Intron transposition

The most commonly purported mechanism of intron gain is intron transposition [34]. Intron transposition has been proposed to occur when a spliced intron reverse splices into either its own mRNA or another mRNA at a previously intron-less position. This intron-containing mRNA is then reverse transcribed and the resulting intron-containing cDNA may then cause intron gain via three different routes: it may undergo “complete” or nearly complete recombination with its original genomic locus, in which recombination between the intron-containing cDNA and the original genomic locus occurs both in exonic and intronic regions (if the gene contains introns), thereby deleting any introns within the region of recombination and causing concurrent intron gain and loss; it may undergo “partial" recombination, in which recombination only occurs between the intron-containing cDNA and an exonic region, avoiding the deletion of introns; or it may retropose into the genome at a different locus and then transfer the gained intron via “partial” or “complete” recombination to the original locus. Partial recombination between the intron-containing cDNA and the original genomic locus is the most commonly discussed route of intron transposition and is illustrated in Figure 1a.

Intron transposition has frequently been favored among the seven proposed mechanisms of intron gain by researchers [34, 37], likely because a transposed intron would already harbor the donor, acceptor, and branch point splicing sequences necessary for splicing [49]. However, the molecular process(es) by which intron transposition may occur, if it occurs at all, are unknown. To an extent, the most crucial and nebulous step in this process, reverse splicing, has been shown to occur [50]. However, this has only been demonstrated under non-physiological conditions using a mutant of Prp22 that fails to release the mRNA from the spliceosome. Furthermore, because a mutant was used that failed to release the mRNA, this process has only been shown to reverse splice introns back into their original position [50]. The process by which a previously spliced intron and spliceosome may bind (or remain bound), recognize an mRNA (based upon sequence motifs, protein complexes loaded onto the mRNA or random interactions), and begin nucleophilic attack and subsequent reverse splicing, is not known. Given that a recent study identified 95 introns with homologous sequences, of which only 11-24% may be novel introns [44], it is possible that interactions between proteins recruited to and deposited on the mRNA during and after splicing [51–53] may recruit spliceosomes “loaded” with previously spliced introns. Thus, reverse splicing may occur frequently, if not entirely, at positions in the mRNA that have already undergone splicing (i.e. only pre-existing intronic positions). This potential preference, if true, may help explain the finding of supposed parallel intron gains [27] if, following an initial intron gain, a different intron is transposed into this intronic position and both intronic sequences are maintained in the species. Alternatively, it is possible that reverse splicing occurs randomly at any position in an mRNA; however, the finding that the majority of suspected reverse splicing events occurred at previously intronic positions [44] argues against this explanation.

Even if intron transposition does occur, it has been suggested that any mechanism of intron gain that relies upon reverse transcriptase (RT) may not be a prevalent mechanism of intron gain. One potential problem is known as the “rate paradox” [34]. This may occur because the presumed most prevalent mechanism of intron loss, RTMIL [35], occurs via a process nearly identical to intron transposition, but does not require reverse splicing. Thus, the difference between the genome-wide rate of intron gain via intron transposition and intron loss via RTMIL should equal the rate of reverse splicing. However, as pointed out by Roy and Irimia, reverse splicing is believed to be an extremely rare process, as no reverse spliced introns have been found in any EST or cDNA sequences [34]. Furthermore, a recent statistical analysis has shown that it is unlikely that RT played a prominent role in intron gain throughout evolution [35].

Transposon insertion

Transposon insertions into genic regions are generally viewed as deleterious mutations. However, one underappreciated outcome of this process may be intron creation. Such an insertion may completely intronize the transposon without disrupting the coding sequence when a transposon inserts into the sequence AGGT, resulting in the duplication of this sequence on each side of the transposon (illustrated in Figure 1b). Three lines of evidence support the generation of introns via this model. One is that the sequence which the transposon inserts into, AGGT (also known as a protosplice site [63]), is believed to be a site of preferential intron gain [63–66]. Another is that the donor and acceptor splice sites created by this insertion, specifically the donor site AG|GT and the acceptor site AG|G (where “|” specifies splice junctions), adhere to the consensus donor and acceptor splice sites found in many organisms [67–71]. Finally, if these potential donor and acceptor splice sites are utilized efficiently by the spliceosome, any transposon may insert into the sequence AGGT in any gene without altering the genes coding sequence (demonstrated in Figure 1b). Alternatively, the transposon itself may harbor strong donor and acceptor splice sites near its boundaries or activate nearby latent splice sites, enabling its precise, or nearly precise, excision by the spliceosome. Indirect evidence of intron gain via transposon insertion has existed for nearly 20 years (see Table 1). However, what is not understood is exactly why these elements are spliced (if not by pure chance), or if any transposons preferentially cause intron gain, potentially due to target site preferences and/or the sequence of the transposon itself.

