Transposable Element DNA and RNA: Drivers of Gene Expression, Evolution and Disease

Summary

Transposable elements (TEs) comprise nearly half of mammalian genomes and have shaped genome architecture, chromatin organization and transcriptional landscapes. Thanks to recent advances in long-read sequencing and functional (epi)genomics, the focus has shifted from TE families to individual TE loci, revealing widespread, locus-specific regulatory roles. While most TEs have lost the capacity to mobilize, they still retain a DNA form and, when transcribed, an RNA form, both of which can affect genome regulation. TEs are alternative promoters, exons, splicing regulators, and 3′-end modulators. They can also act as enhancers, drive 3D genome organization and give rise to lncRNAs that are platforms for transcriptional and chromatin regulators. Mechanistically, TE repression involves DNA methylation, histone modification, phase-separated condensates, RNA modifications, RNA degradation and nuclear compartmentalization, yet this repression can be selectively lifted during development or stress to expand regulatory potential. TEs therefore contribute to cell-type identity, developmental transitions and responses to environmental stimuli, while their dysregulation is linked to human disorders including neurodegeneration, cancer and autoimmune disease. TEs also hold translational promise as biomarkers and tools for gene and cell engineering. In summary, the pervasive integration of TEs as mini-genes, structural scaffolds and regulatory elements redefines our view of the genome: rather than a gene-centric landscape dotted with repetitive “junk”, mammalian DNA is a TE-rich ecosystem in which TEs drive gene regulatory networks and evolution.

Marazzi in brief CELL-D-26-00423

Transposable elements, far from being genomic clutter, have emerged as are not genomic dynamic regulators that shape when, where, and how genes are turned on across development and disease.

The history and rediscovery of “controlling elements”

The discovery of transposable elements (TEs) is one of the most transformative moments in genetics. In the late 1940s, Barbara McClintock, working with maize cytogenetics, observed that certain loci were not static as previously assumed but capable of moving across the genome. She documented how these “controlling elements” could insert near genes and modulate their expression, producing striking phenotypic changes in kernel pigmentation and plant development¹. At the time her ideas were met with skepticism, as they challenged the prevailing dogma of a fixed genome. For decades, McClintock’s findings were dismissed as botanical curiosities until the discovery of mobile DNA in bacteria in the 1960s, which confirmed that DNA mobility was a general biological phenomenon². With the advent of molecular cloning and genome sequencing in the 1970s and 1980s, transposable elements were revealed to be ubiquitous across domains of life. McClintock was awarded the Nobel Prize in 1984³.

The sequencing of the human genome in 2001 marked a turning point as nearly half of our DNA was shown to consist of TE-derived sequences⁴. While initially dubbed genetic “junk,” many TEs have been used (exapted) by their host, and thus have emerged as a force shaping genome architecture and gene regulation. Today, TEs are understood as double-edged swords. They are capable of threatening genomic integrity through mutagenesis and instability, but also serving as sources of genetic novelty and engines of regulatory innovation. In this review, we begin by providing a broad overview of mammalian TE biology and discuss how new technologies are redefining TEs as fundamental “genic” units, a stark move from the analysis of TE as families. We then examine how single TEs influence transcription and co-transcriptional processes, thereby shaping gene regulatory networks. We highlight emerging mechanisms of TE silencing, de-silencing and co-option. We also make distinction, when possible, of their regulatory DNA form vs transcribed RNA form. Finally, we summarize the roles of TEs in human disease and explore emerging strategies to target them for therapeutic benefit in diverse pathological contexts. While we highlight some aspect of TE mobilization linked to somatic mosaicism in specific tissues, we direct the readers to several excellent reviews^5-11 on the mechanisms of TE mobilization, a key aspect of TE biology.

Part 1: Abundance, Evolution and Diversity in mammalian lineages

Perhaps the most striking feature of TEs is their variety (Box 1) and abundance in mammalian genomes. TEs occupy between 27% to 74.5% of mammalian genomes (average ~45.6%) and genome size correlates positively with TE content. Across most taxa, LINEs and SINEs are most abundant: on average, LINEs constitute about 22.6% of mammalian genomes (ranging from 8.2% to 52.8%), while SINEs account for approximately 10.5% (ranging from 0.4% to 32.1%). Other TE classes are less common, including LTR retrotransposons (mean 7.8%, range 2%-17.8%), cut-and-paste (mean 3.5%, range: 0.5%-8.4%) and rolling circle (mean 0.5%, range: 0.01%-19.7%) DNA transposons¹². A “real life” example of the pervasiveness of TEs in our genome is provided in Figure 1, where genomic regions containing a protein coding gene, centromeric and telomeric regions of chromosome 6 are displayed with the relative abundance of TE types.

Box 1. Classification of Mammalian Transposable Elements

TEs in mammals are broadly divided into two major classes, Class I retrotransposons and Class II DNA transposons, distinguished by the type of replication intermediate they employ. They can be further divided into autonomous and non-autonomous elements. Autonomous elements encode all the proteins required for their own (retro)transposition, while non-autonomous elements must parasitize on the machinery of other transposable elements to move around the genome.

Class I retrotransposons

Class I retrotransposons use a “copy-and-paste” mechanism, in which an RNA intermediate is reverse-transcribed into DNA and then integrated into the host genome. Based on their genetic structures and mode of integration, they are subdivided into long terminal repeat (LTR) and target primed non-LTR elements^10,251. They are abundant throughout mammalian genomes.

The LTR class of elements includes endogenous retroviruses (ERVs), likely originating from ancestral germline retroviral infections. Some ERVs remain “intact”, preserving the canonical genetic architectures of integrated proviruses. These carry two 300-1000 bp LTRs flanking a single polycistronic viral open reading frame that encodes the viral Gag, Pol and Env proteins²⁵². The LTRs bear transcriptional and post-transcriptional functions, including promoter activity, polyadenylation and RNA packaging signals, as well as the tRNA primer binding site (PBS) and polypurine tracts required for reverse transcription^10,252,253.

Most ERVs in the mammalian genome have lost replication competence due to accumulations of mutations over evolutionary time^253,254. In mice, only a single copy of the intracisternal A type particle (IAP) remains capable of producing functional retrovirus¹⁰ while other IAP copies propagate via intracellular retrotransposition. In some cases, viral coding regions are mutated, lost (particularly at env)^10,252 or supplanted by host-derived sequences²⁵². Such mutations/deletions are thought to arise via read through transcription, or recombination. Provided such elements retain the appropriate RNA packaging signals, such “defective” ERVs may still propagate in the host genome. Representative examples include the early transposons (ETns)^255-257 which rely on the machinery of the THE1 mammalian apparent LTR retrotransposon (MaLR) elements^140,258 to propagate. In addition, homologous recombination between the 5’ and 3’ LTR of ERVs frequently deletes internal coding regions to produce a solitary LTRs²⁵⁴. Solo LTRs are by far the most abundant LTR-derived sequences in eukaryotic genomes, outnumbering intact ERVs by roughly an order of magnitude. Because solo LTRs can retain regulatory activity, their presence may modulate the function of nearby genes and enhancers, an aspect we will examine in greater detail later in the review.

Non-LTR retrotransposons include Long Interspersed Nuclear Elements (LINEs) as well as the non-autonomous Short Interspersed Nuclear Elements (SINEs). Full length LINEs typically encode two open reading frames (ORF1 and ORF2), flanked by 5’ and 3’ untranslated regions (UTRs). In active elements, the 5’ UTR typically contains internal promoter sequences recognized by host RNA polymerase II²⁵⁹. LINEs propagate via target-primed RT reaction, wherein reverse transcription of the template viral RNA occurs directly at the site of integration. In this process, RNA-bound ORF2p first nicks the target DNA (typically at 5’-TT/AAAA-3’ sites) and then uses the 3’ end to prime RT of the viral RNA, finally directly integrating it into the newly synthesized complementary DNA strand^10,259,260. Because reverse transcription often prematurely terminates, LINE inserts are frequently truncated at their 5’ termini, resulting in loss of promoter activity and autonomy. However, promoter replacement can occur, especially if the element is integrated close to another functional regulatory sequence from the host, or another TE. For instance, R2 elements in insects lack their own endogenous promoters, but rely on co-transcription with the host rRNA repeat for expression.

In contrast, SINE elements (typified by Alus in primates) are non-autonomous elements that encode no proteins of their own, instead parasitizing the RT machinery of LINEs. SINEs typically span 100-600bp. These elements are also transcribed by RNA polymerase III, and the 5’ terminal of SINE elements structurally resemble known RNA polymerase III targets (tRNA, 7SL RNA or 5S RNA). To replicate, transcribed SINE RNA must recruit LINE RT for reverse transcription and integration. To overcome the cis-preference of the LINE RT for its own transcripts, SINE transcripts form complexes with host factors that promote their association with polyribosomes and nascent expressed LINE ORF2p. In addition, about 20% of SINEs carry a region of similarity with the 3’ terminal sequence of LINE. These sequences are thought to facilitate recognition by the RT of some LINEs.

Class II DNA transposons

In mammals, most class II DNA transposons rely on a “cut-and-paste” mechanism to move to new genomic locations, wherein transposases are utilized to excise the transposon and re-integrate it at new sites. Such transposons rely on either the DDE transposases or tyrosine recombinases (YR)^10,261. Autonomous cut-and- paste transposons are typically found with terminal inverted repeats (~10bp – 1kb long) that flank the coding sequences of the transposase and typically rely on host RNA polymerase II for expression. Common cut-and- paste transposons that are still found active in mammals include the Tc1/mariner, hATs and piggyBac families^10,261, which utilize DDE transposases. Active YR-mobilized transposons, Cryptons, have not been identified in mammals^262,263. However, remnants of these TEs are still present in mammalian genomes. For instance, domain of unknown function 3504 (DUF3504) containing genes are thought to have arisen from ancient domestication events of Crypton TEs, that have resulted in the loss of their YR activity. While they are proposed to play roles in transcription regulation or protein-protein interactions^262,263, It is currently still unclear how these domains exactly impact protein function.

