Male Wistar rats (n = 115, 270–300 g, Charles River, Germany), were housed in cages with sawdust bedding and environmental enrichment (plastic shelter, gnawing blocks and paper strips) with food and water ad-libitum. Temperature and humidity were controlled and a 12:12 h reversed cycle (lights off at 6:00 AM) was implemented. All experiments were time matched and began at 09:00, during the lights off cycle in order to capture naturalistic wake behaviour. During the “lights off” period, red light was used to facilitate handling of animals. Animal welfare and weight recording was carried out daily.

Experimental procedures, animal housing and care were carried out in accordance with the Danish legislation according to the European Union regulation (directive 2010/63 of 22 September 2010), granted by the Animal Welfare Committee, appointed by the Ministry of Environment and Food of Denmark.

Naltrexone (Lundbeck, 12 mg/ml) was diluted in 0.9% saline solution and administered subcutaneously (SC) at 1, 3 and 10 mg/kg; clozapine (Novartis, 10 mg/ml) was diluted with 0.5% methylcellulose was administered SC at 0.3, 1, and 3 mg/kg; ketamine (Ketolar, 50 mg/ml, Sigma) was diluted with 0.9% saline and administered SC at 10 mg/kg; Vehicle (VEH) control was 0.9% saline solution.

Rat pharmacologically relevant doses and timing to peak effect of pre-treatment were estimated on the basis of a review of the literature (35, 77–84) in conjunction with application of the “Human Effective Dose conversion formula” (85) in reverse to existing human study data in which the combination of naltrexone plus ketamine have been evaluated (14, 86). Ketamine dose was determined through extensive in-house studies (unpublished) and literature (37, 87) which demonstrate profound modulation of LFPs at 10 mg/kg.

Custom accelerometers were manufactured by Ellegaard Systems and cables by PlasticsOne. Summed accelerometer output [equal to sqrt (X² + Y² + Z²)] was amplified (Precision Model 440; Brownlee, Palo Alto, CA, USA). Each of the 4 recording boxes with their own accelerometer and amplifier were calibrated to ensure equal output.

Depth electrodes (8IE3633SPCXE, E363-3-SPC, Elec.005-125MM SS, 25MM Length) and 6-way pedestals were purchased from PlasticsOne, manufactured by Bilaney Consultants GMBH.

Animals were habituated to placebo rimadyl pellets (Rimadyl MDs, BioServ, Flemington USA) 5 days prior to surgery. On the day of surgery, rats were anaesthetised with 0.25–0.3 ml/100 g subcutaneous (S.C.) injection of 1:1 hypnorm/dormicum and mounted in a stereotaxic frame (David Kopf Instruments, Tujunga, CA, USA) with blunt ear bars. marcain (0.2 ml s.c.) was injected under the scalp, and gel (Neutral Opthta Eye Gel) put on the eyes.

Holes were drilled in the skull for three depth electrodes (Figure 1A) (E363-series; Invivo1/PlasticsOne, Roanoke, VA, USA) in the right infralimbic PFC (AP: +3.0 mm and ML:−0.7 mm from bregma, DV:−3.0 mm from the skull surface), Nucleus Accumbens shell (AP: +1.6 mm and ML: +1.0 mm from bregma and DV:−6.8 mm from the skull surface) and thalamus (AP:−2.8 mm and ML: +0.7 mm from bregma, DV:−4.4 mm from the skull surface) and three screw electrodes (E363-series, 15 mm, Invivo1/PlasticsOne, Roanoke, VA, USA) at vertex (AP:−5.0 mm and ML: +5.0 mm from bregma), auditory cortex (AP:−4.8 mm and ML:−6.4 mm from bregma) and a reference electrode (AP: +8.0 mm and ML: −2.0 mm from bregma). Ends of depth electrodes were cut before use to create an exposed tip. During the procedure, the rat's nails were trimmed to prevent grooming damage to surgical site.

