Background

Transposable elements (TEs) were discovered by Barbara McClintock in the 1940s and described as moving DNA sequences that can cause genomic instability [1]. As she was able to link TE activity with variations in maize kernel colors, she coined them “controlling elements”, underlying their apparent involvement in gene regulation. TEs are nowadays known to be major components of genomes and have been found in every species that has been looked at, including prokaryotes, protists, fungi, plants and animals [2,3,4].

TEs are classified into two main classes according to their transposition mechanism [5, 6]. The transposition of retrotransposons (class I TEs) occurs through the reverse transcription of an RNA intermediate into a cDNA molecule that is subsequently inserted into a new locus [7, 8]. This replicative transposition process, a “copy-and-paste” mechanism called retrotransposition, leads to the expansion of the retroelement family in the host genome. Retrotransposons gather both Long Terminal Repeat retrotransposons (LTRs), with flanking repeated sequences in direct orientation necessary for the expression and integration of the element, and non-LTR retrotransposons, also called Long Interspersed Nuclear Elements (LINEs). Autonomous retrotransposons encode a reverse transcriptase (RT) and other proteins necessary for integration (an integrase for LTRs and an endonuclease for LINEs) and other aspects of transposition [7,8,9]. In contrast, non-autonomous retrotransposons, including Short Interspersed Nuclear Elements (SINEs) that are mobilized by autonomous non-LTR retrotransposons, do not encode any proteins and rely on those produced in trans by autonomous elements to transpose [10, 11]. DNA transposons (class II TEs) do not require the reverse transcription of an RNA intermediate for their transposition [12]. They mostly use a “cut-and-paste” mechanism, the TE copy being excised from its original locus and integrated elsewhere into the genome. Many DNA transposons, including the widespread DDE transposon family, classically encode a transposase (with the DDE motif forming its active site in DDE transposons) and are flanked by Terminal Inverted Repeat (TIR) sequences that are bound by the transposase for excision and integration [9, 12]. Other types of DNA transposons include Helitrons [13, 14], which are rolling-circle DNA transposons with no TIRs encoding a helicase, and Polintons/Mavericks [15, 16], which are self-synthesizing DNA transposons with long TIRs encoding a DNA polymerase. Non-autonomous elements called Miniature Inverted Repeat Transposable Elements (MITEs) are mobilized in trans by related autonomous DNA transposons [12].

Each species genome is characterized by a specific composition in TEs, both quantitatively and qualitatively. For instance, the genome of the maize Zea mays is composed of nearly 85% of transposable elements [17], whereas the genome of the yeast Saccharomyces cerevisiae contains less than 4% of TEs [18]. In unicellular organisms, the genome of Trichomonas vaginalis contains almost exclusively DNA transposons, while almost only retrotransposons are found in Entamoeba histolytica [19, 20]. A marked variability in TE content and diversity has been also observed among vertebrates [21]. Indeed, the genomic amount of TEs ranges from 6% in the pufferfish Tetraodon nigroviridis up to 55% in the zebrafish Danio rerio. Some groups of TEs are found in most vertebrate species (LINE retrotransposons or Tc-Mariner DNA transposons for instance), whereas others are restricted to certain vertebrate sublineages and absent from others, such as the DIRS and Copia retrotransposons that are present in fish and amphibians but absent from mammals and birds [21].

Most TE insertions are thought to be either neutral or deleterious, depending on the context of the genomic region where they are inserted. TE insertions can be deleterious for instance by disrupting open reading frames (ORFs) or by altering gene transcriptional regulations. However, and despite their “selfish” characteristics, TEs are subject to the drift-selection balance and can be positively selected if they are beneficial to the host [12]. Indeed, some insertions have been shown to play a positive role in species evolution by contributing to new regulatory and coding sequences (Fig. 1) [22,23,24,25,26,27,28]. Such a recruitment by the host to fulfil useful functions is called exaptation or molecular domestication. The ability of TE sequences to give rise to evolutionary innovations has been more and more documented in the past years and becomes of growing interest, helped by the recent technological developments in genome sequencing and gene expression profile analysis. The structural and functional characteristics of different TE families might confer them with different potential to be exapted. TEs can contain different functional ORFs encoding proteins with various properties such as endonucleases, integrases, transposases, reverse transcriptases and other proteins with DNA/RNA/protein-binding domains, and diverse transcriptional regulatory sequences such as promoters or enhancers. For example, LINE L1 elements contain an internal RNA polymerase II promotor and encode beside an RT an RNA-binding protein and an endonuclease; SINEs in contrast do not carry any ORF and have an RNA polymerase III promoter; LTR retrotransposons present transcriptional regulatory sequences in their long terminal repeats and generally encode an integrase, a protease, a RNase H and a structural protein called GAG in addition to their RT, with an additional Envelope gene that Endogenous Retroviruses (ERVs) have occasionally kept from their infectious ancestors; DNA transposons can among others code for transposases, helicases and DNA polymerases. These functional ORFs and regulatory sequences can be reused to the host benefits. The mobilome can thus be regarded as an evolutionary toolbox, as TEs bring with them in host genomes sequences encoding proteins able to bind, replicate, cut, rearrange or degrade nucleic acids, and to associate with and modify other proteins, among other biologically relevant properties.

