NanoString Technologies Neuropathology Panel Produces Unreliable Measurements of Mouse Hippocampal Gene Expression

Technologies for measuring gene expression (i.e., the number of RNA transcripts) of large numbers of genes simultaneously in specific tissues have exploded in recent years. Current methods include high-throughput RNA sequencing (RNA-seq), transcript counting platforms like NanoStrings nCounter(R), and spatially resolved techniques based on fluorescent in situ hybridization (FISH). Several studies have evaluated the reliability of these different methods and performance in comparison to one another. Typically, technical reliability, as measured by Pearsons correlation of two measurements of the same sample, is usually well above 90%, and is statistically significant even for small sample sizes (e.g., 8 samples measured twice). We performed an experiment where we aimed to compare hippocampal gene expression between 3 groups (n=5 per group) of young adult male C57BL/6J mice. Before sampling, the groups were treated with either repeated injections of PBS (vehicle), extracellular vesicles taken from the blood plasma of sedentary mice (SedVs) or exercising mice (ExerVs). The hippocampus was dissected, and RNA purified using standard methods. The samples were analyzed using the NanoString Neuropathology panel, that measures 770 genes simultaneously. To estimate reliability, we measured 8 of the samples twice in two separate assays. Surprisingly, only 85 genes showed a significant Pearsons correlation (p<0.05), and none of these met false discovery significance (all q<0.05). To confirm that no errors were made transferring labels, the individual samples were permuted to see whether a different assignment could recover a greater number of positive correlations. Results showed that the original assignment was best suggesting no errors in sample assignments were made. We conclude that the Nanostring neuropathology panel produces unreliable data for mouse hippocampal gene expression.