This rat study investigated the neural correlates of LSD-induced abnormal perceptions while animals navigated a familiar track. It finds that LSD suppresses hippocampal-cortical interactions and degrades internal spatial representations, isolating them from external sensory input, which may underlie the phenomenon of hallucinations.
Lysergic acid diethylamide (LSD) produces hallucinations, which are perceptions uncoupled from the external environment. How LSD alters neuronal activities in vivo that underlie abnormal perceptions is unknown. Here, we show that when rats run along a familiar track, hippocampal place cells under LSD reduce their firing rates, their directionality, and their interaction with visual cortical neurons. However, both hippocampal and visual cortical neurons temporarily increase firing rates during head-twitching, a behavioral signature of a hallucination-like state in rodents. When rats are immobile on the track, LSD enhances cortical firing synchrony in a state similar to the wakefulness-to-sleep transition, during which the hippocampal cortical interaction remains dampened while hippocampal awake reactivation is maintained. Our results suggest that LSD suppresses hippocampal-cortical interactions during active behavior and during immobility, leading to internal hippocampal representations that are degraded and isolated from external sensory input. These effects may contribute to LSD-produced abnormal perceptions.
Lysergic acid diethylamide (LSD) is a potent hallucinogen that produces perceptions uncoupled from the external environment. Prior human imaging work indicates LSD and related hallucinogens alter activity in prefrontal, visual cortical and hippocampal regions and disrupt functional connectivity among them, but the effects of LSD on single‑neuron firing in these areas during behaviour are unknown. Rodent models permit simultaneous in vivo recording of spikes and local field potentials (LFPs) during behaviour, and rodents reliably show a head‑twitch (HT) response to hallucinogens that depends on 5HT2A receptor activation and involves the hippocampus and visual cortex. C. and colleagues therefore recorded spiking and LFP activity from hippocampal CA1 and visual cortical (VC) regions in freely moving rats running a familiar C‑shaped track. The study aimed to determine how systemic LSD (two doses) alters CA1 place‑cell firing, population spatial representations, CA1–VC interactions during active running and immobility, and the relationship of neural changes to behaviour (lap rate, speed, immobility, HTs). The authors also tested whether a 5HT2A antagonist (M100907) blocks LSD effects, and examined LFP events (ripples, high‑voltage spikes) that accompany immobility.
Papers cited by this study that are also in Blossom
Alamia, A., Timmermann, C., Carhart-Harris, R. L. · eLife (2020)
Carhart-Harris, R. L., Erritzoe, D., Williams, T. et al. · PNAS (2012)
Carhart-Harris, R. L., Muthukumaraswamy, S., Roseman, L. et al. · PNAS (2016)
Fantegrossi, W. E., Murnane, K. S., Reissig, C. J. · Biochemical Pharmacology (2007)
Gonza ´lez-Maeso, J., Weisstaub, N. V., Zhou, M. et al. · Neuron (2007)
Adult rats were pre‑trained to run back and forth along a 3.5 m C‑shaped track for condensed‑milk reward. On recording days the behavioural protocol comprised two running sessions (PRE and POST) separated by a multi‑hour sleep period. Fifteen minutes before PRE animals received saline; fifteen minutes before POST they received one of several second injections: high‑dose LSD (0.24 mg/kg; “LSD high”), low‑dose LSD (0.06 mg/kg; “LSD low”), a 5HT2A antagonist M100907 (0.2 mg/kg) alone, M100907 followed by LSD low, or saline as control. POST durations varied to obtain sufficient laps; immobility analyses were restricted to the first 30 min of POST to reduce confounds from session length. Electrophysiology used tetrodes to record spikes and LFPs from CA1 and VC (and ACC LFPs in some animals). Data come from 17 rats with single‑unit recordings across 20 days (781 CA1 neurons, 153 VC cells); additional animals contributed LFP or behavioural data. Spikes were sorted offline; LFP and spike sampling and filtering parameters are reported. Behavioural measures were lap rate, instantaneous running speed, percent immobility, and HT counts; in a subset (N = 8) EMG from neck muscle provided precise, automatically detected HT start/end times (eHTs). Analyses defined active cells as firing >0.5 Hz on at least one trajectory, CA1 interneurons as >5 Hz, and silent cells as <0.5 Hz. Key analyses included: (1) firing‑rate comparisons during running and immobility, with speed effects removed by lap‑by‑lap linear regression to compute residual rates; (2) single‑cell spatial metrics — spatial information (SI), stability (Pearson correlation of firing‑rate curves PRE vs POST), directionality (correlation between rate curves on opposite trajectories), place‑field detection (peak threshold 3 Hz; field boundaries at 10% peak) and within‑field rates (also speed‑corrected); (3) population‑level analyses — population vectors (PVs) and cross‑session/cross‑trajectory PV correlations at spatial bins with ≥5 active cells; (4) oscillatory and phase measures — theta peak frequency/power, theta phase tuning, phase precession and theta sequences; (5) pairwise interactions — normalized cross‑correlograms against shuffled spike trains to obtain Z‑scored coactivity around lag 0, for CA1–CA1 and CA1–VC pairs; (6) multiunit activity (MUA) cross‑correlations between CA1 and VC; and (7) event detection — ripples (100–250 Hz, thresholded at 6 SD, 30–400 ms duration) and cortical high‑voltage spikes (HVSs; 6–12 Hz trough thresholds). Statistical tests (non‑parametric and parametric as appropriate) and sample sizes for key subsets (e.g. numbers of cells per dose group) are reported in the Results.
Behavioural effects: Under LSD rats ran fewer laps, ran more slowly and spent more time immobile in POST than PRE. The median lap rate under LSD was dramatically reduced (authors report an 88% reduction; PRE median 2.3 laps/min, p = 2.4×10‑4), with a smaller reduction in the control condition (46% lower). Running speed (excluding immobility) declined under LSD (54% lower; p = 1.4×10‑10). Percentage immobility in the first 30 min of POST increased from 40% [38%, 53%] to 78% [61%, 86%] under LSD (p = 0.017). Head‑twitch (HT) events were rare in PRE but frequent in POST under LSD: HT rate increased from 0.047 [0.0, 0.12] to 0.58 [0.38, 1.0] HTs/min (p = 4.9×10‑4). HTs were abolished when LSD low followed pre‑treatment with M100907, indicating 5HT2A dependence. Firing‑rate changes: From the set of active cells (365 CA1, 130 VC), CA1 firing rates during running were significantly reduced in POST under LSD high (POST median 0.45 [0.054, 0.89] Hz, p = 9.8×10‑22) and LSD low (PRE 1.2 [0.64, 2.1] Hz to POST 0.92 [0.58, 1.6] Hz, p = 0.015), but not in control conditions. VC neuron mean rates during running were not significantly altered by LSD. Speed‑corrected residual firing rates in CA1 remained significantly lower in POST for LSD high (PRE residual 0.053 [0.0053, 0.10] Hz; POST residual −0.43 [−1.0, −0.047] Hz; p = 3.1×10‑15) and LSD low (p = 0.024), indicating changes were not solely attributable to slowed running. During immobility, CA1 rates were reduced under LSD high (PRE 0.21 [0.073, 0.53] Hz to POST 0.073 [0.023, 0.24] Hz; p = 6.1×10‑14); VC rates during immobility were reduced by LSD high (PRE 3.1 [1.0, 7.0] Hz to POST 2.2 [0.70, 5.3] Hz; p = 0.0097). Putative CA1 interneurons and previously silent CA1 cells also showed rate reductions, suggesting a broad suppression in CA1. Head‑twitch–related activity: Using EMG‑detected eHTs (median duration 236 ms), the authors found CA1 and VC firing rates exhibited a broad increase around eHTs (one‑way ANOVA: CA1 p = 5.0×10‑113; VC p = 1.0×10‑116). VC rate increases peaked ~0.8 s before eHT start and CA1 peaked ~0.4 s after eHT start, consistent with VC activity leading CA1 around HTs. The elevation began seconds before and persisted beyond the brief HT movements. Place‑cell spatial properties: Spatial information (SI) values were not significantly different PRE to POST under LSD or control. Place‑field stability (Pearson correlation of rate curves PRE vs POST) showed no consistent reduction under LSD; notably LSD low sessions displayed higher median stability than LSD high and control (authors report uncertainty about that finding). However, directionality was reduced under LSD: the spatial correlation between rate curves on opposite trajectories increased from PRE to POST under both LSD high and LSD low (e.g. LSD high PRE median −0.0 to POST 0.12; p = 0.017), indicating less direction‑specific firing. The number of place fields per trajectory was unchanged, but median field length was slightly decreased under LSD (small but significant reductions reported), and within‑field firing rates were significantly lower in POST under both LSD doses (example POST under LSD high 3.8 [2.5, 6.5] Hz; p = 4.7×10‑6). Speed‑corrected residual within‑field rates remained significantly reduced (LSD high p = 3.2×10‑21; LSD low p = 6.4×10‑4). Oscillations and fine temporal coding: LSD modestly lowered theta peak frequency (when matched for speed) but did not reduce overall theta power; theta coherence between CA1 and VC LFPs during running was unaltered. Theta phase tuning, phase precession within place fields and theta‑sequence measures were largely preserved for cells remaining active in POST. Population and pairwise analyses: Population‑vector (PV) cross‑session correlations (PRE vs POST) for active cells at locations with ≥5 active cells were not significantly different between LSD and control overall, indicating that for the subset of remaining active cells population maps were broadly preserved. However, cross‑trajectory PV correlation (same location, opposite trajectories) increased under LSD (PRE 0.084 ± 0.029 to POST 0.30 ± 0.05; p = 7.6×10‑4), reflecting ensemble‑level loss of directional specificity. CA1–VC interactions: Pairwise normalized coactivity (Z‑scored cross‑correlograms) for CA1–VC active pairs showed a narrowing of coactivity distributions from PRE to POST under LSD but not control. The PRE/POST correlation of coactivity values was significantly reduced under LSD (LSD high R = 