Get the jump – Do 3′UTRs protect transposable elements from silencing?

Abstract

Eukaryotic genomes contain large numbers of transposable elements and repetitive sequences that are subjected to silencing through epigenetic mechanisms. These involve primarily DNA methylation, chromatin modifications and small RNA. It is known that these transposable elements can affect the expression of neighboring genes; however, little is known about how transposable element silencing depends on the general chromosomal environment at the insertion site. Taking advantage of the vast genomic resources available in Arabidopsis thaliana, a recent report begins to unravel these interactions by identifying insertion sites of one specific MULE element, AtMu1c across the A. thaliana lineage. Among over 30 insertion sites analyzed, a correlation between the loss of epigenetic silencing and the insertion into the 3′end of protein coding genes was found. Here, we discuss details, implications and potential mechanisms of these findings that may be applicable to a much wider set of transposable elements and across diverse species.

Introduction

A large proportion of eukaryotic genomes consists of transposable elements (TE) and derived repetitive sequences.¹ Through mobilization and transposition these sequences create genetic variation that may threaten genome integrity.² In recent years, however, it is becoming increasingly clear that this genetic variation is an important source of functional variation upon which selection acts during evolution.^3-5 All organisms have evolved mechanisms that repress TE activity and transposition to protect genome integrity.^6,7 However, TE often retain some transcriptional activity at least during certain developmental stages. This may be either due to inherent mechanistic limitations of the silencing process (silencing requires low levels of transcription in order to trigger and maintain the silencing machinery) or due to the potentially advantageous effects of TE activity. In addition, TE mobilization may be induced under certain conditions such as strong environmental stress and hybridization.^5,8-10 This may contribute to the adaptation to changing environmental conditions. While there is some knowledge regarding the effects that TE have on the expression of neighboring genes, very little is known about how the chromosomal environment flanking the insertion site of a TE affects its silencing. A recent report has suggested that certain chromosomal environments may be recalcitrant to the establishment of stable silencing.¹¹

In natural accessions of Arabidopsis thaliana, AtMu1c is an ideal model to study TE silencing at the level of single TE copies. While Mutator-like (MULE) transposons in maize (Zea mays) are highly active and are frequently used as tools for insertional mutagenesis, MULE transposons in A. thaliana are subjected to efficient epigenetic silencing. In fact, until recently it was believed that they do not have any transpositional activity. The natural variation of AtMu1c silencing in A. thaliana accessions can be studied as a model case for natural variation in TE silencing within one species. This species provides a unique opportunity to conduct such a study, as by now several hundred accessions have been re-sequenced using next generation sequencing (see www.1001genomes.org).^12-15 In addition, for many of these accessions expression (RNA-seq) and DNA methylation (BS-seq) data are available, allowing a precise estimate of the silencing state of individual TE elements. RNA-seq allows in many cases to ascribe expression to one of several closely related copies of a TE and thus displays much higher resolution compared to expression analysis by RT-PCR or Northern Blotting.

Correlation Between Reduced Silencing and Insertion into the 3′-end of a Gene

In our recent work, we have exploited the differential silencing of the MULE AtMu1c in the A. thaliana accessions Columbia (Col) and Landsberg erecta (Ler) to gain insight into the determinants of differential TE silencing.¹¹ Through expression quantitative trait locus (eQTL) mapping, we found that the differential silencing was likely caused by transposition and a resulting differential chromosomal environment of AtMu1c in Col and Ler.¹¹ In Ler, AtMu1c was inserted at the 3′-end of the protein-coding gene ERD (EARLY RESPONSE TO DEHYDRATION) SIX-LIKE1 (ESL1) in a region with euchromatic character (Fig. 1A). In contrast, AtMu1c(Col) was intergenic and several kilobases away from the next protein-coding gene. AtMu1c copies in other accessions that are closely related in sequence with AtMu1c(Ler) but inserted into intergenic regions had low transcript levels, suggesting that they are targeted for silencing. Thus, the chromosomal environment at the insertion site rather than sequence differences within AtMu1c appears to be responsible for differential silencing. AtMu1c(Col) remained epigenetically silenced in hybrids and recombinant inbred lines. Interestingly, the silencing did not or only to a very limited extent spread to AtMu1c(Ler) when present in the same cell, although there are siRNA with homology to AtMu1c.^11,16

Figure 1.

