Data acquisition and analysis

EEG data were acquired continuously at a sampling rate of 512 Hz from 71 locations using 64 scalp electrodes mounted on an elastic cap and seven external channels (mastoids, temples, and nasion) (Active 2 system; Biosemi^tm, The Netherlands; 10-20 montage). Preprocessing was done using the EEGLAB toolbox (version 2019; Delorme & Makeig, 2004) for MATLAB (version 2019; MathWorks, Natick, MA) (full pipeline can be accessed at: github.com/DouweHorsthuis/EEG_pipeline_auditory_oddball_duration_tones_passive).

Data were downsampled to 256 Hz, re-referenced to TP8 and filtered using a 1 Hz high pass filter (0.5 Hz transition bandwidth, filter order 1690) and a 45 Hz low pass filter (5 Hz transition bandwidth, filter order 152). Both were zero-phase Hamming windowed sinc FIR filters. Bad channels were automatically detected based on kurtosis measures and rejected after visual confirmation (number of channels excluded: controls: M = 5, SD = 1; cystinosis: M = 5, SD = 3). Artifacts resulting from eye blinks and saccades were removed by running an Independent Component Analysis (ICA). After ICA, the previously excluded channels were interpolated, using the spherical spline method. Data were segmented into epochs of − 100 to 400 ms using a baseline of − 50 to 0 ms. These epochs went through an artifact detection algorithm (moving window peak-to-peak threshold at 120 µV). All subjects had trial rejection rates below 30%. No significant differences were found between the number of trials included in the analyses per group (standards: controls: M = 2527, SD = 77; cystinosis: M = 2412, SD = 262, t = 1.62, df = 16.14, p = 0.12, d = 0.59; deviants: controls: M = 444, SD = 17; cystinosis: M = 425, SD = 46, t = 1.47, df = 17.59, p = 0.16, d = 0.53).

The definition of the time windows of interest was based on the typical time of occurrence of the components, and on visual confirmation that amplitudes were maximal in these intervals. Mean amplitude for the N1 was measured between 90 and 130 ms, P2 was measured between 150 and 180 ms, MMN between 200 and 260 (100 to 160 ms post deviance onset), and P3a between 310 and 350 ms. Mean amplitude measures (trial-by-trial for N1 and P2 for standards only, difference in amplitude between standards and deviants for MMN and P3a) were taken at FCz, where signal was at its maximum for both groups. These amplitudes were used for between-groups statistics and correlations. Correlation analyses were computed across and between groups per component. Pearson correlations were performed for most of the variables, given that their distributions were not significantly different from the normal distribution, as tested by the Shapiro-Wilks Normality test [71], implemented using the shapiro.test function of the stats package in R [36]: verbal IQ: W = 0.94, p = 0.06; perceptual reasoning: W = 0.97, p = 0.56; working memory: W = 0.95, p = 0.15; N1 amplitude: W = 0.95, p = 0.19; P2 amplitude: W = 0.97, p = 0.57; P3a amplitude: W = 0.96, p = 0.37. The distributions of cysteamine dosage (W = 0.76, p = 0.01), number of transplants (W = 0.68, p = 0.01), and MMN amplitude (W = 0.92, p = 0.02) were different from normal distribution and thus Spearman correlations were instead used for these variables. All p-values (from post-hoc tests and correlations) were submitted to Holm-Bonferroni corrections for multiple comparisons [72], using the p.adjust of the stats package in R [36].