The classic psychedelic DOI induces a persistent desynchronized state in medial prefrontal cortex

This pre-print (2023) rodent study found that DOI caused changes in brain activity (specifically the mPFC) that differed from normal patterns. During rest, when brain activity is usually synchronized, the drug causes a decrease in synchronization and an increase in gamma activity.

Authors

Abbas, A.
Bartlett, L.
Bretton-Granatoor, Z.

Published

February 27, 2023

Biorxiv

individual Study

Links

Read Paper DOI Google Scholar

Abstract

Administration or consumption of classic psychedelics (CPs) leads to profound changes in experience which are often described as highly novel and meaningful. They have shown substantial promise in treating depressive symptoms and may be therapeutic in other situations. Although research suggests that the therapeutic response is correlated with the intensity of the experience, the neural circuit basis for the alterations in experience caused by CPs requires further study. The medial prefrontal cortex (mPFC), where CPs have been shown to induce rapid, 5-HT2A receptor-dependent structural and neurophysiological changes, is believed to be a key site of action. To investigate the acute neural circuit changes induced by CPs, we recorded single neurons and local field potentials in the mPFC of freely behaving mice after administration of the 5-HT2A/2C receptor-selective CP, 2,5-Dimethoxy-4-iodoamphetamine (DOI). We segregated recordings into active and rest periods in order to examine cortical activity during desynchronized (active) and synchronized (rest) states. We found that DOI induced a robust decrease in low frequency power and decoupled rhythmic activity from neural population dynamics when animals were at rest, attenuating the usual synchronization that occurs during less active behavioral states. DOI also increased broadband gamma power and suppressed activity in fast-spiking neurons in both active and rest periods. Together, these results show that the CP DOI induces persistent desynchronization in mPFC, including during rest when mPFC typically exhibits more synchronized activity. This shift in cortical dynamics may in part underlie the longer-lasting effects of CPs on plasticity, and may be critical to their therapeutic properties.

Unlocked with Blossom Pro

Research Summary of 'The classic psychedelic DOI induces a persistent desynchronized state in medial prefrontal cortex'

Introduction

Classic psychedelics (CPs) produce profound alterations in subjective experience and have re-emerged as candidate therapeutics for treatment-resistant depression and other psychiatric conditions. Their principal psychoactive action is mediated by Gq-coupled 5-HT2A receptors (5-HT2ARs), which are highly expressed in medial prefrontal cortex (mPFC). Previous work links CPs to rapid 5-HT2AR-dependent structural and functional plasticity, but the acute circuit-level neurophysiological changes that precede plasticity are less well characterised, particularly in vivo during freely behaving states. Olson and colleagues set out to characterise how a prototypical 5-HT2A/2C-selective CP, 2,5-dimethoxy-4-iodoamphetamine (DOI), alters mPFC network dynamics in awake, freely moving mice. They tested the hypothesis that DOI produces behavioural state-dependent shifts in cortical synchrony by recording single-unit spiking and local field potentials (LFPs) from ventral mPFC before and after systemic DOI, and comparing activity during segregated active and rest periods.

