The Drosophila fragile X mental retardation protein participates in the piRNA pathway

Mutations in the fragile X mental retardation protein (FMRP, also known as FMR1) cause Fragile X syndrome, which entails cognitive and gonadal defects in humans (de Roodt et al., 2009; Jin et al., 2004a; Jin and Warren, 2000; Tamanini et al., 1997; Verkerk et al., 1991). The members of the FMRP family act in the control of RNA translation in the nervous system (Epstein et al., 2009; Jin et al., 2004a; Jin et al., 2004b) and in the ovaries (Epstein et al., 2009). More interestingly, these proteins have been reported to associate and interact with components of the RNA-induced silencing complex (RISC) including AGO1 and AGO2, two members of the Argonaute protein family, which mediate post-transcriptional control through small RNAs. Thus, RNA interference (RNAi)-mediated pathways are known to be involved in the translational regulation mediated by FMRP (Caudy et al., 2002; Ishizuka et al., 2002; Jin et al., 2004b).

AGO1 is predominantly involved in the microRNA (miRNA)-mediated silencing and AGO2 in the small interfering RNA (siRNA)-mediated silencing (Aravin et al., 2001; Czech and Hannon, 2011). The PIWI subclade of Argonaute proteins, which includes Drosophila Aubergine (Aub), AGO3 and Piwi, interacts with the so-called piRNAs and silences transposable as well as repetitive elements, thereby protecting the genome from instability. The piRNA-mediated pathway was first discovered in the germline (Carmell et al., 2002; Gunawardane et al., 2007; Ishizu et al., 2012; Krusiński et al., 2008; Laskowski and Hillhouse, 2008; Senti and Brennecke, 2010; Thomson and Lin, 2009; Vagin et al., 2006 and references therein) and more recently in the nervous system (Baillie et al., 2011; Brennecke et al., 2008; Lee et al., 2012; Lee et al., 2011; Lukic and Chen, 2011; Perrat et al., 2013; Plaisant et al., 2008; Rajasethupathy et al., 2012; Reilly et al., 2013; Sato and Siomi, 2010; Specchia et al., 2010; Thomas et al., 2012). Although recent high throughput analyses have identified an increasing number of piRNA-related genes (Czech et al., 2013; Handler et al., 2013; Muerdter et al., 2013), the molecular mechanisms underlying the piRNA pathway have not been fully elucidated yet. Furthermore, although the different classes of small RNA display specific features, there seems to be no clear line separating them (Förstemann et al., 2007). Indeed, AGO1 and AGO2 can compete for binding with miRNAs (Laskowitz and Warner, 2008) and ectopic expression of Aub in the soma competes for the siRNA pathway mediated by AGO2 (Specchia et al., 2008). Such promiscuity, the known role of the FMRP protein family in small RNA metabolism (Ascano et al., 2012a; Ascano et al., 2012b; Epstein et al., 2009; Ishizu et al., 2012) and the presence of transposons in the nervous system, a key target tissue of the FMRP protein family (Lee et al., 2011; Perrat et al., 2013), prompted us to hypothesize the involvement of the Drosophila ortholog dFmr1 protein in the piRNA pathway.

One of the most known piRNA-mediated pathways in Drosophila is the one regulating the interaction between crystal [cry, also known as Su(Ste)] and Stellate (Ste, also known as SteXh) (Bozzetti et al., 2012; Bozzetti et al., 1995; Li et al., 2009; Nagao et al., 2010; Vagin et al., 2006). In wild-type (wt) testes, piRNAs mainly produced by the cry locus silence the expression of the Ste RNA. Males lacking this locus exhibit meiotic defects in chromosome condensation and segregation as well as crystalline aggregates, due to the production of the Ste protein (Bozzetti et al., 2012; Bozzetti et al., 1995; Palumbo et al., 1994; Tritto et al., 2003). All of the so far identified cry–Ste modifiers are involved in the piRNA pathway and, beyond the crystalline aggregates in spermatocytes, they also share sterility and germline activation of transposable elements (Li et al., 2009; Nagao et al., 2010; Specchia et al., 2010; Vagin et al., 2006; Zhang et al., 2011).

Here, we provide clear evidence that Drosophila Fmr1 (denoted dFmr1) acts in the piRNA pathway. First, dFmr1 loss-of-function (LOF) mutations exhibit crystalline aggregates in spermatocytes, and derepression of transposable elements in testes and ovaries as well as fertility defects. Second, dFmr1 overlaps and interacts genetically as well as biochemically with Aub. Surprisingly, dFmr1 is required both in the germline and in the somatic compartments of the testis to control the piRNA pathway. The present work provides novel perspectives in the field of RNA metabolism and broadens our understanding of the physiopathological mechanisms underlying the Fragile X syndrome.

The piRNA pathway silences transposons and repetitive elements, and the cry–Ste system provides a faithful readout to study silencing of the latter. As a first attempt to assess the role of dFmr1 in the pathway, we analyzed dFmr1 testes. dFmr1^▵50, dFmr1^▵113 homozygous and dFmr1^▵113/dFmr1^▵50 testes (referred to as transheterozygous throughout the text) exhibit crystals, a phenotype that is absent in wt animals (Fig. 1A–C,G). Interestingly, dFmr1^▵113/+ testes also show crystals, whereas dFmr1^▵50/+ testes do not (Fig. 1D,E). To clarify the different phenotypes of the mutations that were considered to be nulls, we performed western blot analyses on adult testes and on larvae and found that dFmr1^▵50 and dFmr1^▵113 represent a hypomorphic and a null allele, respectively (Fig. 1H; supplementary material Fig. S1), in line with the described size of the deletions, which include the first coding exon in the case of dFmr1^▵50 and the first two coding exons in the case of dFmr1^▵113 (Zhang et al., 2001; Zhang et al., 2004). This likely accounts for the almost complete lethality of homozygous dFmr1^▵113 (but not dFmr1^▵50) flies in our growing conditions.

