Multimodal Analysis of Language Regression Using Genome Sequencing, Neuroimaging and Neurophysiologic Findings in a Cohort of Minimal-to-Nonverbal Autistic Children
Clinical Genetics and Therapeutics
-
Primary Categories:
- Clinical Genetics
-
Secondary Categories:
- Clinical Genetics
Introduction:
This study aims to identify genomic, brain Magnetic Resonance Imaging (MRI), and electroencephalogram (EEG) biomarkers that correlate with a history of language regression in a cohort of minimal-to-nonverbal pediatric patients with Autism Spectrum Disorder (ASD). From an initial study population of 85 minimal-to-nonverbal verbal ASD patients, we established two cohorts: (A) those with language regression (n=28), and (B) those with no documented language regression (n=43). Using these techniques, we identified differentially enriched biological processes in genome data along with differences in EEG impressions between ASD patients with and without language regression.
Methods:
Utilizing the electronic medical record, we identified a cohort of 85 minimal-to-non-verbal patients with ASD/NDD who had clinical WGS available. Individual analysis for EEG reports, MRI impressions, developmental pediatric evaluations, and prenatal histories were performed. A categorization schema was developed by a team of clinical neurologists and geneticists who reviewed all available MRI and EEG data. Developmental screening identified 28 patients with documented language regression (cohort A), 43 patients with no language regression (cohort B), and 14 patients with questionable regression who were excluded from the analysis. Using variant annotation techniques, we identified pathogenic, likely pathogenic, and variants of unknown significance in genes previously implicated in ASD. EEG data analytics was accomplished through clustering and regression analysis using histogram analysis. Gene ontology enrichment analysis using topGO R package was implemented to identify biological processes impacted by mutations across the entire cohort, and separately in patients in cohorts A and B.
Results:
In our cohort of ASD individuals, gene set enrichment analysis identified significantly enriched GO terms related to learning and memory, synaptic function, behavior, brain development, and chromatin remodeling. In the analysis stratified by group, we identified a set of shared enriched terms related to learning, a set of enriched GO terms exclusive to cohort A related to neuron migration and morphogenesis, and a set of enriched GO terms exclusive to cohort B related to astrocyte and glial cell development. The WGS pathogenic + likely pathogenic hit rate was 36% in cohort A, and 31% in cohort B.
In EEG analysis, "Epileptiform Discharges" were more prevalent in cohort A, and both “Slowing” and “Slowing with Epileptiform Discharges” were more prevalent in cohort B. A slightly higher rate of preterm birth was found in cohort A compared to cohort B. No significant associations could be made using MRI impression data.
Conclusion:
Overall, cohort B had more EEG slowing than cohort A (63% vs 32%), which may be a reflection of increased severity of underlying cognitive dysfunction in patients who were never able to establish early language. Though cohort B had more EEGs with epileptiform discharges than cohort A (40% vs. 32%), we found that cohort A had more epileptiform discharges without slowing than group B (22% vs. 10%), which may reflect how network dysfunction from epileptic encephalopathy can result in language regression. These data reconfirm findings from other centers.
A major limitation of the study is the use of MRI text impressions rather than MRI source data. Future directions include using MRI source data and expert neuroradiologists for MRI subtyping schema. Additionally, we will attempt to stratify EEG findings with gene enrichment analysis data within each cohort. Research is ongoing; therefore, the potential cohort size and the significance of our results may change.
This study aims to identify genomic, brain Magnetic Resonance Imaging (MRI), and electroencephalogram (EEG) biomarkers that correlate with a history of language regression in a cohort of minimal-to-nonverbal pediatric patients with Autism Spectrum Disorder (ASD). From an initial study population of 85 minimal-to-nonverbal verbal ASD patients, we established two cohorts: (A) those with language regression (n=28), and (B) those with no documented language regression (n=43). Using these techniques, we identified differentially enriched biological processes in genome data along with differences in EEG impressions between ASD patients with and without language regression.
Methods:
Utilizing the electronic medical record, we identified a cohort of 85 minimal-to-non-verbal patients with ASD/NDD who had clinical WGS available. Individual analysis for EEG reports, MRI impressions, developmental pediatric evaluations, and prenatal histories were performed. A categorization schema was developed by a team of clinical neurologists and geneticists who reviewed all available MRI and EEG data. Developmental screening identified 28 patients with documented language regression (cohort A), 43 patients with no language regression (cohort B), and 14 patients with questionable regression who were excluded from the analysis. Using variant annotation techniques, we identified pathogenic, likely pathogenic, and variants of unknown significance in genes previously implicated in ASD. EEG data analytics was accomplished through clustering and regression analysis using histogram analysis. Gene ontology enrichment analysis using topGO R package was implemented to identify biological processes impacted by mutations across the entire cohort, and separately in patients in cohorts A and B.
Results:
In our cohort of ASD individuals, gene set enrichment analysis identified significantly enriched GO terms related to learning and memory, synaptic function, behavior, brain development, and chromatin remodeling. In the analysis stratified by group, we identified a set of shared enriched terms related to learning, a set of enriched GO terms exclusive to cohort A related to neuron migration and morphogenesis, and a set of enriched GO terms exclusive to cohort B related to astrocyte and glial cell development. The WGS pathogenic + likely pathogenic hit rate was 36% in cohort A, and 31% in cohort B.
In EEG analysis, "Epileptiform Discharges" were more prevalent in cohort A, and both “Slowing” and “Slowing with Epileptiform Discharges” were more prevalent in cohort B. A slightly higher rate of preterm birth was found in cohort A compared to cohort B. No significant associations could be made using MRI impression data.
Conclusion:
Overall, cohort B had more EEG slowing than cohort A (63% vs 32%), which may be a reflection of increased severity of underlying cognitive dysfunction in patients who were never able to establish early language. Though cohort B had more EEGs with epileptiform discharges than cohort A (40% vs. 32%), we found that cohort A had more epileptiform discharges without slowing than group B (22% vs. 10%), which may reflect how network dysfunction from epileptic encephalopathy can result in language regression. These data reconfirm findings from other centers.
A major limitation of the study is the use of MRI text impressions rather than MRI source data. Future directions include using MRI source data and expert neuroradiologists for MRI subtyping schema. Additionally, we will attempt to stratify EEG findings with gene enrichment analysis data within each cohort. Research is ongoing; therefore, the potential cohort size and the significance of our results may change.