Piwi-interacting RNAs (piRNAs) are a class of small non-coding RNAs expressed in the animal gonads. They are implicated in silencing the genome instability threat posed by mobile genetic elements called transposons. Unlike other small RNAs, which use double-stranded precursors, piRNAs seem to arise from long single-stranded precursor transcripts expressed from discrete genomic regions. In mice, the Piwi pathway is essential for male fertility, and its loss-of-function mutations affect several distinct stages of spermatogenesis. While this small RNA pathway primarily operates post-transcriptionally, it also impacts DNA methylation of target retrotransposon loci, representing an intriguing model of RNA-directed epigenetic control in mammals. Remarkably the Piwi pathway components are specifically localized at germinal granule/nuage, an evolutionarily conserved but still enigmatic ribonucleoprotein compartment in the germline. The inaccessibility of the germline for easy experimental manipulation has meant that this class of RNAs has remained enigmatic. However, recent advances in the use of cell culture models and cell-free systems have greatly advanced our understanding. In this review, we briefly summarize our current understanding of the Piwi pathway, focusing on its developmental regulation, piRNA biogenesis and key function in male germline development from fetal spermatogonial stem cell stage to postnatal haploid spermiogenesis in mice.
The germline is the cell lineage that transmits the genetic information to the next generation. The genetic and epigenetic changes in the germline affect embryonic development and subsequent offspring, so the genome stability of germ cells is a critical requirement to maintain both individuals and species. Genome DNA injury generally occurs as a consequence of physical or chemical insults, such as by exposure to ionizing radiation, genotoxic reagents and oxidative stress, etc. Another threat to the genome is encoded by the genome itself, namely mobile transposable elements, which move or amplify themselves and transpose into new genome positions. In mammals, transposable elements and their fossil sequences occupy about half the genome (approximately 45% in humans), as compared to 1–2% of protein coding exonic sequences (Lander et al. 2001; Waterston et al. 2002). Although most of these transposable elements have been silenced by mutations and truncations etc., some elements remain active and cause novel insertional mutations. For instance, LINE1, the most abundant class of retrotransposons in mammals, has approximately 500 000 copies in the haploid genome and about 100 and 3000 of them are estimated to be functional in humans and mice, respectively (Kazazian 2004). Such active transposable elements are typically expressed in the germline, wherein they generate newly transposed copies and pass them onto the next generation.
In order to control such mutagenic transposable elements, host genomes have evolved multiple layers of molecular defense systems. In mammals, one key pathway is DNA methylation. Mutations in a maintenance DNA methyl transferase, DNMT1, lead to transposon desilencing and embryonic lethality (Li et al. 1992; Walsh et al. 1998), whereas DNMT3L plays a more specific role in suppressing retrotransposons in germ cells (Bourc’his & Bestor 2004). Histone modifications also play essential roles in epigenetic silencing of transposable elements (Combes & Whitelaw 2010; Mochizuki & Matsui 2010). Another more adaptive defense mechanism that preferentially targets retrotransposons operates via RNA interference (RNAi) or related systems, which use small non-coding RNA that guides the Argonaute or associated complexes to cut or translationally suppress target mRNA. Germ cells, in particular, are equipped with the Piwi-interacting RNA (piRNA) pathway (Aravin et al. 2006; Girard et al. 2006; Grivna et al. 2006a; Watanabe et al. 2006), wherein the Piwi proteins of the Argonaute family are central effectors for retrotransposon silencing. This small RNA pathway is evolutionarily conserved in many animals and plays a critical role(s) in suppressing retrotransposons during the germline development (Grimson et al. 2008). In mammals, piRNAs are most abundantly expressed in male germ cells, especially during postnatal spermatogenesis, although piRNAs are also detectable to a lesser extent in male fetal germ cells (gonocytes/prospermatogonia) as well as in developing oocytes (Aravin et al. 2008; Kuramochi-Miyagawa et al. 2008; Tam et al. 2008; Watanabe et al. 2008). Notably, all loss-of-function mutations of the piRNA pathway genes reported so far in mice lead to male sterility, while females are fertile, indicating the importance of this pathway in male germline development in mammals. The Piwi pathway acts via both post-transcriptional and transcriptional mechanisms, i.e. at the RNA and epigenetic levels to silence retrotransposons in mice (Aravin et al. 2007a,b; Kuramochi-Miyagawa et al. 2008). This provides an intriguing model of RNA-directed epigenetic control in mammals, in which, unlike plants and yeasts, little is known about molecular function of endogenous small RNA in epigenetic regulation.
In addition to Piwi proteins, recent studies have identified a class of functional components of the piRNA pathway, such as Tudor and helicase family members (Chuma et al. 2003, 2006; Kojima et al. 2009; Reuter et al. 2009; Shoji et al. 2009; Vagin et al. 2009; Wang et al. 2009a; Kuramochi-Miyagawa et al. 2010; Yabuta et al. 2011), and the number is still expanding. These Piwi pathway components form effector RNP assemblies that contain piRNAs. In addition, they are also specifically localized at cytoplasmic compartments in the germline, termed germinal granules/nuage (Eddy 1975). Germinal granules/nuage have been observed for years and implicated in the germline specification in Drosophila and Caenorhabditis elegans elegans but have been an enigmatic structure in mammalian germ cells (Chuma et al. 2009). In this review, we would like to outline the piRNA system focusing on the functional components, their biogenesis, developmental regulation, retrotransposon silencing and subcellular compartmentalization in mammalian spermatogenesis.
