UAP56 Couples piRNA Clusters to the Perinuclear Transposon Silencing Machinery

UAP56 co-localization with the piRNA biogenesis machinery

Drosophila piRNA pathway mutations lead to germline DNA damage and activation of a signaling cascade that blocks asymmetric RNA localization in the developing oocyte (reviewed by Khurana and Theurkauf, 2010). UAP56 has a conserved function in splicing and mRNA export, and strong hypomorphic mutations are lethal (Gatfield et al., 2001; MacMorris et al., 2003; Shen, 2009). However, Drosophila females carrying the uap56^sz15′ allele (an E245K substitution) in trans to hypomorphic UTR deletions are sterile and show remarkably specific defects in asymmetric RNA localization (Eberl et al., 1997; Meignin and Davis, 2008), raising the possibility that UAP56 has a germline function in piRNA production and transposon silencing.

To explore a potential link to the piRNA pathway, we determined the subcellular localization of UAP56 relative to the piRNA cluster associated HP1 variant Rhino (Rhi) and the nuage component Vas (Figure 1). Immunofluorescence labeling and confocal imaging revealed UAP56 diffusely localized throughout germline and somatic nuclei, but also concentrated in distinct nuclear foci in the nurse cells, which produce most of the maternal mRNAs that are deposited in the oocyte (Figure 1A). Strikingly, 98.4% of these foci, defined by UAP56 signal above background, co-localized with Rhi foci. Furthermore, 99.1% of Rhi signal above background co-localized with UAP56 foci (Figure 1A and B; Pearson’s correlation coefficient R = 0.99). In the transcriptionally silent oocyte nucleus, however, Rhi localized to foci but UAP56 was dispersed (Figure 1D). In rhi²/rhi^KG mutant ovaries, which do not express detectable Rhi protein, UAP56 was dispersed throughout the nucleus and did not form foci. Rhi is therefore required for UAP localization to nuclear foci, which may also require transcription.

We next assayed Rhi localization to nuclear foci in uap56 mutants. The uap56^sz15′ allele, when combined with hypomorphic 5’ UTR deletion alleles, is viable but sterile, and leads to spindle class oocyte patterning defects that are characteristic of piRNA pathway mutations (Meignin and Davis, 2008). The point mutation in the uap56^sz15′’ allele does not alter protein expression (Meignin and Davis, 2008), and the uap56²⁸ 5′UTR deletion allele does not change protein structure, but appears to reduce expression (Figure S2A; (Meignin and Davis, 2008). The uap56^sz15′/uap56²⁸ combination thus generates mutant ovaries in which most of the UAP56 is likely to carry the E245K substitution. In these ovaries, UAP56 was dispersed throughout the nurse cell nuclei. By contrast, Rhi was present in distinct foci during early oogenesis (Figure 2A). In later stage egg chambers, however, Rhi was dispersed in the nurse cell nuclei (Figure S2B). Low levels of wild type UAP56, produced by the uap56²⁸ allele, could support Rhi localization during early oogenesis. Alternatively, UAP56 could be required to maintain Rhi localization during later stages of oogenesis, but dispensable for localization early. We favor the former possibility, and speculate that UAP56 and Rhi are co-dependent for localization to nuclear foci.

To directly determine if the E245K substitution alters the function and subcellular localization of UAP56, we generated transgenes expressing wild type uap56 (UAP56-Venus) or the uap56^sz15′ mutant allele (sz-Venus) fused to the fluorescent protein Venus. Both transgenes were driven by the uap56 promoter and integrated into the same chromosomal location, and western blots showed that both fusion proteins were expressed at comparable levels (Figure S2A). The wild type UAP56-Venus fusion rescued the lethality associated with both uap56²⁸/Df and uap56^sz15′/Df combinations, and 75% of the embryos produced by uap56^sz15′/Df females expressing this fusion showed normal dorsal ventral patterning, and a small fraction of these embryos hatched (Table S1). The failure to restore full fertility is likely due to the Venus fusion, as a genomic fragment expressing untagged protein fully rescues the mutant phenotype (Meignin and Davis, 2008). By contrast, the sz-Venus fusion failed to rescue viability in either mutant background (Table S1). To assay fusion protein localization in the ovary, we expressed both transgenes in the viable uap56^sz15′/uap56²⁸ trans-heterozygous background and labeled for Rhi. Wild type UAP56-Venus precisely co-localized with Rhi foci in early egg chambers (Figure 2B),and this transgene rescued Rhi localization to foci during later stages (data not shown). By contrast, the sz-Venus fusion was dispersed in the germline nuclei, did not co-localize with Rhi (Figure 2B), and failed to rescue Rhi foci in late stage egg chambers. The E245K substitution thus disrupts UAP56 co-localization with Rhi and Rhi focus stability.