Another class of DNA transposons found active in mammals, Helitrons, are thought to replicate by a rolling circle mechanism, facilitated by an encoded Rep/Hel protein. Helitrons typically contain distinct ~150bp terminal sequences. During replication, the Rep/Hel protein cleaves and covalently binds to the sense strand at the left terminal sequence (LTS). DNA replication is initiated at the free 3’-OH and facilitated by sense strand displacement by Rep/Hel helicase activity. At the right terminal sequence (RTS), further cleavage and strand transfer results in the generation of an excised circular intermediate containing an RTS-LTS junction. This intermediate is then used for transposition into new sites in the genome^264,265. Within mammalian species, Helitrons have only been detected within the Vespertilionidae bats.

TE abundances in the human genome

Both Class I and Class II TEs are abundantly found within the human genome. The below figure illustrates the genomic coverage of different TEs classes and the genomic structures of the most abundant TE classes (LTRs, SINEs, LINEs) within the human genome.

Box 1 Figure 1. Transposable Element Integrations in the Human Genome.

(A) Proportion of the TE-integrations in the human genome belonging to each class of TE by genomic coverage

(B) Number of duplications in the human genome for each TE class

(C) Features and ORFs of SINEs, LINEs and LTRs

(D) Number of duplications in the human genome of the 5 top families and 15 individual TEs for each class.

Figure 1. Transposable elements are heavily integrated into the mammalian genome and transcriptome.

(A) The OCT4 gene locus overlaps several TEs; (B) TE-integration sites in the centromere of chromosome 6; (C) TE integration sites in a telomeric region of chromosome 6.

A common point of view is that TEs are silenced and relegated to inaccessible region of the genome. This misconception is likely due to a combination of historical and technical reasons: TEs are often grouped together as families which ignores genomic context of individual integrants, and a central focus in the field has been on silencing mechanisms controlling TE expression (see Part 3). In reality, their sheer numbers and in some cases lack of strict preference for integration site mean that TEs populate both accessible DNA regions (euchromatin) where they can influence nearby gene expression, and different flavors of less accessible regions (heterochromatin). There are nonetheless some spatial distinctions among TE locations. For example, primate-specific LINE-1 sequences have limited preference for integration site^13,14 but functionally cluster around lamina-associated domains, anchoring regions of silent chromatin at the nuclear periphery^15,16. They are particularly abundant on the human X-chromosome. ERVs and SINEs are enriched in open chromatin regions, where they provide a rich source of cis-regulatory elements, as describe in greater details in Part 4.

Another key feature of TE biology is their diversity across mammalian lineages. In fact, even when genome TE content is similar amongst mammals, the relative abundance and diversity of individual TE families vary markedly. Comparative genomic analyses indicate a strong phylogenetic component underlying variation in TE composition across mammalian taxa. Variation and acquisition of TEs via horizontal transfer (HT) is rare amongst mammalian families. Differences in TE composition among taxa are largely associated with the rate, number and type of recent TE transposition and expansion events. Notably, expansion events are typically dominated by individual TE types rather than multiple, suggesting lineage- and time- specific bursts of germline transposition activity.

TE diversity across taxa is a consequence of unique evolutionary trajectories that these elements can take once established within their host species. Their capacity to mobilize throughout the genome can inadvertently lead to the recurrent gain or loss of protein coding or regulatory sequences. Such changes may originate from the host itself, or through exchange with other TEs. This penchant for “chimerism” is thought to be a major evolutionary force driving TE diversification. A striking example is the hominoid specific SVA (SINE-R-VNTR-Alu) family of elements. SVAs are composite elements that arose through fusions between Alu retrotransposons, hexamer and variable number tandem repeat (VNTR) sequences and an LTR fragment. Notably, the emergence and sequence divergence of the six SVA subfamilies (A-F) parallel primate evolution and speciation¹⁷. VNTR lengths amongst orthologues differ between species and show a negative correlation with evolutionary time¹⁷. The human-restricted SVA subfamilies E and F remain active, with polymorphism rates ranging from 64% to 93% across different human populations^17,18.

The evolutionary trajectory of TEs within their hosts is also strongly shaped by natural selection pressures acting on host fitness. Given their mobility, TEs can impose a significant fitness cost to their host. This can occur through disruption of gene expression, promotion of ectopic recombination and chromosomal rearrangements, or by diversion of cellular resources and energy. Indeed, de novo TE insertions in the human genome cause or correlate with several human diseases, an aspect that we cover in Part 5. Consequently, hosts have evolved multiple mechanisms to restrict TE activity in all cells and particularly in germ cells (to avoid vertical transfer). Many of these silencing mechanisms are conserved across the taxa, including epigenetic silencing, transcriptional modulation and RNA interference pathways. TEs that ultimately become fixed within the host genome typically do so at loci where their presence exerts neutral or minimally deleterious effects, leading to their enrichment in intergenic or non-coding regions. Over time, many such insertions accumulate mutations and deletions, rendering the TE derived sequences incapable of transposition, although they may still be transcribed by the host. Depending on their genomic context, certain elements can acquire adaptive functions and their DNA form, or transcribed RNA form, can become co-opted as integral components of host regulatory networks.

The abundance of TEs thus redefines our concept of the genome: rather than a gene-centric landscape with some interspersed repeat elements, mammalian DNA is dominated by TE sequences that can, in distinct temporal and/or sequence contexts, act as structural scaffolds, regulatory platforms, and mutagenic threats. Below we will focus on recent insights on four key aspects: i) technological efforts aimed at characterize TE’s positional identity and expression, ii) how TE silencing is executed in the germline and in the soma, iii) how silencing is lifted to permit the expression or cooption of TE as regulatory elements, iv) the role of TEs in individual variation and disease.

Part 2: Technological Advances in Mapping Transposable Elements

TEs may be repetitive in more than one sense of the word: the same or similar sequences may be repeated or interspersed throughout the genome, the nucleotide sequences of TEs may contain low complexity stretches, or both. These repetitive characteristics, along with deviations from reference genomes caused by polymorphisms, make some TEs especially challenging to resolve with functional genomic experiments based on short read sequencing^19,20. TEs can suffer from an intrinsic lack of unique ‘mappability’ (i.e., alignment of reads to individual genomic loci), and may be explicitly ‘blacklisted’ (i.e., excluded) in data processing pipelines. Though TEs are widely transcribed in particular cell types, expression of individual integrants is generally lower than genes, so transcript data may be sparse or noisy. Specialized bioinformatic approaches and an awareness of these challenges are therefore critical to avoid erroneous conclusions when analyzing TEs, especially when mining public repositories of genomics datasets produced with different protocols^20,21. The general mappability problem and some useful potential solutions are visually represented in Figure 2.

Figure 2: Approaches to mapping transposable elements.

Potential solutions to ambiguous mapping of TEs in functional genomics data. Benefits and limitations of each solution are suggested.

Common adoption of paired-end sequencing approaches improves mappability somewhat: as TEs accumulate mutations over time, they are not 100% identical and fragments of several hundred base-pairs can be sufficient to distinguish individual copies^22,23. However, some of the evolutionarily-youngest, polymorphic TE alleles – often the subject of closest scrutiny as potential drivers of genome instability, immune responses or disease – remain invisible to locus-specific analysis even with the longest read lengths afforded by Illumina sequencing, limiting insights of regulation and expression to subfamily level. While this may suffice for certain scientific questions¹⁹, it is now appreciated that expression and regulation of individual TE integrants is highly variable depending on sequence and genomic context in 2- and 3-D^24-29, and the lack of mappability remains an important bottleneck to many functional studies. For example, in the human genome some human-specific LINE-1 (L1Hs), Alu and SINE-VNTR-Alu (SVA) copies are still capable of jumping in the germline, leading to insertional polymorphisms in the population that are not captured by individual reference genomes^7,30,31. SVAs are often not uniquely mappable due to the repetitive sequence of their variable number tandem repeat^18,32. In genomic analysis of inbred mouse strains, high activity of murine L1s and intracisternal A particle (IAP) elements drive polymorphisms that are especially hard to track and control for^23,33.

Analysis of TE transcription, especially with single-cell resolution, presents particular challenges. Signal from readthrough transcription due to activity of neighbouring genes may be impossible to distinguish from TE-driven transcription²¹ unless methods such as CAGE or SMART-seq are implemented to specifically capture the TSS^34-36 And, unlike in DNA sequencing, where read pile-ups typically spread to the flanking genomic regions, signal from transcripts initiated by TEs tend to be confined within the bounds of the TEs themselves and may thus map to multiple genomic coordinates³⁷. Even in pluripotent or transformed cells, which are generally more permissive to TE transcription than differentiated cells due to relative genome hypomethylation^27,38-43, individual TEs are typically expressed at a low level relative to genes and deep sequencing is a general requirement to analyse transcription of individual integrants. In single-cell transcriptomes, the low expression of individual TEs, coupled with the mappability challenges described above (exacerbated by short read lengths of single-cell genomics platforms such as 10X), cause datasets to be especially sparse and may necessitate pseudo-bulk approaches^44-46. A useful strategy is to combine metabolic labelling of nascent TE transcripts⁴⁷ which may allow borders between TEs and downstream genomic regions to be defined. Alternatively, epigenomic measurements such as chromatin accessibility, or profiling histone marks associated with active promoters such as H3K4me3, can serve as orthogonal readouts of TE transcriptional activity^{25,38,45,48-50}. Signals from these approaches, which can be compatible with single-cell formats, often extend into the flanking non-TE regions of the genome, facilitating unique read mapping. Computational tools such as RepeatMasker and numerous other bespoke methods have been critical in the effort to advance locus-specific TE annotation and interpretation. Together, these methods allow researchers to both catalog TE insertions and assess their activity, expression, and regulatory potential at gene- and locus-level resolution.

