Tag Archives: piRNA

The Transposon/piRNA/Chromatin Nexus

Close observation of chromatin states at piRNA-silenced genomic loci demonstrates the power of transposons to change native gene expression.

As reviewed in an earlier post, the Drosophila Piwi/piRNA transposon silencing pathway can be divided into two facets; a complex pathway operating in the germline centred on the Piwi-family argonautes Aubergine and AGO3 localised in peri-nuclear nuage, and a linear pathway operational in the somatic follicle cells. In this linear pathway, piRNAs derived from uni-directional piRNA clusters such as flamenco target Piwi to mediate silencing of a limited subset of retrotransposons. Unlike Aub and AGO3, Piwi is localised to the nucleus, leading to speculation that rather than silencing transposons post-transcriptionally by ‘slicing’ their transcripts, it may act at the transcriptional level. There are many precedents in other organisms for argonautes mediating transcriptional silencing via interactions with chromatin modification and DNA methylation pathways. However, whether one of these silencing modes is employed by Drosophila Piwi was unresolved. A new paper from the lab of Julius Brennecke, generally analysing the linear piRNA pathway active in a cell line derived from the somatic follicle cells surrounding the oocyte (OSC cells) includes important findings for a number of aspects of Piwi-mediated transposon silencing leading to insights on the wider genomic ecology of transposon insertions.

In the first section of the paper, Sienski et al. demonstrate that Maelstrom (Mael), a protein containing putative RNA and DNA binding domains, expressed in both cytoplasm and nuclei and previously implicated in a number of Piwi-pathway effects, acts downstream of Piwi to effect TE silencing. Silencing requires the nuclear localisation of both Piwi and Mael. Further, mutation of the residues necessary for ‘slicer’ activity in Piwi did not de-repress TEs, suggesting a different mechanism for Piwi-mediated silencing.

Sienski et al. go on to marshal three different high-throughput techniques to show that Piwi mediates gene silencing at the transcriptional level. Knocking down (KD) the expression of Piwi pathway factors (piwi, mael) in OSC cells they determined the set of repressed transposable elements (TEs) by comparing RNA levels (RNA-seq). Changes in the steady-state RNA levels were highly correlated with transcription rate as monitored by RNA polymerase II occupancy (ChIP-seq) and levels of nascent RNAs (GRO-seq). Judging by how closely correlated derepression of TEs was to transcription rate, it seems unlikely that the linear piRNA pathway active in follicle cells acts post-transcriptionally at all.

Reasoning that Piwi-mediated transcriptional gene silencing may involve chromatin modification, Sienski et al. profiled the distribution of the repressive histone mark H3K9me3 in OSCs after piwi or mael knockdown. H3K9me3 levels at transposable elements known to be repressed by the piRNA pathway were significantly reduced in the absence of Piwi (and to a lesser extent Mael). This data was from across the genome irrespective of whether the TE was inserted into heterochromatic or euchromatic regions. To negate general effects associated with heterochromatin, the authors looked more closely at TE insertions within euchromatic regions.

Approximate sketch of the patterns of RNA pol II occupancy (ie Transcription), and H3K9me3 at the mdg1 locus after piwi or mael knockdown and normally in control.

Approximate sketch of the patterns of RNA pol II occupancy (ie Transcription), and H3K9me3 at the mdg1 locus after piwi or mael knockdown and normally in control.

At a specific euchromatic insertion of the retrotransposon mdg1, they observed that upon either piwi KD or mael KD, transcription downstream of the insertion strongly increased. However, although this transcriptional bleeding into the surrounding area was similar upon TE derepression due to either piwi KD or mael KD, the pattern of H3K9me3 was very different. Normally this mdg1 insertion displays H3K9me3 in the surrounding 12kb, peaking at the insertion site. This was strongly reduced in piwi KD cells, but in mael KD, H3K9me3 was moderately reduced at the insertion site but had actually spread further downstream (see figure). Similar patterns were observed at nearly all euchromatic mdg1 insertions, as well as other TEs known to be targeted by the linear piRNA pathway active in OSC cells.

Strikingly, most euchromatic H3K9me3 peaks were sensitive to piwi knockdown, whilst 88% of H3K9me3 peaks were found within 5Kb of TE insertions. Piwi-mediated transposon silencing therefore seems to be the main trigger for H3K9 trimethylation in euchromatin.

This transposon silencing mechanism appears to have a major impact on native genes upon TE insertion in their vicinity. An insertion of the retrotransposon gypsy into the first intron of the expanded (ex) gene serves as paradigm for these effects. In OSC cells, the gypsy insertion triggered H3K9me3 spreading into the surrounding 10-12Kb. In control cells RNA polymerase II occupancy was observable at the ex transcription start site (TSS) but weak. Upon piwi or mael knockdown, transcription from the ex TSS was massively increased. As in the earlier mdg1 example, H3K9me3 levels were greatly reduced upon piwi KD but not in mael KD cells. Sienski et al. observed similar effects on the transcription of 28 more genes with nearby TE insertions in OSC cells.