Fig. 1
figure 1

Adaptive mechanisms of TE-derived sequences evolution leading to developmental innovations. After the insertion of a TE: a in an intron of a protein-coding gene, part of the TE can give rise to a new exon (exonization). Splicing sites can either be directly present in the TE sequence or can be acquired by mutations. b part of the TE can form a new host gene and be transcribed from either a flanking host promoter or a promoter derived from the TE sequence itself. c the TE can form a new long non-coding RNA (lncRNA) gene and be transcribed from either a flanking host promoter or a promoter derived from the TE sequence itself. d-e in the upstream region of a coding or RNA gene, the TE can form a new promoter (D) or enhancer (this model also works for TE-derived silencers) (e). f the TE can form an insulator region, which recruits the CCCTC-binding factor (CTCF) and blocks heterochromatin spreading, allowing the expression of downstream sequences. Red boxes correspond to TEs and blue boxes to exapted TE sequences

Full size image

Vertebrates constitute a geographically widely expanded taxonomic group that appeared more than 500 million years ago and has colonized almost all ecological environments [29]. The emergence of vertebrates represents a major evolutionary transition. This group has acquired many derived traits, namely: a unique nervous system composed of a complex brain with forebrain, midbrain and hindbrain specialized regions, and cranial nerves, spinal cord and ganglia; the sensory placodes and the sensory organs they give rise to (olfactory bulbs, vestibular apparatus and otic placode for example); the neural crest, which develops into cranium, branchial skeleton and sensory ganglia; a complex endocrine system allowing the apparition of new hormones and new organs such as the placenta; bones and cartilages contributing to the skull, jaws and vertebrae; paired appendages; adaptive immunity [30,31,32]. These novelties, which subsequently diversified in different sublineages, have contributed to the evolutionary success of vertebrates, allowing them to improve the sense of and the move in their environment, to develop new organs and complexify them, and to turn to extensive predation.

At the origin of vertebrates, two events of whole genome duplications allowed a massive expansion of the gene repertoire [33]. However, the sole emergence of paralogous genes may not explain all the innovations that appeared, and it has been also proposed that regulatory divergence might account for major organismal diversification [34, 35]. Accordingly, the analysis of the genome of the cephalochordate amphioxus, a sister outgroup species of vertebrates, has underlined the specialization of gene expression and the complexification of gene regulation during invertebrate to vertebrate transition, mainly due to the recruitment of new regulatory networks [36]. The precise understanding of the genetic and evolutionary mechanisms underlying this transition is of particular interest, and we propose to explore the role of TEs in this context. Several examples of TE recruitment events crucial for vertebrate development have been documented in the last years. In this review, we discuss the different mechanisms through which TE-derived sequences have played a role in vertebrate genome evolution. We focus on selected examples illustrating the innovative potential of transposable elements as a source of new protein-coding sequences, new small and long non-coding RNA genes and new regulatory elements having driven the evolution of vertebrate development.

TE-derived sequences as new protein-coding sequences

TE exonization

Inserted TE sequences can occasionally be recruited as new exons of pre-existing genes, a process called TE exonization (Fig. 1a). Exonization is defined as the formation of a novel exon from an intronic or intergenic sequence carrying splicing sites. Such new exons can be protein-coding but might also constitute new 5′ or 3′ untranslated regions with possible regulatory functions.

TE exonization is not an anecdotal process and has been largely documented in mammals and other vertebrates, where it occurs more frequently than in non-vertebrate species [37,38,39]. In the human genome, among 233,785 exons, more than 3000 (~ 1%) are derived from TEs [37, 40]. Among them, about 1640 correspond to Alu SINE elements, 640 to LINEs, 310 to MIRs (Mammalian-wide Interspersed Repeats, SINE elements), 300 to LTRs and 230 to DNA transposons [37]. Human exonized TEs are generally alternatively spliced, allowing protein variability [41,42,43]. It was also hypothesized that many TE-derived exons act as post-transcriptional gene regulators instead of being part of the protein-coding sequence itself [40]. The prevalence of Alu elements as TE-derived exons can be linked not only to their high copy number -with 1200,000 copies, they constitute as much as 10% of the human genome [44], but also to the fact that Alu sequences contain many potential splicing sites [45]. Alu elements indeed present up to ten 5′ and thirteen 3′ cryptic splicing sites that can be activated into functional splice sites through mutations or modifications such as adenosine-to-inosine RNA editing [38, 41]. Alu exons often modulate translational efficiency and can lead to lineage-specific regulations of gene translation [46]. Alu exonization can also cause genetic diseases in human such as the Alport syndrome, which is characterized by progressive renal failure, hearing loss and ocular abnormalities [47]. LINEs and to a lesser extent LTR retroelements can be exonized too [48, 49].