0.29; LSD low R = 0.42) versus control (R = 0.62), indicating larger changes in pairwise interactions after LSD. Multiunit CA1–VC cross‑correlation peaks during running were reduced in POST under LSD high (mean correlation peak PRE 0.14 ± 0.01 to POST 0.11 ± 0.01; p = 0.015) and LSD low (PRE 0.11 ± 0.01 to POST 0.073 ± 0.01; p = 0.035), but not under control. Immobilty events, HVSs and ripples: During immobility LSD increased occurrence of cortical high‑voltage spikes (HVSs), a 6–12 Hz spike‑and‑wave event associated with wakefulness‑to‑sleep transition (WST). Percentage time in HVSs in the first 30 min of POST rose from 0.0% [0.0%, 0.82%] to 11% [0.21%, 19%] under LSD (p = 0.0039); median HVS onset in affected animals was 5.9 [3.3, 13.0] min. HVS waveform properties (amplitude, duration, frequency) were quantitatively similar to naturally occurring WST HVSs. CA1 ripple occurrence rate was reduced from PRE to POST under LSD, and within HVS events there were far fewer ripples in POST under LSD than during natural WST (WST: 13 [5.5, 20] ripples/min; POST: 2.5 [0.4, 3.7] ripples/min; p = 6.0×10‑4). Cross‑correlations between CA1 ripple‑band and VC HVS‑band LFPs did not show a clear peak either in WST or POST under LSD, consistent with weakened CA1–VC coupling during HVSs. Awake reactivation: Pairwise awake reactivation within CA1 (correlation between running coactivity and within‑ripple coactivity) was present in PRE and POST across conditions (e.g. LSD high PRE R = 0.23 p = 2.2×10‑30; POST R = 0.20 p = 7.5×10‑10) and did not differ significantly PRE to POST. By contrast, CA1–VC pairs showed no significant awake reactivation correlations in PRE or POST under any condition. Thus, CA1 replay‑like coactivation during ripples persisted under LSD but appeared isolated from coordinated cortical reactivation.
C. and colleagues interpret their results as evidence that LSD degrades hippocampal spatial representations and reduces hippocampal–visual cortical communication. During active running LSD (both doses) substantially lowered CA1 firing rates (mean and within‑field) and reduced directional specificity of place cells at single‑cell and ensemble levels, producing a less precise cognitive map despite preservation of place‑field locations and spatial information measures. The reduced CA1–VC coactivity and weaker CA1–VC MUA cross‑correlations during running indicate a functional miscommunication between hippocampus and sensory cortex that could underlie the degraded spatial representation. During immobility LSD promoted a cortical state resembling the wakefulness‑to‑sleep transition (WST), evidenced by increased cortical HVSs that were quantitatively similar to WST HVSs but occurred without a transition to slow‑wave sleep on the track. Ripples in CA1 were relatively suppressed within HVSs, producing fewer ripple events during cortical HVSs than occur during natural WST; accordingly, CA1 tended to show preserved awake reactivation within ripples but this reactivation was isolated from coordinated cortical reactivation. Around HTs — a 5HT2A‑dependent behavioural signature of hallucinogens — both CA1 and VC showed a broad, transient increase in firing that began seconds before HT onset and persisted after it; the VC peak preceded the CA1 peak, suggesting cortical lead in these transient events. The abolishment of HTs by M100907 confirmed the 5HT2A dependence of that behavioural signature. The authors situate their findings relative to human imaging studies that report hippocampal/parahippocampal reductions and altered hippocampal network connectivity under LSD and to clinical observations that decreased hippocampal–sensory coupling is associated with visual hallucinations in Parkinson’s disease. They note apparent discrepancies with some human resting‑state fMRI reports of increased visual cortical activity under LSD and propose several non‑mutually exclusive explanations: differences between immobility in rodents and resting with eyes closed in humans; fMRI signals may reflect large‑scale LFP events (such as HVSs) rather than changes in single‑unit firing; and much higher per‑kg doses used in rodent studies compared with typical human doses. The authors emphasise that their results point to a specific functional dissociation between sensory cortex and hippocampus under LSD, which could produce internal representations or percepts that are mismatched to external reality and thus contribute to hallucinations. Acknowledged uncertainties and limitations in the extracted text include the difficulty in translating rodent dosage and behavioural states to human experiences, the unclear causes of the unexpected higher stability observed in some LSD low sessions, and the possibility that signals tightly correlated with HT may exist in other brain regions not recorded here. The authors refrain from claiming direct equivalence between the rodent findings and human subjective hallucinations but propose that the identified hippocampal–sensory miscommunication is a mechanistic contributor worthy of further study.
The psychedelic drug lysergic acid diethylamide (LSD) is a potent hallucinogen that produces surreal hallucinations in humans, defined as subjective perceptions uncoupled from external environments. Previous studies in humans suggest that LSD and similar hallucinogens disrupt activities in the prefrontal cortex, visual cortex (VC), and the hippocampus (HP), as well as their functional connections with other regions, during restingand during active tasks. However, how LSD alters firing activities of neurons in these areas in vivo is unknown. Unlike human studies, neuronal firing activity in animals can be recorded in vivo during behavioral responses to psychedelic drugs. We set out to study the effects of LSD on neuronal activities in freely moving rats. Previous behavioral studies in rodents identified a number of behavioral changes, including reduced movement with increased immobility (but sometimes enhanced mobility possibly depending on dosage and timing). In addition, rodents display a unique response to LSD and similar hallucinogens called head-twitching (HT), which is a brief, rapid shake of the head in rats and miceor a head-bobbing motion in rabbits. HT is considered a behavioral signature of the LSD-produced mental state similar to human hallucination. HT in rodents and altered perceptions in humans caused by hallucinogens like LSD require activation of the serotonin 5-hydroxytryptamine-2-A receptor (5HT 2A R), which can be blocked by specific 5HT 2A R antagonists such as M100907. Previous studies show that the LSD-induced HT depends on HP. Furthermore, HP, as well as VC, is also important for visual hallucinations. In this study, we targeted neuronal firing activities, as well as local field potentials (LFPs), in the CA1 area of HP and VC in freely moving rats during their behavioral responses to LSD. HP place cells fire spikes at one or a few places (place fields) of an environment (O'. A population of place cells with place fields covering an environment is believed to encode an internal cognitive map of the environment (O'. Place fields are formed by integrating self-motion cues with external sensory input, especially visual cues. Indeed, during active maze running when CA1 LFPs display prominent theta (6-10 Hz) oscillations, the firing activities of CA1 and VC neurons are correlated. During resting and immobile behavior, firing activity patterns in CA1 are reactivated, possibly for planning or memory recall, at times when CA1 LFPs display high-frequency (100-250 Hz) ripple oscillations. In addition, CA1 and VC are coordinately reactivated during sleep for memory consolidation. Therefore, CA1 and VC are engaged in functional interactions in various behavioral states. Because LSD produces a mismatch between internal perception and external environment, we aimed to study whether the cognitive map in HP and the interaction between HP and VC are altered by LSD in vivo. To this end, we recorded firing activities of CA1 and VC neurons and LFPs in rats while they ran a familiar track before and after systemic injection of LSD. We analyzed whether and how LSD altered CA1 and VC neuronal activities during active running and during immobility on the track, focusing on place-coding properties of CA1 place cells and their interactions with VC neurons.
We trained rats to run laps back and forth (two trajectories) on a familiar C-shaped track (Figure) prior to the recording experiment. On each recording day, rats performed the same running task for two sessions (PRE and POST), separated by a sleep session (Figure). Fifteen minutes prior to PRE, rats received an injection of the saline vehicle. Fifteen minutes prior to POST, rats received another drug administration under various conditions (Figure), namely, injection of LSD at either a high (LSD high ) or low (LSD low ) dose or one of the control injections including the 5HT 2A R antagonist M100907, followed by LSD low , (B) Behavioral procedure is as follows: two track-running sessions (PRE and POST), with each following a drug administration, separated by a sleep session. Prior to PRE, saline was injected. Prior to POST, either LSD high , LSD low , or a control (Ctrl) condition (injection of the 5HT 2A R antagonist M100907 followed by injection of LSD low , M100907 alone, or saline alone) was administered. (C) Linearized spatial trajectories (black lines) and head twitches (HTs; red stars) of an example rat under LSD high in PRE and POST. White gaps, linearization artifacts when the animal made sharp turns at corners. M, two of the corners. (D-G) Lap rate (D), running speed (E), percentage of immobile time (F), and HT rate (G) in PRE and POST under LSD high (n = 10), LSD low (n = 3, except for G, which is n = 4), and the Ctrl (n = 14) conditions. Each dot is a session. Boxplot: median and [25% 75%] range values; same in other figures. *p < 0.05, **p < 0.01, ***p < 0.001. antagonist alone, or saline alone. We recorded 781 CA1 neurons and 153 VC cells from 17 rats on 20 recording days, with most of the rats (n = 14) only recorded 1 day per animal (Table; see STAR Methods). Six additional rats were used only for the behavioral experiment and/or LFP recordings without single-neuron activities, following the same track running and injection schedule.
