2. Retrotransposons and Their Impact on Mammalian Genome Evolution

2.1. Retrotransposons in Mammalian Genomes

TEs are repetitive DNA sequences, typically ranging in length from 100 to 10,000 bp, capable of colonizing new genomic locations with copies of themselves. Based on their mode of transposition, with or without an RNA intermediate, TEs are split into two major classes: the eukaryote-specific class I TEs, or retrotransposons, which mobilize via a “copy-and-paste” mechanism involving reverse transcription of an RNA copy of a source element, and the class II TEs or DNA transposons, mobilizing without reverse transcription mainly via a “cut-and-paste” mechanism [4]. The increase in TE copy number in genomes is primarily due to their capacity for vertical inheritance through the germline, a property giving TEs great potential for the generation of evolutionary novelties. Indeed, an enormous amount of studies since their discovery has mitigated the emphasis on TEs as a useless form of parasitic DNA, by revealing their irreplaceable contribution to genome structure, information content and regulation ^[19][20][19,20], thus substantially corroborating early, far-seeing hypotheses about their role ^[21][22][21,22] (see [23] for a historical perspective). The other side of the coin is that TE mobility has the potential to disrupt functional genetic elements in both germinal and somatic cells, thus leading to disease ^[24][25][24,25]. It is thus not surprising that retrotransposons coevolve with mechanisms counteracting retrotransposition ^[26][27][28][26,27,28]. Each class of TEs comprises different clades/superfamilies and families, the diversity and complexity of which has prompted decades-long identification and classification efforts ^[29][30][29,30]. In particular, class I TEs are typically subdivided into long terminal repeat (LTR) and non-LTR retrotransposons, with the former displaying a close relationship with retroviruses and other reverse-transcribing viruses (Figure 1). TEs of both class I and II can further be classified as either autonomous or non-autonomous, the former having the ability to self-mobilize, the latter relying on the enzymatic machinery of other TEs for mobilization. Such a distinction is especially relevant in the case of non-LTR retrotransposons, which can be broadly divided into the autonomous elements referred to as long interspersed elements (LINEs) and the non-autonomous short interspersed elements (SINEs) [4]. SINEs are known to exploit the LINE retrotransposition machinery for mobilization, often facilitated by 3′ end sequence similarity between LINE and SINE partners ^[31][32][31,32]. Although LTR and LINE retrotransposons are transcribed by the RNA polymerase II machinery, assisted by a few sequence-specific transcription factors (TFs), most SINEs are transcribed by RNA polymerase III due to the presence of internal control regions (A box and B box) recognized by the Pol III-specific basal transcription factor TFIIIC ^[33][34][33,34]. Indeed, the evolutionary origin of most SINEs has been traced back to Pol III-transcribed genes coding for abundant small RNAs, such as tRNA, 5S rRNA and 7SL RNA, all employing TFIIIC as a sequence-specific DNA binding protein essential for transcription complex assembly [35]. Accordingly, SINEs are generally divided into SINE1/7SL, SINE2/tRNA and SINE3/5S (Figure 1), to which the more recently identified SINEU (derived from U1 and U2 snRNAs) have been added [29]. In the last decade, mainly due to high-throughput sequencing and TE annotation advancements, TEs have been disclosed as a major component of vertebrate genomes, strongly contributing to their diversity [36]. In mammals, which are the best-studied vertebrates in terms of TE biology, the mobilome (defined as the whole set of TEs in the genome) generally display distinguishing features among vertebrates. TEs account for more than 50% of the size of many mammalian genomes, with a preponderance of retrotransposons, which are present in extremely large copy numbers, a minimized content of DNA transposons and a low subfamily diversity compared to other vertebrates [36]. For example, in the two most intensively studied mammals, humans and mice, DNA transposons represent approximately 1.2–3% of the genome sequence, to be compared with the 40–45% of retrotransposons. Among the latter, LINEs contribute 20–22% of the genome sequence in both humans and mice, with LINE-1 (or L1) being the most abundant subfamily, contributing ~17% of the genome sequence. The relative abundance of SINEs and LTR retrotransposons differs markedly between the two mammals, however. In humans, LTR retrotransposons and SINEs represent ~8% and ~13% of the genome, respectively, whereas in mice the corresponding values are 12% and 8%, respectively ^[37][38][37,38]. The L1 family, perhaps the most evolutionarily successful retrotransposon family in mammals, has been a resident of their genomes since early in mammalian radiation, and is likely to have undergone recurrent cycles of adaptation and innovation, leading to the persistence of a single successful lineage [39]. In contrast, SINEs, whose expansion is free from the need to code for a retrotransposition machinery, did not expand continuously from a single, evolutionarily successful family. Instead, novel SINEs have arisen multiple times in the evolution of mammals and, more broadly, of vertebrates [36]. The diversity of lineage-specific SINE families that emerged during mammalian evolution is evident when looking at the distribution of SINE families across major mammal groups (see, for example, [13]). In particular, the most numerous human