UAP56 is required for nuage localization of piRNA pathway proteins

Our subcellular localization studies also showed that prominent UAP56-Rhi foci were closely associated with the nuclear periphery, and that nuage foci containing a functional GFP-Vasa fusion appeared to localize directly across the nuclear envelope (Figure 1A). Labeling for Rhi, Vas and nucleoporins (Figure S1A) showed that the pores lie between the nuclear foci containing Rhi and the nuage foci containing Vasa, and that a significant fraction of Vasa signal overlaps with signal for the nuclear pores (Figure S1B). To quantify the relationship between Rhi foci in the nucleus and Vasa foci in nuage, we identified all Rhi foci that were above threshold and within the confocal resolution limit of nuclear pores (the signals overlapped), and for each of these foci, determined if a focus of GFP-Vas was present at the adjacent nuclear periphery. 98.5% of Rhi foci near the pores had an associated focus of Vasa at the cytoplasmic face of the nucleus. We then quantified Rhi in the nuclear foci and Vasa levels in the adjacent nuage foci. Strikingly, the level of Vas was highly correlated with the level of Rhi (Figure 1C; R=0.72, p<2.2e⁻¹⁶). Furthermore, Vasa localization to nuage was disrupted in uap56^sz15′/uap56²⁸ mutants (Figure S3A). By contrast, and Rhi and UAP56 co-localized to distinct nuclear foci in vas mutants (Figure S2C). Rhi and UAP56 thus appear to function at clusters, in a process that is upstream of Vas localization to nuage.

The PIWI proteins Aub and Ago3 localize with Vas in nuage, where they appear to drive piRNA amplification. The founding member of the PIWI family, Piwi, is concentrated in nuclei (reviewed by Klattenhoff and Theurkauf, 2008). The uap56^sz15′/uap56²⁸ transheterozygous combination did not alter Piwi localization to nuclei, but disrupted Aub and Ago3 localization to nuage (Figure 3A). Expression of wild type UAP56-Venus in the uap56^sz15′/uap56²⁸ background partially restored nuage localization of Aub and fully restored nuage localization of Ago3 (Figure 3B). By contrast, transgenic expression of sz-Venus failed to restore nuage localization of either protein (Figure 3B). Aub and Vas localize to the posterior pole of the oocyte, where they are incorporated into pole plasm. Posterior localization of both proteins is disrupted in uap56^sz15′/uap56²⁸ mutants (Meignin and Davis, 2008) (Figure S3B and C). Posterior localization of Aub was restored in mutants expressing the wild type UAP56-Venus transgene, but not in mutants expressing the sz-Venus fusion (Figure S3C). We did not assay Vas localization in these transgenic ovaries, but Aub localization to pole plasm requires Vas (Harris and Macdonald, 2001), suggesting that Vas localization is also restored. The E245K substitution thus disrupts the organization of nuclear and cytoplasmic components of the piRNA biogenesis machinery.