Importantly, dFmr1 expression in testes driven by the T76 driver (Arya et al., 2006; Hrdlicka et al., 2002) rescues the ‘crystal’ phenotype of dFmr1^▵113 heterozygous and homozygous flies as well as the lethality of the homozygous flies (Fig. 1F, T76-Gal4/+;UAS-dFmr1, dFmr1^▵113/+; and data not shown), confirming that these defects depend on the dFmr1 mutation. Thus, dFmr1 regulates the cry–Ste interaction. For the rest of our analysis, we used dFmr1^▵113 heterozygous flies whenever possible but always retested the results using homozygous dFmr1^▵50 or transheterozygous flies.

The presence of crystals suggests a derepression of the Ste locus, as previously found in testes mutant for Aub (Schmidt et al., 1999; Specchia et al., 2008), a known modifier of the cry–Ste interaction (Fig. 1I,J). Compared to controls, dFmr1 mutant animals show an increase of Ste RNA expression, which we assessed by quantitative real-time PCR (qRT-PCR) assays. dFmr1^▵50 homozygous animals show a stronger increase (Fig. 1I,J; 29.7 fold) than that observed in dFmr1^▵113 heterozygous testes (8.7 fold) and, as expected, dFmr1^▵50 heterozygous animals show the lowest increase of Ste RNA levels (6.1 fold). At least an eightfold increase of Ste RNA expression is required to produce crystals (Specchia et al., 2010), likely explaining the absence of the crystalline structures in dFmr1^▵50 heterozygous testes.

Given that the decrease in cry-specific piRNAs induces Ste RNA expression and crystal formation (Aravin et al., 2003; Aravin et al., 2001; Nagao et al., 2010; Specchia et al., 2010), we analyzed the expression of two of the most abundant cry-specific piRNAs in dFmr1 mutant testes by northern blotting and by qRT-PCR, and we found that they decrease compared to what was observed in controls (‘crystal’ and ‘rasi4’; see Materials and Methods). dFmr1^▵50 homozygous and transheterozygous testes display the most severe reduction, dFmr1^▵113 heterozygous testes are less affected and dFmr1^▵50 heterozygous testes result was only significantly different for the qRT-PCR analysis (Fig. 2A,C,D,F and data not shown). Importantly, the levels of both cry-specific piRNAs are rescued upon reintroducing dFmr1 expression in mutant testes (Fig. 2B,E). To assess whether the piRNA pathway affecting transposons is also affected by dFmr1, we next tested the expression of a piRNA specific to the R1 transposable element. In line with the crystal phenotype, we observed a severe reduction of the levels of this piRNA in dFmr1^▵50 homozygous with respect to those found in dFmr1^▵50 heterozygous testes (Fig. 2C,D); dFmr1^▵113 heterozygous animals harbor an intermediate phenotype. Finally, we found no evidence for enhanced levels of piRNA precursors.

We also analyzed transposable element silencing. It is generally accepted that, in the fly gonads, the ‘primary piRNA pathway’ predominantly controls the levels of piRNAs specific to the so-called ‘somatic’ transposable elements and involves Piwi, whereas the secondary piRNA pathway involves AGO3 and Aub, requires a piRNA amplification loop and controls the levels of piRNAs specific to the ‘germline’ transposable elements (Brennecke et al., 2007; Li et al., 2009; Malone et al., 2009; Nagao et al., 2010). Amongst the ‘germline’ group of transposons, we analyzed roo, I and R1; amongst the ‘somatic’ group, we analyzed ZAM, gypsy, 412 and springer. The levels of the transcripts of the ‘germline’ transposons clearly increase in dFmr1^▵50 homozygous testes compared to those observed in dFmr1^▵50 heterozygous ones, which were taken as controls (Fig. 2G, left part). Specific ‘somatic’ transposons are also affected (ZAM levels increase moderately, whereas gypsy, 412 and springer levels do not; Fig. 2G, right part). Germline and somatic transposons are expressed at almost wt levels in dFmr1^▵113 heterozygous testes (Fig. 2H), confirming that the phenotype of these animals is weaker than that of dFmr1^▵50 homozygous ones.

Taken together, these results suggest that dFmr1 behaves as a bona fide member of the piRNA pathway. It controls the levels of cry-specific piRNAs as well as those of piRNAs related to ‘germline’ and ‘somatic’ transposons in testes, hence affecting the expression and/or movement of these transposons.

The movement of transposable elements is one of the molecular causes of DNA instability and sterility (Brennecke et al., 2007; Khurana et al., 2011) and most genes involved in the piRNA pathway identified so far, including all the cry–Ste modifiers, are partially or completely male-sterile (Handler et al., 2013; Li et al., 2009; Muerdter et al., 2013; Pane et al., 2007; Perrat et al., 2013; Schmidt et al., 1999; Specchia et al., 2008; Specchia and Bozzetti, 2009; Specchia et al., 2010; Stapleton et al., 2001; Tomari et al., 2004; Vagin et al., 2006). dFmr1^▵50 homo- and transheterozygous males are completely sterile [Fig. 2I, versus w¹¹¹⁸ ; similar results were obtained upon comparison with Oregon R (Or-R) flies, data not shown]. Importantly, the fertility defects of (heterozygous, transheterozygous and homozygous) dFmr1 mutant males are significantly rescued upon reintroducing dFmr1 expression (Fig. 2I, T76-Gal4/+;UAS-dFmr1, dFmr1^▵113/dFmr1^▵50 and data not shown).

