Retroshuffling the genomic deck

The shape of modern genomes reflects evolutionary forces that have operated on them for millions of years. According to the exon-shuffling hypothesis, which was first suggested in these pages over 20 years ago¹, new genes are assembled from chunks of old ones. But the molecular mechanisms by which these exons (protein-coding sequences) could be glued together remained speculative. A report in Science by Moran and colleagues² now suggests an unexpected and efficient source of such genomic reshufflings. The authors looked at the abundant L1 (LINE-1) retrotransposons — transposable elements that can replicate within the mammalian genome using reverse transcriptase, which copies the retrotransposon RNA into DNA³. L1 usually moves only its own sequence from one genomic location to another (Fig. 1a). But Moran et al. show that it can, with surprising efficiency, co-mobilize a 3′ flanking segment of non-L1 DNA to new genomic locations. In this way, two previously unlinked host genomic segments are juxtaposed in a process referred to as 3′ transduction (Fig. 1b).

Figure 1: Mechanism of 3′ transduction by the L1 retrotransposon, based on the results of Moran et al.².

Read-through transcription by RNA polymerase leads to movement of the DNA that flanks L1. The signal for terminating transcription through L1 is either the weak cleavage/polyadenylation signal of L1 itself (a), or a stronger polyadenylation signal in the 3′ flanking DNA (b). When the resulting complementary DNA integrates into a new genomic locus (green), a simple L1 insertion (a), or a chimaeric insertion (b) is formed. Purple arrows, polyadenylation signals.

The process of 3′ transduction may occur in many eukaryotes, and several possible examples of it have been reported. In the most convincing case, a progenitor human L1 element, which spawned an insertion mutation in the dystrophin gene, was identified based on its unique, transduced 3′ flanking sequence tag⁴. A more ancient event, involving a human CFTR exon, probably occurred around ten million years ago⁵. In mouse cells, an L1 read-through RNA has been found associated with ribonucleoprotein particles⁶. And the structure of a Drosophila gene, jingwei, was proposed to arise from insertion of the complementary DNA for an alcohol dehydrogenase gene into an unrelated gene⁷. There are no retrotransposon sequences in jingwei, but its structure is consistent with a 3′ transduction event in which reverse transcription terminated before retrotransposon sequences were reached (Fig. 1a).

Retroviruses can incorporate cellular genes into their genomes, probably by aberrant read-through transcription into adjacent host DNA. But L1 elements differ from retroviruses in that they do not form the precise 3′ ends that are normally specified by retroviral long-terminal repeats. The signal for 3′ end formation comes instead from cleavage/polyadenylation signal sequences. Although the human L1 element contains a putative polyadenylation signal near its 3′ end, its sequence is not optimal — if it functions at all. This weak polyadenylation signal is supposedly suppressed by the action of more potent polyadenylation signals in the flanking DNA, resulting in the formation of a read-through transcript which could then act as the template for reverse transcription⁴, ⁸. By this hypothesis, the nature of the flanking DNA dictates which 3′ end gets formed. This built-in sloppiness has rendered L1 an efficient exon-mobilizing machine.

Although the exact molecular mechanism of L1 retrotransposition is not known, L1 probably cuts the target DNA and then reverse transcribes its own RNA to DNA^8,9,10. To study L1, Moran et al.² placed a reporter gene containing a splice acceptor — but lacking 5′ exon and promoter sequences — into an intact L1 element. They then introduced this contraption into cultured cells. The modified L1 very efficiently produced derivatives that had mobilized the reporter construct into new genomic locations. As demanded by selection for reporter-gene expression, these new sites were transcriptionally active, and each had ‘trapped’ the reporter-gene exon as part of a different chimaeric messenger RNA expressed in these cells.

As well as showing that L1 can transduce 3′ flanking sequences into target genes, the new work illuminates aspects of L1 retrotransposition. For example, the authors estimate the frequency of L1 retrotransposition into actively transcribed genes to be at least 6%. The 3′ transduction process also provides a convenient explanation for peculiarities of L1 structure, such as the dramatic variability in 3′ untranslated sequences of different L1 subfamilies — they probably arise from different 3′ transduction events, followed by internal deletions. Similarly, it is plausible that 5′ transduction events could occur, explaining the great variability observed among 5′ end sequences in different L1 subtypes. If a cellular promoter read into a full-length or partially truncated L1 element, the result would be a new 5′ end for the element. Finally, 3′ transduction explains the puzzling lack of target-site duplications for many L1 elements.

Moran and colleagues' results² indicate that read-through transcription of L1s is surprisingly efficient, even when the reporter gene's preferred polyadenylation signal is as much as two kilobases from the native L1 3′ end. This L1 read-through transcription — terminated by a strong polyadenylation signal in 3′ flanking DNA, subsequent processing, reverse transcription and incorporation into a new locus — provides a new mechanism for exon shuffling. Furthermore, this type of exon-shuffling mechanism allows fresh insights into the evolution of splicing, allowing us to tackle questions such as how alternative splicing arose, and why terminal exons are often unusually large¹¹ (Fig. 2). Until now, DNA recombination between introns was considered the best mechanism for exon shuffling. But this new mechanism implicates retrotransposons as being pivotal in large-scale modifications of gene structure.

Figure 2: Evolutionary consequences of L1-mediated 3′ transduction.

If exons (open arrows) of gene A are downstream of an active L1 element, they can be incorporated into a chimaeric (L1-A) pre-mRNA produced by L1 read-through transcription. This transcript terminates when it encounters gene A's strong polyadenylation signal. Processing of the pre-mRNA yields an mRNA consisting of L1 fused to the distal A exons. When the resulting L1-A cDNA is inserted into an intron of gene B, a long 3′ A exon including the strong polyadenylation signal is added to B. Even if inserted into the middle of gene B, this long A exon may become the terminal exon of a new hybrid B-A gene (this could help explain why 3′ terminal exons are often much longer than internal exons¹¹). The distal exons of gene B will be lost if cleavage/polyadenylation at the newly introduced gene A signal is 100% efficient (a). But inefficient cleavage/polyadenylation might produce a longer primary transcript that could be alternatively spliced (b). In this way, additional mRNAs can be formed through alternative splicing of the longer primary transcript, and the organism can sample a new gene variant while maintaining the original design. This provides a new hypothesis for the genesis of alternative splicing. (Vertical arrows point to the location of polyadenylation signals.)

Because most of the mammalian genome is made up of DNA that does not encode proteins, the consequences of L1 transporting such ‘non-exonic’ DNA are also notable. Although rearranging exons creates new proteins, L1-mediated shuffling of regulatory sequences may be equally crucial in evolution of the mammalian genome, concocting increasingly complex regulatory circuits. Changes in the timing and cell specificity of gene expression may facilitate fast morphological innovations, some of which could be enough to produce new variants or even species. The process of 3′ transduction is the latest in the series of fanciful twists of logic that typify the retroelements — they have found yet another way to mould not only their own fate, but also that of their host genome.