Germline DNA damage

piRNA pathway mutations lead to germline DNA damage, which is proposed to result from transposon mobilization (reviewed by Klattenhoff and Theurkauf, 2008). γH2Av is a phosphorylated histone variant that accumulates at DNA break sites (Madigan et al., 2002). In wild type egg chambers, γH2Av foci are present in region 2 of the germarium, where meiotic breaks form, and in later nurse cell nuclei undergoing endoreduplication (Jang et al., 2003). Stage 2 and later oocyte nuclei, by contrast, are consistently negative for γH2Av foci. In uap56^sz15′/uap56²⁸ mutants, γH2Av foci persist in oocyte nuclei and appear to be enhanced in the nurse cells (Figure S4A). In uap56^sz15′/uap56²⁸ mutants expressing the wild type UAP56-Venus transgene, by contrast, γH2Av foci were not detected in later stage oocytes (Figure S4A). UAP56 is therefore required to maintain germline genome integrity.

Gene and transposon expression

UAP56 is required for splicing and export of a broad spectrum of mRNAs (Gatfield et al., 2001; Shen, 2009), but our phenotypic and localization data raised the possibility that UAP56 also functions with Vas and Rhi to control transposon activity. We therefore used whole genome tiling arrays to assay gene and transposon expression in uap56^sz15′/uap56²⁸ and vas mutant ovaries, and compared these data with an earlier analysis of rhi mutants (Klattenhoff et al., 2009b). The vasa intronic gene (vig) is contained in a vas intron, and Vig co-purifies with FMRP and components of the siRNA machinery (Caudy et al., 2002), suggesting that it functions in RNA silencing. To specifically disrupt vas, we therefore analyzed a null deletion allele that removes both vas and vig in trans to a point mutation in vas that does not disrupt vig. The Genome Brower screen shot in Figure 4 shows expression of the retrotransposon Blood and neighboring Ent1 gene, which contains three introns, in control, uap56, rhi and vas mutants. In all three mutants, Blood is over-expressed and Ent1 is expressed at levels comparable to the control (Figure 4A). This pattern extends across the genome, as no protein-coding gene showed a statistically significant difference from wild type in uap56, vas or rhi mutants (FDR<0.05) (Figure 4C). By contrast, all three mutations disrupted transposon silencing, with 11 transposon families showing significant over-expression in uap56, rhi, and vas (FDR<0.05; Figure 4B, red points; Venn diagram, FigureS4D). RT-qPCR and northern blotting confirmed that expression of a control protein coding gene (RP49) did not change and that two transposons (Het-A and Blood) were over-expressed (Figure S4B and C). The uap56^sz15′/uap56²⁸ allelic combination, and mutations in vas and rhi, thus disrupt transposon silencing but do not alter gene expression.

piRNA production

To determine if the uap56^sz15′/uap56²⁸ combination blocks production of piRNAs, we deeply sequenced small RNAs from mutant ovaries, which revealed a significant reduction in total piRNAs (Figure S5A and B). piRNAs from opposite strands that overlap by 10nt are characteristic of ping-pong amplification. The bias toward 10nt overlap was reduced in uap56 ^sz15′/uap56²⁸, rhi and vas mutant ovaries (Figure S5C, D and E). The scatter plots in Figure 5A show piRNAs mapping to group 1, 2 and 3 transposons in uap56 ^sz15′/uap56²⁸, vas and rhi mutants compared with wild type controls. In all thee genotypes, piRNAs specific to group 1 transposable elements, which appear to be expressed primarily in the germline, are significantly reduced (Figure 5A, black points). By contrast, the levels of piRNAs linked to group 3 transposons, which are expressed primarily in the somatic follicle cells, were comparable to wild type (Figure 5A, red points). Vas and Rhi both appear to be specific to the germline, consistent with changes in germline piRNAs. However, UAP56 is expressed in both the germline and somatic follicle cells. The uap56²⁸/uap56^sz15′ allelic combination thus appears to disrupt a germline specific branch of the piRNA biogenesis pathway.