This data has a number of ramifications speaking of a complex interplay between transcription, the establishment and maintenance of repressive chromatin states and the Piwi pathway. Firstly, H3K9me3 considered a transcriptionally repressive histone mark is compatible with transcription. In fact, based on it’s pattern in mael KD cells, the authors propose that downstream transcriptional bleeding leads to the spread of H3K9me3. Further, although H3K9me3 has an integral role in Piwi-mediated silencing, it is not the final silencing mark. H3K9 trimethylation is downstream of Piwi action, but is either upstream or acts in parallel to Mael, which mediates an unknown silencing step crucial to Piwi transcriptional gene silencing.

Importantly, this paper has demonstrated the impact that TE insertion and subsequent piRNA pathway transcriptional repression can have on native gene expression. There are two different modes in which the inactivation of Piwi-mediated TE silencing can lead to the transcriptional activation of these loci. Firstly, the spreading of repressive chromatin marks at transposons can suppress RNA polymerase II access to the genes promoter. Alleviation of TE repression hence leads to (re-)activation of gene expression. Conversely, as TEs (especially the long terminal repeats of some retrotransposons) can serve as promoters, the loss of their repressed chromatin state upon piRNA pathway loss, can activate transcription of downstream regions. Although both these modes lead to transcriptional activation after Piwi pathway loss, they demonstrate that transposon insertion can either activate or repress transcription within relatively extensive genomic surroundings. This underscores the scope for transposons to act as regulatory elements, or to produce new chimerical transcripts and hence potential new genes.

These experiments were mainly performed in one cell type that only partially reflects the activity of what is already a subset of piwi/piRNA action during Drosophila oogenesis.  Piwi and Mael are also active in the nurse cells and oocyte, and this paper suggests that they have similar roles within the context of the expanded piRNA pathways active in the germline. It will be interesting to integrate this nuclear-localised transcriptional-silencing aspect of piRNA silencing into the context of ping-pong amplification and bi-directional piRNA cluster transcripts. Further, do these Piwi-mediated chromatin effects in the germline impact on the transcriptional status of TEs and genes later in somatic development? And if not, do other systems have equivalent activity?

This paper underlines again the importance of the arms race between mobile genetic elements and genomic immune systems such as the piRNA pathway on the wider genomic regulatory context. This contest is being observed to have shaped so many aspects of genome organisation throughout evolution that it sometimes becomes hard to differentiate parasitism from regulation. It is clear however, that to understand the evolutionary impact of mobile elements we must also understand the import of the various epigenetic mechanisms controlling their spread. The minutiae of these mechanisms with regard to their targets, plasticity, adaptability, heritability – often different from organism to organism – has major evolutionary significance. Evolution works differently depending on these mechanisms.

Sienski, G., Dönertas, D., & Brennecke, J. (2012). Transcriptional Silencing of Transposons by Piwi and Maelstrom and Its Impact on Chromatin State and Gene Expression Cell, 151 (5), 964-980 DOI: 10.1016/j.cell.2012.10.040

Uploading piRNAs to the Cloud.

A new paper finds a protein linking piRNA transcription with processing in nuage.

The Piwi/piRNA system is responsible for protecting the germline from the mutagenic effects of transposon mobilisation. As summarised in an earlier post, in Drosophila large arrays of transposon fragments, located in pericentromeric and subtelomeric chromatin domains give rise to long piRNA cluster transcripts. These transcripts are then processed to produce the 23-30 nt piRNAs which, when complexed with Piwi-family argonaute proteins effect the post-transcriptional silencing of transposons. Although a more limited piRNA system functions in the somatic follicle cells surrounding the Drosophila egg chamber, the bulk of germline transposon silencing is performed by the system active in the germline siblings of the oocyte – the nurse cells. Here, dual-strand piRNA cluster transcripts are processed in the nuage, a perinuclear electron-dense cytoplasmic structure, where the ‘ping-pong’ system of reciprocal cutting and complexing between the Piwi proteins Aubergine (Aub) and Ago3 leads to piRNA amplification.

Nuage is a hallmark of germline cytoplasm in animals, and appears to be the site of both piRNA processing and transposon silencing. A hierarchy of proteins responsible for the assembly and function of nuage has been revealed by studies in Drosophila. Vasa, a DEAD-box RNA-dependent helicase protein, is required for the localisation of Tudor and other Tudor-domain-containing (Tdrd) proteins. These serve as a platform for the piRNA system, binding Aub and Ago3. Defects in many of these piRNA biogenesis components do not just lead to uncontrolled transposon activity; rather, they affect the asymmetric localisation of RNAs in the developing oocyte – a process by which developmental prepattern is organised. Zheng et al. discovered that weak mutations in the uap56 gene caused similar defects, suggesting a potential role in piRNA biogenesis.

UAP56 is another DEAD-box containing RNA-binding protein. It is ubiquitously expressed, localised in nuclei and has previously been shown to be involved in mRNA splicing and export. Zheng et al. found that in nurse cells it localises to discrete foci in the periphery of the nucleus. This was a similar pattern to that of Rhino (Rhi), a Heterochromatin Protein 1 variant previously shown to associate with piRNA clusters. Indeed, UAP56 and Rhino co-localised ~99% of the time in nurse cell nuclei.  Mutations in either uap56 or rhi caused a failure in the focal localisation of the other protein, showing their co-dependence.