Rats received 0.3 ml each of Norodyl and Noromox SC during the procedure, were placed under a warming lamp for 4 h and provided extra muesli. Rats were closely observed for 10–14-days recovery, sutures removed after 7–10 days. No rats lost >10% pre surgery weight. Animals received rimadyl pellets twice a day for 5 days.

Up until surgery, rats were maintained a normal 12 hr light cycle (lights on at 0600) so that surgery could be performed in full light without disturbing the rat's circadian rhythm. After surgery, the light cycle was reversed (lights off 0600) and 21 days was allowed to elapse between experimental recording in order to allow rats to fully acclimatise.

Rats were anaesthetised with sevoflurane and 0.1 mV passed through the electrodes to create a lesion for histological validation of depth electrode placement. Rats were then decapitated, whole brains extracted, and the brains were placed in labelled, protective bags and frozen at −80°C until cryosectioning. On the day of placement validation, frozen brains were cut at the transverse fissure with a scalpel to remove the cerebellum and mounted with polyethylene glycol & alcohol (OCT Tissue Tek®, Sakure, The Netherlands) to a metal stand, placed in a cryostat (Leica CM3050 S) and 20 μm slices of the lesion sites were taken for examination with an optical microscope. Data from electrodes placed outside of NAc, PFC or thalamus was discarded.

The rats were split into three groups:

Group 1 (n = 50) received VEH + ketamine (10 mg/kg), clozapine (0.3, 1, and 3 mg/kg) + ketamine (10 mg/kg). Each rat was dosed twice (different treatments) following a pseudo-randomised schedule that balanced for drug doses and order with at least 7 days of washout in between to prevent cumulative tolerance.

Group 2 (n = 50) received VEH + VEH, VEH + ketamine (10 mg/kg), naltrexone (1, 3, and 10 mg/kg) + ketamine (10 mg/kg). Each rat was dosed twice (different treatments) following a pseudo-randomised schedule that balanced for drug doses and order with at least 7 days of washout in between to prevent cumulative tolerance.

Group 3 (n = 15) received clozapine (0.3, 1 mg/kg) or naltrexone (1, 3, 10 mg/kg) to quantify peak plasma and brain concentrations.

To facilitate habituation, rats were handled and placed individually into their respective EEG monitoring cage (Acrylic, 30 x 45 x 55 cm) within an electrically shielded, sound-proof box (90 x 55 x 65 cm) for at least 8 h (in <2-h sessions) in the week preceding experimental recording. during habituation, animals were connected to the EEG recording wire with the equipment switched off. Strict sound discipline was observed within the lab, preparation of drugs was performed under conditions that minimised disturbance sound.

On the days of recording, rats were placed into the cage, attached to a 6-pin recording wire on a rotating swivel and allowed to habituate for 120 min. A plastic spring (2.5 cm long when compressed and 2.5 cm diameter) was affixed to the rotating swivel and the recording wire affixed to the spring to allow 5 cm between the base of the cage and the terminal end of the wire. This alleviated the weight stress on the animal, allowed for vertical flexibility and prevented excess wire impeding animal movement. After 90 min of habituation to the recording environment, EEG and accelerometer recording began to establish a 30-min baseline for each session (Figure 1B). After the 30-min baseline recording, animals received a pre-treatment bolus of VEH (saline 0.9%), naltrexone (1, 3 or 10 mg/kg) or clozapine (0.3, 1, or 3 mg/kg) via SC flank injection.

Thirty minutes after pre-treatment, the animals received SC ketamine challenge (10 mg/kg) or VEH. Thirty minutes was selected as the optimal time for pre-treatment(s) to become effective following review of the literature (35, 77–84) and extensive in-house studies (unpublished). Recording of ECoG, LFP, and accelerometers continued for an additional 180 min after which animals were returned to their home cage.