In the context of these challenges, the advent of long-read sequencing technologies - Pacific Biosciences (PacBio) HiFi reads and Oxford Nanopore sequencing - has revolutionized TE biology. The Telomere-to-Telomere (T2T) human genome assembly, completed in 2022, highlighted this transformation: many previously unresolved gaps in sequence and genome function stemmed from repeat-rich TE regions and their epigenetic patterns^51,52. Long reads may span the entire length of repetitive sequences, enabling precise mapping of TE insertions, structural variants and epigenetic modifications with allelic resolution^31,53. For example, Nanopore DNA sequencing can detect methylation and hydroxymethylation at specific TE loci: given the central role of CpG methylation in silencing TE promoters, such experiments have been central in the effort to resolve silencing mechanisms at individual integrants across development^{27,32,38,54-57}. PCR and accompanying amplification biases are typically avoided, with the caveat that sample requirements remain high to achieve good coverage unless enrichment methods^27,58 or perhaps adaptive sampling protocols⁵⁹ are used. Though base-calling accuracy initially limited Nanopore sequencing, accuracy has gradually improved to approach that of PacBio HiFi reads. Together, ultra-long Nanopore and PacBio sequencing allow assembly of centromeric and telomeric repeats, unveiling the contribution of TEs to heterochromatin structure telomere to telomere⁵². Methods coupling CpG methylation information to readouts of histone variants, histone modifications or chromatin accessibility on single Nanopore reads – such as DiMeLo-seq⁶⁰, nanoNOMe⁶¹ and others⁶² – may prove powerful in characterizing chromatin state of polymorphic or otherwise-unmappable TEs. At the same time, the application of long-read RNA sequencing – either passing RNA directly through the nanopore, or analysing cDNA in ONT or PacBio protocols – can resolve full-length TE-driven or TE-containing transcripts and isoforms. Analysis of full-length isoforms in particular tissues and samples has already led to some surprising discoveries of TE exonization with important functional consequences⁶³. For example, long-read analysis of human tissues and tumors identified how a primate-specific Alu element is exonized to generate a truncated alternative isoform of the interferon alpha and beta receptor subunit 2 (IFNAR2), described further below⁶⁴. Recent efforts such as CELLO-seq⁶⁵ and future iterations thereof, that seek to use long reads to measure transcription of TE alleles in single cells, will continue to empower researchers to discover and characterize TE transcriptional activity. The ever-expanding discovery of TE-derived proteins and transcripts in diverse human tissues and clinical^44,66-68 suggest that we may sit at the tip of the iceberg in discovering species-specific biology and signalling pathways driven by TEs. The transition towards a human pangenome reference⁶⁹ promises to further increase our capacity to understand whether and how individual variation is driven by polymorphic TEs, including those residing in the ‘darkest’ regions of the genome.

Part 3: TE Silencing in Germline and Soma.

Transposable elements pose a potent threat to the integrity of the germline and somatic genome, prompting the evolution of sophisticated and overlapping silencing mechanisms. Recent studies have unveiled novel mechanistic insights that refine this defense mechanism, revealing the regulatory complexity orchestrated by RNA, chromatin dynamics, and nuclear architecture. Notably, one common denominator among most silencing mechanisms is the dependency on the transcription of TE that needs to be silenced. Aside from the important exception of KRAB-ZNF proteins, many of which target specific DNA sequences, it has become apparent that other mechanisms of TE silencing require the nascent TE RNA to be used as a platform to recruit silencing machinery. Different flavors of this RNA-based silencing exist.

3.1. Germline Silencing: piRNA Pathway.

The piRNA pathway is the canonical defense against transposons in male and female germ cells⁷⁰. piRNA clusters generate small RNAs that guide PIWI proteins to silence transposons via transcript cleavage and co-transcriptional recruitment of de novo DNA methyltransferases DNMT3A and DNMT3B (and, in the mouse male germline, DNMT3C)^71,72. The latter was thought to be mammal-specific, though recent studies in axolotl have suggested deeper evolutionary origins⁷³. The latest model in the mouse male germline proposes a two-step recognition system: chromatin factors such as SPIN1–SPOCD1 detect active TE promoters, while MIWI2–piRNA complexes engage nascent TE transcripts to instruct SETDB1-mediated H3K9me3 and DNMT3C-driven methylation, ensuring precise suppression of young, active elements^74,75. Genetics further underscores this pathway’s essential role as mutations disrupting piRNA processing cause TE activation, meiotic failure, and infertility^76,77. Female germline control operates through distinct mechanisms. In most mammals, oocytes express PIWIL1 and PIWIL3, generating abundant cytoplasmic piRNAs that suppress TE transcripts post-transcriptionally⁷⁸. While mice appear exceptional as female Piwi knockouts are fertile, studies in golden hamsters and other species reveal that PIWI is critical for oogenesis⁷⁹. Disruption of this pathway leads to ERV upregulation and early embryonic arrest, supporting the idea that piRNA function in female germ cells is broadly essential.

3.2. Somatic Silencing

In somatic tissues, stable TE silencing by promoter CpG methylation is the most well-established mechanism of transcriptional control^80-83. Following the period of epigenome reprogramming after fertilization, silencing of many TE families depends on the KRAB-ZFP-KAP1 corepressor, whose recruitment in early embryonic development leads to SETDB1-dependent H3K9me3 deposition and re-establishment of DNA methylation patterning via context-dependent activity of de novo (DNMT3) and maintenance (DNMT1) DNA methyltransferases^84-91. In principle, stable methylation, at least at symmetric CpG sites, is maintained by DNMT1 across cell divisions. The KRAB-ZFP-KAP1 silencing machinery is sequence-specific: zinc finger arrays encoded by hundreds of KRAB-ZFP transcription factors are rapidly evolving DNA-binding modules with unique ‘zinc fingerprints’ that recognize particular DNA stretches, often in TEs and other repetitive elements^92,93. The attached Kruppel-box repressor domain (KRAB) forms a platform for recruitment of TRIM28/KAP1 dimers and associated effectors that heterochromatinize the underlying TE and prevent transcriptional activation^94,95. Essentially all subfamilies of TEs are specifically bound by at least one (often more than one) KRAB-ZFP^92,93,96. The co-evolution of KRAB-ZFPs and TEs as a driver of biological innovation is a fascinating and complex topic, reviewed in detail recently⁹⁷. While the relationship bears some hallmarks of an ‘arms-race’ model where KRAB-ZFPs evolve in response to emergent waves of TEs that have escaped silencing^96,98,99, it is increasingly apparent that KRAB-ZFPs facilitate integration of TEs into gene networks by restricting their cis-regulatory activities^32,97,100.

TEs and their controllers are among the most rapidly evolving part of non-coding and coding mammalian genomes. Nonetheless, a conceptual gap in models of KRAB-ZFP/TE coevolution is the limit that sequence-specificity imposes: TEs can and do mutate to evade restriction by KRAB-ZFPs, and while new KRAB-ZFPs arise through expansion and diversification in response to waves of TE activity, this still occurs on evolutionary timescales. For example, a structural change in the ancestral primate-specific L1 promoter around 12 million years ago is thought to have allowed subsequent, evolutionarily-younger L1 copies in the human genome (including human-specific elements) to escape silencing by ZNF93⁹⁹. Thus, to preserve genome stability there remains need for a less-restricted surveillance system for somatic TE silencing during development. A major pathway that appears to meet this need centers around the human silencing hub (HUSH) complex and its associated corepressors including the MORC2 chromatin regulator¹⁰¹. Composed of TASOR, MPP8 and Periphilin at its core¹⁰², HUSH recognizes long, intronless transcriptional units, a hallmark of viral and non-viral retroelements including LINE-1 retrotransposons¹⁰³. Structurally HUSH resembles the S. pombe RNA-induced transcriptional silencing complex¹⁰⁴ and indeed several labs have experimentally demonstrated a transcription dependence for HUSH-MORC2-dependent silencing and H3K9me3 deposition^{29,48,103,105}. The dependence of basal transcriptional activity for HUSH targeting and repression, likely via RNA-binding protein Periphilin^106,107, appears to be a general feature across cells and tissues and may be governed by DNA methylation status of the targeted repetitive element⁴⁸: in pluripotency, where L1s and other TEs are transcribed and potentially functional, DNA methylation is generally lower leading to a key role for HUSH-MORC2 in dampening TE expression^108,109. Recent experiments suggest that this pathway is necessary for DNA methylation patterning to those transcribed elements upon exit of pluripotency and differentiation, at least in the neural lineage⁵⁷. So far sequence-specificity of HUSH targeting beyond a preference for A-rich sense strand has not been identified¹⁰³. However, this apparent lack of specificity tallies with a role of HUSH in recognizing invading genetic elements, given the timescales on which new sequence-specific factors evolve. Notably, several gene clusters that contain unusually long exons are also HUSH-targeted, including, in a curious twist, clusters of KRAB-ZFP genes^29,102,110. Together these activities place HUSH and its corepressors at the heart of TE biology and genome evolution.