When Vasa was imaged at the same time, it became apparent that it localised to foci in the nuage directly across the nuclear envelope from UAP-56-Rhi foci. Co-labelling with a nucleoporin showed that in fact UAP56-Rhi foci and Vasa foci directly abut nuclear pores from either side.

In the absence of functional UAP56 the nuage fails to assemble properly; Vasa, Aub and Ago3 all fail to localise. Similar effects are observed in rhi mutants, placing both UAP56 and Rhino upstream of Vasa as extrinsic factors necessary for nuage assembly. The uap56 mutants also fail to produce a large part of the proper complement of piRNAs leading to a consequent mobilisation of transposons. No effects on the level of genic mRNAs were detectable. Due to the failure of nuage assembly, the uap56 mutants also display germline DNA damage and the morphological defects caused by mislocalisation of asymmetric RNAs.

DEAD-box containing proteins act as ATP-dependent RNA clamps. As Rhino is known to associate with dual-strand piRNA clusters, Zhang et al postulated that UAP56 may be binding and stabilising nascent cluster transcripts. Indeed piRNA cluster transcripts could be co-immunoprecipitated with UAP56 and Vasa.

The data therefore suggests an attractive model in which cluster transcripts are passed across the nuclear pore between the two DEAD-box containing proteins, UAP56 and Vasa. The authors term this a nuclear pore spanning piRNA processing compartment. piRNA cluster transcripts must in some way be marked and specifically transported via the trans– nuclear pore compartment.

Running through this work as a consistent undertone are the implicit links to the broader RNA processing systems. The nuage is obviously intricately linked to the differential transportation of RNAs from the nurse cells and around the oocyte. UAP56 has other roles in mRNA splicing and export from the nucleus. What exactly are the links between the germline specific role of UAP56 and the general RNA splicing and export machinery? Zhang et al end with the enticing observation that mutations in two different genes encoding conserved exon junction splicing components also lead to similar asymmetric RNA localisation defects. It appears that the control of piRNA processing and transposon silencing in nuage is intimately linked to broader networks controlling germline specification and the patterning of the oocyte. Although the different strands of these systems are difficult to tease apart, Drosophila oogenesis continues to offer an unparalled paradigm for their investigation. The piRNA system is widely conserved in animals, but there does appear to be quite a lot of plasticity in its specifics. For instance, as discussed at length in this series of posts, in C. elegans, piRNAs are individually transcribed. I’d be very interested to find out whether homologues of Rhino and UAP56 play any role in this system? I’ll riff on the similarities and differences of piRNA systems and their links to development some more in future posts.

Zhang, F., Wang, J., Xu, J., Zhang, Z., Koppetsch, B., Schultz, N., Vreven, T., Meignin, C., Davis, I., Zamore, P., Weng, Z., & Theurkauf, W. (2012). UAP56 Couples piRNA Clusters to the Perinuclear Transposon Silencing Machinery Cell, 151 (4), 871-884 DOI: 10.1016/j.cell.2012.09.040

Lin, H. (2012). Capturing the Cloud: UAP56 in Nuage Assembly and Function Cell, 151 (4), 699-701 DOI: 10.1016/j.cell.2012.10.026

Interacting small RNA pathways in worms 4: 21U-RNAs.

A survey of C. elegans small RNAs from 2006 (Ruby et al) first reported the discovery of a large class of 21nt RNAs with 5’ uridines – 21U-RNAs. The majority of these RNAs mapped to two distinct regions of the C. elegans genome located on chromosome IV. Generally, they were found between genes or in introns, and showed no sense or antisense bias (when within introns). Each 21U-RNA genomic locus shared an upstream sequence motif. This 34bp motif centred on an 8bp core consensus located 38bp upstream from the start of the 21U-coding sequence. Ruby et al had mapped 5,302 unique 21U-RNAs, but using the upstream consensus they identified 10,807 putative 21U-RNA generating loci on chromosome IV. The functional significance of the upstream motifs was not clear; it seemed probable that they could act as promoters, and that 21U-RNAs were individually transcribed, but it was also feasible that these sequences were sites for targeted cleavage of longer transcripts, or that they were recognition sequences for RNA dependent RNA polymerases. Interestingly, the upstream motif was conserved in the related nematode C. briggsae, but none of the 21U-RNAs themselves were. Very few 21U-RNAs matched other sequences in the genome, but in total they encoded a huge diversity of sequence. It appeared that evolutionary pressure was maximising this sequence complexity rather than maintaining sequence identity. The most important questions on 21U-RNAs were therefore clear from the start; what are the targets? What’s the function of the system?

In 2008, parallel studies from the labs of Craig Mello and Eric Miska further characterised 21U-RNAs (Batista et al. Das et al). As well as bringing their total number to 15,722, they discovered that they were solely expressed in the germline and associate with a Piwi-family Argonaute protein, PRG-1. prg-1 null mutants display dramatic reductions in germ cell numbers as well as an independent temperature sensitive fertility defect. 21U-RNAs fail to accumulate in prg-1 mutants, and co-immunoprecipitate with PRG-1. Like many other Piwi proteins, PRG-1 is localised to the perinuclear nuage in the germline (P-granules).