To examine behavioral responses to LSD in our track-running task, we linearized the two trajectories on the track (Figure) and quantified running behavior by the number of laps per minute (lap rate), running speed, and percentage of immobility time in each session. For the quantification of immobility, because the duration of POST varied from animal to animal and was longer than that of PRE, which could naturally lead to more time in immobility, we limited the quantification to the first 30 min of POST, a duration comparable to PRE. The median lap rate in, median [25%, 75%] values, same below unless otherwise specified) was dramatically reduced (88% lower) from PRE (2.3 [1.9, 2.9]; p = 2.4 3 10 À4 , Wilcoxon rank-sum test), whereas the reduction was modest (46% lower) under the control (PRE: 1.7 [0.98, 2.2], POST: 0.91 [0.72, 1.3], p = 1.2 3 10 À4 ) condition (Figure). Although both groups demonstrated reduced laps in POST, the lap rate in POST under LSD was significantly lower than that under the control (p = 5.1 3 10 À5 , Mann-Whitney test) condition. The median running speed (after removing immobility periods) under LSD was also significantly reduced (54% lower) in POST from PREcm/s,cm/s, p = 1.4 3 10 À10 , Wilcoxon rank-sum test) and modestly reduced (29% lower) under the controlcm/s,cm/s, p = 0.0067) condition (Figure). Accordingly, the median percentage of immobile time under LSD in the first 30 min of POST was significantly increased (34% higher) from PRE (PRE: 40% [38%, 53%], POST: 78% [61%, 86%], p = 0.017, Wilcoxon rank-sum test) but not under the control (PRE: 42% [31%, 61%]; POST: 51% [42%, 65%], p = 0.27) condition (Figure). The data suggest that the rats ran fewer laps and ran slower in POST than in PRE, but this change was greater under LSD than the control condition. Furthermore, the animals spent more time immobile in POST under LSD but not under the control condition, consistent with previous studies. We then analyzed the HT events, which were identified from those animals video recorded on the track. HT rarely occurred in PRE but frequently occurred in POST under LSD across all locations of running trajectories (Figure). The HT rate (number of HTs per minute) in POST under LSD was significantly higher than that in PRE (PRE: 0.047 [0.0, 0.12], POST: 0.58 [0.38, 1.0]; p = 4.9 3 10 À4 , Wilcoxon rank-sum test; Figure). No increase in HT was observed in POST under the control condition (PRE: 0.0 [0.0, 0.031], POST: 0.0 [0.0, 0.0], p = 0.25). The HT rate was even higher in POST under LSD low, n = 4 sessions) than that under LSD high, n = 8, p = 0.028, Mann-Whitney test). When LSD was injected at the same low dose after the injection of M100907, HT was abolished in POST (PRE: 0.0 [0.0, 0.038], POST: 0.0 [0.0, 0.0], n = 6, p = 0.50). This result shows that LSD worked as expected from previous reports, inducing 5HT 2A R-dependent HTs (Gonza ´lez-.
We asked how firing activities of CA1 and VC cells were altered by LSD. We first analyzed the CA1 and VC cells that were active (firing rate, >0.5 Hz) on at least one of the trajectories in PRE or POST (active cells). A total of 365 active CA1 and 130 active VC cells were identified (Table). We examined the overall firing rates of these neurons during active running. The firing rates of CA1 cells were lower under LSD in POST (Figure). The median rate was reduced from PRE to POST under LSD high.0] Hz, POST: 0.45 [0.054, 0.89] Hz; p = 9.8 3 10 À22 , Wilcoxon rank-sum test) and under LSD low (PRE: 1.2 [0.64, 2.1] Hz, POST: 0.92 [0.58, 1.6] Hz; p = 0.015) but not under the control (PRE: 1.1 [0.59, 2.5] Hz, POST: 1.3 [0.62, 2.2] Hz; p = 0.57) condition. However, the firing rates of VC neurons were not altered by LSD (Figure). Their median rate was not significantly different between PRE and POST under LSD high] Hz, POST: 3.1 [1.1, 6.9] Hz; p = 0.22), LSD low (PRE: 5.0 [2.5, 10] Hz, POST: 6.1 [0.28, 8.8] Hz; p = 0.63), or the control (PRE: 4.0 [1.7, 8.1] Hz, POST: 3.7 [1.2, 8.9] Hz; p = 0.063) condition. A question is whether the firing rate changes in CA1 resulted primarily from the speed changes between PRE and POST. To address this question, we removed the effect of speed by using a regression analysis (see STAR Methods) and computed the average residual rate for PRE and POST (Figure). We found that the average residual rates of CA1 neurons remained significantly lower in POST than in PRE under LSD high (PRE: 0.053 [0.0053, 0.10] Hz, POST: À0.43 [À1.0, À0.047] Hz; p = 3.1 3 10 À15 , Wilcoxon rank-sum test) and LSD low (PRE: 0.033 [À0.059, 0.11] Hz, POST: À0.20 [À0.58, 0.18] Hz; p = 0.024) but not under the control (PRE: À0.026 [À0.20, 0.21] Hz,Hz; p = 0.34) condition (Figure). The average residual rates of VC cells were not significantly different between POST and PRE under LSD high (PRE: 0.023 [À0.10, 0.23] Hz, POST: À0.18 [À1.6, 0.43] Hz; p = 0.55), LSD low (PRE: 0.017 [À0.31, 0.66] Hz, POST: À0.52 [À1.5, 0.60] Hz; p = 0.81), or the control (PRE: À0.14 [À0.53, 0.079] Hz, POST: 0.26 [À0.49, 2.2] Hz; p = 0.25) condition (Figure). Thus, the effects of LSD on CA1 cell firing rates during running remained even after removing the speed modulation. We also examined firing rates of active CA1 and VC cells during immobility. The median rate of CA1 cells was significantly lower in POST than PRE under LSD high (PRE: 0.21 [0.073, 0.53] Hz, POST: 0.073 [0.023, 0.24] Hz; p = 6.1 3 10 À14 , Wilcoxon rank-sum test) but not under LSD lowHz, POST: 0.25 [0.076, 0.49] Hz; p = 0.45) or the control (PRE: 0.21 [0.063, 0.50] Hz, POST: 0.21 [0.071, 0.62] Hz; p = 0.73) condition (Figure). Similarly, the median rate of VC cells in POST was significantly reduced from that of PRE under LSD high (PRE: 3.1 [1.0, 7.0] Hz, POST: 2.2 [0.70, 5.3] Hz; p = 0.0097) but not under LSD lowHz; p = 0.82) or the control] Hz, POST: 3.0 [0.79, 8.0] Hz; p = 0.36) condition (Figure). Taken together, LSD reduced the firing rates of active CA1 and VC cells in a behavior-and dose-dependent manner. LSD low had a relatively moderate effect, reducing the rates of CA1 active cells during running. LSD high had a broader effect, reducing rates during both running and immobility and also reducing rates of active VC cells during immobility. In addition, we found that the firing rates of putative CA1 interneurons (firing rate >5 Hz) and those CA1 cells that were inactive (firing rate <0.5 Hz) during running (silent cells) were also reduced by LSD during running (Figure) and during immobility (Figure), suggesting that LSD appeared to affect all cell types in CA1.
Because HT is considered a behavioral signature of 5HT 2A Rmediated hallucinations, we examined the behavior and neural activities surrounding HTs. We took advantage of the electromyography (EMG) signals recorded from the neck muscle in a group of rats under LSD high and LSD low (N = 8, Table). EMGs displayed a large deflection whenever an HT event occurred, which allowed us to precisely determine its start and end times (Figure). Because HTs were rare in PRE and in control rats, we analyzed only the EMG-detected HT events (eHTs) in POST under LSD (LSD high and LSD low combined). The median duration of eHTs was 236ms (Figure). As expected, the velocity of head movement increased immediately at eHT start times and lasted slightly after eHT end times for a total duration of $300 ms (Figure). The animals' movement along the one-dimensional linearized trajectory slowed down between 100 ms before and 100 ms after eHT start times but resumed afterward (Figure). The analysis indicates that HTs occurred mostly when rats were running along a trajectory, slowed, head-shook quickly, and then continued running. To understand whether neural activities were altered during eHTs, we computed the average firing rates of active CA1 and VC cells triggered by eHT start times. We found that both CA1 and VC cells significantly increased their rates around eHTs (CA1: p = 5.0 3 10 À113 ; VC: p = 1.0 3 10 À116 , one-way ANOVA; Figuresand). The increase appeared to be broad (>4 s), starting before the eHT start times and ending well beyond the eHT end times. In addition, it appeared that VC rates peaked 0.8 s before the eHT start times, but CA1 rates peaked 0.4 s after the eHT start times, suggesting that the VC changes led the CA1's changes. Therefore, a temporary, broad firing rate increase occurred before and during HTs, despite an overall rate reduction in POST under LSD.
