Cbp80 is needed for the expression of piRNA components and piRNAs

Cap binding protein 80 (Cbp80) is the larger subunit of the nuclear cap-binding complex (nCBC), which is known to play important roles in nuclear mRNA processing, export, stability and quality control events. Reducing Cbp80 mRNA levels in the female germline revealed that Cbp80 is also involved in defending the germline against transposable elements. Combining such knockdown experiments with large scale sequencing of small RNAs further showed that Cbp80 is involved in the initial biogenesis of piRNAs as well as in the secondary biogenesis pathway, the ping-pong amplification cycle. We further found that Cbp80 knockdown not only led to the upregulation of transposons, but also to delocalization of Piwi, Aub and Ago3, key factors in the piRNA biosynthesis pathway. Furthermore, compared to controls, levels of Piwi and Aub were also reduced upon knock down of Cbp80. On the other hand, with the same treatment we could not detect significant changes in levels or subcellular distribution (nuage localization) of piRNA precursor transcripts. This shows that Cbp80 plays an important role in the production and localization of the protein components of the piRNA pathway and it seems to be less important for the production and export of the piRNA precursor transcripts.


Introduction
piRNAs act in a gene silencing mechanism that controls the expression and translocation of transposable genetic elements (TE). This activity is particularly important in the germline where it is needed for maintaining genome integrity [1] [2]. piRNAs form a group of small RNAs, ranging in size from 23 or 24 to 30-32 nucleotides. They induce RNA-mediated gene silencing by complementary pairing with target sequences. piRNAs are bound by Argonaute proteins of the PIWI clade. In Drosophila these are the P element-induced wimpy testis protein (Piwi), Aubergine (Aub) and Argonaute 3 (Ago3). These three genes and proteins are essential for the production of the piRNAs [1] [2]. piRNA-related processes in Drosophila are active in a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 both the male and the female germline, but most piRNA studies focused on ovaries. Two different, but related, piRNA pathways operate in these organs. The primary piRNA biogenesis pathway is active in the germline and in the somatic cells of the ovary. In both cell types it follows similar routes and involves some common factors. While somatic cells only express Piwi, germ cells express all three Drosophila Piwi clade Argonaute proteins, and through them they additionally produce piRNAs by the so called secondary piRNA pathway or ping-pong amplification cycle that it is active only in the germline [1] [2] [3] [4].
Many genes coding for piRNA factors had been identified in screens for genes that are essential for ovarian development and female fertility (e.g. [5]). However, while the genes had been known for quite some time, they were only much later linked to the piRNA pathway [1] [2] [6]. Recent high throughput screens have revealed a number of additional candidate genes required for the piRNA pathway in the soma and in the germline [7] [8] [9]. Among these are the genes encoding the nuclear cap-binding complex (nCBC), Cbp80 and Cbp20 [7]. The Cbp80/Cbp20 heterodimer interacts primarily through residues in the Cbp20 subunit with the cap structure of mRNAs and most snRNAs. Cbp80 stabilizes this interaction and mediates further interactions with other proteins such as importins, the REF (RNA and export factor binding protein)/Aly protein, CTIF, a MIF4G-domain protein involved in translation, and the NMD (non-sense mediated decay of mRNAs) factor Upf1 [10]. Binding of nCBC to the cap takes place early in the transcription cycle and is a prerequisite for the binding of additional protein components of the RNP.
In this study we investigated in more detail the role of Cbp80 in the piRNA pathway. We found that knocking down Cbp80 in the germline altered the expression of several piRNA pathway components and it interfered with nuclear localization of the piRNA pathway component Piwi and with the nuage localization of Argonaute 3 (Ago3) and Aubergine (Aub), two other piRNA biogenesis factors. Importantly, mRNA levels from germline TEs became elevated in the ovary. Sequencing small RNAs after knocking down Cbp80 in the germline further revealed that Cbp80 is required for the primary production of germline piRNAs as well as for the secondary mechanism, the ping-pong amplification. Cbp80 is therefore involved in the expression of Piwi pathway components and in both piRNA biosynthesis pathways that are active in the female germline.
To construct the C-terminal Cbp80::Venus fusion reporter, the Venus ORF was amplified with primers containing Xba1 sites, and Cbp80 was amplified with a forward primer containing a BamH1 site and a reverse primer containing an Xba1 site. Both sequences were first cloned into the pCRTopo vector. Cbp80 was cut with BamH1 and Xba1 and subcloned into the pUASp-K10 vector to generate pUASP-C-Cbp80-K10 vector. Then Venus was cut from the pCRTopo vector with Xba1 and cloned into the Xba1 site of pUASP-C-Cbp80-K10 to generate pUASP-Cbp80::Venus-K10 (Cbp80::Venus). For the N-terminal fusion construct Venus was amplified with primers containing Not1 and BamH1 sites, and Cbp80 was amplified with primers containing BamH1 and Xba1 sites. Both were initially cloned into the pCRTopo vector. Cbp80 was then cut with BamH1 and Xba1 and subcloned into these sites in the pUASP-K10 vector to generate pUASP-N-Cbp80-K10 vector. Then Venus was cut from the pCRTopo vector with Not1 and BamH1 and cloned into these sites in the pUASP-N-Cbp80-K10 to generate pUASP-Venus::Cbp80-K10 (Venus::Cbp80).