Exonization of intronic insertions is influenced by multiple factors. In the human genome, exonization is promoted by large intron size, high intronic GC content, and, importantly, by the presence of young transposable elements, in particular close to transcription starting sites [50]. These factors might contribute to a decrease of RNA polymerase II elongation rate and to a reduction of spliceosomal efficiency, allowing an increase of the “window of opportunity” for spliceosomal recognition and thus for exonization. Other mechanisms inhibit Alu exonization. It has been shown in human that the RNA-binding protein hnRNP C prevents Alu exonization by avoiding the binding of splicing factor U2AF65 to Alu cryptic exons, thus blocking Alu splicing sites; this prohibits Alu exon inclusion that would potentially lead to the formation of aberrant transcripts [51]. The binding of hnRNP C to Alu RNA is highly dependent on two poly(U) tracts present in Alu sequences inserted and transcribed in antisense orientation compared to the gene. These poly(U) arise from the antisense transcription by the gene promoter of the Alu terminal poly(A) and the internal poly(A) linker separating the two arms of Alu sequences (Alu are dimeric elements). Point mutations in these Alu poly(U) sequences are sufficient to impair the binding of hnRNP C [51]. Thus, the accumulation of mutations preventing hnRNP C binding can favor Alu exon inclusion.

Some examples illustrate well how intronic TEs can drive transcriptome and proteome diversification through the formation of lineage- and tissue-specific alternative exons. The vertebrate lamina-associated polypeptide 2 gene (tmpo for thymopoetin) encodes several membrane protein isoforms including LAP2β suggested to control nuclear lamina dynamics at the nuclear periphery by binding specifically to B-type lamins. Another isoform, the mammalian-specific LAP2α protein, has a domain derived from the gag ORF of a DIRS1-like retrotransposon [52]. Unlike other isoforms, LAP2α is a non-membrane protein that binds to A-type lamins in the nucleoplasm [53]. This isoform is implicated in nuclear organization dynamics during the cell cycle [54, 55]. A mutation in the TE-derived domain of LAP2α has been associated with dilated cardiomyopathy in humans [56].

In mammals, the gene prl3c1 belonging to the prolactin gene family encodes a cytokine expressed in uterine decidua and implicated in the establishment of pregnancy. In rodents, this gene has acquired a novel transcript variant in a common ancestor of the house mouse Mus musculus, M. spretus and M. caroli through the insertion of a composite TE into its first intron [57]. The inserted TE, which consists of an LTR element interrupted by a LINE, gave rise to an alternative promoter and an alternative first exon. In contrast to the “classical” transcript, the new variant is expressed in the Leydig cells of the testis. The variant protein shows a different intracellular localization and modulates the growth of testes and their capacity to produce testosterone and sperm. Such a TE co-option might contribute to the diversity of testicular development and functioning.

The rtdpoz-T1 and rtdpoz-T2 retrogenes, specifically expressed in testis and in the develo** embryo in rat, and supposed to encode nuclear scaffold proteins functioning as transcription regulators, have multiple exons deriving from TE sequences [58, 59]. For example, rtdpoz-T1 has 5 out of 8 exons and an alternative polyadenylation signal that are derived from various TEs, mainly L1 and ERVs. These TE-derived exons may be implicated in the translational regulation of these transcripts, notably through the formation of upstream ORFs [59].

The vertebrate insulin-like growth factor 1 (IGF-1) is a hormone involved in the development and growth of many tissues. IGF-1 plays a role for instance in synapse maturation and skeletal muscle development. Three isoforms of IGF-1 are known, IGF-1Ea, IGF-1Eb and IGF-1Ec [60]. The IGF-1Ea isoform is conserved among vertebrates, whereas the two others are mammal-specific and coincide with the insertion of a MIR-b SINE element that allows the formation of a fifth exon [61]. This fifth exon adds a disordered tail to IGF-1, which is highly suspected to be the source of post-translational modifications and regulatory functions. This allows a lineage-specific regulation of IGF-1.

Finally, the exonization of an Alu-J SINE element has been linked to the evolution of hemochorial placentation in anthropoid primates [62]. Hemochorial placentation is a placental implantation specific to rodents and higher order primates. In this type of placenta, the maternal blood is separated from the fetal blood by only one barrier, the chorion. This may optimize nutrient and gas exchange but makes the immune tolerance more challenging. The chorionic gonadotropin (CG) is a heterodimeric glycoprotein hormone formed by an alpha subunit, the glycoprotein hormone alpha (GPHA), and a beta subunit CGB [63]. CG is involved in the regulation of ovarian, testicular and placental functions. An Alu-J is inserted in the gpha gene in anthropoid primates, and its alternative exonization induces the formation of a GPHA isoform called Alu-GPHA that contains an additional N-terminus [62]. This isoform is only expressed in chorionic villus tissues and placenta, while the GPHA isoform without the Alu is expressed in other tissues. In human, the heterodimer Alu-hCG formed with the subunit Alu-GPHA shows a longer serum half-life and has a better trophoblast invasion activity compared to hCG, allowing the improvement of placenta implantation and invasion.

