Causes and Consequences of Varying Transposable Element Activity: An Evolutionary Perspective

Abstract

Transposable elements (TEs) are genomic parasites found in nearly all eukaryotes, including humans. This evolutionary success of TEs is due to their replicative activity, involving insertion into new genomic locations. TE activity varies at multiple levels, from between taxa to within individuals. The rapidly accumulating evidence of the influence of TE activity on human health, as well as the rapid growth of new tools to study it, motivated an evaluation of what we know about TE activity thus far. Here, we discuss why TE activity varies, and the consequences of this variation, from an evolutionary perspective. By studying TE activity in nonhuman organisms in the context of evolutionary theories, we can shed light on the factors that affect TE activity. While the consequences of TE activity are usually deleterious, some have lasting evolutionary impacts by conferring benefits on the host or affecting other evolutionary processes.

1. INTRODUCTION

Appreciable parts of nearly all eukaryotic genomes—approximately half of the human genome (82), for example—are derived from parasitic repetitive DNA called transposable elements (TEs). When active, TEs propagate by making copies of themselves and jump into new positions in genomes, promoting their own transmission over that of other genes in the host genome. This strategy is highly evolutionarily successful, and TEs are now found in almost all eukaryotic genomes (195).

TEs are classified into higher-order groups based on their replicative mechanisms and into families based on sequence similarity (197) (see the sidebar titled Taxonomy of Transposable Elements). While some species have dozens of active TE families, humans have only four, and much of the TE-derived sequence in the human genome comes from families that are no longer active. Nevertheless, TE activity—the copying and jumping movement of TEs—is not a relic of the past in humans. New TE insertions occur in approximately 6% of births (61), and as a cumulative result of this TE activity, every pair of individual human genomes differs by at least 1,000 TE insertions (i.e., polymorphic TEs) (175, 176). Some of this TE activity results in harmful mutations, such as disrupted genes (117), ectopic regulatory sequences (145), chromosomal rearrangements (8, 121), or altered epigenetic regulation (37). Indeed, polymorphic TE insertions contribute to genetic diseases (74, 144), and TE activity is associated with degenerative conditions related to aging and cancer (27, 71). Accordingly, an appreciation of the causes and consequences of TE activity is necessary for a thorough understanding of the genetics of human health.

In this review, we address what determines TE activity from an evolutionary perspective. TE activity is evolutionarily labile and can differ at multiple levels—among TE families (126), among copies of the same family (25), among cell types (113), among individuals (166), and between humans and closely related primates (110). Many molecular mechanisms underlying this variation in TE activity have now been discovered (3, 47), but here we ask why this variation exists in the first place. Our evolutionary perspective is grounded in decades of work by evolutionary geneticists to understand TE activity. Multiple models have been proposed to understand the evolutionary dynamics of TEs, including analytical theories with empirically testable quantitative predictions (reviewed in 35, 106). We summarize insights from some of these theories and evaluate how they bear on what we know about variation in TE activity as well as what we know about the associated evolutionary consequences that may be relevant to human genetic variation, health, and evolution.

2. WHAT TRANSPOSABLE ELEMENT ACTIVITY IS AND HOW TO ESTIMATE IT

Before delving into evolutionary theories for variation in TE activity, which we define as the insertion of TE sequences into new genomic locations, we first describe the wide variety of approaches used to estimate TE activity. Because TE activity leads to new TE insertions, intuitively, one might assume that TE activity is easily estimated by quantifying genomic TE abundance. However, the relationship between TE activity and abundance is not straightforward (see the sidebar titled The Complex Relationship Between Transposable Element Activity and Abundance). Instead, TE activity is estimated by many alternative methods.

As TE activity in the germline results in new, heritable TE copies, it is most directly quantified by identifying de novo inherited TE insertions. In model organisms that are easy to rear and have short generation times, reasonable estimates of TE activity can come from whole-genome sequencing of a handful of lines or strains propagated for many generations under conditions where they accumulate new mutations, including new TE insertions (1, 81). In humans, where this mutation accumulation approach is impractical, an alternative approach is to sequence genomes from a large number of parent–offspring trios: TE copies present in the offspring but absent from both parents are assumed to be due to transposition (13, 61). Sequencing three generations, instead of two, allows some insertions to be classified as occurring in gametogenesis (or early embryogenesis) or during later development and resulting in germline mosaicism (61, 163). While readily applicable to humans, this approach is expensive, requiring a large number of related individuals and high-quality genomic data suitable for detecting structural variants. Germline TE activity can also be inferred indirectly based on the abundance of intermediate products necessary for transposition, such as TE mRNAs and proteins (101). However, as many other processes may also influence the abundance of these surrogates of TE activity, they may not directly reflect the rate of new TE insertions (165, 204).

Finally, relative rates of germline TE activity can be inferred using comparative methods, mainly by comparing TE variation from different individuals or species or from different copies of the same TE family. Species-specific TE insertions indicate that the insertion likely occurred since the common ancestor of the species and, accordingly, was active at the time. Similarly, TE copies with low population frequencies, or those present in only a few individuals, should have been active very recently. These approaches have been widely applied to identifying recent TE activity in humans and other primates (58, 67, 126, 199). Comparisons of sequences between TE copies in the same genome can also imply recent activity. Once a TE is inserted into a new genomic location, it accumulates new single-nucleotide mutations. By assuming the rate of substitution is relatively constant over time, one can use the number of mutations to infer the age of TE copies, and low sequence divergence implies recent TE activity (14). However, these approaches should be applied with caution because other evolutionary processes, particularly natural selection, can influence both the presence/absence polymorphism of TEs and their sequence variation (96).

Other than heritable germline insertion, TE activity can also occur in the soma. Somatic activity can be estimated by identifying de novo TE insertion events in whole-genome sequences, but this comes with inherent challenges. Mainly, most somatic insertions are present only in a small subset of cells, which makes it difficult to distinguish true de novo insertions from technical artifacts, especially with short-read sequencing (57, 179). Ongoing technical developments, including long-read (173) and single-cell (16) sequencing, should improve estimates of somatic TE activity in the near future. Beyond these genome-wide approaches, reporters cleverly designed to signal new TE insertions have been helpful in pinpointing the activity of focal TE families, especially in human cell lines (128, 138) and nonhuman model organisms (32, 130).

