < 0.0001). Post-hoc tests revealed this was driven by a reduction in P2 potentiation in the placebo group at the Final visit relative to Baseline (placebo within-group t(98) = 5.41, FWE-c p = 0.0001); the LSD group showed no significant within-group change. No between-group differences at either Baseline or Final were found. Computational modelling (TCM/DCM): The re-estimated DCMs explained > 99% of variance for both Baseline vs. Treatment and Baseline vs. Final fits. Parameters reaching 'very strong' evidence (posterior probability > 0.99) differed between groups. In the Baseline vs. Treatment analysis, the LSD group showed relatively greater modulation than placebo in inhibitory connections SI→SP (superficial interneuron to superficial pyramidal), SI→SS (superficial interneuron to spiny stellate), intrinsic SP→SP self-gain of superficial pyramidal cells, and AMPA channel decay modulation. Conversely, placebo showed relatively greater modulation in excitatory feedforward SP→DP (superficial to deep pyramidal), intrinsic SS→SS, and NMDA channel decay. In the Baseline vs. Final analysis, the SI→SS modulation remained greater in the LSD group, while SP→DP feedforward and NMDA decay modulation remained greater in the placebo group. These modelling results indicate laminar-specific shifts in excitatory and inhibitory microcircuit parameters associated with LSD versus placebo across acute and sustained contrasts.

Murphy and colleagues interpret the data as showing no clear LSD-induced modulation of canonical ERP peak measures of LTP (N1b, P2) either acutely or after six weeks of microdosing; instead, the placebo group exhibited unexpected session-to-session changes (increased early potentiation at Treatment and reduced late P2 potentiation at Final) while the LSD group remained comparatively stable. The absence of a conventional P2 enhancement like that reported after full-dose ketamine may reflect several factors the authors discuss: the timing of the Treatment recording (2.5 h post-dose may be too early relative to downstream plasticity markers), the very low LSD dose used, and the use of healthy participants rather than a clinical sample in which plasticity-related gains might be more evident. By modelling source-localised LFPs with a biologically grounded thalamo-cortical model, the investigators identified laminar and receptor-channel parameter changes that were not apparent at the scalp ERP level. Acute LSD relative to placebo was associated with greater modulation of inhibitory inputs to layer 4 spiny stellates and superficial pyramidal cells, and increased intrinsic self-gain of superficial pyramidal cells; placebo showed relatively greater modulation of superficial-to-deep (layer 2/3 → layer 5) feedforward connections and NMDA decay. The authors link these findings to existing frameworks: reduced modulation of feedforward SP→DP under LSD may reflect maintained sensitivity to bottom-up signals, consistent with predictive-coding models in which psychedelics relax top-down priors. The increased GABAergic-type parameters under LSD may reflect LSD-driven suppression of neuronal firing and serotonin-mediated enhancement of interneuron activity, and differential AMPA/NMDA modulation aligns with some preclinical reports of low-dose LSD potentiating AMPAR responses. The authors emphasise caution. Although the TCM provided a finer-grained signature of drug-related changes, many laminar effects were confined to the modelled primary visual cortex LFP and did not produce a corresponding scalp-ERP effect; this raises questions about sensitivity, robustness and generalisability. Limitations acknowledged include potential mistiming of recordings relative to the plasticity time course, the low microdose employed, uncertain test–retest reliability of the Teyler visual LTP protocol across repeated sessions, and the model's lack of an explicit 5-HT2A receptor parameter — a potentially relevant omission for serotonergic psychedelics. The authors recommend future work varying dose and post-dose intervals, attempting replication with counterbalanced session orders, testing other non-invasive LTP modalities (auditory, tactile, TMS), extending modelling to other cortical regions, and relating LTP measures to higher-order cognitive outcomes such as memory performance.

Analysis of conventional ERP peaks did not show evidence that LSD microdoses modulated visually induced LTP, but dynamic causal modelling of the full evoked waveform detected laminar microcircuit changes in visual cortex consistent with LSD altering feedforward aspects of neural plasticity. The authors conclude that a model-based source-level approach may be more sensitive than traditional ERP peak analyses for detecting the effects of low psychedelic doses, and that these modelling-derived signatures warrant further replication and exploration across doses, time windows, and modalities.