Germ cell development in mammals
In mice, germline determination occurs among a population of pluripotent epiblast cells depending on inductive signals from surrounding somatic cells (Mclaren 2003). This specification process gives rise to primordial germ cells (PGCs), which are fate-committed to form future gametes at around the gastrulation stage of embryonic development. PGCs are allocated at an extra-embryonic tissue, allantois, then migrate through the embryo proper to reach at the gonadal primordia at mid-gestation, wherein they start male or female germ cell differentiation. In the male, PGCs enter into G1/G0 arrest and become pro-spermatogonia, while in the female they enter into meiosis and are arrested at meiotic prophase I. Prospermatogonia then resume mitotic proliferation soon after birth to become postnatal spermatogonial stem cells, followed by the first wave of spermatogenic differentiation, which gives rise to meiotic spermatocytes, haploid round spermatids, then mature spermatozoa (Yoshida 2010) (Fig. 1). Through the continuous proliferation of spermatogonial stem cells and subsequent differentiation, functional spermatozoa are produced throughout the male life. In the female, in contrast, the number of oocytes is determined before birth and with puberty, primordial oocytes periodically start to grow and progressively mature into functional eggs (Handel & Eppig 1998).
During such germ cell differentiation, dynamic epigenetic reprogramming takes place, which includes global DNA demethylation in PGCs, followed by de novo establishment of DNA methylation during male and female germ cell differentiation (Mochizuki & Matsui 2010). More specifically, PGCs after their fate determination exhibit a progressive reduction in genome-wide cytosine methylation, then following their entry into developing gonads, parental imprinted genes as well as other coding genes and also transposable elements become highly demethylated (Hajkova et al. 2002). How such DNA demethylation proceeds, i.e. whether the process is active or passive has yet to be determined. Subsequently after the genome-wide erasure of DNA methylation, de novo methylation including paternal and maternal imprinted patterns is re-established in fetal prospermatogonia/gonocytes in the male, and in postnatal growing oocytes in the female (Schaefer et al. 2007). At this de novo DNA methylation stage in fetal prospermatogonia/gonocytes in the male, the piRNA machinery first establishes retrotransposon silencing through the Piwi pathway mediated DNA methylation (see below).
The central core of all small RNA pathways is a member of the Argonaute protein family and its associated small RNA (Carmell et al. 2002). The Argonaute family can be broadly classified into the Ago and Piwi clades. Ago clade members are found in almost all organisms studied, from prokaryotes to eukaryotes, with notable exceptions like the budding yeast Saccharomyces cerevisiae. Ago proteins are ubiquitously expressed and bind approximately 21 nt RNAs like microRNAs (miRNAs) and small interfering RNAs (siRNAs). The Piwi clade members on the other hand are found exclusively in animal gonads and associate with approximately 24–30 nt piRNAs (Ghildiyal & Zamore 2009). The mouse genome encodes three Piwi proteins: MIWI (PIWIL1), MILI (PIWIL2) and MIWI2 (PIWIL4). Mili is detected in both sexes, while the other two are expressed only in the male germline (Kuramochi-Miyagawa et al. 2001). Mili shows a broad expression profile from embryonic male germ cells at 12.5 dpc (days post coitum) to post-natal haploid round spermatids (Kuramochi-Miyagawa et al. 2004). The other two have more restricted expression domains, with Miwi2 present from 14.5 dpc to 3 days after birth (P3) (Aravin et al. 2008), while Miwi expression commences only during meiosis in pachytene spermatocytes (at P14) and peaks in the round spermatids (Deng & Lin 2002).
Argonaute proteins are characterized by their signature PAZ, MID and PIWI domains (Parker & Barford 2006). Crystal structures of archael AGO proteins in complex with small nucleic acid guides reveal how the PAZ domain recognizes the small RNA 3′ end, while the 5′ end is anchored in the MID domain of the Argonaute (Wang et al. 2009b). Most small RNAs have a 5′ monophosphate and 3′ hydroxyl ends, while in a few cases like plant siRNAs and miRNAs or Drosophila siRNAs, the 3′ terminal nucleotide is modified with a 2′-O-methyl marker (Yu et al. 2005; Horwich et al. 2007). Such a 3′ end modification is a universal feature of piRNAs in all organisms studied so far (Kirino & Mourelatos 2007b). Structural insight into recognition of the modification by the Piwi proteins have come from crystallographic and Nuclear Magnetic Resonance (NMR) solution structures of Piwi PAZ in complex with methylated RNA ligands (Simon et al. 2011; Tian et al. 2011). The overall twisted β-barrel architecture of the Piwi PAZ domain is similar to that described for other Argonautes (Sashital & Doudna 2010). From structures of the bound and un-bound states it is clear that the methylated 3′ end of piRNAs is accommodated into a pre-formed hydrophobic pocket. Surprisingly, the methyl marker contributes only modestly to the interaction with Piwi PAZ and provides no dramatic increase in binding affinities. This suggests that the specific loading of piRNAs to Piwi proteins is not dictated by recognition of the piRNA modification. The guide nucleic acid residue at position 1 is singled out for base-specific contact by the Argonaute (Wang et al. 2009b). Indeed, crystal structure of human Ago2 MID domain highlighted the importance of a rigid loop in specifying the bound nucleotide (Frank et al. 2010). It remains to be seen whether the strong preference for a uridine as the first nucleotide (U1-bias) in piRNA sequences is a consequence of specific recognition by Piwi MID domain or an outcome of piRNA biogenesis mechanisms (Aravin et al. 2006; Girard et al. 2006).
