A Novel Approach to Comparative RNA-Seq Does Not Support a Conserved Set of Orthologs Underlying Animal Regeneration

Abstract

Molecular studies of animal regeneration typically focus on conserved genes and signaling pathways that underlie morphogenesis. To date, a holistic analysis of gene expression across animals has not been attempted, as it presents a suite of problems related to differences in experimental design and gene homology. By combining orthology analyses with a novel statistical method for testing gene enrichment across large data sets, we are able to test whether tissue regeneration across animals shares transcriptional regulation. We applied this method to a meta-analysis of six publicly available RNA-Seq data sets from diverse examples of animal regeneration. We recovered 160 conserved orthologous gene clusters, which are enriched in structural genes as opposed to those regulating morphogenesis. A breakdown of gene presence/absence provides limited support for the conservation of pathways typically implicated in regeneration, such as Wnt signaling and cell pluripotency pathways. Such pathways are only conserved if we permit large amounts of paralog switching through evolution. Overall, our analysis does not support the hypothesis that a shared set of ancestral genes underlie regeneration mechanisms in animals. After applying the same method to heat shock studies and getting similar results, we raise broader questions about the ability of comparative RNA-Seq to reveal conserved gene pathways across deep evolutionary relationships.

Keywords: RNA-Seq; evolution; regeneration.

Fig. 1.

Cases of animal regeneration included in this study. a) The six animals analyzed in this paper, organized by their evolutionary relationships. The region of each organism undergoing regeneration is highlighted in red and is described underneath the image of each animal. The RNA-Seq sampling regime from each study is visualized with a bar; each time point that was sampled is represented by a notch in that bar. Despite the different absolute time ranges, the studies are comparable in that the time points span the early key stages of regeneration: starting with wound healing (red) and transitioning into blastema formation/cell proliferation (blue). b) A simplified overview of the methodology used to define deCOGS. A more detailed version is provided in supplementary fig. S2, Supplementary Material online.

Fig. 2.

An UpSet plot demonstrating the number of overlapping deCOGs shared across all six data sets. This plot focuses on overlaps of four or more of the six data sets. The number of deCOGs common across all six cases (160) is highlighted in orange. Additional deCOGs that are recovered when individual case studies are removed are highlighted in blue. The data used to generate this figure are provided in supplementary table S2, Supplementary Material online.

Fig. 3.

Jaccard distance matrices based on the presence/absence of COGs across taxa. a) Matrix derived from all COGs as assigned by OrthoFinder. b) The same analysis, but restricted to deCOGs The data used to generate this figure are provided in supplementary table S3, Supplementary Material online.

Fig. 4.

Evolutionary (phyletic) origin of deCOGs. The total number of deCOGs recovered at each node of the evolutionary tree is indicated by a bar chart to the right. Novel deCOGs at each node are broken down by their phyletic origin; for example, deCOGs that are a “bilaterian novelty” contain genes that have no significant sequence similarity to genes outside of the Bilateria. The data used to generate this figure are provided in supplementary table S4, Supplementary Material online.

Fig. 5.

The presence of Wnt genes in the six RNA-Seq data sets analyzed (produced by OrthoFinder). Wnt genes were recovered as a single deCOG in our analysis, which can be subdivided into a minimum of 13 previously described subfamilies. The presence/absence of these subfamilies in each taxon is demonstrated by silhouettes. Gray silhouettes show the subfamily is present in the organism's transcriptome; black silhouettes show that the subfamily is present and differentially expressed in the relevant RNA-Seq study. Note that no subfamily is present and differentially expressed across all taxa. The data used to generate this figure are provided in supplementary table S5, Supplementary Material online.

Fig. 6.

The presence of deCOGs within the stem cell pluripotency network. The network has been reproduced and simplified from KEGG pathway 04550. The color of each box indicates the number of data sets with one or more differentially expressed genes within the relevant COG. Red arrows indicate pathways that are specific to “primed” stem cells (e.g. human embryonic stem cells, human-induced pluripotent stem cells, and mouse epiblast–derived stem cells); gray arrows indicate pathways also found in “naïve” stem cells (e.g. mouse embryonic stem cells and mouse-induced pluripotent stem cells). The data used to generate this figure are provided in supplementary table S6, Supplementary Material online.