3.3. Role of TE transcription in somatic and germ silencing

Across systems, TE transcription has emerged not only as a failure of repression but as a necessary precursor to stable silencing^111-113. The latter seems paradoxical, but not if we consider the complex of RNAPII and nascent RNA as a landing pad for the silencing machinery. Such a complex can be more or less stable, likely depending on whether RNAPII is in a paused or prematurely terminated stable state¹¹³, a distinction that is often neglected but that has important mechanistic implications. It is interesting to consider that piRNA-directed silencing and other modes of TE repression may require transcription for fundamentally different reasons: in the Piwi system, nascent TE transcripts are essential because proteins must physically engage RNA as it is being transcribed. The base-pairing of piRNAs to nascent TE RNA provides the sequence specificity that recruits Piwi complexes to chromatin, triggering co-transcriptional repression, H3K9me3 deposition (via SETDB1), and transcript cleavage¹¹⁴. Thus, without ongoing transcription, Piwi cannot “find” its target because its recognition is RNA-based rather than DNA-based. In somatic silencing, nascent transcription is needed for nucleation and scaffolding as short bursts of Pol II transcription at repeats (or meiotic genes) generate RNA molecules that recruit RNA-binding adaptors^112,115, RNA-processing factors (e.g., termination factors in fission yeast¹¹⁶ and mammals¹¹¹), and chromatin writers like Suv39/Clr4 to seed H3K9me3. Once this initial RNA-dependent nucleation occurs, heterochromatin can self-propagate via read-write mechanisms even in the absence of further transcription¹¹⁷.

3.4. RNA Modifications and RNA Decay Pathways in TE Silencing

Emerging evidence reveals that RNA modifications on nascent TE transcripts are integral components of their silencing, linking RNA surveillance to heterochromatin establishment. Many TE RNAs, including LINE-1 and certain ERVs, acquire m6A marks co-transcriptionally. These marks recruit m6A readers such as YTHDF proteins, which promote RNA decay, limit translation of TE-encoded proteins, and facilitate co-transcriptional deposition of H3K9me3 by stabilizing interactions between RNA-bound adaptors and chromatin modifiers. Loss of METTL3 or METTL14 leads to derepression of LINE-1 and impaired heterochromatin maintenance¹¹². TUTases are enzymes that add terminal uridines to TE RNAs, marking them for degradation. This process decreases TE RNA half-life and is particularly important for LINE-1 3′ UTR fragments generated during failed reverse transcription or premature termination. Uridylation thereby couples RNA decay to restriction of retrotransposition¹¹⁸.

RNA degradation exerts also a quality control on TE expression that is inherently linked to silencing. In fact, the nuclear RNA exosome, together with the NEXT (Nuclear EXosome Targeting) complex, provides a co/post-transcriptional layer of transposable element control, particularly targeting long terminal repeat (LTR) retrotransposons⁴⁷. The NEXT complex—comprising RBM7, ZCCHC8, and MTR4 recognizes nascent TE transcripts and recruits the RNA exosome, which then degrades them in the nucleus^63,119. This process prevents accumulation of potentially deleterious TE RNAs, limiting their cooption as regulatory element and their exonization⁴⁷. An association between NEXT and HUSH has been documented¹¹⁹, but how NEXT-mediated degradation may be coupled to transcriptional surveillance is a fundamental question. Are TEs that escape chromatin-based silencing rapidly targeted by this pathway, providing a fail-safe mechanism to control LTR activity? This seem to be the case in pluripotent cells but has not yet been established in differentiated cells⁶³. The idea that cells have evolved RNA degradation as a common solution to the general problem of having too many TEs (and some of them being transcribed) makes sense when one considers that a vast proportion of TEs have engaged RNAPII (as measured by PRO-seq⁵²). In fact, the idea that every RNA made in a cell is functional contradicts the evolutionary definition of function, which requires evidence of purifying selection. Current estimates suggest that only about 10% of the mammalian genome shows signs of selection¹²⁰. Some regions are transcribed because the act of transcription may be useful even if the RNA sequence itself has no selected role, and such transcripts may occasionally become raw material for new functional lncRNAs¹²¹. TEs nonetheless persist in mammalian genomes because complex organisms have small effective population sizes, making natural selection too weak to eliminate mildly harmful, unnecessary sequences, which instead drift to high abundance. As a result, cells have evolved global RNA quality-control systems to manage the large amount of nonfunctional transcript “noise”. From an evolutionary stand point, this noise is fodder for evolution, and TE that have bypassed silencing may ultimately provide some beneficial function.

In summary, a unifying principle emerging from germline and somatic studies is that TE silencing is often tied to transcription: nearly every major repressive pathway - piRNA, HUSH, heterochromatin nucleation, RNA decay, and RNA-modification systems - depends on nascent TE RNA as the entry point for recognition. This requirement may seem paradoxical given that most TEs in mammalian genomes are no longer capable of retrotransposition, yet their transcription and translation still pose risks through genome instability, R-loop formation, dsRNA accumulation, and activation of innate immune pathways. Thus, the silencing machinery that evolved to block or limit retrotransposition is maintained because it now serves the broader role of controlling the transcriptional byproducts of fossil TEs. At the same time, the pervasive insertion of TEs within introns, enhancers, and regulatory landscapes means that complete repression is neither possible nor desirable: host genomes continuously balance the need to suppress harmful expression with the adaptive opportunity to co-opt TE sequences as regulatory elements. This tension between defense and innovation defines the modern architecture of TE silencing, perhaps explaining why transcription-dependent mechanisms remain deeply embedded in genome regulation. It also suggests that we can either evolve cell-, tissue- type mechanisms that bypass silencing when it is useful to use them. We highlight this defense-innovation concept below, in relationship to the many instances in which a TE has become a regulatory element or part of a gene.

Part 4. TE Re-expression and Roles as Regulatory Elements

TE insertion sites in the host genome can directly influence nearby gene activity or local recombination frequencies. When reactivated, these insertions can be co-opted as regulatory elements to fine-tune developmental, immune and stress-responsive transcriptional programs within the host¹²². TEs have acquired roles as regulatory elements affecting transcription-factor binding sites or chromatin topology, have been exonized, or act as alternative promoters. When TEs are exonized within the internal regions of a gene, they become components of exons or regulators of splicing efficiency. When exonized at the 3′ end, they can modulate cleavage and RNA maturation by functioning as alternative polyadenylation sites. The exaptation of TEs exemplifies the concept of evolutionary tinkering, where nature prefers to reuse existing resources rather than build new ones from scratch¹²³. Evolving regulatory sequences de novo is inefficient if one can repurpose sequences that already contain regulatory information. Many TEs are information-rich because the genetic elements or viruses from which they evolved relied on host cellular machinery to regulate their own expression¹²⁴. They used host transcription factors, thereby evolving TF-binding sites, and used host RNA polymerases, thereby evolving regulatory regions controlling co- and post-transcriptional events. Even though most TEs are now unable to transpose, the vast majority retain regulatory information that can be used either in housekeeping contexts or in a cell-type-specific manner. That TE retained regulatory regions across evolutionary timescales remains a strong argument for TE functionality. In this part, we use an operational definition to discriminate different classes of regulatory element derived from TEs.

4.1. TEs as transcription factor binding sites

The timing and extent of TE reactivation are shaped primarily by two factors: the host cell’s current transcription factor (TF) repertoire and its epigenetic landscape. External (i.e. environmental, infection or heat stress) or internal triggers (i.e. differentiation programs) that alter either of these two factors have the capacity to reactivate TEs.

TEs are enriched in transcription factor binding sites (TFBS). ChIP-sequencing studies have showed that a significant proportion of genomic binding sites for many transcription factors (TFs) originate from TEs (~4-50%, mean: 20% for a given TF)¹²⁵. For a given TF, such TE-derived TFBS are typically associated with specific TE families and are often lineage specific. TE-expansion events and mutations are thus thought to play major roles in shaping species-specific transcription regulatory networks. Given this, re-expression of TEs is partially correlated with the TF abundances, leading to cell-type and/or cell-state specific reactivations of individual TE families. For example, LTR5HS-containing HERV-K elements bear consensus binding sites for OCT3/4 and SOX2^39,126. During human embryonic development, preferential transactivation of these elements over related LTR5a and LTR5b containing elements is thought to be in part attributed to high expression of these pluripotency factors. Similarly, transcriptional induction of TEs can also be achieved by the use of exogenously expressed TFs, such as with SOX2¹²⁶ or with CRIPSR/dCas9-based approaches⁴⁹. However, TF expression alone is usually not sufficient. Most TE loci are kept transcriptionally silenced via DNA methylation, chromatin condensation or repressive histone modifications. TE reactivation must therefore also rely on epigenetic changes that occur during the host cell’s shift between physiological states. For example, reduced DNA methylation or RNA degradation⁴⁷, such as that occurring during embryogenesis, oncogenic transformation, or after treatment with DNA methylation inhibitors¹²⁷, all lead to TE upregulation.

Chromatin remodeling activities also facilitate TE activation. This can come either through attenuation of key epigenetic silencing pathways or recruitment of pioneer transcription factors that circumvent repressive chromatin structures. For example, in mammalian cells, the Krüppel-associated box zinc finger proteins (KRAB-ZFPs) and KAP1 epigenetically silence TE loci via deposition of H3K9me3 and heterochromatinization, as described above. Alternatively, pioneer transcription factors such as SOX2 and FOXA1 can promote TE expression by penetrating compacted chromatin and displacing nucleosomes at TE-derived regulatory sequences. Indeed, regulation of α-fetoprotein (Afp) during stem cell differentiation and endoderm lineage specification is partially regulated by a composite FoxA1/p53/SBE (smad binding element) sites within an upstream enhancer composed of a B2-type SINE and a MER1 elements. During retinoic acid induced differentiation, FoxA1 displaces histone H1 at this enhancer to bind to the TFBS, thus activating Afp expression¹²⁸.

Aside from rewiring TFBS networks and gene structures, TEs also contribute extensively to the 3-dimensional (3D) architecture of the genome. Mammalian genomes are demarcated into large topologically associating domains (TADs) through the actions of cohesin complex and CCCTC-binding (CTCF) proteins. Such loops ensure appropriate promoter-enhancer contacts within individual TADs and thus are important for establishing tissue- and cell-specific transcriptional programs. In both mice and humans, multiple TE families are enriched at TAD boundaries. TE-derived CTCF binding sites account for ~35% of all sites, and have been shown to contribute to species- and cell type-specific variability in chromatin looping in mice and humans^129-131. Deletion of TE-derived loop anchors can eliminate TAD boundaries, leading to mis-regulation of gene expression¹³².