3. VARIATION IN TRANSPOSABLE ELEMENT ACTIVITY IN HUMANS AND OTHER ORGANISMS

The evolutionary lability of TE activity is demonstrated by large differences in TE activity at multiple levels (reviewed in 83). Currently or recently active TE types differ among taxa, with mammals tending to have active long interspersed nuclear elements (LINEs) and short interspersed nuclear elements (SINEs) (136) and plants tending to have active long terminal repeats (LTRs) (149) (see the sidebar titled Taxonomy of Transposable Elements). There are also smaller-scale differences in TE activity, with specific TE families having varying activity between closely related species (97). Further, not only which but also how many TE families are active at any given point in time differs among species (136, 149). This variation in TE activity has many causes, both proximate and evolutionary, which we address in Section 4.

Even within humans, TE activity varies substantially among TE families and among copies of the same family. Almost half of the human genome is occupied by TE sequences, with more than 4 million TE copies annotated (82), but only a very small fraction (<0.05%) appear capable of transposition (125). For example, there are no active class II TEs in humans, even though they have contributed many relic sequences, amounting to ~3% of the human genome sequence (140), and at least a few genes of critical importance [e.g., THAP genes (116) and RAG1/2 (91)].

The active, or potentially active, TE families in humans are all retrotransposons. Only one of these families is an endogenous retrovirus (ERV). Despite constituting ~8% of the human genome, most ERVs have long been inactive, with the exception of a low-abundance ERV, HERV-K, which bears sequence similarity to infectious beta-retroviruses (122). HERV-K has been active in the human lineage, as evidenced by human-specific HERV-K insertions (9). At least a few dozen polymorphic HERV-K insertions (11) further imply recent, though not necessarily ongoing, transposition activity.

Most TE activity in humans is due to three LINE/SINE families (42). LINE-1 is the most abundant of these, occurring in thousands of copies, and constitutes 17% of the human genome sequence (82). LINE-1 is also highly diverse and is further divided into subfamilies based on sequence similarity, with each subfamily likely derived from a highly active ancestral sequence (20, 94, 110). Full-length LINE-1 encodes ORF1p, which provides RNA chaperone activity (119), and ORF2p, which has endonuclease and reverse transcriptase activities and enables autonomous transposition (59, 120). However, surprisingly few LINE-1 copies (~80–100) are currently active (25).

The two remaining active elements, Alu and SVA, are nonautonomous SINEs that parasitize proteins encoded by LINE-1 (see Section 4.2). In terms of copy number, Alu is the most abundant human TE, with more than 1 million copies, but because the full-length copies are only 300 bp, this amounts to only ~10% of the human genome sequence. Very few of these Alu elements are capable of retrotransposition (41, 50). Interestingly, an Alu insertion may have played a key role in the loss of tails in the ape lineage (203). The remaining active element, SVA, contains sequences with homology to both Alu and HERV-K (169) and makes up a minor part of the genome (0.15%). However, SVA shows evidence of relatively high recent activity—SVA is specific to hominids, with two subfamilies specific to humans (137), and ~60% of the ~2,700 SVA copies are intact, suggesting that they have recently arisen and may be transposition competent (187).

4. EVOLUTIONARY CAUSES FOR VARIATION IN TRANSPOSABLE ELEMENT ACTIVITY

Nearly universally across eukaryotes, TEs have an arsenal of silencing mechanisms directed at them (see the sidebar titled Silencing Mechanisms of Transposable Elements in Animals), yet they have persistent activity. While TE-derived sequence is beneficial to the host in some cases (reviewed in 29, 64), the replication activity we focus on here is usually parasitic and has clear harmful effects (77, 117). Accordingly, selection at the host level should favor the evolution of defense mechanisms that suppress TE activity. So why do TEs still jump, despite the selection pressure on hosts to completely suppress them?

From an evolutionary perspective, the reasons TEs are still active fall into three main classes. First, host silencing mechanisms can have pleiotropic effects. The core machinery generating small RNA targeting TEs, for example, has other functions in germline development (e.g., 208). Host-mediated TE silencing can sometimes propagate to nearby sequences and have harmful off-target effects on host genes (37). Further, TE silencing mechanisms usually target many TEs, which may limit their ability to evolve to fully suppress any particular TE families or subfamilies. Second, this question is related to another core question in evolutionary biology: Why do mutations still occur? Both mutation and transposition generate genetic variation that is the source material for adaptive evolution, but that does not explain why they persist—evolution is short-sighted, and selection mainly sees their immediate harmful effects. The reason why mutation and transposition still occur likely lies in the limits of natural selection, as reducing already low mutation and transposition rates might have a benefit that is too small to matter (112). Specifically, when selection only weakly favors a beneficial allele, it can be overwhelmed by genetic drift, causing the extinction of the allele in spite of its benefits. As a rule of thumb, to overcome random loss, a selective benefit has to be at least as strong as the reciprocal of the effective population size, and stochastic effects are strongest in small populations.

Finally, TEs are still active because, while selection on the host might favor complete suppression, selection on the TE sequences favors selfish TE activity. These opposing forces set the stage for evolutionary theories asking such questions as, When do we expect selection for parasitic TE activity to dominate? When do we expect host suppression to evolve? And what do these predictions reveal about variation in TE activity within and between species? We first describe the evolutionary genetic models addressing these questions, and then contrast and compare their predictions with empirically observed variation in TE activity.

4.1. Evolutionary Models for Variation in Germline Transposable Element Activity

Two types of theoretical models attempt to provide general explanations of the evolution of germline TE activity: (a) short-term models for the evolution of host alleles that alter TE activity and (b) long-term models for host–TE coevolution. Here, we describe the core ideas and major predictions of these models.

4.1.1. Short-term models for the evolution of host alleles altering transposable element activity.

Early theoretical work developed a series of analytical models under the assumption that the number of active TEs remains stable within species, despite active TE replication (33, 102). In these equilibrium models, new TE copies created by transposition are balanced by the removal of TEs via excision or by purifying selection against copies that are harmful (transposition–selection balance). Within this framework, a seminal theoretical study (34) used analytical solutions to explore how and why TE activity might evolve.

Charlesworth & Langley (34) studied under what biological conditions TE suppression is expected to evolve by calculating the strength of selection (s) favoring alleles suppressing TE activity. Intuitively, alleles that reduce TE activity, whether host or TE encoded (hereafter referred to as suppressor alleles), are expected to be favorable to the host because they reduce its genome-wide TE burden. However, in an equilibrium population that reproduces sexually, randomly mates, and undergoes meiotic recombination, this benefit can be very small: s is proportional to the number of active TEs copies (n) and the square of the TE transposition rate (u²) (34). With transposition rates estimated to be ~10⁻⁴–10⁻⁵ per copy per generation (1, 35, 81, 192), this model predicts that selection is far too weak (~10⁻⁸–10⁻¹⁰) to overcome the stochastic effect of genetic drift under biologically realistic conditions, and that suppression will not evolve.

