Enhanced Beta2-band Oscillations Denote Auditory Hallucination in Schizophrenia Patients and a Monkey Model of Psychosis

Abstract

An electroencephalographic (EEG) signature of auditory hallucinations (AHs) is important for facilitating the diagnosis and treatment of AHs in schizophrenia. We recorded EEG from 25 schizophrenia patients with recurrent AHs. During the period of AHs, EEG recordings exhibited significantly elevated beta2-band power in the temporal region, as compared to those recorded in the absence of AHs or during stimulation with verbal sounds. We further generated methamphetamine-treated rhesus monkeys exhibiting psychosis-like behaviors, including repetitive sudden searching actions in the absence of external intrusion, suggesting the occurrence of AHs. Epidural EEG beta2-band power in the temporal region of these monkeys was enhanced immediately after methamphetamine treatment and positively correlated with the frequency of sudden searching actions. Thus, the enhancement of temporal beta2-band oscillations represents a signature for AHs in both patients and a monkey model of psychosis, and this monkey model can be used for developing closed-loop neuromodulation approaches for the treatment of refractory AHs in schizophrenia.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12264-023-01100-2.

Introduction

Schizophrenia is a severe psychiatric disorder that affects ~1% of the population worldwide [1], and over half of patients suffer from auditory hallucinations (AHs), even during antipsychotic medication [2, 3]. In recent decades, studies of schizophrenia symptoms have focused mostly on cognitive disorders such as deficits in working memory and executive functions. However, impairment of basic sensory functions such as an auditory processing deficit underlying AHs also contributes substantially to the psychosocial disabilities of these patients [4]. Refractory AHs are closely associated with recurrence and exacerbation of the overall schizophrenic condition, thus treatments for AHs may help the clinical management of schizophrenia [4].

Identification of a robust physiological signature marking the occurrence of AHs is crucial for the development of online close-loop neuromodulation methods for treating AHs. Transcranial direct-current stimulation (tDCS) that inhibits the activity of the left temporoparietal cortex reduces AHs in patients [2, 5–7], although negative findings have also been reported using similar approaches [8–10]. Most previous neuromodulation studies have not considered the timing of neuromodulation relative to the occurrence of AHs, because of the lack of a reliable AH signature. There is increasing evidence that closed-loop “online” therapy, in which neuromodulation is applied during the occurrence of the symptom, may be more effective in modulating dysfunctional neural circuits. This is exemplified by the finding that online neuromodulation can result in a rapid and sustained alleviation of depression [11] and recovery of movements after stroke [12]. In this study, we identified similar EEG signatures for the occurrence of AHs in schizophrenic patients and for psychotic searching actions in methamphetamine-treated monkeys. And this signature could be useful for developing online close-loop neuromodulation methods aiming at alleviating AHs.

In previous studies on AHs, functional magnetic resonance imaging (MRI) has shown that cerebral activation in the right inferior frontal area [13], superior temporal gyrus [14], Broca’s area [15], and the ventromedial prefrontal cortex [16] are particularly relevant to the occurrence of AHs. The high temporal resolution of EEG makes it most suitable for marking the time of AHs for the potential application of closed-loop neuromodulation. Previous studies of the EEG pattern of schizophrenia patients with AH symptoms have yielded disparate results, showing elevated EEG power in the alpha-band [17, 18], beta-band [19, 20], or theta-band [21]. Magnetoencephalography (MEG) recording at the time of AHs also shows a reduction in the beta-band power [22]. This apparent discrepancy may result from the fact that some studies were performed by comparing patient populations with and without AHs, and EEG recordings were not acquired at the time of spontaneous AHs. In this study, we recorded the scalp EEG of schizophrenia patients during the period that the patient reported AHs, in the absence of AHs, and during verbal sound stimulation. We found that the elevation of beta2-band oscillations is the most prominent signature for AHs.

Monkey models of psychosis that exhibit AHs may be useful for developing neuromodulation treatment of AHs in schizophrenia patients. Animal models for psychiatric disorders are difficult to establish because monkeys cannot report the occurrence of AHs like patients [23]. Some studies have used audition-related false behaviors as an indication of AHs in mice [23], but it remains unclear whether such false behaviors correspond to AHs. A behavioral phenotype reflecting AHs requires further validation by AH-related neural activity in the brain. In this study, we generated a monkey model of psychosis exhibiting sudden searching actions in the absence of intrusion, suggesting the occurrence of AHs. This monkey behavior was found to be associated with enhanced beta2-band EEG power similar to that found in schizophrenia patients during a period of frequent AHs. Our result supports the notion that beta2-band enhancement is a reliable signature of AHs and the methamphetamine-treated monkey model may be useful for developing online neuromodulation methods for treating AHs in schizophrenia patients.