(A) Schematic overview of the chromosomal environment of AtMu1c in the A. thaliana accessions Col, Ler, and Qar-8a. Blue arrow, AtMu1c; gray arrows, protein-coding genes; red boxes, 3′UTR of protein-coding gene into which AtMu1c is inserted. (B) Summary of the chromatin characterization of AtMu1c(Col) and AtMu1c(Ler). AtMu1c(Col) is subjected to RdDM and show hallmarks of heterochromatin (histone H3 K9 dimethylation, DNA methylation, siRNAs). AtMu1c(Ler) displays characteristics of expressed genes such as histone H3 K4 trimethylation.

Characterizing the chromatin state of AtMu1c(Ler) and AtMu1c(Col) in the parental lines and in near isogenic lines, we found that AtMu1c(Ler) chromatin resembled that of actively transcribed genes, showing enrichment in histone H3 K4 trimethylation, but little or no H3 K9 dimethylation and low levels of CHG or CHH DNA methylation (H = A, T or C) (Fig. 1B).¹¹ In contrast, AtMu1c(Col) was associated with heterochromatic features such as enhanced DNA methylation and H3 K9 dimethylation but little H3 K4 trimethylation. In addition, the level of 24 nucleotide siRNA complementary to AtMu1c was higher in Col than in Ler. This is in line with the idea that AtMu1c(Col) is silenced by a classic RNA-directed DNA methylation (RdDM)-type mechanism, whereas AtMu1c(Ler) is subjected to no or only limited silencing. Given that both insertions likely have existed for a large part of the A. thaliana lineage (based on sequence polymorphisms and the number and relatedness of accessions that share the respective insertions), it is reasonable to expect that stable silencing should have been established in both AtMu1c copies alike. This suggested that silencing of AtMu1c(Ler) is either actively prevented by the location or that some essential component is missing at AtMu1c(Ler).

From the bioinformatic analysis of genome and transcriptome data of 217 accessions,¹³ we identified 4 other accessions with the AtMu1c(Ler)-type of insertion (that contained no additional AtMu1c copies).¹¹ All of them showed high AtMu1c expression. Whether there is a correlation with the structure of the flanking chromatin remains to be investigated. In addition, we identified one additional insertion within the 3′UTR of a protein-coding gene (At1g77525) from the Qar-8a accession. AtMu1c(Qar-8a) also showed loss of silencing. In the case of AtMu1c(Ler) and AtMu1c(Qar-8a) the insertion is within the annotated 3′UTR of the protein-coding gene. However, 3′RACE and RT-PCR experiments suggested that the transcripts terminate immediately before the AtMu1c insertion regardless of its presence (Ref. 11 and data not shown). Furthermore, we did not find evidence for read-through transcription into the TE.¹¹ None of the other insertions that were analyzed had similarly high transcript levels. Together, these experiments indicate that AtMu1c can escape epigenetic silencing by inserting into specific genomic locations, such as the 3′-end of protein-coding genes. It will be interesting to see whether this observation extends to other TE families and to identify determinants that mediate this position-dependent escape from silencing.

Possible Mechanisms for the Loss of Silencing

What are possible mechanisms for the observed escape from silencing? As both transcription and silencing happen in the context of chromatin,^6,17 it is possible that an unidentified chromatin property at the 3′-end of genes interferes with the establishment of stable silencing. Stable TE silencing though RdDM involves DNA methylation and decoration of histone tails with repressive modifications such as H3 K9 dimethylation.⁶ It is possible that a mechanism exists to exclude such modifications at the 3′-end of genes in order to prevent the adjacent gene from being silenced. The nature of this chromatin property is an exciting area of study.