SINEs, represented by the SINE1/7SL Alu elements (~1.1 × 10⁶ copies), are primate-specific. A comparative study of mobile element insertions in human and great ape genomes revealed that, during recent human/great ape evolution, the most variable form of genetic variation is represented by Alu retrotransposition, with remarkable increases and decreases occurring over very short evolutionary times [40]. The recently discovered, composite SVA elements—evolutionarily young, SINE-derived retrotransposons which include subfamilies restricted to the human lineage—are also specific to primates ^[1][41][42][1,41,42]. In mice, the most numerous SINE family is represented by the 7SL-derived, monomeric B1 elements (~5.6 × 10⁵ copies), closely followed by the tRNA-derived B2 elements (~3.5 × 10⁵ copies) and by the B1-and tRNA-derived B4 elements (3.9 × 10⁵ copies) ^[37][43][44][37,43,44]. More generally, drastically different retrotransposon landscapes characterize the genomes of even closely related taxa, and speciation events and the expansion of new retrotransposon families are often correlated, all pointing to the mobilome as a driver of organism diversification [45]. This does not exclude the existence of conserved retrotransposon subfamilies that appeared at very early stages of mammalian evolution, such as the tRNA-derived SINEs referred to as mammalian-wide interspersed repeats (MIRs), which were actively propagating prior to the radiation of mammals and before placental mammals separated. Although they are retropositionally inactive, MIRs still represent the second most numerous SINE subfamily in humans ^[46][47][48][46,47,48]. Concerning LTR retrotransposons, as mentioned above, they derive from ancestral retroviral infections sustained by exogenous retroviruses that have now gone extinct, except for a few exceptional examples of ongoing endogenization [49]. Having originated from proviral integrations, these elements display a typical retroviral structure—presenting two LTRs that flank the three main genes gag, pro-pol and env—and are hence also named endogenous retroviruses (ERVs) (Figure 1). ERVs are present in all vertebrate genomes, constituting around 10% of the diverse species’ DNA, and have provided important contributions to their hosts over the course of evolution [50]. As suggested by comparative studies, the numerous ERV lineages found in modern mammal genomes arose from multiple independent events of genome invasion, also affecting the host germline, followed by the vertical inheritance of ERVs as host alleles. As a relevant number of such events occurred after the divergence of mammalian orders, each mammalian order tends to have its own distinct ERV content, composition and history, with some ERVs being unique even to individual genera or species, and the same diversification trend also applies to vertebrates as a whole ^[51][52][51,52]. In the case of primates, for example, lineage-specific ERV insertions have been observed in the genomes of African great apes that are absent from human and Asian ape genomes [53]. Even though the retrotransposition activity of human ERVs (HERVs) is presently very limited or absent [54], there is growing evidence that HERVs are widely expressed in human tissues, even in the absence of protein production, which has led to an intense study of their possible roles in human pathologies, including cancer, autoimmune disorders and infectious diseases ^[55][56][57][55,56,57]. ERVs are usually divided into three classes based on their affinity to exogenous animal viruses: class I (gammaretrovirus- and epsilonretrovirus-like), class II (betaretrovirus-like) and class III (spumaretrovirus-like). Concerning individual ERV group classification, it is still incomplete—also due to the relatively recent availability of assembled genome sequences for many vertebrates—and sometimes controversial, given that ERVs are not always named based on phylogenetic and taxonomical criteria. A recent work performed on the human genome with the software RetroTector employed a multi-step classification approach, identifying ~3300 reasonably intact HERV loci that were divided in 31 taxonomical groups, plus 39 “non-canonical” clades showing high degrees of mosaicism and recombination events [58]. Such a comprehensive genomic analysis, complemented by the available detailed characterizations of individual HERV groups (see, for instance, ^{[59][60][61][62]}[59,60,61,62]), represent an ideal background to evaluate HERV expression in human tissues and its variation in diseased contexts [63]. The high proportion of retrotransposons in mammalian genomes, exceeding 90% of all TEs in humans and 95% in mice and rats, together with the presence of at least one family of currently accumulating retrotransposons in most mammals [38], has attracted the greatest attention onto this TE class. An impressive body of studies in the last two decades have addressed the role that retrotransposons played in mammalian, and particularly in human, evolution by facilitating the appearance of genomic novelties. As new evidence accumulated, authoritative reviews were published at various times, covering in great detail the genomic impact and the different emerging aspects of retrotransposons ^{[6][8][13][37][38][52][54][64][65][66][67][68][69][70][71][72][73][74][75][76][77][78]}[6,8,13,37,38,52,54,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78]. In a nutshell, it is thought that retrotransposons contributed to the generation of genomic novelties in two main ways: (i) indirectly, through the promotion of genomic rearrangements; and (ii) directly, through exaptation of retrotransposon-derived sequences.

