Optimization of RNA-Sequencing Analysis and a Role for the Epidermis in Sensation

Abstract

Our skin is our largest organ, covering our entire bodies, and is the first point of contact for numerous external stimuli. Recent evidence in mammalian models suggests that epidermal cells that comprise our skin may be playing an active role in sensing and responding to these stimuli, and for this dissertation, I sought to understand how widely conserved this epidermal sensation is. To begin, I used transcriptome analysis of different cell types in the skin to identify how skin may be able to respond to different stimuli. However, RNA-Sequencing (RNA-Seq) is still a relatively new technique with disparate analysis pipelines in place, and so I first set out to optimize methods for RNA-Seq analysis, which I did over the course of multiple bioinformatics projects. I assayed the effects of varying parameters for quality-based trimming on gene expression estimates, identifying that short reads remaining after trimming are the main driver of bias in expression estimates. I quantified the performance of 495 different differential expression workflows, using the metrics of recall and precision to describe a performance tradeoff observed across workflows. With current sequencing costs still a limiting factor for many bulk RNA-Seq experiments, I compared the tradeoff of sample size versus read depth for 30 high performance differential expression workflows and found inflection points at which performance degraded substantially. Together, these three projects allow me to make three broad recommendations. First, aggressive quality-based trimming is detrimental to RNA-Seq expression estimates and therefore should be avoided. Second, the choice of analysis tools at each step of differential expression analysis will affect the results obtained and therefore tools chosen should be tailored to the specific question being asked. Finally, differential expression workflows perform better with larger sample sizes and higher read depths but when the two are constrained, obtaining sample sizes of at least six per group should be prioritized. With these methods in place, I contributed to RNA-Seq analyses for multiple collaborators, which are presented in this work in short vignettes. After describing these transcriptome methods optimizations and applications, I return to the initial topic of epidermal sensation. By using an invertebrate, Drosophila melanogaster, I show that the ability of epidermal cells to induce aversive behaviors and respond directly to mechanical stimuli is conserved across evolution. In fruit flies and in mammals, epidermal cells ensheathe the dendrites of somatosensory neurons and these epidermally-induced behaviors are mediated by canonical neuronal circuitry. Overall, I describe a novel function for invertebrate epidermal cells and show that D. melanogaster provides a fruitful system for investigating epidermal-neuronal interactions.