Given the firing rate changes during running, we asked how the place-coding properties of CA1 place cells were affected by LSD. We plotted a spike raster of example place cells and their firing rates at positions along a linearized trajectory (rate curve) in PRE and POST under LSD high and LSD low; Figure). The plots suggest that place field locations (legend continued on next page) Cell Reports 36, 109714, September 14, 2021 5 Article were relatively preserved between PRE and POST, but firing rates within place fields were reduced, especially under LSD high . In addition, place cells appeared more likely to fire at the same locations along the two opposite trajectories (directions), i.e., the directionality of place fields was reduced under LSD (Figuresand; Figure). We used several measures to quantify these observations, starting with overall properties of firing rate curves, followed by properties of individual place fields. These measures were computed from firing activities during active running, excluding those occurring during immobility and at both ends of the track. First, the spatial tuning of a place cell on an active trajectory was quantified by spatial information (SI; Figure). The median SI values were not significantly different between PRE and POST under LSD high (PRE: 2.0 [1.4, 2.4] bits/spike; POST: 1.9 [1.4, 2.6] bits/spike, n = 121; p = 0.46, Mann-Whitney test), LSD low (PRE: 1.9 [1.4, 2.4] bits/spike; POST: 2.0 [1.5, 2.6] bits/spike; p = 0.33), or the control (PRE: 1.6 [1.2, 2.5] bits/spike; POST: 1.8 [1.2, 2.3] bits/spike; p = 0.81) condition. Second, we quantified the stability of place cell firing locations on active trajectories between PRE and POST (Figure). The median stability values were significantly different among LSD high (0.59 [0.21, 0.85]), LSD low (0.85 [0.61, 0.94]), and the control (0.71 [0.22, 0.91]) condition (p = 5.6 3 10 À7 , Kruskal-Wallis test). Post hoc comparisons indicated that this difference was due to the higher median stability of LSD low , relative to that of LSD high (p = 5.3 3 10 À8 , Mann-Whitney test) and to that of the control (p = 0.0021), with no difference between LSD high and the control (p = 0.10). The reason for the seemingly higher stability under LSD low is unclear; it could be due to other factors such as the amount of experience on the familiar track (although rats were all trained on the track for >4 days). Nevertheless, our data show that LSD did not reduce the stability of place field locations. Third, to quantify place cell firing directionality, we computed a spatial correlation between a cell's rate curves on the two opposite trajectories in a session (Figure). A higher correlation means lower directionality. We found that the median correlation value was significantly increased from PRE to POST under both LSD high (PRE: 0.0 [À0.12, 0.34]; POST: 0.12 [À0.034, 0.50]; p = 0.017, Mann-Whitney test) and LSD low (PRE: 0.096 [À0.097, 0.59]; POST: 0.33 [0.053, 0.67]; p = 0.049) but not under the control condition (PRE: 0.15 [À0.12, 0.43]; POST: 0.11 [À0.098, 0.45]; p = 0.95). Thus, LSD reduced the directionality of CA1 place cells. Fourth, we detected individual place fields on each active trajectory and quantified their changes from PRE to POST by the number of place fields per trajectory, field length, and withinfield firing rate. The number of fields was similar between PRE and POST under LSD high [PRE: 1.1 ± 0.04 (mean ± SE), n = 244 cell 3 trajectories; POST: 1.1 ± 0.6, n = 133; p = 0.80, Student's t test), LSD low (PRE: 1.1 ± 0.06, n = 101; POST: 0.98 ± 0.08, n = 98; p = 0.30), and the control (PRE: 1.2 ± 0.08, n = 85; POST: 1.2 ± 0.09, n = 87; p = 0.74) condition. The median field length (Figure) was slightly, but significantly, decreased from PRE to POST under LSD highcm;cm; p = 7.2 3 10 À4 , Mann-Whitney test) and LSD lowcm;cm; p = 0.023) but not under the controlcm; POST: 50cm; p = 0.55) condition. The median within-field firing rate (Figure) was significantly lower in POST than in PRE under both LSD high] Hz; POST: 3.8 [2.5, 6.5] Hz; p = 4.7 3 10 À6 , Mann-Whitney test) and LSD low.9] Hz; POST: 4.9 [3.0, 7.0] Hz; p = 0.029) but not under the control (PRE: 5.2 [3.1, 8.9] Hz;Hz; p = 0.91) condition. To address the issue that the lower within-field rate might be due to slower speed under LSD, we removed the effect of speed on within-field rates in PRE and POST. We found that the median residual within-field rate in POST remained significantly lower under LSD high (PRE: 0.19 [0.058, 0.42] Hz, POST: À2.1 [À4.0, À0.76] Hz, n = 161 fields; p = 3.2 3 10 À21 , Wilcoxon rank-sum test) and LSD low (PRE: 0.14 [À0.099 0.34] Hz, POST: À0.87 [À2.5, 0.26] Hz, n = 69; p = 6.4 3 10 À4 ) but remained similar under the control (PRE: 0.090 [À0.92, 0.77] Hz, POST: À0.28 [À1.9, 2.2] Hz, n = 61; p = 0.80) condition (Figure). Taken together, these results indicate that place cell firing activities were largely stable under LSD, with similar SI and no reduction in stability. However, the spatial tuning became significantly less precise, with reduced directionality and a lower within-field firing rate. Previous studies show that place cells at different locations of CA1, either in superficial versus deep layers of the pyramidal layeror in proximal versus distal ends of CA1, differ in their firing rates or bursting properties. We asked whether LSD differentially affected place cells with these different characteristics. First, we separated the place cells recorded under LSD into three groups, as follows: those with stable fields between PRE and POST, those with unstable fields, and those active in PRE but silent in POST. The firing rates of the POSTsilent group were lower in PRE to begin with than those of the other two groups, whereas the burst index was not significantly different among the groups (Figure). This result is consistent with a broad rate reduction in all place cells. Second, we divided those cells that fired on both running trajectories in PRE and POST under LSD into two groups, namely, those with reduced directionality and those with directionality unchanged or increased. We found no difference in either firing rate or burst index between the two groups (Figure). In addition, we examined the identified recording sites in CA1 for rats under LSD high , LSD low , and the control condition and found they were comparable (Figure). Therefore, our results suggest that LSD affected place cells broadly in CA1, consistent with the rate change of CA1 silent cells and interneurons during running (Figure). Because some drugs such as cannabinoids can affect LFPs and fine spike timing of place cells, we examined how LSD affected theta oscillations in the CA1 LFPs, as well as theta phase tuning and phase precession of place cells. LSD appeared to lower theta peak frequencies, but not theta power, when compared at similar speed ranges between PRE and POST (Figuresand). But, the coherence in the theta band between CA1 and VC LFPs was unaltered (Figure). The mean theta phase and phase variance (a measure of the degree of theta phase tuning) of active cells were not significantly affected by LSD (Figuresand). Theta phase precession within CA1 place fields was apparently intact under LSD (Figures). In addition, theta sequences, measured by the pairwise correlation between place field distances and firing timing intervals within theta cycles, appeared normal under LSD (Figure). Therefore, theta oscillations were modestly affected, but theta phase precession and theta sequences of place cells remained under LSD.
After analyzing individual place cells, we examined how spatial representations of CA1 place cell populations were altered by LSD. We already showed that firing rates of CA1 cells reduced with LSD, necessarily leading to fewer active cells during trajectory running in POST. Here, we focused on those locations with a sufficient number of active place cells (R5) and asked whether the remaining active cells in POST encoded the locations similarly as in PRE at the population level. We constructed population vectors (PVs) from the active cells' rate curves and computed a correlation between PVs at the same location of the same trajectory between PRE and POST (crosssession PV corr ) or between PVs at the same location in the same session but between two opposite trajectories (cross-trajectory PV corr ). The cross-session PV corr quantifies how stable PVs were between PRE and POST. Plots of example trajectories suggest similar PV corr at the same locations under all conditions (Figure). We computed an average PV corr across all locations of a trajectory and compared the average PV corr for all trajectories between different conditions. We found that the cross-session PV corr did not significantly differ between LSD (LSD high and LSD low combined) and control conditions (LSD: 0.52 ± 0.04; control: 0.55 ± 0.07; p = 0.72, Student's t-test; Figure). By breaking down the LSD group to LSD high and LSD low , we showed that the cross-session PV corr under LSD low was modestly higher than that under LSD high , but it did not reach significance (LSD high : 0.46 ± 0.06, n = 16 trajectories; LSD low : 0.64 ± 0.08, n = 6; p = 0.099). This result suggests that population-level spatial representations were largely stable under LSD for the active cells in POST, consistent with the result of stable individual place cells (Figure). The cross-trajectory PV corr quantifies how similar PVs on one trajectory were to its opposite trajectory in the same session. Higher cross-trajectory PV corr means lower directionality at the population level. Plots of example trajectories suggest higher cross-trajectory PV corr at the same locations in POST than that in PRE under LSD high and LSD low but not the control condition (Figure). There was a significant increase in the average cross-trajectory PV corr from PRE to POST under LSD (PRE: 0.084 ± 0.029; POST: 0.30 ± 0.05; p = 7.6 3 10 À4 ; Student's ttest) but not from that of the control (PRE: 0.080 ± 0.05; POST: 0.14 ± 0.03; p = 0.35) condition (Figure). The increase under LSD was consistent in both the LSD high (PRE: 0.10 ± 0.05, n = 9 sessions; POST: 0.44 ± 0.07, n = 8; p = 0.0011) and LSD low (PRE: 0.043 ± 0.05, n = 3; POST: 0.21 ± 0.03, n = 3; p = 0.049) groups. Therefore, consistent with the reduced directionality in individual place cells, LSD reduced the directional specificity of spatial representations at the population level as well.