Analogue LFP/ECoG signals were amplified (Precision Model 440; Brownlee, Palo Alto, CA, USA) and converted to a digital signal (CED Power 1401, Power 1 (625 k Hz, 16 bit) and CED Expansion ADC16; CED, Cambridge, England) at a sampling rate of 1 k Hz. LFP/ECoG signals were band-pass filtered at 0.01–300 Hz. Spike2 was used to simultaneously record inputs from microelectrodes, cameras and accelerometers, this ensured synchronised timestamps across file types.

Animal behaviour was recorded in parallel with a video camera and an accelerometer (custom-made with ADXL335Z, Analogue Devices) during each recording session. The accelerometer was fixed inside the plastic docking connector at the terminal end of the recording tether which screws onto the thread of the rodent's electrode headstage. The recorded video was used to qualify whether the animal was active or inactive, the latter here being defined as a state with no visible body movement with the exception of occasional micromovements of the nose and head. The accelerometer signal was then reviewed in parallel with the video recording, and an ad hoc threshold for distinguishing activity from inactivity was determined. For further processing the signal was smoothed with a gaussian kernel and divided into 7-s segments bins with 1-s overlap. If the signal during a segment was above the threshold for at least 60% of the time the segment was determined to be from an active period, correspondingly if the signal was below the threshold for at least 60% of the time the segment was determined to be from an inactive period. Segments that fulfilled neither criteria were left unclassified.

This study was intended to test whether the investigated drugs and behavioural states affect LFP and ECoG signals. This hypothesis was measured by consideration of EEG profiles across the following frequency bands: Delta (0.1–4 Hz), Theta (4–10 Hz), low Beta (10–20 Hz), high Beta (20–30 Hz), low Gamma (30–60 Hz), high Gamma (60–130 Hz), HFO (130–160 Hz), and Ultra High Frequency Oscillations UHFO (160–200 Hz) separated by behavioural state. To avoid power line interference, 2-Hz sections of frequency centred at 50, 100, and 150 Hz were excluded from analysis.

Analysis was carried out in MATLAB (MathWorks, Natick, MA). Signals were divided into consecutive 2-s segments with 1-s overlap. To minimise influence of artefacts, 2-s segments in which the signal exceeded ± 7 standard deviations (SD) from the mean were excluded from analysis. Furthermore, through comparison with the outcome of the behavioural state classification, each 2-s segment was assigned to either the active or inactive motor state or left unclassified. Next, a spectrogram with time and frequency resolution of 1 s and 0.5 Hz, respectively, was produced for each brain area by applying the Fast Fourier transform (FFT) to each 2-s segment. A spectrogram is a time series of power spectral densities and allows assessment of the spectral content of a signal over time, such as the presence of oscillatory activity in certain frequency bands.

When analysing the raw power, the logarithm was taken, otherwise each power spectral density was normalised to the baseline by dividing with the average power spectral density during the stable 30-min baseline period immediately prior to injection. The baseline-normalised spectral content was then converted to decibel (dB). Next, the power spectral densities were averaged over non-overlapping consecutive 10-min bins, positioned such that the time of injection is at 0 min, thereby producing spectrograms with 10-min time resolution. The steps of baseline normalisation and 10-min averaging were done both disregarding the behavioural state as well as only considering power spectral densities from segments classified as active or inactive, respectively. As a final step, grand averages were produced for each combination of brain area, behavioural state and treatment group.

Statistical analysis was conducted for averages over certain time intervals (10–30 min for pre-treatment, 40–70 min for ketamine challenge) and/or the already outlined frequency bands (see Tables 1–3). To investigate whether there were any significant treatment effects compared to the VEH + ketamine group, repeated measures analysis of variance (RM-ANOVA) was performed using MATLABs fitglme function with subsequent multiple comparison correction using Tukey's honest significant difference (HSD). P < 0.05 were considered significant. The fitted generalised linear mixed effects (GLME) model included an intercept and a factor for the treatment group, as well as a random-effects intercept for each animal to account for animal-specific variations. If applicable (i.e., when averaging only over a time interval or frequency band), the model also included a factor for the frequency/time bin and its interaction with the treatment group.