The so far identified mutations of the piRNA pathway also exhibit female fertility defects associated with transposon activation (Czech et al., 2013; Ishizu et al., 2012). We therefore analyzed dFmr1^▵50 homozygous female gonads for the activation of the same transposons as those analyzed in males and found that the levels of I, roo and R1 transcripts increase even more than in testes (Fig. 2J, left part). As in males, the somatic transposons behave differently from one another: ZAM and gypsy are activated at a high rate, whereas springer and 412 remain almost at the same level as in wt animals (Fig. 2J, right part). Accordingly, the fertility of dFmr1 mutant females is significantly reduced (dFmr1^▵50 heterozygous, transheterozygous and homozygous females, Fig. 2K; we obtained similar results with Or-R, data not shown). The fertility of dFmr1^▵50 heterozygous and homozygous females is rescued upon reintroducing dFmr1 expression (data not shown).

In summary, dFmr1 affects transposon activation and fertility in males and in females.

To gain insights on the mode of action of dFmr1, we analyzed its profile of expression in the adult testes. dFmr1 is expressed in the germ cells, as confirmed by double labeling using the anti-Vasa germline-specific marker (Lasko, 2013 and references therein; supplementary material Fig. S2). It widely accumulates in the cytoplasm, as expected from its described role in translational control, and it overlaps with Vasa in the nuage, a perinuclear compartment where the other components of the piRNA pathway accumulate (Anand and Kai, 2012; Ishizu et al., 2012; Pek et al., 2012). A prominent colocalization occurs at the level of the piNG bodies, structures in which piRNA components are located and perform their function (Kibanov et al., 2011) (supplementary material Fig. S2C–H, see arrowheads).

Aub constitutes one of the best-known players of the piRNA pathway and its loss triggers the formation of crystals in testes (Schmidt et al., 1999) (Fig. 1I,J). As for the other proteins involved in the piRNA pathway, Aub is specifically expressed in the germline, and localizes to the ‘nuage’, accumulating preferentially at the piNG bodies (Kibanov et al., 2011; Li et al., 2009; Nagao et al., 2010; Pek et al., 2012; Specchia and Bozzetti, 2009; Zamparini et al., 2011; Zhang et al., 2012; Zhang et al., 2011). Double labeling with anti-Aub and anti-Fmr1 antibodies (Fig. 3A–L) confirms the overlap at the level of the nuage (Fig. 3E–L) and strong colocalization in the piNG bodies (Fig. 3G–L). Within the testis, Aub levels are highest in the apical tip (Fig. 3B), which corresponds to the earliest developmental stages. To further characterize Aub distribution throughout the testis, we used anti-Aub antibody staining and a marker for the fusome, a germline structure that changes morphology during germ cell differentiation, being globular in stem cells and gonialblast while branched in spermatogonia (Fig. 3P–R, red arrows in Fig. 3Q and testis schematic in Fig. 3S). The highest levels of Aub occur in stem cells, spermatogonia and possibly early spermatocytes. dFmr1 expression extends to the late stages. Aub expression therefore seems to colocalize with dFmr1 in a specific territory that corresponds to that of the early spermatocytes (Fig. 3A–C). dFmr1 does not seem to affect Aub expression, which still accumulates and localizes properly in dFmr1^▵50 homozygous testis (Fig. 3M–O).

We also analyzed the expression profile of Aub and dFMR1 in ovaries from wt adults (Fig. 4A–Q). The two proteins colocalize in the germarium, mostly at the position of the stem cells (Fig. 4A–D) and later on they overlap at the nuage of the nurse cells (Fig. 4E–P). By stage 6, the two proteins also colocalize in the oocyte cytoplasm (Fig. 4M–P), which will eventually contain the polar granules, where Aub, dFmr1 and Vasa accumulate (Costa et al., 2005; Hillebrand et al., 2007; Thomson et al., 2008). dFmr1 does not seem to affect Aub expression, which still accumulates and localizes properly in dFmr1^▵50 homozygous ovaries (Fig. 4R–U).

As a further evidence for a role of dFmr1 in the piRNA pathway, we asked whether the two proteins interact biochemically, as has been shown for dFmr1 and other members of the AGO family (Ishizuka et al., 2002; Jin et al., 2004a; Jin et al., 2004b; Zhang et al., 2004). Aub and dFmr1 cotransfection assays in S2 Drosophila cells, which are devoid of Aub, and immunoprecipitation with an anti-dFmr1 antibody revealed the presence of the Aub protein in the immunoprecipitate (Fig. 4V). A GST pulldown assay using a GST–dFmr1 fusion protein and the in vitro translated full-length Aub product demonstrates the direct interaction between the two proteins (Fig. 4W).

In summary, Aub and dFmr1 interact biochemically and colocalize in specific territories and cellular compartments of testes and ovaries.

The above data prompted us to analyze the genetic interaction between dFmr1 and Aub. First, we asked whether altering the levels of Aub modifies the effects of dFmr1 loss. Germline Aub overexpression rescues the dFmr1-mediated crystal phenotype showing that the two proteins work in the same pathway and act in the same direction (Table 1, genotypes 1–3). The data also suggest that Aub might act downstream of dFmr1. This was confirmed by doing the reverse experiment: dFmr1 overexpression (Table 1, genotypes 4,5) cannot rescue the crystal phenotype triggered by the aub knockdown (KD) (Table 1, genotypes 6,7). Finally, we analyzed the crystal phenotype of double heterozygous flies for dFmr1 and different Aub alleles. aub^HN and aub^QC42 are considered as strong hypomorphs that do not exhibit crystals in heterozygous testes but do so in transheterozygous conditions (Table 1, genotypes 9–11). As predicted from the mutant phenotypes of the single genes, aub^HN/+; dFmr1/+ and aub^QC42/+; dFmr1/+ flies exhibit crystals (Table 1, genotypes 12,13; Fig. 5A). We then analyzed aub^sting testes, which exhibit crystals in transheterozygous conditions with aub^HN and aub^QC42 and not in heterozygous conditions, as the two other alleles (Table 1, genotypes 14–16). aub^sting homozygous animals are viable and also present the crystal phenotype (Table 1, genotype 17). Strikingly, aub^sting/+; dFmr1/+ flies do not exhibit crystals (Table 1, genotype 18; Fig. 5B). Moreover, aub^sting/+ rescues the other phenotypes observed in dFmr1 mutant animals as well, namely, it reduces the expression of the Ste RNA transcript and that of transposable elements (supplementary material Table S1). Furthermore, aub^sting/+ rescues the male fertility defects linked to dFmr1^▵113 heterozygous, transheterozygous and homozygous conditions (Fig. 5C; data not shown). Finally, aub^sting/+ animals that are homozygous mutant for dFmr1 show no crystals and aub^sting/+; dFmr1^▵113/dFmr1^▵113 animals are fully viable (Table 1, genotypes 19,20).

