Pupillometric neuronal gain indicates the locus coeruleus to underlie predictive coding in autism

349 

Abstract Submission 

Nico Bast¹, Luke Mason², Emily Jones³, Tobias Banaschewski⁴, Christine Freitag⁵

¹Goethe University Frankfurt, Frankfurt, Hessen, ²Institute of Psychiatry, Psychology and Neuroscience, King’s College, London, London, UK, London, United Kingdom, ³Centre for Brain and Cognitive Development, Birkbeck, University of London, London, UK, London, United Kingdom, ⁴Department of Child and Adolescent Psychiatry, Central Institute of Mental Health, Mannheim, Mannheim, Germany, ⁵Goethe University Frankfurt, Frankfurt, Germany

Nico Bast  
Goethe University Frankfurt
Frankfurt, Hessen

Luke Mason  
Institute of Psychiatry, Psychology and Neuroscience, King’s College, London, London, UK
London, United Kingdom

Emily Jones  
Centre for Brain and Cognitive Development, Birkbeck, University of London, London, UK
London, United Kingdom

Tobias Banaschewski  
Department of Child and Adolescent Psychiatry, Central Institute of Mental Health, Mannheim
Mannheim, Germany

Christine Freitag  
Goethe University Frankfurt
Frankfurt, Germany

Predictive coding describes sensory processing as an updating of acquired priors to minimize costly prediction errors. Prediction errors are weighted based on an expected precision of the sensory input. Different sensory processing in autistic individuals has been suggested to include inflated prediction errors caused by increased precision weighting which (over-)emphasizes sensory input. This altered predictive coding might contribute to sensory phenomena reported by autistic individuals. The locus-coeruleus norepinephrine (LC-NE) system is a mechanism of arousal regulation that has recently been outlined to modulate neuronal gain in sensory processing.

We hypothesized that altered LC-NE functioning contributes to an increased precision weighting in autism. Matched groups of autistic (n=139) and non-autistic (n=88) individuals were assessed during an auditory oddball task (trials: k = 1400) with pupillometry and electroencephalography. The task was entirely passive and included trials of frequent standards (likelihood: 82%, tone duration: 50ms, tone pitch: 1000Hz), pitch oddballs (6%, 50ms, 1500Hz), length oddballs (5%, 100ms, 1000Hz), and pitch+length oddballs (6%, 100ms, 1500Hz). Pupillometric measure of baseline pupil size (BPS) and stimulus-evoked response (SEPR) were applied to index LC-NE tonic and phasic activity, respectively. Electroencephalography assessed amplitudes of mismatch negativity to assess an established index of prediction errors. Measures were modeled per trial to capture changes in precision weighting with task progression. A computational model assessed neuronal gain for standards. Linear mixed models were applied to investigate group differences and further confirmed by Bayesian posterior estimates.

Higher LC-NE tonic activity was associated with increased mismatch-negativity-associated amplitudes (rs = -.20 - -.22). LC-NE tonic activity differed between groups. Autistic versus non-autistic individuals showed a higher initial increase (autistic: Δβ = 0.20 [0.17, 0.22], non-autistic: Δβ = 0.11 [0.08, 0.14]) and overall attenuated decrease (autistic: Δβ = 0.02 [-0.01, 0.06], non-autistic: Δβ = -0.09 [-0.13, -0.05]) with task progression (trials: 1-1400). This was supported by Bayesian posteriors (autistic: b = 0.16 [0.13, 0.18], non-autistic: b = 0.00 [0.00, 0.00]). Higher LC-NE tonic activity was further supported by a higher computational estimate of neuronal gain for standards (group difference: d = 0.41 [0.09, 0.73]). Autistic versus non-autistic individuals were further characterized by increasing LC-NE phasic activity to pitch oddballs and decreasing mismatch-negativity-associated amplitudes to standards and length oddballs with task progression.

We conclude that higher LC-NE tonic upregulation is a mechanism of increased precision weighting that contributes to a different updating of priors in autistic individuals. This might reflect an arousal upregulation during sensory processing. The different prior updating is characterized by increasing precision-weighted prediction errors to pitch oddballs (SEPR) and decreasing "unweighted" prediction errors to standard trials and length oddballs (MMN-amp), which reflects different sensory processing. LC-NE functioning is outlined as a neurophysiological mechanism of predictive coding that might underlie sensory phenomena in autism.

Neurodevelopmental/ Early Life (eg. ADHD, autism) ¹

Bayesian Modeling

EEG/MEG Modeling and Analysis

Other Methods

Attention: Auditory/Tactile/Motor ²

Autism

Electroencephaolography (EEG)

Norpinephrine

Other - pupillometry