TE molecular domestication to form new protein-coding genes

TEs can give rise to new functional host genes, a process known as molecular domestication (Fig. 1b). In the human genome, more than hundred protein-coding genes are thought to be derived from TEs [64, 65], representing about 0.5% of the complete set of human protein-coding genes. For example, the mammalian centromere protein B (CENP-B) is derived from the transposase of a pogo-like DNA transposon [66, 67]. Like its transposase ancestor, this protein is able to bind DNA. CENP-B is involved in centromere formation during both interphase and mitosis, and directs kinetochore assembly. Ty3/gypsy LTR retrotransposons have given rise to several multigenic gene families including the Paraneoplastic (PNMA, also called Ma genes, 15 genes), MART (12 genes) and SCAN families (56 genes) [68,69,70,71]. Overall, at least 103 genes derived from GAG proteins of Gypsy LTR retrotransposons have been identified in mammalian genomes, 85 being present in the human genome.

TE domestication and lymphocyte development

Two important TE-derived proteins in jawed vertebrates are RAG1 and RAG2 (Recombination Activating Gene 1 and 2) that together catalyze the V(D)J somatic recombination, a mechanism essential for the establishment of the vertebrate immune repertoire [72]. This genetic recombination, which takes place in develo** lymphocytes, is at the basis of the adaptive immune system, since it allows the formation of diverse antibodies and T-cell receptors capable of specifically recognizing a great variety of pathogens. Pathogen recognition is ensured by the antigen-binding domain, which is encoded after assembling gene segments called variable (V), diversity (D) and joining (J). The joining of different V, D and J segments generates, in association with additional mutational processes, the great diversity of antibodies that can be produced by a jawed vertebrate.

RAG1 and RAG2 lymphoid-specific endonucleases are key enzymes for this somatic recombination. Both proteins associate as a recombinase to introduce double-strand breaks in DNA at recombination signal sequences (RSSs) that frame each V, D and J gene segment. This DNA cleavage resembles the transposition mechanism of DNA transposons in early steps. Indeed, the rag1 and rag2 genes have been derived from a RAG transposon related to Transib DNA transposons approx. 500–600 million years ago [73,74,75]. The RSSs recognized by RAG1/RAG2 might be derived from the TIRs of the ancestral transposon. The hypothesis is that, at the basis of deuterostomes, a Transib element originally containing only a rag1 transposase might have captured an additional rag2 ORF, leading to a RAG transposon with increased transposition activity [76]. By comparing vertebrate RAG proteins to a RAG transposon from the amphioxus genome that carries both rag1- and rag2-like genes [76, 77], putative key mutations in the domestication process, that impaired the transposition ability of the rag genes in the post-cleavage steps, have been identified [78]. This example of molecular domestication illustrates well how a specific genomic context may favor the selection and domestication of a transposable element. Indeed, for the emergence of the V(D)J recombination, the insertion of a TE with its RSS sequences into a gene encoding an immunoglobulin-domain receptor protein was probably a prerequisite to the formation of the ancestral fragmented antigen receptor gene [78].

TE domestication and brain development

Several retrotransposon-derived genes are implicated in vertebrate brain development, such as members of the PNMA, MART, SCAN and ARC gene families, that are all derived from gag genes of Ty3/gypsy LTR retrotransposons [68,69,70,71].

The pnma10 gene (aka sizn1/zcchc12/pnma7a) from the PNMA gene family is involved in mouse forebrain development and mutations are associated with X-linked mental retardation in human [79]. The pnma5 gene shows a neocortex-specific expression in primate adult brain particularly in the association areas [80]. Higher order association areas are primate-specific areas responsible for the integration of multiple inputs such as somatosensory, visuospatial, auditory and memory processes; they contribute to perception, cognition and behavior [81]. The pnma5 gene is also present in mice but its neocortex-specific expression is not conserved. Thus, pnma5 is thought to be one of the major genes involved in the expansion and specialization of association areas in the primate brain [80].

The protein encoded by the eutherian gene sirh11 (aka mart4/rtl4), which belongs to the MART gene family, has conserved the gag zinc finger domain necessary for its binding to nucleic acids [70]. Sirh11 is of crucial function for cognition [82]. Indeed, mice sirh11 knockout mutants show impulsivity, attention and working memory defects as well as hyperactivity, suggesting a critical role in behavior. As this gene is present in eutherians only and could have conferred an essential advantage for competition by develo** cognitive functions, it has been suggested to have played an important role in eutherian evolution [82].

The placental mammal gene peg3 (zscan24) from the SCAN gene family has been also shown to be involved in mouse behavior [70]. This gene is paternally expressed during embryonic development and in adult brain. Its inactivation leads to growth retardation and abnormal maternal behavior for nest building, pup retrieval and crouching over pups, which can cause offspring death [83]. Moreover, mutant mothers present milk ejection defects. This phenotype has been related to a reduced number of oxytocin neurons. Growth retardation and abnormal maternal behavior are suggested to be due to impaired neuronal connectivity [83].

Finally, the arc tetrapod gene was shown in mice to be essential for synapse maturation and synaptic plasticity, and is involved in major neuronal processes of learning [70, 84]. Arc mutations have also been linked to several human disorders such as Alzheimer’s disease, Angelman neurodevelopmental disease, schizophrenia and autism among others, highlighting the crucial role of the arc gene in brain development and functioning [85,86,87,88,89,90,91,92]. The ARC protein has conserved structural properties similar to those of GAG proteins. Particularly, it forms capsid-like structures that transport RNA molecules across synapses and thus mediate intercellular communication between neurons [93]. Interestingly, arc-like genes called darc have been identified as duplicated copies in the genome of Drosophila melanogaster. Although tetrapod arc and Drosophila darc genes have been formed from Ty3/gypsy retrotransposons by independent molecular domestication events, they present similar properties of mRNA trafficking, suggesting evolutionary convergence [93, 94].