Similar results were found in females, that is, aub^sting/+ reduces the expression of transposable elements observed in homozygous dFmr1^▵50 ovaries (data not shown) and rescues the fertility defects observed in homozygous and heterozygous dFmr1^▵113 females (Fig. 5D). Finally, aub^sting/+ rescues dFmr1^▵113/dFmr1^▵113 induced lethality in females as well.

aub^sting was previously shown to trigger ectopic Aub expression in somatic adult tissues (Specchia et al., 2008) and should be therefore considered as a neomorph in these territories. Given the known germline requirement of the piRNA pathway linked to the cry–Ste system, the rescue data with the aub^sting allele were quite surprising, prompting us to ask whether, in this allele, Aub is ectopically expressed in the somatic compartment of the gonad and whether the piRNA-mediated Ste silencing also has a somatic requirement.

Western blot analyses showed that the Aub protein is not detected in aub^HN/aub^QC42 flies but is present at low levels in aub^sting animals (Fig. 6A). Immunolabeling assays clarified these data: aub^sting testes show reduced overall levels of Aub in the germline and high levels in the hub, a somatic territory normally devoid of Aub and in direct contact with germ cells (Fig. 6B,E). This was further validated by the colocalization of Aub and the hub marker E-Cadherin in the aub^sting testes (Fig. 6B–G). This result suggested that the piRNA pathway regulating cry-Ste also has a somatic requirement. We hence specifically expressed Aub in the hub (upd-Gal4) and found that this indeed rescues the dFmr1-mediated crystal phenotype (Table 1, genotype 8). Thus, the neomorphic behavior of the aub^sting mutation (loss of function in the germline and gain of function in the hub) likely accounts for the allele-specific interaction observed with dFmr1. In line with this finding, dFmr1 is also expressed in the hub (Fig. 6H–J). Thus, the aub^sting allele made it possible to reveal a somatic requirement of the piRNA-mediated cry-Ste regulation in the testis.

We then assessed the tissue-specific requirement of dFmr1 in the piRNA pathway that controls crystal formation. Germline-driven dFmr1 expression rescues the crystal phenotype of heterozygous dFmr1^▵113 testes (Table 1, compare genotypes 21–23) and c135-Gal4-dependent dFmr1 KD triggers crystal formation (Table 1, genotype 24). Thus, dFmr1 works in the male germline, like all the so far identified members of the piRNA pathway controlling the cry–Ste interaction, including Aub [(Bozzetti et al., 2012; Ishizu et al., 2012; Nishida et al., 2007 and references therein]. We noticed that the nos-Gal4 driven dFmr1 KD does not trigger the crystal phenotype, even upon using two doses of the driver line (Table 1, genotypes 25,26). Upon closer inspection of the GFP-driven expression, we believe that the different behavior depends on the much earlier expression of the nos-Gal4 compared to that of the c135-Gal4 driver (supplementary material Fig. S3). This suggests that dFmr1 is required at a specific stage for the piRNA-mediated cry–Ste regulation.

The above data imply that dFmr1 protects from DNA instability in the germline. However, the rescue of the dFmr1-mediated crystal phenotype obtained with somatic Aub expression (aub^sting or with upd-Gal4 driven expression) implies that this piRNA pathway also requires dFmr1 in somatic cells of the testis, directly or indirectly. This is also in line with the observation that ‘somatic’ transposons like ZAM, as well as ‘germline’ transposons are activated in dFmr1 mutant animals (Fig. 2G,H). The formal demonstration that somatic dFmr1 accumulation directly affects the cry–Ste system, however, comes from the following assays: (1) dFmr1 KD in the hub leads to the formation of crystals (Table 1, genotype 27; Fig. 7A), and (2) dFmr1 expression driven by upd-Gal4 rescues the crystal phenotype of dFmr1 testes (Table 1, genotype 28; Fig. 7B). Of note, aub KD in the hub does not trigger crystal formation, further emphasizing the specificity of the dFmr1-mediated phenotype (Table 1, genotype 29). Thus, dFmr1 is required in the hub to suppress the formation of crystals in the male germline and high levels of somatic Aub or dFmr1 cause the recovery of the dFmr1-dependent crystal phenotype. Importantly, even though somatic Aub expression rescues the dFmr1-mediated crystal phenotype, this strictly depends on the endogenous Aub: somatic Aub expression no longer rescues the dFmr1-dependent crystal phenotype if the germline expression of Aub is lowered (Table 1, compare genotypes 28–30; Fig. 7C). Taken together, these results show that dFmr1 as well as Aub are required in the male germline to prevent crystal formation and dFmr1 is also required in the somatic cells of the male gonad. To the best of our knowledge, this is the first example of a case of a gene required in the hub to suppress crystal formation in the germline.