4.2. TE-gene chimeras: TEs as promoters, exons and terminators.

Besides TFBS, TEs can also impact host gene architectures through alternative splicing and polyadenylation (Figure 3). TE insertions often carry cryptic splice or polyadenylation sites that, when recognized by the host machinery, give rise to new exons and transcript isoforms. Indeed, with the advent of new sequencing technologies and analyses pipelines, thousands of TE exonization events occurring in mammalian cells have been uncovered,^64,133-136 with a significant proportion of recently acquired splice signals associated with TEs¹³⁶.

Figure 3. Examples of TE as: 1 promoter, 2 exon, cryptic exon, poison exon, terminator.

(A) MTC_Mm exonizes in mouse oocytes to produce a truncated Dicer1 isoform that regulates TE-integration through negative feedback. (B) L2b provides an alternative splice acceptor site in ALS/TARDBP proteinopathy-associated isoforms of UNC13A. (C) MER58A provides a polyA site to the GALNT9 gene locus. (D) Antisense AluJb provides a splice acceptor site, a splice donor site, and 18 amino acids to the protein-coding region of the RAN gene locus.

In evolutionary terms, TE-derived promoter exaptation represents a robust mechanism for generating regulatory innovation without the need to evolve new promoters from scratch. This mechanism is very pervasive for LINE1 and ERVs. A paradigm case is the Dicer1 gene in mouse where an intronic endogenous retroviral insertion (an MT-C element) acts as an alternative promoter that drives expression of a truncated Dicer1 isoform specifically in oocytes¹³⁷. This shorter protein lacks the N-terminal helicase domain of the full-length enzyme and exhibits enhanced catalytic activity in the germline context. In the human placenta transcriptional network, multiple host genes — including immunomodulatory or trophoblast-specific genes — are under control of ERV-derived LTR promoters (e.g., LTR10A, MER-derived LTRs¹³⁸). For instance, the PRL (prolactin) gene uses a lineage-specific ERV (a MER39 LTR) in primates to drive decidua-specific expression, distinct from its pituitary promoter¹³⁹. In preimplantation development, there is a pervasive usage of MT-2 element as promoter of ZGA genes^140,141.

TE-derived sequences are also frequently found embedded within internal exons of protein-coding genes, reflecting the long evolutionary dynamic of TE insertion into introns followed by sporadic exonization events. Such exonization may be deleterious as TE-derived exonic fragments often introduce premature stop codons, disrupt reading frames, or encode low-complexity, aggregation-prone peptide stretches that compromise protein stability. To mitigate these risks, cells deploy multiple layers of repression to protect against gene traps and prevent inappropriate inclusion of TE-derived fragments into mature mRNAs. A key component of this surveillance system is the Scaffold Attachment Factor B (SAFB) protein family, which binds AT-rich or repetitive intronic regions enriched for TE insertions.¹⁴² SAFB proteins interact with components of the spliceosome and co-transcriptional RNA processing machinery to enforce exon definition boundaries, suppressing weak or cryptic splice sites within TE sequences. Recent data have also identified a non-canonical ‘SOS’ splicing mechanism that specifically recognizes RNA hairpins made by the terminal inverted repeats of DNA transposons and removes these sequences from nascent transcripts¹⁴³. Despite this, based on our analysis roughly 11% of TE-chimeras identified in the human transcriptome are integrated within internal exons⁶³.

TE can also provide alternative polyadenylation to genes. While it is likely that this event is more pervasive than estimated via phylogenetic analysis, there are clear examples of this mechanism¹⁴⁴. In a classic study, it was shown that transcripts of two human genes HHLA2 and HHLA3 use a HERV-H LTR as their 3′ polyadenylation signal. Indeed, all cDNA clones for these genes terminated within the LTR, indicating that the retroviral insertion provided the functional poly(A) site for the host gene¹⁴⁵. Among ~10,000 Alu retrotransposons inserted in 3′ untranslated regions (3‘ UTRs) of protein-coding genes, ~1% (≈ 107 events) act as active polyadenylation sites, demonstrating that SINEs can be exapted to provide cleavage/polyadenylation signals¹⁴⁶. In primates, an Alu-Jr element found within the last intron of the interferon alpha receptor 2 (IFNAR2) generates a truncated IFNAR2 isoform (IFNAR2-S) through alternative splicing and polyadenylation. This shorter isoform lacks the receptor’s intracellular signaling domain, acting as a decoy receptor that competes for full length IFNAR2 and inhibits type I interferon signaling and the cells resistance to viral infection⁶⁴. Overall, such splice site capture and exon shuffling events have occurred recurrently across evolution to help generate novel gene products¹⁴⁷.

Independently of exonization events, TE insertions within introns can also influence splicing outcomes in host cells. For example, in human cells, circular RNA (circRNA) backsplicing can be facilitated by long-range reverse-complementary base pairing of inverted Alu repeats located in the flanking introns of circularized exons of nascent transcripts¹⁴⁸. Similar mechanisms involving the B1/B2/B4 SINE elements and Alu elements have been observed in mice¹⁴⁹ and humans¹⁵⁰ respectively. These TE-driven modifications to pre-mRNA secondary structure have thus contributed to species specific expression of circRNA expression. Inverted Alu repeats in the human genome can promote human-specific exon skipping through the formation of RNA stem loop structures in the flanking introns of skippable exons¹⁵¹.

4.3. TE as enhancers

Multiple transposable element families have been co-opted as enhancers at distinct stages of mammalian development. This is a consequence of a TE recruiting TFs, and been selected to influence short or long range gene expression. In mouse preimplantation embryos, MERVL and MT2 elements function as enhancers for 2-cell–stage genes, and required for activation of cleavage-stage transcriptional programs¹⁵². RLTR13 elements act as enhancers in mouse embryonic stem cells, recruiting pluripotency factors such as OCT4, SOX2, and NANOG to regulate core stem cell genes¹⁵³. In human pluripotent stem cells, HERVH elements serve as enhancer–promoter elements that maintain naïve pluripotency¹⁵⁴, while HERVK elements are transiently activated during early embryogenesis and contribute to developmental gene regulation³⁹. During cortical development, for example, primate-specific SVA and LINE-1 insertions acquire enhancer chromatin signatures and regulate nearby neurodevelopmental genes, providing a mechanism for lineage-restricted transcription¹⁵⁵. TE have also been co-opted as enhancers outside of development. In the mammalian immune system, endogenous retroviruses such as MER41 act as interferon-inducible enhancers by providing STAT1 and IRF binding sites, directly regulating genes including AIM2, IFI6, and GBP family members during antiviral responses¹⁵⁶. Similarly, Alu elements function as stimulus-responsive enhancers in human macrophages and epithelial cells, where they recruit NF-κB and AP-1 to amplify inflammatory gene expression¹⁵⁷. In the liver, LINE-1 and ERV-derived sequences acquire enhancer-associated chromatin marks and bind HNF4α and CEBP factors, contributing to hepatocyte-specific transcriptional programs¹⁵⁸. In hormone-responsive tissues, MER20 DNA transposons act as enhancers for progesterone- and estrogen-regulated genes in endometrial stromal cells, integrating transposable elements into adult reproductive physiology¹⁵⁹. A recent study found ~515 hominoid-specific TEs (~250 human-specific, mostly LTR5Hs) acting as enhancers in iPSC-derived human CNCCs¹⁶⁰. When ~75% of these were repressed by CRISPR interference, many neural-crest migration genes were dysregulated, and CNCC migration was impaired — implicating these TEs in craniofacial regulatory networks. Moreover, genome-wide studies indicate that ~77% of ape-specific CREs, and nearly all human-specific CREs, overlap TEs. Among these, LTRs and SVAs are the most commonly recruited TE types in recently evolved CREs, suggesting widespread TE co-option as regulatory elements in primate evolution^161,162.

4.4. TE as lincRNA, ssRNA and dsRNA.

Besides exerting their influence through modifying the host genome architectures, the transcriptional products of TEs can themselves also impact the cellular transcriptomic landscape. A substantial proportion of long intergenic noncoding RNAs (lincRNAs) originate from TE-derived sequences, reflecting both the bias of lincRNA loci toward TE-rich genomic regions and the evolutionary trajectories through which new genes arise. LincRNA promoters, splice junctions, and exonic segments are frequently built from LINE, SINE, and ERV insertions, providing latent regulatory modules that can drive low-level, tissue-restricted transcription. Such “proto-genes” typically emerge as noncoding transcripts because TE-derived open reading frames are initially incomplete, disrupted, or lack the regulatory architecture needed for productive translation. A model can be envisioned whereby lincRNAs represent intermediate evolutionary states along the continuum from noncoding transcription to de novo protein-coding gene birth. In this evolutionary tinkering framework, TEs act as rich raw material as their promoters initiate transcription, their sequence complexity seeds novel exons, and their mutational plasticity facilitates the eventual acquisition of structured ORFs. Lx9c11-RegoS and LINC01876 exemplify how TE-derived long noncoding RNAs can be co-opted for distinct, lineage- and context-specific biological functions^44,163. In mice, the LINE-1 element Lx9c11 is required for the neogenesis of the lncRNA Lx9c11-RegoS, which regulates expression of Schlafen family genes and acts to restrain antiviral hyperinflammation; CRISPR-Cas9 deletion of Lx9c11 leads to a lethal immune response following viral infection, demonstrating a direct, fitness-relevant role for a transposable element-derived regulatory RNA in host defense¹⁶³. In contrast, the human-specific L1-derived lincRNA LINC01876 is expressed exclusively during brain development, where L1 promoters are dynamically active, and its silencing causes reduced cerebral organoid growth and premature neural progenitor differentiation⁴⁴. Together, these examples illustrate how exapted LINE-1 sequences can give rise to functional lincRNAs that shape species-specific immune and developmental programs.