Our results so far indicate that LSD degraded spatial representations in CA1, with reduced firing rates and reduced directionality. To understand whether a miscommunication between HP and VC might underlie the degradation, we analyzed interactions between CA1 and VC cells under LSD. For each pair of CA1-VC cells active during running (active pair), we computed a normalized cross-correlogram relative to those computed from the shuffled spikes of the two cells. We then defined coactivity for the pair as the average correlation value within the time bins around the time lag 0 [À0.1, 0.1] s. Higher coactivity means a higher degree of firing together between the two cells during running. We examined whether coactivity values of CA1-VC active pairs were altered between PRE and POST under different conditions. Examples of normalized cross-correlograms suggest that the cell pairs with high coactivity in PRE maintained high coactivity in POST under all conditions, but with relatively lower peaks especially under LSD high (Figure). Indeed, the distribution of coactivity was narrowed from PRE to POST under LSD high and LSD low , but it was unchanged under the control condition (Figure). The narrower distribution could be due to overall changes in firing rates from PRE to POST under LSD. To examine more specific alterations in coactivity beyond the effects of rate changes, we computed a correlation between coactivity values in PRE and POST for all active pairs. There was a significant correlation in coactivity between PRE and POST under all conditions (LSD high : p = 6.3 3 10 À33 , Pearson's r; LSD low : p = 4.7 3 10 À18 ; control: p = 3.1 3 10 À33 ; Figure), suggesting that the VC-CA1 coactivity patterns overlapped significantly between PRE and POST. However, the correlation was significantly smaller under LSD high (R = 0.29, p = 2.1 3 10 À11 , Fisher's exact test) and under LSD low (R = 0.42, p = 1.6 3 10 À4 ) than that under the control (R = 0.62) condition. This result was true even when the CA1-VC active pairs were downsampled with a matched number of pairs and similar firing rates across the three condi-tions (Figure). The result indicates that the PRE/POST coactivity patterns were less overlapped under LSD than under the control. Therefore, more changes in CA1-VC interactions had occurred from PRE to POST under LSD. To further understand how the CA1-VC interaction changed, we examined the multiunit activities (MUAs) in CA1 and VC, which included all sorted and unsorted spikes recorded in an area. We computed the cross-correlation between normalized CA1 and VC MUAs during running. The average cross-correlogram in PRE, over all running periods and all rats under the same condition, had a clearly defined peak under LSD high , LSD low , and control (Figure). However, the peak was reduced in POST under LSD high and LSD low but not under the control condition (Figure). The mean correlation value at the peak was significantly lower in POST under LSD high (PRE: 0.14 ± 0.01; POST: 0.11 ± 0.01; p = 0.015; Student's t-test) and LSD low (PRE: 0.11 ± 0.01; POST: 0.073 ± 0.01; p = 0.035) but not under the control (PRE: 0.20 ± 0.01; POST: 0.18 ± 0.01; p = 0.11) condition. The result suggests that LSD reduced the strength of CA1-VC interactions.
After analyzing neural activities during running, we switched to the immobile behavior on the track, given the significantly enhanced immobility with LSD (Figure). As expected, the CA1 LFPs displayed ripple events in PRE and POST under all conditions, which were accompanied by bursts of spikes in MUAs of CA1 cells (Figure). The occurrence rate of ripples was reduced from PRE to POST under LSD with small changes in ripple frequency, duration, and amplitude (Figure). Unexpectedly, the VC LFPs frequently displayed a high-amplitude spike-and-wave event (Figure; Figure), called highvoltage spikes (HVSs), which naturally occurs during the wakefulness-to-sleep transition (WST) in rodents. Here, HVSs were observed during immobility on the track under LSD and accompanied by highly synchronized bursts in MUAs of VC cells (Figure). We analyzed properties of individual cortical HVS events in PRE and POST, as well as those in WST (in a sleep box prior to the LSD injection). For this analysis, we used cortical LFPs recorded from VC in 12 rats and those recorded from the anterior cingulate cortex (ACC) in 5 rats. Previous studies show that HVS events are highly synchronized across many cortical areas, including sensory and frontal cortices. Because our purpose was to examine the occurrence and properties of HVSs under LSD compared to those naturally occurring during WST, we combined the animals with cortical LFPs in VC and ACC and performed within-animal comparisons in HVS properties across sessions. To quantify HVS occurrence, we computed the percent of time in HVSs among total immobility time in PRE and POST. The percentage of time in HVSs was dramatically increased in the first 30 min of POST from that in PRE under LSD (PRE: 0.0% [0.0%, 0.82%]; POST: 11% [0.21%, 19%]; p = 0.0039, Wilcoxon rank-sum test) but not under the control (PRE: 0.0% [0.0%, 1.2%]; POST: 0.0% [0.0%, 0.0%]; p = 0.25) condition (Figure). The first HVS event occurred as early as 48 s on the track with a median onset time of 5.9 [3.3, 13.0] min in the 9 rats (out of 11) under LSD that displayed HVSs within the first 30 min of POST (Figure). The HVS occurrence plateaued around 12 min from the POST start (Figure). Furthermore, we did not observe any transition to slow-wave sleep (SWS) on the track. The rapid HVS onset and lack of SWS indicate that the occurrence of HVSs was not simply due to the rats falling asleep on the track. We compared properties of HVS in POST under LSD to those occurring in WST without LSD. Judging from their waveforms (Figure), the HVSs in POST appeared qualitatively similar to those in WST. Quantifying individual HVS events under LSD in POST did not reveal a significant difference in their median amplitude, duration, or frequency from those in WST (Figure). Given the prominence of ripples in CA1 and HVSs in the cortex, we asked how the two types of events interacted. We first examined the interaction at the broad LFP and MUA levels. Consistent with our previous study showing weak CA1-VC interactions during WST and unlike SWS, the cross-correlogram between CA1 LFPs in the ripple band and VC LFPs in the HVS band did not show an obvious peak in WST, and this result was not altered in POST under LSD (Figure). Similarly, the average VC MUAs triggered by CA1 ripple peak times did not show a clear response in WST or in POST under LSD (Figure). We then counted the number of ripples within each HVS event and computed the ripple occurrence rate per minute of HVS. We found a significant reduction in ripple rate within HVSs in POST under LSD from that in WST (WST: 13 [5.5, 20] ripples per min; POST: 2.5 [0.4, 3.7] ripples per min; p = 6.0 3 10 À4 ; Mann-Whitney test; Figure). Therefore, LSD during immobility promoted a cortical state similar to WST but with an even weaker interaction with CA1 ripples, consistent with the reduced CA1-VC interaction under LSD during running.
It is known that place cell activities during running are replayed within ripples during normal immobility on the track (awake replay), which can be quantified by a pattern analysis like Bayesian decoding that requires a large number of place cells. In our experiments, the number of cells simultaneously active on a trajectory was limited, especially under LSD high when CA1 rates were reduced. Therefore, we used a pairwise approach as in previous studies. For all pairs of CA1 cells active during running and with non-zero rates within ripples, we computed a correlation between their coactivities during running and those within ripples in the same session. A significant correlation means that cell pairs co-activated during running also coactivated within ripples, which is referred to as awake reactivation. We found that awake reactivation of CA1 pairs occurred in PRE and POST under all conditions (Figure). There was a significant correlation between running and ripple coactivities in PRE and POST under LSD high (PRE: R = 0.23, p = 2.2 3 10 À30 , Pearson' r; POST: R = 0.20, p = 7.5 3 10 À10 ), LSD low (PRE: R = 0.32, p = 2.7 3 10 À17 ; POST: R = 0.29, p = 4.0 3 10 À16 ), and the control (PRE: R = 0.38, p = 3.0 3 10 À17 ; POST: R = 0.44, p = 2.8 3 10 À25 ) conditions. The correlation values did not significantly differ between PRE and POST under LSD high (p = 0.24, Fisher's exact test), LSD low (p = 0.29), or the control (p = 0.12) condition. The same result was observed even when the CA1 pairs were downsampled to match the number of pairs and firing rates between PRE and POST (Figure). Thus, awake reactivation during ripples in CA1 persisted under LSD. We also examined whether such awake reactivation occurred across CA1-VC cell pairs. For this analysis, to account for the possibility that VC cells might fire spikes preceding or lagging CA1 ripples, we computed the cross-correlogram between a CA1-VC cell pair within a time window of [À1, 1] s from each ripple trough time and took the average value with the time lag window [À100, 100] ms as their coactivity. Our analysis found no awake reactivation across the two areas (Figure). There was no significant correlation between running and ripple coactivities for CA1-VC pairs in PRE or POST under LSD high (PRE: R = 0.022, p = 0.35, Pearson' r; POST: R = 0.044, p = 0.15), LSD low (PRE: R = 0.052, p = 0.30; POST: R = 0.069, p = 0.21), or the control (PRE: R = À0.0012, p = 0.98; POST: R = À0.021, p = 0.71) condition. The same was observed when the CA1-VC cell pairs were downsampled (Figure). Furthermore, such a lack of awake reactivation was also observed within HVS events in POST under LSD high (Figure). Thus, unlike cell pairs within CA1, the cells across CA1 and VC did not coordinately reactivate in ripple or HVS events, suggesting that the awake reactivation within CA1 under LSD was isolated without proper interactions with VC.