Materials and Methods

Experimental Design

This work comprised both human and monkey model studies. In the human study, we recruited schizophrenia patients with refractory AHs and recorded their EEG during the patient-reported occurrence of AHs, in the absence of AHs, and during verbal sound stimulation, together with EEG recorded from healthy participants. In the monkey model study, we used methamphetamine (Meth) administration to generate monkeys with psychosis, monitored their repetitive sudden searching actions, and recorded wireless epidural EEG signals from them before, during, and after the period during which the monkey exhibited searching behaviors.

Participants

For clinical experiments, we recruited 33 schizophrenia patients with refractory AHs. Recording of EEG from 25 patients (41.5 ± 15.6 years; mean ± SD) was analyzed, and EEG data from 8 patients were excluded due to either the large EEG artifacts caused by patients’ movement or failure in reporting AH occurrence. All patients were diagnosed by two experienced psychiatrists based on the criteria of schizophrenia in the fifth edition of the Diagnostic and Statistical of Mental Disorder (DSM-5), using the following inclusion criteria: (1) aged 18–65 years, (2) able to communicate verbally, (3) no history of neurologic disorders, serious physical illness, craniotomy surgery, and epilepsy, and (4) no suicidal or self-harm tendency. Each patient’s age, education, and handedness were collected and the Auditory Hallucinations Rating Scale (AHRS) was used to evaluate their AH severity, with higher AHRS scores corresponding to more severe AHs (Table 1). The antipsychotic medications used by patients included fluphenazine, clozapine, trifluoperazine, haloperidol, and chlorpromazine.

Table 1.

Clinical and demographic characteristics of patients.

	Patients (mean ± SD, n = 25)	Healthy controls (mean ± SD, n = 20)	Significance (P, t test)
Age in years	41.5 ± 15.6	37.1 ± 10.4	P = 0.28
Education in years	9.8 ± 4.0	12.8 ± 6.4	P = 0.07
Gender (female/male)	5/19	5/15	NA
Right-handed	25/25	20/20	NA
AHRS score	27.4 ± 5.2	–	–

Twenty healthy controls (37.1 ± 10.4 years; mean ± SD) were recruited through advertising. All healthy subjects met the inclusion criteria as follows: (1) aged 18–65 years, (2) able to communicate verbally, (3) no history of neuropsychiatric disorders, serious physical illness, craniotomy surgery, or epilepsy, (4) no suicidal or self-harm tendency. All participants signed the informed consent form before enrollment. This study was approved by the Ethics Committee of Shanghai Mental Health Center and was carried out in accordance with the principles of the Declaration of Helsinki.

Human EEG Recording

In schizophrenia patients, EEG was recorded at the time the patient often experienced AHs, under two conditions during daytime: 30-min eye-open resting state and 10-min listening to verbal sounds. During the recordings, patients were asked to keep still, and orally report when AHs occurred. To ensure the validity of the AH report, the procedure was performed by an experienced psychiatrist together with an EEG expert. The psychiatrist recorded the specific period of AH occurrence. In the healthy participants, EEG was recorded during the same two conditions described above. The recording room was sound-attenuated. The procedures of EEG recording included: (1) a 32-channel EEG cap (Greentek, CM-HC01-2037) was placed according to scalp landmarks; (2) two reference electrodes and the ground electrode were placed at the bilateral mastoid processes and Fpz, respectively; (3) EEG was recorded with an amplifier (Compumedics, Grael 4K-EEG) at a sampling rate of 4096 Hz, with electrode impedance maintained below 10 kΩ to reduce noise.

Generation of Monkey Model of Psychosis

Three adult male rhesus monkeys (macaque, 5–8 years old) were used in this study. The monkeys were single-housed in stainless-steel cages in which water was available ad libitum. They were weighed twice a week and fed enough food, fruit, and vegetables daily to maintain healthy body weights without becoming obese as determined by daily inspection and periodic veterinary examinations. All experimental procedures were approved by the Animal Committee of the Institute of Neuroscience and Center for Excellence in Brain Science and Intelligence Technology, Chinese Academy of Sciences. Animal care protocols were in line with the US National Institutes of Health Guide for the Care and Use of Laboratory Animals.

Each monkey was first habituated to wear a nylon collar and sit in a standard monkey training chair for a few weeks in the regular housing. After the adaptive behavior training, surgical implantation of intravascular catheters for Meth administration was performed under general anesthesia. The monkeys were anesthetized with an intra-muscular injection of Zoletil-50 (5 mg/kg). The anesthesia was maintained during surgery with isoflurane (~1.75%) mixed in oxygen (O₂) by a Veterinary Inhalation Anesthesia Machine (RWD R620-S1, China). Basic physiological vital signs, including heartbeat, oxyhemoglobin saturation, carbon dioxide level, and rectal temperature were monitored concurrently (MINDARY iPM12, China) [24]. A silicone rubber catheter (POLYSITE 3007ISP, France) was implanted in the internal jugular or femoral vein and exited in the mid-scapular region [25]. Monkeys were allowed to recover from surgery for 2 weeks before initiating the subsequent experiments.

