Abstract
Ion channels play a crucial role in various physiological processes by selectively permitting the passage of ions, such as in neuronal signaling, muscle contraction, and cellular volume regulation. The patch clamp technique is the gold standard for investigating ion channel activities and cellular electrophysiological properties. Patch clamp facilitates clinical diagnosis, drug development, and neuroscience research by precisely measuring ionic currents. However, the recording and analysis are expertise-intensive, time-consuming, and labor-intensive procedures. This thesis aims to advance measurement and data analysis for automated patch clamp systems to address those challenges from two aspects: 1) achieve a high success rate in tight lipid-glass seal, enabling the automated acquisition of whole-cell recordings, and 2) automate the kinetic characterization of multiple ion channels. The main contributions of this thesis are summarized as follows:First, a lipid-glass seal strategy has been proposed to achieve a high success rate of gigaseal formation. Gigaseal minimizes the leakage current between the micropipette tip and cell membrane, allowing for accurate monitoring of the biological ion channel current. A correlation between the bath current and the seal current is identified by analyzing electrical signals during the formation process. Further experiments indicate that the sealing current's critical point originated from the cell membrane's sealing limit to the micropipette micro-opening. Based on the critical point, an algorithm is developed to optimize the threshold of the sealing current for the formation of a gigaseal. HEK 293 cells and C2C12 cells are employed to validate this approach. The seal threshold strategy improves the success rate of gigaseal formation to 95.9% and achieves an 80.8% yield of whole-cell recordings.
Second, enhanced with the lipid-glass seal strategy, automated patch clamp was further explored for rapid and efficient measurement of ion channel dynamics. A seal current-based real-time feedback is designed to detect the extent of cell membrane deformation and control the progress of the measurement process. The duration required to perform a single whole-cell recording has been reduced to 3-5 minutes. The stability and accuracy of the automated patch clamp are demonstrated by an arithmetic mean error of 0.029 and a mean-square error of 0.31, derived from comparing the calculated sealing threshold values with the actual measured values. Additionally, the design offers a systematic approach for single and dual-channel measurement. The single-channel configuration has been used to investigate electrically stimulated myotube contractions of differentiated myoblast cells. Cell signaling between coupled cells can be evaluated using the synchronized dual-channel configuration.
Third, an artificial intelligence framework has been proposed for characterizing multiple ion channel kinetics using whole-cell recordings. The framework integrates machine learning for anomaly detection and deep learning for multi-class classification. The anomaly detection excludes recordings that are incompatible with ion channel behavior. The multi-class classification combined a one-dimensional convolutional neural network, bidirectional long short-term memory, and an attention mechanism to capture the spatial and temporal patterns of the recordings. After training on 240 datasets, anomaly detection achieved an accuracy of 97.22% on 144 recordings. The multi-class classification model was trained using 139 datasets and achieved an accuracy of 97.58% in classifying 124 recordings into six categories. This framework effectively identifies recording features and understands the interactions between response currents by segmenting recordings into distinct phases based on ion channel characteristics.
Fourth, the artificial intelligence framework has been utilized for Alzheimer's disease drug screening and nanomatrix-induced neuronal differentiation. In the drug screening, the framework demonstrates the kinetics of various ion channels under memantine conditions. Memantine shifted the activation threshold of the potassium channel from 0 mV to 20 mV. The peak negative current of the sodium channel was significantly reduced from -6.64 pA/pF to -2.52 pA/pF. This study highlights the voltage-dependent inhibitory effects of memantine on endogenous channels, along with its antagonistic interactions with potassium, magnesium, and calcium ion channels. In experiments involving nanomatrix-induced differentiation, the framework characterized the specific phenotypic differentiation of neural stem cells. The classifier reveals the impact of differentiation conditions on sodium and potassium channels, which are associated with action potentials. The analysis results validate the functional properties of differentiated neurons for Parkinson's disease treatment.
In summary, the combination of the lipid-glass seal strategy and the artificial intelligence framework offers a novel solution in automated patch clamp for whole-cell recording acquisition and ion channel kinetics analysis. Based on this approach, cell physiological characteristics in drug screening and induced neuronal differentiation are systematically investigated. This study shows promise in enhancing the efficiency and accuracy of biomedical research.
| Date of Award | 17 Dec 2024 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Wai Chiu King LAI (Supervisor) |