piRNA-guided slicing specifies transcripts for Zucchini-dependent, phased piRNA biogenesis

See allHide authors and affiliations

Science  15 May 2015:
Vol. 348, Issue 6236, pp. 812-817
DOI: 10.1126/science.aaa1039

Spreading small RNAs to protect the genome

In animals, PIWI-interacting RNAs (piRNAs) are small noncoding RNAs that protect our germ lines from the ravages of transposons. To do this, piRNAs target and cleave transposon RNAs. Synthesis of piRNA is initiated by a cut made in a long, single-stranded precursor RNA. The piRNAs can also undergo a self-perpetuating amplification cycle (see the Perspective by Siomi and Siomi). Han et al. and Mohn et al. now reveal that piRNA biogenesis can also spread in a strictly phased manner from the site of initial piRNA formation. Spreading piRNA synthesis greatly increases their sequence diversity, potentially helping them to target endogenous and novel transposons more effectively.

Science, this issue p. 817, p. 812; see also p. 756


In animal gonads, PIWI-clade Argonaute proteins repress transposons sequence-specifically via bound Piwi-interacting RNAs (piRNAs). These are processed from single-stranded precursor RNAs by largely unknown mechanisms. Here we show that primary piRNA biogenesis is a 3′-directed and phased process that, in the Drosophila germ line, is initiated by secondary piRNA-guided transcript cleavage. Phasing results from consecutive endonucleolytic cleavages catalyzed by Zucchini, implying coupled formation of 3′ and 5′ ends of flanking piRNAs. Unexpectedly, Zucchini also participates in 3′ end formation of secondary piRNAs. Its function can, however, be bypassed by downstream piRNA-guided precursor cleavages coupled to exonucleolytic trimming. Our data uncover an evolutionarily conserved piRNA biogenesis mechanism in which Zucchini plays a central role in defining piRNA 5′ and 3′ ends.

The Piwi-interacting RNA (piRNA) pathway silences transposable elements (TEs) during animal gametogenesis. This pathway uses PIWI-clade Argonaute proteins bound to piRNAs that are ~23 to 30 nucleotides (nt) long; these piRNAs act as sequence-specific guides to specify targets via base-pair complementarity (1, 2). piRNAs are processed from single-stranded precursor transcripts via two biogenesis pathways, both of which are initiated by endonucleolytic definition of piRNA 5′ ends: 5′ ends of primary piRNAs (mostly loaded into Drosophila Piwi) are generated by Zucchini (38). They carry a 5′ uridine (1U) but otherwise appear to be derived randomly from their precursors (913). 5′ ends of secondary piRNAs [loaded into Drosophila Aubergine/Argonaute3 (Aub/AGO3)] are specified via piRNA-guided slicing. Here, reciprocal cleavages of complementary transcripts (ping-pong cycle) define piRNA pairs, whose 5′ ends display a 10-nt offset (14, 15). piRNA 3′ ends lack a nucleotide preference, and the molecular events underlying their formation are elusive. It is thought that after 5′ end definition, longer piRNA intermediates are loaded into PIWI proteins (16) and subsequently trimmed at their 3′ ends to mature piRNAs (17).

How are transcripts selected as piRNA biogenesis substrates? We noticed ectopic piRNA production from mRNAs in fly ovaries mutant for the Rhino-Deadlock-Cutoff complex (1820). The corresponding piRNA profiles initiate abruptly (e.g., row) (Fig. 1, A and B), and some of these initiation sites are also apparent in wild-type (WT) ovaries [green fluorescent protein (GFP) control]. Consistent with suggestions that piRNA-guided transcript cleavage initiates 3′-directed piRNA biogenesis (21), we identified a target site in the row transcript with complementarity to a 1360 transposon piRNA (trigger-piRNA) (Fig. 1C). The 5′ end of this piRNA maps 10 nt away from the 5′ end of the most abundant row piRNA (responder-piRNA). Whereas the 1360 trigger-piRNA occupies Aub and AGO3 equally, the row responder-piRNA resides mostly in Aub (Fig. 1, B and C) (22). In contrast, row piRNAs originating downstream of the responder-piRNA (trail-piRNAs) occupy Piwi and, only to a small extent, Aub (Fig. 1B). We identified 570 mRNAs with similar characteristics. On average, responder-piRNAs predominantly populate Aub but are also present in AGO3 and Piwi (Fig. 1D). Trail-piRNAs instead are funneled into Piwi (~85%) and moderately into Aub (~15%) but not into AGO3. When considering all piRNAs mapping within 200 nt downstream of the trigger site, 70% of Aub/AGO3-bound piRNAs, but only 7% of Piwi-bound piRNAs, correspond to the responder-piRNA (Fig. 1D).