The ability of ectopic Aub expression in the hub to rescue the dFmr1 crystal phenotype led us to extend our analysis of this member of the piRNA pathway to another somatic tissue. Recent studies have documented the expression of piRNAs in the Drosophila nervous system and the occurrence of aub-dependent transposition in memory-relevant neurons in the brain (Lee et al., 2011; Perrat et al., 2013). Furthermore, small RNA and in particular piRNAs are involved in the epigenetic control of synaptic plasticity in Aplysia (Rajasethupathy et al., 2012). We hence turned our attention to this process, which is tightly controlled by FMRP (Bassell and Warren, 2008). The fly larval neuromuscular junction (NMJ) represents the classical Drosophila model for synaptic plasticity and dFmr1 larvae display NMJs that are longer than the wt ones (Zarnescu et al., 2005) (Fig. 7D, compare columns 1,5 and Fig. 7E,I).

Loss of Aub does not affect NMJ morphology (Fig. 7D, columns 1,2 and Fig. 7E,F) and Aub is not detectable in the larval somatic tissues (Fig. 7L), paralleling the fact that Aub is not detected nor required in the somatic tissue of the gonad. Interestingly, however, aub^sting induces ectopic Aub expression in the soma and affects the NMJs, which are extremely and moderately undergrown in homozygous and heterozygous animals, respectively (Fig. 7D, columns 3,4 and Fig. 7G,H,L). The NMJ phenotype was further confirmed by inducing the expression of Aub with the Actin-Gal4 driver (supplementary material Fig. S4 and data not shown). Finally, and most interestingly, aub^sting rescues the dFmr1 phenotype, as seen in double mutant animals (Fig. 7D, columns 4–6 and Fig. 7H–J), showing that somatic Aub expression counteracts the dFmr1 loss of function phenotype, as for the hub (the somatic part of the testis).

The mode of action of Aub and the overall role of the piRNA pathway in the brain is still under intense investigation. Typically Aub is not uniformly distributed in different areas of the brain and so neither is the level of transposition (Lee et al., 2011). To assess whether Aub is expressed below threshold levels in the somatic tissues of the larva and whether its impact onto the NMJ is only detected upon removing both Aub and dFmr1, we analyzed double heterozygous animals and found that the NMJs look identical to the wt ones (data not shown). These data confirm that Aub is neither expressed nor required at the NMJ.

One possible explanation for the above results is that Aub might directly or indirectly impact on the pathway of other AGO proteins acting on synaptic growth. Previous work has shown that AGO1 interacts genetically with dFmr1 at the NMJ, the double mutant aggravating the overgrowth phenotype observed in the dFmr1 mutant animals (Coles et al., 2008). Thus, AGO1 seems to control NMJ growth negatively, as does the forced expression of Aub in the soma. We therefore tested whether the NMJ overgrowth due to the loss of AGO1 is rescued by expressing Aub in the soma and found that the NMJ of AGO1^k08121/aub^sting double mutant animals are significantly shorter than those of AGO1^k08121/+ animals (Fig. 7K). Again, NMJs that lack one dose of Aub and one of AGO1 are similar to those that only lack one dose of AGO1, supporting the hypothesis that Aub does not play a significant role at the NMJ on its own (Fig. 7K).

Here, we show for the first time that dFmr1, a member of the FMRP protein family, is involved in the piRNA pathway controlling the cry–Ste system. This is the first protein that has been found to be required both in the hub and in the germline for the function of this pathway.

dFmr1 is a translational regulator and its role in the miRNA pathway is widely accepted (Cziko et al., 2009; Jin et al., 2004a; Jin et al., 2004b; Zarnescu et al., 2005). We here provide several lines of evidence that dFmr1 can be considered as a ‘bona fide’ member of the piRNA pathway that keeps repetitive sequences and transposons silenced (Brennecke et al., 2007; Vagin et al., 2006). First, dFmr1 mutant testes display crystalline aggregates, as do other mutants of the piRNA pathway (Schmidt et al., 1999; Specchia et al., 2010; (Aravin et al., 2003; Brennecke et al., 2007; Nagao et al., 2010; Nishida et al., 2007). Second, the levels of cry-specific and transposon-specific piRNAs dramatically decrease in dFmr1 mutant testes. Third, as a consequence of this decrease, the Ste RNA is produced and, in addition, transposons are expressed at higher levels than in wt animals. Fourth, dFmr1 mutant animals display fertility defects, a phenotype shown by several mutations affecting the piRNA pathway. The fact that recent screens did not identify dFmr1 as a member of the somatic piRNA pathway could be due to the heterogeneous phenotypes observed with the somatic transposons (present work) and/or to the material used for those assays (Czech et al., 2013; Handler et al., 2013). The cry–Ste system thus proves very efficient for identifying new members of this important pathway.

The movement of transposable elements is one of the molecular causes of DNA instability and sterility (Brennecke et al., 2007; Khurana et al., 2011; Klattenhoff et al., 2007; Peng and Lin, 2013; Tamanini et al., 1997). Considering that human patients mutant for FMRP also display defects in male and female gonads (Sherman, 2000; Tamanini et al., 1997), it will be interesting to characterize the activity of transposons and repetitive sequences in the gonads of mice or humans that are mutant for the FMRP pathway, although there might be no observable defects in mammals because they express three members of the FMRP family versus the single ortholog in fly. Finally, we speculate that mutations affecting the piRNA pathway might also induce gonadal defects in humans.