During experimental treatment, monkeys were moved to a ventilated and sound-attenuating chamber fitted with an operation panel (AniLab Software & Instruments) in front of the training chair. The panel was equipped with a hand-pressing lever and two signaling lights of different colors. The white and green lights signal the trial onset and reward delivery via lever pressing, respectively. All data on light signals, lever pressing, as well as dose and duration of Meth delivery were collected with commercial software (AniLab Software & Instruments).

Self-administration of Meth was performed by monkeys via lever pressing for intravenous injection of Meth (0.1 mg/kg) in response to the white light signal. The number of Meth-lever presses for the monkey to receive one Meth injection increased gradually from 1 to 15. The green light was turned on to signal Meth delivery. A 60-s timeout period (inter-trial interval, ITI) followed each Meth delivery, and no Meth injection was provided during the ITI, regardless of whether the monkey pressed the lever.

After the monkey had learned to press the lever 15 times to acquire Meth, we proceeded with 3–4 rounds of chronic Meth treatment. During the first 2–3 rounds, Meth was self-administered by the monkey for 7 days (2 h per day) with the total amount of Meth received by the monkey at ~1 mg/kg per day, a dose lower than that normally taken by Meth abusers (~2–10 mg/kg per day). During the last round (3rd or 4th round) Meth was self-administered together with experimenter-initiated supplementary injections to increase the total dose received by the monkey to 1.5 mg/kg per day. This extended the range of Meth dose treatment, facilitating later analysis of Meth dose-dependence of monkey behaviors. On each day of Meth administration, we collected EEG signals and recorded monkey behaviors before, during, and after the two-hour Meth treatment. Following each 1 week of Meth treatment, there was a 3-week Meth-free period, simulating a drug abstinence period.

Monkey Epidural EEG Recording

Surgical implantation of EEG electrodes for epidural EEG recording was performed under general anesthesia, as described before. Two EEG electrodes (bipolar electrodes) were implanted at the surface of the dura mater via metal nails screwed through the skull over the left temporal region and frontal lobe (first quarter along the middle line between the brow and occipital ridges), corresponding roughly to the T5 and Fz locations for human EEG recording, respectively. In addition, an electromyography (EMG) electrode was implanted in the monkey’s trapezius muscle. The wireless implant (L-03, Data Sciences International, DSI) was implanted subcutaneously on the monkey’s back [26]. Monkeys were allowed to recover from surgery for 2 weeks before initiating the subsequent experiments. All animals received intra-muscular injections of sufentanil (2.5 μg/kg) twice a day for 3 days or longer if they exhibited signs of discomfort.

Data Preprocessing

Human and monkey EEG data were analyzed by EEGLAB and FieldTrip toolboxes in MatLab software (2019a, MathWorks Inc.), using customized codes. The Harvard Automated Processing Pipeline for EEG was used to reduce noise and remove artifacts from EEG signals [27]. The EEG signals were first filtered at 1–250 Hz, together with a notch filter at 50 Hz to remove line noise. The bad recording channels were then rejected based on their high impedance or displacement during recording. Eye- and muscle-generated EEG artifacts were removed using wavelet-enhanced principal component analysis, followed by further artifact removal with the Multiple Artifact Rejection Algorithm when independent components had a >50% chance to be labeled as artifacts. Finally, the rejected bad channels were interpolated using spherical splines. The cleaned EEG signals were re-referenced by the average signal from all recording channels. The processed EEG signals were segmented into three periods: resting (without AH), AH period, and verbal sound exposure periods. The AH period refers to the time of frequent AH occurrence based on self-reports. The “no AH” period corresponds to at least 5 min of continuous data without AH occurrence, and the verbal sound exposure period refers to a period without AH when the patient was exposed to verbal sounds from standard Chinese comic shows. The noise and artifacts of epidural EEG signals collected from monkeys were manually removed based on visual inspection of raw epidural EEG signals and their spectrograms.

EEG Data Analysis

The time-frequency analysis, power analysis, and phase-locking value (PLV) analysis were used to identify EEG signatures collected from human participants and monkeys. Multi-taper spectral analysis (3 tapers, discrete prolate Slepian sequences, ±1 Hz smoothing) was used in time-frequency analysis.

For the EEG analysis, the power spectral density of EEG signals was measured by Welch’s method [28], using sine sliding overlapping time windows, with 512 sample points in each window and 250 overlapping sample points between adjacent windows. The power of EEG signals (Pow) was estimated by the composite trapezoidal rule in the delta band (1–4 Hz), theta band (4–8 Hz), alpha band (8–12 Hz), beta1 band (12–20 Hz), and beta2 band (20–30 Hz). The task-related power (TRP) between the activation period (AHs or verbal sound exposure) and the resting period was computed for each electrode (i) based on the formula TRP_i = 10∗Log (Pow_i, _activation) − 10∗Log (Pow_i, _reference). To compare with previous results, we grouped the values of power differences at 30 electrodes into the values for five cortical areas (frontal, central, temporal, parietal, and occipital) and two hemispheres (left and right). In the left hemisphere, the areas were defined as follows: frontal (Fp1, F3, F7), temporal (FT7, T3, TP7, T5), central (FC3, C3), parietal (CP3, P3), and occipital (O1). In the right hemisphere, the corresponding even-numbered electrodes were included. The midline electrodes (Fz, FCz, Cz, CPz, Pz, Oz) were not included in the analysis [29].