Methods

Subjects were eight adult C57BL/6J mice (age 12–24 weeks). A subset of four animals received saline injections one week after DOI and were used for saline comparisons. Surgical implantation placed a 28-microwire bundle (14 stereotrodes of 13 µm tungsten wire) into left ventral mPFC; a ground screw was placed over the cerebellum and a reference screw over orbitofrontal cortex. Animals recovered for one week before recording. Behavioural sessions consisted of a 15–30 minute baseline in a novel open field followed by intraperitoneal injection of racemic DOI at 5 mg/kg (or saline in the subset) and a further 60 minutes of free behaviour. Video (30 fps) was synchronised to neural data and processed with DeepLabCut to extract head/body/tail positions and derive velocity. A single velocity threshold (visually determined across animals) was used to binarise active versus rest periods; short micro-states were collapsed using sequential 5 second windows to produce stable behavioural epochs. Electrophysiology employed a Digital Lynx system with LFPs bandpass filtered 1–1,000 Hz and sampled at 2 kHz, spikes filtered 600–6,000 Hz and sampled at 32 kHz. Single units were clustered with KlustaKwik and curated by waveform, inter-spike-interval distribution, isolation distance and L-ratio. Neurons with baseline firing <0.5 Hz were excluded. Neurons were classified as high (>5 Hz) or low (<5 Hz) baseline firing rate groups. LFP analysis compared a baseline window (-15 to -5 min pre-injection) with an experimental window (+25 to +60 min post-injection). Data for each window were concatenated separately for active and rest periods producing four conditions: baseline active, baseline rest, experimental active, experimental rest. Continuous wavelet transform (Morse wavelet) calculated power from 1–120 Hz, and power was averaged into canonical bands: delta (2–4 Hz), theta (6–10 Hz), beta, low gamma (30–50 Hz) and high gamma (reported in text as 50–90 Hz). Power per animal was normalised and group comparisons used Wilcoxon signed-rank tests. Single-unit analyses included firing rates binned in 1 ms and summarised per the same four conditions. To classify neurons as increasers/decreasers after drug, a bootstrapping shuffle procedure generated 1,000 surrogate spike trains; observed changes outside the 97.5% bounds were labelled significant. Absolute Z-scored firing dynamics were computed in 60 s bins relative to the 10 min baseline. Spike entropy used the inter-spike-interval distribution (50 bins, 5 ms width) to compute a normalised Shannon entropy. To probe the relationship between LFP and population spiking, the authors performed a latent dynamical analysis: smoothed firing rates were reduced by PCA (first 10 PCs), yielding a low-dimensional latent manifold which was correlated with band-specific LFP power using canonical correlation analysis (CCA). Statistical comparisons of canonical correlations again used Wilcoxon signed-rank tests. The extracted text does not report exact p-values or effect sizes for most comparisons.

Results

DOI altered mPFC oscillatory power in a behavioural state-dependent manner. Alignment of velocity to LFP showed the expected baseline pattern of elevated delta (2–4 Hz) during rest and higher theta (6–10 Hz) during activity. Approximately 5 minutes after DOI injection (5 mg/kg i.p.) the normal rest-associated increase in delta power was significantly reduced relative to baseline; this decrease was not observed after saline. DOI also reduced the usual active-state increase in theta power, whereas beta power showed no effect. In contrast, DOI produced a significant increase in high gamma power (50–90 Hz) that occurred during both rest and active periods; no gamma increase followed saline. Single-unit activity showed preferential modulation of high firing rate neurons. Average firing rate did not change systematically with behavioural state after DOI or saline. When neurons were split by baseline firing (>5 Hz versus <5 Hz) and sorted by change magnitude, the mean firing rate of high-firing neurons was significantly lower in the experimental period after DOI compared to baseline, while low-firing neurons showed no significant mean change. A larger proportion of high-firing neurons decreased their firing after DOI compared with low-firing neurons. The absolute magnitude of DOI-induced firing changes (after Z-scoring) was greater in high-firing neurons. Low firing neurons nonetheless showed state-dependent effects, with a higher proportion increasing firing during active periods; because low-firing cells displayed mixed directions of change their population average did not shift markedly. DOI disrupted the coupling between LFP band power and population spiking dynamics. Using latent dynamical CCA, the authors found that DOI reduced the canonical correlation between the low-dimensional spiking manifold and LFP power across all frequency bands during rest. During active periods the disruption was selective to the theta band. Saline produced no change in canonical correlations. In addition, entropy analyses of inter-spike intervals revealed that low firing rate neurons exhibited a significant decrease in spike-interval entropy after DOI compared to baseline, while high firing neurons trended toward increased entropy; saline had no effect. These results together indicate a DOI-induced decoupling of population spiking from dendritic-input-related LFPs, especially during rest. The extracted text does not provide detailed numerical statistics (e.g. exact p-values or confidence intervals) for every reported comparison.