piRNA clusters that produce uniquely mapping piRNAs from both genomic strands dominate in the germline, while the major somatic piRNA cluster, flam, and cluster 2 produce unique piRNAs almost exclusively from one genomic strand (Brennecke et al., 2007; Gunawardane et al., 2007). Mutations in rhi nearly eliminate piRNAs from dual-strand clusters, but do not block piRNA production by flam or cluster 2 (Klattenhoff et al., 2009b). The uap56²⁸/uap56^sz15′ allelic combination produces an essentially identical reduction in dual strand cluster expression (Figure 5B). Mutations in vas also block expression of dual-strand clusters, and do not alter flam expression. However, vas mutations also reduce piRNAs linked to cluster 2, which appears to be expressed in the germline and soma (Figure 5B; (Malone et al., 2009). To directly compare cluster expression in the three mutant backgrounds, we generated pair-wise scatter plots covering the top 50 clusters (Figure 5C). This analysis showed that uap56^sz15′ / uap56²⁸ and rhi²/rhi^KG produce nearly identical changes in cluster expression (R=0.96), and that very similar changes are produced by vas mutants (R=0.83 for both vas-uap56 and vas-rhi comparisons; Figure 5C). UAP56, Rhi and Vas thus represent nuclear and nuage components of a dual-strand cluster expression system.

The majority of piRNAs in the Drosophila germline are derived from transposons and other repeats, but a subset map to gene transcripts. Chromosomal profiles of uniquely mapping 23 to 29 nt RNAs in uap56^sz15′ / uap56²⁸ mutants revealed an increase in small RNAs from euchromatic sites (data not shown). Consistent with this observations, the scatter plots in Figure 5D show that 23 to 30 nt genic RNAs increase significantly in both rhi and uap56 mutants. By contrast, vas produced a modest decrease in these genic RNAs. These data were normalized to sequencing depth, and reduced transposon mapping piRNAs could therefore artificially inflate the genic piRNA pool. We therefore normalized to total miRNAs and repeated the analysis, which confirmed the increase in genic 23 to 30 nt RNAs in uap56 and rhi mutants (Figure S5I-K). Most of these species are derived from 3’ UTRs (Figure S5L). This 3’UTR bias is also present in wild type and vas mutants, but the high signal in uap56 and rhi compresses profiles and obscures this bias.

In wild type ovaries, genic small RNAs show a peak at 21nt, which is characteristic of endo-siRNAs, and a peak at 26nt, which is consistent with piRNAs bound to Piwi (Brennecke et al., 2007; Figure S5F). In uap56^sz15′ / uap56²⁸ mutants, by contrast, the length distribution was much broader, with a shoulder between 20–22nt (Figure S5G). Some of these species could be non-specific mRNA breakdown products. Following binding to Piwi proteins, piRNAs are modified by the addition of a 2’-O-methyl group at the 3’ terminus (Horwich et al., 2007; Saito et al., 2007), which renders these species resistant to oxidation. To estimate the fraction of genic small RNAs that are bound to Piwi proteins, we therefore deep sequenced small RNAs after oxidation. These studies revealed a significant increase in oxidation-resistant genic small RNAs in uap56 mutants, with a length distribution characteristic of Piwi binding (Figure S5H). Mutations in uap56 thus reduce cluster piRNAs and increase genic piRNAs, suggesting that the specificity of processing has been compromised.

piRNA precursor binding

Clusters are proposed to produce long precursor RNAs that are processed into primary piRNAs in the nuage. Strand specific RT-qPCR showed that the uap56^sz15′ / uap56²⁸ combination significantly reduced the steady state level of plus and minus strand RNAs from two regions of the major 42AB cluster (Figure S6A), but did not reduce precursor RNAs from uni-strand cluster 2 or flam (Figure S6B). Mutations in rhi produce similar changes in cluster transcript abundance (Klattenhoff et al., 2009b). DEAD box proteins can function as ATP-dependent RNA clamps (Linder and Jankowsky, 2011), and we speculated that UAP56 binds and stabilizes transcripts derived from germline clusters. To test this hypothesis, we immunnoprecipitated wild type UAP56-Venus and sz-Venus fusion proteins from ovary extracts and assayed RNA binding by random primed strand-specific deep sequencing. To quantify enrichment, genome mapping fragments in the input and precipitated fractions were normalized to noncoding RNAs (mainly ribosomal RNAs), which show only background binding to the beads.

