Genetic Overlap Between Midfrontal Theta Signals and Attention-Deficit/Hyperactivity Disorder and Autism Spectrum Disorder in a Longitudinal Twin Cohort

Background Cognitive control has been strongly linked to midfrontal theta (4–8 Hz) brain activity. Such control processes are known to be impaired in individuals with psychiatric conditions and neurodevelopmental diagnoses, including attention-deficit/hyperactivity disorder (ADHD) and autism spectrum disorder (ASD). Temporal variability in theta, in particular, has been associated with ADHD, with shared genetic variance underlying the relationship. Here, we investigated the phenotypic and genetic relationships between theta phase variability, theta-related signals (the N2, error-related negativity, and error positivity), reaction time, and ADHD and ASD longitudinally in a large twin study of young adults to investigate the stability of the genetic relationships between these measures over time. Methods Genetic multivariate liability threshold models were run on a longitudinal sample of 566 participants (283 twin pairs). Characteristics of ADHD and ASD were measured in childhood and young adulthood, while an electroencephalogram was recorded in young adulthood during an arrow flanker task. Results Cross-trial theta phase variability in adulthood showed large positive phenotypic and genetic relationships with reaction time variability and both childhood and adult ADHD characteristics. Error positivity amplitude was negatively related phenotypically and genetically to ADHD and ASD at both time points. Conclusions We showed significant genetic associations between variability in theta signaling and ADHD. A novel finding from the current study is that these relationships were stable across time, indicating a core dysregulation of the temporal coordination of control processes in ADHD that persists in individuals with childhood symptoms. Error processing, indexed by the error positivity, was altered in both ADHD and ASD, with a strong genetic contribution.


EEG processing and analysis
Following average referencing, Adaptive Mixture ICA (AMICA) was used to calculate ICA components (1) using the nsgportal plug-in on the high performance computing resources available on Neuroscience Gateway (NSG, nsgportal.org)(2).The ICLabel algorithm was used for automatic detection and removal of ICs representing ocular artefacts (3).Continuous data was divided into epochs.Incongruent-incorrect trials were response-locked to incorrect responses with epochs -900 to 600 ms based on the time of the button press.Congruent-correct and incongruent-correct trials were stimulus-locked to correctly answered stimuli with epochs from -500 ms to 1000 ms relative to the onset of the target stimulus.All trials were baseline corrected using baseline -900 ms to -600 ms for response-locked and -400 ms to -100 ms for stimulus-locked trials.

Calculation of ITC
ITC was calculated over a window of 3 cycles at the mean frequency of 6.9 Hz (435 ms) starting 100 ms after the flanker stimulus, simultaneous with the onset of the central target stimulus, and ending before the response on average.The measure used for ITC is equivalent to the Phase Locking Value (PLV) defined in Lachaux et al (4), where the phase locking value is calculated between the trial central midline waveform and the target stimulus impulse signal, in the 3 cycle window starting 100 ms after the flanker stimulus to account for perceptual delay.
The formula for computing the complex normalized phase associated with an individual trial j is, Where sFCz,j(k) is jth trial epoch of the EEG FCz channel time series, k100 is the sample corresponding to the central target onset (100 ms after the flanker onset), e i2πfk/T is the complex exponential, f is the mean of the participant maximal ITC theta frequencies, k is the sample index, and T is the number of samples in 3 cycles of the mean max ITC theta frequency f.Then the ITC for a group of N trials is calculated length (absolute value) of the mean of the normalized complex phasors,

Questionnaire measures
In addition to the DIVA and ADOS described in the main manuscript two other questionnaires were included in the analyses: -Barkley Adult ADHD Rating Scale-IV (BAARS): is an empirically developed self-rating scale, based on DSM diagnostic criteria.It is suggested to evaluate the most reliable underlying dimensions of the symptom list for adults (5).Here we used the overall scale ranging from 18 to 72.
-Social Responsiveness Scale 2nd edition (SRS): is a self-reported scale measuring deficit in social behaviour associated with autism and their severity (6).Here we used the raw sum scores ranging from 0 to 195.

Twin Modelling
Since the sample was selected on affection status of either ADHD or autism, some corrections are needed within the standard twin model (7).First, one of the selection variables always needs to be included so that bivariate models are the smallest models to be fitted.Second, fixing the thresholds to the population prevalence of 4% for ADHD (8) and 1% for autism (9).The third adjustment is fixing twin correlations and heritability parameters for ADHD and autism to population estimates: a 2 = 0.76, c 2 = 0, e 2 = 0.24, rMZ = 0.76, rDZ = 0.38, using the same estimates for ADHD and autism (9,10).It was necessary to include ADHD and autism variables in each model to avoid bias due to the enriched study sample.Throughout this study we will only refer to the a 2 , c 2 and e 2 estimates in the model selected on ADHD because the estimates did not vary whether they were examined in the model with ADHD or autism.Due to the inclusion of the two dichotomous variables (ADHD and ASD), the genetic models were calculated once with each.However, the estimates did not vary; thus we only showed the estimates obtained with ADHD unless Rph, Ra, Rc and Re with ASD are being reported (11).

Table S1 :
The number of participants that have valid diagnostic and questionnaire, EEG and reaction time measures after artefact rejection prior to outlier detection as well as the number of outliers detected for each measure are outlined in the table below.Prior to twin analysis, outliers in EEG and reaction time measures were excluded based on the 2*IQR (interquartile range) criterion.

Table S2 :
The variances for adulthood questionnaires and latency measures estimated by twin analysis.

Table S3 :
Rc and Re between RTV, RTM, and ITC.

Table S4 :
Rc and Re between EEG and behavioural measures and childhood questionnaires (CAST, CPRS, SDQ) and adulthood conditions (ADHD, ASD).