Trade-off between Transcriptome Plasticity and Genome Evolution in Cephalopods

Abstract

RNA editing, a post-transcriptional process, allows the diversification of proteomes beyond the genomic blueprint; however it is infrequently used among animals for this purpose. Recent reports suggesting increased levels of RNA editing in squids thus raise the question of the nature and effects of these events. We here show that RNA editing is particularly common in behaviorally sophisticated coleoid cephalopods, with tens of thousands of evolutionarily conserved sites. Editing is enriched in the nervous system, affecting molecules pertinent for excitability and neuronal morphology. The genomic sequence flanking editing sites is highly conserved, suggesting that the process confers a selective advantage. Due to the large number of sites, the surrounding conservation greatly reduces the number of mutations and genomic polymorphisms in protein-coding regions. This trade-off between genome evolution and transcriptome plasticity highlights the importance of RNA recoding as a strategy for diversifying proteins, particularly those associated with neural function. PAPERCLIP.

Keywords: ADAR; Epitranscriptome; RNA editing; RNA modifications; cephalopods; genome evolution; neural plasticity; proteome diversity.

Figure 1. Extensive recoding is an invention of coleoid cephalopods

(A) The species studied span the cephalopod evolutionary tree, as well as sea hare (Aplysia californica) as an outgroup (top). For comparison, a representative tree for vertebrates is shown (bottom), constructed based on divergence times estimated in (Hedges et al., 2006). (B) Tens-thousands of A-to-I editing sites (identified as A-to-G DNA-RNA mismatches) are detected in squid, sepia and the two octopus species (see Tables S1–S4 for more details). The noise level (estimated by the number of G-to-A mismatches) is rather low. In contrast, in nautilus and sea hare no enrichment of A-to-G mismatches is observed (inset). (C) The nucleotides neighboring the detected editing sites, show a clear pattern consistent with known ADAR preference (Alon et al., 2015; Eggington et al., 2011; Kleinberger and Eisenberg, 2010) for the extensively recoded coleoid species – squid, sepia, and the two octopus species – but not in nautilus or sea hare. The motif is characterized by under-representation of G upstream to the editing site (relative location −1) and over-representation of G in the downstream base (The height of the entire stack of letters represents the information content in bits, the relative height of each letter represents its frequency).

Figure 2. Proteomic validation of recoding by RNA editing

We analyzed peptides identified by mass spectrometry analysis of two squid tissues, looking for evidence of recoding. For each site covered by one or more peptides, we marked whether the edited, non-edited or both versions of the peptide are observed. The distribution is presented, binned by the predicted RNA editing level (as measured from RNA-seq data). In parentheses are the numbers of recoding sites analyzed in each editing-level bin. The proteomic recoding level follows closely the predicted RNA editing level. Altogether, this experiment validated protein recoding in 432 sites in two tissues: (A) Squid stellate ganglion, where 320 of the 3,204 single-site peptides (10.0%) were shown to be edited. (B) Squid giant axon (giant fiber lobe), where 283 of the 2,741 single-site peptides (10.3%) were shown to be edited.

Figure 3. Editing in Octopus bimaculoides

(A) A-to-I editing sites were found within coding sequences of Octopus bimaculoides using three methods: the genome-free method (alignment to de-novo transcriptome), the genome-dependent approach using REDItools (Picardi and Pesole, 2013), and identification of hyper-edited reads (Porath et al., 2014). Overall, the three methods identified 170,825 unique AG sites in Octopus bimaculoides coding sequences (38,066 hyper-editing sites do not overlap those found by the other methods). See Methods for analysis of the differences between the results of the first two methods. (B) RNA editing levels, measured across the whole transcriptome (see Table S5) by the editing index (weighted average of editing levels over all editing sites identified in the transcriptome, see Methods). Levels vary across tissues and are highest for neural tissues (see Table S6). Unlike mammals, a sizable fraction of editing events (11–13% in neural tissues) results in recoding events. Annotation of transcripts and repeats is based on (Albertin et al., 2015). (CNS= central nervous system; ANC=Axial nerve cord; OL=Optic Lobe; Sub=Subesophageal ganglia; Supra=Supraesophageal ganglia; PSG=posterior salivary gland; ST15=stage 15 embryo) (C) The number of editing sites in coding region is comparable to the number found in introns. (D) Unlike the case in mammals, editing is not exceptionally enriched in specific repeat families in Octopus bimaculoides, as measured by the editing index (here defined as the editing level averaged over all, edited and unedited, adenosines in each specific repeat family). (E) Protocadherins is a gene family known to be principally expressed in the brain, important for mediating combinatorial complexity in neuronal connections and are thought to play a role in diversifying neural circuitry (Chen and Maniatis, 2013). It was impressively expanded in Octopus bimaculoides (Albertin et al., 2015). A large number of protocadherins are found in the assembled transcriptomes for the four coleoid species (127–251 open reading frames), but not in nautilus (28 open reading frames). (F–G) Protocadherins contain significantly higher numbers of AG sites (F) and are edited at higher levels (editing level summed over all sites and normalized by ORF length), in all four coleoid species but not in nautilus (G).