While counterintuitive, this result can be understood by considering that a suppressor only gradually reduces TE burden in the background over generations, while sex and recombination act every generation to decouple suppressor alleles from this reduced TE burden. Hosts with or without a suppressor allele will thus suffer roughly equal TE burdens, resulting in little benefit to suppression (Figure 1). This theoretical prediction also suggests that self-regulation of TEs is difficult to evolve in panmictic, freely recombining populations.

Figure 1.

Short-term model of the emergence of suppressors with and without recombination. (a) Each transposable element (TE) insertion (shown as black bars on the chromosomes) has the capacity to copy and insert itself into a new genome location with a transposition rate of u (per copy per generation). The emergence of a suppressor allele (red bar) reduces this transposition rate. (b) With frequent recombination and thus low genetic linkage, the suppressor may be constantly recombined onto new genetic backgrounds with many TE insertions (shown as chromosome blocks with different colors). As a result, the suppressor allele is unable to enjoy a selective benefit despite its ability to reduce TE activity and prevent the occurrence of new TE insertions. (c) According to Charlesworth & Langley (34), the selective benefit of a suppressor (s) is proportional to the number of TE insertions in an individual (n) and the square of transposition rate (u²), shown by the colored lines. This theoretical prediction suggests that, under the typically observed transposition rates, the benefits of suppressors are too weak to overcome the impacts of genetic drift (dotted lines) for a broad range of realistic effective population sizes (N_e). Accordingly, suppressor alleles are unlikely to evolve in recombining populations with low transposition rates. (d) Without recombination, suppressors remain associated with their genetic background, so that they enjoy the long-term benefits of preventing the occurrence of additional TE insertions. (e) Under the scenario described in panel d, the selective benefit of a suppressor (s) is instead proportional to n and u. Thus, suppressors can evolve under a broad range of combinations of transposition rates and population sizes.

However, there are at least two scenarios where enhanced suppression of TE activity readily evolves. As hinted at above, a newly arising suppressor allele can spread through the population when there is a strong linkage between the suppressor allele and TEs on its genetic background, as can occur due to asexual reproduction, a selfing mating system, inbreeding, or lack of meiosis (e.g., somatic cells) (see Section 4.3; Figure 1). In these cases, suppressor alleles remain associated with a reduced TE burden long enough for the host to enjoy the benefits of suppressing TE activity. Suppressor alleles can also spread when they reduce TE activity that causes dominant lethal or sterile mutations. Suppressing this kind of TE activity has strong selective benefits, even if the rate of mutation is low, as it immediately protects the offspring of parents bearing the suppressor allele (34).

In marked contrast, TEs that escape host suppression, or increase their transposition rates in some other way, are expected to enjoy a much stronger selective benefit than suppressor alleles—proportional to u rather than u², as for the host- or TE-encoded suppressor alleles discussed above (34). Accordingly, TE variants that increase their own activity are expected to spread easily.

4.1.2. Long-term models for host–transposable element coevolution.

It is important to note that the analytical model of Charlesworth & Langley (34) assumes an equilibrium TE copy number under transposition–selection balance and thus is not designed to model cases where TE copy number increases or fluctuates dramatically over time. To describe those cases, instead, there are several related verbal models, which describe long-term host–TE coevolution as an arms race. These host–TE arms race models resemble those for host–pathogen coevolution, such as the Red Queen hypothesis (183), and describe the dynamics of antagonistic interactions between hosts and TEs over long evolutionary timescales (196). In an arms race, the focus of selection cycles between the host and the TE, with the host responding to an active TE by evolving suppression and the TE responding to host suppression by escaping it and regaining activity, which triggers a new evolutionary response from the host, and so on (Figure 2a). These alternating cycles between hosts and TEs lead to variation in TE activity through time, predicting bursts of TE activity followed by periods of quiescence.

Figure 2.

Long-term model of transposable element (TE) activity and suppression. (a) A new, unsuppressed TE family (①, black bar) can lead to many new copies of the TE (②, additional black bars) due to high rates of transposition events (black arrows). This high TE burden decreases individual fitness, driving the emergence of suppressors of transposition (③, red Xs). Silenced TEs then accumulate mutations over time (④), potentially losing activity to a point where the entire family can no longer transpose (⑤, gray bars). The entire cycle can then be reinitiated with a mutation reactivating a preexisting TE insertion or the introduction of a novel TE family through horizontal transfer. (b) Multiple bouts of transposition and suppression, as described in panel a, lead to a lumpy distribution of pairwise sequence divergence between copies of a TE family. That is, a recent burst of transposition due to an unsuppressed TE family will initially lead to many similar copies of a TE with low pairwise sequence divergence. Bursts of transposition that happened long ago will result in copies of TEs that have accumulated many mutations, thus showing high sequence divergence with other TE copies. (c) Phylogenetic signatures generated by a TE that recently escaped host suppression (Esc) give rise to many highly similar copies (in radial form, this would yield a star-like phylogeny). (d) Horizontal transfer (HT) of a TE leads to incongruence between a species phylogeny (top) and a phylogeny of TE sequences (bottom). Shown is horizontal transfer of a TE from species C (black) to species A (blue), which is sister to species B (orange). The resultant TE tree shows that, in a phylogenetic tree, sequences of TE insertions found in species A are nested within those of species C and missing from species B.

An elaboration of the arms race models, the life cycle models (17, 18), was inspired by the evidence that TEs occasionally horizontally transfer between species through mostly yet-to-be-uncovered mechanisms (69). When TEs horizontally transfer to naive hosts, or hosts that do not have such a family of TEs, transposition rates can be extremely high (up to 0.1 per copy per generation) (52; reviewed in 35), leading to a burst of TE replication and likely strong selection for host suppressor alleles. TEs with suppressed activity may then have one of several fates: escaping host suppression and entering a new cycle of the host–TE arms race, becoming inactive relics in the host, or being domesticated by performing some useful or even essential function for the host. In this life cycle model, TEs maintain their long-term parasitic functions through repeated horizontal transfer into new species. The major prediction of this model is that the level of activity of a TE family depends strongly on the time since a TE family has invaded: Recently invaded families will be at their peak of activity, while TEs from long-ago invasions should be mostly silenced or domesticated (18) (Figure 2b).

