Transcriptional profiling predicts behavioral performance in experimental mouse model of traumatic brain injury (TBI)

Abstract

Traumatic brain injury (TBI) is a prevailing cause of disability and death. There are no curative therapies or robust predictors of long-term neurodegenerative disease. Treatment is challenging due to poor understanding of the cellular responses and mechanisms underlying TBI. We linked transcriptional profiling, cytokine profiling, and behavioral scoring in experimental TBI to discover new insights into immune-mediated mechanisms of TBI. Mice received ipsilateral moderate controlled cortical impaction. Luminex analyses showed upregulation of inflammatory cytokines in injured cortex, hippocampus, and spinal cord. Neurobehavioral evaluations were performed, then hippocampi isolated at 28d post-TBI for RNAseq. Pathway analyses of upregulated genes were functionally enriched for glial activation, proliferation, and neuronal development. Genes involved in synaptic transmission, glutamate receptor signaling, and neuronal migration were downregulated. Partial least squares (PLS) regression successfully modeled the relationship between differentially expressed genes and behavioral scores (R2=0.94). A core 10-gene signature predicted TBI severity and behavioral score. PLS discriminant analysis (PLSDA) classified TBI severity. VIP scores were used to define a 20-gene signature in TBI based on high VIP score, high fold change, and contribution to biological function. A PLSDA model using only these 20 genes was cross-validated by bootstrap method (AUROC=0.93), indicating that these genes may identify the critical TBI-altered processes. Altogether, our analyses revealed TBI-related biological pathways, potential therapeutic targets, and gene signatures as biomarkers for TBI severity and behavioral outcomes.