Tandem genomic duplication

Although widely underappreciated, the tandem genomic duplication of an exonic segment has recently emerged as a genuine and potentially prevalent mechanism of intron gain. Due to the similarity between consensus donor and acceptor splice sites, both of which closely resemble the sequence AGGT, the tandem genomic duplication of an exonic segment harboring the sequence AGGT generates two strong potential splice sites. If these splice sites are utilized by the spliceosome, the sequence between the original and duplicated AGGT will be spliced, affording the “precise” generation of an intron, i.e. the creation of an intron without alteration of the coding sequence of the gene (illustrated in Figure 1c). As discussed in the Transposon Insertion section above, use of AGGT as both the donor and acceptor splice site conforms to the most common splice site consensus sequences found in a variety of organisms [67–71], and AGGT has been found to be a site of preferential intron gain [64, 65]. In contrast to precise intron gain, in which the coding sequence of the gene remains unaltered, tandem genomic duplication may also result in “imprecise” intron gain, in which the coding sequence is altered. This may occur when latent splice sites within the duplicated region are utilized, resulting in the addition of nucleotides to the coding sequence of the gene. Alternatively, the duplication may result in the activation of latent splice sites near the duplicated sequence, resulting in the removal of nucleotides from the coding sequence.

Unlike any other proposed mechanism of intron gain, in vivo evidence exists to support the occurrence of this mechanism. The ability of this mechanism to have produced a novel intron nearly 500 million years ago in the ancestor of jawed vertebrates was recently tested and verified in vivo[56]. Not only has this process been shown to be a feasible mechanism of intron creation in vivo, but a plethora of indirect evidence in support of this mechanism has also been identified in a number of eukaryotes (Table 1). It should be noted, however, that many of these intron gains were imprecise and resulted from the activation of latent splice sites within the duplicated segment [57].

Intron Gain during Double-Strand Break Repair (DSBR)

Double-strand breaks (DSBs) are genomic lesions in which nearby phosphodiester bonds are severed on both strands of the DNA double helix. Such lesions may be caused by ionizing radiation, reactive oxygen species, or cellular processes. Because the repair of these lesions is paramount to cell viability, organisms have evolved two ubiquitous, genetically distinct, well conserved processes to repair these breaks: homologous recombination and non-homologous end joining (NHEJ) [72]. The former results in the precise repair of the break and requires an undamaged template, such as a sister chromatid. Thus, homologous recombination occurs primarily during the S and G2 portions of the cell cycle [73]. NHEJ, on the other hand, requires no template. If the DNA ends are undamaged and complementary, NHEJ faithfully restores the break point junction. However, if the DNA ends are damaged or are not complementary, NHEJ may insert or delete nucleotides from the break point junction [45, 46]. DSBR by NHEJ was recently implicated in intron gain when researchers identified short direct repeats flanking 43% of gained introns in Daphnia[27]. These repeats suggest that these introns were gained by the insertion of nucleotides during the repair of staggered DSBs by NHEJ [27] (illustrated in Figure 1d). Because NHEJ has been shown to preferentially insert mitochondrial DNA [74, 75], further support for this model was garnered when the authors identified a gained intron that was homologous to the mitochondrial 16S ribosomal subunit [27]. How or why NHEJ preferentially uses mitochondrial DNA [74, 75], why these sequences appear to integrate preferentially into genic regions [58], and why these sequences may function as introns, if not purely by chance, are topics that await further investigation.

Currently, indirect evidence for this proposed mechanism consists of gained introns with high similarity to mitochondrial DNA, rather than nuclear DNA (Table 1). In an attempt to determine the frequency by which introns without similarity to mitochondrial DNA may be gained via NHEJ, researchers have also quantified the number of repeats found flanking gained introns. In comparison to conserved introns, some studies have found gained introns to be enriched for repeats near their splice junctions [22], while others have not [23, 44]. The current incongruence of gained introns to be preferentially flanked by repeats may be the result of a number of factors, such as differences in NHEJ among species [30, 76–78] and cell types [79], insufficient sample sizes, or repeats and/or introns being inserted via other mechanisms. Alternatively, sequences of DNA inserted by NHEJ that are long enough to potentially form introns [80] may not frequently be flanked by direct repeats (see, for instance, inserts >30 bp long in [58, 75, 81, 82], however, also see one insert in [83]). Finally, it should be noted that other studies have identified a number of gained introns that are flanked by repeats [27, 28]. However, these numbers must be compared to the number of conserved introns flanked by repeats. This ensures that the number of repeats found flanking gained introns is significantly higher than the background level of repeats found to naturally flank introns. For instance, in Drosophila we found that 25% of gained introns were flanked by direct repeats ≥ 5 bp; however, 26% of conserved introns were flanked by repeats of the same size [23], indicating that while many introns are flanked by repeats in Drosophila, there is no bias for gained introns to be flanked by repeats in Drosophila.