4.1.3. Reconciling the short- and long-term models: Where do they differ, and what do they explain?

The short- and long-term models yield strikingly different pictures of TE activity: While the short-term equilibrium models treat TE activity as parasitic but stable components of eukaryotic genomes, the long-term models predict bursts of TE activity followed by periods of quiescence.

In humans and primates, phylogenies of LINE-1 and Alu elements show patterns consistent with historical bursts of activity. A few ancestral copies of these families give rise to most of the descendant copies (20, 94, 110) (Figure 2c), which is consistent with the small number of currently active LINE-1 and Alu elements (25) (see Section 3). In some organisms, these bursts may be due to newly horizontally transferred TE families with high rates of activity (Figure 2d)—particularly in insects, bats, and carnivores (136, 146, 150). But for humans, a recent horizontal transfer is unlikely to play much of a role, as both LINE-1 and Alu were acquired long ago (21, 49). Instead, changes in TE activity through time may be due to host–TE arms races. As might be expected with such an arms race, genes involved in host silencing have high rates of evolution or rapid gene copy number expansion in many organisms. A classic example of this phenomenon is the Krüppel-associated box (KRAB) domain zinc finger proteins (KZNFs), the DNA-binding component of transcriptional repressor complexes that regulate many TEs in tetrapods (reviewed in 26). KZNFs show many signs of involvement in an arms race, including being extensively duplicated and showing more copies in genomes with higher TE content (68, 177). Careful unpicking of the evolutionary history of several of the hundreds of human KZNFs and their targeted TEs shows evidence of TE–KZNF coevolution at the molecular level. For instance, the proliferation of some LINE-1 subfamilies has coincided with the evolution of the loss of a binding site for ZNF93, suggesting that this deletion allows escape from transcriptional repression (87).

If TE activity changes through time, as predicted by the long-term models, what is left for the short-term models to explain? In fact, quite a bit. At any given point in time, for most organisms, much TE activity appears to be regulated and stable, with low rates of transposition. Under these conditions, the short-term model provides insight into selection acting on changes in TE activity—selection for host alleles reducing TE activity should be weak, with a baseline rate of insertion inevitable, while selection for TE variants to escape host silencing can be strong (34) (see Section 4.1.1). Stitched together, the short- and long-term models provide a unified picture of the evolution of TE activity. That is, under the short-term model, strong selection favoring TEs that escape host silencing may lead to elevated activity and increased TE copy number. In turn, these are the conditions that lead to strong selection for TE-suppressing alleles. Accordingly, the changing balance of selective forces under the short-term model can yield coevolutionary patterns that look remarkably like an arms race over the long evolutionary time.

Likewise, horizontal transfer of TEs may result in TE activity that is high enough to drive the spreading of host suppressor alleles. Note, however, that the duration of any single burst of TE activity may be too ephemeral to allow for the fixation of host suppressor alleles, as suggested by theoretical analysis (107) and empirical experimental evolution studies (191). Accordingly, the repeated burst of activity of TE families, as pointed out by the life cycle model, may be needed (17, 18).

Further, predictions from the short-term models have led to novel insights into why TE activity varies at multiple biological levels. Specifically, the short-term model suggests the following:

1. Suppression of TE activity evolves easily in asexual or selfing species: Comparisons of closely related species with contrasting mating systems—for example, in plants (2), insects (10), and nematodes (51)—generally find evidence of reduced TE activity in the selfing lineages. Also, obligatory asexual eukaryotes, such as bdelloid rotifers, have massively duplicated TE suppression genes (133, 157). However, there are cases where selfing lineages have higher TE activity (reviewed in 70), contrary to the prediction of the short-term model. This may result from the impacts of other factors inherent to these breeding systems on TE activity (202), including weaker selection on host silencing due to reductions in effective population sizes. Indeed, an experimental evolution study in Caenorhabditis elegans found that small laboratory populations suffered compromised RNA-mediated suppression of TEs (15).

2. TE suppressors are strongly favored when TEs cause dominant lethal or sterile mutations: Severe, dominant TE-induced mutations are probably rare, but there are known examples. For instance, P elements, a type of class II TE, can induce sterility in Drosophila (159), most likely due to DNA damage in oocytes. A similar process may occur in mammals, wherein LINE-1 activity appears to lead to loss of oocytes during fetal development (118). P element–induced sterility might have selectively favored the rise of truncated variants that suppress P element activity (151), even though piwi-interacting RNAs (piRNAs) are likely the main suppressors of the induced sterility (23). In yeast, which has a haploid phase where all mutations are expressed (and thus effectively dominant), retrotransposition of Ty1 is inhibited by a Ty1-encoded protein (160).

3. Suppression mechanisms for TE activity should target large numbers of TE sequences: The model predicts that the selection coefficient for host suppressor alleles is proportional to TE copy number (see Section 4.1.1). Accordingly, suppressor alleles that reduce the activity of many TE families, and thus a large number of TE copies, should enjoy a larger selective benefit than those that target a specific TE family or even a few TE variants. Consistently, despite differences in details, sequence-dependent posttranscriptional and transcriptional silencing mechanisms that target most, if not all, TE families are common across eukaryotes (see the sidebar titled Silencing Mechanisms of Transposable Elements in Animals). While the actual targeting molecules, such as piRNA sequences with homology to specific TEs or KZNFs with high binding site specificity, could be TE-family specific, other mechanisms allow them to maintain the ability to target a large number of TEs. For instance, piRNAs seem to be able to quickly expand their repertoire (48, 170, 207), allowing a broad spectrum of suppression activity. KZNFs also can often suppress many TEs by targeting conserved regions shared across many TE copies (75, 211). Other evidence consistent with the prediction comes from the slow evolution of KZNFs in the face of new TE variants that evade suppression. Specifically, no human-specific LINE-1 subfamily appears to be suppressed by a specific KZNF (30), consistent with the idea that the escaping TEs reach a high copy number before suppression can efficiently evolve.