PLV analysis was applied as follows [30]. First, the EEG signals were segmented into 2-s epochs and each epoch at two different electrodes was first filtered for two frequency bands, denoted as $x_{f_A}(t)$ and $x_{f_P}(t)$. Second, the time series of the amplitude envelope and phases of $x_{f_A}(t)$ and $x_{f_P}\left(t\right)$ were obtained from the Hilbert transform, denoted as $A_{f_A}(t)$ and ${\phi }_{f_p}(t)$. Third, the time series of phases of $A_{f_A}(t)$ was obtained from the Hilbert transform again, denoted as ${\phi }_{f_A}(t)$. The $A_{f_A}(t)$ was then filtered into the same frequency range of ${\phi }_{f_p}(t)$. Finally, the PLV was calculated by the formula $PLV=|\frac{{\sum }_{t=1}^n{e}^{i({\theta }_{\mathit{At}}-{\theta }_{\mathit{Pt}})}}{n}|$, where n is the number of sample points in each segment, t is a sample point, and ${\theta }_{\mathit{At}}$ and ${\theta }_{\mathit{Pt}}$ is the phase angle of ${\phi }_{f_A}(t)$ and ${\phi }_{f_p}(t)$ at sample point t, respectively.

A permutation test was used to quantify the significance of PLV by comparing the observed PLV to the surrogate PLV in each 2-s epoch. The surrogate PLV was calculated between the original phase time series and the shuffled amplitude time series. The shuffled amplitude time series was constructed by reversing the order of two sample segments divided at random time points. Shuffling was repeated 200 times, corresponding to 200 different surrogate PLVs in each 2-s epoch. Finally, the PLV across epochs was extracted by averaging the PLV of each 2-s epoch when its significance level was <0.05.

Statistical Analyses

Statistical tests were applied using GraphPad Prism and MatLab. For human participants, the one-sample t-test was used to compare the mean TRP between AHs (or sound exposure) and the hypothetical mean (0). Holm-Sidak corrected multiple t-tests were used to compare the mean TRP between AHs and sound exposure (Fig. 1C–F).

Fig. 1.

Enhancement of temporal beta2-band oscillations accompanying auditory hallucinations (AHs) in schizophrenia patients. A Power spectrogram of EEG activity at channel FT7 (left temporal region) of schizophrenia patient ^#010 during resting state (no AHs), AH period, and sound stimulation. Note the enhancement of beta2-band EEG power during the AH period (red arrow). B Power spectra of EEG during the three periods shown in A. Note that the red arrow indicates significant differences between the AH period and the other two periods. C–F Summary of theta-, alpha-, beta1- and beta2-band EEG power changes (C, D, mean power of AH epochs minus resting state vs sound stimulation state minus resting state) and power values (E, F) at various brain regions in AH patients and healthy controls. Note the significant difference between the enhanced temporal beta2-band power of AH periods and sound stimulation periods relative to baseline power (no AH periods) (red arrow, “*”, t = 2.88, P = 0.029, Holm-Sidak corrected multiple t-tests). Error bars, SEM.

For analyses of monkey data, group differences in Meth intake dose (Fig. 2C) and post-Meth abnormal behaviors time (Fig. 3B–D) were compared by univariate analyses of variance (ANOVAs). Tukey’s multiple comparisons tests were used for multiple comparisons between the means of different groups. Linear regression analysis was applied between the duration of post-Meth abnormal behaviors and the total dose of Meth injection (Fig. 3E), and between the frequency of searching actions and beta2-band power, and further examined with Spearman rank correlation tests (Fig. 4E). The mean temporal beta2-band power under the Meth use and matched abstinence periods at different times were compared by Holm-Sidak corrected multiple t-tests (Fig. 4C, D). Finally, the paired t-test was used to compare the strengths of average meaningful PLV between different periods of Meth-use monkeys and schizophrenia patients (Fig. 5B–G).

Fig. 2.

Paradigm and setup for monkey training with Meth use. A Paradigm for monkey training with Meth use. Each round consists of a 1-week Meth-use period (gray, self-administration, SA) and a 3-week abstinence period (white). There are 3–4 rounds in the whole paradigm. The daily dose of Meth received on each day of the Meth-use period is labeled. B The monkey was habituated to sit in a standard training chair. Meth can be obtained by pressing the bar under the green light during Meth-use periods. C The amount of daily Meth used by the monkeys in each Meth-use round. The daily intake in each round was calculated by averaging daily intake during the last 5 days of the Meth-use period. During the last round, Meth was self-administered together with experimenter-initiated supplementary injections to increase the total dose received to 1.5 mg/kg per day (see Materials and Methods). *P < 0.05; **P < 0.01; ***P < 0.001; error bars, SEM.

Fig. 3.