Discussion

Olson and colleagues interpret their findings as showing that systemic DOI produces persistent desynchronization in ventral mPFC, including during rest when cortex is normally more synchronized. They highlight four principal circuit-level changes observed after DOI: attenuation of rest-related delta synchrony, a broadband increase in high gamma (50–90 Hz) power irrespective of behavioural state, a reduction in firing rate among fast-spiking putative parvalbumin interneurons, and a decoupling of spiking from LFP during rest. The authors situate these results in the context of previous work: decreased delta power during rest aligns with studies using DOI in anaesthetised rats and with human EEG studies of DMT and psilocybin. The observed gamma increase agrees with some mouse studies but contrasts with reports of gamma decreases in anterior cingulate cortex; the investigators note that regional differences within mPFC and the use of anaesthesia in other studies may account for apparent discrepancies. They also note that earlier ex vivo work showed strong 5-HT2AR effects on fast-spiking interneurons and shifts toward more asynchronous synaptic transmission, which dovetails with the present in vivo observation of reduced fast-spiking activity and increased broadband gamma. Mechanistically, the authors propose that DOI-induced suppression of fast-spiking interneurons could disorganise gamma-related coordination and prevent the normal transition to low-frequency synchrony during rest, producing an ‘‘aberrant desynchronised’’ cortical state that persists into periods typically dominated by synchrony. Such a persistent desynchronised state might promote synapse-specific long-term plasticity, and the investigators link this to prior reports of 5-HT2AR-dependent structural plasticity (dendritic spine growth and increased synaptogenesis) that persist after single CP exposures. They further suggest that elevated broadband gamma in mPFC could be an early translational neurophysiological correlate of antidepressant efficacy, drawing parallels with ketamine and some CP studies. The authors acknowledge complexity and remaining uncertainties. They emphasise the need for further in vivo work in awake animals to map effects onto specific behaviours and to determine which neuronal subpopulations and long-range inputs drive the observed changes. Differences between brain subregions, anaesthetic confounds in prior studies, and the heterogeneous responses among low-firing neurons are cited as factors complicating interpretation. Overall, the investigators conclude that rapid DOI-induced circuit changes in mPFC bias cortical dynamics toward persistent desynchronization, a state that may underlie the subsequent plasticity and experiential effects attributed to classic psychedelics.

View full paper sections

INTRODUCTION

Classic psychedelics (CPs) induce profound and multifaceted effects on experience after consumption. The resulting experience is often described as "mystical", and for many is rated as highly meaningful. In recent years, there has been rekindled interest in the potential of CPs, particularly psilocybin, as possible therapeutic agents for treatment-resistant depression and other psychiatric disorders. This work has added to longstanding evidence that CPs can chronically alleviate depressive symptomatology in a dose-dependent manner mediated by the intensity of the experience. Classic psychedelics exert their main psychoactive effects through interaction with Gq-coupled 5-HT2A receptors (5-HT2ARs), and pretreatment with the selective 5-HT2AR antagonist ketanserin greatly reduces the subjective effects of the CPs psilocybin and LSD. Expression of these receptors is enriched in medial prefrontal cortex (mPFC) where roughly half of all excitatory neurons and 20-30% of all GABAergic interneurons express 5-HT2ARs. Their activation by both CPs and non-psychedelic 5-HT2AR agonists induces a diverse range of downstream molecular actions that do not clearly differentiate CPs from nonpsychedelic 5-HT2AR agonists. The most consistent long-term effect across compounds seems to be a potent initiation of structural and functional dendritic plasticity, which is similarly blocked by application of ketanserin. However, the rapid, circuit-level neurophysiological effects which precede this plasticity are less well understood. Thus, investigating how CPs affect neural circuit function might help determine how they acutely modulate experience and aid in identifying key neural activity which confers superior long-lasting therapeutic properties. Previous ex vivo electrophysiology studies in the mPFC have identified complex acute effects of 5-HT2AR activation on the intrinsic excitability and synaptic activity of different neuron subtypes. Whole-cell recordings from layer 5 GABAergic neurons have shown that 5-HT2ARs modulate activity most strongly in fastspiking interneurons, the main inhibitory subtype on which they are expressed. More mixed effects were found in pyramidal neurons of rat mPFC, in which 5-HT2ARs were found to either hyperpolarizeor depolarizemembrane potential. Synaptic activity impinging onto mPFC pyramidal neurons is also modified by 5-HT2ARs, which increase the frequency and amplitude of spontaneous and evoked inhibitoryand excitatorypostsynaptic currents. Furthermore, 5-HT2AR activation has been shown to bias these synapses toward a more asynchronous mode of evoked neurotransmitter release, which can alter the timing of communication between neurons. Taken together, these acute ex vivo effects of 5-HT2AR activation would be expected to alter local field potentials (LFPs), spiking, and synchrony in mPFC circuits in vivo. However, to our knowledge only a few studies have investigated these phenomena in freely behaving rodents. In the dorsal anterior cingulate subregion of the mPFC, systemic administration of the 5-HT2A/2C-selective CP DOI altered the firing rate of a subset of neurons and decreased gamma power. More ventrally, in the prelimbic and infralimbic cortices, CPs produced the opposite effect on gamma power, implying subregion-specific mechanisms of CPs in the mPFC. This effect was reported to be significantly modulated by the behavioral state of the animal, which suggests a more complex interaction between CPs and cortical activity than previously thought. Studying these behavioral state-dependent effects of CPs on neural network activity is especially intriguing in more ventral mPFC, which is proposed to be analogous to the human subgenual and pregenual ACC, two brain regions known to be dysregulated in major depressive disorder. Importantly though, neither of the studies in ventral mPFC mentioned above recorded spiking and LFP simultaneously in response to CP administration. Learning how CPs regulate the behavioral state-dependent relationship between spiking and LFP in this region would provide a more complete picture of the complex network effects that may contribute to their therapeutic properties. To that end, we recorded single units and LFPs in the ventral mPFC of freely behaving mice before and after administration of DOI, with the goal of studying its effect on spiking, brain rhythms, and population activity patterns. We predicted that DOI would lead to a behavioral state-dependent shift in synchronized population dynamics and associated brain rhythms. To assess this, we segregated recordings by behavioral stateactive versus restand compared pre-and post-DOI mPFC rhythmic power, single neuron firing, and measures of population dynamics. We found that activity in fast-spiking neurons, LFP at multiple frequencies, and the synchrony between neural population dynamics and LFP across a range of frequency bands were all disrupted after administration of DOI in a manner that depended on behavioral state.

