Objective

Due to pervasive transcription and pervasive translation in these yeast, new transcripts and ORFs can quickly appear in non-genic sequences and become exposed to selection. This process, known as de novo gene birth, can lead to the appearance of new genes with entirely novel functions. Our objective was to identify and characterize putative de novo genes in baker’s yeast to further understand the phenomenon of de novo gene birth. To correctly classify putative de novo genes via the taxonomic conservation of these unique sequences, we need comparable data for a set of closely related species. Due to the similarity of molecular pathways to more complex eukaryotes coupled with their ease of growth in the lab, budding yeasts have proved to be a popular group of organisms for experiments ranging from experimental evolution to genetic engineering. We selected these 11 species based on their sparse taxonomic distribution, their amenability to growth in a custom rich media, the availability of genome assemblies, and their inclusion in previous studies of de novo genes in yeast. We have used novel transcripts assembled from our RNA-Seq data, taken together with the reference annotations, to generate a more complete transcriptome for each of the eleven species surveyed. We have estimated the time that each S. cerevisiae transcript originated in the yeast phylogeny using homology searches and genomic synteny [1]. As organisms modify their expression and translation of genes in response to stress, we sequenced the transcriptomes of all 11 species of yeast in both rich media and oxidative stress conditions to capture potential transcriptome variability.

The availability of complete gene annotations is key for genome-wide studies. The transcript assemblies provided contain hundreds of transcripts that were not present in the available annotations, and thus provide a more complete view of the gene content of each organism than previous annotations. These transcriptomes can be used as a basis to discover new encoded proteins, to study the evolution of yeast gene families and to investigate the changes in gene expression across different Saccharomycotina species. The addition of the ERCC Spike-into all samples also allows for the benchmarking of different de novo transcriptome assembly protocols.

Data description

We grew 11 species of yeast in two conditions:

  1. 1.

    Rich medium The yeast were grown in 20 mL of a custom rich medium [2], which was shown to accommodate various species of yeast, in 50 mL Erlenmeyer flasks at 30 °C. Cells were harvested in log growth phase at an OD600 of approximately 0.25.

  2. 2.

    Oxidative stress The same isogenic populations of yeast were grown in parallel, identical to the first condition. However, 30 min prior to harvesting the cells, hydrogen peroxide was added to a final concentration of 1.5 mM; we used a time period of 30 min to maximize the cellular response to stress [3], and a concentration of 1.5 mM H2O2 as we observed the yeast to grow approximately twice as slowly at this concentration.

After extraction, purification, and polyA selection of the RNA, synthetic spike-in transcripts from the ERCC RNA Spike-in kit [4] were added to each sample in order to assess the performance and limitations of our pipeline. After library preparation, the libraries were pooled into two batches (normal/stress) and sequenced in one lane on the Illumina HiSeq 2500 (paired-end, stranded, 50 bp long). This generated > 20 million high-quality strand-specific read pairs per sample (Table 1).

Table 1 Overview of data files
Full size table

After taking some quality control measures with our raw RNA-Seq data, we mapped the reads to their respective genomes (Table 1) and assembled de novo transcriptomes using the program Trinity version 2.1.0 [5]. We created a non-redundant set of features from the reference annotations combined with our de novo assembled transcripts; de novo assembled transcripts which correspond to annotated features according to Cuffmerge version 2.2.0 [6] were discarded, while those that did not were considered to be novel; we identified an average of 700 novel transcripts per species [1] (Table 1). The majority of these novel transcripts were found to be expressed in both conditions, but dozens of transcripts were only expressed in one condition or the other. Using the ERCC RNA Spike-in [4], we calculated that the lower limit of detection for annotated features in our pipeline was 2 TPM, and the lower limit of expression necessary to reliably assemble novel transcripts was 15 TPM; over half of the unannotated transcripts that we assembled were expressed above this conservative threshold of 15 TPM in at least one of the two conditions.

Limitations

A limitation of this dataset is that there are not multiple replicates for each species/condition, except for L. waltii, which has two replicates in each condition. We also would like to acknowledge that the concentration of hydrogen peroxide we used to induce an oxidative stress response (1.5 mM) was higher than the concentration used in other studies of oxidative stress response in yeast (0.1–1 mM).