To understand how LSD dissociates a subject's internal perception from the external environment, we have investigated LSDinduced alterations in HP place cell activities and their interactions with VC neurons in freely behaving rats. As rats actively ran on a track, LSD high and LSD low lowered the firing rates of CA1 place cells and reduced their directionality at both the individual cell and ensemble levels. Despite the overall reduction of firing rates in CA1, both CA1 and VC cells temporarily increased their rates before and during HT events. Importantly, the interaction between CA1 and VC neuronal activities was reduced by LSD during running at both doses. During immobility on the track, LSD promoted a cortical state similar to that of WST, which was observed by the occurrence of cortical HVSs. However, different from WST, LSD in immobility further weakened interactions between ripples and HVSs but left the awake reactivation within CA1 preserved. Therefore, LSD reduces normal communication between HP and the sensory cortex, which consequently degrades the HP cognitive map during active running and promotes a state with isolated HP reactivation during immobility. These findings may contribute to the dissociated perceptions produced by LSD. Our data support that LSD alters HP spatial representation. The firing rates of CA1 place cells, both mean firing rates and within-field rates, were significantly reduced under LSD high and LSD low . The reduction persisted even after accounting for rate changes due to speed modulation. For those place cells that remained active under LSD, their directionality was reduced in both high and low doses, i.e., they tended to fire at the same locations despite the different trajectory directions. As such, place cells became less differentiated between the two directions. Despite these changes, the rate curves of place cells were stable, as measured by spatial correlation, indicating that their firing locations were relatively unaltered. The reduced directionality and stability of place cell activities were also observed at the ensemble level, measured by PV correlations. Therefore, our data suggest that LSD induces a less precise, degraded spatial representation of the external environment. Our finding suggests that this degraded spatial representation may be due to an abnormal interaction between HP and VC. It is well known that visual information is a crucial modality driving the formation of place fields. During track running, it has been shown that firing activities of CA1 place cells and VC neurons are functionally correlated. Our data here show that the coactivity between CA1-VC cell pairs changed more in POST than that in PRE under LSD high and LSD low , indicating a miscommunication between CA1 and VC cells. More importantly, the cross-correlations between CA1 and VC MUAs were weaker under LSD high and LSD low . Therefore, the CA1-VC interaction involved in place cell activities during running is likely reduced by LSD. The reduced functional interaction between CA1 and VC cells appears to occur during immobility as well. When animals stop on a track, ripple events take place with highly synchronized population bursts of CA1 place cells. We found that ripple rates and other parameters were modestly altered by LSD. In the cortex, we found that LSD dramatically enhanced the occurrence of HVSs, a type of event that normally occurs in WST as the brain shifts to a more internally generated state. During HVSs, cortical neurons are highly synchronized, whereas CA1 ripples occasionally occur. Our previous study suggested that HVSs and ripples during WST produce a weak, transitional interaction between the cortex and HP that eventually leads to a strong correlation between the two areas for memory consolidation in SWS. Here, we found that there were fewer ripples during HVS under LSD, suggesting a reduction in cortical-hippocampal consonance. In addition, CA1 cells had reduced firing rates during immobility under LSD high and LSD low , and VC neurons also had lower rates with the high dose. Thus, like during running, the CA1-VC communication during immobility is also reduced by LSD. In addition, our data confirm awake reactivation of CA1 activity patterns within ripples during immobility but provide no support for awake reactivation for the joint activity patterns between CA1 and VC. Importantly, LSD high and LSD low did not alter the lack of CA1-VC reactivation in our experiment. Because LSD enhanced immobility on the track, this finding further supports the idea that LSD reduces CA1-VC communication, namely, this time through the promotion of a behavioral state lacking coordinated reactivation, which possibly leads to isolated CA1 awake reactivation in the absence of engagement from the sensory cortices. Our study helps to understand how neural activity changes during the HT behavior. HT is believed to be a behavioral signature of a brain state similar to hallucination in humans, because the 5HT 2A R agonists that are hallucinogenic in humans also evoke HT in rodents, whereas those 5HT 2A R agonists that are non-hallucinogenic in humans do not evoke HT. Our analysis indicates a temporary increase in the firing rates of CA1 and VC cells when HT occurred, despite the overall reduction in the firing rates in POST under LSD. This increase was broad, starting seconds earlier than and lasting beyond HT events. We did not find specific, short-term changes immediately before HT in our data. However, it is possible that there exist signals in other brain areas that are more tightly correlated with HT. Nevertheless, based on our data, we speculate that the increase in firing rates is a temporary compensation for the overall rate reduction under LSD and that HT is a coping behavior in response to internal percepts associated with the temporarily increased neural activities in HP and VC. One of our findings is that LSD greatly promotes the occurrence of HVSs. Because HVSs naturally occur during WST, we propose that LSD produces a WST-like state on the track as animals become less engaged with the task and descend into immobility. Our analysis shows similar quantitative properties between LSDinduced and WST HVSs, suggesting that the LSD-induced state is similar to WST. However, unlike the HVS events in WST that are normally followed by SWS, we did not detect any SWS on the track even when strong HVSs occurred repeatedly under LSD. The observation suggests the cortex perpetuates a highly synchronized state without going to sleep. It is unknown how this cortical state is related to LSD-induced hallucinations. However, hypnagogic imagery does occur often in WST in humans, and a recent study shows that slow oscillatory events in the retrosplenial cortex are involved in the effect of dissociative drugs such as ketamine. In addition, the forward propagation of alpha waves, a potential signature of HVSs in humans, is enhanced by the hallucinogen N-Dimethyltryptamine or DMT. It is possible that this LSD-induced state similar to WST may contribute to hallucinations under LSD. A previous human fMRI study shows that activities in the parahippocampal gyrus are reduced and the functional connectivity in the hippocampal-prefrontal network is disrupted by LSD in an active task. Another study found that the functional connectivity between HP and VC is lower in patients with Parkinson's disease who display visual hallucinations. These findings are consistent with our results that CA1 neurons had lower firing rates and lower functional interaction with VC during running. When humans rested awake with eyes closed, fMRI studies found that VC is hyperactive under LSD, with enhanced functional connectivity with some brain areas, which seems inconsistent with our result that VC neurons reduced their rates during immobility. However, our finding may not necessarily contradict human studies. It is unclear whether the resting state in humans with eyes closed is the same as immobility in rodents. Second, it is also unclear whether the increase in fMRI signals directly translates to an increase in firing rates of individual neurons. It is possible that large LFP signals such as HVS events under LSD underlie the increased fMRI signals. Finally, the dosage of LSD used in rodent studies including ours (60 mg/kg or 240 mg/kg) is much higher than that in human studies (single dose of $100 mg), which could lead to differences in neurophysiological responses. Overall, our work reveals neurophysiological alterations that can advance our understanding of how LSD produces its powerful reality-altering effects. The reduced HP-VC interaction leads to degraded spatial representations during active tasks, resulting in an altered cognitive map different from the external environment. The temporary, compensatory increase in firing rates of VC and HP cells around HTs may further alter the sensory and memory processing under LSD. Furthermore, LSD prolongs the immobility behavior and promotes a WST-like state with enhanced cortical HVSs, during which HP reactivates spatial representations in isolation without the participation of the sensory cortex. Our findings contribute to the neural circuit mechanism of LSD-induced hallucinations by identifying a specific functional dissociation between sensory and memory circuits through VC and HP miscommunication, which may produce abnormal spatial representations of the external world and/or abnormal sensory percepts misaligned with external reality.