2.2. Retrotransposons as Drivers of Genomic Rearrangements

Their indirect contributions to genomic novelties derive from the fact that retrotransposons, due to their high copy number and high sequence homology within families, are a relatively frequent substrate of unequal recombination events producing gene and/or exon duplication, shuffling or deletion. As an additional mechanism, retrotransposons sometimes carry flanking genomic sequences with them (a process referred to as 5’ or 3’ transduction) thus potentially introducing new copies of genes/exons into new locations [54]. The generation of retrogenes is a further indirect consequence of the presence of autonomous retrotransposons, the machinery of which may be exploited by mRNAs or other RNAs to generate new copies of their coding sequences [7]. Independently from the mechanisms of their generation, new gene copies have great potential for neofunctionalization favoring phenotypic evolution [79], a property that, contrary to what has been prevailingly thought, is likely to also apply to pseudogenes and retropseudogenes [80].

2.3. Retrotransposon Exaptation as a Source of Genomic Novelties

The exaptation of retrotransposed sequences, consisting in their cooption for a current function out of a hitherto neutral evolution mode, is a well-documented phenomenon ^[2][74][2,74]. In general, large-scale DNA editing of retrotransposons, by simultaneously generating large numbers of mutations, may have accelerated their exaptation during mammalian evolution [81]. In a similar vein, inverted SINE repeats being part of longer RNAs may have promoted RNA editing by adenosine to inosine deamination, thus generating potential novelties in both coding and regulatory sequences [82].

For simplicity, two major exaptation modes can be distinguished. According to the first mode, retrotransposon-derived sequences become physical and functional parts of transcription products, even being eventually translated into protein sequences. The second mode consists in the co-optation of retrotransposon-derived sequences as transcription regulatory elements or 3D genome boundary elements. This exaptation mode, allowing retrotransposon sequences to exert their influence without becoming incorporated into gene products, might have had an even wider influence on genome evolution ^[83][84][83,84].

