Altogether, prediction of ES incidences from ancestral reconstructions and comparison with extant species are consistent with the increase in ES frequency observed in metazoans, particularly from bilaterian ancestors.

Then, we used this predictive framework to analyze ES transitions in animals by comparing the incidence of ES in reconstructed premetazoan and postmetazoan ancestors (Fig. 7c). To do so, we used previously published estimates of intron density in ancestral genomes obtained by comparative genomic analyses [26, 27], which reported an increase in intron density from 7.85 to 12.37 introns/gene between the origin of Holozoa (Urholozoa) and animals (Urmetazoa) (Table 1). As a proxy for ancestral intron lengths, we used means obtained from extant species’ phylogenetically independent contrasts [60] (see “Methods”; Additional file 1: Figure S24). Under these assumptions, the Urmetazoa (12.37 introns/gene of ~ 850 bp mean length) would have an IES ~ 0.30 (Fig. 7c). Conversely, the intron-poorer Urholozoa would reach IES ~ 0.30 only if it had an average intron of > 1000 bp, which exceeds by an order of magnitude the average intron lengths of extant unicellular holozoans (120–571 bp; Additional file 1: Figure S2) [27] and our ancestral estimation (~ 481 bp). Therefore, under reasonable assumptions for ancestral genome architecture, the unicellular ancestors of animals likely had lower ES levels than the Urmetazoa (Fig. 7c, d).

Thus, relative expression of core spliceosomal components seems adjusted to the number of introns to be spliced in each species.

It has been proposed that, in species with extremely long introns (e.g., mammals and G. gallus), a differential in GC content between exons (GC-richer) and flanking introns (AT-richer) can act as a compositional mark to assist the recognition of splice sites

Interestingly, we observed similar patterns in the ichthyosporean S. arctica (with higher ES rates than other holozoans; Fig. 2e) and the chlorophyte V. carteri (also with higher ES rates than other chlorophytes; Fig. 2f). Their ES-positive exons had more heterogeneous splice sites than ES-negative ones, were shorter, and had higher intron-to-exon length ratios. Furthermore, the majority of S. arctica and V. carteri introns were longer than their cognates in their close unicellular holozoan (Fig. 4e) and chlorophyte relatives [38] (Fig. 4f), respectively, suggesting recent intron lengthening events. Moreover, both S. arctica and V. carteri have relatively high intron densities (3.22 and 3.88 introns/CDS kbp; Fig. 1b, c and Additional file 1: Figure S2) that derive from recent, lineage-specific intron gain processes at the root of ichthyophonid Ichthyosporea [27] and Chlorophyceae plus Trebouxiophyceae [26] (Table 1). As the median CDS length is relatively constant across eukaryotes (~ 1400 bp [57]) and is independent of intron content, higher intron densities at the species level usually imply the presence of shorter exons (Fig. 1c, Additional file 1: Figure S19). Thus, V. carteri and S. arctica seem to have independently acquired ES-conducive genome architectures (higher intron densities, shorter exons flanked by longer introns) that contribute to their ES-richer AS profiles when compared to their closest relatives (Fig. 2e, f). On the other hand, the most notable exceptions to these patterns were the multicellular phaeophyte E. siliculosus (which exhibited low ES frequencies despite having unusually long introns; Fig. 4c), the charophyte Klebsormidium netis, and B. natans. In these species, ES was associated with short exons, but, unusually, also with short introns (Fig. 4a).

we investigated the effect of global length of genes and transcripts; the length of the alternatively spliced exons and its flanking introns (for ES) or vice versa (for IR); intronic and exonic GC content and the differential in GC content between introns and exons; the strength of the 5′ and 3′ splice site definition; the intron density (introns per gene and base pairs of introns per base pairs of coding sequence [CDS]); the relative position of the AS event along the gene (from the start codon); and the mean transcript expression level (using cRPKMs from pooled RNA-seq experiments).

Previous studies have linked the level of ES and IR within a genome to differences in traits such as the length of exons and introns, intron density, sequence composition, splicing site homogeneity, or other cis signals [16, 50,51,52,53].

we divided exons or introns between those that had lengths divisible by three (henceforth ‘3n’) and those that did not,

Bigelowiella natans

Therefore, these data show that (i) bilaterians, and vertebrates in particular, have a consistently higher ES frequency than their close relatives and other eukaryotes, and (ii) some non-bilaterian animals and unicellular holozoans have experienced relative increases in ES frequency as well.

On the other hand, the ichthyosporean Sphaeroforma arctica had the highest FES,sp among unicellular holozoans (Fig. 2e; p < 1e − 09, oKS), which indicates a clear lineage-specific departure from the low incidence of ES in other unicellular holozoans

This indicates that our approach for ES quantification yields robust results independently of the experimental approaches used in different RNA-seq experiments.

performed de novo RNA-seq for three phylogenetically key species: the placozoan Trichoplax adhaerens, the holozoan Sphaeroforma arctica, and the intron-rich excavate Naegleria gruberi.

we track the frequency of both main modes of AS (IR and ES) across all major eukaryotic lineages

Third, relative increases in ES frequency have also been identified in other phylogenetically scattered eukaryotes—e.g. in plants, Volvox carteri, or the chlorarachniophyte Bigelowiella natans [10, 35,36,37,38].

scarce

ctenophores

placozoans

poriferans

the earliest eukaryotes already exhibited splicing-rich transcriptomes yielding multiple mRNA variants per gene

last eukaryotic common ancestor (LECA)

Consistently, and although the extent to which ES transcripts are translated and functional is still under debate [19, 20], many ES-derived isoforms have been found to be physiologically relevant in animals (reviewed in [8, 21]), for example, by tuning protein–protein interaction networks [22,23,24]. In contrast, IR events have been linked to down-regulation of gene expression via the nonsense-mediated decay (NMD) pathway [4,5,6, 25], nuclear retention [7] or intron detention [3] in a wide variety of eukaryotes.

First, exon skipping in early animals became enriched for frame-preserving events. Second, bilaterian ancestors dramatically increased their exon skipping frequencies, likely driven by the interplay between a shift in their genome architectures towards more exon definition and recruitment of frame-preserving exon skipping events to functionally diversify their cell-specific proteomes.

indicating co-evolution between genome architectures and exon skipping frequencies

bilaterians

We used RNA-seq data to quantify exon skipping and intron retention frequencies across 65 eukaryotic species, with particular focus on early branching animals and unicellular holozoans.

The sequence of a full-length RE includes a portion encoding two proteins, GAG and POL

Neural expression of the eMIC domain is regulated transcriptionally through the expression of Srrm3 and Srrm4 in vertebrates, both containing the eMIC domain, but through post-transcriptional processing of Srrm234 in non-vertebrates

eMIC

Evolution of novel genes in three-spined stickleback populations

The avoidance of aggregation represents a critical force in protein evolution.

sequence differentiation (FST

the early stages of gene emergence are of particular interest and need to be further investigated at the level of populations.

we take advantage of the genomic and transcriptomic datasets available for three-spined sticklebacks to study the distribution of transcripts between populations.