4.2. Ecological Models for Variation in Transposable Element Activity

In addition to host–TE interactions, TEs can compete with other TEs for resources, such as host factors or transposition machinery (65), and this can similarly influence TE activity. Such TE–TE interactions are analogous to ecological interactions between species and can drive the diversification of TE families in a genome (185). To avoid competition, TEs may fill empty ecological niches in the host, such as different cell types or developmental stages. Even when this means that TEs may have to diversify their activity into somatic cells, some TEs can still ensure their transmission to the next generation—especially virus-like LTRs, which can cross cell–cell boundaries. In Drosophila, for example, ZAM retrotransposons are transcribed in somatic ovarian follicle cells and then infect and invade the neighboring oocyte (206). Similar activity has also been proposed for HERV-K in humans (12). Recent work (167) beautifully illustrates how TE–TE competition can drive LTRs in Drosophila to diversify into distinct expression niches. Specifically, closely related LTR families avoid coexpression in the same cells, which may allow individual LTRs to avoid cross-mobilization of copies from closely related TE families.

Competition between TEs can also limit TE activity. For example, if host proteins are rate limiting, then competition between TEs for host factors may preclude the evolution of hyperactive TE variants. A TE copy encoding a hyperactive transposition protein may not benefit from an increase in transposition rate if host resources are sequestered by other TEs in the same cell (18). Another common form of TE–TE interaction, hyperparasitism, is found throughout the tree of life. A prime example is the maize Dissociation element, a nonautonomous version of the Activator element, first discovered by McClintock (121). Hyperparasitism is also found in humans—the nonautonomous TEs Alu and SVA are hyperparasites of LINE-1, as they depend on proteins encoded by LINE-1 for transposition. Models of autonomous TEs and their hyperparasites show that, when there is no reduction in the activity of nonautonomous copies, hyperparasitism can lead to the eventual loss of active autonomous copies due to competition for the transposition machinery (92, 105). The best direct support for this theoretical prediction comes from a Drosophila experimental evolution study that tracks autonomous and nonautonomous copies of Mariner, a class II TE, over generations (156). The introduction of nonautonomous copies caused their autonomous counterparts to suffer reduced activity and eventually led to their extinction and complete loss of Mariner activity. It is worth noting that, while the nonautonomous TEs may seem to benefit hosts by reducing the load of autonomous TEs, this long-term benefit is very likely to be outweighed by the immediate deleterious effects of nonautonomous insertions.

4.3. Evolutionary Models for Somatic Transposable Element Activity

The short- and long-term models above are focused largely on germline TE activity, but TE activity and host silencing also occur in the soma, with variation in somatic TE activity having potentially important consequences for human health. Below, we discuss some of the potential evolutionary factors that influence somatic TE activity.

4.3.1. Somatic transposable element activity is expected to be low.

As new TE insertions must be inherited in order for TE copy number to increase through evolutionary time, somatic insertions are usually inconsequential to the long-term success of the TE. In fact, they may be self-defeating if they prevent host reproduction. Thus, the optimal strategy, for both host and TE, may be complete TE inactivity in the soma (73). An extreme form of this strategy is found in ciliates, which have two nuclei: an inherited germline micronucleus and a somatic macronucleus that arises from the germline micronucleus in each generation (reviewed in 22). As the macronucleus forms, TE-rich portions of the genome are expelled through programmed genome elimination.

Though far less dramatic, TEs in multicellular organisms also typically display quiescence in somatic activity. For example, in mammals, some TEs incorporate binding sites for silencing transcription factors and regulatory proteins expressed in the soma (76, 109). Suppressor alleles for somatic TE activity could be either host or TE encoded, and it is difficult to distinguish whether these are examples of evolved self-regulation of TEs or host-mediated suppression. One interesting case is somatic suppression of the P element by P element somatic inhibitor (PSI). In Drosophila melanogaster, PSI targets P element mRNA precursors. This action results in unspliced P element transcripts that encode a truncated transposase that suppresses P element activity (36, 171), preventing the demise of both the host and the P elements it carries. While the suppression mechanism is unequivocally host derived, this function of PSI is unlikely to be an evolved host response, as the P element invaded the genome of D. melanogaster through horizontal transfer only within the last century (40). Indeed, PSI protein sequence in populations before and after the invasion of P elements shows little difference and is highly conserved across species (107). Rather, the somatic silencing might have been an opportune happenstance that made the P element uniquely suitable for invading the species’ genome, unburdened by the deleterious impact of somatic activity.

4.3.2. Somatic transposable element activity has been widely observed.

While low activity in the soma may be the norm for healthy individuals, there is mounting evidence that somatic activity is commonplace in diseased states. Here, we are specifically referring to TE activity that results in somatic mosaicism, rather than somatic expression that results in germline insertion, as discussed in Section 4.2.

Elevated TE activity in somatic tissues has been observed in many cancers (27) and is associated with aging (71). It remains challenging to determine whether somatic TE activity plays a causal role in any cancer or aging-related disease, but some evidence suggests that it does. Comparative analysis between long-lived, cancer-resistant species and their relatives suggests a correlation between those traits and TE activity (153). Experiments in model organisms also found a causal relationship between aging-associated elevated TE activity and decreased life span (201). In humans, associations between elevated TE expression and several aging-associated diseases, particularly neurological ones, have been widely reported (e.g., 164; reviewed in 152). The proximate causes for the aging-associated dysregulation of TEs likely relate to the destabilization of heterochromatin as individuals age (111). The ultimately evolutionary causes may have their basis in theories of senescence, which are based on the idea that selection to maintain organismal function is relaxed once an individual has aged past a normal reproductive life span (reviewed in 63).

However, TEs can also be active in normal cells. For instance, elevated LINE-1 expression in preimplantation embryos has been found in mammals (86), including humans (129), and has been suggested to play an important developmental role in mice (86, 148). If this LINE-1 expression results in transposition, as has been suggested to occur in human embryos (66, 129, 182), the consequence may be a selfish increase in inherited LINE-1 copies. Yet retrotransposition in embryos may more commonly result in only some cells of the developing individuals carrying the insertion, leading to somatic (or germline) mosaicism (89, 154, 182). Importantly, such mosaicism resulting from embryonic LINE-1 activity can itself be harmful (98).

Elevated TE activity has also been observed in neuronal tissues (44, 130). This somatic TE activity is suggested to confer beneficial functions for the host—particularly in memory formation in mice (4)—and to have other functional (though not necessarily beneficial) effects, such as leading to behavioral differences between monozygotic twins (172). Interest in the role of TE activity in generating neural diversity (7, 55, 181) was spurred by early work estimating impressive TE transposition rates in the brain, in the range of tens per cell (181). Later studies that addressed some tricky methodological issues concluded that the true neuronal insertion rate is likely orders of magnitude lower, with estimates as low as 0.2 insertions per cell (57, 179).