THE CP DOI MODULATES LFP POWER IN A BEHAVIORAL STATE-DEPENDENT MANNER

To examine the interaction between behavioral state and the effect of the CP DOI on cortical activity, we implanted mice with 14 tungsten stereotrodes in the left mPFC. Electrodes were connected to an electronic interface board that facilitated the acquisition of synchronized electrical and video recordings of mPFC neural activity using a Neuralynx acquisition system (see Methods for further details). Mice were placed in a box and allowed to freely explore for a 15-30 minute baseline period, after which they were injected intraperitoneally (i.p.) with 5 mg/kg of DOI or saline. Mice continued free exploration for another 60 min as recordings continued. We used DeepLabCutto calculate the position and velocity of the animal throughout the recording session to segregate the animal's behavioral state into active and rest periods (see Methods for details), as LFP is known to be modulated by behavioral state. Alignment of velocity traces to wavelet power spectra of mPFC LFP recordings during baseline (Figure) revealed the typical relationship of higher delta (2-4 Hz) and lower theta (6-10 Hz) power during rest states and subsequent reversal of this relationship during active states (Figure). We did not see state-dependent modulation of gamma power (Figure); however, this is not surprising given that our behavioral state classification segregates between lighter rest periods (very low velocity but still interacting with the environment) and active periods. Deeper rest and sleep are associated with decreased gamma power, but we rarely captured deep rest periods as the mice were mostly engaged during recording. After classifying behavioral state-dependent dynamics, we examined the effect the CP DOI had on these dynamics. DOI-treated animals revealed that the typical increase in low frequency delta (2-4 Hz) power which occurs during rest periods, a hallmark of cortical synchronization, is significantly decreased approximately 5 min after DOI injection, but not after saline injection (Figure). Averaging normalized low frequency power spectra across animals showed that delta power was significantly decreased after DOI injection (Figure) but not saline injection. We also found the typical increase in theta power during active states was decreased only after DOI but not saline injection (Supplementary Figure). Neither DOI nor saline affected beta power (Supplementary Figure). Alignment of velocity traces with higher frequency spectrograms revealed that systemic injection of DOI led to a significant increase in gamma power (50-90 Hz) in the mPFC (Figure). The DOI-induced increase in gamma power occurred during both rest and active periods (Figure). No increase in gamma power was seen after saline injection (Figure). Taken together, the CP DOI disrupts the typical behavioral statedependent modulation of oscillatory activity in mPFC.