Meth-Induced AH-like sudden searching actions in rhesus monkeys. A Images of typical AH-like sudden searching actions in monkey P28 after Meth use (see Movies S1 and 2 for monkeys P28 and P31, respectively). B, C The frequency of sudden searching actions in each hour (h) after post-Meth return to the home cage during the last Meth-use period (B), as well as the corresponding daytime during the last Meth abstinence period (C). Data from video-recorded monkeys (P28 and P31) are summarized. D Durations of post-Meth AH-like behaviors of each monkey during each Meth-use period. E Correlation between the duration of post-Meth AH-like behaviors and the daily Meth intake doses. The dashed line represents the linear regression with a correlation coefficient of 0.5653. *P < 0.05; **P < 0.01; “***”, P < 0.001; error bars, SEM.

Fig. 4.

Region-specific enhancement of beta2-band EEG power associated with Meth use and post-Meth AH-like sudden searching actions. A (Left): Reconstructed skull of monkey P28 based on CT images, with the locations of temporal and frontal electrodes (red circles), and example EEG spectrograms at the temporal and frontal electrodes during one Meth-use day (right). The white bar indicates the time when Meth was used. B Power spectra of the EEG activity shown in A. Note the significant elevation of beta2-band power (red arrow) in the temporal but not frontal region after Meth use. C, D Beta2-band EEG power recorded during Meth use (Meth) and extinction (Extin.) periods. (Upper) Example heat maps of peak beta2-band EEG power (see text) at the temporal region of monkey P28 and P31 for each day over a 1-week session of Meth use (“Meth”, red bars) and 3-week sessions of Meth extinction (“Extin.”, gray bars). Data are from the two rounds using low (~1.0 mg/kg) or high (~1.5 mg/kg) daily doses of Meth. (Lower) Beta2-band EEG power at the temporal region over days during Meth use (red) and Abstinence (black) periods. Note the average frequency of sudden searching actions for each minute after post-Meth return to the home cage (during the last Meth access period) are plotted in purple lines. Vertical shadow rectangles (gray) label the times when Meth was used. *P < 0.05; **P < 0.01; Student’s t-test. E Correlation analysis between the frequency of sudden searching actions and the power of temporal EEG in all 10-min bins in each monkey (black points, P28; gray points, P31) during Meth-use periods. Dashed lines represent significant linear correlations with coefficient r = 0.53 in monkey P28 (Spearman rank correlation tests, P = 0.0001) and r = 0.53 in P31 (Spearman rank correlation tests, P = 0.0001), respectively; error bars, SEM.

Fig. 5.

Changes of band-specific coupling between frontal low-frequency phase and temporal beta2 amplitude in Meth-treated monkeys. A Diagram of the data epochs from the continuously recorded epidural electrophysiological signals in monkeys during Meth use and extinction periods. B–D Phase locking analyses between frontal low-frequency (B, delta; C, theta; D, alpha) phase and temporal beta2-band amplitude in monkeys (P28 and P31) during the epochs before (B), during (D) and after (A) Meth intake during Meth-use periods (Meth) (n = 11, data points for monkeys P28 (6) and P31 (5)) and the corresponding epochs without Meth intake during Meth extinction periods (Extin.) (n = 14 and 7 data points for monkeys P28 and P31, respectively). E–G Phase locking analyses between frontal (F7) low-frequency and temporal (T5) beta2-band oscillations in schizophrenia patients with auditory verbal hallucinations during resting state (Resting), auditory verbal hallucination period (AVH), and sound stimulation period (Sound exposure). Note that in schizophrenia patients and Meth-use monkeys, phase locking between frontal delta-band and temporal beta2-band is enhanced during AVH or AH-like periods. *P < 0.05, ***P < 0.001, paired t tests.

Results

Enhanced Temporal Beta2-band Oscillation Accompanying AHs

We first examined the EEG power spectrograms and average power at different frequencies for schizophrenia patients at three periods: “Resting” (without AH), “AHs” (with frequent AHs), and “Sound exposure” (verbal stimulation in the absence of AHs). The EEG signals at the left temporal area (FT7) from one patient are illustrated in Fig. 1A and B. The data revealed an elevated power in the beta2-band (~20–30 Hz) during the period of frequent AHs, as reported by the patient. The spectral power at four frequency bands for EEG data collected from five brain regions of both hemispheres of 25 patients was analyzed to determine the difference between the average power level during “AHs” and “Sound exposure” and that of “Resting”. The results are summarized in Fig. 1C and D, showing a significant elevation of power in the beta2 band (SEM, n = 25; difference values vs 0, t = 3.06, P = 0.005; one sample t-test) at the left temporal area. Furthermore, the mean power difference during AHs was also significantly larger than that during sound exposure at the left temporal area (SEM, n = 25, t = 2.88, P = 0.029; Holm-Sidak corrected multiple t-tests) (Fig. 1C, D). Thus, AHs in schizophrenia patients occurred with significant power enhancement of temporal beta2-band oscillation in the left hemisphere, in line with the hemispheric dominance of language processing. Finally, we applied spectral power analysis to similar EEG recordings from 20 normal participants and found no difference in beta2-band power during the resting period between schizophrenia patients and healthy controls in either the left or right hemisphere (Fig. 1E, F). This indicates that the enhanced EEG beta2-band power during AHs is specific to schizophrenia patients.

