Abstract

Introduction

Long INterspersed Element 1 (LINE-1 or L1) is a transposable element that has accumulated over time to now account for about one third of human genomic DNA [1]. L1 is classified as an autonomous retrotransposon, as it encodes the enzymatic machinery necessary for its own mobilization via an RNA intermediate. In a round of L1 retrotransposition (Fig. 1A), a donor L1 is transcribed from its original genomic location, and the L1 mRNA translated to give rise to two different proteins. ORF1p is a nucleic acid binding protein [2, 3], and ORF2p harbours endonuclease [4] and reverse transcriptase [5] activities critical for the generation of new insertions. ORF1p and ORF2p exhibit strong cis-preference for binding their encoding mRNA to form the L1 ribonucleoprotein particle (RNP), a hypothesised retrotransposition intermediate [2, 6–10]. The L1 RNP then enters the nucleus, where the ORF2p endonuclease nicks genomic DNA to expose a free 3′ hydroxyl residue, which the ORF2p reverse transcriptase activity uses as a primer to initiate reverse transcription of its associated L1 mRNA, resulting in a new L1 insertion [4, 8, 11]. This process is known as target-site primed reverse transcription (TPRT) [12]. L1 insertions end in a poly-A tail [13] and are frequently flanked by variable length target-site duplications, a hallmark of the retrotransposition process [14].

Open in a separate window

Figure 1

Illustration of retrocopy mechanism and detection strategy. A: Generation of L1 insertions and gene retrocopies by the L1 machinery. The typical L1 retrotransposition pathway is indicated by black arrows; gray arrows denote the less-frequent mobilization of cellular mRNAs. Retrotransposition begins with the transcription of a full-length L1 in the genome. The L1 mRNA (wavy black line) is exported from the nucleus and translated, giving rise to the L1 encoded proteins ORF1p (green circles) and ORF2p (blue oval). ORF1p and ORF2p exhibit a strong cis-preference for binding their encoding mRNA, resulting in formation of the L1 ribonucleoprotein particle (RNP). Occasionally, the L1-encoded proteins mobilise cellular mRNAs (dashed red line) in trans. Regardless of the RNA template, insertions generated by the L1-encoded enzymatic machinery undergo target-site primed reverse transcription, resulting in a new L1 copy (blue rectangle), or a gene retrocopy (multi-coloured rectangle), at a distinct genomic location. L1 insertions and retrocopy insertions bear the hallmarks of target-primed reverse transcription, including poly-A tails and target-site duplications (purple arrows). B: Characteristics and detection of gene retrocopies. A typical parent gene (above; coloured rectangles denote exons) contains introns (grey lines) and resides at a particular genomic location (heavy black line). Paired-end sequencing reads (dashed lines) wherein one end maps to a gene (red and green rectangles) and the other to its known genomic location (black rectangles) are termed concordant. Conversely, paired-end sequencing reads wherein one end maps to a gene (red and green rectangles), but the other end maps to a distal genomic location (blue rectangles), are termed discordant (denoted by red X's). Discordant paired-end reads are indicative of a gene retrocopy (below), and allow mapping of the retrocopy to its genomic location. Gene retrocopies are distinguished from parent genes by their discrete genomic location (heavy blue line), the presence of retrotransposition hallmarks (target site duplications (TSDs), purple arrows; and a poly-A tail, An) and a lack of introns. Sequencing reads which span exon-exon junctions (bi-coloured rectangles) are also indicative of gene retrocopies.

Although cis-preference is the prevailing rule for template selection by the L1 enzymatic machinery, other transcripts can occasionally be retrotransposed in trans [6, 9, 15–20]. For example, the L1-encoded proteins are responsible for the addition to the human genome of more than one million copies of the nonautonomous retrotransposon Alu, and approximately 2,700 copies of the composite retroelement SVA [1]. The L1 enzymatic machinery can also reverse transcribe small RNAs such as the U6 snRNA [21, 22], as well as cellular mRNAs [6, 9, 23] (Fig. 1A).

L1 mobilization of cellular mRNAs is responsible for ~8,000 to 17,000 retrotransposed gene copies identified in the human genome reference sequence [24–27]. Here, we will refer to any retrotransposed gene copy as a ‘retrocopy’. Retrocopies bear the hallmarks of L1-mediated retrotransposition, including L1 endonuclease preference for the loose consensus sequence 5′-TTTT/AA-3′ [28], flanking target-site duplications, and, as mature mRNAs are mobilised by the L1 machinery, retrocopies lack introns and contain a poly-A tail [29, 30]. A promoter sequence is typically not delivered upon retrotransposition of a cellular mRNA; hence, most retrocopies are not expressed, and presumably lack functionality [29]. These retrocopies are known as processed pseudogenes. Occasionally, a retrocopy may insert into a genomic context permissive of its expression [31–33], giving rise to potentially functional transcripts. We will refer to such an insertion as a ‘retrogene’.