In addition, given the repetitive nature of their sequences, expression of TEs, particularly the LTR, LINE and SINE elements, can be associated with long double stranded RNA (dsRNA) formation. These RNAs can have several functions in host cells, including that of immune adjuvants, transcription inhibitors as well as altering subcellular compartmentalization. When expressed, TE-derived long dsRNAs are immunogenic and recognized by host dsRNA sensors such as RIG-I¹⁶⁴, MDA5^127,165-167 and ZBP1^168,169. These RNAs can serve as an internal cellular signal that can be used by the host cell to coordinate cellular responses to the environment. This can occur in both physiological or pathological contexts (see Figure 4). For example, during hematopoietic differentiation, TE derived dsRNAs are thought to induce RIG-I like receptor (RLR) signaling to potentiate hematopoietic stem and progenitor cell (HSPC) formation¹⁶⁴. Ectopic expression of TEs was shown to enhance HSPC development in zebrafish¹⁶⁴. On the other hand, TE-derived dsRNA can also contribute to host responses under pathological conditions. During HSV-1 or IAV viral infection, disruption of transcription termination (DoTT) often results in readthrough transcription and 3’ elongation of known genes^170,171. Inverted Alu repeats found in these aberrantly terminated host transcripts were shown to form dsRNA Z-RNA structures that were preferentially recognized by and activate host ZBP1, driving ZBP1-dependent apoptosis of the host cell¹⁷².

Figure 4. A summary of TE roles in human biology.

The mechanisms by which TE can alter genome and transcriptome function are listed (grey slices) and follow the description of such mechanistic aspects in Part 3 of this review. The single or combined TE-dependent mechanism has downstream effect in contributing to gene expression changes and/or genome instability, often in a cell type-specific manner (yellow slices). Effects impact organismal variation and disease (red slices).

In addition to binding of TE-derived RNAs to nuclear and cytoplasmic sensors, TE-derived transcriptional products and ncRNA can directly interfere with global transcription machinery. Alu and B2 SINE RNAs have been shown to directly bind to RNA polymerase II and inhibits the formation of functional initiation complexes after heatshock in human and mouse cells respectively^173-175. This activity is thought to occur in trans, with exogenously expressed B2 RNA sufficient to repress transcription in in vitro systems^173-175. This repression can apparently be lifted through self-cleavage of these RNAs via their ribozyme-like properties^176,177, affording the cell an additional transcriptional regulatory layer during stress responses or alterations in cellular states.

Besides directly binding RNA polymerase II, TE-derived RNAs may also impact transcription by physically disrupting the distribution of host cell machinery via phase separation. De-repression of endogenous retroviruses, but not GFP mRNAs in pluripotent cells via TRIM28 knockdown leads to relocation of transcriptional condensates away from super-enhancers and a corresponding downregulation of pluripotent gene expression¹⁷⁸. In as similar vein, LINE1 antisense RNA has also been shown to interact with the nuclear matrix protein MATR3 via phase separation to redistribute TADs in cell nuclei¹⁷⁹. Alu-containing RNAs are also important for maintaining nucleolar structure and function as their knockdown results in the dispersion of nucleolar condensates and altered localization of nucleolin and nucleophosmin¹⁸⁰.

4.6. TEs effect on GRN and phenotypic plasticity

Overall, the ability of TEs to reshape the genetic architecture of their host cells has important implications for the establishment of gene regulatory networks (GRNs) controlling cell identity (Figure 4). One striking example of the effect of TEs on GRNs is during zygotic genome activation (ZGA). In mouse MERVL and its MT2 LTRs drive the 2-cell (2C) program by acting as transient promoters and enhancers for ZGA genes and 2C-specific chimeric transcripts^47,141; a parallel burst of ERV activity occurs in humans around the 4-cell stage¹⁸¹, under the control of DUX- and OBOX family transcription factors¹⁸². At the same time, LINE1 RNA is thought to interact with nucleolin and KAP1/TRIM28 to organize chromatin and facilitate the exit from the 2C-like state, acting as a structural RNA that represses MERVL and promotes progression toward morula and blastocyst lineages¹⁸³. Together, these TE-derived activities provide flexible, rapidly evolving regulatory elements that establish stage-specific transcriptional programs in early embryos, while coordinated re-silencing ensures proper lineage specification as development proceeds. Notably, using antisense oligonucleotides (ASOs) to downregulate the many active multi-copies of LINE1 and MERVL, it was recently shown that the full-length RNA forms of both elements initiate the developmental cell transitions and it essential for preimplantation development^184,185. Moreover, the GRN dictated by MERVL RNA expression in ES (pluripotent to totipotent-like transition) can be mimicked by suppressing spliceosomal activity¹⁸⁶ or RNA exosome-mediated RNA degradation^47,63. These puzzling evidences suggests underapreciated mechanisms of TE transcription and cotranscriptional events controlling cell identity.

Notably, there are very few examples in which direct genetic evidence has been obtained for the functional roles of full-length TE RNAs. Addressing this gap has been technically challenging, but CRISPR approaches and ASO delivery now make such investigations feasible.

Another example of co-option of TEs in GRN is their role in host defense against infection. In primates, binding sites for STAT1 and IRF binding sites are enriched within several families of LTR retrotransposons, particularly the MER41 ERV family. Insertion of these elements near genes can confers STAT1/IRF-dependent expression of such genes, thereby rewiring transcription programs controlled by Interferon signaling. Indeed, a conserved MER41 insertion located ~220bp upstream of the human AIM2 gene contributes a functional STAT1-binding site and acts as a IFNγ-inducible enhancer of AIM2 in human cells. Deletion of this MER41-derived STAT1 binding site reduces IFNγ-dependent AIM2 expression, lowering inflammasome activity in response to viral infection¹⁵⁶. Overall, this demonstrates how such ERV insertions can become integral parts of inflammation-dependent GRNs. In species, such as mice, that lack the insertion, AIM2 is expressed constitutively rather than in a stimulus-dependent manner. Remarkably, multiple mammalian lineages have independently co-opted STAT1 binding sites related (MER41-like) and unrelated ERVs in their specific IFN-regulatory gene networks¹⁸⁷. TEs have also been co-opted to regulate immune responses is linked to the regulation of their repression. Recently, a newly characterized HUSH2 complex linked retroelement silencing to induction of an immune response: this alternative to the canonical HUSH complex is thought to directly repress interferon-stimulated genes^188,189. Under conditions of retroelement activity, HUSH sequesters the limited pool of protein subunits shared by both complexes, leading to loss of HUSH2-mediated repression and thereby innate immune stimulation.

The examples above illustrate how, over long evolutionary timescales, TEs can be co-opted by host organisms and integrated into essential GRNs. However, on much shorter timescales, such as within the lifespan of an organism, TEs and their episodic reactivation or re-expression can also have significant consequences for survival of individual cells and for the organism. A clear vertebrate example comes from the domestic dog. Multibreed association analyses have shown that the short-leg phenotype (chondrodystrophy) is influenced by the presence of FGF4 retrogenes inserted into CFA18 and CFA12^190,191. Based on their flanking target-site duplications and poly(A) tails, these retrogenes appear to have arisen from spliced FGF4 mRNA retrotransposed by the LINE-1 machinery^190,191. Expression of these retrogenes (FGF4L1 and FGF4L2) is driven primarily by a CpG island within the transposed FGF4 5′ UTR. Acting in a semidominant fashion relative to endogenous FGF4, these retrogenes cause inappropriate activation of FGF-dependent GRNs in chondrocytes, thereby contributing to chondrodystrophy. Retrogene copy number is also proposed to have additive effects on intervertebral disc calcification¹⁹² and is associated with intervertebral disc degeneration in certain breeds such as Corgis and Dachshunds. Thus, a single retrotransposition event can contribute substantially to morphological and phenotypic variation within a species. Beyond FGF4L1 and FGF4L2, numerous other functional and chimeric retrogenes have been identified in dogs, underscoring the role of TEs as major generators of genetic diversity and evolutionary novelty¹⁹³.

Part 5: Implications of Transposable Elements in Human Variation and Disease

TE activity is correlated with, or causative for, multiple human diseases¹⁹⁴. However, as described above TEs present a complex balance of genomic threat and regulatory innovation such that the phrase ‘TE activity’ is ambiguous and means different things to different researchers. On the one hand, TE promoters and enhancers drive or affect transcriptional activity in multiple ways, and transcriptional deregulation of TEs and genes are often tightly interconnected, including in pathological states. TE transcriptional activity is a necessary but not sufficient prerequisite to retrotransposition. On the other hand, to some ‘TE activity’ can indeed mean transposition, an activity can be further subdivided depending on whether it occurs in the germline or soma, leading to polymorphisms that are either inherited or not. As described above, germline transpositions are substantial drivers of individual variation in the human population. These polymorphic TEs can cause genetic diseases including hemophilia (L1), X-linked Dystonia Parkinsonism (SVA) and others. Somatic retrotransposition has been extensively documented in cancer and some somatic tissues, but the physiological relevance of TE-driven genetic mosaicism – and more broadly whether TEs are a cause or consequence of disease in these instances – is an extremely challenging topic to address. Below we summarize how TE activity – transcription and transposition – is linked to human variation and disease.