RNA isolation from ovaries and real time qPCR
Ovaries were dissected from 3-4 days old females. Total RNA from ovaries displaying the "d" phenotype (Fig 1) was extracted using TRIzol (Life technologies). cDNA for analysis of the Germline-specific knock down of Cbp80 was performed either with the driver pCog-GAL4 and an shRNA against Cbp80 (shCbp80) or by the GAL4-nos.NGT40 driver combined with UAS-Dcr2 and a dsRNA against Cbp80 (dsCbp80). The knockdown leads to 3 phenotypic classes with the frequencies shown in (A'): normal wild-type phenotype [n], ovaries that appear to lack germ cells [gc -], and partially developing ovaries that mostly degenerate transposon expression levels shown in Fig 1C, was prepared via oligo(dT) priming from 1μg of total RNA. For the analysis of the Piwi components and transposons levels in the remaining experiments, 100ng of total RNA, treated with the Turbo DNA-free kit, were reverse transcribed using the SuperScript III reverse transcriptase and oligo dT primers (Invitrogen). piRNA precursors transcripts were analyzed essentially as described [24]. 100 ng RNA, treated beforehand with Turbo DNA-free kit, were reverse transcribed using the SuperScript III reverse transcriptase (Invitrogen). Specific RT primers for two regions of clusters 42AB, region A and 1-32 (plus strand primers were used) and for the flam locus were mixed with rp49-RT specific primers. All primers have been described already [24]. Real time PCR was carried out using MESA GREEN qPCR MasterMix Plus for SYBR (Eurogentec) with a Qiagen Rotor-Gene Q according to the manufacturer's instructions. Cycle threshold (C( T )) values were determined by the second differential maximum method as calculated by the Rotor-Gene software. Calculation of relative mRNA levels was done by using the 2 [-ΔΔC( T )] method [25], where the C( T ) values of the mRNA levels were normalized to the C( T ) values of rp49, Tub or BicD mRNAs in the same sample. C( T ) values used were the means of triplicate repeats. To test for statistical significance, we first applied a Box-Cox transformation [26]. Finally, a ttest was performed on the transformed data to obtain p-values [27]. Most primers used for RT-PCR were described previously [24] [6] [28]. Primers for Burdock were 5' CGGTAAAATCGCTTCATGGT 3' and 5' ACGTTGCATTTCCCTGTTTC 3'. Cbp80, Ago3, rhi and zuc primers for qPCR were designed in a way that one primer spanned an intron. Primers for Cbp80 were 5' GGATGAGGGCTATGATCATC 3' and 5' TCTAGGTTCGATTCCACGGA 3'; for rhi 5' ATTCCGAAGTGGAGAGCATG 3' and 5' CGTCATTCATCTGGTAGCAG 3'; for Ago3 5' CAATTGGTACGACAGGGTAC 3' and 5' TGAGCGTACATACAACAAGC 3'; for zuc 5' TATGCGTCCGTGCTATAGCA 3' and 5' CCACATTCGTTGGAATTCCG 3'; for Aub 5' CGGTCATCCGGAATTTCCTCATATA 3' and 5' CGTGCATATCAATAGGTGGTATGTG 3' and for piwi 5' TAACGCCGAAAGATACTCATCAATC 3' and 5' TATCCCAACTTGCAATT-CAGTTGGA 3'.

Western analysis
Western blotting was carried out with mouse anti-Piwi P3G11 (1:1,000; [18]), mouse anti-Aub 4D10 (1:2,000; [19]), rabbit anti-Cbp80 (1:1,000; [21]), mouse anti-alpha tubulin primary antibodies (1:250; Developmental Studies Hybridoma Bank), mouse anti-BicD 1B11 (1:10; [29]), mouse anti-Cdk7 (1:10 mix of 20H5 and 19E7.2 clones; [30]) and rabbit anti-Clc (1:3000, [31]). Horseradish peroxidase-conjugated secondary antibodies were from GE Healthcare. To analyze the effects of Cbp80 knockdown and the expression of Piwi components, ovaries with the phenotype "d" were used (Fig 1A). after stages 7-9 [d]. (B-B') Ovaries showing the "d" phenotype upon Cbp80 knockdown were tested for Cbp80 levels by Western blotting. The sample contained only a minor fraction of the other phentypes. In (B) wild-type ovaries and wild-type ovaries expressing a Myc-tagged Cbp80 were used as controls, and knock down was performed with the shRNA. 2 different amounts of the shCbp80 samples were loaded. In (B') ovaries expressing a dsRNA against Cbp80 or GFP were used. Tubulin was used as a loading control. (C) Fold increases in RNA levels of indicated TEs upon germline-specific RNAi-mediated knock down of piwi and Cbp80 (shRNA against Cbp80). The germline GAL4 driver alone was used as control. Fold-changes in RNA levels relative to control were normalized to rp49 levels. Error bars indicate SD; n = 3, with 2 biological replicates. (D) Fold increases in RNA levels relative to rp49 of the same TEs upon knock down of Cbp80 using the dsRNA. Control ovaries expressed the dsGFP RNAi construct. The upregulation of the TE was also observed when Tub and BicD were used to normalize the reads (S3 Fig). *p<0.05; **p<0.01; ***p<0.001.