DOI PREFERENTIALLY MODULATES HIGH FIRING RATE NEURONS

To investigate effects of DOI on the activity of individual neurons in mPFC, we recorded single units before and after injection of DOI or saline and separated neurons into high (>5 Hz; enriched for fast-spiking parvalbumin inhibitory neurons) and low (<5 Hz; enriched for pyramidal neurons) firing rate groups in accordance with their baseline activity. We initially compared average firing rates of these groups during rest and active periods but found that average firing rate did not vary as a function of behavioral state after DOI or saline (Supplemental Figuresand). We then sorted them within each group by the magnitude of the change in firing rate after DOI or saline and generated heat-maps of sorted neuron firing rates over the course of the experiment (Figure). The mean firing rate of high firing rate neurons was significantly lower after DOI compared to the baseline period, while there was no significant change in mean firing rate seen with low firing rate neurons (Figure). Breaking the groups down further, we noticed that a larger proportion of high firing rate neurons decreased after DOI as compared to low firing rate neurons (Figure). Interestingly, the relative changes in firing rate for low firing neurons were dependent on behavioral state, showing a significantly higher proportion of cells increasing their firing rate during active periods, suggesting a complex relationship between circuit dynamics and behavioral state (Figure). Accordingly, the magnitude of the change effected by DOI on each neuron, irrespective of direction of change, was significantly larger for high firing rate neurons even after rescaling by Z-scoring (Figure). It is important to note that low firing rate neurons also significantly changed their overall magnitude of firing, however, because the population of low firing neurons is heterogeneous as to direction of firing rate changes, the average firing rate does not change. Overall, DOI has a predominantly inhibitory effect that is larger in terms of the proportion of neurons inhibited and the magnitude of inhibition in high firing rate neurons relative to low firing rate neurons. Furthermore, that effect of DOI on single neuron spiking is modulated by behavioral state, albeit in a complex way.

DOI DISRUPTS THE CHARACTERISTIC RELATIONSHIP BETWEEN LFP AND SPIKING ACTIVITY DURING REST PERIODS

LFPs, which are predominantly a reflection of dendritic input, exhibit a characteristic relationship with spiking which depends on a brain region's afferent and local connectivity. If DOI is indeed disrupting rest-related synchronization within cortex, we predicted that we would see a disruption of that characteristic relationship. To examine this possibility we performed latent dynamical canonical correlation analysis (CCA). Latent dynamical CCA analysis takes advantage of ongoing neural data and is well-suited for examining the relationship between LFP power and spiking over longer periods of time, unlike phase locking analyses which are better suited for event-centered, trialized data during stereotyped behaviors. During latent dynamical CCA, the dimensionality of population spiking is reduced and a correlation between that reduced signal and bandpass-filtered LFP is calculated to quantify the relationship between the latent dynamics of population spiking and LFP power. Figuresshow an example LFP and the first dimension of the projected population spiking activity before and after DOI for both rest and active periods. In the example, the signals appear less correlated after DOI only during rest. Latent dynamical CCA before and after DOI or saline during both rest and active states revealed that DOI disrupted the correlation between spiking and all frequency bands during rest, and only the theta frequency band during active periods (Figure), as would be expected if DOI induced an active-like desynchronization that persisted into the rest period. In contrast, saline did not affect the canonical correlations between band-specific power and population activity (Supplementary Figure). We also wanted to investigate if spiking patterns changed as a result of decoupling from the LFP during rest that occurs after DOI. To asses this, we calculated the entropy of each neuron's distribution of inter-spikeintervals. We found that low firing rate neurons exhibited a significant decrease in entropy after DOI compared to baseline, whereas high firing rate neurons were unchanged, albeit trending toward increased entropy (Supplemental Figure). Saline had no effect on spike entropy (Supplementary Figure). Taken together, these results suggest a fundamental change in network dynamics after the CP DOI.