Figure 4. Extensive recoding is conserved across coleoid cephalopods

(A) Tens-thousands sites are conserved across species (see Table S7). The closer the species are evolutionarily, the higher the number of conserved sites. (B) Virtually all (97.5–99%) mismatches conserved across species are A-to-G, resulting from A-to-I editing. Manual inspection of the few non-A-to-G mismatches appearing in multiple species suggests that they either result from systematic erroneous alignments, or they are actually editing sites that were mistakenly identified as G-to-A mismatches due to insufficient DNA coverage. (C) The majority of editing sites is conserved between the two octopus species, and even the most distant species share a sizable fraction of their sites. (D) In contrast, only 36 human recoding sites (1–2% of human recoding sites) are shared by mouse, and a similar number is shared between Drosophila melanogaster and D. mojavensis (Yu et al., 2016) (diverged at later times than squid-sepia). (E) Interestingly, 1146 AG modification sites (in 443 proteins) are conserved and shared by all four coleoid cephalopod species. Of these, 887 are recoding sites and 705 are highly edited (>=10% editing) recoding sites (in 393 proteins). (F) Some proteins include multiple highly-edited recoding sites (see Table S8). Of note are Uromodulin, α Spectrin (previously reported to harbor the highest number of recoding sites in squid (Alon et al., 2015)), and Calcium-dependent secretion activator 1 (CAPS1) with 14, 8 and 7 strong shared recoding sites, respectively. Recoding in CAPS1 was found to be conserved in vertebrate species from human to zebrafish (Li et al., 2009). (G) Not only are the locations of editing sites conserved, but their editing levels are correlated as well. Editing levels in 887 recoding sites shared by all species are highly, positively and significantly correlated in all pairs of coleoid cephalopod species (p<1e–75 for all pairs; see Supp. Fig. 1 for three additional pairs). Correlation is higher the closer the species are to each other in evolutionary terms, with Pearson rho = 0.95 for the two octopus species.

Figure 5. Signs for positive selection of recoding by editing

(A) The fraction of recoding sites among all editing sites in coding region increases with editing levels (top), as well as the fraction of recoding sites among all conserved sites (bottom). Red horizontal dashed line represents the recoding fraction expected assuming neutrality. (B) Editing levels are higher in conserved recoding sites. Distributions of editing levels in four groups of putative A-to-I editing sites: recoding and conserved (Rec+, Cons+), recoding and non-conserved (Rec+, Cons−), conserved sites that cause a synonymous change (Rec−, Cons+), and non-conserved synonymous sites (Rec−, Cons−). Horizontal red lines mark the median level, and yellow diamonds mark the mean. Conservation and non-synonymity are both positively correlated with higher editing levels, as well as their interaction (ANOVA, p-value<1.0e–162). Data presented here for squid (conserved sites are conserved in sepia), but the results are similar and significant for all species. (C) In contrast with the case in humans, highly edited sites tend to be more conserved: the fraction of conserved sites rises with the editing level for all species pairs, but more dramatically for the closely related octopuses and the sepia-squid pair. (D) Highly conserved regions of the transcriptome are enriched in editing sites, further attesting for positive selection of RNA editing. Density of editing sites (number of AG sites normalized by length) is higher for 112 recoding regions that are highly-conserved across the four species (>95% identity; average length 1382bp), compared with all other, less conserved, regions (Wilcoxon p-value<0.001 for all species). Error bars represent the S.E.M.

Figure 6. Conserved and species-specific editing sites affect protein function

Unedited (wt) and singly-edited versions of the voltage-dependent K⁺ channels of the K_v2 subfamily were studied under voltage-clamp (see Table S9). (A) (i) Current traces resulting from a voltage step from −80 mV to 40 mV for the wt Sepia K_v2.1 and the same construct containing the sepia-specific I529V edit, lying within the 4^th transmembrane domain (green), showing that I529V accelerates the rate of slow inactivation. (ii) Time constants for slow inactivation determined by fitting single exponentials to traces similar to those in panel (i) at different activating voltages (Vm). (B) (i) Tail currents measured at a voltage (Vm) of −80mV, following an activating pulse of +20 mV for 25 ms. Traces are shown for the wt K_v2.1 channels from squid, sepia and Octopus vulgaris. (ii) Tail currents for the same channels edited at the shared I-to-V site in the 6^th transmembrane span, following the same voltage protocol. (iii) Time constants from single exponential fits to tail currents obtained at various negative voltages (Vm) (following an activating pulse to 20 mV for 25 ms) show that the unedited channels close at distinct rates, (iv) but the edited versions close at similar rates. N = 5 ± s.e.m. for all data plotted in this figure.

Figure 7. RNA editing slows down cephalopod genome evolution

(A) Inter-species mutations are purified from genome loci surrounding conserved recoding sites (data shown for sites shared by squid and sepia). Depletion of mutations extends up to ~100bp of shared recoding sites (left). As a control, we show the mutations density (mutations/bp) around random non-edited adenosines from the same transcripts (right). Yellow – synonymous change; light green – non-synonymous; dark green – deletions. (B) Genomic polymorphisms are depleted near editing/recoding/conserved-recoding sites in squid, attesting to reduced genome plasticity. Effect is stronger for recoding sites, and even more so for the conserved recoding sites. (C) GC-content is elevated near editing sites in squid, allowing for more stable double-stranded RNA structures. The effect is even stronger in conserved sites. Dashed line represents the baseline GC level in the entire ORFome, and error bars represent the S.E.M. See Supp. Fig. 3 for analyses similar to those presented in panels A–C in other species.