Small RNA libraries and bioinformatics analysis
Libraries were prepared with RNA extracted from ovaries. Partial Cbp80 knockdown in the germline was achieved by pCog-Gal4 driving shRNA-Cbp80 (TRiP line). pCog-Gal4 driving shmCherry (TRiP line) in flies expressing mCherry-Jupiter in a Jupiter + background was used as control. For the Cbp80 knockdown we selected underdeveloped ovaries (phenotype "d"; Fig  1). RNA was extracted with TRIzol and treated with DNase I (amplification grade, Invitrogen) according to the manufacturer's instructions. Libraries were prepared by Fasteris (Geneva, Switzerland) using the Illumina small RNA kit and a polyacrylamide size selection of 18-30 nt. The Drosophila-specific depletion of 2S rRNA was also performed.
Bioinformatics analyses of small RNAs were performed as described [6]. The adapter sequences were removed from the reads using Trimmomatic version 0.32 [32] without applying any further filter on base quality. Reads with lengths between 19 and 28 bp were then collapsed and aligned to the Drosophila genome (dm3) with Bowtie1 version 0.12.9 [33]. No mismatches and no multiple mapping were allowed. The reads per million (RPM) normalization allowed us to compare our control and experimental conditions in an unbiased way. Additionally, we also normalized knockdown conditions on the basis of the coverage of the flamenco locus. This normalization turned out to be very close to the RPM one, confirming that the flamenco locus had a very similar coverage in both the knockdown and the control. The ping-pong effect was investigated by analyzing the relative frequency (Z-score) of overlaps between reads on different strands as described in [6]. A 10 bp overlap was scored as a signature of the ping-pong effect.

Results
Cbp80 is required for the biosynthesis of piRNAs An RNAi screen for genes required for repression of transposable elements (TEs) indicated that Cbp80 might also be required for the repression of different types of TEs [7]. Because single RNAi and high throughput screening results can sometimes be misleading due to off-target effects, we studied the function of Cbp80 in oogenesis by targeting a different region of the Cbp80 mRNA in the female germline. Because Cbp80 is involved in general gene expression, it seemed likely that Cbp80 is required for several different processes in oogenesis. To find conditions that would allow sufficient development of the germline while still showing a phenotype that resulted from reduced Cbp80 function, we tested different Gal4 drivers. We monitored the expression of these drivers with a Venus::Cbp80 reporter in the UASP vector (S1 Fig). This method revealed that matα4-GAL4 [34] drove expression in the vitellarial stages but not in the germarium. pCog-GAL4 [14] on the other hand, drove expression in the germarium as well as in later stages.
To knock down Cbp80 in the germline we then used the different Gal4 drivers in combination with a transgenic RNAi project (TRiP) line that expresses an shRNA against Cbp80 (shCbp80) [35]. Using the strong but late expressing maternal tubulin driver matα4-GAL4 resulted in ovaries with no apparent abnormalities. In contrast, pCog-GAL4, a driver that is expressed already in the germarium, produced ovaries that could be categorized into three distinct morphological types, normal ovaries, ovaries that lacked germ cells and ovaries that showed partial development until mid-oogenesis, but then mostly degenerating egg chambers after stages 7-9 (phenotype "d"; Fig 1A and 1A'). These phenotypes and the correlation of the phenotypes with the expression patterns of the Gal4 drivers suggest that the early expression of Cbp80 is particularly important for the survival of the germline. Driving Cbp80 RNAi during the germarial stages blocked germline development efficiently, whereas even the strong matα4-GAL4 driver, which is expressed only during the vitellarial stages, did not cause any visible defects when used to knock down Cbp80 specifically in the germline.