2.3.1. Retrotransposon-Derived Sequences within Gene Products

As to the first mode of action, there is solid evidence that SINE (in particular, primate-specific Alu) exonization contributes to both untranslated and protein-coding regions of mRNAs [85], as well as portions of long noncoding RNAs [86], to which the embedded Alu can confer new regulatory functions [87]. Alu-derived exons are often the site of alternative splicing, due to the presence in the Alu body of multiple cryptic splice sites [88]. SVA-mediated transduction events, involving alternative mRNA splicing at cryptic splice sites, have been found to promote exon shuffling and thus genomic novelty [89]. At the same time, cells have evolved precise mechanisms to control Alu and the incorporation of other retroelements within mRNA sequences via their cryptic splice sites, as their incorrect presence might induce devastating physiological responses [90]. Moreover, exonic SINE sequences embedded into the 3’ UTR of mRNAs participate in different layers of post-transcriptional gene regulation, which may also involve intermolecular base-pairing with SINE sequences embedded in lncRNAs ^{[43][88][91][92]}[43,88,91,92]. Alu SINEs embedded into precursor transcripts were also found to promote the formation of circRNAs [93], a complex family of eukaryotic regulatory transcripts under intense study ^[94][95][94,95]. There is also abundant evidence for TE-derived microRNAs, some of which are potentially involved in human evolution and disease ^{[96][97][98][99][100]}[96,97,98,99,100].

In the case of autonomous retrotransposons, which contain protein-coding sequences in their body, several striking cases of exaptation of retrotransposon-encoded proteins as new host proteins have been documented. A remarkable example is represented by syncytins, an ensemble of Env proteins coded by different ERVs in the genome of various vertebrates, that through a process of convergent evolution led to the development of the placenta in eutherian mammals ^[55][101][55,101]. In fact, the union between maternal and fetal cells to constitute the placental syncytiotrophoblast—the main site of trophic exchanges during pregnancy—is mediated by the fusogenic activity of syncytins, which changed from being mechanisms of viral entry to exerting physiological activity domesticated to serving the host biology [55]. Some syncytins are indeed thought to have a role in other placenta-associated functions, such as the establishment of maternal immune-tolerance against the fetal allograft through their natural immune-suppressive properties, which in ancestral infections likely guaranteed their immune escape ^[102][103][102,103].

2.3.2. Retrotransposons as a Source of cis-Regulatory Sequences

Given the centrality of cis-regulatory elements, and particularly of enhancers, in orchestrating organ-, tissue- and cell type-specific gene expression both during development and in adult organisms [104], it has been argued that the “vast majority of the genetic changes responsible for the evolution of morphology occur at pre-existing cis-regulatory elements” [105], and that TE-mediated cis-regulatory network rewiring has been one of the key mechanisms for the appearance of such changes [6]. In the last 10–15 years, the exaptation of TE-derived sequences (especially retrotransposon-derived) as cis-regulatory elements has been well documented by a rapidly growing body of studies, the majority of which have focused on mammalian genomes, characterized by the overwhelming prevalence, in terms of both amount and activity, of retrotransposons over DNA transposons. Retrotransposon-derived cis-regulatory sequences have been reported to play several roles in gene regulation as promoters, enhancers, silencers and boundary elements ^[2][83][2,83]. In general, due to their own replicative needs, retrotransposons have evolved cis-acting sequences mimicking those of the host, a fact that predisposes them to cis-regulatory activity [76]. Although rwesearchers are still far from a comprehensive picture of the multiple layers of TE-derived regulatory novelties and their integration with the whole genomic background of mammalian evolution, various cis-regulatory modes of TE exaptation have begun to be clearly portrayed (Figure 2).