21U-RNAs are therefore the piwi-interacting RNAs (piRNAs) of C. elegans. Like piRNAs in vertebrates and Drosophila, they are 5’ monophosphorylated, have 5’ uridine residues, and are 3’ modified. Most importantly they interact with a Piwi-family AGO, and are implicated in germline functions. However, they do display some major differences not encountered in piRNAs from other clades; they are only 21nt long (whilst those in vertebrates and Drosophila are 24-30nt in length) and most strangely they appear to be individually transcribed. As with 21U-RNAs, piRNAs in other animals are generated from large genomic clusters. However, in Drosophila and mammals individual piRNAs  derive from larger cluster transcripts. Another important difference with other piRNA systems was the lack of a ping-pong piRNA amplification system.

As the main known role for PIWI/piRNA systems is the silencing of transposable elements, Das et al. and Batista et al. looked for evidence of a similar function for the PRG-1/21U-RNA system. Das et al. screened worms mutant for both prg-1 and the closely related gene prg-2 for signs of transposon desilencing. The only transposon found to be affected was Tc3.  Expression of the Tc3 transposase mRNA was higher, and the reversion rate of mutations caused by Tc3 insertions increased 1000 fold in the double mutants (interestingly, the reversion rate was up only 100 fold in prg-1 single mutants; the only evidence of prg-2 having a functional role in the 21U-RNA system). Das et al. identified a single 21U-RNA that mapped to the sense strand of the Tc3 transposase gene. They then found that a large number of endogeneous siRNAs (ie WAGO associated 22G-RNAs) targeted against both the transposase gene and the terminal inverted repeats (TIR) were strongly depleted in the prg1+2 mutants. Batista et al reported slightly different findings: They identified a different 21U-RNA mapping to the TIR of Tc3 in the same orientaion as the transposase gene. When they searched for endo-siRNAs against Tc3 that were depleted in the prg-1 mutants, they only found that those targeted against the TIR were affected.  The discrepancies between the two studies could indicate slightly different roles for the two PRG proteins.

Importantly, both studies found that the PRG-1/21U-RNA system functions upstream of the WAGO/22G-RNA system in transposon control. The specifics were rather hazy though. The 21U-RNAs matching Tc3 sequence were both orientated sense to the transposase gene, and so would not be able to base-pair with the transposase mRNA. The 21U-RNA identified by Batista et al. wasn’t even directed against a part of the transposon expected to be transcribed. However, the 22G-RNAs that were sensitive to prg-1 function were generally antisense to the direction of transposase transcription. These findings suggested a model in which transcription from downstream genomic regions may generate antisense Tc3 transcripts which would be recognised by a PRG-1/21U-RNA complex. This would then trigger the RdRP-dependent synthesis of 22G-RNAs against the transposon transcripts, that would then silence the transposons most probably through alterations to chromatin structure.

Linking the PRG-1/21U-RNA system to the WAGO/22G-RNA was a major step, and could be a general mechanism for 21U-RNA action. The 22G-RNA mediated stage of the process can be viewed as equivalent to the ping-pong amplified secondary piRNA stage in Drosophila transposon silencing. However, these findings raised important questions; If this is a system for the control of mobile elements, why was only one transposon found to be desilenced in prg-1 mutants, and so few 21U-RNAs found to match transposon sequences? And what about the vast number of 21U-RNAs without identity to other genomic sequences? A similar phenomenon is seen in mammals, in which a huge pool of ‘pachytene’ piRNAs without known targets or functions, are found. To explain this enigma, Batista et al. suggested that 21U-RNAs could base-pair imperfectly with their targets. microRNAs are able to recognise their targets by base-pairing within a ‘seed’ region driving less stringent pairing between the rest of the molecules. If 21U-RNAs work by a similar mode, the whole worm transcriptome could be under piRNA-directed regulation. Only modest changes in general gene expression were found in the prg-1 mutants, so it’s unclear what this global genome regulation or surveillance system really means.

I get the impression that a dichotomy of interpretations has arisen between the two principal labs studying 21U-RNAs. This can be broadly expressed as Eric Miska considering 21U-RNAs as a mechanism primarily directed against genomic parasites, whilst Craig Mello favours a broader ‘global surveillance system’ interpretation. The next post, on the recent papers from these labs, will try to dissect these different interpretations and clarify the current state of knowledge of C. elegans piRNAs.

Stop Press: A new paper has just cleared up the contentious issues of the function of the upstream motif, and whether 21U-RNAs are individually transcribed. Cecere et al. show that the upstream motifs are indeed promoters. Forkhead family transcription factors bind the motif, and drive the separate expression of the thousands of 21U-RNAs.

