Machine Learning Resolves Functional Phenotypes and Therapeutic Responses in KCNQ2 Developmental Epileptic...

Area of Science:

• Neuroscience
• Genetics
• Stem Cell Biology

Background:

• Pathogenic KCNQ2 variants lead to KCNQ2-developmental and epileptic encephalopathy (KCNQ2-DEE), a severe neurological disorder with no effective treatments.
• KCNQ2 encodes the K_V7.2 potassium channel, crucial for neuronal excitability, but its precise role in KCNQ2-DEE pathophysiology is unclear.
• Understanding K_V7.2 dysfunction is essential for developing targeted therapies for KCNQ2-DEE.

Purpose of the Study:

• To investigate the pathophysiological mechanisms of KCNQ2-DEE using patient-derived neurons.
• To identify functional neuronal phenotypes and biomarkers associated with KCNQ2-DEE.
• To evaluate the therapeutic potential of K_V7 activators in a precision medicine context.

Main Methods:

• Utilized human induced pluripotent stem cell (iPSC)-derived neurons from KCNQ2-DEE patients and CRISPR/Cas9-corrected isogenic controls.
• Employed microelectrode arrays (MEAs) to record extracellular spikes from a large number of neurons over an extended period.
• Applied supervised and unsupervised machine learning algorithms to analyze complex neuronal firing patterns and identify disease-specific features.

Main Results:

• Identified dyshomeostatic enhancement of Ca²⁺-activated small conductance potassium (SK) channels as a common mechanism in KCNQ2-DEE neurons.
• Discovered irregular spike timing and enhanced bursting as functional biomarkers for KCNQ2-DEE, influenced by genetic background.
• Demonstrated that retigabine, a K_V7 activator, can rescue disease-associated phenotypes, although with variable efficacy.

Conclusions:

• SK channel upregulation is a critical mechanism underlying KCNQ2-DEE.
• MEA recordings combined with machine learning provide a powerful platform for dissecting KCNQ2-DEE phenotypes and identifying biomarkers.
• This approach facilitates the evaluation of precision medicine interventions for KCNQ2-DEE in personalized neuronal models.