Spatiotemporal trajectory analysis and validation of microglia activation in traumatic brain injury

Spatial transcriptomics (ST) is an innovative technology that holds tremendous potential for transforming the field of tissue biology research. By simultaneously capturing multiple types of spatial data, including gene expression values, spatial distance information, and tissue morphology, ST enables a comprehensive understanding of biological samples. However, the effective integration of these diverse data types remains a challenge. In this study, we present stLearn, a collection of three computational-statistical algorithms specifically designed to exploit the combined power of gene expression, spatial distance, and tissue morphology data. Our aim is to unlock novel insights into tissue maintenance, development, and disease. The first algorithm, known as pseudo-time-space (PSTS), employs a spatial-graph-based approach to uncover the spatial relationships between cells' transcriptional states in dynamic tissue contexts. To demonstrate the effectiveness of stLearn, we utilize traumatic brain injury datasets to investigate the spatio-temporal dynamics of microglia activation. By applying the PSTS algorithm to a well-established mouse model of acquired brain injury, we successfully reconstruct the spatial trajectory of microglia activation following insult, thereby validating this key component of stLearn. Overall design: We conducted an experimental study to investigate the spatio-temporal dynamics of microglia in the mouse brain following injury. The Visium ST data were acquired from the mouse brain at 3 days post-injury. To segment the brain regions, we employed stSME-based clustering, fine-tuning the clustering parameters using the Allen Mouse Brain Atlas. Next, we identified microglia-containing spots by examining the expression of marker genes, specifically Fcrls and Tmem119. To analyze the transcriptional dissimilarity between the injury site and different brain regions, we applied the PSeudo-Time-Space (PSTS) algorithm, which assigned a PTS score. Our analysis revealed that the hypothalamus region exhibited the greatest transcriptional dissimilarity compared to the injury site. To visualize the spatial trajectory of microglia activation, we utilized the PSTS algorithm to predict the minimum spanning tree connecting the damaged site and the hypothalamus. The resulting trajectory displayed the activation of microglia along the dorsoventral axis of the injured brain, with the directionality of transcriptional changes indicated by the arrows. Additionally, we compared the results with the mouse naive brain as a control to provide evidence for the spatio-temporal dynamics of microglia.