Some Argonautes are small RNA-guided nucleases (Slicers) that can catalyze cleavage of target nucleic acids identified by base-pair interactions. This property is housed in the RNase-H-like fold of the PIWI domain and determined by the presence of a catalytic triad DDH (Asp-Asp-His) needed for coordination of magnesium ions (Martinez & Tuschl 2004; Meister & Tuschl 2004; Song et al. 2004; Wang et al. 2009b). Catalytic activity has been biochemically demonstrated for some Piwi proteins (Lau et al. 2006; Saito et al. 2006; Gunawardane et al. 2007; Reuter et al. 2011). The in vivo importance of Piwi catalysis is highlighted by the fact that single point mutation of MILI and MIWI catalytic motif results in infertile mice, phenocopying the complete knock-outs (De Fazio et al. 2011; Reuter et al. 2011). A similar catalytic triad mutation of MIWI2 did not affect fertility, suggesting that MIWI2 could potentially be catalytically inactive (De Fazio et al. 2011). However, this needs to be directly examined by biochemical assays. MIWI slicer activity was investigated using purified endogenous MIWI complexes and shown to require extensive base-pairing between the piRNA guide and the target RNA (Reuter et al. 2011). Mismatch between the first nucleotide of the piRNA and the target did not affect catalysis, as expected from its insertion into the MID domain of the Argonaute. Base-pairing between target RNA and piRNA residues at positions 2–21 were absolutely essential for MIWI catalysis, reflecting a need to form at least two turns of an A-helix to produce the catalytically competent conformation.
Piwi proteins are unique among Argonautes in using post-translational modifications to enable complex assembly. Piwis are post-translationally modified by symmetrical dimethyl (sDMA) marks at their N-terminus (Kirino et al. 2009; Reuter et al. 2009; Vagin et al. 2009), which in flies is catalyzed by the protein arginine methyltransferase 5 (PRMT5) (Kirino et al. 2009). Such marks are recognized by Tudor domains of a large family of mainly gonad-specific Tudor domain-containing proteins (TDRDs) (Heo & Kim 2009; Handler et al. 2011) (see below).
Deep sequencing analysis of piRNA libraries indicated a very complex nature of the RNA sequences (Aravin et al. 2006; Girard et al. 2006; Grivna et al. 2006a). This corresponds to over several million individual piRNA sequences compared with a few hundred miRNAs. Mapping of piRNA reads have revealed a clustered genomic origin for these sequences, with most (>90%) piRNAs within a library mapping to approximately 100 genomic windows. The clusters typically range from a few kilobases (kb) to over a 100 kb in length. Most mouse clusters show profound strand asymmetry, with reads arising from only one strand within a cluster (uni-strand cluster). When piRNAs map to both strands, they do so in a non-overlapping manner (bidirectional cluster) (Aravin et al. 2006; Girard et al. 2006). In embryonic germ cells, piRNAs map to both strands (dual strand clusters), but do not show the typical phasing observed for siRNAs on double-stranded precursors (Aravin et al. 2008).
One unifying feature of piRNAs in all organisms is the presence of repeat-derived sequences. Intergenic regions and genic transcripts (mainly exons) also contribute to the piRNA pool (Aravin et al. 2006, 2008; Girard et al. 2006; Brennecke et al. 2007). Depending on the stage at which they are expressed during spermatogenesis two distinct piRNA populations are identified in mice: the pre-pachytene and pachytene piRNA pools. Pre-pachytene piRNAs are enriched (approximately 80% in MIWI2) in repeat-derived sequences and associate with MIWI2 and MILI (Aravin et al. 2008). Pachytene piRNAs on the other hand, have a higher proportion (approximately 70%) of intergenic, unannotated sequences, with diminished contribution from repeat-derived sequences (approximately 25%) (Aravin et al. 2006; Girard et al. 2006; Reuter et al. 2011). Pachytene piRNAs enter MILI and MIWI in pachytene spermatocytes and round spermatids. Unlike miRNAs, sequence of piRNA-producing loci are not conserved, however, their genomic location seems to be strikingly conserved from rodents to humans. The functional implication of such synteny is presently unclear (Girard et al. 2006).
Primary piRNA biogenesis
The RNase III enzyme Dicer is key to the biogenesis of siRNAs and miRNAs from double-stranded RNA precursors (Siomi et al. 2011). However, the longer size of piRNAs already suggested a distinct biogenesis mechanism, which is independent of Dicer requirement (Vagin et al. 2006; Houwing et al. 2007; Saito et al. 2010). The observation that piRNAs map exclusively to one of the cluster strands, led to the suggestion that long single-stranded precursors are originators of piRNAs (Brennecke et al. 2007). Such strand-specific transcripts are detected by reverse-transcription coupled polymerase chain reaction (RT-PCR) and RNA-seq analysis (Lau et al. 2006; Reuter et al. 2011). Genetic support for this proposal comes from Drosophila, where a P-element insertion into the putative promoter of a piRNA cluster abolished piRNA production from a 180 kb region downstream (Brennecke et al. 2007). Production of piRNAs from such long single-stranded precursors is poorly understood and is tentatively termed as primary piRNA biogenesis (see below).