Is this low but appreciable level of somatic mosaicism driven by TE activity likely to have beneficial functions in the host? Normally, to respond to natural selection, mutations should be inherited. While new TE insertions in neurons are not directly inherited, they could still be reproducible across generations if they were restricted to specific genome locations, in the same cell type, at the same developmental time point (155). It is unclear whether LINE-1 activity in the neurons meets this criterion, as unlike other LINE-like elements, LINE-1 targets very simple motifs and, thus, has low site specificity (127). LINE-1 does appear to target genes in at least one part of the brain (100), possibly due to the associated open chromatin (7). However, this nonspecific gene targeting seems more likely to disrupt, rather than improve, gene function. A recent theoretical study suggested an alternative way in which new TE insertions might be beneficial, even in the absence of site specificity (115). Somatic mutations could facilitate the exploration of fitness landscapes and be beneficial when the somatic mutations directly confer cellular advantages (115). This scenario requires that certain mosaic cell types enhance overall neuronal performance, which is currently unknown. In any case, it is clear that TE activity in neurons comes at a cost later in life, given the abundant evidence for the role of TE activity in neurodegenerative diseases (152).

5. EVOLUTIONARY CONSEQUENCES OF VARIATION IN TRANSPOSABLE ELEMENT ACTIVITY

While most TE activity is harmful, even disease causing, some can be beneficial. There are many examples of single TE insertions that have conferred advantageous phenotypes by changing gene regulation [e.g., melanism in the peppered moth (184)] and cases where TE-encoded proteins have been co-opted by hosts [e.g., to perform a key function in pregnancy in placental mammals (104, 124)], both of which have been extensively reviewed elsewhere (29, 39, 60, 64, 88). Here, we focus on the beneficial functions of TEs uniquely mediated by their transposition activity—either directly through insertions of new TEs or indirectly through the distribution of regulatory elements across the genome.

5.1. Evolutionary Adaptation Directly Mediated by Transposable Element Activity

Cases where TE activity is suggested to provide beneficial function—priming innate immunity against foreign viruses in Drosophila (190), preimplantation development in mammals (86, 162), and training self-recognition in the thymus in humans (103)—usually require TE expression or a TE-encoded protein, rather than insertion activity. Another prime example of potential adaptive TE activity is the programmed genome rearrangement mediated by active, full-length class II telomere-bearing elements (TBEs) in the ciliate Oxytricha (132) (see Section 4.3), which minimizes the harmful consequences of high TBE abundance in the Oxytricha genome for both the host and TE (88). Similar to the above cases, this host function depends on TBE-encoded transposase and cis-elements, instead of actual transposition. Therefore, the host may benefit more from co-opting the function from components of inactive TBE, enjoying the beneficial function without suffering from the deleterious insertion activity, as seen in the ciliates Paramecium and Tetrahymena (186).

New insertions from TE activity are usually detrimental, at least with respect to host fitness, in large part due to the randomness of insertions. Accordingly, for TE activity to directly confer predictable and long-term benefits across generations, we argue that insertions should usually target specific genomic locations. Drosophila telomeric TEs, families from the Jockey clade of the LINE-order TEs, are perhaps the best-known example of target-site-specific TE domestication. In most eukaryotes, telomeres shorten with each cell division and are regenerated with the telomerase enzyme (72). In Drosophila, this regenerative function is instead performed by telomeric TE families jumping specifically into chromosome ends (141). These telomeric TEs are not only necessary for telomere regeneration but also sufficient—they can establish new telomeres at broken chromosome ends (178). Because new insertions of telomeric TEs are targeted mainly to chromosome ends, this new telomere maintenance mechanism properly functions without the usual harmful effects associated with TE activity. While this is a classic story of TE domestication, recent evidence suggests something other than domestic bliss: Telomeric TEs and the host proteins that interact with them show rapid evolution and high rates of turnover across the Drosophila phylogeny (108, 161), characteristic signs of an antagonistic arms race between hosts and TEs.

There are other potential cases where TE activity contributes to maintaining the structure and function of the genome, though none are as well established as the telomeric TEs. Centromeres, for example, harbor repeated insertions and rapid expansion of specific TE families (28, 31). A recent study also suggested a beneficial role for the activity of R2, a LINE-order TE family that is highly conserved in animals and targets rDNA arrays in a site-specific manner, to facilitate the regeneration of rDNA copy number through unequal crossing over (131). It is worth noting, though, that compared with other genomic regions, telomere, centromere, and rDNA arrays are highly repetitive in their sequence compositions, and insertions of a TE are less likely to immediately abolish essential functions. Indeed, the centromere-targeting activity of those specific TE families was recently suggested to result from TEs’ selfish activity (78, 200). Also, individual rDNA copies with R2 insertions are usually silenced, and the expression from other rDNA copies can oftentimes compensate to ensure proper cellular functions (53, 180, 209). The suggested regeneration effect thus might occur only in extreme cases when R2 insertions are desilenced due to large numbers of rDNA copies with R2 insertions. Accordingly, in either case, the beneficial effects of site-specific TE insertions could be a secondary consequence of selfish TE activity targeting safe havens that minimize their damage to the host while maximizing TE transmission to the next generation.

5.2. Evolutionary Adaptation Indirectly Mediated by Transposable Element Activity

Unlike typical mutations, TEs can exponentially amplify in genomes, as each new insertion is itself a source for more transpositions. In this way, TE activity can elicit rapid chromosome- or genome-wide changes by providing recurring substrate distributed across many loci. The capability for TEs to disperse the same regulatory elements across the genome and induce concerted evolutionary changes in the regulatory networks was hypothesized early on (24) and has been extensively discussed (39, 60, 64). Here, we discuss some other adaptive changes driven by this unique amplifying property of TE activity.

In a stunning case of chromosome evolution, this property of TE activity underlies the rapid acquisition of chromosome-wide dosage compensation in Drosophila miranda (54, 210). In flies, dosage compensation involves hypertranscription of the X chromosome in males, which is achieved through recognition of a sequence motif scattered across the chromosome. In D. miranda, dosage compensation rapidly arose on a new X chromosome that was previously autosomal (210). Rather than mutations creating recognition motifs independently, ISX, a class II TE containing sequences similar to the recognition motif, was repeatedly inserted across the new X chromosome, establishing chromosome-wide dosage compensation (54). The involvement of rapidly expanding TEs in the evolution of dosage compensation was also recently reported in fish (123).