Fig. 1 Aub/AGO3-mediated slicing triggers phased piRNA biogenesis.

(A) Normalized piRNA populations from Rhino-depleted ovaries or from control ovaries mapping to the row mRNA. ppm, parts per million. (B) As in (A), but piRNAs bound to Aub/AGO3/Piwi (22) are shown individually. (C) Alignment of the 1360 trigger-piRNA with the row mRNA. The inverted triangle denotes the slicer cleavage position. (D) Metaplots showing profiles of Aub/AGO3/Piwi-bound piRNAs (from Rhino-depleted ovaries) at genic trigger sites. Profiles represent the median of normalized values; responder peak: 100%. (E) Heat map indicating piRNA levels (Σ Aub/AGO3 5′ ends) mapping antisense (≤six mismatches) to 570 mRNAs with trigger events. Position 0 denotes the responder-piRNA 5′ end. The binary histograms show the percentage of transcripts with a cloned 5′ end of indicated piRNAs mapping in sense/antisense orientation at nucleotide resolution. (F) Heat maps indicating AGO3/Aub/Piwi-bound piRNA 5′ end levels from Rhino-depleted ovaries in a window centered on 570 genic responder-piRNAs. The corresponding binary histograms indicate the percentage of transcripts that exhibit a cloned piRNA 5′ end at a given position.

Only 8% of the 570 triggered transcripts harbor fully complementary target sites for Aub/AGO3-bound piRNAs at the responder position (fig. S1A). However, when allowing up to six mismatches, piRNA 5′ ends exhibit a high likelihood to align precisely 10 nt offset from responder-piRNA 5′ ends (Fig. 1E). We could identify trigger-piRNAs in ~75% of the cases (fig. S1, B to G) (22). Based on this, slicing characteristics of Aub/AGO3 follow known features of Argonaute proteins (2325). Trigger-piRNAs also conform to rules of heterotypic ping-pong (14, 15) (fig. S1E).

The 5′ ends of trail-piRNAs are immediately downstream of responder-piRNA 3′ ends (Fig. 1D), and trail-piRNAs in Piwi and Aub display pronounced phasing (Fig. 1F). The phase is ~27 nt, which is precisely one average piRNA length (23 to 30 nt) (Fig. 1F). Both phasing accuracy and piRNA levels decrease with increasing distance from the trigger site, which suggests that piRNA biogenesis occurs in a 3′-directed, processive fashion. To investigate whether slicing generally initiates directional and phased piRNA biogenesis, we analyzed TE-mapping piRNAs from WT ovaries. Piwi/Aub/AGO3-bound piRNAs mapping in the vicinity of abundant ping-pong piRNAs display patterns that are virtually identical to the ones described above (fig. S2, A and B). Biogenesis of Piwi-bound piRNAs (by definition, primary piRNAs) is therefore largely a consequence of piRNA-guided target cleavage.