How certain genomic regions are defined as piRNA clusters is not clear. In C. elegans a specific sequence motif is detected upstream of piRNA-producing regions, but similar nucleotide sequence identifiers are not detected in other organisms (Ruby et al. 2006). It is possible that the cell recruits the transcription machinery to these genomic regions via specific chromatin marks or by incorporation of these regions into a transcription factory (Osborne et al. 2004) for a coordinated expression from these loci. This might then facilitate co-transcriptional tagging of the cluster transcripts with specific RNA-binding proteins that can chaperone the precursors to cytoplasmic nuages, which are putative sites for piRNA biogenesis (see below).
Within a cluster, individual piRNAs are randomly spaced and distributed across the breadth of the genomic window. But in many regions with high piRNA density, almost every single nucleotide position serves as the 5′ end of piRNAs. This implicates the action of two nucleases: an endonuclease that breaks down the long precursor transcript, followed at some loci with a 5′–3′ exonuclease digestion (Fig. 2A). These pre-piRNAs are then loaded into Piwi proteins, where the 5′ of the pre-piRNA is sampled by the Piwi MID domain. The observed U1-bias of piRNAs could very well be an outcome of the sequence preference of the endonuclease activity. However, recent studies point to an inherent property of the Argonaute MID domain itself in specifying this bias. First, crystal structure of the Argonaute2 MID domain reveals the presence of a rigid loop that can provide base-specific recognition of the first small RNA nucleotide, and displays a preference for uridines (Frank et al. 2010). Second, in an in vitro piRNA loading assay with Bombyx mori BmN4 cell lysates, a U1-containing 50 nt RNA was preferentially loaded into a Piwi protein (Kawaoka et al. 2011). Third, in mouse piRNA libraries, although piRNAs starting with every consecutive nucleotide is detected within a high-density cluster region, those starting with a U show higher abundance, suggesting that U1-containing piRNAs are somehow selectively captured and enriched in Piwi proteins.
Subsequent to incorporation into Piwi proteins, the pre-piRNAs are 3′ trimmed by an exonuclease activity, which is sterically hindered by the bound Piwi protein, whose footprint determines the size of the mature piRNA. This is consistent with the observation that >90% of pachytene piRNAs bound by MILI (approximately 26 nt) and MIWI (approximately 30 nt) share identical 5′ ends, but differ at their 3′ termini (Aravin et al. 2006; Reuter et al. 2011). Biochemical proof for such an exonuclease activity comes from in vitro piRNA loading assays performed using insect cell-free systems, where a Piwi-loaded 50 nt pre-piRNA was 3′ trimmed down to the mature size of piRNAs (Kawaoka et al. 2011). Although of unknown identity, 3′ trimming is the outcome of a magnesium-dependent exonuclease. This 3′ end definition is closely followed by 3′ end methylation of mature piRNAs by the RNA methyltransferase Hen1 (Horwich et al. 2007; Kirino & Mourelatos 2007a; Saito et al. 2007; Kawaoka et al. 2011). The modified piRNA 3′ end is then inserted into the Piwi PAZ domain to complete primary piRNA biogenesis.
Secondary piRNA biogenesis
Initially described in Drosophila, this pathway involves slicer-mediated cleavage of sense transposon transcripts by the fly Piwi protein Aubergine (Aub), generating the 5′ end of a new secondary piRNA that enters Ago3 (Brennecke et al. 2007; Gunawardane et al. 2007). These sense piRNAs guide Ago3 to cleave antisense cluster transcripts to generate the same initial Aub-associated antisense piRNA. Although Piwi slicer activity is implicated in the 5′ end generation, the 3′ end generation of piRNAs is believed to use similar trimming and methylation activities, as described above for primary piRNAs. Two characteristic signatures are observed in piRNA populations displaying the amplification cycle: first, a significant (>50%) proportion of Aub- and Ago3-bound piRNAs show overlap of their first 10 nucleotides, which is an outcome of the biochemistry of Argonaute slicer activity; second, while Aub piRNAs show a strong U1-bias, Ago3 piRNAs display a preference for A at the 10th position (A10-bias). This model is genetically supported by the observed collapse of secondary piRNAs in ago3 mutants, and an overall reduction in antisense piRNAs entering Aub (Li et al. 2009).
The importance of secondary biogenesis for transposon silencing is highlighted by the fact that MIWI2 accommodates most of the repeat-derived antisense piRNAs (Aravin et al. 2008). Indeed, in MiliDAH and Tdrd1 mutants, LINE1 (L1) elements are selectively upregulated, as MIWI2 is the repository of most antisense piRNAs to L1 elements (Aravin et al. 2008; Reuter et al. 2009; Vagin et al. 2009; De Fazio et al. 2011). As new secondary piRNAs are sourced from existing sense and antisense transposon transcripts, the piRNA amplification pathway can shape the piRNA profile to reflect the threat from active transposons. This is comparable to the adaptive immune system, where piRNAs targeting active transposon elements get quickly amplified for mounting an effective response (Aravin et al. 2007a; Brennecke et al. 2007). The role of maternally contributed fly piRNAs in initiation of the piRNA amplification cycle to counter newly acquired paternal transposons highlights the utility of having an adaptable small RNA-driven immune system (Blumenstiel & Hartl 2005; Brennecke et al. 2008).