Ruby JG, Jan C, Player C, Axtell MJ, Lee W, Nusbaum C, Ge H, & Bartel DP (2006). Large-scale sequencing reveals 21U-RNAs and additional microRNAs and endogenous siRNAs in C. elegans. Cell, 127 (6), 1193-207 PMID: 17174894

Das PP, Bagijn MP, Goldstein LD, Woolford JR, Lehrbach NJ, Sapetschnig A, Buhecha HR, Gilchrist MJ, Howe KL, Stark R, Matthews N, Berezikov E, Ketting RF, Tavaré S, & Miska EA (2008). Piwi and piRNAs act upstream of an endogenous siRNA pathway to suppress Tc3 transposon mobility in the Caenorhabditis elegans germline. Molecular cell, 31 (1), 79-90 PMID: 18571451

Batista PJ, Ruby JG, Claycomb JM, Chiang R, Fahlgren N, Kasschau KD, Chaves DA, Gu W, Vasale JJ, Duan S, Conte D Jr, Luo S, Schroth GP, Carrington JC, Bartel DP, & Mello CC (2008). PRG-1 and 21U-RNAs interact to form the piRNA complex required for fertility in C. elegans. Molecular cell, 31 (1), 67-78 PMID: 18571452

Cecere G, Zheng GX, Mansisidor AR, Klymko KE, & Grishok A (2012). Promoters Recognized by Forkhead Proteins Exist for Individual 21U-RNAs. Molecular cell PMID: 22819322

Interacting small RNA pathways in worms 1: Introduction

A cluster of new papers, in the journals Cell and Science, discuss the links between piRNAs and various endogenous siRNA pathways in the nematode worm C. elegans. Emerging from these experiments is a picture of a genome-wide surveillance system capable of differentiating between self and non-self nucleic acids. The commonalities and differences between these papers require rather detailed analyses. I’m therefore intending to write a series of posts; first covering some of the background information on these small RNA systems and then getting onto the new findings.

A panoply of small RNA molecules, involved in diverse cellular functions have been discovered in the wake of the initial observation of RNA interference (RNAi). Originally RNAi described the mechanism by which genes could be specifically silenced by the exogenous application of cognate double-stranded RNAs. Nowadays, the term RNAi is more generally applied to gene silencing pathways involving the three major classes of small RNAs; microRNAs (miRNAs), small-interfering RNAs (siRNAs), and piwi-interacting RNAs (piRNAs). A common feature of all these small RNAs is that they complex with members of the Argonaute (AGO) family of proteins. Embedded within AGOs, the small RNAs act as guides; base-pairing with specific target RNAs that can then be cleaved by the RNase H endoribonuclease activity of the AGO protein. However, not all argonautes act by this ‘slicing’ activity; gene silencing can also be achieved by interactions with pathways involved in chromatin modification, or the inhibition of transcriptional elongation. Meanwhile the list of non-silencing roles of AGOs and their complexed small RNAs continues to grow; chromosome segregation, double-strand break repair, programmed genomic rearrangement etc.

The synthesis of both miRNAs and siRNAs generally involves the recognition of dsRNA and its cleavage by Dicer enzymes. miRNAs are derived from short stem-loop structures found in transcripts. siRNAs are the main effectors of the ‘classical’ RNAi. Exogenous dsRNA molecules are cleaved by Dicer into 20-30nt siRNAs that are loaded onto AGOs. However, this pathway also targets endogenously formed dsRNA, which can derive from hairpin structures in transcripts, or by base-pairing between transcripts produced either from separate loci, or by bidirectional transcription at individual loci. Hence, endogenous siRNAs generally target transposons or other repetitive sequences, but can target genes as well.

This basic siRNA pathway is present in most animals, but more complicated systems exist in plants, fungi and nematode worms. These ‘secondary siRNA’ systems wrest on the use of RNA dependent RNA polymerases (RdRPs) to amplify siRNAs against targets recognised by primary siRNAs. In the cases of plants and fungi the dsRNAs produced by RdRPs are again processed by Dicer enzymes and then loaded onto AGOs. However, various populations of RdRP-produced siRNAs in C. elegans do not require Dicer cleavage.

As noted above, the key effectors of these small RNA pathways are argonaute proteins. The numbers of AGOs present in organisms varies widely. The Drosophila genome encodes 5 AGOs; 3 of these are involved in the piRNA system, whilst the other two specialise in either miRNAs or siRNAs. In contrast, C. elegans has 27 AGOs. This reflects the presence of various additional networks of endogenous- siRNAs. Deep sequencing in C. elegans has revealed a large diversity of different varieties of small RNAs, with major peaks at 21, 22, and 26 nt. Different types of sRNAs have different biases in relation to their predominant 5′ residue, 3′ modifications, and extent of 5′ phosphorylation.

This series of posts will ignore many classes of C. elegans siRNAs and instead concentrate on two varieties of 22nt endo-siRNAs with 5′ guanosine residues (22G-RNAs). 22G-RNAs associated with the AGO CSR-1 have been shown to play critical roles in chromosome segregation during meiosis and mitosis. Another population of 22G-RNAs that associate with various worm-specific AGOs (WAGOs) have been implicated in the long-term silencing of transposons and other genomic loci. The piRNAs of worms – 21U-RNAs – display some critical differences with those found in Drosophila and mammals. However, understanding their role in C. elegans may help to explain some of the outstanding questions about their functions thoughout animals.

