(G-273) Building an Automated Program to Assess Hippocampal Activity After Chronic Pain

Waveform analysis of large nerve populations are important to understand how the brain processes information. Extracellular recordings of large nerve populations can be achieved with micro-electrode arrays (MEA) which produce large waveform data sets, however, reliable waveform analysis is computationally challenging. Single sessions of recordings can result in thousands of waveforms to be evaluated with features than can vary subtly or dramatically across recording sessions or experimental groups. For example, chronic pain conditions alter hyperexcitability¹ and neuroplasticity² (the brain’s ability to adapt) in the hippocampus. The long-term potentiation (LTP) experiments to explore these changes using an 8x8 MEA can generate thousands of waveforms for a single sample and tens of thousands of waveforms across an experiment. Characterizing these waveforms is time consuming, involving manually identifying features like waveform minima, maxima, slope, and timestamps based on semiquantitative assessments in proprietary software. Being proprietary, most commercial analysis software are inflexible, expensive, and difficult to import/export for further analysis. Furthermore, errors in proprietary analyses are difficult to identify and fix. As such, new automated open-source waveform analysis methods will enable faster, more flexible, and accessible tools to improve analytics from ever-growing neuroscience data sets.

This work produced a total of nine hippocampus MEA datasets in healthy and injured mice to develop an open-source, Python-based program capable of analyzing standard LTP waveform measures. In addition to immensely reducing the person-hours required to process and analyze MEA datasets, this software allows for greater control over analysis parameters to ensure full methodological transparency.

Nine adult female C57b/6 mice were deeply anesthetized (isoflurane) prior to euthanasia. Brains were dissected, chilled in Ca₂⁺ free artificial cerebrospinal fluid (ACSF), oxygenated (95%O₂/5%C0₂), coronally sectioned (350µm, Bregma -1.94 mm), then stabilized for 120 minutes in ACSF at 35ºC. Sections were immobilized on an 8x8 MEA (MED-A64MD1, Alpha Med Scientific) then a standard LTP protocol³ producing evoked field excitatory postsynaptic potentials (fEPSP) waveforms was performed, inducing potentiation with standard theta-burst stimulation⁴ (TBS). fEPSPs were collected every 40 seconds with three measurement phases: baseline (10 minutes pre-TBS), TBS (2 minutes following TBS), and 45 post-TBS (45 to 60 minutes following TBS) (Figures 1 & 2).

The automated analysis program (LTP.py) was written in Python using several common modules, including Pandas (pd), NumPy (np), and MatPlotLib (mpl). LTP.py imports a .csv of raw waveform data (timepoints and voltage) into a data frame, then injects it into a “Raw” class object along with the relevant experimental information. Based on global constants regarding instrument parameters such as sampling rate and trace duration, the program determines how many rows to parse for each voltage trace, creating a list of “Trace” class objects for each channel. Experimental phases and waveform measures are calculated as member variables of the Raw and Trace objects upon initialization. Manual calculations of each fEPSP waveform minimum and slope between 10-40% of the minimum were used as ground truth for the automated program. T-tests were used to determine differences between manual and automated analyses.

LTP.py reduced waveform analysis time from one hour to under one minute. LTP.py automated calculation of fEPSP minimum replicated manual calculations (p < 0.001, 1.3x10^-6 ± 3.7x10^-5% error). The average slope calculation had 0.72 ± 5.67% error. While this average is low, the individual trace errors in some cases exceed 20% error (Figure 3). More investigation is required to determine the cause of these slope calculation errors as traces with particularly high error ( >20%) still correctly identified the 10-40% interval. Ultimately, some discrepancies between the automated and manual method may never be explained as the manual method was performed using proprietary commercial MEA software that has not disclosed the precise method used to calculate waveform features.

This work presents a promising new opensource tool to automatically calculate fEPSP waveform features. fEPSPs are a foundational electrophysiological measure used by thousands of neuroscientists worldwide and present a significance analytical challenge. Improving the accessibility, transparency, and flexibility of automated analysis methods will empower how fEPSPs can be used in many fields to better understand brain function.

Special thanks to Angel-Rose L. Villegas and Dr. Heidi Kloefkorn for their guidance on this project, and the whole Kloefkorn Lab team for their continuing support. Thank you to Dr. Kathy Magnusson and Dr. Kenton Hokanson for providing consultation, training, and allowing us to use their facilities. We are also grateful to Oregon State University's College of Science for their SURE award, funding from which contributed to training and materials costs.

1. Guo, Hua, Yuqing Wang, Lihua Qiu, Xiaoqi Huang, Chengqi He, Junran Zhang, and Qiyong Gong. “Structural and Functional Abnormalities in Knee Osteoarthritis Pain Revealed With Multimodal Magnetic Resonance Imaging.” Frontiers in Human Neuroscience 15 (2021). https://www.frontiersin.org/articles/10.3389/fnhum.2021.783355.

2. Zhao, Xiao-Yan, Ming-Gang Liu, Dong-Liang Yuan, Yan Wang, Ying He, Dan-Dan Wang, Xue-Feng Chen, et al. “Nociception-Induced Spatial and Temporal Plasticity of Synaptic Connection and Function in the Hippocampal Formation of Rats: A Multi-Electrode Array Recording.” Molecular Pain 5 (September 22, 2009): 55. https://doi.org/10.1186/1744-8069-5-55.

3. Liu, Ming-Gang, Rui-Rui Wang, Xue-Feng Chen, Fu-Kang Zhang, Xiu-Yu Cui, and Jun Chen. “Differential Roles of ERK, JNK and P38 MAPK in Pain-Related Spatial and Temporal Enhancement of Synaptic Responses in the Hippocampal Formation of Rats: Multi-Electrode Array Recordings.” Brain Research 1382 (2011): 57–69. https://doi.org/10.1016/j.brainres.2011.01.076.

4. Larson, John, and Erin Munkácsy. “Theta-Burst LTP.” Brain Research 1621 (September 24, 2015): 38–50. https://doi.org/10.1016/j.brainres.2014.10.034.