Until now, the members of the piRNA pathway controlling the cry–Ste interaction, including Aub, have been described as being required in the male germline (Brennecke et al., 2007; Cook et al., 2004; Klattenhoff and Theurkauf, 2008; Nagao et al., 2011; Pane et al., 2007). Surprisingly, the conditional dFmr1 rescue and KD experiments demonstrate that dFmr1 controls the piRNA pathway both in the germline and in the somatic cells of the gonad, which raises questions as to the somatic contribution of other members of the piRNA pathway in the male gonad. The phenotypes induced by somatic Aub expression (Table 1, genotypes 8,18–20) also suggest that the hub expresses one or more AGO proteins that are involved in the somatic piRNA-mediated Ste silencing and that interact with dFmr1; however, the only other protein of the Piwi clade present in the somatic tissue, Piwi, does not participate in Ste silencing (Vagin et al., 2006). Based on preliminary data, we propose that AGO1 might be one such protein. First, AGO1/+ testes display Ste-made crystals (Table 1, genotypes 31–33; Fig. 7M,N), as do testes expressing UAS-AGO1 RNAi driven by the upd-Gal4 driver (Table 1, genotype 34). Second, aub^sting rescues the AGO1-mediated crystal phenotype (Table 1, genotypes 35,36). Third, AGO1 and dFmr1 interact biochemically (Ishizuka et al., 2002) and are known to interact genetically in the ovaries to control germline stem cell maintenance, as well as in the nervous system, where they modulate synaptic plasticity (Bassell and Warren, 2008; Jin et al., 2004b). Taken together, these data suggest that AGO1 contributes to the piRNA pathway that controls the cry–Ste system in the somatic part of the gonad.

The finding that Aub somatic expression affects the NMJ and counteracts the AGO1 loss of function phenotype was also unexpected. Recent work has documented the activation of piRNA pathway in the nervous system [in flies (Perrat et al., 2013), mice (Plaisant et al., 2008), humans (Lukic and Chen, 2011) and molluscs (Rajasethupathy et al., 2012)] and it has been proposed that synaptic plasticity, cognitive functions and neurodegeneration might involve the control of genome stability, even though the precise mode of action and impact of this pathway are not completely understood (Baillie et al., 2011; Lee et al., 2011; Li et al., 2013; Saxena et al., 2012; Tan et al., 2012). Because Aub is not required in the larval somatic tissues, its ectopic expression could affect the NMJ by replacing AGO1 in its known role on the miRNA pathway. However, AGO1 might also affect the NMJ through the piRNA pathway, much in the same way as AGO1 loss of function affects a piRNA pathway in the gonad. Even though AGO1 has been previously described as being exclusively involved in the miRNA pathway, some degree of overlapping between different RNAi pathways has been recently described: (1) the double-stranded-RNA-binding protein Loquacious (Loqs) is involved in the miRNA pathway and in the endogenous siRNA pathway (Kim et al., 2009; Zhou et al., 2009), (2) AGO1 and AGO2 can compete for binding with miRNAs (Laskowitz and Warner, 2008), and (3) ectopic expression of Aub in the soma competes for the siRNAs pathway mediated by AGO2 (Specchia et al., 2008). In addition, miRNAs have been demonstrated to have a role on easi-RNA biogenesis in plants (Creasey et al., 2014). In a similar manner, AGO1 could act on piRNAs through its activity on the miRNA pathway. Although future studies will clarify the connection between AGO1 and the piRNA pathway, the present data provide novel perspectives in the field and could have a broad relevance to diseases affecting cognitive functions.

Our expression, genetic and biochemical data indicate that Aub and dFmr1 interact directly. dFmr1 has been proposed to bind specific cargo RNAs (Bassell and Warren, 2008; Dictenberg et al., 2008; Estes et al., 2008; Siomi et al., 1994; Siomi et al., 1995) and the human FMRP binds small RNA, in addition to mRNAs (Ascano et al., 2012b; Costa et al., 2005; Cziko et al., 2009; Jin et al., 2004b; Yang et al., 2009). Similarly, we speculate that the Aub–dFmr1 interaction allows the targeting of piRNAs to the transcripts of repetitive sequences and transposable elements, dFmr1 providing the molecular link between small RNAs and AGO proteins of the RISC.

The Aub and dFmr1 proteins colocalize and likely interact in the piRNA pathway in a specific stage of testis development and also have additional functions that are independent from each other. Typically, dFmr1 accumulates at high levels in more differentiated cells of the testis, where Aub is not detectable, likely accounting for the axoneme phenotype described in dFmr1 testes (Zhang et al., 2004). In the future, it will be interesting to analyze whether the other genes involved in the piRNA pathway in testis are also required at specific stages, as also recently found in the ovary (Phillips et al., 2012).

Finally, FMRP proteins work in numerous molecular networks, show complex structural features (TUDOR, KH, NLS, NES RGG domains) and are characterized by widespread expression and subcellular localization (cytoplasm, nucleus, axons, dendrites, P bodies), providing versatile platforms that control mRNA and small RNA metabolism (e.g. translation, degradation and transport). Understanding whether FMRP proteins interact with other members of the piRNA pathway, whether this interaction is modulated physiologically and how does the interaction with this pathway compare with that observed with other AGO proteins will clarify the role and mode of action this family of proteins in small RNA biogenesis and metabolism.

The biogenesis of the piRNAs requires two pathways. The primary pathway involves Piwi and predominantly occurs in the somatic tissues (Li et al., 2009; Malone et al., 2009). The ping-pong pathway involves Aub, as well as AGO3, and predominantly occurs in the germline, where Aub is thought to bind an antisense piRNA, to cleave the sense transcript from an active transposon and to produce a sense piRNA that is loaded onto AGO3 (Brennecke et al., 2007; Ishizu et al., 2012; Li et al., 2009; Malone et al., 2009; Nagao et al., 2010; Nishida et al., 2007). The AGO3–piRNA complex binds complementary transcripts from the piRNA cluster, producing the so-called secondary piRNAs by an amplification loop (Brennecke et al., 2007). Although the piRNA pathways have emerged as a very important tool to understand the role of RNA metabolism in physiological and pathological conditions, the relationship and interactions among the involved proteins are not simple to interpret, mostly because not all the players have been characterized. Moreover, recent data support the hypothesis that the somatic and the germline piRNA pathways share components: for example, shut down (shu), vreteno (vret) and armitage (arm) affect primary as well as ping-pong pathways in ovaries (Olivieri et al., 2010; Saito et al., 2009). Our work calls for a role of dFmr1 in both piRNA pathways at least in testes. Based on the alignment of the human, mouse and fly FMRP family members, we speculate that dFmr1 participates to piRNA biogenesis as a Tudor domain (TDRD) containing protein (Maurer-Stroh et al., 2003) (Fig. 7O,P).