TE domestication and placenta development

TE molecular domestication probably played crucial roles in the appearance and diversification of placenta development during mammalian evolution (Fig. 2). For instance, the MART genes peg10 (aka mart2/rtl2) and peg11 (aka mart1/rtl1) are placental genes derived from gag and partial pol sequences of Sushi Ty3/gypsy LTR retrotransposons [95, 96]. Peg10 influences the development of the spongiotrophoblast and labyrinth layers, which are the cell layers separating the embryo from the maternal tissues of the placenta, and peg11 maintains the fetal capillary endothelial cells. Mutation of the sirh7 (aka mart7/rtl7/ldoc1) gene leads to dysregulation of placental cell differentiation and maturation linked to placental hormone overproduction [97].

Fig. 2
figure 2

The different evolutionary contributions of TE-derived sequences to placental development. a Major TE co-option events in placental development. Molecular domestication of several TEs (Ty3/gypsy, ERV) has led to the formation of genes essential for placental development (peg10, peg11 and syncytins). Alu exonization in gpha gene has improved placenta implantation and invasion. Co-option of TEs (ERVs) as promoter regions has led to placental regulatory circuits for several genes such as leptin and pleiotrophin. Co-option of TEs as enhancers has allowed the rewiring of placental gene networks, such as ERVs which have led to progesterone and cAMP responsive enhancers regulating placental endometrial cell gene (ECG) network. ECPs: proteins encoded by ECGs. The regions of the TE source of the co-opted sequence are represented in red in TEs and the resulting host sequences are represented in different blue/green shades. b Roles of the TE co-options in human placental development. The arrows illustrate the function of the proteins encoded by the genes presented in A. Baby and pregnant woman illustrations are from https://smart.servier.com

Full size image

Syncytin genes also play a central role in placenta development. They are derived from endogenous retrovirus envelope (env) sequences, which encode membrane proteins that allow viral fusion with the target cells necessary for infection. The SYNCYTIN proteins have kept some properties of the ancestral ENV proteins. They are able to promote cell-cell fusion, allowing trophoblast differentiation and the formation of the syncytiotrophoblast tissue, which triggers the exchange of nutrients and gases between mother and child [98,99,100]. Moreover, some SYNCYTIN proteins play a role in maternal immune tolerance, this being probably linked to the capacity of parental retroviruses to target and repress immune cells thanks to the immunosuppressive activity of the ENV protein [101,102,103]. Indeed, at least one human (SYNCYTIN-2) and one mouse SYNCYTIN (SYNCYTIN-B) show immunosuppressive activity in vivo in mouse [104].

Among placental mammals, 14 different syncytin genes have been identified in different lineages presenting various placenta structures characterized by different invasion levels of the uterus by trophoblast cells. The different syncytin genes, their expression and their properties may play a role in the placental morphological diversity observed among mammals. In sheep, the env gene of a very recently endogenized Jaagsiekte Sheep Retrovirus (JSRV), present at ca. 20 copies in the genome, has functions similar to those of syncytin domesticated genes [105]. This env gene indeed contributes to trophectoderm (first epithelium of the mammalian embryo) development and leads to pregnancy loss when downregulated. This might represent an example of a retrovirus gene being on the way of molecular domestication. Additionally, the human gene suppressyn has also been identified as an ERV env-derived gene [106]. Its protein product acts as a regulator of SYNCYTIN by binding to SYNCYTIN-1 receptor, thus inhibiting SYNCYTIN-1-mediated cell fusion.

Interestingly, syncytin genes in different lineages are not orthologous and have been formed by independent events of molecular domestication of ERV envelope genes, testifying for a fascinating case of convergent evolution. This underlines how TEs can represent (almost) ready-to-use molecular material that can be repurposed independently several times during the evolution of different lineages. In addition, it has been recently demonstrated that ERV env sequence captures are not specific of eutherian mammals, since other syncytin genes of independent origins have been found in marsupials and even in some viviparous lizards [107, 108].

Mammalian placenta evolution through the molecular domestication of several different retrotransposon and retrovirus genes has been proposed to follow a “baton pass” mechanism [109]. First, the early birth and high conservation of the three LTR retrotransposon-derived genes peg10, peg11 and sirh7 among mammals suggest that they could be at the origin of the primitive placenta at the base of placental mammals. Subsequently, an ancestral gene responsible for cell fusion may have been substituted by syncytin gene(s), which might have then replaced one another, ensuring or even improving the function and the performance of the previous syncytin gene, and allowing placenta morphological innovations [109, 110].

Placenta appears thus to be the place of multiple events of TE co-option. Some studies suggest that these domestications may have been facilitated by the hypomethylation of DNA in placenta compared to other tissues, allowing higher TE expression and subsequent easier TE recruitment [111, 112].