First of all, many binding sites for diverse TFs are contributed by retrotransposons, as mainly revealed by genome-wide TF occupancy mapping by chromatin immunoprecipitation coupled with high throughput sequencing (ChIP-seq) [106]. Although some of the TF binding sites carried by TEs are justified by their need to employ host TFs for their own life cycle, others may have been acquired independently through TE propagation mechanisms [34]. Molecular evolution studies have revealed waves of expansion of the TF target repertoire over the course of vertebrate evolution, with TEs majorly contributing to such expansions [107]. TFs tend to bind to TE-provided cognate sites in a species-specific manner, in line with the expansion of different TE subfamilies at different evolutionary timepoints [83]. A striking example of how the evolutionary recruitment of TE-derived TF binding contributed to mammalian evolution is provided by the TE-dependent transformation of the uterine regulatory landscape in the evolution of mammalian pregnancy [108]. An emerging topic that is potentially highly relevant to the exaptation of TE-binding TFs, is that of Krüppel-associated box domain zinc finger proteins (KRAB-ZFPs). The great expansion and diversification in mammals of these TFs has been correlated with the invasion of new endogenous retroelements, which require specialized mechanisms of repression via the binding of specific KRAB-ZPs and subsequent recruitment of the KAP1 corepressor [28]. It is thought that the arms race between KRAB-ZFPs and their target retroelements, facilitated by the evolutionary plasticity conferred on both contenders by the repetitive organization of their genes, favored retroelement domestication, allowing them to develop cis-regulatory functions, to which KRAB-ZFPs have the potential to directly contribute as enhancers or promoter-binding TFs ^{[28][71][109][110]}[28,71,109,110].

A second, more complex mode of TE exaptation for cis-regulatory purposes is represented by TE-derived clusters of TF binding sites, exemplified by the contribution of species-specific, composite enhancers to mouse placental development by rodent endogenous retroviruses [111]. In addition, mouse-specific LTRs have been found to carry multiple pluripotency TF-binding sites (specifically, ESRRB-, KLF4- and SOX2-binding motifs) regulating gene expression in a mouse embryonic stem cell (ESC)-specific manner, thereby distinguishing ESCs in mice from ESCs in other species [112]. In a similar vein, recent hominoid-specific LTR and SVA retrotransposons were shown to host enhancers that were active in human naive ESCs and embryonic genome activation [110]. Systematic studies of TEs’ contribution to enhancer function have benefited greatly from high-resolution profiling of the regulatory epigenome, such as the profiling of DNase hypersensitivity, histone H3-lysine 4 mono-methylation (H3K4me1) and histone H3-lysine 27 acetylation (H3K27ac) as typical enhancer chromatin signatures [113] and by the use of a chromatin characterization software such as ChromHMM [114]. A recent comprehensive quantification of the epigenomic status of TEs across many human tissues and cell types revealed that approximately one quarter of the human regulatory epigenome is composed of retrotransposed sequences, with motif-enriched LTRs being particularly favorable substrates for the evolution of new host regulatory elements [115]. In other studies, based on epigenomic profiling, evolutionary novelties in primate gene regulation were similarly found to have TEs as the primary source, with a major contribution from ERV-derived sequences ^[116][117][116,117]. Accordingly, a subset of ERV sequences were found to be significantly enriched in cis-regulatory elements, having a critical role in primate liver gene regulation [117]. A fascinating example of ERV contribution in the shaping of entire regulatory pathways is represented by the interferon (IFN) transcriptional network, a crucial innate antiviral system which also serves as a fundamental effector to initiate and maintain adaptive immunity. Chuong and coauthors showed that ERV insertions had a central role in its evolution and amplification, accounting for the independent dissemination of a wide number of IFN-inducible enhancers in many mammalian genomes, which are required for the correct functioning of different immune responses [118]. A similar scenario is found for p53 tumor suppressor factor, of which the genomic binding sites in humans overlap in more than one-third of cases with ERV elements [119]. Of note, these binding sites are primate-specific and not present in other mammals, further demonstrating that TEs are able to shape important regulatory networks in a species-specific manner. An intriguing observation, consistent with the previous ones, is that of the pervasive function of an ape-specific class of ERV-derived LTRs, LTR5HS, as early embryonic enhancers, regulating hundreds of human genes [120], and the strong contribution of ERV and L1 retrotransposon families to species-specific differences in enhancer activity between chimpanzee and human cranial neural crest cells ^[83][121][83,121]. Epigenome profiling also allowed researchers to distinguish between older retrotransposon copies displaying most of the features of de facto enhancers and younger copies that seem instead to be configured as proto-enhancers, serving as a repertoire for the de novo evolutionary birth of enhancers [122]. Despite the scarcity of studies, an intriguing retrotransposon feature favoring their exaptation as enhancers is their intrinsic capability of generating functional non-protein-coding RNAs (ncRNAs) that could overlap with the so-called enhancer RNAs (eRNAs) [123], thereby raising the possibility that many eRNAs could be generated through TE-derived ncRNAs.