Ketting RF (2011). The many faces of RNAi. Developmental cell, 20 (2), 148-61 PMID: 21316584

See also these related posts:
small silencing RNAs. I: Piwi-interacting RNAs.
RNAi and Chromatin Modification
Lamarckian inheritance of antiviral response in Nematodes.
Double-strand break interacting RNAs (diRNAs)

piRNAs in the brain: epigenetics and memory

An exciting new paper in Cell, links Piwi-interacting RNAs (piRNAs) to long-term memory via the epigenetic regulation of gene expression by DNA methylation.

Two different novel findings are especially important: piRNAs had been thought to be a germline specific mode of genetic control, specifically a type of genetic immunity against the mobilisation of transposable elements. Rajasethupathy et al demonstrate that in the sea hare, Aplysia, piRNAs are expressed in the CNS and other somatic tissues. Secondly, this paper demonstrates specific dynamic de novo methylation of the promoter of a gene regulating neuronal plasticity in response to neurotransmitter- mediated excitation. This provides an epigenetic mechanism by which memories can persist by molecular encoding.

To investigate microRNAs expressed in the Aplysia CNS, Rajasethupathy et al had constructed a small RNA library. Surprisingly, they found that ~20% of the sequence reads from this library were ~28nt long, and showed a bias towards having 5′ uridine residue. This fitted with them being Piwi-interacting RNAs (piRNAs) rather than microRNAs. When they were mapped to the Aplysia genome, it was clear that the piRNAs were generated from piRNA clusters (see previous introductory post). After constructing more libraries and deep sequencing, Rajasethupathy et al. identified 372 distinct Aplysia piRNA clusters. Certain piRNAs are found far more commonly than surrounding piRNAs from the same cluster, indicating that an amplification process is occurring during piRNA biogenesis. Although overall piRNA content was highest in the ovotestes (ie the germline and associated somatic tissues), various piRNAs were found to be enriched in the CNS, as well as other somatic tissues analysed.

Consistent with the presence of piRNAs in the CNS, Rajasethupathy et al. were able to clone a full-length cDNA for Piwi protein from Aplysia CNS. Using an antibody against Piwi, they were then able to co-immunoprecipitate piRNAs with Piwi protein from the CNS. By separating cell nuclei from cytoplasm and then western and northern blotting against Piwi and piRNAs respectively, they showed that both were primarily found in nuclei.

To briefly comment on the piRNA aspect of this paper; a number of outstanding questions arise. In the best characterised model systems, piRNA amplification occurs by the ‘ping-pong’ mechanism in which reciprocal recognition and cleavage reactions between sense and antisense piRNAs complexed with the (cytoplasmic) Piwi-related proteins Aubergine and AGO3 (Drosophila terminology), leads to selective amplification of piRNAs with a tell-tale 10nt offset. Neither of these proteins, nor the 10nt offset, nor the ratio between sense and antisense piRNAs are mentioned by Rajasethupathy et al. I imagine that these questions would’ve been looked into and mentioned if found, therefore it seems that the mechanism of piRNA amplification in the Aplysia CNS is potentially novel.

To explore possible functions of piRNAs in the Aplysia CNS, the researchers used a co-culture system that monitors changes in synaptic plasticity in response to stimulation by the neurotransmitter serotonin (5HT). In this assay, two sensory neurons synapse with a motor neuron. Changes in the strength of one sensory-motor synapse are monitored by electrophysiological recording from the motor neuron. This system measures long-term facilitation (LTF): ie. changes in synaptic strength in response to 5HT stimulation. LTF is considered to be a memory-related phenomenon, however, it is contentious just how well it serves as a paradigm for long-term memory.

Knockdown of Piwi (and hence of complexed piRNAs), by the injection of antisense oligonucleotides into one of the sensory neurons, significantly impaired LTF, whilst overexpression of Piwi enhanced it. To investigate how these Piwi effects on LTF were mediated, Rajasethupathy et al. looked at the expression of proteins known to be regulating synaptic plasticity in response to Piwi knockdown. Only one of the assayed proteins was responsive: CREB2, a transcriptional repressor, known to be a major inhibitory constraint on LTF, was upregulated in response to Piwi knockdown. Interestingly, an even greater increase in CREB2 mRNA was observed.

The fact that Piwi knockdown led to an increase in CREB2 mRNA, and it’s nuclear localisation, suggested that rather than acting post-transcriptionally (ie by degrading mRNAs as in Drosophila), Piwi/piRNA complexes appeared to be inhibiting CREB2 gene expression at the DNA level.  It is known that in mice Piwi/piRNA complexes act to silence transcription by facilitating DNA methylation. Rajasethupathy et al therefore asked whether CREB2 regulation by Piwi occurred via DNA methylation.