To investigate how tight piRNA 3′ end formation is coupled to 5′ end formation of the immediate downstream piRNA, we analyzed piRNA length and 3′ end patterns. On average, piRNAs exhibit a broad length distribution (Fig. 2, A to C) (14). We noticed that individual piRNAs sharing the same 5′ end often have distinct length profiles and 3′ ends (fig. S3A). We therefore grouped all piRNAs with unambiguous 5′ ends into length cohorts (Fig. 2, A to C, and fig. S3B) (22). Length profiles of AGO3-bound piRNAs are the least defined, whereas those bound to Piwi display the most accurate 3′ ends, irrespective of whether they are processed in the germ line or the soma (fig. S3, C to E). No obvious nucleotide bias within piRNAs correlates with their distinct length groups. However, uridine (U) residues are enriched immediately downstream of dominant piRNA 3′ ends (Fig. 2, D and E, and fig. S3, F to J). Length cohorts with two dominant species display a downstream U bias for both prominent 3′ ends (e.g., cluster 7 in fig. S3H and cluster 6 in fig. S3J). The downstream U bias is strongest for Piwi-bound piRNAs and weakest for AGO3-bound piRNAs, where it is only evident for the most accurately defined length populations (Fig. 2E and fig. S3G). As Piwi/Aub-bound piRNAs display a 1U bias and primary piRNAs exhibit strong phasing, the pronounced downstream U bias implies that piRNA 3′ end formation simultaneously specifies the 5′ end of the downstream piRNA. When Piwi-bound piRNAs are aligned at their dominant 3′ ends, neighboring piRNA 5′ ends are highly enriched at the +1 position, irrespective of the tissue (fig. S4A).

Fig. 2 Biogenesis of flanking primary piRNAs is coupled.

(A to C) (Top) piRNA length histograms. IP, immunoprecipitation. (Bottom) Heat maps showing length groups for piRNAs in Piwi/Aub/AGO3. 5′ ends with similar length profiles were grouped using K-means clustering. n denotes the number of analyzed piRNAs. (D) Sequence logos indicating nucleotide biases for piRNA length clusters defined in (A). The gray shaded area denotes the piRNA body. For Aub/AGO3, see fig. S3. (E) Sequence logos indicating nucleotide biases for Piwi/Aub/AGO3-bound piRNAs at 5′ ends, position 10, and 3′ ends (only 3′ ends where ≥50% of piRNAs terminate were considered). (F) Heat maps displaying somatic Piwi-bound piRNA 5′ or 3′ ends in a window around the most abundant TE piRNAs. The sum of each line was scaled to 100% and sorted according to piRNA level at position 0. (G) Heat maps displaying 5′ or 3′ end counts of piRNAs mapping to positions 20 to 32 downstream of major piRNA 5′ ends [detail from (F)]. The left plot is sorted as in (F); the right plot is re-sorted for dominant piRNA length species (see also fig. S4B). (H) Sequence logos indicating uridine frequency along piRNA sequences of different length cohorts [defined in (G); black line indicates dominant piRNA 3′ ends].

To further investigate 3′ end formation, we focused on somatic Piwi-bound piRNAs. We selected the 5000 most abundant TE antisense piRNA 5′ ends, sorted them for their abundance, and displayed 5′ and 3′ ends of all piRNAs mapping in their immediate vicinity (Fig. 2F). The 3′ ends of neighboring piRNAs are enriched precisely 1 nt upstream of selected 5′ ends (position 0, dashed arrow in Fig. 2F). The 5′ ends of downstream piRNAs instead display a fuzzy enrichment around +27 nt (blue arrow head in Fig. 2F). However, re-sorting the heat map according to increasing length of the dominant piRNA species for each 5′ end at position 0 resolves the strong enrichment of 5′ ends precisely 1 nt downstream of the dominant 3′ ends (Fig. 2G and fig. S4B).

The obtained piRNA length cohorts (23 to 29 nt) (Fig. 2G and fig. S4B) display a strong downstream U bias (Fig. 2H). U residues are depleted upstream of 3′ ends of unusually long piRNA cohorts (27 to 29 nt), and an inverse trend is seen for the 23-nt piRNA cohort. This suggests that 3′ ends of Piwi-bound piRNAs are defined by an endonuclease that cleaves immediately upstream of a U residue and 23 to 30 nt downstream of the piRNA 5′ end. Asymmetrically distributed U residues within the cleavage window force the generation of atypically long or short piRNAs.