Strikingly, RNAs from the major germline piRNA cluster at 42AB were the most highly enriched species in the UAP56-Venus precipitated fraction (34 fold: Figure 6 and Figure S6G, orange point), and piRNAs from 42AB are nearly eliminated by the uap56 mutation (Figure 6). Transcripts from flanking protein coding gene show relatively modest enrichment, and transcripts from the somatic flamenco cluster were not enriched (Figure S6G, blue point). RT-qPCR confirmed that transcripts from cluster 1/42AB are enriched in the precipitated fraction, while transcripts from cluster 2, flam are not (Figure S6E). Across the transcriptome, RNAs from the top 40 piRNA producing clusters showed a 7.9-fold enrichment, and species from all 142 clusters identified by Brennecke et al. (2007) showed an average 5.2 fold enrichment. Protein coding genes, by contrast, showed no average enrichment (Figure S6H). The uap56^sz15′ mutation caused a modest, but statistically significant, reduction in RNA binding for both cluster transcripts and mRNAs (Figure 6A, S6E, G and H). By contrast, this substitution does not significantly alter gene expression (Figure 4C), but nearly eliminates piRNA production by germline clusters (Figure 5B) and blocks UAP56 co-localization with Rhi (Figure 2). The E245K substitution thus appears to specifically block an interaction between UAP56 and the piRNA biogenesis machinery.

Primary piRNAs show a significant strand bias relative to the embedded transposon fragments that make up clusters, with 60% +/− 24% accumulating anti-sense to the mobile elements (Brennecke et al., 2007). The longer RNAs that immunoprecipitate with UAP56, by contrast, do not show strand bias relative to the transposons (47% +/− 23% antisense). For example, the region of the 42AB cluster composed of GATE transposon fragments in Figure 6B shows typical anti-sense piRNA strand bias, but no clear bias in UAP56-bound RNAs. This difference in strand bias is highly significant (p-value = 1.123e-06 by Wilcoxon signed rank test (paired)). Strand bias in the piRNA pool is proposed to originate from post-transcriptional ping-pong amplification of primary piRNAs (Brennecke et al., 2007). We therefore speculate that UAP56 binds to precursor RNAs that have not entered the piRNA amplification cycle.

The DEAD box protein Vas localizes across the nuclear envelope from UAP56 and is required for production of piRNAs from germline clusters. To determine if cluster transcripts associate with Vas, we immunoprecipitated a functional Vas-GFP fusion from wild type ovaries (Johnstone and Lasko, 2004) and assayed for cluster transcripts by RT-qPCR (Brennecke et al., 2007; Klattenhoff et al., 2009). Anti-GFP immunoprecipitation from the parental strain was used as a negative control. As shown in Figure S6F, RNAs from the 42AB cluster, but not from flam or uni-strand cluster 2, were significantly enriched by Vas-GFP immunoprecipitation. The DEAD box is an ATP dependent RNA binding domain, suggesting that this interaction is direct. Steady state cluster transcript levels were not reduced in vas mutants, and may be somewhat elevated (Figure S6C, D). These observations suggest that Vas binds cluster transcripts as they exit the nucleus, and deliver these RNAs to the piRNA biogenesis machinery.