TDRDs are regions of about 60 amino acids that were first identified in a Drosophila protein called Tudor (Adams-Cioaba et al., 2010; Ramos et al., 2006; Siomi et al., 1994; Siomi et al., 1995). In the recent years, the requirement of TDRD proteins in piRNA biogenesis and metabolism has become evident (for a review, see Ishizu et al., 2012). Typically, the founding member of the family, Tudor, binds AGO proteins and helps them interact with specific piRNAs (Nishida et al., 2009). Among the different TDRD proteins, fs(1)Yb works in the primary pathway (Qi et al., 2011; Saito et al., 2010; Takata et al., 2008); Krimper, Tejas, Qin/Kumo, and PAPI work in the ping-pong pathway (Lim and Kai, 2007; Liu et al., 2011; Patil and Kai, 2010); and Vret works in both systems (Handler et al., 2011; Zamparini et al., 2011). PAPI, the only TDRD protein that has a modular structure closely related to dFmr1 (two KH domains and one TDRD), interacts with the di-methylated arginine residues of AGO3 and controls the ping-pong cycle in the nuage. At least during the early stages of testis development, dFmr1 might interact with Aub in a similar way. Given that TDRDs are involved in the interactions between proteins and in the formation of ribonucleoprotein complexes, future studies will assess whether RNAs mediate the Aub–dFmr1 interaction.

In conclusion, the discovery of dFmr1 as a player in the piRNA pathway highlights the importance of the fly model. Our data also add a new perspective to understanding the role and mode of action of this protein family and the physiopathological mechanisms underlying the Fragile X syndrome.

Flies were raised at 25°C on standard food. Or-R and w¹¹¹⁸ flies were used as controls. All the stocks are described in FlyBase, as well as the expression profile of the drivers. They are listed in supplementary material Table S2.

We generated the UAS-aub vector and the transgenic line, which contains the complete AF 145680 coding sequence of the aub gene. We obtained UAS-aub, dFmr1^▵113 flies by recombination.

Testes were dissected in Ringer's modified solution (182 mM KCl, 46 mM NaCl, 3 mM CaCl₂, 10 mM Tris-HCl pH 7.5). Testes were fixed with methanol, washed with PBST (1× PBS, 1% Triton X-100, 0.5% acetic acid) for 15 min, washed with 1× PBS for 5 min three times and incubated with the polyclonal mouse anti-Ste antibody (1:100) (Bozzetti et al., 1995). Samples were washed with 1× PBS for 5 min three times, incubated 2 h with 1:100 FITC-conjugated anti-mouse-IgG antibody (Jackson) and examined by epifluorescence microscopy (Nikon-Optiphot 2). Whenever necessary, DAPI was used at 100 ng/ml for nuclear labeling.

An optimized protocol to detect small RNA analysis was employed (Pall and Hamilton, 2008). Total RNA extracted from freshly dissected testis cells was fractionated on 10% poly-acrylamide gel in MOPS–NaOH (pH 7), 7 M Urea and transferred onto Amersham Hybond-NX nylon membrane (GE Healthcare).

RNA cross-linking was performed, as described previously (Pall and Hamilton, 2008). DNA oligonucleotides complementary to the sequence of piRNAs crystal, rasi4 and R1 and to 5S rRNA were ³²P-labeled and used as probes. The sequences of the oligonucleotides are: crystal, 5′-UCGGGCUUGUUCUACGACGAUGAGA-3′; rasi4, 5′-CGGUGUUCGACAGUUCCUCGGG-3′; and R1, 5′-CUCGGAGUCCGACUCAGGAGAUCGGA-3′.

The crystal piRNA corresponds to the that identified as Su(Ste)-4 previously (Nishida et al., 2007); the rasi4 piRNA corresponds to one of the most abundant crystal-related piRNA used previously (Specchia et al., 2010). The R1 piRNA correspond to one of the piRNA related to R1 identified previously (Nishida et al., 2007).

Total RNA was extracted from 30 mg of male or female gonadal tissues using the RNAqueos-4 PCR Kit (AMBION) reagent following the manufacturer's protocol as described previously (Specchia et al., 2010). Samples were incubated with DNase I RNase free (AMBION) (1 Umg⁻¹ RNA) at 37°C for 30 min (100 μl). DNase-treated RNA was precipitated at −80°C overnight and after centrifugation (10,000 g for 15 min) it was dissolved in 50 μl of nuclease-free water. The RNA concentration and purity were determined photometrically.

5 μg of total RNA were used as a template for oligonucleotide dT-primed reverse transcription using SuperScriptIII RNaseH-reverse transcriptase (Invitrogen), according to manufacturer's instructions. qRT-PCR was performed in the SmartCycler Real-time PCR (Cepheid) using SYBR green (Celbio) according to the manufacturer's protocol. For quantification of the transcripts, we used the 2ΔΔC_t method as reported previously (Specchia et al., 2010). Primers are reported in supplementary material Table S3.

One young male was mated to three control virgin females, and three virgin females were mated with three control males. Ten individual males and 30 females were tested for each genotype. After 4 days, the crosses were transferred to a fresh vial. The parental flies were removed from the last vial after an additional 4 days. The number of the adult progeny from each vial was counted.

For comparisons between two measurements a two-tailed Student's t-test was used to show statistical significance.