TE domestication and the diverse roles of the ZBED family

The ZBED gene family derives from hAT DNA transposons, and more precisely from the BED zinc finger domain of their transposase, which is involved in DNA binding [113]. This gene family is implicated in various aspects of tissue or organ development in vertebrates. For example, the mammalian ZBED3 binds to the AXIN protein to form a complex that regulates the Wnt/β-catenin signaling pathway, which is essential for embryogenesis and carcinogenesis [114]. In addition to the BED domain, zbed1, zbed4 and zbed6 also kept the DDE catalytic domain of the ancestral TE transposase, which contains an ⍺-helical domain and a dimerization domain. Present in bony vertebrates, zbed4 is proposed to be involved in retinal morphogenesis and in the functioning of Müller retinal glial cells by activating the transcription of genes expressed in Müller cells or by regulating their nuclear hormone receptors [115]. The placental mammal gene zbed6 encodes a transcription factor essential for muscle development. A single nucleotide (nt) mutation in an igf2 intronic sequence prevents the repression of this gene by ZBED6, leading to an increase in muscle growth and heart size and to a decrease in fat deposition [116]. ChIP-sequencing experiments have revealed about 1200 additional putative genes targeted by ZBED6, with particular enrichment in genes involved in development, cell differentiation, morphogenesis, neurogenesis, cell-cell signaling and muscle development. Finally, the vertebrate gene zbed1 is implicated in cell proliferation by regulating several ribosomal protein genes [117, 118].

TEs as a source of new non-coding RNA genes

TE-derived small non-coding RNAs

TE sequences can be a source of small non-coding RNAs (sncRNAs) (Fig. 1c). Several studies have shown that some sncRNAs can derive from TEs, such as microRNAs (miRNAs) [119] and Piwi-interacting RNAs (piRNAs) [120]. These sncRNAs generally constitute TE silencing factors, but they have also shown abilities to regulate host gene expression by sequence complementarity through mRNA degradation and translation inhibition (Fig. 3a). sncRNAs can also induce DNA methylation of the loci close to the nascent mRNA their target. This can induce heterochromatinization, which can spread in the targeted genomic region and thus can potentially lead to the transcriptional repression of neighboring genes (Fig. 3a) [121].

Fig. 3
figure 3

Functions of TE-derived non-coding RNAs. a Mechanisms of action of TE-derived small non-coding RNAs (sncRNAs) through sequence complementarity. TE-derived sncRNAs are formed by fragmentation of TE-derived transcripts [122, 294], siRNAs being generated through the cleavage of the successive precursors pri-miRNAs and pre-miRNAs [122]. TE-derived sncRNAs, associated to proteins (RNA-induced silencing complex for miRNAs [122], PIWI proteins for piRNAs [150]) form double-stranded RNAs with complementarity to some RNAs of the host transcriptome, this leading to the cleavage of RNAs (1) and to the inhibition of translation (2). sncRNAs also mediates the heterochromatinization of TEs to silence them after the recruitment of DNA and histone methyltransferases (3). This heterochromatinization can spread to neighboring regions, altering their expression. b Evolution and function of the xist gene. Top: the human xist lncRNA gene has been formed after ancient insertions of several TEs (red boxes) into the ancestral protein-coding lnx3 gene, which is still present in chicken. lnx3 blue boxes represent the exons homologous to xist exons and dark grey boxes other exons. ** molecular pathways in human evolution. Cells. 2019;8(2):130." href="/article/10.1186/s13100-020-00229-5#ref-CR213" id="ref-link-section-d177312572e2296">213]. Olfaction, color vision, fertilization, cellular immune response, amino and fatty acids metabolism and detoxification were found to be particularly enriched for retrotransposon-derived gene regulation, i.e. mainly pathways with strong lineage/species specificity. The analysis of the association between TEs and active/repressed chromatin marks across 24 human tissues showed that SINEs and DNA transposons are enriched in globally active regions, while LTRs show a more tissue-specific enrichment [214]. Moreover, TEs enriched in tissue-specific regulatory regions present binding sites for tissue-specific TFs, and their expression correlates with the tissue-specific expression of neighboring genes. This indicates that TEs can serve as a major source for regulatory sequence turnover in a tissue-specific manner, as observed in human and mouse [214, 215].

In addition to enhancers and silencers, TEs can form new gene promoters. As much as 11 and 16% of RNA polymerase II binding sites have been estimated to be derived from TEs in mouse and human genomes respectively [228, 229].

An insertion of the MER130 SINE is involved in the development of the neocortex, a mammalian-specific structure responsible for the implementation of cognitive, emotive and perceptive functions [230]. This TE works as an enhancer of critical neocortical genes. A tetrapod LF-SINE-derived enhancer controls the islet-1 (isl1) gene, which encodes a transcription factor essential for tetrapod brain development, particularly for motor and sensory neuron differentiation [231, 232].

Interestingly, a new regulatory function has been identified for SINEs in mouse neurons [233]. In neurons, synaptic activity influences gene expression through epigenetic modifications and the recruitment of regulatory proteins. SINE sequences located close to activity-regulated genes act as regulators for their expression. In response to neuron depolarization, these SINE sequences are acetylated, inducing the binding of the transcription factor TFIIIC. TFIIIC recruitment allows activity-dependent transcription, the relocation of inducible genes to transcription factories (i.e. specific nuclear foci where stimulation-responsive genes are expressed), as well as dendritogenesis [233]. In this context, the binding of TFIIIC to SINEs mediates the coordination of the nuclear architecture, allowing activity-dependent gene expression.