To experimentally test the phased biogenesis of primary piRNAs, we generated transgenic flies expressing a piRNA biogenesis reporter harboring a piRNA target site (Fig. 3A). Without a target site, no piRNAs are produced (Fig. 3B). Insertion of a single complementary target site for an Aub/AGO3 piRNA represses GFP expression (fig. S5), triggers the generation of a ping-pong responder-piRNA, and forces 3′-directed, phased biogenesis of trail-piRNAs (Fig. 3C). Trail-piRNA levels and phasing accuracy decrease with increasing distance from the trigger site but are detectable for several hundred nucleotides. To probe the effect of U residues on biogenesis patterns, we constructed a sensor with four Us in regular 26-nt intervals downstream of the trigger site. This increases levels and phasing accuracy of reporter-derived piRNAs (Fig. 3D). Spacing of single Us in decreasing intervals of 28 to 23 nt dictates piRNA 5′ ends but also defines piRNA 3′ ends precisely 1 nt upstream (Fig. 3E). These findings reinforce the notion that 3′ and 5′ ends of adjacent Piwi-bound piRNAs are formed via a single endonucleolytic cleavage upstream of a U residue.

Fig. 3 Primary piRNA biogenesis is continuous and guided by uridine residues.

(A) Cartoon of the piRNA biogenesis reporter with target site for an Aub/AGO3-bound piRNA. (B and C) Normalized small RNA 5′ end profiles from a reporter lacking a piRNA target site (B) or from a reporter with target site for an Aub/AGO3-bound piRNA. bp, base pairs. (C) Black denotes antisense reads, red indicates sense reads, and blue bars represent U residues. Numbers indicate normalized piRNA counts. Histograms indicate length profile and 1U bias of trail-piRNAs. (D) Similar to (C), but the reporter contains four U residues every 26 nt. (E) Similar to (C), but the reporter contains single U residues in the indicated intervals.

The endonuclease Zucchini generates primary piRNA 5′ ends (38) and is therefore also the main candidate for their 3′ end formation. A direct test of this is precluded, as primary piRNAs depend on Zucchini. Thus, we asked whether Zucchini is involved in 3′ end formation of secondary piRNAs whose 5′ ends are formed via slicing. Many Aub/AGO3-bound secondary piRNAs are generated in Zucchini-depleted ovaries (26), and we used their 5′ ends as anchor points (fig. S6A). In WT ovaries, piRNAs mapping to these 5′ ends display the characteristic nucleotide bias at the first position (1U for Aub piRNAs) and the 10th position (10A for AGO3 piRNAs) (Fig. 4A). They also exhibit a downstream U bias. In Zucchini-depleted ovaries, the lengths of these piRNAs and therefore their 3′ ends are altered and the downstream U bias is lost, implicating Zucchini in 3′ end formation (Fig. 4A and fig. S6, B and C). Thus, the downstream U bias is most likely a fingerprint of Zucchini, making it the central nuclease in piRNA biogenesis. The less severe effect on AGO3-bound piRNAs and their weaker downstream U bias suggest that many AGO3-bound piRNAs are Zucchini-independent or are resected at their 3′ ends in WT ovaries.

Fig. 4 Zucchini is involved in 3′ end formation of Drosophila and mouse piRNAs.

(A) Sequence logos for nucleotides at 5′ ends, position 10, and 3′ ends of Aub- or AGO3-bound piRNAs in control or Zucchini-depleted ovaries. (B) Normalized WT small RNA 5′ end profile for a reporter (similar to Fig. 3D) with a single target site for a Zucchini-independent piRNA. (C) As in (B) but for piRNAs isolated from Zucchini-depleted ovaries. (D and E) As in (C), but the reporter contains target sites for two (D) or three (E) Zucchini-independent piRNAs. (F) Heat maps showing length cohorts of Mili-bound piRNAs from Tdrkh heterozygous mouse testes. (G) Sequence logos showing nucleotide composition around dominant piRNA 3′ ends for length cohorts defined in (G). For individual length clusters, see fig. S6. (H) Frequency of Mili-bound piRNA 5′ ends around aligned dominant piRNA 3′ ends from Tdrkh heterozygous testes. (I to K) As in (F) to (H) but from Tdrkh mutant testes.

