The goal of this observational study is to compare subjects with at-risk-mental-state, early psychosis, schizophrenia, depression, and autism spectrum disorders, with healthy controls (N = 21 x 6). The main questions it aims to answer are:

• are EEG microstate anomalies associated with diagnosis, clinical and functional prognosis, both in resting conditions and during sleep ?
• are EEG microstates anomalies associated with differences in sensorimotor integration, prosodic and conversational, interoceptive, and narrative self ?
• an ancillary study will be to see whether in healthy controls EEG microstate properties vary under light hypnosis conditions.

Participants will:

• undergo deep phenotyping based on psychopathology and neuropsychological assessments
• undergo a high-resolution EEG (64 electrodes) with a resting period and a sensorimotor task; and healthy controls will have a light hypnosis period.
• undergo a recording of the characteristics of their voice (tone, prosody)
• undergo a one-night polysomnography
• undergo MRI and biological sampling for multi-omic analyses
• undergo a virtual reality experience

• Aim of the study: EEG microstates translate the resting-state temporal dynamics of neuronal networks throughout the brain. Here, the investigators aim to see whether EEG microstate anomalies could constitute markers of psychiatric disorders.

• Methods: six groups of 21 participants each will be included. There will be five groups of participants with psychiatric disorders (at-risk-mental state - ARMS, first-episode psychosis - FEP, schizophrenia - SCZ, major depressive disorder - MDD, and autism spectrum disorders - ASD) and one group of healthy controls. Our main objective is to test differences in means between the groups, at rest and during sleep, for each of the variables characterizing each of the microstates (duration, frequency, occupation time) as well as, secondarily, EEG measures of connectivity (somatosensory evoked potentials), cortical excitability (alpha-band power), and prosodic and conversational linguistic measures.

  • Regarding the microstates measures: a five minute eyes-closed resting-state EEG with 64 channels will be recorded (as part of the larger task including the sensorimotor task described below). A minimal preprocessing will be done with the MNE EEG software on Python, which includes a bandpass filter between 0.5 and 40 Hz, rereferencing to the mean, and visual and automatic correction for artifacts. Each recording will be visually reanalyzed by clinical neurophysiologists to ensure it is indeed an alpha-dominant, resting rhythm without any residual artifact. Microstate analysis will be done using the Pycrostates package. Global field power (GFP) will be determined for each participant. Only EEG topographies at GFP peaks will be retained to determine microstates' topographies, through a modified K-means clustering. For each subject the same number of GFP peaks will be extracted and concatenated into a single data set for clustering. A combined score will be used to compute the optimal number of clusters. The resulting clusters will be backfitted to each individual maps. Temporal smoothing will be used to ensure that periods of inter-peak noise, of low GFP, did not interrupt the sequences of quasi-stable segments. For each subject, three parameters will be computed for each microstate class: frequency of occurrence ("occurrence"), temporal coverage ("coverage") and mean duration. Occurrence is the average number of times a given microstate occurs per second. Coverage (in %) is the percentage of total analysis time spent in a given microstate. Mean duration (in ms) is the average time during which a given microstate was present in an uninterrupted manner (after temporal smoothing).
  • Regarding the linguistic measures: each participant undergoes a semi-structured interview with a trained experimenter. Both the participant and the interviewer wear head-set AKG-C544L condenser microphones, connected via AKG MPA VL phantom adaptors to a Zoom H4n Pro Handy recorder. Speech is digitally recorded at a sampling rating of 44000 Hz (16-bit). The distance between the mouth and the microphone is kept as constant as possible (2 cm) to assure consistent levels of vocal loudness. The interviews are done in a quiet room to limit environmental noise; the two interactants are placed as far as possible, to prevent crosstalk (i.e. speech of the interviewer caught by participant's microphone and vice versa). The .wav files obtained from the recordings are annotated using the Praat software and subsequently analysed with Praat and R. Prosodic features are extracted using the Prosogram tool (a set of Praat scripts, open-source) and a new modified version of scripts from the Prosogram tool. Turn-taking variables are extracted with new combined Praat and R scripts.
  • Regarding the sensorimotor intergration measures: the sensorimotor integration is investigated using a visuo-haptic task. On each trial, the participant, seated in front of a screen, has a visual instruction (a point to the right or left of the screen). The task consists of pressing one of the two buttons positioned on each side of the body with the index finger of the corresponding hand according to the visual instruction. A vibrotactile stimulator (small speakers wired to an Arduino electronic card modulated by an amplifier) is applied to the first dorsal interosseous muscle of both hands. 400 msec before the visual instruction, one of the two hands receives a tactile cue (vibration) on one hand for 100 msec. This cue is more or less reliable depending on the block. In some blocks, it is quite reliable, since 90% of the trials present the vibration and visual instruction congruently (indicating the same hand). Another condition is composed of only 50% of the congruent trials, and in this case, the tactile cue is not reliable. Two blocks with 70% congruent cases are carried out intermediately. Finally, a baseline block which does not contain any tactile cues is presented at the beginning and the end of the task. The order of the 90% and 50% blocks is randomized. The tactile and visual stimuli are generated with a MATLAB script. Each block consists of 100 trials, in total 500 trials. Electroencephalographic (EEG) data is recorded throughout the task, using a 64-channel EEG cap (from Biosemi) in order to record the electrical brain activity. The setup is coupled to an eyetracker, to control that the participant is fixating the cross at the center of the screen during each block.
  • Regarding the multidimensional self and episodic memory task (task design: Laboratoire Mémoire, Cerveau et Cognition): at baseline, participants will be submitted to self-reported questionnaires assessing their sense of minimal Self on 8 domains (Multidimensional Assessment of Interoceptive Awareness - Version 2) and sense of narrative Self on 5 domains (Tennessee Self Concept Scale - Short Form, Present). They will undergo a neuropsychological test assessing their visual episodic memory performance (Family Pictures from Wechsler Memory Scale-III). They will rate their current emotional state on a visual analogue scale on 4 domains (Mood Visual Analogue Scale). Following each of the two navigation sessions in virtual reality, which consist in a walk through a virtual city where participants encounter daily life events that aim to be incidentally encoded in episodic memory, associated with different levels of self-reference, participants will be submitted to self-reported questionnaires assessing their sense of embodiment on 4 domains (Embodiment Questionnaire), their sense of presence on 4 domains (Igroup Presence Questionnaire), and their cybersickness on 2 domains (Simulator Sickness Questionnaire). They will rate again their current emotional state on a visual analogue scale (Mood Visual Analogue Scale). Finally, participants will undergo two episodic memory tests: a free recall task and a recognition task. The free recall will be based on a verbal interview of 20 minutes, during which participants will be asked to recall all the events that they remember encountering in the virtual city. For each event, they will be asked to recall systematically and the most precisely possible: what was the event, where and when it happened during the navigation, in which of the two navigation it happened (source), who was the referent according to which the personal significance of the event was assessed, objective (perceptive) and subjective (phenomenological) details of the event, and if the event was vividly relived or felt merely familiar (Remember/Know procedure). The recognition test will be performed on a computer and programmed using the Python module Neuropsydia. All 32 encountered events mixed with 16 lures which were not encountered will be displayed successfully in a random order on a computer screen. For each event, several questions will be asked successfully and participants will click on what they consider the correct answer among several propositions: did they encounter the event (Yes/No), and if yes where it happened (among several possible localisations on a picture of the zone where the event occurred), when it happened (replacing the event in the chronological order with two other events), in which navigation (first or second navigation), and who was the referent (Me/Other). For each event, participants will also rate on scales ranging from 0 to 100: the degree of reliving or familiarity of the event (100 = Remember, 0 = Know), the perspective of the memory (100 = first-person perspective, 0 = third-person perspective), its vivacity, fidelity, emotional intensity, strength of associated bodily sensations, episodic self-reference, and semantic self-reference.