Meth-treated Monkeys Show Behaviors Suggesting AH Occurrence

AH is a psychotic symptom prevalent among human Meth abusers. We generated addiction-like psychotic behaviors in rhesus monkeys by chronic i.v. self-administration of Meth, and found that these monkeys exhibited abnormal searching behaviors in the absence of external intrusion, suggesting the occurrence of AHs. Three monkeys were each subjected to 4 rounds of Meth treatment over 4 months (Fig. 2A). In each round, the monkey received i.v. injection of Meth during the first week, by pressing a lever 15 times following the trial onset light to receive one injection for a period of 2 h (Fig. 2B, see Methods). This one-week Meth injection was followed by 3 weeks of Meth abstinence. We found that the total dose of self-administered Meth increased in the 2^nd and 3^rd rounds in two monkeys (P8 and P31). In the 4^th round, besides self-administered Meth, the monkey also received an experimenter-initiated Meth injection to elevate the total dose received by the monkey to a level of 1.5 mg/kg per day, close to the dose used Meth addicts (Fig. 2C).

During the period of the Meth treatment, we noticed that Meth-treated monkeys displayed several abnormal behaviors after returning to the home cage, such as increasing frequency of teeth grinding, self-scratching, and sudden head and body turning towards a particular direction, apparently in search of something (Fig. 3A, Video S1–2). The latter occurred in the absence of any intruder or sound stimulus in the room, suggesting these behaviors were due to AHs. Such sudden searching actions often occurred while the monkey appeared to be in a drowsy state. Notably, these monkeys also failed to show the usual responses to food delivery or human calls which occurred when the monkey was free-moving in the home cage, but not on the monkey chair. For two monkeys (P28, P31), we obtained complete video records of their behaviors. During the last round of Meth treatment (total dose 1.5 mg/kg per day), the frequency of these searching actions was very high (>60/min for monkey P28) in the first hour after returning to the cage. The data showed that the frequency of the searching actions was high after the Meth treatment, gradually declined to a significantly lower level after 4 h (SEM, n = 14, F = 32.03, P = 0.0001; one-way ANOVA, Tukey's multiple comparison tests), and disappeared after 12 h (Fig. 3B), when the usual responses to food and human calls recovered. During the days with no Meth treatment, we observed very low levels of this searching behavior throughout the daytime (Fig. 3C). Linear regression analysis showed a significant day-to-day correlation between the duration of post-Meth searching actions (Fig. 3D) and the total dose of Meth administered (n = 52, r = 0.57, P = 0.0001; Spearman rank correlation tests) (Fig. 3E). Taken together, we surmised that this Meth-induced searching behavior reflects AHs.

Enhanced Beta2-band EEG Power in Meth-treated Monkeys

To examine changes in brain activity associated with the sudden searching actions of Meth-treated monkeys, we recorded epidural EEG from the temporal and frontal regions of two monkeys (P28, P31). The locations for temporal and frontal electrodes were determined with computed tomography (CT) scanning for each monkey, as illustrated for P28 in Figs 4A and S1, together with the spectrograms for EEG recorded by the two electrodes in monkey P28 during a single day of the last round of Meth treatment. During the Meth-injection period, we found a clear elevation of the spectral power in the beta2-band (20–30 Hz) in the temporal but not the frontal region during awake hours before sleep onset. This was further quantified by calculating the average power over the 30-min period before and after Meth use at various frequencies. The results showed a clear elevation of beta2-band power in the temporal region, but only a slight elevation at higher EEG frequencies in the frontal region (Figs 4B and S2).

We further examined the changes in the beta2-band power at the temporal region of the monkey for each day during the last two rounds of 1-week Meth treatment and the 3-week Meth abstinence. This was done by calculating the average beta2-band power over 10-min windows throughout the day, and the results are displayed in pseudo-color in Fig. 4C, D for monkeys P28 and P31, respectively. During the night-time, the beta2-band power generally showed a substantial reduction, although the reduction was much more pronounced during the Meth-extinction days. The data for power changes throughout each day were averaged for Meth-treatment days and Meth-extinction days, together with the average number of search actions (per minute) observed after the Meth treatment each day. The results showed that during each day of Meth treatment, the beta2-band power gradually increased during the Meth treatment, reached the highest level within 30 min after the end of Meth treatment and the monkey’s return to the cage, and subsided to the baseline level a few hours later. No such increase in the beta2-band was found during the Meth- extinction days. Linear regression analysis between the average beta2-band power and the average frequency of searching actions for each 10-min bin over the 5-h post-Meth period of each Meth-treatment day showed a significant positive correlation (monkey P28: n = 190, r = 0.53, P = 0.0001; monkey P31: n = 136, r = 0.53, P = 0.0001; Spearman rank correlation tests) (Fig. 4E).