2.3.3. Involvement of Retrotransposons in Three-Dimensional Genome Architecture

Chromosome contacts within the nuclear space, recently revealed at unprecedented resolution by HiC and complementary approaches [124], exert a wide and still largely unexplored influence on gene regulation by demarcating regulatory districts in a highly dynamic way. At a large scale within nuclei, chromosomes segregate into regions of preferential long-range interactions that form two mutually excluded types of chromatin, referred to as “A” and “B” compartments [125], the formation of which has been recently linked to homotypic clustering of L1 and B1/Alu, respectively [126]. At a scale of tens to hundreds of kilobases, chromosomes fold into domains with preferential intradomain interactions known as topologically associating domains (TADs), which harbor the potential to influence enhancer function and thus gene regulatory networks ^{[127][128][129][130][131][132]}[127,128,129,130,131,132]. TAD demarcation is achieved by specific regions called TAD boundaries, which are enriched for the occupancy of CCCTC-binding factor (CTCF), a zinc finger DNA binding protein also known to mediate the formation of chromatin loops [133]. SINE retrotransposons have also been found to be enriched at TAD boundaries ^[134][135][134,135]. Curiously, in rodents (but not in humans) B2 SINE retrotransposons have been shown to carry CTCF binding motifs, and therefore rodent B2 SINEs can contribute to clustered CTCF sites at TAD boundaries, thus helping in the maintenance of genome organization [136]. However, the rapid expansion of rodent SINEs might provide excessive CTCF sites throughout the genome, therefore critically increasing the possibility of genome mis-folding due to the creation of aberrant CTCF sites. In this context, a complex formed by CHD4, ADNP and HP1 chromatin proteins (ChAHP complex) has been shown to play a role in the maintenance of evolutionarily conserved spatial chromatin organization via the buffering of novel CTCF binding sites that emerge through SINE expansion [137]. Moreover, SINE and other retrotransposons have been proposed to participate in the establishment of species-specific chromatin loops by introducing novel binding sites for architectural proteins, including CTCF [138]. CTCF might also participate, together with other proteins, in the DNA methylation and histone modification boundary activity recently attributed to currently active copies of mouse B2 SINEs, which might be involved in the epigenomic and phenotypic diversification of mouse species [139].

The contribution of retrotransposons to chromatin regulatory domains is not limited to providing CTCF binding clusters. MIR retrotransposons, for example, have been shown to provide regulatory sequences, functioning as insulators in the human genome independently from CTCF [140]. The presence of binding sites for the multi-subunit DNA binding protein TFIIIC is a distinguishing feature of SINEs, and TFIIIC bound to Alu elements has been shown to influence gene regulation through its chromatin looping and histone acetylation capacities ^[141][142][141,142]. In the case of SINEs exapted as enhancers or TAD boundaries, their regulatory function might even take advantage of their Pol III-dependent transcription, which was recently demonstrated to occur with a marked cell-type specificity ^{[123][143][144]}[123,143,144]. Retrotransposon transcription has also been shown to be required for the cell type- and species-specific chromatin architecture remodeling properties recently attributed to the primate-specific HERV-H TE family of LTR retrotransposons [145].