Detailed methods are provided in the online version of this paper and include the following: Behavioral procedure and drug conditions Animals were pre-trained to run back and forth (two trajectories) for food reward (condensed milk) along a C-shaped track, which was 3.5 m long. The pre-training lasted for at least 4 days until they reliably and consistently ran at least 20 laps on the track. After the pretraining, a behavioral procedure was conducted for one or more days per animal. Each daily procedure began with a rat resting in a sleep box followed by a subcutaneous injection of 0.5 mL saline. Fifteen minutes following the injection, the rat ran for food rewards back and forth along two trajectories on the now familiar (after pre-training) C-shaped track for 20 -30 minutes (PRE). The rat then slept in the sleep box for $3 hours between sessions, followed by another administration of 0.5 mL under various drug conditions. Fifteen minutes after the second drug administration, the animal performed the same task for 20 -70 minutes on the track (POST). The variation in the duration of POST was needed to collect sufficient number of running laps (R6) in a session, especially for the animals injected with LSD high who's running activities were greatly reduced. However, longer sessions might lead to increased immobility behavior or immobility-related neural activities that were drug-irrelevant. To eliminate this possibility, we restricted our analysis on immobility-related neural activities within the first 30 minutes in POST. The drug condition for the second administration was the injection of a high dose of LSD at 0.24 mg/kg (LSD high ), a low dose of LSD at 0.06 mg/kg (LSD low ), or one of the control injections including the injection of saline alone, the injection of a 5HT 2A R antagonist M100907 alone at 0.2 mg/kg, or the injection of M100907 at 0.2 mg/kg together with the low dose of LSD at 0.06 mg/kg. In the M100907 and LSD together condition, M100907 was administered 15 minutes prior to the injection of LSD. M100907 at this concentration is known to block 5HT2ARs at even high concentrations of LSD. If an animal was used in this procedure for more than one day, the condition on the first day was always saline or M100907 (Table). Electrophysiological recordings were performed during this behavioral procedure in 23 rats across 28 days of experiments. The other 2 days' experiments did not have neural recordings, but their behavioral data were included in the behavioral analysis.
Recordings were performed using tetrodes made by twisting 4 fine nichrome wires (diameter 13 mm; Sandvik Palm Coast, Palm Coast, FL), as previously described. During the two to four weeks post-surgery, tetrodes were descended slowly to the target regions: CA1, VC, or ACC. Neuronal spikes and LFPs, as well as the EMGs, were acquired using a Digital Lynx system (Neuralynx, MT). For the ACC recordings, only the LFPs were used in this study for analyzing cortical HVS events. In 19 animals, LFP signals were sampled at 2 kHz with a broad-band filter (0.1 -1 kHz); in other 4 animals LFPs were sampled at 4 -8 kHz. Spikes were identified using a preset threshold of 50 -70 mV from signals filtered within 600 Hz -9 kHz and sampled at 32 kHz. Animal positions were tracked using two diodes mounted to the tetrode drive and recorded by an overhead camera. Position data were sampled at 33 Hz with a resolution of $0.25 cm per pixel.
After the recording, animals were euthanized by pentobarbital overdose (200 mg/kg). For recording site identification, electrical lesions were made at each recording site by passing a current of 30 mA for 15 s. Brains were fixed with 10% formalin and sectioned at 90 -300 mm thickness. Sections were stained with 0.2% cresyl violet or 1% sodium sulfide nonahydrate (for acetylcholineesterase activity). Recording locations in CA1 and ACC were verified from the lesion marks in the cresyl violet stained sections. Recording sites in VC were determined from the sodium sulfide nonahydrate staining as shown in previous studies.
For the majority of analyses on neuronal activities, we considered LSD high and LSD low conditions separately because the number of neurons recorded under each condition (LSD high : N = 435 CA1 cells, N = 99 VC cells; LSD low : N = 177 CA1 cells, N = 22 VC cells; Table) was sufficient for the study of dose-dependent responses to LSD (see details below). However, for behavioral analyses and certain neuronal analyses that required individual rats as samples, we combined LSD high and LSD low together, due to the limited number of animals in each condition (LSD high : N = 10 rats, LSD low : N = 4 rats). Although the combination did not permit the study of dosedependent responses in these measures, it still allowed us to examine the effects of LSD in comparison to the control. For various drug conditions in the control group, behavioral and neuronal activity results were similar and thus combined unless specified otherwise. Statistical details of the experiments including N, statistical tests, and their descriptive statistics are found in the Results section of the paper. Significance is defined in the figure legends as *p < 0.05; **p < 0.01; ***p < 0.001.
The behavior of an animal during tracking running was quantified for each session by the number of running laps, running speed, the amount of time the animal was immobile, and the number of head twitches (HTs). A running lap was the time period when the animal ran from one end of the track to the other. Instantaneous running speed was calculated for every time point using the position data and smoothed by a Gaussian window with a sigma of 0.5 s. Slow-speed or stopping periods were removed from speed calculations by removing events when the animal's speed was below 10 cm/s for at least e2 Cell Reports 36, 109714, September 14, 2021 Article 0.5 s. Immobility periods throughout the entire session including the ends of the trajectory were used for ripple analysis and percent of immobility in a session. These periods were identified by setting a speed threshold below the mean speed at the reward sites, with the same threshold applied to PRE and POST for each animal. If two neighboring immobility periods (either within or outside the running laps) had a gap smaller than 0.5 s, they were combined into a single immobility period. A head twitch (HT) was visually identified from the recorded videos of the running sessions. Across most animals, two experimenters independently scored and compared counts for reliability. In addition to visual observation, in a subset of sessions (N = 8) under LSD high or LSD low , the start and end times of each HT was precisely determined automatically from deflections in the filtered EMG signal recorded from the rat neck muscle. These EMG-identified HTs were referred to as eHTs. In this case, raw EMG signals were band-pass filtered within 20 -60 Hz to remove the movement-related artifact. eHTs were detected when at least 2 peaks exceeded 8 standard deviations (SDs) from the filtered trace baseline with a minimum inter-peak time of 0.1 s. The eHT start and end times were assigned when the signal first and last exceeded a threshold of 3.5 SDs.
Our analyses on single-unit data were performed on a total of 781 CA1 neurons and 153 VC cells from 17 rats on 20 recording days (Table). Single units were sorted offline using custom software (xclust, M. Wilson at MIT, available at GitHub repository:). The VC cells were mostly located in the primary visual cortex V1 (140) and the rest (13) were located in the neighboring secondary visual cortex. We identified those CA1 and VC cells that were active on at least one trajectory in at least one of the track-running sessions (PRE or POST) with a minimum firing rate of 0.5 Hz (active cells). CA1 silent cells were defined as those with rate < 0.5 Hz and putative interneurons in the CA1 as firing rate > 5 Hz in both trajectories during running. Firing rates of CA1 and VC active cells, CA1 silent cells, and CA1 putative interneurons were analyzed during running and during immobility. For VC cells, we did not analyze those silent (< 0.5 Hz; N = 15) or those with very high rate (> 25 Hz, N = 8) due to their small numbers. Further analyses described below were only performed on active CA1 (N = 365) and active VC (N = 130) cells.
We analyzed the spatial firing properties of active CA1 cells. Spikes occurring at any stopping periods were excluded from the analysis. For each cell on its active trajectory in a session, we constructed a firing rate curve by computing the average firing rate across all laps at each position along the trajectory, excluding the positions within $10 cm from the reward sites. The rate curves had a spatial bin size of 1 cm and were smoothed using a Gaussian window with a sigma of 3 bins. Spatial information (SI) was computed from the firing rate curve according to the established formula as in previous studies. Stability was the Pearson correlation between the two firing rate curves on the same trajectory between PRE and POST. For this calculation, a cell was included if it was active on the trajectory in at least one session (rate > 0.5 Hz) and fired at least one spike in the other session. Directionality was measured by a Pearson's correlation between the two firing rate curves on the two trajectories in the same session. In this case, a cell was included if it was active on one trajectory (rate > 0.5 Hz) and fired at least one spike on the other. For a CA1 cell on an active trajectory in a session, its place fields were identified using a threshold of 3 Hz for peak rates. Boundaries of a place field were determined by 10% of its peak rate. Fields with a gap smaller than 5 cm were combined. The within-field firing rate of a field was the mean rate of the rate curve within the field's boundaries. For comparing firing rates between PRE and POST sessions, we removed the effect of running speed on firing rates. In this case, for each cell active on a trajectory we computed its mean firing rate r i and mean running speed s i during every lap i on the trajectory in both sessions. We performed a linear regression between the lap-by-lap rate and speed: r = ks + b. We then computed a residual rate r i ' = r i -(ks i + b). The residual firing rate for PRE or POST was the average value of r i ' for those laps in PRE or POST. The within-field rates were similarly corrected for the effect of speed. In this case, the lap-by-lap firing rate and speed were computed within the boundaries of each place field. In addition, we computed a burst index for CA1 active cells, which was the percentage of spikes of a cell during running that occurred within 10 ms of another spike.
To compute theta peak frequency and theta power, we divided running laps of a session into different events with different running speeds (< 20, 20 À40, 40-60, > 60 cm/s) with minimum length of 2 s. We then used multi-taper method to estimate the power spectral density (PSD) of CA1 LFPs. Theta peak frequency (at 0.25 Hz resolution) and total theta power were obtained for each speed range of a session. Not all speed ranges were available in a given session, especially for POST under LSD when rats tended to run slower. Coherence was computed between CA1 and VC LFPs during running and then averaged over the theta band (6 -10 Hz). To obtain theta phases of spikes of a CA1 place cell, CA1 LFPs were filtered within [6 10] Hz and theta peaks and troughs during active running were identified. Theta phases were assigned according to spike times relative to their nearest theta peaks (360 /0 ) and troughs (180 ). Theta phase properties for cells active in a session were computed according to circular statistics. Theta phase precession for spikes within a place field was quantified by the maximum linear correlation between phases and positions within the field (O'. Only the fields with a minimum of 20 spikes were included in the analysis. To quantify theta sequences, we selected those place cells with the peak distance of their overlapping place fields < 40 cm and identified their cross-correlation peak time as a measure of firing interval within theta cycles. The correlation between the field distance and the theta firing interval among all overlapping place fields in a session was computed and compared between PRE and POST.