For each recording, the time the animal spent in the active and inactive behavioural state, respectively, was also calculated during non-overlapping 10-min bins, and grand averages were calculated for each treatment group. Statistical differences were assessed similar as for the spectral power in a certain frequency band, i.e., by using a GLME model with an intercept, a factor for the treatment group and the time interval and their interaction, and random-effects intercept for each animal, followed by Tukey's HSD. An animated visualization of the fundamental principles behind LFP recording, our recording procedure and some of the locomotor state differences is provided in the Supplementary Material.

To determine if the selected doses of naltrexone, clozapine and ketamine resulted in translationally relevant concentrations in the blood and brain of subjects, a drug exposure study was performed. Satellite animals (n = 3 per dose per drug) were treated by subcutaneous (SC) injection with Clozapine (0.3, 1, or 3 mg/kg) or Naltrexone (1, 3, or 10 mg/kg) then terminal venous blood and whole brain samples were taken at 1 h for exposure determination. In brief, plasma was isolated from whole blood and whole brains were isolated according to a previously described protocol (87). The brain tissue was prepared for extraction by dilution in buffer (1:5 w/v in deionised water) then homogenised by isothermal focused acoustic ultra-sonication using a Covaris instrument [Covaris E220x, 3.5 min at a bath temperature of 7°C with a peak power of 500 W and average power of 250 W (1,000 cycles per burst, duty cycle 50%)].

Total drug concentrations (Naltrexone or Clozapine) were determined in plasma and brain samples using high performance liquid chromatography coupled with tandem mass spectrometry (LC-MS/MS). The plasma (25 μL) and brain homogenate (25 μL) samples were precipitated with acetonitrile (4 volumes), centrifuged (3,500 g, 20 min, 5°C) and the supernatant (50 μL) diluted with water (3 volumes) before injection on the LC-MS/MS system. Drug concentrations were determined from calibration lines of known concentrations spiked into control plasma or brain homogenate and extracted under identical conditions. Bioanalysis was performed using a Waters Aquity UPLC coupled to a Waters XevoTQXS detector. A Waters Acquity UPLC HSS C18 SB, 1.7 μm, 30 × 2.1 mm column was used operating at 40°C. Mobile phase A consisted of 0.1% Formic Acid in water and mobile phase B of 0.1% Formic Acid in Acetonitrile. The LC flow rate was 0.6 mL/min. Analytes were separated on the LC column using a gradient. From 0 to 0.5 min the gradient was held at 2% mobile phase B. From 0.5 to 2 min B changed from 2 to 95% and was held at 95% until 2.5 min. Thereafter, between 2.5 and 2.7 min, B changed to 2% and was held at 2% from 2.7 to 4 min. Electrospray ionisation-MS (ESI-MS) was performed in positive MRM mode. For ketamine, clozapine, and naltrexone the parent:daughter [M+H]⁺ ions: 327.09⁺→ 270.08⁺ and 342.17⁺→ 270.15⁺ were selectively monitored for quantification, respectively.

Clozapine pre-treatment elicited oscillatory activity throughout the recording regions during Inactive epochs. Interestingly, the mid-dose (1 mg/kg) induced Inactive LFP power across the broadest range of frequency bands and brain areas (Figures 1F,G, Table 2 and Supplementary Tables 2, 3). Clozapine dose dependently increased Delta [0-4 Hz] activity in the Inactive Thalamus, PFC, and most substantially in the AC (p = 0.0006; p = 0.004; p = 0.0009). During Inactivity, clozapine (1 and 3 mg/kg) substantially enhanced spectral power in frequency bands between 30 and 60 Hz and across all electrodes.

Clozapine's effects on Active spectra were primarily depressive. In the Active PFC and Thalamus, activity in several frequency bands (low and high beta [10–20 Hz; 20–30 Hz] and low y [30–60 Hz]) were depressed by clozapine (3 mg/kg). Suppression in Active epochs was eclipsed when analysing both motor states.