Finally, TE-derived sequences can be involved in neural gene cis-regulation through epigenetic modifications [267], as proposed for SINE invasion in dog, rodent and opossum genomes [265]. Accordingly, multiple TEs can form chromatin loop anchors in a species-specific manner: in human, LTR, LINE and DNA transposons mostly contribute to CTCF anchors, while in the mouse SINEs, and particularly the B2 SINE family, are the main contributors [264]. Interestingly, the ChAHP complex (a protein complex constituted by the chromatin remodeler CHD4, the transcription factor ADNP and heterochromatin-binding protein HP1) binds at younger, less divergent SINE B2 elements and competes with CTCF for binding, buffering the genome architecture rewiring, associated with SINE B2 expansion in mice [268]. Most TE-derived CTCF anchors are cell-type specific, showing the potential of TEs to influence cell-type specific expression programs. TE-derived anchors are also hypomethylated, consistent with the fact that CTCF only binds unmethylated DNA.

In hominid pluripotent stem cells, HERV-H elements have been shown to be able to form TADs [269]. Deletion of HERV-H sequences induces the loss of their corresponding TADs and leads to a reduction of transcription of upstream genes. Conversely, the insertion of novel HERV-H copies is able to form new TADs. Repression of HERV-H transcription induces TAD loss, suggesting an importance of HERV-H expression in TAD formation [269]. In the human genome, insulators can also arise from MIR retrotransposons, but in a CTCF-independent manner [270]. They are characterized by an RNA Pol III transcription and various histone modifications that can directly impact chromosomal organization.

In mouse, the SINE B2 repeat has been linked to organogenesis through its dynamic insulator activity [271]. Bidirectional transcripts of a SINE B2-derived sequence located upstream of the murine growth hormone gene (gh) are synthetized using both Pol II and Pol III promoters. These transcripts act as boundary elements by perturbing chromatin structure and inducing chromatin modifications, resulting in a change from heterochromatin to a permissive euchromatic state in this region. This transcription is both tissue- and time-specific and is responsible for the developmentally controlled expression of the gh gene, which promotes pituitary gland development [271]. SINE B1 elements also have insulator properties and can form heterochromatic barriers [272, 273]. It has been shown that B1 transcripts influence the chromatin state of proximal genes between embryonic stem cells and fibroblast cells, suggesting a primordial role of B1 elements in cell differentiation.

In addition to insulators, local chromatin structure is influenced by so called super-enhancers, which correspond to clusters of enhancers associated with Mediator complexes (transcriptional coactivators) that trigger the tissue-specific expression of genes [274]. A novel group of lncRNAs has recently been shown to interact with super-enhancers. These “super-lncRNAs” are able to form RNA:DNA:DNA triplex structures at specific sites within super-enhancers. Interestingly, approx. 40% of super-lncRNA binding sites in super-enhancers overlap with TEs, with SINEs and particularly Alu elements being the major contributors [274]. Moreover, it has been demonstrated that some lncRNAs can act as platforms interacting with several proteins and DNA [275]. For example, ** demonstrates functional domains in the noncoding RNA **st. Proc Natl Acad Sci U S A. 2001;98(16):9215–20." href="/article/10.1186/s13100-020-00229-5#ref-CR277" id="ref-link-section-d177312572e2767">277, 278]. Thus, super-lncRNAs can possibly transport major regulators such as transcription factors and Mediator complexes to super-enhancers, influencing chromatin organization and driving surrounding tissue-specific gene expression.

Conclusions

In this review, we present an overview of the multiple TE resources and functionalities that can be co-opted by host genomes (Fig. 4). TEs can be the source of developmental innovations through their recruitment as new coding sequences and new ncRNAs, and by acting as regulatory sequences, even if TEs are probably less active in gene regulation than expected from their abundance in vertebrate genomes [215]. Particularly, TEs have been instrumental to the evolution of brain, placenta, immunity and embryonic development in vertebrates. The pace of TE recruitment in vertebrate developmental program remains to be investigated. According to the developmental gene hypothesis for punctuated equilibrium, developmental regulatory genes essential for organism morphogenesis are extremely conserved and intolerant to mutations, maintaining an equilibrium state [279]. Changes might not be progressive but rather punctuated, this being often due to transposable elements accumulation and co-option as regulatory sequences to give rise to bursts of morphological innovations and species divergence.

Fig. 4
figure 4

Timing of recruitment of selected TE-derived sequences in vertebrate development. Selected examples are summarized in boxes corresponding to the different types of co-option. These examples are plotted with colored dots onto the vertebrate phylogeny, indicating their timing of appearance and phylogenetic distribution (circles correspond to ancestral events with orthologous sequences in the species, triangles correspond to convergent events). Silhouette images from http://phylopic.org.