Insertion of a Group II intron

Group II introns are self-splicing introns found in bacterial genomes and the organellular genomes of many eukaryotes. A number of remarkable similarities exist between group II introns and spliceosomal introns, such as their method of excision from a primary transcript, the sequence of their 5’ and 3’ ends, and structural similarities [84]. Even as our understanding of introns lay in its infancy, these similarities gave rise to the idea that spliceosomal introns and group II introns may be evolutionarily related [85]. Subsequent investigations have shown that it is likely that, following endosymbiosis between an α-proteobacteria and its archaeal host, many group II introns were transferred from the genome of the α-proteobacteria to the archaeal genome. The invasion and subsequent degradation of these elements may have imposed selective pressures which eventually gave rise to various hallmarks of eukaryotic cells, such as the nucleus [86, 87], nonsense-mediated decay [60, 88], and spliceosomal introns themselves [1, 89, 90].

In light of the findings that group II introns were likely the progenitors of spliceosomal introns, the retrohoming, or insertion, of a group II intron into a nuclear gene was proposed to cause recent spliceosomal intron gain [38] (Figure 1e). While it is widely believed that group II introns originally gave rise to spliceosomal introns, a recent in vivo assay demonstrated that the insertion of a group II intron into a nuclear gene nearly abolishes gene expression [60]. Thus, the evolution of nucleus-cytosol compartmentalization and nonsense-mediated decay following the initial invasion of group II introns (and subsequent creation of spliceosomal introns) may now impede intron gain via this mechanism. In accordance with these findings, indirect evidence of a recent intron gain via the insertion of group II intron has never been identified (Table 1). Therefore, it is unlikely that the insertion of a group II intron into a nuclear gene is a mechanism of recent spliceosomal intron gain. It is interesting to note, however, that the insertion of a group I intron into a nuclear gene does not appear to affect gene expression [60].

Intron transfer

Intron transfer has been hypothesized to result in intron gain when a paralog or pseudogene gains an intron and then transfers this intron via recombination to an intron-absent location in its sister paralog (illustrated in Figure 1f) [31]. Although indirect evidence has been found to support intron transfer (Table 1), this mechanism does not explain how the initial intron was gained, but rather explains how once gained an intron may propagate to other paralogs. It is possible that initially an imprecise intron gain occurs in a paralog or pseudogene and, following suppressor mutation(s), this intron is transferred via recombination to another paralog. Such a mechanism may allow introns to be initially gained imprecisely in a duplicated, unessential copy of a gene. Then, following suppressor mutation(s) in this neutrally evolving, extra intron-containing duplicate gene, recombination between the intron-containing region in the duplicated gene and the functional gene may result in precise or near precise intron gain in the essential copy of the gene. This process would allow an initially deleterious intron gain event in one paralog to result in non-deleterious intron gain in both paralogs without imposing a negative fitness cost to the host. If true, this suggests that intron gain rates should be higher in paralogous genes than genes without paralogs. This prediction has been confirmed in a broad range of species [15, 91, 92].

Intronization

Intronization is the process by which mutations create novel introns from formerly exonic sequence. Thus, unlike other proposed mechanisms of intron gain, this mechanism does not require the insertion or generation of DNA to create a novel intron. In the most commonly discussed route of intronization, mutations, generally point mutations, forge novel splice sites from exonic sequence [41, 42], resulting in the formation of a novel intron (Figure 1g). An alternative route of intronization has been proposed to occur when, in an exonic segment flanked by latent splice sites, a premature termination codon is generated via mutations [93]. The spliceosome may then act upon these latent splice sites in a mysterious process known as nonsense-associated altered splicing [94–97], removing the premature termination codon from the transcript and saving it from destruction via nonsense-mediated decay. This differentially spliced product (i.e. lacking the premature termination codon) has then been proposed to persist until subsequent mutations facilitate efficient utilization of these latent splice sites by the spliceosome, resulting in the formation of a novel intron [93].

Only the most commonly discussed route of intronization, by which mutations forge novel splice sites, has garnered indirect evidence (Table 1). This may be because the alternative route of intronization is much more difficult to detect. Alternatively, this route may not occur. Regardless, nonsense-mediated decay may play some role in enabling intron gain, potentially by facilitating the persistence of an initially infrequently spliced gained intron that harbors a premature termination codon, as a recent study in Drosophila found that novel introns were enriched for in-frame stop codons [28], although it should be noted that similar results were not found in Aspergillus[22]. Furthermore, it is possible that the point mutations that generate potential splice sites during intronization may, at least initially, be utilized infrequently. Thus, the ability to detect intronization events relies heavily upon accurate gene annotation. Additionally, as these point mutations may not initially generate strong splicing signals, alternative splicing may occur using these or other latent splice sites, resulting in transient bouts of intronization. Such a process may confound both the process of genome annotation and analyses aimed at identifying intron gains and losses. Therefore, identification of many of these events has been limited to species with deep transcriptome profiling [41, 42] or in analyses of recently created genes [61, 62]. Indirect evidence of intronization may exist in fungi [22] as well as other species, but has yet to be identified. Further analyses armed with extensive RNA-seq data may shed further light on the prevalence of intronization in other species.