Tudor family genes in the piRNA pathway
The Piwi proteins are central catalytic components of the piRNA system, but they do not act alone. Growing body of evidence is emerging that the Piwi proteins form a larger effector complex, as the AGO proteins and miRNAs form the RISC complex to regulate mRNA translation and/or stability (Ghildiyal & Zamore 2009). The first protein identified to interact with the Piwi protein in mice, MILI, is TDRD1, a germline-specific member of Tudor family proteins (Chuma et al. 2003, 2006; Kojima et al. 2009; Reuter et al. 2009; Vagin et al. 2009; Wang et al. 2009a). The Tudor family genes in general encode a tudor domain(s), which recognizes arginine dimethylation of target proteins (Chen et al. 2011) and they structurally resemble chromo, agenet, MBT, PWWP and bromo domains (Maurer-Stroh et al. 2003), which all recognize protein modifications. In the mammalian genome, there are approximately 30 genes that encode a tudor domain(s), with some members, such as Smn1 and Trp53bp1, being widely expressed in both somatic and germ cells. Another specific subclass of tudor family genes, termed tudor domain containing (TDRD) genes, are preferentially expressed in the germline, including TDRD1 as mentioned above (Chuma et al. 2003, 2006). TDRD1 encodes four tudor domains and a MYND domain, and shows a similar expression pattern as MILI. TDRD1 association with MILI is dependent on arginine dimethylation at the MILI N-terminus (Kojima et al. 2009; Reuter et al. 2009; Vagin et al. 2009; Wang et al. 2009a). Another TDRD member, TDRD9 (a homologue of Drosophila Spindle-E/homeless) interacts with MIWI2 (Shoji et al. 2009), suggesting that the Piwi-Tudor association is a conserved feature and each interaction has its own binding specificity. The two TDRD members, TDRD1 and TDRD9, cooperate non-redundantly in the Piwi pathway, and their loss-of-function mutations exhibit male-specific sterility showing meiosis defects similar to Mili and Miwi2 mutations, with LINE1 retrotransposons being upregulated and piRNA profiles affected. DNA methylation is also reduced at LINE1 promoters, indicating that TDRD1 and TDRD9 are essential components of the epigenetic regulation by the Piwi pathway, too.
Recent studies identified other TDRD members, including TDRD5 and TDRD7, to participate in the Piwi pathway and/or retrotransposon silencing (Hosokawa et al. 2007; Tanaka et al. 2011; Yabuta et al. 2011). In Tdrd5 mutants, haploid spermiogenesis is affected, LINE1 retrotransposon is desilenced and cognate DNA methylation is reduced, with MIWI2 localization being altered (see below), indicating that TDRD5 is a functional component of the Piwi pathway, although the piRNA profile has not been examined in this mutant. Similarly, Tdrd7 mutation also desuppresses LINE1 retrotransposons, but unlike other Tdrd genes, available data did not detect any defect in the Piwi pathway including piRNA biogenesis, DNA methylation and subcellular localization of the Piwi pathway components. Also, the developmental kinetics of LINE1 activation is different from those of other Tdrd and Piwi mutants. Currently, the detailed mechanism of the TDRD7 action is unclear, but Tdrd family genes may operate in several distinct pathways to act against retrotransposons. At the molecular level, TDRD proteins most likely function as scaffolds to assemble macromolecular complexes through tudor domains as well as other domains in each TDRD member, like MYND (TDRD1), LOTUS (TDRD5 and 7) and helicase (TDRD9) domains. Given that all Tdrd mutants reported so far lead to male sterility, such scaffolding activity of the TDRD proteins, including the proper assembly of the Piwi effector complexes, is indispensable in male germline development. The role of Tudor family genes in retrotransposon regulation is evolutionarily conserved as exemplified by the involvement of Drosophila Krimper, Tejas, Yb and Spindle-E in the fly piRNA pathway (Siomi et al. 2010). Together, tudor family members are now emerging as key conserved factors that ensure the germline integrity of diverse animals.
Additional factors for piRNA biogenesis and function
Biochemical purifications of mouse Piwi complexes and mutational studies have identified additional factors implicated in the piRNA pathway. This includes three proteins with helicase domains. The putative mouse RNA helicase MOV10L1 or its fly orthologue Armi are required for primary piRNA biogenesis (Frost et al. 2010; Haase et al. 2010; Olivieri et al. 2010; Saito et al. 2010; Zheng et al. 2010). They are believed to be involved in loading of primary piRNA precursors to Piwi proteins, as 25–70 nt piRNA precursor intermediates (piRNA intermediate-like, piR-IL) were detected in Armi complexes (Saito et al. 2010). piR-ILs sequences encompass mature piRNA sequences, but do not always align with the piRNA ends. They were shown to carry a 5′ hydroxyl and 2′–3′ cyclic phosphate groups, suggesting that Armi-associated piR-ILs still require further processing at both ends to become mature piRNAs. Similar piRNA precursors are not yet described in MOV10L1 complexes.
The fly Vasa or the Mouse Vasa Homologue (MVH/DDX4) is a demonstrated helicase, which when deleted abolishes secondary piRNA biogenesis in both flies and mice (Malone et al. 2009; Kuramochi-Miyagawa et al. 2010). The exact role of this helicase activity is presently unclear. Similar to Piwi proteins, Vasa also carries sDMA marks on its N-terminus (Kirino et al. 2010), which might allow it to integrate with Piwi complexes by plugging into tudor scaffold proteins. Finally, mouse TDRD9 or its fly orthologue SPN-E contains both helicase and tudor domains and is also implicated in secondary piRNA biogenesis, although its exact role is unclear (Malone et al. 2009; Shoji et al. 2009).