DISCUSSION

In this study, we recorded mPFC single units and LFP before and after administering the psychedelic DOI during freely moving behavior, which was segregated into rest and active periods. We saw four key circuitlevel changes in mPFC after DOI. First, the usual rest-related synchronization that typically occurs in rodents, which is associated with an increase in delta power, was attenuated after DOI. Second, we observed an increase in high gamma power (50-90 Hz) irrespective of whether the mouse was active or at rest. Third, there was a significant decrease in the firing rate of fast-spiking, putative parvalbumin inhibitory neurons which also occurs regardless of behavioral state. Fourth, we see a decoupling of mPFC spiking from the LFP during rest. Overall, our findings suggest that DOI induces aberrant desynchronization that persists into rest periods, when brain activity is typically more synchronized. Our results showing decreased delta power in mPFC during rest agree with other studies using the CP DOI in anesthetized rats, and with human EEG studies using DMTand psilocybin. DOI also induced a broadband increase in gamma power, another effect reported elsewhere in mice. This is in contrast to Wood et al., who observed a decrease in gamma power. However, their recordings were done in the anterior cingulate cortex, which is known to be functionally and anatomically different from more ventral mPFC. A more recent study in the anterior cingulate cortex, which performed experiments in head fixed mice, similarly found reduced cortical synchrony but a general increase in population firing. Furthermore, our data showing decreased fast-spiking activity after DOI is inconsistent with previously published results showing the opposite. The reason for this discrepancy may be due to the chloral hydrate used to anesthetize animals for recording, which is known to directly affect the serotonergic system. Different behavioral states are canonically characterized by oscillatory activity in distinct frequency bands. In rodent and monkey cortex, rest and quiet wakefulness are normally associated with synchronized, low-frequency oscillations, which transitions to higher-frequency, desynchronized activity during more active engagement with the environment. Fast-spiking inhibitory neurons are known to organize the higher-frequency gamma rhythms, and blocking the firing of these neurons desynchronizes cortical networks and increases broadband gamma power. Therefore, it is plausible that the aberrant desynchronized activity that we see during rest periods is partly due to DOI disrupting the dynamics of fastspiking neurons, which prevents the normal transition to more synchronized brain states during rest/quiet wakefulness. More active cortical states with decreased delta power and increased gamma power normally reflect a shift toward more local computations. These enhanced gamma power states are theorized to be optimal for inducing synapse-specific long-term plasticity, the precision of which relies on feed-forward inhibition involving parvalbumin positive, fast-spiking interneurons. Since DOI seems to be modulating these neurons, the observed desynchronized rest state could lead to long-lasting changes in associative plasticity, which may underlie some of the experiential or therapeutic effects of CPs. This desynchronized state induced by DOI is consistent with the theory that psychedelics effectively enable neural networks to escape their strongest previously entrained patterns of activity, leading to a more labile and dynamic state. As a precedent for this idea, several studies have shown long-lasting 5-HT2AR-dependent plasticity following CP administration. An in vitro study showed that when exposed to CPs, rat cortical neurons displayed an increase in dendritic arbor complexity, dendritic spine growth, and new functional synapse formation. In vivo, a single dose of psilocybin resulted in dendritic growth and synaptic activity in layer 5 apical dendrites of the mouse mPFC. Both of these findings persisted for at least 24 hours after drug removal and were dependent on 5-HT2ARs. Slice studies report a more rapid effect of 5-HT2AR activation in which DOI strengthens NMDA currents but weakens AMPA currents within minutes, while gating the induction of spiketiming dependent depression in mPFC at thalamocortical synapses. These long-lasting neuroplastic changes induced by CPs may be caused at least in part by the rapid circuit changes occurring after CP administration reported in this article. It is apparent that the circuit level effects of DOI are complex, and our results highlight the need for future in vivo work on this subject in awake, freely behaving animals. Studying the effect of DOI during behaviors that depend on circuits and cell subpopulations expressing 5-HT2A receptors will be particularly valuable. The effect of DOI on high broadband gamma is especially intriguing for future study, as the same phenomenon is seen in humans with ketamineand other CPs like psilocybin. This could be related to the known effectiveness of ketamine and several CPs for alleviating treatment-resistant depression, which supports the idea that elevated broadband gamma power in the mPFC may be an early translational neurophysiological correlate of antidepressant efficacy. Follow up work will require more detailed experiments to definitively determine which neuronal subpopulations are most affected by DOI and explore whether these effects are being driven predominantly locally or due to effects on one or more long range inputs. Overall, our results suggest that classic psychedelics like DOI induce rapid circuit level changes in the mPFC. These circuit changes fundamentally alter the typical dynamics associated with different behavioral states, favoring persistent cortical desynchronization. This shift toward desynchronization may be a correlate of the experiential effects of psychedelics and of a more labile neural state which drives subsequent neuroplastic changes observed after psychedelic administration.

ANIMALS

All procedures were conducted in accordance with the US NIH Guide for the Care and Use of Laboratory Animals and approved by the New York State Psychiatric Institute Institutional Animal Care and Use Committee at Columbia University. Eight adult (age 12-24 weeks) C57BL/6J (Jackson Labs, stock number 000664) experimental mice were used for DOI experiments, while a subset of these mice (n = 4) received saline injections one week following DOI administration and were used for all saline analyses.

