Declining RNA integrity in control autopsy brain tissue is robustly and asymmetrically associated with selective neuronal mRNA signal loss

Abstract

RNA integrity numbers (RINs) are a standardized method for semi-quantification of RNA degradation, and are used in quality control prior to transcriptional profiling analysis. Recent work has demonstrated that RINs are associated with downstream transcriptional profiling, and correction procedures are typically employed in bioinformatic analysis pipelines to attempt to control for RINs influence on gene expression. However, relatively little work has been done to determine whether RINs influence is random, or is specifically targeted to a subset of mRNAs. We tested the hypothesis that RIN would be associated with a robust transcriptional profile seen across multiple studies. To test this, we downloaded subsets of raw transcriptional data from six published studies. We only included control, non-pathological post-mortem human brain tissue (n = 383 samples) in which independent subjects RIN values were also reported. A robust set of mRNAs consistently and significantly correlated with RIN across multiple studies, appearing to be selectively degraded as RIN declines. Many of the affected gene expression pathways are related to neurons (e.g., vesicle, mRNA transport, synapse, and mitochondria), suggesting that neuronal synaptic mRNA may be particularly vulnerable to degradation. Subsequent analysis of the relationship between RIN and vulnerable mRNA expression revealed most of the decay occurred over a relatively narrow RIN range of 7.2-8.6, with RIN values > 8.6 showing a ceiling effect, and those < 7.2 showing a floor effect on gene expression. Our data suggests that the RIN effect is pathway selective and non-linear, which may be an important consideration for current bioinformatic RIN correcting procedures, particularly in datasets in which declining RIN is confounded with a pathology under study (e.g., in Alzheimers disease).