Numerous studies have demonstrated the importance of retrotransposed gene copies in genome evolution and function. Protein-coding retrogenes represent a significant contribution to the ongoing generation of new genes over evolutionary time [34–36]. Gene retrocopies can also influence parent gene expression; for example retrogene-derived siRNAs have been demonstrated to regulate gene expression in mouse oocytes [37, 38] and in the protozoan parasite T. brucei [39]. An expressed pseudogene in mouse has also been demonstrated to regulate the mRNA stability of its homologous coding gene [36], and retrogene mRNAs can compete with target gene transcripts for miRNA binding, acting as a molecular ‘sink’ and interfering with miRNA-mediated gene regulation [40]. Although an instance of a retrocopy insertion associated with human disease has yet to be identified, expression of an FGF4 retrogene causes short-legged chondrodysplasia in certain breeds of domestic dog [40]. Conversely, retrocopy insertions have the potential to confer protection from infectious disease: in owl monkeys, a retrocopy of the Cyclophilin A cDNA into the TRIM5 gene confers resistance to HIV [32].

Retroelement activity in humans represents an important source of inter-individual genetic variation. Indeed, numerous studies have taken advantage of advances in sequencing technology to uncover novel retrotransposon polymorphisms in the human population (reviewed in [41]). However, the extent of variation between humans with respect to retrocopy insertions is relatively unexplored. To address this issue, three very recent studies by Abyzov et al. [42], Ewing et al. [43], and Schrider et al. [44] analysed whole genome sequencing data generated by the 1,000 Genomes Project Consortium [45], as well as other sources of human genome sequences including The Cancer Genome Atlas (TCGA) [46], to uncover novel polymorphic retrocopies. Below, we compare the retrocopy discovery strategies and results of the three studies, and summarise the biological and evolutionary insights gleaned by each group. We then present our opinions on retrocopy discovery criteria, and speculate regarding the potential impact of germline and somatic retrocopies on inter- and intra-individual human genomic diversity.

Strategies for identifying novel retrocopies

Bona fide retrocopies bear distinguishing features relative to their parent genes: the presence of exon-exon junctions, a genomic address distant from the locus of the parent gene, and the structural hallmarks of L1-mediated retrotransposition. The three studies reviewed here exploited these characteristics to identify and validate novel retrocopies from whole genome sequence data (Table ​(Table1).1). Discordant paired-end reads, in which one end matches exonic sequence of a gene, but the other end aligns to a region of the genome distinct from the known locus of that gene, are indicative of retrocopy insertions, and can be used to map retrocopies to the genome (Fig. 1B). Sequencing reads that span exon-exon junctions are another means to identify retrocopies, since such reads would not arise from intron-containing parent genes (Fig. 1B).

Interpretations from novel retrocopy discovery

Each of the three studies extended its findings to draw conclusions about L1 biology, human diversity, and genome evolution. Abyzov et al. observed that their set of novel retrocopies was enriched for parent genes that already had known retrocopies present in the human genome reference, leading to the hypothesis that genes involved in specific biological processes may be prone to mobilization by the L1 machinery. Analysis of parent gene functional categories led to the conclusion that genes involved in nuclear envelope breakdown and re-formation, or expressed during the M or M/G1 phase of the cell cycle in HeLa cells, are more likely to be reverse transcribed by the L1 machinery. The authors posited that this correlation indicates a coupling of L1-mediated retrotransposition to cell division, as transcripts highly expressed during cell division would be more likely to undergo retrotransposition. Interestingly, Ewing et al. also observed an enrichment for parent genes that already have retrocopies in the human genome reference, but reported no specific enrichment for cell cycle genes. However, the identification by Ewing et al. of novel retrocopies in tumours is consistent with retrotransposition taking place under conditions of accelerated cell division.

Identifying polymorphic retrocopy insertions can provide insights regarding the contribution of retrocopy insertions to human diversity. Indeed, Abyzov et al. and Ewing et al. found that retrocopy insertions can be population-specific. Furthermore, from 58 segregating retrocopy insertions among 1,025 individuals, Ewing et al. calculated an approximate rate of retrocopy generation at one insertion per 5,177 individuals per generation. Thus, retrocopy insertions contribute to human diversity. Ewing et al. also estimated a rate of retrocopy formation in chimps at one per every 6,804 individuals per generation. A recent study suggested that L1 retrotransposition has differentially influenced human and chimp genome evolution [49]. The ongoing generation of retrocopies by the L1 retrotransposition machinery may likewise contribute to differences in genome evolution between humans and chimps.