For all variables, the investigators will apply a repeated measures ANOVA, and use the following contrasts:

1. "ARMS, FEP, SCZ, ASD, MDD" vs. "Healthy subjects" (microstates are tested as markers of general psychopathology);

2. "ARMS, FEP, SCZ" vs "ASD, MDD" (microstates are tested as specific markers of psychosis; equivalently, the specificity of this signature for depression and ASD will be tested)

3. "ARMS" vs. "FEP" vs. "SCZ" (microstates are tested as evolutionary markers);

4. Finally, depending on the rate of transition to psychosis among UHRs, a comparison of "UHR-T" vs "UHR-NT" can be made (microstates are tested as predictive markers of psychosis All subjects will undergo a deep phenotyping including neuropsychology, psychopathology, neurological soft signs scales, as well as structural MRI, and genetic and epigenetic measures.

   • Hypothesis: imbalances in EEG microstates C and D are expected to be more pronounced across the spectrum of psychosis and in ASD compared to controls, MDD, and ARMS, and be associated with anomalies in somatosensory, interoceptive, and language characteristics.

Inclusion Criteria:

• age between 15 and 30 years
• subjects meeting CAARMS criteria for stage Ia or Ib mental states at risk (for stage Ia, mild or nonspecific symptoms of psychosis or severe mood disorder, and mild functional impairment; for stage Ib, moderate symptoms below intervention threshold and moderate functional impairment)
• subjects meeting any DSM-5 criteria associated with a first onset of psychotic symptoms (first episode psychosis)
• subjects satisfying DSM-5 criteria for depressive disorder
• subjects meeting DSM-5 criteria for autism spectrum disorder
• healthy control subjects recruited from the general population

Exclusion Criteria:

• suicidal risk
• severe or non-stabilized somatic and neurological disorders
• epilepsy
• head trauma
• IQ below 70
• for healthy control subjects, a family history of psychosis is an exclusion criterion
• bipolar disorder
• obsessive-compulsive disorder
• substance use disorder, except for cannabis, tolerated up to 5 joints/day.