Clozapine largely reversed ketamine's effects on lower bands, and enhanced effects >60 Hz. Ketamine induced depression of Theta [4–10 Hz] was completely ameliorated by clozapine in the Inactive state. In low beta [10–20 Hz], where ketamine induced suppression was more profound, clozapine partially returned LFP power towards baseline throughout the AC, PFC and Thalamus (3 mg/kg: p = 0.0006; p = 0.0007; p = 0.0009) (Supplementary Tables 6, 7). In the AC for example, ketamine depressed low beta to 40.18% of baseline, and 3 mg/kg clozapine returned this to 84.14% of baseline. A similar relationship, though of a lower magnitude, was also displayed in neighbouring frequency band high beta [20–30 Hz]. Reversal of beta suppression was exclusively seen in the Inactive state. By contrast, Active beta depression at the PFC and NAc (Table 3 and Supplementary Tables 6, 7) was exacerbated by clozapine (3 mg/kg) (p = 0.015; p = 0.005).

Ketamine-induced power in higher frequencies was synergistically enhanced by clozapine. Robust, dose-dependent increases were seen to ketamine-induced y [60–130 Hz] and HFO [130–160 Hz] in the Inactive NAc [3.05dB to 3.90dB, p = 0.00097; 4.63–8.51 dB, p = 0.00096, respectively]. Clozapine (3 mg/kg) also dose dependently reversed ketamine-induced depression of low y in the Active PFC, returning it almost to baseline. Analysis of LFP without separating by locomotor state rendered this effect invisible (Supplementary Table 3).

Increasing doses of clozapine also modulated the peak frequency of ketamine-induced spectra in the NAc. Clozapine increased peak power, but downshifted HFO peak frequency [from 151 to 143 Hz] and low y [58 Hz to 51 Hz] (Figure 2A). Interestingly, clozapine dose and peak HFO exhibit a biphasic relationship – 1 mg/kg clozapine peak HFO was higher than either 0.3 or 3 mg/kg. The nadir of beta suppression was also downshifted by clozapine, from 18 to 15 Hz.

Naltrexone (time 0) reduced oscillatory power globally in the acute pre-treatment phase (10–30 min) across a broad range of frequency bands (Figures 1F,G, Table 4 and Supplementary Tables 4, 5). Naltrexone decreased Inactive high beta [20–30 Hz] power and a biphasic relationship was seen between dose strength, with the mid dose (3 mg/kg) inducing the greatest depression [NAc, p = 0.001; PFC, p = 0.006; Thalamus, p = 0.0008]. Increasing doses of naltrexone depressed all frequency bands >30 Hz during Inactive epochs and across all electrodes. Active HFO power was also reduced below baseline at every electrode (10 mg/kg/Active; AC, p = 0.0007; NAc, p = 0.0009; PFC, p = 0.04; Thalamus, p = 0.0009).

Naltrexone pre-treatment did not significantly alter ketamine-induced beta depression in the Inactive or Active state (Table 5 and Supplementary Tables 8, 9). Non-modulation of low beta [10–20 Hz] was consistent at all electrodes and states (10 mg/kg: Inactive: AC, p = 0.99/; NAc, p = 0.99; PFC, p = 0.99; Thalamus, p = 0.39; Inactive: AC, p = 0.77; NAc, p = 0.19; PFC, p = 0.83; Thalamus, p = 0.74). In bands y and above, naltrexone reduced ketamine-induced power in the Inactive PFC (10 mg/kg, low y, p = 0.006; high y, p = 0.001; HFO, p = 0.0009; UHFO, p = 0.007), though the resulting LFP power remained substantially higher than baseline. Similar, but less consistent suppression was observed at other electrodes, and during Active epochs (Supplementary Tables 8, 9).