S2 cells were cultured in Schneider cell medium (Gibco BRL) plus 10% fetal calf serum. They were grown in monolayers at room temperature and transiently cotransfected with pPAC-dFmr1 (Schenck et al., 2002) pPAC-Gal4 and pUAS-aub vectors using Effectene reagent (Qiagen). S2 cells were scraped in Schneider cell medium and washed two times with 1× PBS. The pellet was resuspended in buffer A (400 mM KCl, 50 mM Tris-HCl pH 7.5, 20% glycerol, 5 mM DTT, 1 mM PMSF, 0.1% NP40 and 1× protease inhibitor cocktail) and cells were lysed by freeze and thaw three or four times. After adding three volumes of buffer A without KCl, the lysate was centrifuged at 12,000 g for 5 min. The precleaned supernatant (INPUT) was incubated with mouse anti-dFmr1 antibody [5A11, Developmental Studies Hybridoma Bank (DSHB)] for 2 h at 4°C and then with Protein-G–sepharose (GE Healthcare) overnight at 4°C. Beads were pelleted and the supernatant (SN) was saved for western blotting. After three washes in RIPA buffer, the beads were added to 1× Laemmli's loading buffer (50 mM Tris-HCl pH 6.8, 2% SDS, 100 mM dithiothreitol (DTT), 10% glycerol and 0.1% Bromophenol Blue]. The samples were boiled, pelleted and run on a 10% SDS-polyacrylamide gel. The gel was transferred at 250 mA onto a nitrocellulose membrane (Amersham). After blocking for 1 h in TBST (800 mM NaCl, 100 mM Tris-HCl pH 7.5, 0.1% Tween 20) with 5% dried milk powder, the membrane was incubated with rabbit anti-Aub antibody [Abcam or a gift from G. Hannon, Cold Spring Harbor Laboratory (CSH), New York] diluted 1:1000 in TBST plus 5% dried milk powder for 1 h at room temperature or overnight at 4°C. The membrane was rinsed three times for 5 min in TBST followed by incubation for 1 h at room temperature with horseradish peroxidase (HRP)-conjugated anti-rabbit-IgG antibody (1:5000, Jackson). For western blot analysis, protein extracts were added to 2× Laemmli's loading buffer, boiled and run in a 10% SDS-polyacrylamide gel. For loading control, we used mouse anti-β-tubulin antibody (1:4000).

For gonadal extracts, testes were dissected in cold Ringer's solution (130 mM NaCl, 4.7 mM KCl, 1.9 mM CaCl₂, 10 mM Hepes pH 6.9) and mashed in lysis solution (50 mM NaCl, 50 mM Tris-HCl pH 7.5, 320 mM sucrose, 10% glycerol, 1% Triton X-100, and 1× protease inhibitor cocktail. They were centrifuged at 12,000 g for 10 min at 4°C. The amount of total protein was determined by Bradford assay.

For larval extracts, crude protein was extracted from carcasses of third-instar larval stage, in buffer (50 mM NaCl, 50 mM Tris-HCl pH 7.5, 10% glycerol, 320 mM sucrose, 1% Triton-X, Roche^® protease inhibitor cocktail), from S2 cells in buffer (100 mM KCl, 20% glycerol, 50 mM Tris-HCl pH 7.5, 5 mM DTT, 1 mM PMSF, 0.1% NP40, Roche^® protease inhibitor cocktail). After homogenization the samples were kept on ice for 10 min and centrifuged at 12,000 g for 10 min. Equal amounts of protein were loaded for SDS-PAGE on 8% gels. Each western blot was repeated three times.

Drosophila gonads were dissected in Ringer’s solution and fixed with 4% paraformaldehyde for 20 min, washed with PBT (1× PBS with 0.5% Triton X-100), blocked with 5% NGS for 1 h and incubated with mouse anti-dFmr1 (1:300, DSHB), rabbit anti-Aub (1:300, gift from Brennecke), rabbit anti-Vasa (Santa Cruz Biotechnology), mouse anti-αSpectrin (DSHB), rat anti-E-Cadherin DHSB (1:100) antibodies at 4°C overnight. Samples were washed with PBST and then incubated for 2 h with the secondary antibodies against mouse IgG, rabbit IgG or goat IgG conjugated to FITC, Cy3 or Alexa Fluor® dyes (1:500, Jackson). All the samples were examined and captured using a laser-scanning confocal microscope (Zeiss LSM 700 on Axio imager M2).

The Aub coding sequence was cloned in frame, using the following primers: 5′-ATGAATTTACCACCAAACCCTGT-3′, 5′-GTTAGCTGCGGCGTTGTC-3′, 5′-GAACGCCTTCGCATGTTC-3′ and 5′-TTACAAAAAGTACAATTGATTCTGCAG-3′. Luciferase was used as negative control of the dFmr1 interaction and CYFIP as the positive control (Schenck et al., 2003). All are in vitro translated from the T3 and T7 promoters while radiolabeled with ³⁵S. GST tagged dFmr1 and GST alone overexpression was induced by 0.5 mM IPTG, 2 h at 37°C. Protein was purified as described before (de Santa Barbara et al., 1998). Purified GST and GST-dFmr1 were pulled down with in vitro translated proteins in 200mM NaCl containing buffer.

Dissection, antibody labeling of third-instar larvae and quantification of NMJ phenotypes were performed as described previously (Schenck et al., 2003). At least two rounds of independent assays were performed per genotype; for each round at least ten larvae of normal body size were dissected and 30 NMJs were analyzed for each assay. The synapse type-Ib on muscle 4, segments A2–A4 was analyzed. Pictures were taken at 40× with a of Zeiss Axiophot 2 microscope and imported in the in-house developed TCS/timt software that quantified synaptic length by automatic measurement. Statistical significance was calculated using ANOVA and the Newman-Keuls method for post hoc pair-wise analyses.

We thank J. Brennecke and G. Hannon for providing anti-Aub antibody; M. Van Doren and Y. Zhang for Drosophila strains; the Developmental Studies Hybridoma Bank and FlyBase for material and information. We thank O. F. Karatas and E. Green for sharing unpublished data as well as C. Delaporte and N. Arbogast for technical help.