The same ovarian phenotypes, although with different frequencies were also observed by expressing a dsRNA that targets a different sequence of Cbp80 (dsCbp80) (Fig 1A'). This dsRNA and a Dicer-2 construct (UAS-Dcr2) were driven by the GAL4-nos.NGT40 that is also expressed from the germarial stages on (S1 Fig). The fact that different lines and different RNAi constructs targeting different parts of Cbp80 cause the same phenotypes strongly argues that the observed ovarian phenotypes are not due to off target or background effects. We interpret the phenotypic defects seen in the partially developing group ("d" in Fig 1A) as partial loss of Cbp80 function phenotypes. Because these egg chambers still developed through the stages where specific phenotypes in early and mid-oogenesis can be studied, we focused our subsequent studies on this group. Consistent with this interpretation, these test conditions also revealed partial knockdown of Cbp80 protein by Western blotting in underdeveloped egg chambers (d), but not in the ones with wild-type appearance (n) (Fig 1B, 1B' and S2 Fig).
To test whether germline RNA levels of transposable elements (TEs) are affected by Cbp80, we performed a Cbp80 knockdown specifically in the female germline using pCog-GAL4 to drive shRNA expression against Cbp80. To isolate the material for the TE expression analysis we dissected ovaries that showed the "d" phenotype ( Fig 1A; see Methods section for details). Using RNA isolated from dissected wild-type and mutant ovaries we then measured levels of mRNAs of marker transposable elements. Furthermore, in order to be able to compare the Cbp80 knockdown results with the effect of knocking down genes that are known to be required for the repression of TEs, we also knocked down piwi in the germline using the driver combination matα4-GAL4; nos-GAL4. Interestingly, like piwi knockdown, Cbp80 knockdown in the germline led to de-silencing of the germline TEs blood, HET-A and TAHRE ( Fig 1C  and S3 Fig). However, a quantitative comparison between 2 different knockdown experiments is not possible. Confirming the role of Cbp80 in repressing TEs, we also observed a similar upregulation of TEs using ovaries expressing the dsRNA against Cbp80. Furthermore, normalization to diverse mRNAs had little influence on the results (Fig 1D and S3 Fig). The combination of both results therefore provides good evidence that the Cbp80 knockdown phenotype is not due to an off-target effect and that Cbp80 is required for repression of TE expression in the female germline of Drosophila, a conclusion that is also consistent with the result observed by Czech et al. [7].
In Drosophila, Piwi proteins and Piwi-interacting RNAs (piRNAs) form the basis of the small RNA-mediated immunity against selfish genetic elements in the gonads [3]. To test whether Cbp80 contributes to the piRNA pathway, we isolated and sequenced small RNAs from ovaries with reduced levels of Cbp80 mRNA in their germline. As control, we also prepared a library from ovaries expressing a mCherry-Jupiter fusion construct and shRNAs against mCherry (shmCherry) in their germline. Both knockdowns were performed using the germline-specific driver pCog-Gal4. The phenotypic composition of the dissected Cbp80 knock-down-ovaries was the same as described in the previous experiment.
We analyzed the piRNA sequence reads mapping to the somatic flamenco (flam) locus and the ones mapping to the germline locus 42AB (Fig 2A and 2B). Criteria for piRNA identification and bioinformatics approaches were chosen according to [6] and are described in the Methods section. Small RNA reads (of 19 to 29 bp) from the 42AB locus were normalized to the number of reads mapping to the flam locus, which is unaffected by germline-specific knockdowns. Most piRNAs were in the size range from 24-28 nucleotides (nt, Fig 2C). The number of reads of this size uniquely mapping to the germline-specific, dual-strand 42AB cluster were reduced upon Cbp80 RNAi treatment compared to the control knockdown ( Fig 2C). On the other hand, small RNA reads of this size derived from the flam locus were unaffected as expected (Fig 2C). Levels of unique piRNAs mapping to 42AB were reduced around 2.8 fold upon Cbp80 knockdown ( Fig 2D). This reduction was observed when data was normalized to the total number of small RNA reads or to the non-affected flam locus. Interestingly, Cbp80 knockdown not only affected levels of piRNAs derived from the dual-strand 42AB cluster, but also the piRNAs derived from the 20A cluster, which is an uni-strand cluster (S4 Fig).
We also used the sequence data to analyze whether Cbp80 plays a role in the ping-pong amplification cycle of the piRNA pathway. The ping-pong signature is defined as the frequency of reads from opposite strands that overlap by 10 nt. Ovaries in which genes involved in the ping-pong amplification pathway were knocked down (spn-E, aub, del and shu) showed a reduced frequency of ping-pong signatures, whereas knockdown of genes that participate mainly in the primary biogenesis pathway (armi and piwi) have no effect on it [7] [6]. Interestingly, Cbp80 knockdown also resulted in a reduced frequency of ping-pong pairs (Fig 2E and  2F). These results therefore show that Cbp80 plays an important role in piRNA biogenesis by affecting both the primary and the secondary piRNA biogenesis pathway.