Population vectors (PVs) were constructed for every spatial bin of a trajectory from rate curves of all cells on a day active on the trajectory. Only PVs at those locations with at least 5 active cells were included in the PV correlation analysis. A PV correlation was the Pearson's correlation between the two PVs at the same spatial bin either for two sessions (PRE and POST, cross-session PV corr ) or for two trajectories (cross-trajectory PV corr ). The PV correlations for all the spatial bins and all animals were combined and compared among different experimental conditions.
To quantify pairwise interactions within CA1 cell pairs or between CA1 and VC cell pairs, we computed a normalized cross-correlation of two spike trains, relative to the shuffled spike trains. We first computed a spike-count cross-correlation using the two original spike trains within all events under consideration (e.g., all running laps in PRE, or ripple events in POST). Then, each of the two spike trains was independently shifted by a random time within each event and a cross-correlation was computed using the shuffled spike trains. The shuffling was repeated for 200 times. The mean and standard deviation (SD) were computed at each time bin from all the shufflegenerated cross-correlations. The normalized correlation value at each time lag was the Z-score of the original correlation value relative to the shuffle mean and SD. For within-CA1 cell pairs, we computed their normalized cross-correlations during running laps, with stopping periods within laps removed. The bin size was 10 ms and the cross-correlation was smoothed by nearest-neighbor averaging among ± 10 bins (100 ms). Coactivity between two cells was the average Z scores around time lag 0 ([-100 100] ms). For within-CA1 cell pairs, we also computed their normalized cross-correlations within ripple events. In this case, the bin size was the same 10 ms, the cross-correlation was smoothed using ± 4 bins (40 ms), and coactivity was the average Z-score among the time lagms. For CA1-VC pairs, we computed their normalized cross-correlations and coactivities during running similarly as within-CA1 pairs. But for ripple events, the CA1-VC cross-correlation was performed in time intervals of [-1 1] s from the ripple peak times and coactivity was the average Z-score among the time lag [-100 100] ms, to account for the possibility that VC neurons might activate earlier or later than the CA1 ripples. Awake reactivation was identified by correlating coactivities of all cell pairs during running with those within ripples. In this case, we included those cell pairs that contained at least one cell active during running (rate > 0.5 Hz) and the other fired at least one spike during running.
All spikes in a recording area (CA1 or VC) were included in the multiunit activities. The spikes in a given area were counted in 10 ms time bins. The spike counts were normalized to values within [0 1] with 0 meaning no spikes and 1 meaning maximum spike count. The spike counts were then smoothed by a Gaussian kernel with a sigma of 3 bins into a MUA activity curve with time. A cross-correlation between CA1 and VC MUAs was computed within lags of [-1 1] s for running periods longer than 2 s without immobility. The mean MUA cross-correlation for a session under a condition was the average cross-correlation among all running periods in all animals under the same condition. For analyzing fine temporal relationship between VC MUAs and CA1 ripples, we used ripple peak times (time 0) to trigger VC MUAs within ripple events and the average VC MUAs were computed at time lags withinms (to identify possible peaks within an HVS cycle of $100 ms), relative to baseline periods ofms and [100 120] ms.
We computed the average velocity, linearized 1-dimensional (1D) position, and firing rates of CA1 and VC cells triggered by the EMGidentified head-twitching (eHT) start times. We aligned the start times of all eHTs in POST under LSD high and LSD low as time 0 s. For triggered averages of velocity and 1D-position, we considered every 50 ms bin (time lag) within a time interval of [-1 1] s around the eHT start times. The HT-triggered average velocity at a time lag was the average across all values of velocity at the time for all eHTs. For HT-triggered average 1D-positions, we aligned the 1D positions at time lag 0 s as 0 cm and averaged all 1D-positions relative to this 0 position at every time lag for all eHTs. For HT-triggered average firing rates, we considered a time interval ofs with a bin size of 50 ms. At every time lag, firing rates of all running-active cells in CA1 or VC at the time lag were averaged across all cells and all eHTs. High voltage spike (HVS) and ripple event detection HVS events were detected as described in previous work. Cortical LFP was filtered with a band-pass of 6 -12 Hz. HVS were identified if the filtered LFP trace exceeded a trough threshold set as 6 SD below the baseline trace with at least 4 troughs below this threshold and with a maximum inter-trough interval of 250 ms. The start and end time was assigned at the point when the signal first and last exceeded a threshold of 2.5 SD. The peak time was determined as the time with the lowest trough amplitude within the HVS event. Neighboring HVS events with a gap between them less than 0.5 s were combined into a single event. Detected events were visually examined to ensure reliable and consistent identification, and parameters may have been slightly adjusted in rare cases in which HVS events were mis-identified. e4 Cell Reports 36, 109714, September 14, 2021
We detected ripple events from a band-pass filtered (100 -250 Hz) hippocampal LFPs. Ripple events were identified by a trough threshold exceeding 6 SD from baseline. The start and end times of ripples were identified as the moment the amplitude of the filtered LFP trace crossed 2.5 SD and only ripple events between 30 and 400 ms in duration were included in the analysis. If neighboring ripple events were separated by fewer than 30 ms, they were combined into a single event. In addition, we also computed the cross-correlation between ripple-filtered (100-250 Hz) CA1 LFPs and HVS-filtered (6 -12 Hz) VC LFPs as a measure of possible fine temporal relationship between CA1 and VC activities during HVS events. -4 ). There were no obvious differences in burst index among the groups (P = 0.063). This result is consistent with the idea that LSD broadly reduced firing rates of CA1 cells such that the cells with low rates in PRE were more likely to go silent in POST. (B) Same as (A), but for two groups of PRE-active cells: those becoming more bidirectional (Bi-dir: directional correlation POST-PRE > 0.2; N = 24) and other cells (Others: directional correlation POST-PRE < 0.2; N = 69). There were no differences in either firing rate (P = 0.50, Mann-Whitney test) or burst index (P = 0.83) between the two groups. The cells with increased bi-directionality in POST under LSD did not appear to form a special group. (C) Tetrode recording sites (red arrows) in the CA1 of example animals under LSDhigh, LSDlow and the control (Ctrl) condition. The tetrodes were implanted to the same targets in all animals under all conditions. Scale bar: 0.6 mm. The correlation here is used to quantify theta sequences of CA1 place cells with theta cycles. The number of pairs was relatively small, due to the requirement of overlapping place fields on the same trajectories. As a result, the regression was not statistically significant under Ctrl, however, its trend and the correlation value were comparable to LSD. Importantly, the correlation values under LSD in both PRE and POST were statistically significant and they were not different from each other (P = 0.38, Fisher's exact test). Overall, theta oscillations appeared to be modestly impacted, but no obvious differences were found in theta phase tuning, phase precession or theta sequences for those cells that remained active during running in POST under LSD. The PRE-POST correlations in LSDhigh (R = 0.29) and LSDlow (R = 0.36) were significantly lower than that of Ctrl (R = 0.51; P = 0.010, 0.048 to LSDhigh, LSDlow respectively, Fisher's exact test). Bottom: cumulative distributions of firing rates under all 3 conditions in PRE and POST for the downsampled CA1 and VC cells. There were no significant differences in firing rate across the conditions in PRE/POST for CA1/VC cells (P > 0.14, Kruskal-Wallis test). (C) Coactivities during running and within HVS events for CA1 -VC cell pairs. The coactivity was computed within HVS events in POST under LSD high , since total HVS event duration was short in PRE and under other conditions. Each dot is a pair. Note the non-significant correlation between RUN and HVS coactivities, suggesting the lack of awake reactivation across CA1 and VC within HVS events.
Create a free account to open full-text PDFs.
Liechti, M. E. · Neuropsychopharmacology (2017)
Müller, F., Dolder, P. C., Schmidt, A. et al. · NeuroImage (2018)
Palhano-Fontes, F., Andrade, K. C., Tófoli, L.F. et al. · PLOS ONE (2015)
Preller, K. H., Razi, A., Zeidman, P. et al. · PNAS (2019)
Roseman, L., Leech, R., Feilding, A. et al. · Frontiers in Human Neuroscience (2014)
Schmid, Y., Enzler, F., Gasser, P. et al. · Biological Psychiatry (2015)
Schmidt, A., Müller, F., Lenz, C. et al. · Psychological Medicine (2017)
Tagliazucchi, E., Roseman, L., Kaelen, M. et al. · Current Biology (2016)
Papers in Blossom that reference this study
Ekins, T. G., Brooks, I., Kailasa, S. et al. · Biorxiv (2023)
Breant, B., Mengual, J. P., Bannerman, D. et al. · Biorxiv (2022)
Kwan, A. C., Olson, D. E., Preller, K. H. et al. · Nature Medicine (2022)