The enzyme responsible for methylation of cytosine residues in CG dinucleotides, DNA methyltransferase (DNMT), was known to be expressed in the Aplysia CNS. Inhibition of DNMT activity (using a pharmacological reagent) led to a strong increase in the level of CREB2. In the normal LTF experiments, CREB2 levels are reduced 12 hours after exposure to 5HT and remain low until 48hrs. This downregulation of CREB2 was abolished when DNMT activity was inhibited, as was 5HT-dependent LTF. This led the researchers to search for CpG islands in the promoter region of CREB2 that could be sites for DNA methylation mediated transcriptional control. Indeed, they identified a CpG island in the CREB2 promoter region that normally exists in both methylated and unmethylated states. After 5HT exposure, this CpG island is almost entirely methylated, whilst in the presence of the DNMT inhibitor it becomes almost entirely unmethylated. This 5HT-dependent methylation of the CREB2 CpG island requires Piwi, as it was abolished when Piwi was inhibited. The authors then went on to search for candidate piRNAs that could be responsible for mediating this effect, by searching for those with complementarity to the CREB2 promoter. They identified 4, one of which, aca-piR-F, when knocked down caused an increse in CREB2 expression. Notably, Rajasethupathy et al. did not demonstrate the expected result that aca-piR-F knockdown would lead to demethylation of the CREB2 CpG island, although this experiment was surely attempted.

In conclusion, this paper offers a broad outline for a mechanism of memory encoding; it connects neurotransmitter synaptic stimulation with the stable epigenetic marking of the transcription state of an important regulator of neuronal plasticity, via the action of Piwi/piRNA complexes. It should be noted that ‘epigenetic’ is used in this context in a loose definition with reference only to a stable marking of cellular state. Strictly ‘epigenetic’ should refer to heritable non-genetic changes, but as neurons do not divide that is inapplicable. In this case DNA methylation is a relatively long-lasting mark. However, for instance the change in CREB2 expression with respect to 5HT-stimulated long term facilitation only lasts a couple of days – does this correspond to the level of methylation in the promoter? In which case, one gets an impression that DNA methylation and demethylation are highly dynamic processes in the Aplysia CNS. Currently there are three modes proposed to resolve the difficult question of how memories can persist for a long time, whilst the cellular components that must mediate them have a high rate of turnover: Prion-like synaptic marks, autoregulatory loops that can maintain a cell state whilst their components come and go, and epigenetic mechanisms that can alter gene expression in a long term manner. This paper shows a clear example of the latter mode, but the apparent dynamism of DNA methylation in this system suggests a lack of permanence.

Although I like the way this paper has ranged over a large terrain and connected so many disparate elements, by necessity it raises many questions and leaves many aspects of the work unmentioned. I’ve already mentioned some questions about Aplysia piRNAs; no doubt a fully annotated Aplysia genome will answer some of them. A few other questions and future directions spring to mind: The authors haven’t quite shown that DNA methylation is responsible for the transcriptional silencing at the CREB2 promoter, only correlated it. Likewise the mode by which Piwi/piRNA complexes act to promote DNA methylation is unclear. A wider question is the nature of DNA methylation in Aplysia and other invertebrates. Some invertebrates show virtually no DNA methylation (C. elegans, Drosophila) whilst the majority display mosaic patterns quite different from those found in vertebrates. This suggests functional differences, and without deeper knowledge of the role of DNA methylation in Aplysia it is difficult to guess how widely applicable these findings are in other systems. Likewise the finding that piRNAs are acting at the level of DNA methylation, previously only found in mammals, raises questions about the state of affairs in other invertebrate model systems. Also, do Aplysia piRNAs only act on DNA methylation, or post-transcriptionally aswell.? Future studies will no doubt also look at how this type of regulation corresponds to histone marks, and try to synthesise the different levels of regulation. Perhaps the most important take home message is that piRNAs are more than a germline specific immunity against tranposons. Just how widespread these other roles are is an open question.

Rajasethupathy, P., Antonov, I., Sheridan, R., Frey, S., Sander, C., Tuschl, T., & Kandel, E. (2012). A Role for Neuronal piRNAs in the Epigenetic Control of Memory-Related Synaptic Plasticity Cell, 149 (3), 693-707 DOI: 10.1016/j.cell.2012.02.057


small silencing RNAs. I: Piwi-interacting RNAs.

Three major classes of small RNAs involved in gene silencing have been found in animals: microRNAs (miRNAs), small-interfering RNAs (siRNAs) and Piwi-interacting RNAs (piRNAs). miRNAs are involved in the regulation of mRNA stability and translation, whilst the main purpose of the siRNA and piRNA pathways appears to be the defense of the cell and genome from viruses and transposable elements. Unlike the other two systems that are ubiquitously active, the piRNA pathway is generally only active in germline cells, the most important locus of defense against transposons.

A common feature of all three pathways is the formation of RNA-induced silencing complexes (RISCs), composed of a small RNA bound to an Argonaute family protein. The small RNA guides RISC to specific target RNAs, resulting in target silencing (generally by the Argonaute protein ‘slicing’ the cognate RNA). A key stage in the miRNA and siRNA silencing pathways is the recognition of double stranded RNAs, and their cleavage by Dicer proteins. This is not a feature of the piRNA system. Another difference is that piRNAs range from 22nt to 30nt in length, whilst siRNAs and miRNAs are 21 or 22-24nt long respectively. When piRNAs were first discovered they were called repeat-associated small-interfering RNAs (rasiRNAs). However, as they are not always associated with repeat sequences and as they bind a specific clade of Argonaute proteins, the PIWI family, they were subsequently renamed.

The piRNA system in Drosophila

A Drosophila melanogaster egg chamber. The large nurse cell nuclei are visible in the upper half, whilst the follicle cells cover the oocyte in the lower half.

