High-throughput sequencing for measuring transcription: What it can and cannot tell an ecologist

Background/Question/Methods

Historically, many organisms were genetically intractable.  Today, many of these same organisms can be investigated using Next Generation sequencing (NextGen), which does not require prior knowledge of an organism’s genome or the development of a full genetic map.  Data produced by NextGen can help address basic questions about population structure, migration, and evolutionary history. NextGen can also reveal the genetic basis of functional differences between populations or species through a technique known as RNAseq.  RNAseq techniques isolate RNA from the desired organism or tissue and then sequenced.  RNAseq simultaneously reveals both the genes expressed in a particular tissue and their RNA abundance.  The power and the limitations of this technique, however, are not well understood.  We use a combination of experiments and simulation to show the strengths and the weaknesses of RNAseq when applied to ecologically interesting organisms.  Specifically, we looked at patterns of gene expression differences between two species of spadefoot toad – Spea bombifrons and Spea multiplicata – and their hybrids.  These species are known to naturally hybridize in sympatry and we sought to identify which genes are exchanged and retained.  We followed this analysis with a series of simulations to estimate the power of detecting differences in expression among moderate to lowly expressed genes across a broad taxonomic range.

Results/Conclusions

RNAseq effectively captures a snapshot of the patterns of expression for highly expressed genes. We show that these data can be used to estimate the abundance and differential expression of transcripts and isoforms. Using standard bioinformatic techniques, we can then infer functional differences among the organisms. In Spea, for example, there is clear evidence of asymmetric gene exchange in sympatric populations that is biased toward several specific functional categories of genes. We also show that the power of these techniques is limited by the complexity of the RNA sample.  Specifically, the power of RNAseq to detect low to moderately expressed genes is limited. To address this problem, we develop an analytical method for estimating the power of a particular experiment and suggest several “good practices” for the design and implementation of an RNAseq experiment on non-model organisms.