Expression of Cbp80 during oogenesis
Most of the processes involving the nCBC take place in the nucleus and in the vicinity of the nuclear envelope. To study the localization of Cbp80 in the germline, we stained wild-type ovaries with anti-Cbp80 antibodies (Fig 3A). Cbp80 protein signal was mainly localized in the nucleus, although some of it was present in the cytoplasm as well. We also generated transgenic flies expressing Myc-tagged Cbp80 using the GAL4>UASP system. Overexpressing this construct with the ubiquitously active actin-GAL4 and tub-GAL4 drivers did not appear to be toxic because these flies were viable and fertile. The matα4-GAL4 germline-specific driver was thereafter used to express UASP-myc::Cbp80 in the germline. In addition, we expressed Cbp80 as fusion protein tagged with Venus at the N-and C-term, respectively. We will refer to the tagged fusion proteins as Myc::Cbp80, Venus::Cbp80 and Cbp80::Venus, respectively. Staining Drosophila ovaries with tag-specific antibodies and direct assessment of Venus fluorescence showed primarily a nuclear signal in the germline, too (Fig 3A). Cbp80 signal is often seen in the vicinity of the DNA, but also along the nuclear envelope. In order to find out whether the latter localization corresponds to the nuage, a perinuclear organelle that is involved in piRNA production, we also stained these ovaries for the nuage marker Aub (Fig 3A). The three different staining experiments revealed that the Aub signal was peripheral to the Cbp80 signal, suggesting that Cbp80 does not accumulate in "nuages". Staining ovaries of the same genotypes mCherry (as control). (C) Plots showing the size distribution of small RNAs derived from each strand of the 42AB and the flam clusters. Small RNA reads derived form the 42AB cluster were normalized to small RNAs mapping to the somatic flam, which is unaffected in this germline-specific knockdown. Number of small RNA reads of the characteristic size for piRNAs (23-29 nt) mapping to the germline-specific, dual-strand 42AB cluster were reduced upon Cbp80 knockdown. The same results were obtained when normalizing to the total number of small RNA reads. Small RNA reads derived form the flam cluster were normalized to the total number of small RNA reads. (D) A histogram showing the relative levels of 42AB derived piRNAs upon Cbp80 knockdown compared to the control knockdown. The data is normalized to the number of reads from the flam locus. Differences between control (shmCherry) and shCbp80 treatment are highly significant (p-value is < 2x10 -16 using a chi-square test). (E) Relative abundance of sense-antisense piRNA pairs overlapping by 10 nt (compared to the total number of sense-antisense pairs mapping to the flam or the 42AB clusters, in the small RNA libraries of the Cbp80 and the control knockdowns. Differences between 42AB control (shmCherry) and 42AB levels upon shCbp80 treatment are highly significant (p-value is < 2x10 -16 using a chi-square test). for the nuclear envelope protein Lamin also showed that the intense Cbp80 signal resided within the nucleus, next to the nuclear envelope and the nuclear pores (Fig 3B; S5 Fig). Interestingly, this nuclear Cbp80 distribution pattern in ovaries strongly resembled the one of Piwi (Fig 3C). Role of Cbp80 in piRNA biogenesis

Cbp80 knockdown affects nuclear localization of Piwi
The argonaute sub-family proteins Piwi, Aub and Ago3 play important roles in piRNAinduced silencing. Piwi is normally enriched in the nucleus, a feature reported to be important for its silencing rather than its slicer activity [36] [37]. Aub and Ago3 are cytoplasmic proteins possessing slicer activity and they are key players in the piRNA amplification loop [1] [2] [3] [4]. To learn more about the role of Cbp80 in the piRNA pathway, we tested the effect of Cbp80 knockdown on Piwi localization in ovaries. To be able to correlate Piwi expression and distribution with the efficiency of the Cbp80 knockdown we stained the ovaries simultaneously for Cbp80 and Piwi. To account for staining differences between samples, the nuclear Cbp80 signal in the surrounding somatic follicle cells served as reference signal because we do not expect it to change upon germline specific knock down of Cbp80. In wild-type control egg chambers Piwi nuclear localization was clearly detected from early oogenesis stages on (Fig 4A). Upon germline knock down of Cbp80 we observed that egg chambers that displayed a clearly reduced Cbp80 signal in the germline also showed reduced nuclear localization of Piwi (Fig 4A and  4B). The reduced nuclear levels of Piwi protein were observed upon Cbp80 knockdown with either shRNA or with the long dsRNA targeting different regions of the Cbp80 mRNA. Furthermore, the strength of the reduction correlated with the reduction of the levels of Cbp80 (Fig 4A and 4B). Interestingly, of the known components of the piRNA pathway, the ones that are needed for the primary piRNA biogenesis pathway are also needed for the nuclear localization of Piwi [ [40], further supporting the notion that Cbp80 is essential for piRNA biogenesis.
Given the role of Cbp80 in the piRNA pathway and in Piwi localization, and the fact that both proteins show similar nuclear accumulation patterns, we also tested for direct interactions between Cbp80 and Piwi. However, yeast 2-hybrid experiments failed to detect a physical interaction between Cbp80 and Piwi (S6 Fig) suggesting that Cbp80 affects Piwi localization indirectly.

