The Norway spruce genome sequence and conifer genome evolution

We generated >1 billion RNA-Seq reads and used transcript assemblies of these in combination with public expressed sequence tags (ESTs) and transcripts to perform ab initio prediction of protein-coding genes, which identified a high confidence set of 28,354 loci with >70% coverage by supporting evidence from the total set of 70,968 predicted loci. A notable characteristic of the predicted gene structures was the presence of numerous long introns (Fig. 1b), with mean intron length being higher than in most available plant genomes, although similar to the repeat-rich genomes of Vitis vinifera and Zea mays17, 18. The longest intron in the high-confidence genes was 68kb (Supplementary Table 2.6), and 2,384 high-confidence genes contained 2,880longer than5-kb introns (20 of which we confirmed by PCR amplification; Supplementary Information 2.14), 2,679 of which contained a repeat, suggesting that repeat insertions account for intron expansion. By contrast, exon size was consistent among the species considered (Supplementary Information 2.6.3). Numerous genes (~30%) remained split across scaffolds owing to assembly fragmentation, and as such, the longest introns were not represented in the P.abies 1.0 assembly. Long introns (either individual or cumulative intron length) did not influence expression levels (Fig. 1c) and introns containing repeats have not contracted despite a lack of recent repeat activity (see below).

a, Gene family loss and gain in eight sequenced plant genomes (Arabidopsis thaliana, Populus trichocarpa, Vitis vinifera, Oryza sativa, Zea mays, Picea abies, Selaginella moellendorffii and Physcomitella patens). Gene families were identified using TribeMCL (inflation value 4), and the DOLLOP program from the PHYLIP package was used to determine the minimum gene set for ancestral nodes of the phylogenetic tree. We used plant genome annotations filtered to remove transposable elements. Orphans refers to gene families containing only a single gene. Blue numbers indicate the number of gene families. b, Boxplot representation of length distribution for the 10% longest introns in the same eight genomes. c, Scatter plots of cumulative intron length against log10 expression calculated as fragments per kilobase per million mapped reads (FPKM) for high-confidence gene loci (top, coloured orange) and green for lncRNA loci (middle, shaded green). The bottom panel shows a histogram of cumulative intron size in the two sets of loci. d, Distribution of small (1824-nucleotide (nt)) RNAs and their co-alignment-based colocation to genomic features (repeats, high-confidence genes and their promoter/UTRs). CDS, coding sequence.

Analysis of gene families in the high-confidence gene set and seven sequenced plant genomes (five angiosperms: Arabidopsis thaliana, Populus trichocarpa, Vitis vinifera, Oryza sativa and Zea mays, and two basal plants: Selaginella moellendorffii and Physcomitrella patens) identified 1,021 P. abies-specific gene families (Fig. 1a and Supplementary Information 2.8). P. abies-specific families included over-representation of Gene Ontology categories involved in DNA repair and methylation of DNA and chromatin (Supplementary Information 2.8). As for most draft genomes, these results probably overestimate gene numbers19 and will be refined as we further improve the genome assembly.

A common mechanism leading to genome size expansion is the occurrence of a whole genome duplication (WGD) event. We calculated the number of synonymous substitutions per synonymous site (Ks) of paralogues within the high-confidence genes but found no evidence for any recent WGD; there was a clear, exponential decay in the number of retained paralogues with increasing Ks values (Supplementary Information 2.9 and Supplementary Fig. 2.6). However, a population dynamics model that takes into account both small- and large-scale modes of gene duplication20 suggested the presence of a small peak (around Ks of 1.1), which, considering the slow substitution rate of conifers, might represent the ancient WGD predating the divergence of angiosperms and gymnosperms (350Myr ago21).

Previous examinations of small genomic subsets indicated that conifer genomes contain numerous pseudogenes5, 6, 22, 23. The gene-like fraction of the P.abies 1.0 assembly was identified by alignment of RNA-Seq reads and de novo assembled transcripts (Supplementary Information 2.10). Within this subset of the genome, loci with valid spliced alignments of de novo assembled transcripts or the presence of a high-confidence gene were also identified. The high-confidence gene set represented 27Mb of protein-coding sequence, whereas 72Mb of regions were identified with a valid spliced alignment or a high-confidence gene. In stark contrast, 524Mb of gene-like regions were identified by less stringent alignments. The presence of such a large gene-like fraction lacking predicted gene structures supports the presence of numerous pseudogenes.

Recent ENCODE publications24, 25 characterized numerous long non-coding RNA (lncRNA) loci in the human genome, but this class of RNA remains largely uncharacterized in plants. Using short-read de novo transcript assemblies, 13,031 spruce-specific and 9,686 conserved intergenic lncRNAs were identified (Supplementary Information 2.4.3). In common with the ENCODE results, P. abies lncRNA loci contained fewer exons, were shorter (Fig. 1c), and had more tissue-specific expression than protein-coding loci (Supplementary Fig. 2.8).

There has been conflicting evidence about the presence of 24-nucleotide short RNAs (sRNAs) in gymnosperms26, 27, 28, 29, a class of sRNA that silence transposable elements by the establishment of DNA methylation30. Across 22 samples, we identified numerous 24-nucleotide sRNAs, but these were highly specific to reproductive tissues, largely associated with repeats but present at substantially lower levels than in angiosperms (Fig. 1d and Supplementary Fig. 2.10). By contrast, 21-nucleotide sRNAs were associated with genes, repeats and promoters/untranslated regions (UTRs) (Fig. 1d). De novo microRNA (miRNA) prediction identified 2,719 loci, including 20 known miRNA families, with target sites predicted within the high-confidence gene set for 1,378 of these (Supplementary Information 2.13). Furthermore, 55 known miRNA families had >5 aligned sRNA reads and mature miRNAs, representing 49 known families aligned to the genome (Supplementary Information 2.13).

Read the original:
The Norway spruce genome sequence and conifer genome evolution