To experimentally test Zucchini’s involvement in 3′ end formation of secondary piRNAs, we constructed a biogenesis reporter with a single target site for a Zucchini-independent piRNA. In WT ovaries, this reporter gives rise to a prominent ping-pong responder and to trail-piRNAs (Fig. 4B). As expected, loss of Zucchini ablates trail-piRNAs (Fig. 4C and fig. S2C). But although trigger-piRNA levels and reporter silencing were unaffected, levels of the responder-piRNA also dropped by a factor of >30 (Fig. 4C and fig. S6D). Zucchini is therefore centrally involved in 3′ end formation of ping-pong piRNAs. The piRNA target site in this reporter is from the F-element, one of the TEs that maintain high levels of secondary piRNAs in Zucchini-deficient ovaries (fig. S6E). How, then, are F-element piRNAs generated independently of Zucchini? The central difference between the F-element and the reporter is that the latter contains only a single Aub/AGO3 target site. We added a downstream target site for a second Zucchini-independent piRNA, which would generate a ~50-nt-long piRNA intermediate. This leads to recovery of the first but not the second responder-piRNA in Zucchini-depleted ovaries (Fig. 4D and fig. S6F). Introducing a target site for a third Zucchini-independent piRNA leads to recovery of the first two responder-piRNAs (Fig. 4E and fig. S6G). A downstream slicer cleavage can therefore bypass Zucchini’s role in piRNA 3′ end formation. This must be coupled to exonucleolytic trimming, consistent with the existence of an activity that resects intermediates to mature piRNAs (17).

To uncover an involvement of mouse Zucchini (MitoPLD) (8, 27) in piRNA 3′ end formation, we searched for a downstream U bias in published piRNA data sets. Although mouse piRNA populations group into well-defined length cohorts, they lack a downstream U bias (Fig. 4, F and G, and fig. S7A) (28). Similarly, piRNA 3′ ends and adjacent 5′ ends display no coupling signature (Fig. 4H), suggesting fundamental differences in piRNA 3′ end formation between flies and mice. However, Mili-bound primary piRNAs are extensively trimmed at their 3′ ends (28). Although mature Mili-bound piRNAs are ~24- to 28-nt-long, Mili associates with ~30- to 40-nt piRNA intermediates harboring mature 5′ ends in Tdrkh/Papi mutants (fig. S7, B and C). This suggests that Tdrkh recruits an exonuclease to PIWI proteins to facilitate piRNA precursor trimming (28, 29). Trimming might therefore occlude the downstream U bias defined by MitoPLD. Indeed, Mili-bound piRNA intermediates from Tdrkh mutants display a strong downstream U bias and exhibit strong coupling of 3′ ends with subsequent 5′ ends (Fig. 4, I to K, and fig. S7D). Hence, coupled biogenesis of neighboring piRNAs is conserved between Drosophila and mice.

Altogether, a model for piRNA biogenesis can be drawn (fig. S8). A critical first step is the specification of piRNA precursors, leading to the endonucleolytic definition of a piRNA 5′ end. In the fly germ line, the dominating process is piRNA-guided target slicing, which specifies the 5′ end of a responder-piRNA. 3′ end formation of this piRNA can occur via a second slicer cleavage event, which liberates a piRNA intermediate for exonucleolytic trimming. Alternatively, 3′ end formation is catalyzed by Zucchini, which cleaves the precursor upstream of a U residue. Zucchini-mediated 3′ end formation promotes phased and 3′-directed primary piRNA biogenesis. In flies, Zucchini cleavage products seem directly compatible with Piwi binding. In mice, MitoPLD cleavage products are too long, making 3′ end trimming essential. Notably, how primary piRNA biogenesis initiates in Drosophila ovarian somatic cells or adult mouse testes remains elusive.

Supplementary Materials

Materials and Methods

Figs. S1 to S9

Tables S1 and S2

References (3039)

References and Notes

  1. See materials and methods and other supplementary materials on Science Online.
  2. Acknowledgments: We thank J. Gokcezade for fly injections, Harvard TRiP and Bloomington stock centers for flies, D. Jurczak for bioinformatics support, the Campus Science Support Facilities Next Generation Sequencing unit for Illumina sequencing, and P. Zamore for sharing unpublished data. This work was supported by the Austrian Academy of Sciences, the European Community (European Research Council grant 260711EU), the Austrian Science Fund (grant Y 510-B12), and Swiss National Science Foundation and Human Frontier Science Program fellowships to F.M. Sequencing data sets are deposited at Gene Expression Omnibus (accession numbers GSE64802 and GSE55842).
View Abstract

Stay Connected to Science

Navigate This Article