Schrider et al. applied their findings to investigate aspects of genome evolution. For example, in support of the hypothesis that natural selection drives gene movement on and off of the X-chromosome [50], they found that fixed functional retrogenes are more likely to have arisen from movement to or from the X-chromosome than novel, polymorphic retrocopies. In addition, their analysis revealed a depletion of fixed retrogenes in introns, while novel retrocopies do not show such a bias. This finding is consistent with the hypothesis that intronic insertions are generally deleterious, whereas intergenic insertions are neutral [31]. Schrider et al. also presented evidence that some of their novel retrocopy insertions can give rise to expressed chimeric transcripts, and that positive selection may act upon retrocopies.

Perspectives, insights and future directions

How important is a mapped insertion site?

Perhaps the most important distinction among discovery strategies is whether identification of the genomic insertion site is required to call a novel retrocopy. Ewing et al. adhered to the most stringent requirements, as their discovery strategy required discordant reads mapping a retrocopy to a genomic locus. Perhaps unsurprisingly, Ewing et al. reported the smallest cohort of novel retrocopies; however, about 56% of their novel retrocopies identified from 1,000 Genomes Project Consortium data were also discovered by at least one of the other two studies. This is a higher rate of overlap than Abyzov et al. (25%) or Schrider et al. (42%), suggesting that calls made based on insertion site may be more reproducible than those made on exon-exon junctions alone. Furthermore, as shown in Fig. 2, among insertions detected by Abyzov et al. or Schrider et al. but not confirmed in another study, the vast majority (99/111 and 38/41, respectively) were called based on exon-exon junctions alone. On the other hand, the datasets generated by Abyzov et al. and Schrider et al. share 17 retrocopies, seven with a mapped insertion site, that were not detected by Ewing et al. Thus, more conservative discovery criteria may result in reduced sensitivity. In general, however, we argue that the ‘gold standard’ for calling novel retrocopies should entail identification of a genomic location. This criterion is particularly important in cases where the parent gene has one or more known retrocopies in the genome, as reads arising from a known retrocopy could be falsely identified as evidence of a novel retrocopy because of sequencing or alignment errors. Ideally, novel retrocopies should also be demonstrated to bear the hallmarks of L1-mediated retrotransposition.

Open in a separate window

Figure 2

Overlap in retrogene cohorts discovered from 1,000 Genomes Project Consortium data by the three studies. Only retrogenes absent from the human genome reference, but present in 1,000 Genomes Project Consortium data, are represented. Nine retrocopies present only in TCGA data discovered by Ewing et al. are excluded from this comparison. Within each segment, black numbers indicate the total number of novel retrogenes; below, red numbers indicate the number of retrogenes for which an insertion site was mapped by at least one study. For retrocopies with known insertion sites, overlap was confirmed by comparing insertion site coordinates. For those without known insertion sites, only gene names were used. Segments are not drawn to scale.

Conclusions and outlook

As more deep-sequenced genomes become available for analysis, the number of known polymorphic retrocopies in the human population will likely increase. Indeed, high-coverage genomes yield many more novel retrocopies than pooled low-pass genomes: for example in six deeply sequenced individuals, Abyzov et al. detected 17 novel retrocopies (2.8 retrocopies per individual), while analysis of 968 shallowly sequenced genomes yielded only 132 additional retrocopies (0.13 retrocopies per individual). Thus, analysis of low-coverage genomes likely results in a gross underestimation of novel retrocopies in the human population. Furthermore, the three studies reviewed here relied on short, paired-end sequencing reads for mapping putative retrocopies to the human genome reference. PacBio single-molecule real time (SMRT) sequence technology was recently used to sequence complex genomic regions, generating sequence reads of ~1.8 kb average length [66]. Genome sequences generated using this technology will no doubt facilitate confident mapping of transposable element and retrocopy insertions, as individual reads have the potential to span extensive sequence stretches on both sides of an insertion-genome junction.

The three studies highlighted above reveal that, although rare, retrocopies generated by retrotransposition of cellular mRNAs represent an important component of human genetic diversity. Furthermore, their findings provide insight into L1 biology and genome evolution. Indeed, it is interesting to speculate on whether novel retrocopies uncovered by the studies reviewed here will ultimately become fixed functional retrogenes. A 2005 study by Marques et al. estimated a rate of one new functional retrogene per million years in the human lineage [32], suggesting that most novel retrocopies will ultimately be eliminated through negative selection or genetic drift [31]. Generation of retrocopies in somatic cells has important implications for human health and disease, and future studies will no doubt shed light on the consequences of novel retrocopy insertions occurring alongside retroelement insertions in cancer and in the mammalian brain.

Note added in proof, post peer-review

De Boer et al. have reported a male patient afflicted with chronic granulomatous disease resulting from the L1-mediated retrotransposition of a transcript of the TMF1 gene into the X-linked CYBB gene [67]. This is the first documented case of human disease caused by a retrocopy insertion.

Acknowledgments

References