5.1-. Individual Variation

Population-genetic studies show that polymorphic TE insertion sites are widespread in the human genome, though most occur at low allele frequencies and are restricted to particular populations or continental groups^195,196. Much of this variation involves Alu insertions, with smaller contributions from L1 and SVA elements. Polymorphic TE loci are typically found in non-genic regions, consistent with strong purifying selection against insertions that disrupt coding sequences. Expression quantitative trait locus (eQTL) analyses indicate that these variants frequently exert cis-regulatory effects on gene expression and chromatin accessibility, and they can influence the coordination of GRNs, including immune-related regulatory networks^195-197. Long-read sequencing has further revealed that TEs contribute to roughly 30% of structural variants (SVs) in the human genome^31,53,198. Some of these SVs appear to have arisen recurrently in different populations, with identical SV alleles present on distinct haplotype backgrounds. Such deletions are thought to result from homology-directed repair or non-allelic homologous recombination (NAHR) mediated by sequence similarity between flanking TEs (e.g., Alu or L1 elements), which facilitates rearrangements³¹.

An ever-increasing set of whole-genome sequencing of different populations, crucially employing long-read technologies, is likely to uncover clear instances of TE-driven phenotypic variation. Indeed, where such data is effectively integrated with population registry data this can be a powerful combination. For example, a polymorphic SVA insertion in the ASIP gene was identified to shape human pigmentation¹⁹⁹, and an ever-increasing number of other SVA and Alu insertions are causative for genetic human diseases (see below). Collectively, recent findings support the view that TE polymorphisms contribute substantially to phenotypic variation among individuals and across human populations, likely through effects on gene regulation or through the introduction of structural variation at genomic loci. Notably, polymorphic TE insertions show higher linkage disequilibrium with disease-associated SNPs than with non-associated SNPs, suggesting that they may participate in shaping genetic risk to disease²⁰⁰. However, more direct mechanistic evidence is needed to determine how these events influence cellular plasticity and survival. Emerging genome-editing tools, including CRISPR-based approaches, will be essential for dissecting these relationships and experimentally testing causal roles for TE activity in phenotype and disease.

5.2-. Cancer

The functional consequences of TE expression for patient outcomes appear to be highly cancer- and context-dependent²⁰¹. For example, a TE–chimeric transcript formed by HERVH and Calbindin in lung squamous cell carcinoma was shown to promote tumor initiation by inhibiting early senescence; however, at later stages of tumor progression, expression of the same transcript was associated with increased inflammation and suppression of tumor growth²⁰². Similarly, L1 ORF1p expression was reported to reduce immunostimulatory Alu expression in TP53 mutant tumors, enhancing tumor fitness by masking the tumor from the immune system²⁰². Together, these findings highlight the dual and stage-specific roles that TE activity can play during tumor evolution and growth.

Large-scale sequencing studies across diverse tumor types indicate that approximately 50% of cancers harbor evidence of active somatic retrotransposition of L1^203-207, as reflected by characteristic insertional and structural genomic footprints. Re-expression of HERVs²⁰⁸ has also been observed, albeit at lower frequencies. The majority of somatic L1 insertions originate from a limited number of germline-encoded, retrotransposition-competent L1 loci. Mobilization of these elements further enables trans-mobilization of non-autonomous retrotransposons, including Alu and SVA elements, as well as retrogenes, all of which depend on the L1-encoded machinery (Box 1). TE activation in cancer is thought to result primarily from global DNA hypomethylation and/or genetic disruption of pathways that normally repress TE expression.

In general, most TE retrotransposition events in cancer appear to have minimal impact on tumor fitness. However, a subset can directly contribute to tumorigenesis. Non-allelic homologous recombination between TEs, especially the LINEs can result in large-scale structural rearrangements. Such events can produce megabase-scale deletions encompassing tumor suppressor genes or, conversely, lead to amplification or dysregulation of oncogenes, thereby reshaping tumor evolutionary trajectories. Besides structural variation, reactivated TEs can also promote tumorigenesis through direct insertional mutagenesis. Somatic L1 insertions into the APC locus have been reported in colorectal cancer and are thought to contribute to early tumorigenic events by disrupting APC function^209-211. TE insertions can serve as sources of enhancers/promoters, rewiring tumor-specific GRNs^212,213. In addition, TE exonization events have been identified across multiple tumor types^41,214, with some shown promote tumor growth^213,214.

Beyond direct oncogenic effects, TE reactivation profoundly influences tumor–immune interactions. Elevated TE expression correlates with DNA damage responses, inflammatory signaling, immune infiltration, and increased tumor immunogenicity²¹⁵. TE-derived peptides are presented on MHC-I molecules and recognized by the adaptive immune system, identifying TEs as a significant source of tumor-associated neoantigens^215-217. Interestingly, TE expression in the immune system may also have functional consequences for tumors as well. In exhausted tumor infiltrating T cells, LINE1-containing chimeric transcripts aberrantly accumulate on chromatin, dampening T cell effector function and resulting in dysfunctional responses to the tumor. Use of antisense oligonucleotides to downregulate such transcripts can reverse this phenotype, boosting their cytolytic activity and cytokine production²¹⁸.

These immunogenic properties have spurred interest in using TEs for cancer therapy, with proposals that TEs expression can be used as biomarkers⁴¹, immunotherapy targets^41,219 or even vaccination targets²²⁰ against cancer. In addition, pharmacologic reactivation of TEs has been explored as a strategy to enhance anti-tumor immunity via viral mimicry. This concept is supported by the observation that epigenetic therapies, often used in cancer therapies, frequently also themselves induce TE re-expression and lead to the accumulation of immunogenic RNA species that activate innate immune sensing pathways and potentiate immune-mediated tumor control¹²⁷. However, given the strong context dependence of TE reactivation effects, a deeper understanding of its functional consequences will be essential for safely leveraging this strategy.

5.3-. Neural development and disease

It has been suggested for decades that TE expression is elevated in neural tissues including the human brain. While proving that somatic retrotransposition of TEs such as L1s occurs at sufficiently high frequency as to be of physiological relevance in normal brain function is difficult, transcriptional TE activity appears to be a general feature of neurodevelopment and neurodegeneration^{38,44,56,221-227}. In addition, epigenetic changes including in DNA methylation patterns are a direct or indirect consequence of particular monogenic disorders including Rett syndrome (caused by a mutation in methyl-CpG binding protein 2^228,229) and neurodevelopmental syndromes caused by mutations in DNA methyltransferases^230,231, and may influence TE activity. Proof of concept mechanistic data on how transposons cause neurological disease remains sparse but numerous correlations to neurodegenerative disorders have been identified. In ALS and frontotemporal dementia, mutations in TDP-43 and FUS lead to nuclear depletion of RNA-binding proteins that normally suppress TE-derived transcripts. This results in toxic accumulation of LINE-1 RNA and proteins, exacerbating neuronal stress²³². Though caveated by the challenges of TE mappability, higher levels of expression of L1s and ERVs in the brain has been observed by several labs, both in ageing cohorts²³³ and in neurodegenerative disorders including Alzheimer’s disease²³⁴, Parkinson’s disease²³⁵ and traumatic brain injury²³⁴, raising the possibility that TE activation contributes to neuroinflammation. Concrete examples where transposons are causative for neurological disorders come from SVA retrotransposons^18,236,237. For example, in X-linked Dystonia Parkinsonism (XDP), affected individuals bear a polymorphic SVA element in intron 32 of the TAF1 gene. The intronic SVA interferes with expression and processing of TAF1 transcripts in neural lineages of patients^238,239. TAF1 encodes an essential transcriptional regulator, mutations or deregulation of which have been associated with neurodevelopmental defects²⁴⁰. SVA-driven changes in TAF1 expression and splicing therefore represents a plausible, causative genetic mechanism for XDP pathogenesis²³⁸. Recent studies in cellular models of XDP showed that the disease-causing SVA allele is covered by DNA methylation and H3K9me3-heterochromatin, and that loss of SVA-induced ‘mini-heterochromatin domains’ leads to an aggravation of the XDP molecular phenotype, directly linking the chromatin status of the XDP-SVA to disease progression³². Whether the interplay between TE chromatinization, gene deregulation and epigenome variations in relevant tissues may be a common feature of TE-associated disorders is unknown. However, it is noteworthy that SVAs, which are CG-rich and powerful gene regulatory elements, are increasingly linked to human phenotypes and disease – an association only set to increase as clinical and population genetic datasets expand with technologies better able to resolve their highly repetitive sequences.

5.4-. Aging

Aging represents another context of TE activation. During cellular senescence, heterochromatin erosion leads to widespread de-repression of LINE-1 and ERVs. Their transcripts contribute to senescence-associated secretory phenotype (SASP) and chronic inflammation, hallmarks of organismal aging²⁴¹. While this has spurred the idea that interventions that reinforce TE silencing, such as reverse transcriptase inhibitors, could mitigate age-associated inflammation, it is unclear whether TE expression is physiologically relevant or not. TE in aging, similarly to the expression of TE in cancer is still a descriptive event with many open questions, like: is the level of TE de-repression enough to mimic a real infection and causing inflammation? How persistent, in time, is the upregulation of TE considering that the cell is often sensitized²⁴² to immune stimulation and uses negative feedback to stop responding?