Maelstrom (MAEL) and GASZ are additional factors demonstrated to participate in the piRNA pathway (Soper et al. 2008; Ma et al. 2009). MAEL is a high molecular group box (HMG-box) protein, while GASZ is reported to stabilize MILI and help form cytoplasmic granules in which piRNA pathway components are detected. Their mechanistic roles are currently unclear. Finally, Zucchini (Choi et al. 2006; Saito et al. 2009; Haase et al. 2010; Olivieri et al. 2010; Huang et al. 2011; Watanabe et al. 2011a) is a member of the phospholipase D family and implicated in the piRNA pathway in both flies and mice. Although long suspected to be a nuclease, biochemical assays failed to confirm this. Interestingly, zuc is required for primary piRNA biogenesis in flies, while in mice it only impacts overall levels of piRNAs.
Mechanisms of the Piwi pathway action
Transposon silencing is the central role of the piRNA pathway. In the embryonic germline, cytosolic MILI likely uses its slicer activity to not only birth MIWI2 piRNAs, but also destroys transposon mRNAs (De Fazio et al. 2011) (Fig. 2B). In contrast, nuclear MIWI2 initiates transcriptional silencing by promoting DNA methylation of transposon promoter regions. The piRNA-guided MIWI2 likely gains access to transposon promoters via base-pairing with nascent transcripts from the target loci, similar to what is described for the RNA-induced transcriptional silencing (RITS) complex in fission yeast (Motamedi et al. 2004). The possibility that MIWI2 is perhaps not a slicer only strengthens such a hypothesis as its retention at target loci is enhanced by non-cleavage of the RNA tether. How MIWI2 recruits the DNA methylation machinery or other necessary chromatin modifiers is presently unknown. In post-natal stages, repeat-derived pachytene piRNAs guide MIWI slicer activity for a surveillance role in round spermatids by post-transcriptionally destroying L1 transposon mRNAs that escape silencing (Reuter et al. 2011). Thus, Piwi slicer activity is required for both initiating transcriptional silencing of transposons in the embryo, and for its maintenance at the post-transcriptional level after birth. The question then is why MILI, which is loaded with similar piRNAs as MIWI, fails to complement the loss of MIWI catalytic activity? Possible explanations could come from the abundance of MILI ribonuceloprotein (RNP) that sharply drops from pachytene spermatocytes to round sparmatids, and due to differences in complexes associating with the two Piwi proteins (Vagin et al. 2009). The function of pachytene piRNAs arising from unannotated genomic regions is presently unknown.
piRNAs are also implicated in establishment of genomic imprints on mouse Rasgrf1 locus via slicing of a non-coding RNA traversing the silenced domain (Watanabe et al. 2011b). However, the recent work on Piwi slicer activity raises questions as to which Piwi is the slicer for this presumably nuclear event (De Fazio et al. 2011). piRNAs in flies are also implicated in control of genic transcripts via slicer cleavage and deadenylation-mediated decay (Nishida et al. 2007; Rouget et al. 2010). However, in mice there is no evidence suggesting a regulation of mRNA levels by Piwi proteins (Unhavaithaya et al. 2009; Reuter et al. 2011). Nevertheless, translation control by MIWI piRNPs is a reported possibility, but it is not known how target mRNAs are selected, and whether piRNAs have a role in this process (Grivna et al. 2006b; Unhavaithaya et al. 2009).
Spermatogenesis defects by Piwi pathway mutations in mice
In mice, all loss-of-function mutants of Piwi pathway genes identified so far are male-specific sterile. Although piRNAs and several Piwi pathway factors are detectable in oocytes, the reason why females are not affected is currently unclear (Tam et al. 2008; Watanabe et al. 2008). One explanation would be that endogenous siRNAs are generated in mammalian oocytes and the canonical RNAi pathway functions to suppress retrotransposons (Tam et al. 2008; Watanabe et al. 2008). Another speculation is that mutant oocytes are actually affected to certain degrees, which however, do not lead to overt cellular phenotypes. Such possibilities need to be carefully explored in future studies.
In the male, Piwi pathway mutants show spermatogenesis defects and the phenotypes are observed mainly at two distinct developmental stages. First, the earlier phenotype is seen during meiosis of spermatocytes and this group of mutants include Mili, Miwi2, Tdrd1, Tdrd9, Mvh, Mael, Mov10l, Gasz, zucchini/Mitopld/Pld6 and Gtsf1/Cue110 (Kuramochi-Miyagawa et al. 2004, 2010; Chuma et al. 2006; Carmell et al. 2007; Soper et al. 2008; Ma et al. 2009; Shoji et al. 2009; Yoshimura et al. 2009; Frost et al. 2010; Zheng et al. 2010; Huang et al. 2011; Watanabe et al. 2011a). Note that in the Gtsf1/Cue110 mutant the link between transposon derepression and the piRNA pathway is not clearly established (Yoshimura et al. 2009). The late phenotype, on the other hand, becomes apparent post-meiotically in haploid spermatids in Miwi and Tdrd5 mutants (and a Tdrd1 mutant, which shows both spermatocyte and spermatid defects) (Deng & Lin 2002; Chuma et al. 2006; Yabuta et al. 2011). In both groups of mutants, germ cells are severely degenerated and no functional sperm are produced, resulting in complete male sterility. Remarkably, in the former group of mutants, wherein the cellular phenotype is evident in postnatal spermatocytes, distinct molecular changes are already seen at a much earlier stage, in fetal prospermatogonia (gonocytes) during embryonic development. In these mutants, LINE1 and/or IAP retrotransposons are clearly activated in prospermatogonia, while at this embryonic stage, there is no detectable defect in cell viability. The biogenesis of piRNAs per se (Mili) or the sequence profiles (Miwi2, Tdrd1, Tdrd9, Mvh, Mael, Gasz, Mov10l, zucchini/Mitopld/Pld6) are also affected, and LINE1 and/or IAP loci are hypomethylated in mutant prospermatogonia. The reason for the delayed phenotype is currently unknown, but one of the simplest explanations is DNA demethylation. That is, the epigenetic change in prospermatogonia is transmitted to postnatal spermatogonial stem cells and then to following meiosis of spermatocytes, in which trans-acting factor(s) trigger retrotransposon expression to a moderate level in the wild-type (non-toxic), but to a much higher detrimental level in the mutants, wherein epigenetic suppression i.e. DNA methylation does not work. Another possibility is that unusual “open” chromatin conformation of the retrotransposon loci in the mutants, which are quite repetitive in the genome, may fail to suppress non-allelic homologous pairing and/or recombination between the repetitive sequences or may lead to aberrant meiotic chromosome condensation, which should then lead to meiotic checkpoint activation followed by apoptosis.