Since EEG signals could be contaminated by EMG-recorded muscle activity and monkey’s spontaneous locomotion, we also examined the power associated with the EMG recorded from the neck muscles (Fig. S3A) and locomotion monitored with a bio-signal cassette (Fig. S3B). Consistent with the enhanced head-turning behaviors, both EMG and locomotor activity increased for a few hours after post-Meth return to the cage. However, a similar level of EMG increases in the morning of all recording days with Meth use was not accompanied by any enhanced EEG power at the temporal region (Fig. S3C). Thus, the beta2-band temporal EEG signals we recorded after Meth use reflected temporal brain activity rather than muscle activity. Taken together, we conclude that the enhancement of temporal beta2-band EEG power represents a specific signature for AH-like behavior after Meth use.

We computed the PLVs of EEG channel pairs between the frontal and temporal areas in monkey models (Fig. 5A–D) and patients (Fig. 5E–G, S4A–C) during different periods. During and after Meth intake, we found that the PLVs between frontal delta-band and temporal beta2-band oscillations were significantly higher than the values of no-Meth use (n = 11, “B” vs “D”, t = 5.59, P = 0.0002; “B” vs “A”, t = 7.47, P = 0.0001; paired t tests) (Fig. 5B). Meanwhile, the PLVs between frontal alpha-band and temporal beta2-band oscillations were significantly lower than the values of no-Meth use (n = 11, “B” vs “D”, t = 5.62, P = 0.0002; “B” vs “A”, t = 4.51, P = 0.001; paired t tests) (Fig. 5D). However, we found no significant PLV changes between different periods in patients during different periods (Fig. S4A–C). According to whether the content of the AH was semantic speech, we subgrouped patients into an auditory verbal hallucination (AVH) group (n = 19) and an auditory non-verbal hallucination group (n = 6). In the AVH group, the PLVs between frontal (F7) delta-band and temporal (T5) beta2-band oscillations were significantly higher during AH periods than that during resting (“AHs” vs “Resting”, t = 2.44, P = 0.025; “AHs” vs “Sound exposure”, t = 2.26, P = 0.037; paired t tests) (Fig. 5E). As a control, there was no significant frontotemporal PLV change between “Resting” and “Sound exposure” periods both in AVH patients (Fig. 5E) and non-verbal AH patients (Fig. S4A–C). These results suggest that more synchronous frontal activity, as reflected by delta oscillations, may play a role in elevating beta2-band activity at the temporal region.

Discussion

In this study, we found a significant enhancement of the beta2-band EEG activity in the left temporal brain region when schizophrenia patients reported frequent occurrences of AHs. Notably, this enhanced temporal beta2-band activity was not found during external sound stimulation or the resting state, showing the specificity of this EEG signature for AHs. A similar elevation of temporal EEG activity was also found in Meth-treated rhesus monkeys. In these monkeys, we found that an enhanced beta2-band power was highly correlated with a stereotyped sudden searching action that implicates the occurrence of AHs in these monkeys. Taken together, our results show that enhanced temporal beta2-band EEG oscillations could represent a signature for AHs.

Given the evolutionary proximity of monkeys to humans, the development of non-human primate models for psychiatric disorders could greatly facilitate the studies of their pathogenesis and treatment [31]. Amphetamine-type stimulants induce prevalent psychotic and hallucination behaviors in humans [32, 33], as well as in monkey [34, 35] and rodent [36] models of schizophrenia. Meth is a highly potent amphetamine-type stimulant and results in enhanced dopamine release [37]. Meth-induced psychosis is usually clinically indistinguishable from acute psychotic episodes in schizophrenia patients [38]. It is plausible that Meth-induced psychoses share common neurological mechanisms with schizophrenia and Meth-induced psychotic-like changes closely mimic the positive symptoms of schizophrenia [34, 36]. In this study, we found that rhesus monkeys displayed striking stereotyped AH-like behaviors during the Meth use period (Fig. 3A, Video S1–2). Consistent with the EEG signature in patients with AHs, Meth-induced AH-like symptoms are accompanied by enhanced temporal beta2-band oscillations in monkeys (Fig. 4C, D).

Psychosis-like behaviors have previously been reported in monkeys treated with amphetamine-type stimulants [34, 35, 39], including hallucination-like “wet-dog shakes” and “visual tracking of invisible objects”, which resemble the sudden searching actions found in the present study. Beyond qualitative description, we applied quantitative analysis of the behavior and associated cortical activity and compared the latter with similar EEG recordings during AHs in schizophrenia patients. Our findings provide a basis for further analysis of AH-associated brain activity in the monkey model. Further multi-channel recording methods such as the electrocorticogram from Meth-treated monkeys could reveal the mechanism underlying the enhanced temporal beta2-band oscillations.