5.5-. Immune disorders

Links between transposable element (TE) expression and immune or inflammatory disease are increasingly observed, yet in most contexts the evidence remains largely correlative²⁴³. Elevated TE transcription frequently coincides with activation of interferon, cytokine, and antiviral gene programs across autoimmune disorders, chronic inflammation, and aging tissues, consistent with the idea that TE-derived nucleic acids can engage innate immune sensors. Aicardi–Goutières syndrome (AGS)²⁴⁴ illustrates both the promise and the limitations of this framework: while strong genetic evidence implicates defects in nucleic acid metabolism and sensing (e.g., TREX1, RNaseH2, SAMHD1) as drivers of chronic type I interferon signaling, the precise endogenous ligands responsible remain incompletely defined. TE-derived RNAs or DNAs have been proposed as potential contributors based on indirect evidence, but direct causal links to specific TE species are still lacking. More broadly, TE derepression, as mentioned earlier, often accompanies epigenetic instability, cellular stress, DNA damage, or local immune activation itself²⁴⁵, complicating efforts to distinguish cause from consequence. One notable example is the recent evidence linking RNaseH2 to DNA damage which is the primary driver of neuropathology in models of AGS²⁴⁶. More over, unfortunately, clinical trials with reverse transcriptase inhibitor do not produce desirable effect in AGS²⁴⁷. Together, these observations suggest that while TE expression and sensing may amplify inflammatory responses, rigorous genetic and temporal dissection is required to establish when TEs act as primary drivers rather than downstream correlates of immune dysregulation.

5.6-. Rare disease

A large fraction (ca. 90%) of rare diseases remains undiagnosed because causative variants lie outside protein-coding regions, where interpretation is far more challenging than for coding mutations. Notably, as described above, much of the non-coding human genome is composed of TE–derived sequences with potential regulatory information. By simple numerical consideration, the sheer abundance of transposable elements within the non-coding genome makes it likely that mutations affecting TE sequences or their regulatory activity represent an underappreciated source of rare genetic disease. Encouragingly, recent institutional initiatives and coordinated consortia are beginning to prioritize the systematic study of non-coding genetic variation, recognizing its central role in unresolved rare disease^248-250. Identification of ncRNA variants linked to disease will also help AI models by providing “ground truth” that we are currently missing. The application of large language models and other AI-based frameworks to genomic interpretation offers a powerful new avenue to integrate regulatory annotations, evolutionary context, and clinical phenotypes, accelerating progress toward the imperative goal of delivering molecular diagnoses to patients.

Conclusion and Future Perspectives

The story of TEs exemplifies one of the biggest paradigm shifts in modern biology. Once selfish genomic parasites or junk DNA, TEs are now understood as dynamic and essential components of the genome. They propel innovation, sculpt regulatory landscapes, influence evolution, and contribute to health and disease. Advances in long-read sequencing, single-cell epigenomics, and spatial genomics have allowed us to view TEs not as amorphous families, but as locus-specific regulatory elements whose activity can be dissected with unprecedented resolution. These tools has helped, and will help, to understand that TEs participate in cycles of repression, reactivation, and co-option, which shape chromatin architecture, transcriptional programs, and phenotypic diversity.

Mechanistic studies have redefined how we think about TE repression. Beyond DNA methylation and histone modifications, we now recognize that silencing often involves phase-separated condensates, RNA modifications such as m6A, spatial sequestration into heterochromatic nuclear compartments and ribonucleolysis. We are also starting to entangle the intricate feedback loops linking innate immunity to epigenetic defences and how controlled derepression of TEs is a source of regulatory novelty. TE-derived enhancers can be selectively engaged by pioneer transcription factors, hypoxia-inducible regulators, or viral mimicry pathways, thereby allowing cells to reconfigure their regulatory circuits in response to stress, infection, or developmental transitions. These insights position TEs not as passive passengers but as versatile regulatory modules, operating at multiple levels of gene regulation. They provide new chromatin loop anchors through CTCF binding, generate enhancer RNAs that stabilize transcriptional networks, donate splice sites that diversify protein isoforms, and give rise to lncRNAs that likely act as scaffolds for chromatin-modifying complexes. In this light, TEs emerge as multi-dimensional architects of genome plasticity, enabling rapid adaptation and innovation across evolutionary and cellular time scales.

Yet, key challenges remain. One frontier lies in understanding locus-specific TE activity—how individual insertions are differentially regulated and co-opted in specific tissues or disease contexts. Another is deciphering how host defences balance the tension between TE suppression for genomic stability and TE reactivation for regulatory innovation. The intersection of TE biology with aging, neurodegeneration, cancer, and immune regulation offers fertile ground for biomedical discovery. In parallel, the translational potential of TEs is beginning to unfold: as gene therapy vectors, as immunotherapy adjuvants, and as biomarkers that report on cellular stress, viral infection, or malignant transformation (Box 2).

Box 2. Use of TE based technologies in human health and disease.

Given their intrinsic capacity to excise and insert genetic material within their host genomes, TEs have become increasingly attractive toolsets for both functional genomics and genome-engineering applications. Several bacterial and eukaryotic TEs have already been repurposed for these applications. For example, Tn5²⁶⁶, which was originally characterized in E. coli, is extensively utilized in multi-omics workflows, including that of genome and transcriptome sequencing. Likewise, DNA transposons such as piggyBac^267,268, are routinely used in laboratory settings to stably express transgenes. Both PiggyBac and Sleeping Beauty DNA transposons have also been evaluated in clinical trials as non-viral gene transfer vectors, with multiple studies demonstrating their feasibility and potential^269-275.

Despite these advances, the widespread use of TE-based technologies into clinical contexts is hindered by issues related to insufficient control over insertion site specificity and concerns about adverse effects associated with insertional oncogenesis and genome instability^276-279. Most TE families do exhibit broad target or integration preferences, often showing low nucleotide-sequence specificity or biases towards particular chromatin states (e.g., actively transcribed regions, open chromatin or AT-rich regions). Overall, this unpredictability poses obstacles for therapeutic use, where highly targeted and precise genomic integration are preferred and desired.

To overcome this, work has focused on tuning transposase activity durations and copy numbers of transgenes incorporated into cell²⁸⁰. In addition, other strategies have focused on engineering programmable TEs using CRISPR-based targeting strategies. In one such approach, Cas9 was fused to the transposase encoded by the rice Pong transposable element. This fusion protein enabled targeted insertion of mPing, a non-autonomous TE in rice that requires Pong for mobilization, into the CRISPR/Cas9 induced double stranded breaks. This system could deliver up to 8.9kb of cargo, at higher efficiencies than CRISPR/Cas9 mediated HDR or NHEJ knock-ins, underscoring the utility of transposase-assisted gene editing²⁸¹. CRISPR-guided approaches have been developed in mammalian systems using other transposases such as Sleeping Beauty²⁸² and piggyBac^283,284 transposases to introduce both enhancers as well as transgenes into recipient genomic loci. In a similar vein, CRISPR-associated transposases (CASTSs), bacterial systems that use RNA-guided nuclease-deficient CRISP-Cas systems to guide insertions by Tn7-like transposases^285,286, have also been developed and optimized for use in mammalian cells to deliver large payloads (~15kb) into recipient genomes^287-289.

More recently, several groups have also proposed to adapt site-specific retroelements, particularly members of the non-LTR R2 retroelement family, to insert transgenes via TPRT into the recipient genomes^290-293. Although mammals do not bear active copies of R2, they do retain conserved target sites within the multicopy ribosomal RNA (rRNA) gene loci. Transgene insertion into these sites can be achieved via transfection of mRNAs encoding an R2 protein as well as a template RNA bearing a 3’ end complementarity to target sites^290,291. Further protein engineering of R2 proteins may also be refine or expand target specificity, improving efficiency and precision of insertion events. While promising, additional work is still required needed to improve full-length transgene retention, control copy number, and assess whether the rDNA loci truly represent a suitable safe-harbor site for transgene integration.

In some cases, the inherent ability of transposons to excise themselves has also been exploited to enhance genome-editing workflows. For example, for CRISPR/Cas9-based prime editing (PE) or gene-editing, prolonged expression of the Cas9 and pegRNA/sgRNA generally improves editing efficiencies. However, existing transient transfection protocols using ribonucleoprotein complexes or episomal vectors often have short expression windows and require extensive optimization per cell line. To counter this, Cas9 expression cassettes flanked by piggyBac inverted terminal repeats can be transiently integrated into the genome using piggyBac transposase^294-296. After editing, the entire cassette can be excised by expressing integrase-deficient piggyBac transposase. Doing so, entire cassettes are precisely excised, leaving a minimal 4-bp TTAA genomic scar. This strategy thus minimizes genomic perturbation while enabling the prolonged expression necessary for efficient editing, thus improving the feasibility of adapting such approaches for clinical use.

It is worth noting that while issues related to target-site integration specificity have also been reported for viral gene-therapy vectors (e.g. lentiviruses, AAV), these have only retained an advantage over the non-viral DNA-transposon based approaches discussed above due to their higher delivery and integration efficiencies. Currently, non-viral DNA-transposon based vectors require less efficient physical (e.g. direct injection or electroporation) or chemical-based (e.g. cationic lipids, lipid nanoparticles) methods for delivery. A further bottleneck also lies in the relatively low insertion efficiencies observed across the different platforms, likely compounded by delivery-related limitations such as reduced cellular uptake, insufficient nuclear entry and constraints associated with transposition efficiency^297-299.

In principle, if non-viral TE-based systems can be sufficiently optimized, their development as gene therapy vectors could offer many advantages over traditional viral vectors, including improved biosafety profiles, the capacity to integrate larger genetic cargo, lower production costs, greater scalability of manufacturing to GMP standards^297-299.

Overall, these developments illustrate the rapid evolution of TE-based genome engineering from basic research tools to programmable platforms capable of precise, large-cargo genomic integration. Continued advances in protein engineering and targeting specificity will be essential for realizing the full therapeutic potential of such transposon- and retroelement-derived technologies.

In many ways, this trajectory vindicates Barbara McClintock’s vision of the genome as a dynamic, responsive system rather than a static canvas. What once seemed like molecular noise is now recognized as a reservoir of regulatory potential, mobilized by cells to adapt, evolve, and survive. TEs are not mere dead body remains of evolutionary accidents. They are architects of genomic resilience and creativity, guiding both evolution and the immediate responses of cells to their environments. As research continues, it becomes clearer that to understand the genome is to read the languages written into its non-coding part, which is dominated by TEs.