The “late” phenotype group of Piwi pathway mutants (Miwi and Tdrd5) shows post-meiotic defects in haploid spermatids. In Miwi null knock-out, spermiogenesis is arrested at a round spermatid stage with the piRNA level being greatly reduced (Deng & Lin 2002), but the molecular link between the piRNA biogenesis and spermatid development has not been clarified. Recently, we showed that MILI acts to suppress LINE1 retrotransposons in round spermatids depending on its slicer activity (Reuter et al. 2011) and this LINE1 regulation by MIWI does not involve epigenetic change (DNA methylation), but has a significant impact on the cleavage of LINE1 transcripts by sequence complementarity to piRNAs. MIWI therefore mainly acts post-transcriptionally to silence retrotransposons in postnatal spermiogenesis, while MILI and MIWI2 establish both transcriptional and post-transcriptional repression of retrotransposons in fetal germ cells in embryos. MIWI has also been shown to associate with translational machinery (Grivna et al. 2006b) and is required for the expression of its target mRNAs involved in spermiogenesis (Deng & Lin 2002). MIWI likely functions through several distinct modes of action to target retrotransposon transcripts and possibly other RNA species.
Tdrd5 mutation also causes a round spermatid arrest (Yabuta et al. 2011). However, despite the post-meiotic spermiogenesis phenotype, fetal prospermatogonia are clearly affected similarly to Mili and Miwi2 mutants, i.e., LINE1 retrotransposon is upregulated and the localization of MIWI2 and TDRD9 is altered at the embryonic stage. Tdrd5 mutants also show DNA demethylation at LINE1 loci. Together, these observations indicate that TDRD5 participates in the Piwi pathway in fetal prospermatogonia, while the cellular phenotype appears much later in postnatal spermatids. Currently, we do not have a good explanation for this long lag period. The level and extent of retrotransposon activation in this mutant may be relatively moderate as compared with other mutants, or TDRD5 may have more diverse roles other than retrotransposon silencing and piRNA pathway regulation. A more detailed understanding on how the Piwi pathway and other cellular processes crosstalk and integrate in spermatogenesis awaits future investigations.
Subcellular RNP assembly of Piwi pathway components in the germline
Another compelling feature of the Piwi pathway is its close association with germinal granules/nuage. Germinal granules/nuage are an evolutionarily conserved cytoplasmic RNP characteristically observed in the germline of many animals (Eddy 1975). These structures have been recognized for years and have been implicated in the germline specification of several representative model animals, such as Drosophila and C. elegans, on the basis that germinal granules in oocytes/early embryos of these species are concentrated with maternal germline determinants and are asymmetrically segregated to prospective germ cells (Saffman & Lasko 1999; Seydoux & Braun 2006; Strome & Lehmann 2007). In mammals, however, the germline determination occurs among pluripotent epiblast cells depending on inductive signals from surrounding somatic cells (Mclaren 2003), and in accordance with this, germinal granules are not observed in early mammalian embryos. Instead, they become discernible at later stages of germ cell differentiation, i.e. during spermatogenesis and oogenesis (Eddy 1975; Chuma & Pillai 2009). These mammalian germinal granules are classified into two types, (i) inter-mitochondrial cement (inter-mitochondrial material/bar, also termed pi-body), which is assembled among clusters of mitochondria in fetal prospermatogonia, postnatal spermatogonia and meiotic spermatocytes in the male and in developing oocytes in the female, and (ii) chromatoid bodies, which are a more conspicuous form of mammalian germinal granules that develop in meiotic spermatocytes and haploid spermatids. Such germinal granule/nuage structures in mammals have long been recognized but remained unaddressed for years especially at the molecular level.