A thematically similar evolutionary scenario may also underlie dosage compensation in eutherian mammals, including humans. Unlike in flies, dosage compensation involves X inactivation that randomly silences one of the two X chromosomes in females. This chromosome-specific inactivation is enhanced by insertions of LINE-1, which are enriched on the X chromosome relative to the autosomes and can be up to fivefold higher in certain X chromosome segments (6). Regions on the X chromosome with relatively few LINE-1 insertions are enriched for genes that are escapers of X inactivation, suggesting that the absence of LINE-1 insertions attenuates silencing (38). Incidentally, LINE-1 enrichment on the X chromosome dates to the evolutionary origin of eutherian mammals (56), coinciding with the transition from the ancestral mechanism of paternal X imprinting to the derived state of random X inactivation. Because the marsupial X chromosome is not enriched for LINE-1, the burst of LINE-1 activity and resulting chromosome-wide distribution of interspersed cis-elements in eutherians likely facilitated the transition to this new mechanism of dosage compensation (6).

Adaptation involving TE activity may also play a complex role in the degeneration of Y chromosomes. Male-limited Y chromosomes in primates and Drosophila accumulate TEs, other repetitive DNA, and mutations that disable gene functions, likely due to the loss of recombination (reviewed in 5). Once Y-linked genes suffer the accumulation of such mutations, selection may favor the shutdown of the expression of Y chromosome to avoid the deleterious effects of maladapted Y-linked alleles (135). Selection for the predicted loss of transcription of Y-linked genes, mediating by TEs, is potentially occurring on the newly formed Y chromosome in D. miranda. On this former autosome, TEs accumulate near genes, and likely because of their proximity to regulatory factors promoting gene expression that interferes with silencing, these TEs are poorly suppressed (193, 194). The resulting elevated TE activity allows TEs to further amplify on the male-limited neo-Y, increasing the mutational burden specifically in males (194). To restore male fitness, selection may favor the silencing of Y-linked genes in order to silence nearby TEs, likely mediated through the spread of TE silencing to adjacent sequence, a mechanism similar to LINE-1-mediated X inactivation. These examples illustrate how the amplifying capacity of TE activity on specific chromosomes and sexes profoundly shapes the evolution of the genome beyond just singular changes to neighboring loci.

5.3. Contribution of Transposable Element Activity to Reproductive Isolation Between Species

In addition to adaptation, TE activity has long been thought to have the potential to affect another important process in evolution—speciation. Species arise whenever groups of interbreeding individuals no longer successfully reproduce, either because they no longer breed freely with one another or because any resulting offspring are evolutionarily unfit. Here, we focus on genetic incompatibilities between species, where a gene or protein functions abnormally in the wrong species background, because such incompatibilities are the most likely kind to arise from TE activity. It is worth noting, though, that the details of how any particular species formed historically are difficult to unpick. Speciation can involve geographic and ecological as well as genetic factors, and genetic incompatibilities can be a cause or consequence of speciation (45). However, once genetic incompatibilities are entrenched between species, they can act to maintain stable reproductive isolation.

Examples of incompatibilities induced by strain- or species-specific TEs show that TEs can quickly establish postzygotic reproductive isolation. In an extreme example, the horizontal transfer of a TE family carried genetic elements across species boundaries, resulting in incompatibility between the recipient species and a close relative that lacks them (198). Similarly, some recently horizontally transferred TEs, such as the P element in Drosophila, induce sterility in hybrids between invaded and uninvaded strains (95). However, the observed rapid spread of invading P elements (e.g., 80) and the rapid evolution of silencing to suppress them (191) suggest that invading TEs alone are unlikely to result in long-term stable reproductive isolation in the absence of other mechanisms.

Other evidence consistent with a direct contribution of TEs to reproductive isolation comes from elevated TE activity sometimes seen in hybrids (reviewed in 79). Such elevated activity could also be due to incompatibility-driven failures of TE silencing, which suggests the possibility that TE-related genetic incompatibilities arise indirectly through silencing mechanisms that suppress TE activity. In fact, in Drosophila, where genetic incompatibilities are best understood, several studies have shown that incompatibilities involve TE silencing machinery. For example, the rapid evolution of Rhino, a key player in the production of TE-silencing piRNAs, leads to its failure to properly interact with protein partners from closely related Drosophila species (142), likely causing the incompatibility seen in between-species hybrids. Similarly, expression of heterospecific piRNA proteins in a D. melanogaster background resulted in an elevation of piRNAs targeting host genes (189). This autoimmunity of TE silencing, in fact, has been suggested as the underlying explanation for the rapid evolution of the TE-silencing piRNA machinery, rather than a simple host–parasite arms race (19, 43). The contrast with viruses makes the distinction clear: Some viruses encode proteins that directly target host suppressor proteins, potentially triggering bouts of adaptive evolution of the host proteins (134). Yet, unlike viruses, TEs are unlikely to directly target host machinery without compromising their own chances of survival—a TE will also suffer extinction if it excessively harms the host (43). Instead, the piRNA machinery may be evolving rapidly as a result of recurrent bouts of selection to balance TE silencing, as part of the arms race model discussed above, and to avoid deleterious off-target silencing of host genes (19).

Accordingly, examples of genetic incompatibilities that involve TEs are perhaps a subset of conflict-induced hybrid incompatibilities (46). Specifically, these hybrid incompatibilities are a byproduct of genetic divergence—homologs that have accumulated a large number of differences between species may no longer interact normally with other genes in the genome (45). Arms races and genetic conflict, including those mediated by TEs, naturally lead to rapid genetic divergence (see Section 4.1.3) and, thus, escalate opportunities for genetic incompatibilities to evolve.

6. FUTURE OUTLOOK

Here, we have attempted to describe the complex, nuanced factors affecting TE activity. An evolutionary perspective allows us to shed light on the determinants of TE activity in humans, through insights into why and how TE activity evolves between species. Early theories for the evolution of TE activity, despite predating much of our knowledge of TE biology, give context to current empirical observations and suggest new hypotheses to test. Also, by looking at other organisms, we can consider the potential for future or unknown impacts of TE activity on humans. For example, as for other mammals, the risk of humans horizontally acquiring new TEs is likely low, but endogenization of transmissible viruses in other vertebrate species (188), as well as sporadic cases of human individuals with transmissible viruses in germ cells (114), suggests that it is not impossible. Similarly, piRNA silencing of TEs is suggested to evolve to avoid off-target autoimmune effects in flies (189), raising the possibility of similar effects in humans. Finally, TE activity in humans continues to evolve—currently, TE activity in humans is at a historical low (110). Investigation of declining TE activity in the broader context of evolutionary theory with a comparative approach may shed light on the ultimate causes.