The E245K substitution disrupts a conserved surface residue

DEAD box proteins are composed of N- and C-terminal domains joined by a linker, and the ATP and RNA binding sites span these domains (reviewed by (Linder and Jankowsky, 2011). In the ATP-bound “closed” state, these domains form an RNA clamp. The studies described here indicate that the E245K substitution in uap56^sz15′ defines a domain that is critical to piRNA production and transposon silencing. To define the location of this substitution on UAP56, we threaded the Drosophila sequence into the crystal structure of the ATP bound form of human UAP56 (Shi et al., 2004; Zhao et al., 2004). This modeling places E245 at the protein surface, near the linker between the two UAP56 domains (Figure 7A). This residue appears to form a salt bridge between the end of an α-helix and β-sheet, and a structurally analogous salt bridge appears to stabilize the closed ATP and RNA binding conformation in the related DEAD box protein Dpb5(Montpetit et al., 2011). This bridge is disrupted in the open form, which is engaged with the nuclear pore and the Gle1 activator and dose not bind RNA (Montpetit et al., 2011). The E245K substitution in uap56^sz15′ may therefore contribute to the observed reduction in mRNA and cluster transcript binding (Figure S6E, G and H). By contrast, this substitution nearly eliminates piRNA production and co-localization with Rhi, but does not disrupt steady state mRNA expression (Figure 2, 5B and 4C). Reduced RNA binding cannot explain this specificity. We therefore propose that this surface residue lies in a domain of UAP56 that interacts with the piRNA machinery, but is dispensable for mRNA export and processing. This residue is conserved from humans to fission yeast (Figure S7), raising the possibility that this domain mediates interactions with the small RNA silencing machinery in other systems.

Fly stocks and transgenic lines

All the stocks and crosses were raised at 25°C on cornmeal medium using standard conditions. OregonR (OreR) and w¹¹¹⁸ were used as control strains. The uap56^sz15′/CyO, uap56²⁸/CyO and UASp-UAP56-GFP stocks were obtained from Ilan Davis (University of Oxford). uap56^Df/CyO, vas^D5/CyO, vas^PH/CyO, rhi² (rhi⁰²⁰⁸⁶)/CyO and rhi^KG (rhi^KG00910) /CyO were obtained from Bloomington Stock Center.

The UASp-UAP56-GFP transgene was constructed using a full length UAP56 cDNA from the Gold collection (BDGP23644). Coding sequences were recovered by PCR amplification using a sense primer (5′-GGTACCATGGCCGACAATGACGATC-3′) that spanned the translation initiation codon, and an antisense primer that introduced a unique Nhe1 restriction site upstream of the stop codon (5′-GCTAGCGCGTCCCTCAATGAATGTAG-3′). The PCR product was cloned into pGemT (Promega) and sequenced. 2 silent mutations at position 22 (GAA to GAG) and 87 (TCC to TCT) of the protein were detected. The UAP56 cDNA was fused to the N-terminus of GFP (from vector KS-GFP kindly provided by A. Vincent). A KpnI-BamHI fragment encoding the UAP56-GFP fusion was cloned downstream of the promoter in the UASp transformation vector (Rorth, 1998). Transgenic lines were obtained in w1 background, and transgene expression was induced using the germline-specific nanos Gal4 vp16 driver.

The UAP56-Venus transgene was engineered as follows: The UAP56 promoter and 5’ UTR (695bp) were cloned upstream of a full-length cDNA, which was fused in frame to a multifunction affinity purification tag followed by Venus. (Ma et al. 2012, Mol. Cell Proteomics, in press). The fusion was followed by uap56 3’ UTR (1255bp). Site directed mutagenesis was used to generate a G to A substitution at position 733 of the uap56 coding sequence, producing the E245K substitution found in the uap56^sz15′ allele. The wild type and uap56^sz15′ fusion genes were introduced into the attB vector, and independent transformants were obtained by site-direct phiC31 integration into chromosomal locus 68A4 (Ni et al., 2008).

Immunohistochemistry

Ovaries were immunolabeled using the Buffer A staining protocol, as described earlier (Klattenhoff et al., 2009b). The following primary antibodies and dilutions were used: rabbit anti-Aub (1:1000)(Brennecke et al., 2007), rabbit anti-Ago3 (1:250)(Li et al., 2009), rabbit anti-Piwi (1:1000), rabbit anti-UAP56 (1:1000)(Eberl et al., 1997), guinea pig anti-Rhi (1:500)(Klattenhoff et al., 2009a), and rabbit anti-Vas (1:5000)(Liang et al., 1994). Anti-UAP56 and anti-Vas were gifts from P. Lasko. A rabbit polyclonal antibody against γ-H2Av (1:500, Rockland) was used to detect DNA DSBs. A mouse monoclonal antibody against nuclear pore complex (1:1000, Covance) was used to label nuclear envelope. TOTO-3 dye (Molecular Probes) was used at 1:500 to label DNA after RNase One (1:500, Promega) treatment.