SURGICAL PROCEDURES

Mice were anesthetized with 1%-3% vaporized isoflurane in oxygen (1 L/min) and placed in a stereotaxic apparatus. A craniotomy was made to allow for implantation of a 28 microwire bundle (14 stereotrodes; 13 micron tungsten wire, California Fine Wire) implanted in left mPFC (-0.35 ML, +1.85 AP, 1.3 below brain surface). A ground screw was placed over the cerebellum and a reference screw was placed over the orbitofrontal cortex. Electrodes were connected to a 32-channel Omnetics electrode interface board using gold pins (Neuralynx). Electrode placements were confirmed using an electrolytic lesion (5 mA, 10 s). Mice were allowed to recover for one week post-surgery before behavioral testing and were monitored closely during recovery.

BEHAVIOR

Animals were recorded in a novel open field environment and were allowed to move freely. After 15 minutes of baseline recording mice received a 5 mg/kg i.p. injection of racemic 2,5-dimethoxy-4-iodoamphetamine (DOI) and then placed back in the open field environment to freely behave for 60 min. A subset of animals (n=4) were exposed to the same experiment as described above except that saline was injected.

NEURAL RECORDING

A Digital Lynx system (Neuralynx, Bozeman, MT) was used to amplify, band-pass filter (1-1000 Hz for LFPs and 600-6000 Hz for spikes), and digitize the electrode recordings. LFP sampling rates were 2 kHz and spikes were collected at 32 kHz. Single units were clustered based on the first two principal components (peak and energy) from each channel using Klustakwik (Ken Harris) and visualized in SpikeSorter3D (Neuralynx). Clusters were then visually inspected and included or eliminated based on waveform appearance, inter-spike interval distribution, isolation distance, and L-ratio.

SEGREGATION BY BEHAVIORAL STATE

All data analyses was done using custom scripts in MATLAB (Mathworks) unless otherwise identified. Video was recorded at 30 fps and synchronized to neural data (Digital Lynx) .Videos were then imported into DeepLabCut (DLC,), and positional markers were manually set for head, body and tail for 200 randomly selected frames throughout the entire recording. DLC was then run and average position data from the head, body and tail was calculated per frame for each animal. Due to the imperfect nature of DLC and errors in position, data was cleaned by removing any points that were outside of the physical space the animal could move, and those values were filled by the previous value. Velocity outliers were further removed via a hampel identifier and smoothed via a moving window of 5 fps. Each velocity trace was then synchronized to the existing video and manually inspected. Upon further inspection, large outliers that were not identified by the hampel identifier were removed using a threshold value of 120-fold greater than the threshold value for active/rest. The active/rest threshold velocity value was identified base on visual inspection of velocity trace and video agreement across all animals, and one threshold for all animals was assigned. This threshold was then applied to the velocity trace resulting in a binary vector, 0 denoting rest and 1 denoting active. This vector still resulted in 0 or 1 periods that only last a few frames. These micro rest/active states were too small to perform further analyses on. As a result, sequential 5 second windows were used to identify the consensus of 0s or 1s in each window, changing to either all zeros or all ones for a given window.

LOCAL FIELD POTENTIAL ANALYSES

Power: LFP data was segregated for the baseline window (-15 min to -5 min prior to DOI or saline injection) and for the experimental window (DOI or saline, +25 min to +60 min after injection). For both the baseline window and the experimental window, bins corresponding to the "active" and "rest" periods were segregated and respectively concatenated into 4 conditions, (1) baseline active, (2) baseline rest, (3) experimental active, and (4) experimental rest, where experimental refers to DOI or saline. Each condition was normalized by the root mean square of the whole signal, and the analytic signal was calculated via the continuous 1-D wavelet transform (CWT using Morse wavelet) with frequency limits 1-120 Hz, symmetry parameter gamma (γ) = 3, time-bandwidth product equal to 60, and 10 voices per octave. CWT uses L1 normalization. We calculated the power of the LFP by taking the square of the absolute value of the analytic signal. The power was then segregated into canonical frequency bands: delta (2-4 Hz), theta (6-10 Hz), beta, low gamma (30-50 Hz) and high gamma . For spectrograms in Figures 1, 2 and 3 the entire non-segregated signal was used with which power was calculated as described above, with the smoothed non-binarized velocity trace. For quantification of LFP each animal's power was averaged across conditions and then compared using Wilcoxon signed-rank tests.

