Based on piRNA sequence signatures and biogenesis factor requirements, two distinct cytoplasmic piRNA biogenesis pathways have been recognized: primary and secondary biogenesis. In the Drosophila ovarian environment, primary biogenesis is operational in both the somatic follicle cells and the germline, while secondary biogenesis - known as the `ping-pong' cycle - is specific to the germline.
Primary piRNA biogenesis
A combination of transcriptome sequencing, mapping of transcription factor binding sites, and genetic analyses have revealed that piRNA clusters are transcribed into long single-stranded transcripts [4]. Such precursors are then exported to cytoplasmic granules called nuage (also called Yb bodies in the fruitfly ovarian soma) where most piRNA biogenesis factors reside. Primary processing describes the path taken by such precursors to end up as approximately 24 nucleotide primary piRNAs [3]. In the ovarian soma, a single type of Piwi protein, PIWI, receives primary piRNAs. Although we do not have a clear picture, it is believed that the precursor RNA is randomly broken up by unknown nucleases into fragments, which are then loaded onto a Piwi protein, such that the 5′ end of the fragment (pre-piRNA) is inserted into its MID domain (Box 1). Primary piRNAs display a strong preference for a 5′ uridine (1U-bias), but primary processing generates piRNAs without any such preference. So it is likely that 1U-bias is an outcome of the nucleotide preference of the MID domain that allows the Piwi protein to bind and enrich certain piRNAs. Subsequently, another unknown 3′-5′ exonuclease (tentatively called Trimmer) cleaves the 3′ end to generate the mature piRNA. The 3′ end is then modified by a 2′-O-methyl modification catalyzed by the RNA methyltransferase Hen1.
To date, Zucchini is the only protein known to act as a nuclease during primary processing. Zucchini homodimerizes, presumably when anchored on the outer membrane of mitochondria, to form an endonuclease active on single-stranded RNAs [5], but its exact substrate or product in the piRNA pathway is not known (Figure 1b). Other primary piRNA biogenesis factors include RNA helicases Armitage (Armi) and Yb, the Hsp90 co-chaperone Shutdown, and an RNA-binding protein Vreteno. Molecular details of how these factors function are yet to be determined.
One major question to consider is how the primary processing machinery distinguishes Pol II-transcribed piRNA cluster transcripts from other capped and polyadenylated noncoding RNAs. One possibility is that the nuclear history of transcription from the cluster promoter is conveyed by a co-transcriptionally loaded RNA-binding protein. However, the transcription factor Cubitus interruptus (Ci) that transcribes the flamenco cluster in the ovarian soma may not be enough to signal this, as Ci transcribes other protein-coding genes that do not enter the piRNA pathway. Alternatively, the information might be contained in the precursor RNA itself in the form of a specific sequence or an RNA structural element. The latter possibility is attractive given that the piRNA processing machinery is largely conserved across the animal kingdom, so it might identify conserved features in piRNA precursors.
It is interesting to note that before reaching the nuage, cluster transcripts like that of flamenco may transit through another perinuclear structure called Dot-COM/Flam-body that is in the proximity of Yb bodies [6]. Once PIWI is loaded, it becomes licensed for nuclear import. The bound piRNAs are then proposed to identify target genomic loci by annealing to nascent transcripts, eventually leading to deposition of silent H3K9me3 marks on the target chromatin. To enable this transcriptional gene silencing (TGS), PIWI collaborates with a zinc-finger protein Gtsf-1 and a HMG box-containing protein Maelstrom.
Secondary piRNA biogenesis: the `ping-pong' cycle
Germline piRNA biogenesis is more complex than in somatic cells as all three Piwi proteins (Aubergine, Ago3 and PIWI) are involved, as opposed to PIWI alone. Most factors implicated in the follicle cells for primary processing also act in the germline, loading primary 1U-containing piRNAs into Aubergine (Aub) and PIWI. This results in the loading of essentially antisense repeat piRNAs into the two proteins [3]. Thus, the primary processing machinery is somehow able to distinguish the different Piwi proteins present in the germline and deliver piRNAs only into specific proteins. Once loaded, PIWI goes into the nucleus, while Aub remains in the cytoplasm to recognize and slice complementary cluster transcripts or sense transposon targets. The latter step leads to post-transcriptional gene silencing (PTGS) of transposable elements in the cytoplasm. In a process unique to the germline, a specialized machinery of proteins then converts one of the cleavage fragments into a sense-oriented secondary piRNA that is loaded into Ago3 [3]. Such sense Ago3 piRNAs can then target antisense cluster transcripts to slice and generate the exactly same antisense Aub-bound piRNAs that initiated the process. This feed-forward piRNA amplification loop - the ping-pong cycle - functions as an adaptive arm of the transposon defense system, as every time a target is destroyed more (silencing) antisense piRNAs are generated.
Secondary piRNA biogenesis links post-transcriptional gene silencing of transposon transcripts to biogenesis of a new piRNA, and this requires a number of factors unique to the germline environment (Figure 1c). Recent studies in Bombyx mori (Silkworm) BmN4 cell culture have shed light on the molecular role of the conserved DEAD box RNA helicase Vasa in linking the two events [7]. Vasa is shown to function as an ATP-regulated RNA clamp that anchors a piRNA Amplifier complex on transposon transcripts. Such a complex is composed of Vasa, the two ping-pong Piwi partners, the Tudor domain protein Qin/Kumo, antisense piRNAs and complementary sense transposon transcripts. The purpose of the Amplifier complex is to bring the two ping-pong partners into close proximity such that an Aub-generated transposon cleavage fragment can be transferred to Ago3. Such an exchange is triggered by hydrolysis of ATP within Vasa. The multiple Tudor domain protein Qin/Kumo is known to promote Aub-Ago3 interactions to ensure heterotypic ping-pong between the two proteins [8]. The specificity of the process is evident from the fact that Ago3 fails to become loaded with piRNAs in cell culture environments where only primary processing is active. In fact, Krimper is shown to sequester unloaded Ago3 in specialized Krimp bodies, presumably to prevent illegitimate entry of RNA sequences into Ago3. A number of other factors are shown to be essential for the ping-pong cycle, but their molecular functions are not clearly understood (Figure 1c).