Previous brain imaging studies have provided a wealth of information on the abnormal structural and functional connectivity of schizophrenia patients with AHs [40–44]. However, the pathogenic mechanism underlying schizophrenia-associated AHs remains poorly understood. During psychotic episodes, schizophrenia patients show heightened dopamine synthesis [45] and (drug) impulse-evoked dopamine release [46, 47]. According to the dopamine-mediated motivational salience hypothesis, the heightened dopamine release in the brain results in an aberrant assignment of salience to internal representations of auditory percepts, leading to AHs [48–50]. In a mouse model of AHs, erroneous responses in an auditory-cued task, which was proposed to represent AHs, were caused by abnormal D2 dopamine receptor activation in the thalamocortical pathway in the superior temporal lobe [51]. In rodent models of cocaine addiction, aberrant dopamine release in the nucleus accumbens and prefrontal cortex (PFC) has been found [52–55]. Thus, the AH-like behaviors we observed in Meth-treated monkeys may be attributed to heightened dopamine levels in the PFC. In our EEG analysis of Meth-treated monkeys, we found enhanced beta2-band power and increased phase-amplitude coupling between the phase of frontal delta waves and the amplitude of temporal beta2-band power during and after the period of Meth use. Generally speaking, low-frequency brain oscillations would be associated with a wide range of brain activation, while high-frequency brain oscillations tend to reflect the activity of localized brain regions [56]. Therefore, we speculate that this AH associated with β2-band oscillation enhancement reflects the activity change of a specific small area of the brain in the temporal region. It is speculated that this brain region may be the primary auditory cortex. Meanwhile, the PFC is known to be involved in selecting salient auditory signals via its functional interaction with superior temporal cortices [57, 58]. And the increased phase-amplitude coupling between the phase of frontal delta waves and the amplitude of temporal beta2-band power may reflect the disturbed functional interaction between PFC and superior temporal cortices. The heightened dopamine level associated with schizophrenia or Meth addiction may lead to this abnormal pattern of PFC-temporal cortical activity or thalamic inputs to the auditory cortex [51, 59], causing AH-related auditory percepts.

The phenomenon of AH represents an impaired perceptual process in which prior auditory memory stored in the brain is consciously accessed out of context and perceived to be real auditory input. In developing an animal model for studying AHs, the main difficulty lies in the experimental access to the auditory perception of the animal. Using behavioral actions of the animal as an indication of visual perception (awareness) has been achieved by using binocular rivalry and flash suppression paradigms [60, 61], in which a monkey reports the perceptual access of the image via a learned lever-pressing action. The sudden searching action of Meth-treated monkeys was used in the present study to indicate an auditory perception of sounds that were not present. Several lines of evidence suggest that AHs occurred in these monkeys after the Meth treatment. First, a similar head-turning searching action normally occurred when intruding sounds appeared in the animal room, indicating that this behavior reflects the auditory perception of intruding sounds, and this searching action was rarely observed in the silent environment for monkeys not treated with Meth. Second, during the period of frequent AH-like searching actions after Meth treatment, the monkeys showed reduced responsiveness of normal sensory stimuli as associated with the presence of a care-taker and food, consistent with the notion that AHs involve the suppression of normally salient sensory stimuli [62, 63].

The application of tDCS that inhibits the activity in temporal-parietal cortical regions has been found to reduce AHs in humans [2, 5, 64]. However, the current use of neural modulation therapy is mainly “offline”, where the timing of modulation does not correspond to the occurrence of the targeted symptom [65]. A few reports have shown that the effect of neuromodulation depends on the brain state of the patient [11, 66], and closed-loop neuromodulation is now considered to be a more promising approach for the neuromodulation of some refractory psychiatric diseases, as exemplified by the improved depression treatment with closed-loop “online” deep brain stimulation [11] and enhanced efficacy of cognitive training with concurrent anodal tDCS time-locked to the task performance [67–69]. The enhanced temporal beta2-band power accompanying AHs in schizophrenia patients represents a promising signature for the diagnosis of AHs. Future closed-loop “online” neuromodulation therapy for AHs could be developed to reduce the temporal beta2-band power during the period of AHs.

The present study of EEG signatures associated with AHs was limited in several aspects. In the patient study, the sample number included for analysis was relatively small, due to the irregular occurrence of AHs in schizophrenia patients, and the lack of cooperation from these patients in properly reporting the occurrence of AHs. In addition, the definition of AH periods based on patients’ reports was relatively imprecise, resulting in large variations in the EEG samples. In the study of Meth-treated monkeys, the EEG recording sites were limited to the frontal and temporal lobes, due to the availability of wireless EEG recording implants for monkeys, which are required for prolonged monitoring of free-moving monkeys. Finally, our results showed the correlation between the enhanced beta2-band oscillations and AH-like behaviors in monkey models. Whether these oscillations play a causal role in the generating of AHs remains to be examined. Nevertheless, this Meth-treated monkey model may help to elucidate the neural mechanism underlying AHs.

Supplementary Information

Below is the link to the electronic supplementary material.

Conflict of interest

The authors declare that there are no conflicts of interest.