UCLA

Mathematical modeling and simulations were employed to guide successfully the optimization of electroenzymatic biosensors for glutamate and for choline and to clarify how well such sensors can monitor rapid neurotransmitter signaling in the brain. Improvements in methods for accurate neurotransmitter detection in behaving animals are vital for the study of many neurological processes. However, the spatial and temporal resolution of even the most optimized sensors will be insufficient to record neurotransmitter dynamics directly in vivo, thereby causing potential confusion in the analysis of sensor data. Much of this confusion arises from the effects in vivo of both neurotransmitter mass transport and clearance on sensor recordings. These effects historically have been considered too complex or insignificant to investigate but must be included in the proper analysis of sensor data. Surprisingly, a chemical engineering approach for the analysis of mass transfer-influenced sensor kinetics has remained largely unexplored until now, where the value of the chemical engineering perspective is demonstrated through the study of electroenzymatic sensor performance in vivo.To increase spatial and temporal resolution and minimize uncertainties in data analysis, a model of an electroenzymatic glutamate sensor was used to reveal the theoretical limits of sensor performance and to suggest biosensor modifications that led to a six-fold increase in sensitivity to 320 nA M-1 cm-2 and an order-of-magnitude reduction in intrinsic response time (in the absence of external mass transfer limitations) to 80 ms while maintaining excellent selectivity. However, the improved response times were still an order of magnitude slower than the simulated response time at optimal sensitivity. Due to the importance of sensor response times for the proper analysis of data collected from measurements of transient glutamate release events in the brain, the difference in these response times was investigated with an adapted mathematical model that accounts for noncatalytic, reversible binding of glutamate to the proteins within the sensor immobilized enzyme coating. With reasonable parameter estimates, the new model was shown to fully resolve the prior discrepancy in sensor response time, and a set of future experiments was proposed to confirm the significance of this phenomenon. The sensor apparent Km, which is a measure of the linear calibration range, was shown to be influenced by O2 diffusion into the enzyme coating at relatively high glutamate concentrations rather than to be reflective of the immobilized enzyme kinetics, as was commonly believed. Such oxygen limitation is observed when bulk O2 concentrations are less than 33% of the local glutamate concentration. Similar simulations of choline sensors led to highly selective choline sensors with the improved sensitivity and intrinsic response time of 660  40 nA M-1 cm-2 and 360 � 50 ms, respectively. It also was determined that O2 may become limiting for bulk O2 concentrations <60% of the local choline concentration. Simulations of glutamate sensor response to glutamate transients showed that optimized sensors would have difficulty resolving Gaussian shaped transient glutamate concentrations present for <40 milliseconds in vitro and that optimized sensors in vivo may provide improved temporal resolution, although they are not likely to respond to glutamate concentration changes >30 �m from the sensor surface due to diffusional mass transfer limitations and glutamate uptake from the brain extracellular space. Detailed analysis of sensor response in vivo showed that variability in the possible rate of H2O2 clearance from brain extracellular space could result in a sensor recording that exceeds by 300% the value expected from sensor calibration in vitro. It was also shown that apparent response times could be significantly slower in vivo than in vitro due primarily to slow diffusion in the brain and that the decay in sensor response is not reflective of glutamate uptake rates in any meaningful way without extensive contextual details. Three-dimensional models additionally served to provide insight into the role of electrode size and enzyme deposition area in adequately detecting glutamate release from precisely defined and localized increases in neuronal activity, representative of typical changes in neuron firing frequencies. These results highlighted the benefits of sensor miniaturization and suggested probe design modifications to improve spatial resolution and detect glutamate release from smaller populations of active synapses. Modeling and simulations of electroenzymatic sensors in vivo have demonstrated the value in bringing a chemical engineering approach to bear on the optimization of such sensors and on the interpretation of sensor recordings. Further, detailed mathematical modeling that includes descriptions of chemical dynamics with mass transfer equations could prove broadly valuable in future neuroscience research.