Flies use master lists of DNA sequences from transposons to identify and silence these virus-like, genomic parasites. How the lists themselves escape the fate of their transposon targets has now been solved. See Letter p.54
Transposons are virus-like genomic parasites that propagate by inserting new copies of themselves into the DNA of sperm and egg precursor cells, ensuring the parasites' inheritance. Because transposons can damage DNA when they jump, animals deploy PIWI-interacting RNAs (piRNAs) — short RNA molecules based on snippets of transposon sequence. In fruit flies, piRNAs direct protein complexes to initiate the construction of a DNA–protein complex called heterochromatin that represses transcription of transposon DNA. The piRNA sequences match both the DNA sequences from which they derive and the transposons themselves. Paradoxically, the piRNA-producing DNA must be actively transcribed, whereas transposon transcription must be suppressed. How can identical DNA sequences be transcribed in one location but silenced in another? Andersen et al.1 describe on page 54 how fruit flies solve this problem.
In flies and other insects, piRNAs derive from piRNA clusters — large regions of DNA that act as a record of the transposon sequences that have invaded the species over its evolutionary history2. It has been thought that the enzyme RNA polymerase II (Pol II) transcribes piRNA clusters into long RNAs that are then sliced up into individual piRNAs of 22–30 nucleotides. The piRNAs bind to a protein called Piwi, and this complex binds to newly transcribed transposon RNA that is associated with transposon DNA in the genome3. Piwi, with the help of other proteins, enables enzymes to modify the DNA-packaging protein histone H3 by adding three methyl groups to one of its amino acids, lysine 9 (a modification dubbed H3K9me3)4. The transposon DNA becomes tightly packaged around the modified histones to form dense heterochromatin, the structure of which is thought to inhibit further transposon transcription3,5,6.
H3K9me3 is typically bound by a protein called HP1 that helps to build heterochromatin. But in fruit-fly piRNA clusters, H3K9me3 is bound by an HP1 variant called Rhino7,8,9,10. Unlike HP1, Rhino assembles a protein complex that promotes transcription of piRNA clusters11,12,13,14.
In protein-coding genes, the binding of transcription factors to DNA sequences called promoters directs construction of a transcriptional initiation complex that tells Pol II where to begin synthesizing messenger RNA and which of the two complementary strands of the DNA helix to use as a transcription template (Fig. 1a). Bizarrely, piRNA clusters lack promoters and other typical hallmarks of coherent transcription. How, then, does Rhino facilitate cluster transcription?
Andersen et al. report that piRNA-cluster transcription involves a strikingly different mechanism. The authors show that Rhino and another protein, Deadlock, bind to a protein that the authors dub Moonshiner, which is found only in fly eggs and sperm. Moonshiner is a variant of the core transcription-initiation protein TFIIA-L. The authors find that Moonshiner binds other proteins that are part of the typical transcription-initiation complex, aiding the assembly of an alternative complex. This complex, in turn, binds Pol II, causing transcription to start not at one site on one DNA strand in the piRNA cluster, but incoherently, at many sites on both strands (Fig. 1b).
Formation of this atypical transcription-initiation complex allows piRNA clusters to be packaged into 'silent' heterochromatin, as well as actively transcribed in both directions, generating what is known as sense RNA from the strand typically transcribed during transposon mRNA production, and antisense RNA from the other. Such widespread transcription might explain why searches for long transcripts from Rhino-bound piRNA clusters have been fruitless: no single transcript has been long or abundant enough to convince researchers that it was biologically important.
The study has other implications, too. In flies, most piRNAs are antisense2. This has been attributed to a key feature of piRNA production called the ping-pong amplification loop. The current model for ping-pong amplification proposes that, whenever a silent transposon becomes active, it begins to make sense RNA through a promoter-mediated transcription mechanism. Sense transposon mRNAs are then cut into pieces by nuclease enzymes, which are guided by the antisense piRNAs made from piRNA clusters. Cleavage of the transposon RNA not only blocks transposon activity, but also produces sense piRNAs. The ping-pong model proposes that these sense piRNAs then direct nucleases to cut the antisense RNAs transcribed from piRNA clusters — generating more copies of the original antisense piRNAs1,15. But Andersen and colleagues' findings suggest that this model might require revision.
To explain, Rhino binds to H3K9me3 not only at piRNA clusters, but also on individual transposons throughout the genome11. If the Rhino on transposons binds Moonshiner, transposons would be expected to be transcribed from random positions on both DNA strands. Such transcription might prevent coherent transcription from transposon promoters, blocking production of functional transposon mRNA. Instead of being 'silent', individual transposons might make the unusual type of RNA seen at piRNA clusters. This model predicts that transposons should generate roughly equal amounts of sense and antisense RNA.
If these sense and antisense transposon RNAs were used equally in the ping-pong cycle, they would not drive production of antisense piRNAs, the predominant type of piRNA. Because transposons are present in many copies in the genome, both inside and outside piRNA clusters, it is difficult to assign bits of transposon RNA to a specific location. Thus, incoherent transcription at individual transposons might have been overlooked. But if these Rhino-dependent transposon transcripts exist, new explanations for how piRNA amplification creates an antisense bias in the overall piRNA population will be needed.
Andersen et al. explain how flies transcribe 'silent' heterochromatin such as that found at piRNA clusters, but their findings do not reveal how piRNA-producing DNA is transcribed in other insects or in mammals, which do not make Rhino. However, the authors argue that they may have uncovered a general principle: poorly evolutionarily conserved, cell-lineage-specific, heterochromatin-binding proteins can recruit evolutionarily conserved proteins such as core transcription-initiation factors. In this way, transcription can occur without promoters.
Perhaps the fundamental difference between heterochromatin and the more loosely packed euchromatin, which readily supports conventional transcription, is not that one is silent and the other active. Rather, euchromatin is coherently transcribed, with RNA synthesis beginning and ending at well-defined sites, whereas heterochromatin transcription is incoherent, arising from both DNA strands and beginning and ending at no particular location. Both heterochromatin and euchromatin have a voice, but speak different languages.Footnote 1
Andersen, P. R., Tirian, L., Vunjak, M. & Brennecke, J. Nature 549, 54–59 (2017).
Brennecke, J. et al. Cell 128, 1089–1103 (2007).
Czech, B. & Hannon, G. J. Trends Biochem. Sci. 41, 324–337 (2016).
Yu, Y. et al. Science 350, 339–342 (2015).
Grewal, S. I. S. & Elgin, S. C. R. Nature 447, 399–406 (2007).
Bühler, M. & Moazed, D. Nature Struct. Mol. Biol. 14, 1041–1048 (2007).
Vermaak, D., Henikoff, S. & Malik, H. S. PLoS Genet. 1, e9 (2005).
Volpe, A. M., Horowitz, H., Grafer, C. M., Jackson, S. M. & Berg, C. A. Genetics 159, 1117–1134 (2001).
Klattenhoff, C. et al. Cell 138, 1137–1149 (2009).
Yu, B. et al. Cell Res. 25, 525–528 (2015).
Zhang, Z. et al. Cell 157, 1353–1363 (2014).
Mohn, F., Sienski, G., Handler, D. & Brennecke, J. Cell 157, 1364–1379 (2014).
Le Thomas, A. et al. Genes Dev. 28, 1667–1680 (2014).
Chen, Y.-C. A. et al. Mol. Cell 63, 97–109 (2016).
Gunawardane, L. S. et al. Science 315, 1587–1590 (2007).