Full size image

Concerning the formation of new genes, Ohno proposed in 1999 that gene duplication is the main mechanism sha** evolutionary transitions [33]. New genes can also be formed from scratch, but this mechanism is very rare. We show here that TEs are a major source of material for the birth of novel protein-coding and RNA genes. In the absence of events of whole genome duplications, it has been estimated in primates that 53% of new genes originate at least partially from TE exaptation (mostly in primate-specific regions) compared to 24% from gene duplication and 5.5% de novo from non-coding sequences (the origin of the last 17.5% is still unclear) [280]. The contribution of TEs in this process is thus quantitatively important, in addition to the new functions they provide to the genome.

Several characteristics could modulate the propensity of TEs to be exapted. First, the different characteristics of each TE, such as the presence/absence of internal promoters, protein-binding motifs and ORFs encoding proteins with various properties, might favor the domestication of certain families depending on the needs of the host. For instance, ERVs have greater capacities to become gene regulatory drivers than most other TE families [215]. This has been proposed to be linked to the frequent loss of functional internal genes in ERVs, which abolish their transposition ability but leaves LTRs in genomes that can be readily repurposed. ERVs are frequently non-repressed in hypomethylated tissues, this also possibly facilitates their recruitment. Second, the age of the TE sequences might also be of importance. Repressive silencing being relaxed in old TEs, the repression of younger elements in the genome might limit their chance to be recruited by the host. Third, the activity, copy number and diversity of a TE family probably influence its evolutionary potential for the host. Even if low copy number elements can also lead to important innovations, as shown for the Izanagi transposon in the sex determination cascade of the medaka fish [236], high copy number and diversity of TEs might increase the probability of generating an element advantageous for the host at both sequence and localization levels. On the other hand, maintenance of transposition activity and recombination opportunity with other TE copies might hinder the fixation of a beneficial TE-derived sequence at a specific position in the genome. Fourth, the insertion preferences of TEs or the strength of the selection pressure against their maintenance certainly impact their possible recruitment. While TEs inserting or better tolerated in gene-poor regions will probably undergo less counter-selection, they might be often silenced in heterochromatin. On the other hand, TE preferential insertion or tolerance in gene-rich regions might be more frequently deleterious but could also increase the chance of generating a beneficial combination between TE and host sequences [27]. This might for example be the case for Alu elements in primates, which are probably better tolerated than LINEs in gene-rich regions due to their smaller size and therefore more frequently recruited in exaptation processes. The major factor influencing the co-option of a TE is probably the context of its insertion, as proposed for the domestication of the Transib-like DNA transposon at the origin of the V(D)J recombination [281]. A significant part (36.5% in the human genome) of TE-derived genes are positioned head-to-head to a host gene and share with him a bidirectional promoter containing a CpG island [282]. Since CpG islands correspond to open and actively transcribed chromatin regions, these promoters could be targeted by TE insertions and would provide them with a permissive transcriptional context for their expression, favoring the TE recruitment by the host as new transcribed sequences. TE domestication might also be facilitated by an insertion close to a promoter, or when the insertion results in a fusion with a host gene, with the TE possibly benefiting from the regulatory elements of the linked host gene if this gene is expressed in the germ line [64, 283, 284]. Fifth, if a novel TE is acquired by horizontal transfer, it will transiently escape the repression mechanisms of the host, bringing new evolutionary potentialities and recruitment opportunities.

Developmental pathways are closely linked to those causing cancer. Illustrating this, several examples of TE-derived developmental innovations have also been associated to cancer formation. The human syncytin-1 gene, involved in immunomodulation and cell-cell fusion in placenta, is expressed in several cancers such as colorectal and breast cancers, and endometrial carcinoma [285,286,287]. Several genes of the PNMA family have also been implicated in cancers, such as pnma5 or pnma7a, which acts as an oncogene in thyroid cancers [288, 289]. Finally, the RAG1/RAG2 recombinase, which catalyzes the V(D)J recombination, is a driver of the genetic instability linked to lymphoblastic leukemia [290].

To conclude, Barbara McClintock’s initial model [1] is now widely illustrated. In addition to form “controlling elements”, TEs are also a rich source of new host coding and RNA sequences. Most current examples illustrating the role of TE-derived sequences in vertebrate developmental innovation stems from mammals, but it is reasonable to think that TEs play also a major role in the evolution of other vertebrate species, which generally present even a higher diversity of transposable elements compared to mammals [21]. More studies in other vertebrate sub-lineages are therefore needed. For instance, an accumulation of TE sequences in the Hox gene clusters has been recently reported in four species of squamates (green-anole lizard, slow-worm, corn snake and gecko), which contrasts with the extremely conserved structure of Hox clusters in other vertebrates [291, 292]. It has been suggested that these TEs may provide new coding and non-coding regions or novel regulations of transcription to the cluster genes. The emergence of such elements inside the Hox clusters may explain the observed morphological diversity of squamates, but this hypothesis must now be tested at the functional level [292, 293]. The accurate characterization of the whole mobilome of multiple and divergent vertebrate species, i.e. the accurate and complete genome-wide identification and annotation of TEs and TE-derived sequences in genomes along with their evolutionary and functional characteristics, is an ongoing challenge that will allow to better assess the impact of TEs on vertebrate evolution.