Normal nuage localization of Aub and Ago3 requires Cbp80
Localization of Aub to the nuage region around nurse cell nuclei was also strongly affected in egg chambers with reduced Cbp80 signal in the germline (Fig 5A and 5B). In control egg chambers the nuage ring of Aub is clearly detected and its signal intensity clearly exceeds the cytoplasmic one (Fig 5A). In contrast, in the germline of most Cbp80 knockdown egg chambers with lowered Cbp80 signal, we did not observe accumulation of Aub in a ring around the nuclear surface (Fig 5A and 5B). On the other hand, egg chambers expressing the RNAi, but showing no reduction of Cbp80 (due to inefficient knockdown), show mostly a normal localization of Aub (Fig 5B). Because this was observed with both RNAi lines, we conclude that Aub localization to the nuage depends on Cbp80 function. Similarly, Ago3 enrichment in the immediate vicinity of the nurse cell nuclear envelope was also reduced by the Cbp80 knockdown (Fig 5C and 5D). The vast majority of control egg chambers and knockdown egg chambers retaining normal levels of nuclear Cbp80 showed a clear signal of Ago3 as a ring around the nurse cell nuclei. In contrast, egg chambers with reduced levels of germline Cbp80 show no discernable perinuclear ring of Ago3 signal even though expression of Ago3 was detectable in the cytoplasm of these egg chambers (Fig 5C and 5D). The fact that reducing levels of Cbp80 also affects the localization of Ago3 and Aub may also explain the requirement for Cbp80 for the ping-pong amplification pathway, which is less dependent on Piwi, but depends on Ago3 and aub [41] [3].

Effect of reduced Cbp80 on levels of RNAs and proteins involved in piRNA biogenesis
Because Cbp80 plays a role in transcription, stability and nuclear export of RNAs, we considered the possibility that Cbp80 is primarily involved in piRNA precursor biogenesis and that defects at this step could lead to mislocalization of Piwi protein. We performed RNA in situ experiments using Stellaris probes designed to detect the piRNA precursors for the 42AB and the 20A clusters (S7 Fig; [13]). In order to be able to correlate possible phenotypes with the reduction of Cbp80 levels, we adopted the protocol for simultaneous labeling of RNAs by in situ hybridization and labeling of proteins by antibody staining. Precursor transcripts were mainly detected inside the nuclei and in the perinuclear nuage region where their processing  Fig 1A) upon Cbp80 knockdown were dissected and co-stained with anti-Cbp80 antibodies (green signal) and anti-Piwi antibodies (red signal). DNA staining (Hoechst) is in blue. Egg chambers with reduced Cbp80 signal in the germline, but normal levels of nuclear Cbp80 signal in the surrounding follicule cells display a healthy normal appearance, but show mislocalization of Piwi protein to the cytoplasm. While rare cases of degenerating or damaged wild-type egg chambers can also show delocalization of Piwi in nurse cells and follicle cells, we did not observe this in healthy wild-type egg chambers.  . Cbp80 knockdown ovaries of the "d" phenotype were used (Fig 1A). Egg chambers were co-stained with anti-Cbp80 antibodies (green signal) and anti-Aub (A, red signal), and anti-Ago3 (C, red signal), respectively. DNA staining (Hoechst) is in blue. (A) The localization of Aub (red), to the nuage is affected by knocking down Cbp80 with two different RNAi lines (shCbp80 and dsCbp80). The level of takes place. To be able to discriminate between these two compartments we also co-stained the ovaries with Lamin. Clear blobs of cluster staining in the nurse cell nuclei were observed even in the absence of Cbp80, arguing that the transcription and stability of precursors transcripts is not strongly affected upon Cbp80 knockdown (S7 Fig). Confirming these results by quantitative RT-PCR, we found that precursor transcript levels from two regions of cluster 42AB (region A and 1-32) were also not reduced upon Cbp80 knockdown (S8 Fig). By analyzing the subcellular localization of the transcript precursors, we observed very few dots of the 42AB and 20A probes outside the nuclei in the nuage region and larger blobs in the nuclei where the transcripts are clustered at the site of transcription (S7 Fig). The signal, at least for the 42AB probe was specific since it disappeared in egg chambers where Rhino (Rhi) was knocked down and Rhi is essential for transcription of this dual strand cluster (S7 Fig; [13]). The low levels of precursor RNAs in the nuage are probably due to the fact that their nuclear export and processing are tightly coupled and every precursor transcript gets immediately chopped up when exiting the nucleus. Cytoplasmic (nuage) precursors would then not be abundant enough to quantify an effect of Cbp80 knockdown on precursor RNA export. In any case, trying to quantify these effects we did not observe a clear reduction of the nuage signals. In summary, although we cannot rule out an effect on piRNA precursor export, the nuclear export of the piRNA precursors is clearly not completely abolished upon Cbp80 knockdown, (S7 Fig). We also considered the possibility, that Cbp80 may be needed for the normal expression of mRNAs encoding protein components of the Piwi pathway. Because Cbp80 might be needed for the stability and transcription of many mRNAs, including the ones coding for subunits of RNA Pol1, 2 and 3, we standardized specific mRNA levels measured by qPCR to total RNA levels. In this way mRNA levels in the Cbp80 knockdown were compared to the control samples (Fig 6A). In addition, we normalized the levels of Piwi component mRNAs with the set of mRNAs used previously (S9 Fig). Germline knock down with the dsRNA against Cbp80 consistently led to a significant reduction of piwi mRNA regardless of whether normalization was performed with total RNA levels or the different control mRNAs (Fig 6A and S9A Fig). A significant reduction of the mRNA levels for other Piwi components (Ago3 and Aub) was also observed when results were normalized to total RNA levels, but their reduction was less clear when specific genes were used for normalization (Fig 6A and S9A Fig). In contrast to these results, knocking down Cbp80 with the shRNA had no clear effects on the levels of Piwi component mRNAs (Fig 6B and S9B Fig). The variability might be due to differential effects of Cbp80 knockdown on the expression of individual genes.