Little is yet known about the functional significance of such Piwi pathway localization to germinal granules/nuage, but with respect to inter-mitochnodrial cement, it has been found that this structure is absent or greatly reduced in a group of Piwi pathway mutants, including Mili, Tdrd1 and Mvh, suggesting that inter-mitochnodrial cement among mitochondrial clusters is actually under the control of the Piwi pathway (Chuma et al. 2006; Kuramochi-Miyagawa et al. 2010). Recent studies also provided another evidence for a possible link between the Piwi pathway and mitochondria per se. Zucchini/MITOPLD/PLD6, which belongs to the phospholipase D family, is essential for male fertility and its loss-of-function mutations cause severe defects in piRNA biogenesis and retrotransposon suppression (Huang et al. 2011; Watanabe et al. 2011a). Intriguingly, Zucchini/MITOPLD/PLD6 is located on the outer membrane of mitochondria, and its overexpression (in somatic cells) facilitates mitochondrial fusion through the lipid signaling (Choi et al. 2006), while its loss-of-function disrupts inter-mitochondrial cement and affects the distribution of mitochondria, too (Huang et al. 2011; Watanabe et al. 2011a). Zucchini/MITOPLD/PLD6 thus provides an important link that associates the Piwi pathway to mitochondria, as well as inter-mitochondrial cement. Remarkably, this type of germinal granules/nuage among mitochondria is conserved in divergent animals (Eddy 1975), and whether it has a similar correlation to the Piwi pathway is an intriguing issue to be established.
Chromatoid bodies, as compared to inter-mitochondrial cement, contain even more Piwi pathway factors, which include (i) the components of inter-mitochondrial cement in (pro)spermatogonia and spermatocytes (MILI, TDRD1 etc.); (ii) processing body proteins, such as TDRD9 and MAEL as well as more ubiquitous GW182 and AGO2 (Kotaja et al. 2006a; Tanaka et al. 2011); and (iii) Piwi pathway proteins that start to be expressed in and/or after meiosis, such as MIWI. However, despite such characteristic composition, a functional involvement of chromatoid bodies in the Piwi pathway is not at all known. In several mutants of the Piwi pathway genes, such as Miwi and Tdrd5, chromatoid bodies are structurally reduced in size and are fragmented (Kotaja et al. 2006b; Reuter et al. 2011; Yabuta et al. 2011), but their overall architecture, both structurally and compositionally, is retained, although the Piwi pathway is significantly disturbed in the mutants (Grivna et al. 2006b; Reuter et al. 2011; Yabuta et al. 2011). In addition, recently, a series of chromatoid body mutants were obtained by making Tdrd7, Tdrd6 and double mutants of them (Tanaka et al. 2011). In the Tdrd7 mutant, chromatoid bodies and processing bodies, which usually are structurally fused during meiotic division in the wild-type (see Tanaka et al. 2011 for detail of chromatoid body biogenesis), remain aberrantly segregated, leading to a peculiar premature form of chromatoid bodies, which adjoin processing bodies in the mutant spermatids. Tdrd6 mutants then show a chromatoid body deficiency with reduced size and amount similar to Miwi and Tdrd5 mutants at a later stage of spermiogenesis (Vasileva et al. 2009; Tanaka et al. 2011). Further, Tdrd7 and Tdrd6 double mutations cause a complete lack of chromatoid bodies from their initial assembly stage in spermatocytes to the later maintenance process in spermatids. In all these Tdrd7 and Tdrd6 mutants, which exhibit severe deficiencies in chromatoid bodies in common, the piRNA levels and sequence profiles are not significantly affected. This unexpectedly unveiled that chromatoid bodies or at least their structural integrity are not a prerequisite for piRNA biogenesis per se, and until now, we still do not know why and how almost all the Piwi pathway components are assembled at this characteristic cytoplasmic compartment in the spermatid. Chromatoid bodies have been shown to have features similar to processing bodies, stress granules and also aggresomes (Haraguchi et al. 2005; Kotaja et al. 2006a; Tanaka et al. 2011). These structures in common are thought to assemble and sequester target RNAs and/or proteins into massive aggregates for degradation or later reuse (Bhattacharyya et al. 2006). Chromatoid bodies may function through aggregation and sequestration of unnecessary cellular machineries, including Piwi pathway components, which probably have to be turned off, to avoid unnecessary processing of RNAs etc., after a proper developmental time window. If so, the two forms of mammalian germinal granules/nuage, i.e., inter-mitochondrial cement and chromatoid bodies, play rather different roles in the regulation of the Piwi pathway. The detailed understanding of the Piwi-small RNA pathway, its RNP assembly and subcellular compartmentalization again await future intense investigations.
We touched briefly on the piRNA system in male germline development in mice. So far, our knowledge of this small RNA pathway is still limited and partial. Key unsolved issues include the primary processing of piRNA precursors, the target selection and specificity if any, and the epigenetic functioning of the piRNA pathway in fetal germ cells in mice. At the same time, the role(s) of piRNA is still expanding. The Piwi pathway in Drosophila targets several endogenous mRNAs in addition to retrotransposon transcripts (Robine et al. 2009; Saito et al. 2009; Rouget et al. 2010). In mice, genomic imprinting of a paternal imprinted gene, Rasgrf1, is regulated by the Piwi pathway through a neighboring non-coding RNA containing a retrotransposon sequence (Watanabe et al. 2011b). The Piwi pathway likely has an effect on endogenous mRNAs and possibly non-coding RNAs other than retrotransposons in trans and in cis. Further, there are several reports that piRNA and Piwi pathway factors are detectable in somatic cells (Wu et al. 2010; Wang et al. 2011; Yan et al. 2011), although whether such expression might have any physiological function or not is still unclear. Together, further comprehensive and careful characterization of the Piwi-small RNA system will uncover novel and intriguing molecular aspects present but hidden in the germline and perhaps beyond.
We would like to thank members of our laboratories for their patient support and helpful discussion during the preparation of this manuscript. S.C. is also grateful to Norio Nakatsuji for his support and encouragement. R.S.P is grateful to Jordi Xiol and Zhaolin Yang for helpful discussions.