SINGLE UNIT ANALYSES

Firing Rate: Spike times sampled at 32 kHz were imported into MATLAB, each neuron having an associated spike train with spike times given to the nearest nanosecond. Spikes were binned in 1 ms bins, and we calculated the baseline firing rate in Hz using the same LFP baseline window. Neurons with baseline firing rates below 0.5 Hz were omitted from further analysis. The remaining neurons were classified as either "high" or "low" firing rate neurons based on their average baseline firing rates, with 5 Hz being the cutoff between the two groups. Average firing rate for each of the same four conditions used in the LFP (1) baseline active, (2) baseline rest, (3) experimental active, and (4) experimental rest was calculated, where experimental refers to DOI or saline. Wilcoxon signed-rank tests were used to compare firing rates of neurons across conditions. Firing Rate Changes: A wide range of responses to drug administration were observed in neuronal activity in the form of firing rate changes. To categorize neurons as increasing their firing rate relative to baseline or decreasing their firing rate relative to baseline, referred to as "increasers" or "decreasers," spike trains were bootstrapped to ensure the observed change in activity was statistically significant. Bins from the baseline and experimental period were concatenated into a single spike train, which was used to generate 1,000 shuffled spike trains by randomly selecting a number of bins equivalent to either the length of the baseline spike train or the length of the experimental spike train to create a shuffled baseline or experimental spike train, respectively. The difference in these spike trains was calculated for each pass, and the observed difference was compared the difference of the collection of shuffled spike trains. If the observed difference was greater or less than 97.5% of the shuffled differences, the cell was labeled as an increaser or decreaser, respectively. Neurons not significantly different from the shuffled spike train were labeled as no change. To test proportions of increasers and decreasers across conditions we bootstrapped the proportions 10,000 times. Absolute Z-Scored Firing Rate: To assess the firing rate dynamics in experimental conditions and examine firing rate changes from baseline, spikes were binned into 60 second bins. Firing rates in each bin for each neuron were Z-scored relative to the neuron's mean firing rate across the 10 minute baseline period. Z-scored firing rates were averaged separately across the population of low-and high-firing rate neurons, respectively, and then the absolute value was taken. Average absolute value Z-scored firing rates after DOI were then compared against firing rates after saline for both populations of neurons using Wilcoxon signed-rank tests. Spike Entropy: Spike times sampled at 32 kHz were imported into MATLAB, each neuron having an associated spike train with spike times given to the nearest nanosecond, and then to spike times in seconds. The inter-spike-interval (ISI) distribution for each neuron was calculated using MATLAB's built in histogram function with 50 bins at a width of 5 milliseconds. A probability mass function (PMF) was then calculated by dividing the count in each bin by the sum of all counts, and probabilities of zero were excluded. The normalized Shannon entropy, h, was calculated from the PMF. Wilcoxon signed-rank tests were used to compare entropy of neurons across conditions.

LOCAL FIELD POTENTIAL AND POPULATION ANALYSIS

Latent Dynamics: To analyze how local field potentials influence ongoing mPFC spiking population dynamics, we examined the correlation between latent population activity and LFP. For a more in-depth review of this analysis see Gallego-Carracedo et al. (2022). First LFP power in the mPFC was calculated as stated in our LFP methods above. Spikes and LFP power were binned in 30 millisecond bins and smoothed with a Gaussian kernel with a width of 50 milliseconds. The firing rate was square-root transformed. This yields an n by T matrix, where n = number of neurons, and T time points. A low dimensional manifold of this matrix was made by performing PCA and taking the first 10 PCs, yielding a matrix X. The smoothed firing rates of n neurons were then projected onto the low dimensional manifold matrix X created with the PCA, yielding an m by T latent manifold L, where m is the dimension of the manifold. Finally, canonical correlation analysis was performed between the latent manifold L and the LFP power at a given frequency band, yielding correlation values from 0 to 1. This analysis was performed for each condition (1) baseline active, (2) baseline rest,experimental active, and (4) experimental rest, where experimental refers to DOI or saline. Wilcoxon signedrank tests were used to compare the canonical correlations that resulted from the latent dynamical CCA across conditions for each frequency band.