Even if the effect on mRNA levels of Piwi components is mild, we clearly observed that levels of Piwi and Aub proteins were strongly reduced upon knockdown of Cbp80 (Fig 6C and  6D). Furthermore, as seen in two independent knockdown assays, these proteins were clearly reduced when normalized to Tubulin or to total protein levels (Fig 6C and 6D). Interestingly, Cbp80 knockdown can be judged by comparing the Cbp80 signal in the huge nurse cell nuclei to the signal in the surrounding somatic follicel cell nuclei of the same egg chamber (internal control). This ratio is lower in the knock-down situation than in the controls. Egg chambers with Cbp80 knockdown show no clear nuage ring of Aub staining. Plots on the right of the pictures display the fluorescence signal intensity for each channel across the nurse cell nucleus shown. Scale bars: 7.5 μm in control and 10 μm in Cbp80 knockdown egg chamber pictures. (B) Quantification of perinuclear Aub accumulation in the germline where Cbp80 expression is normal or only slightly knocked down (Cbp80+), or where it is efficiently knocked down (Cbp80-, where no or only very weak Cbp80 signal is seen in the germline). The presence of the Aub ring in the nuage correlates with the presence of Cbp80. (C) In control egg chambers Ago3 (red) signal is slightly concentrated in the immediate vicinity of the nuclear envelope of the nurse cells, the nuage. Upon knock down of Cbp80, this localization is reduced or lost. Cbp80 staining (green) reveals efficiency of Cbp80 knockdown. Scale bars: 7.5μm. (D) Quantification of the correlation between Cbp80 levels and Ago3 localization to the nuage as done in B) for Aub localization.
https://doi.org/10.1371/journal.pone.0181743.g005 levels of Cyclin dependent kinase 7 (Cdk7) were also strongly reduced upon Cbp80 knockdown and Piwi proteins had previously been implicated in the upregulation of the expression of Cdks and Cyclins and in enhancing cell proliferation in several cancers [42][43][44][45]. In contrast to Piwi and Aub proteins, levels of Bicaudal-D (BicD) and Clathrin light chain (Clc), two proteins unrelated to the Piwi pathway, were only mildly affected by the Cbp80 knockdown ( Fig  6D). The effect on them was similar to the effect on the Tubulin control. The stronger effect on the levels of Piwi proteins suggests that Cbp80 affects piRNA production most likely through its requirement for the correct expression and normal localization of these protein factors of the Piwi pathway.

Discussion
Knocking down Cbp80 with a set of different germline-specific drivers produced a variety of phenotypes. Cbp80 is clearly essential for germline survival because its knockdown can cause the formation of rudimentary ovaries that lack germ cells. Surprisingly, however, the requirement for germline development and survival seemed to be restricted to the early, the germarial stages of oogenesis. Gal4 drivers that drive Cbp80 RNAi expression during the germarial stages, and uniquely these drivers, blocked germline development efficiently. For instance, the nos-Gal4 driver is expressed during early stages of germarial development and it caused an early and almost complete disappearance of the germline. The pCog-Gal4 driver promoted high expression levels of genes under UASP control in the later phase of germarial development (late in region 2 and in region 3), and the knockdown of Cbp80 with this driver caused the formation of rudimentary ovaries, but still allowed many egg chambers to develop to stage 8 or 9, and some even further. In contrast, the matα4-GAL4 driver is a very strong germline driver, but it is only active slightly later in ovarian development, in the vitellarial stages. Surprisingly, even this strong driver did not cause any morphologically visible mutant phenotypes when it was used to knock down Cbp80. Germline cells therefore seem to have a particularly high need for Cbp80 during the period when nos-Gal4 and pCog-Gal4 are expressed in the germarium, but they seem to be less dependent on Cbp80 for their development and survival during the later phases when matα4-GAL4 is active. Interestingly, a stage-specific requirement for piRNA pathway genes has also been reported by others [46]. These authors found that aub, vasa and spn-E are necessary in very early stages of oogenesis within the germarium, but they appear to become less important for efficient TE silencing in later stages (whereas piwi, ago3, mael appear to be required throughout oogenesis). The requirement for these genes seems to coincide with a period of restarting the piRNA production after the germ cells have moved beyond the stage with reduced Piwi, called the 'Piwiless pocket' [46]. Coinciding with this overlapping stage specific requirement for Cbp80 and the genes involved in the production of the protein components of the piRNA pathway, we also observed that the expression levels (and localization) of Piwi and Aub protein depends much more strongly on Cbp80 than the expression of control proteins unrelated to the Piwi pathway. While we do not know the basis of this differential requirement, and this represents one of the limitations of this study, this differential effect seems to argue against a random pleiotropic effect.