The piRNA transposon silencing system has been most comprehensively analysed during oogenesis in the fruitfly, Drosophila melanogaster. Within a Drosophila egg chamber, the germline cells (fifteen nurse cells and the oocyte) share a common syncytial cytoplasm. They are surrounded by a layer of somatic follicle cells, which exchange developmental signals and nutrients with the germline cells. The Drosophila genome harbours over a hundred transposon families, including representatives of all three major classes (LTR and non-LTR retrotransposons, and DNA elements). Some retrotransposons, such as the gypsy family, form viral particles that have been shown to be able to invade the germline from the follicle cells via cellular transport vesicles. Therefore the germline is under threat from transposable elements primarily from within, but also from the somatic follicle cells. Two different variants of the piRNA system function in the germline and the somatic follicle cells: a more complicated system involving three PIWI family Argonaute proteins and a piRNA amplification system functions in the germline, whilst a simpler system involving only one PIWI protein works in the follicle cells to silence a more limited repertoire of retrotransposons.

The piRNA pathway in somatic follicle cells

Approximately 70% of somatic piRNAs map to transposons or transposon fragments. Of these 90% are antisense to active transposons. Mapping piRNAs to genomic sequence has yielded a great insight into genomic structure and the piRNA system of transposon control: piRNAs are derived from large clusters of densely packed, inactive transposon copies and fragments. piRNA clusters are a conserved feature of piRNA biology. They generally span dozens to hundreds of kilobases and are located in the heterochromatin associated with centromeres or telomeres. In the case of Drosophila somatic follicle cells two piRNA-clusters dominate: The flamenco locus and cluster 20A. Follicle cell piRNAs from these clusters are derived from one DNA strand, meaning that transcription is unidirectional. In flamenco and cluster 20A, the transposon fragments are generally oriented antisense to the direction of transcription, explaining the strong antisense bias of somatic follicle cell piRNAs. A P-element insertion at the beginning of the flamenco cluster blocks piRNA production from the whole 180kb cluster, suggesting that the formation of long single stranded transcripts of piRNA clusters is a necessary stage of piRNA biogenesis. However, the mechanisms of piRNA generation are not clear. It appears likely that the long piRNA precursor transcripts are stochastically cut into smaller fragments. Piwi then selectively binds fragments with a 5′ uridine (75% of Piwi-bound piRNAs have a 5′ uridine residue), and the pre-piRNAs are then 3′ trimmed to generate the final piRNA.

The germline piRNA pathway and ping-pong amplification.

In addition to Piwi, Drosophila ovarian germline cells express two related PIWI family Argonaute proteins: Aubergine (Aub) and AGO3. Unlike Piwi, which is localised to the nucleus, Aub and AGO3 are associated with an electron-dense peri-nuclear region of cytoplasm called nuage. Most importantly, they act together in a sophisticated piRNA amplification loop that is dependent on target expression, termed the ping-pong cycle. In a simplified version: Aub complexed with an antisense piRNA targets and slices a sense transcript of an active transposon, resulting in the production of novel sense piRNA species which are loaded onto AGO3. The AGO3-piRNA complexes then cleave complementary piRNA cluster transcripts, resulting in the production of novel antisense piRNA to be complexed with Aub. The ping-pong cycle results in the amplification of sets of antisense and sense piRNAs that are 10nt out of register with each other, suggesting the site of Aub slicer activity and providing a useful signal that shows that ping-pong amplification has occurred.

In the germline, more piRNA clusters are active, representing a larger spectrum of transposons. They are also expressed bi-directionally. An outstanding question is why this doesn’t trigger ping-pong amplification? The most likely reason is that the processes of piRNA biogenesis and transposon silencing are tightly localised and regulated. The roles of other proteins in these processes are starting to be understood. Proteins containing Tudor domains appear to be very important in the localisation and function of Aub and AGO3 in the nuage.

Many other intriguing aspects of piRNA biology are yet to be understood. Although the bulk of piRNAs are directed against transposons, some are involved in the regulation of cellular mRNAs. These piRNAs are derived from mRNAs rather than cluster transcripts: Are these transcripts marked in some way to be processed into piRNAs? The links between the primary piRNA biogenesis pathway and the ping-pong amplification system are also poorly understood. An interesting aspect of the piRNA system active in mouse spermatogenesis, is that the nucleus localised mouse PIWI family protein MIWI2 has been implicated in guiding de novo DNA methylation at transposon loci. Is this a more widespread phenomenon?

The piRNA system has been likened to an acquired immune response and works together with the (more acute response) siRNA pathway in transposon silencing. Future posts will discuss the other small RNA systems, and go further into piRNA biology.

Senti, K., & Brennecke, J. (2010). The piRNA pathway: a fly’s perspective on the guardian of the genome Trends in Genetics, 26 (12), 499-509 DOI: 10.1016/j.tig.2010.08.007

Khurana, J., & Theurkauf, W. (2010). piRNAs, transposon silencing, and Drosophila germline development The Journal of Cell Biology, 191 (5), 905-913 DOI: 10.1083/jcb.201006034

of further interest: piRNAs in the brain: epigenetics and memory