1. Chromosomes and Gene Expression
  2. Developmental Biology
Download icon

Maternally inherited piRNAs direct transient heterochromatin formation at active transposons during early Drosophila embryogenesis

  1. Martin H Fabry
  2. Federica A Falconio
  3. Fadwa Joud
  4. Emily K Lythgoe
  5. Benjamin Czech  Is a corresponding author
  6. Gregory J Hannon  Is a corresponding author
  1. CRUK Cambridge Institute, University of Cambridge, Li Ka Shing Centre, United Kingdom
Research Article
  • Cited 0
  • Views 1,593
  • Annotations
Cite this article as: eLife 2021;10:e68573 doi: 10.7554/eLife.68573

Abstract

The PIWI-interacting RNA (piRNA) pathway controls transposon expression in animal germ cells, thereby ensuring genome stability over generations. In Drosophila, piRNAs are intergenerationally inherited through the maternal lineage, and this has demonstrated importance in the specification of piRNA source loci and in silencing of I- and P-elements in the germ cells of daughters. Maternally inherited Piwi protein enters somatic nuclei in early embryos prior to zygotic genome activation and persists therein for roughly half of the time required to complete embryonic development. To investigate the role of the piRNA pathway in the embryonic soma, we created a conditionally unstable Piwi protein. This enabled maternally deposited Piwi to be cleared from newly laid embryos within 30 min and well ahead of the activation of zygotic transcription. Examination of RNA and protein profiles over time, and correlation with patterns of H3K9me3 deposition, suggests a role for maternally deposited Piwi in attenuating zygotic transposon expression in somatic cells of the developing embryo. In particular, robust deposition of piRNAs targeting roo, an element whose expression is mainly restricted to embryonic development, results in the deposition of transient heterochromatic marks at active roo insertions. We hypothesize that roo, an extremely successful mobile element, may have adopted a lifestyle of expression in the embryonic soma to evade silencing in germ cells.

eLife digest

Maintaining the integrity of DNA, which encodes all of the instructions necessary for life, is essential for ensuring the survival of a species, especially when genetic information is transferred across generations. DNA, however, contains selfish, mobile elements, called transposons, that move around the genome, hence their nickname ‘jumping genes’. Their movement, a process by which these elements also multiply within genomes, can muddle an organism’s DNA if the transposon happens to land in the middle of a gene, creating a mutation which renders the gene inactive. Transposons have also been linked to the development of cancer, which is a group of diseases driven by accumulating genetic mutations.

Animals have evolved various ways of protecting their DNA against transposons. These are especially important in developing egg cells and sperm, known collectively as germ cells. These cells can produce small fragments of RNA, a molecule similar to DNA, which are able to identify and disarm transposons. While it is known that these small RNAs effectively protect adult gonads from DNA damage, it has been unclear how germ cells formed during the beginning of life are protected.

To find out more, Fabry et al. used a combination of genetic sequencing, protein binding and imaging studies to look at the activity of small RNAs, called piRNAs, which are passed on from the mother to her progeny.

By studying the gene expression levels in fruit fly embryos, Fabry et al. showed that certain transposons become highly active in the first few hours of embryo development, posing a potential threat to DNA integrity. The experiments also identified clear signs in the embryos of an active mechanism for controlling transposons that resembles the small RNA system known from adult germ cells. Fabry et al. removed the piRNAs from the embryos and found that without piRNAs, transposons were more active. This indicates a direct role of these small RNAs in controlling transposons in early development and evidence for a maternally inherited defence system in early embryos.

This study provides insights into the control of transposons in fly embryos. More research is needed to find out whether these embryonic mechanisms are conserved in other animals, including humans. Studying the intrinsic mechanisms that prevent DNA damage and protect our genome could, in time, help to identify new approaches to possibly treat and prevent diseases involving genetic mutations.

Introduction

Transposable elements (TEs) are mobile genomic parasites that can change their genomic position or increase in copy number, and therefore pose a threat to genome integrity. Many TEs have evolved mechanisms that promote their activity specifically in gonads, thereby introducing new insertions that are inherited by future generations (Kim et al., 1994; Leblanc et al., 2000; Wang et al., 2018). Accumulation of insertional mutations in germ cells can lead to decreased population fitness and increased risk of disease (Hancks and Kazazian, 2016; Payer and Burns, 2019). Germ cells, however, harbor protective systems that substantially decrease the likelihood of transposition events.

In animal gonads, the main transposon defense mechanism is the PIWI-interacting RNA (piRNA) pathway (reviewed in Czech et al., 2018; Ozata et al., 2019). At its core, this system depends on 23- to 30-nt piRNAs to distinguish transposon-derived RNAs from host-encoded transcripts and to direct their associated PIWI proteins to active TE targets. In Drosophila, PIWI-guided repression involves cytoplasmic post-transcriptional mRNA cleavage by Aubergine (Aub) and Argonaute-3 (Ago3) and nuclear P-element-induced wimpy testis (Piwi) that engages nascent transposon transcripts and instructs co-transcriptional gene silencing (coTGS) through heterochromatin formation. coTGS requires additional factors acting downstream of Piwi, including Panoramix (Panx), Nuclear Export Factor 2 (Nxf2), NTF2-related export protein 1 (Nxt1), and Cut-up (Ctp), that together form the PICTS complex (also known as SFiNX) (Batki et al., 2019; Eastwood et al., 2021; Fabry et al., 2019; Murano et al., 2019; Schnabl et al., 2021; Sienski et al., 2015; Yu et al., 2015; Zhao et al., 2019). PICTS interfaces with general chromatin silencing factors including Su(var)205/HP1a, SETDB1/Eggless (Egg), and its co-factor Windei (Wde), Su(var)3–3/Lsd1, and its co-factor coRest, Mi-2, Rpd3, Ovaries absent, and Su(var)2–10 (Czech et al., 2013; Handler et al., 2013; Koch et al., 2009; Muerdter et al., 2013; Mugat et al., 2020; Ninova et al., 2020; Osumi et al., 2019; Rangan et al., 2011; Sienski et al., 2015; Yang et al., 2019; Yu et al., 2015). While the detailed mechanisms of transcriptional silencing remain to be established, loci targeted by Piwi are decorated in repressive chromatin marks including trimethylated H3K9 (H3K9me3) (Klenov et al., 2014; Le Thomas et al., 2013; Rozhkov et al., 2013; Sienski et al., 2012; Wang and Elgin, 2011). Loss of Piwi in Drosophila ovaries results in de-repression of TEs and correlates with a severe reduction in H3K9me3 deposition at their corresponding genomic loci.