TE activity usually has harmful consequences, but, at the same time, its replicative nature can drive large-scale innovations that are not possible with other types of mutations. A fuller understanding of both the good and bad of TE activity is likely to be accelerated by the rapid advances in long-read (99) and single-cell (e.g., 85) sequencing technology. In addition, understanding how TEs affect health may come from studies of the long-term effects of antiviral drugs that inhibit LINE-1 reverse transcriptase and thus reduce human TE activity (143). The sum of these continued efforts in refining theoretical models for evolutionary causes of TE activity and empirical understanding of consequences for genome function may hold the key to understanding how the abundant genomic parasites shape our genomes, health, and individuality.

TAXONOMY OF TRANSPONSABLE ELEMENTS.

To classify TEs into types, the system currently most used is from Wicker et al. (197), where classes are the highest level, followed by orders, superfamilies, and subfamilies. There are two classes (62): class I retrotransposons, which insert new copies via retrotranscription of an RNA intermediate, and class II TEs, which directly excise from and reinsert into the genome. Classes are divided into orders. The class II orders include terminal inverted repeats (TIRs), which often use terminal inverted repeats as transposase binding sites, and Helitrons, which use a rolling circle mechanism to transpose (90). Class I comprises three major orders: long terminal repeats (LTRs), which contain long terminal repeats; long interspersed nuclear elements (LINEs); and short interspersed nuclear elements (SINEs), the latter two of which lack long terminal repeats and are sometimes referred to as non-LTRs. LTRs use retrovirus-like proteins to jump, and in fact, some of them are known as endogenous retroviruses (ERVs). LINEs and SINEs are of great interest due to their dominant presence in mammalian (especially human) genomes.

THE COMPLEX RELATIONSHIP BETWEEN TRANSPOSABLE ELEMENT ACTIVITY AND ABUNDANCE.

TE abundance shows large differences among species; for example, across assembled vertebrate genomes, TE-derived DNA varies from 5% in pufferfish (168) to 70% in poison dart frogs (158). Insects also show a wide range of TE content (1.6–81.5%) (174), and plant TE content ranges up to 90% (147). While observed variation in TE activity may be an important contributor to these differences, TE abundance also depends on historical TE activity and forces that act to remove TEs, including purifying selection that removes harmful TE sequences. Genomes with high TE abundance often contain many relics of ancient, inactive TEs, so that TE abundance may reflect historical, instead of current, TE activity (see Section 3). In addition, some species may suffer systematically more detrimental effects from TEs, such as from accidental off-target effects of TE silencing (84), resulting in stronger selection removing these TEs. Finally, all else being equal, species with large effective population sizes are expected to experience more effective selection purging their TEs (see Section 4).

SILENCING MECHANISMS OF TRANSPOSABLE ELEMENTS IN ANIMALS.

Animals silence TEs both transcriptionally and posttranscriptionally. In the germline, the primary system is mediated by a class of 24–32-nt small RNAs called piwi-interacting RNAs (piRNAs), which recognize TE transcript sequences through complementarity, enabling a family of Argonaute proteins to recognize and slice TEs (139). In addition, piRNA targeting can also induce the deposition of repressive epigenetic marks, such as DNA methylation and H3K9 methylation, at TEs in the genome, achieving transcriptional silencing (47). In tetrapods, a separate TE silencing mechanism involves a diverse class of transcription regulators known as Krüppel-associated box (KRAB) domain zinc fingers (KZNFs), which can directly bind to TEs. These proteins target different TE families by recognizing specific DNA motifs and engaging the cofactor KAP1 through the KRAB domain, which in turn recruits effectors for heterochromatin formation (205). An analogous sequence-specific transcriptional silencing system, zinc finger–associated domain zinc fingers (ZAD-ZNFs), may silence TEs in insects (93).

Glossary

TE activity

insertion of TE sequences into new genomic loci through either copy-and-paste or cut-and-paste mechanisms

Polymorphic

characterized by differences between individuals within a population; here, we use the term specifically to refer to TE insertions being present or absent in particular genomic locations

TE abundance

the amount of TE-derived DNA in a genome or the copy number of TEs in a genome

Autonomous

describes TEs that encode their own enzymes necessary for their own transposition, though some may also use host-encoded factors

Nonautonomous

describes TEs that rely on the products of autonomous TEs for transposition; these are often incomplete versions of autonomous TEs

Genetic drift

changes in allele frequencies due to random effects introduced by finite sampling (e.g., random gamete sampling) rather than selection

Effective population size

the size of an ideal population that would have the same amount of genetic drift as the actual population

Purifying selection

removal of genetic variants that have deleterious fitness effects and are harmful in a population

Transposition–selection balance

an equilibrium where the rate of TE births by transposition is equal to the rate of TE removal through purifying selection

TE burden

reduction in host fitness caused by the presence of deleterious TEs

Panmictic, freely recombining populations

idealized populations that are often used in mathematical models to approximate sexual eukaryotes; loci in these populations are statistically independent

Red Queen hypothesis

the hypothesis that evolutionary change is driven largely by unending antagonistic coevolution between agents to maintain fitness rather than improve it

Horizontal transfer

the movement of genetic material between organisms, including between species, other than the vertical transmission from parent to offspring

Ecological niche

the environmental factors that define where and when an organism (or TE) lives and reproduces

Fitness landscape

the set of complex relationships between genotypes/phenotypes and fitness, conceptualized as peaks (high fitness) and valleys (low fitness) in multidimensional space

Dosage compensation

a process that compensates for differences in the copy number of genes (e.g., between sex chromosomes)

Genetic incompatibility

faulty genetic interactions that cause low fitness in F1 or later-generation hybrids

Postzygotic reproductive isolation

a barrier preventing the hybridization of species due to reduced fitness in the hybrid offspring caused by genetic incompatibilities

Heterospecific

belonging to a different species; here, we use the term specifically to refer to genetic material from another species in transgenic organisms

Autoimmunity of TE silencing

an off-target effect of piRNA machinery that silences host genes instead of TEs