The 3′-end of a gene is also the region of 3′-processing of nascent transcripts, as 3′-end processing occurs in close proximity to transcription.^18–20 Signal sequences in the nascent transcript are bound by 2 multiprotein complexes called CPSF and CstF.^19,20 The nascent transcript is then cleaved and the poly(A) tail added by Poly(A) polymerase. Transcription typically proceeds beyond the polyadenylation site before it is terminated.¹⁸ Improperly processed transcripts are subjected to rapid degradation through the XRN and exosome pathways.²¹ Thus, another possibility for the escape from silencing may be that the 3′-end processing machinery or the mRNA degradation machinery prevent silencing of TE inserted into 3′-UTRs. This may be achieved by the rapid degradation of improperly processed transcripts that may otherwise trigger the silencing machinery.

How Frequent are 3′UTR Insertions?

Among the 32 novel AtMu1c insertions that were newly identified in our study,¹¹ 4 were associated with annotated 3′UTRs. - Expression could be analyzed only for 2 of those as the remaining 2 co-occurred together with other insertions.- Is this more or less than expected assuming random insertion into the genome? According to TAIR genome annotation, the A. thaliana genome is 119.146 Mb long and contains 27206 nuclear protein-coding genes.²² With an average 3′UTR being 233 bp long,²³ we expect 5 % of insertions to occur in annotated 3′UTRs. Interestingly, we observed a rate of 12% (4/32). It is unknown whether AtMu1c has any insertion site preferences. For maize AtMu1c it has been shown that integration preferentially occurs into regions of open chromatin at the 5′end of protein-coding genes.^24-26 As the AtMu1c insertions in our study were identified from natural populations, it is also not clear how the outcome has been shaped by natural selection, which is expected to eradicate TE insertions with detrimental effects on neighboring genes. In fact, we did not find an effect on neighbor gene expression for AtMu1c in Ler or Qar-8a. Interestingly, only one of the 32 AtMu1c insertions disrupted the coding sequence of a gene by inserting into an intron and there the insertion was truncated.

How Does Chromosomal Location Affect the Silencing State of TE?

It is long known that plant chromosomes contain constitutive heterochromatin that is so densely packed that it can be distinguished from euchromatin cytologically. Such heterochromatin is located around the centromeres and is packed with TE and repetitive sequences that are deeply and stably silenced, but contains very few genes.^27,28 Almost all AtMu1c insertion sites are outside the pericentromeric heterochromatin (Fig. 2). A few studies have investigated the interaction between genes and neighboring TE at a global scale in plants, however, none of them addressed the effect of nearby genes on TE silencing. This may in part be due to the difficulty of studying the silencing level of individual TE. Hollister and Gaut reported a negative effect of methylated but not unmethylated TE on nearby genes.²⁹ Similarly, Maumus and Quesneville found a negative effect of flanking repeats with complementary siRNA on neighboring genes.³⁰ Meanwhile, the effect of neighboring genes on TE silencing remains unclear. The analysis of natural variation of low copy TE may prove useful in addressing the question how chromosomal location affects the silencing state of TE as closely related elements with different insertion sites can be analyzed individually.

Figure 2.

Schematic overview of the distribution of AtMu1c insertion sites along the 5 chromosomes of A. thaliana. AtMu1c insertion sites are marked with a black dash. The approximate location of the pericentromeric heterochromatin³¹ is shown as dark gray rectangles. Most AtMu1c insertions are outside the pericentromeric heterochromatin.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Acknowledgments

We thank M. Lenhard for helpful comments on the manuscript.

Funding

We acknowledge funding from the Alexander-von-Humboldt-Foundation through a Sofja-Kovalevskaja-Award to I.B..