Several groups provided evidence that in Drosophila noncanonical transcription coupled with splicing-and termination inhibition discriminates piRNA precursors from mRNAs and ensures their correct processing [13] [47] [48]. The current model proposes that dual-strand Aub and Cdk7 were strongly reduced upon Cbp80 reduction. On the other hand, levels of Tub, Clc and BicD were less affected. For Cbp80 knockdown samples were extracted from egg chambers showing the phenotype "d" (Fig 1A). https://doi.org/10.1371/journal.pone.0181743.g006 Role of Cbp80 in piRNA biogenesis cluster transcription is achieved by read-through transcription from convergent neighboring genes or by noncanonical transcription initiation by RNA polymerase II (Pol II). In both situations, piRNA-mediated recruitment of Piwi to a dual-strand cluster locus leads to H3K9 trimethylation and the subsequent recruitment of the RDC complex consisting of Rhi, Deadlock (Del) and Cuttoff (Cuff). The cited authors also proposed models in which the upstream transcript undergoes 3 0 end processing and the binding of the RDC complex to chromatin brings it in close proximity to the newly formed 5 0 end of a nascent piRNA precursor transcript. The binding of the RDC complex then prevents processing and degradation of the transcript. This step would distinguish piRNA precursors from mRNAs, which are bound by CBC, spliced, exported and subsequently bound by eIF4E that promotes translation initiation. It is interesting to analyze our results in the context of this model. Cbp80 knockdown did not significantly alter the expression of piRNA precursors from the dual strand cluster 42AB and it even led to higher expression of transposable elements. In contrast, Cbp80 knockdown interfered with expression and localization of the protein components of the piRNA pathway. It therefore seems that in the piRNA biogenesis pathway Cbp80 functions mainly in the production and localization of the protein components of the piRNA pathway because these factors differentially dependent on Cbp80. Anti-Lamin staining allowed us to classify perinuclear dots from the 42AB and 20A transcripts as residing in the 'nuage' region (if they were within approx. 1μm of the Lamin signal) or inside the nucleus. Dots overlapping with the Lamin signal were not counted. The percentage of transcripts in the 'nuage' (relative to transcripts in the 'nuage' and the nucleus) was ploted for control and Cbp80 knockdown. No significant differences were observed between them. (C) Ovaries expressing specifically in the germline (MTD-Gal4 driver) shRNAs against Rhi or white (as control) were used to test the specificity of the 42AB probe. Ovarioles were stained at the same time for Lamin, Rhi and the 42AB and 20A piRNA precursor transcripts. Left pictures show confocal images of nurse cell nuclei stained for Lamin (blue) and Rhi (green). Right pictures show the signals for Lamin (blue), 42AB transcripts (red) and 20A transcripts (green). The signal for the 42AB probe was lost upon Rhi knockdown, which affects transcription from this cluster, confirming the specificity of the probe and the in situ protocol used. Expression of the 20A cluster is not affected (as expected). (PDF) S8 Fig. Levels of 42AB precursor transcripts are not reduced upon Cbp80 knockdown. Ovaries expressed in their germ line shRNAs against Cbp80 (trip line) and shRNAs against mCherry (as control), respectively, under the pCog-Gal4 driver. Control flies expressed also a Jupiter-mCherry fusion gene in a jupiter + background. Levels of precursor transcripts for two regions of the 42AB cluster, regions A and 1-32, and the flam locus were measured by qRT-PCR. Fold expression levels relative to the expression of rp49 are shown for each sample. Error bars represent + /-SD of 2 control and 3 biological knock down replicates. While there is a high variability between the different biological samples (probably due to the phenotypic differences between the knock down samples and the size-matched wild-type stages), no reduction in the expression of piRNA precursors was observed upon Cbp80 knockdown. (PDF)