The width of peak HFO that ketamine affected was also modulated by naltrexone. Animals pre-treated with saline saw significant ketamine-induced power in a moderate band [135–167 Hz], 1 mg/kg naltrexone widened the band of affected frequencies by 43.8% [121–167 Hz] vs. saline, whilst 10 mg/kg naltrexone thinned affected HFO 84.4% [152–157 Hz] vs. saline (Figure 2B).

In agreement with previous locomotor-state-separated EEG analyses (37), power in frequencies above 30 Hz were broadly enhanced by ketamine, particularly in the Inactive state. Drug effects in the gamma band largely resemble those in HFO albeit with a lower magnitude, therefore as in other NMDAR antagonist LFP studies (35, 36, 38, 39, 49) we focus the discussion on effects in the HFO band. Ketamine induced-HFO were further strengthened by clozapine across both locomotor states. Increased HFO power can represent asynchronous activity in several distinct local neuronal populations, and/or circuit(s) that have become dysregulated (101, 102). Such asynchrony was indicated by the broader peak of spectral power/greater spectral entropy observed with increasing doses of clozapine in the present study (101, 103). Whilst it could be hypothesised that circuit desynchronisation occurs from clozapine (104) and ketamine (2) possessing opposing affinities for NMDAR on GABAergic interneurons, it has been demonstrated that the firing rate of local GABAergic interneurons in the rat thalamus and PFC are not significantly altered by ketamine (87). Ketamine potentially drives HFO through increased firing of excitatory pyramidal neurons (105–107). According to the “direct” hypothesis, ketamine-induced, NMDAR-dependent plasticity-related protein synthesis seen in pyramidal neurons (108, 109) is responsible for increased excitatory drive (107, 110).

Clozapine has affinities for several receptors that could recruit additional neuronal populations, generating more power yet less synchrony in the HFO band compared to ketamine alone. Agonism at NMDAR on local GABAergic interneurons, known generators of fast rhythmic activity in their own right (106), is one example. Clozapine additionally increases the firing of dopaminergic neurons in the ventral tegmental area by 100% (111), which innervates two structures this study observed broadband HFO increases within: the PFC (112) and NAc (113). However, single unit electrophysiology studies are necessary to characterise the precise neuronal sub-populations that are recruited during acute ketamine and clozapine exposure vs. ketamine alone.

Our findings indicate a clear difference in LFPs between ketamine, and ketamine plus naltrexone; a combination that is suspected to block RAAD effects (11, 14). While ketamine's RAAD effects are suspected to be driven through transient excitation of pyramidal neurons and synaptogenesis in key brain structures such as the PFC (114–119), the precise mechanistic pathway(s) through which improvement manifests is not yet fully elucidated. In addition, mechanisms have been identified through which opioid blockade could prevent RAAD (120) including BDNF upregulation and synaptogenesis (121), which is blocked by naltrexone (122); and acute agonism at mu-opioid receptors situated on neurons in the lateral habenula, dorsal raphe nucleus and ventral tegmental area. Inhibition of these neurons, via ketamine's antagonism at NMDAR and agonism at mu-opioid receptors, triggers downstream disinhibition of serotonergic and dopaminergic neurons in the PFC and NAc (120, 123–127). In this proposed circuit, as increasing doses of naltrexone block mu-opioid receptor agonism by ketamine, less excitatory disinhibition manifests in the PFC and NAc. Accordingly, we report a dose-dependent decrease of ketamine-induced HFO in these locations. If future studies confirm that naltrexone blocks ketamine's RAAD properties, increased HFO in the PFC and NAc should prove to be valuable biomarkers for antidepressant drug research.

Whilst naltrexone and clozapine had opposite effects in this band, it is important to be cautious drawing direct comparisons between the two until more acute studies have been conducted. One important limitation of this study is the exclusion of behavioural outcome measures for depressive and psychotomimetic symptoms. Thus, we can only say that in drug combinations that block RAAD effects in humans, we see suppression of ketamine induced HFO. Investigation in human subjects and in pre-clinical depression models to characterise the relationship between HFO amplitudes and RAAD effects is recommended.