(G-245) Machine Learning-Based Morphological Analysis of V2a Interneurons in the Spinal Cord

V2a interneurons play a crucial role in coordinating motor circuits within the spinal cord and have been recognized as the predominant interneuron type that mediates locomotor recovery after spinal cord injury. However, V2a interneurons represent a large cardinal class of interneurons with many distinct subtypes. The specific type(s) of V2a interneuron that helps restore function has yet to be identified. While biochemical identifiers of developmental patterns and transcriptional profiles have been recently used to further sub-categorize V2a interneurons, the morphological diversity of these interneurons remains unclear. Understanding the structure of V2a interneurons may help elucidate their role in functional recovery after SCI. To address this gap in knowledge, we developed a pipeline to investigate V2a interneuron morphologies in the tissue-cleared mouse spinal using two methods: (1) extracting structural features from sparsely labeled V2a interneuron light-sheet fluorescent microscopy (LSFM) images and (2) machine learning based clustering of structural-features to identify morphological sub-populations of V2a interneurons.

For this study, two adult mouse lines were used: Confetti x Chx10-Cre (Chx10::Confetti) and MORF3 x Chx10-Cre (Chx10::MORF3). Chx10::Confetti mouse spinal cords have ~25% of V2a interneurons labeled with a distinct fluorescent protein, while Chx10::MORF3 mouse spinal cords have a reporter protein expressed in ~1-5% of V2a interneurons. The spinal cords were cleared using a modified iDISCO+ protocol. The cleared spinal cords were then imaged using the Zeiss Z1 light-sheet microscope and stitched in Imaris. Structural features of the V2a interneuron cell bodies (somas), including volume and principal axis lengths, were extracted from the images using Cellpose and MATLAB. These features were input into a K-means clustering algorithm to group V2a interneurons that have similar structural features.

Automatic segmentation and feature extraction were successful for V2a interneuron somas in the Confetti mouse spinal cord. K-means clustering of the raw structural features of the V2a interneuron somas displays possible clusters related to the longest principal axis length. Further feature engineering is required to reveal more defined morphological clusters of these somas. Preliminary clearing and imaging of the MORF3 cords were also successful and individual V2a interneuron morphologies were readily observed. However, the labeling process and image analysis still need to be optimized for future studies, including enhancing the signal-to-noise ratio and improving the segmentation accuracy of the labeled cells.

While preliminary results are promising, further imaging and analysis of the Chx10::Confetti and Chx10::MORF3 mouse spinal cords is needed to determine if morphology alone is sufficient for clustering V2a interneurons into subgroups. Future studies will compare how the V2a interneuron morphological subgroups may alter following spinal cord injury and locomotor recovery.

Use of the Texas A&M Microscopy and Imaging Center is acknowledged. Funding was provided by Mission Connect, a program of TIRR Foundation.