Total RNA isolation and array analyses

Total RNA was extracted from 2–4 day old ovaries from OreR, uap56^sz15′/uap56²⁸, rhi²/rhi^KG and vas^D5/vas^PH females using the RNeasy Kit (Qiagen) and the manufacturer’s instructions. Tilling array analysis was performed as described previously (Klattenhoff et al., 2009b). Data can be accessed through GEO reference series GSE35638.

Small RNA sequencing

Total RNA was extracted from ovaries dissected from 2–4 day old females using MirVana kit (Ambion). 18–29 nt small RNAs were gel purified following 2S rRNA depletion and libraries were prepared as described previously (Klattenhoff et al., 2009b). Sequencing was performed using a Solexa Genome Analyzer (Illumina, San Diego, CA). Small RNA sequence analysis was performed as described previously(Klattenhoff et al., 2009b; Li et al., 2009). Data can be accessed through GEO reference series GSE35638.

Strand-specific Reverse Transcriptase qPCR

Strand-specific RT-qPCR for cluster transcripts was performed as described previously (Klattenhoff et al., 2009b). Signals were normalized to 18s rRNA after subtracting no RT primer background.

RNA Immunoprecipitation and Strand-specific RNA Sequencing

Whole ovaries were dissected from 2–4 day old flies in 1×PBS Buffer and homogenized on ice in Lysis Buffer (50mM Tris-HCl, pH7.5, 150mM NaCl, 1mM EDTA, 0.5% NP-40) with 40U/ml RNasin Plus (Promega N261), 1× Proteinase Inhibitor Cocktail (Sigma) and 1mM PMSF. The lysate was sonicated three times for 5 seconds each at 35% amplitude using a Diagenode Bioruptor, and then centrifuged at 14,000rpm for 15min at 4°C in a table top microfuge. 10% of the supernatants were saved as input samples and the rest were incubated with GFP Trap-A beads (Chromotek) for 2hr at 4°C. The beads were separated from the supernatants after incubation and washed three times with Lysis Buffer at room temperature. UAP56-Venus and sz-Venus were immunoprecipated by ANTI-FLAG^® M2 Affinity Gel (Sigma A2220) and eluted with 200µg/ml FLAG^® Peptide (Sigma F3290). Three biological replicates were performed for each IP. Total RNA was extracted from the input lysate and the beads using RNeasy Micro Kit (Qiagen) using the manufacturer’s protocol. Strand-specific RNA sequencing was performed after depletion of ribosomal RNAs using Ribo-Zero (Epicenter Inc) by a modification of the procedure of Parkhomchuk et al. (Parkhomchuk et al., 2009), described in detail elsewhere (Z. Zhang and P. D. Zamore, manuscript in preparation). Data can be accessed through GEO reference series GSE35638.

Quantification of transposon transcripts levels by RT-qPCR

Total RNA was prepared from whole ovaries using RNeasy Mini Kit (Qiagen#74104). Oligo(dT)₂₀primer was mixed with an 18s rRNA-specific control primer and first strand cDNAs were synthesized using SuperScript™III reverse transcriptase (Invitrogen#18080-093) following the manufacture's protocol. The resulting cDNAs were used as templates for quantitative real-time PCR using the primers indicated (Supplemental Table 2). qPCR reactions were performed using a StepOnePlus™ System (Applied Biosystems) and SYBR Green I (Qiagen#204145). The expression level of transcripts was measured relative to the 18s rRNA internal control. Background obtained with no RT primer reactions was subtracted. Three technical replicates were performed for each RT primer. Graphs show the average and standard deviation (SD).

Other procedures

Western blots were quantified using an Odyssey Infrared Imaging System (LI-COR). Rabbit polyclonal anti-UAP56 antibody was used as 1:1000, and the mouse monoclonal anti-actin antibody (DSHB, JLA20) was used as 1:20. Northern blots ere performed as described previously (Klattenhoff et al., 2007).