Piwi and Aub, and to a lesser degree Ago3, have been detected as maternally deposited proteins in Drosophila embryos (Brennecke et al., 2007; Brennecke et al., 2008; Gunawardane et al., 2007; Mani et al., 2014; Megosh et al., 2006; Rouget et al., 2010). Considering that pluripotent progenitor cells give rise to multiple cell lineages, including the germline, maintaining genome integrity during the early stages of embryogenesis is potentially critical. Consistent with their adult roles, maternally inherited PIWI proteins have been observed in the pole plasm of syncytial embryos and in pole cells, the germ cell progenitors, after cellularization (Brennecke et al., 2008; Dufourt et al., 2017; Mani et al., 2014; Megosh et al., 2006).

Though, in adult flies, the piRNA pathway is restricted to the gonad, during the early phases of embryogenesis Piwi is also present in somatic nuclei (Brennecke et al., 2008; Mani et al., 2014; Megosh et al., 2006). This has long been taken as an indication that the piRNA pathway could play roles also in the developing soma, for example, helping to establish its epigenetic landscape (Gu and Elgin, 2013; Seller et al., 2019; Yuan and O'Farrell, 2016). However, probing piRNA pathway function during early embryogenesis has been hampered by a lack of suitable experimental approaches. Disrupting Piwi or other piRNA pathway factors in the female parent either via mutation or RNAi leads to oogenesis defects and often results in sterility or patterning defects that would confound the outcome of analyses (Cox et al., 1998; Czech et al., 2013; Handler et al., 2013; Khurana et al., 2010; Klattenhoff et al., 2007; Klenov et al., 2011; Li et al., 2009a; Malone et al., 2009; Mani et al., 2014; Muerdter et al., 2013; Pane et al., 2007; Park et al., 2019). RNAi-mediated depletion in embryos or generation of homozygous mutant embryos carrying piRNA pathway defects enables analysis of later developmental stages (Akkouche et al., 2017; Gu and Elgin, 2013; Marie et al., 2017), but not time windows where maternally deposited proteins predominate and generally drive development.

Here, we exploit a conditional protein degradation strategy to explore the function of maternally deposited piRNAs during Drosophila embryonic development. We find that Piwi-piRNA complexes present in the embryo are primarily derived from the oocyte, whereas components of the PICTS complex are both maternally deposited and zygotically expressed. An embryonic burst of transposon expression in somatic cells as the zygotic genome becomes active precedes the transient decoration of normally active elements in repressive chromatin marks. Rapid and efficient degradation of maternally deposited Piwi protein in embryos leads to earlier and increased activity of zygotically expressed TEs in concert with loss of repressive marks during the affected developmental stages. Although loss of transposon control in the embryonic soma does not result in an overt morphological phenotype, our results suggest that the piRNA pathway indeed plays a role in regulating the somatic chromatin structure during early embryogenesis. Through these mechanisms, a wave of expression, primarily of the roo transposon, is attenuated, though substantial expression of the TE remains.

Results

A transient burst of transposon expression during Drosophila embryogenesis

The maternal deposition of Piwi, Aub, and Ago3, noted more than a decade ago (Brennecke et al., 2007; Brennecke et al., 2008; Gunawardane et al., 2007; Mani et al., 2014; Megosh et al., 2006; Rouget et al., 2010), has long suggested a possible role for the piRNA pathway during embryogenesis. Prior studies have indicated that maternal instructions transmitted via piRNAs are important for defining piRNA clusters in the subsequent generation and/or provide critical information for gaining control over at least some transposons in daughters (Akkouche et al., 2013; Akkouche et al., 2017; Brennecke et al., 2008; de Vanssay et al., 2012; Hermant et al., 2015; Khurana et al., 2011; Le Thomas et al., 2014a; Le Thomas et al., 2014b). Both of these functions are relevant in gonadal cells. Yet, prior studies highlighted the presence of maternally deposited Piwi protein in the somatic nuclei of developing embryos (Brennecke et al., 2008; Mani et al., 2014; Megosh et al., 2006), leading to suggestions that piRNAs might help set the global epigenetic landscape of the embryonic soma (Gu and Elgin, 2013). To investigate the role of the piRNA pathway during embryogenesis, we first focused on its most well-established role, that of transposon control. Toward this end, we first characterized the expression of transposons throughout Drosophila embryogenesis by RNA-seq and quantitative mass spectrometry (Figure 1A).

Figure 1 with 2 supplements see all
A transient burst of transposon expression during Drosophila embryogenesis.

(A) Schematic of Drosophila embryogenesis indicating Bownes stages and collected time points. (B) Bar graphs showing contribution of transposon derived reads to the transcriptome of control w1118 embryos at the indicated time points in percent. Error bars show standard deviation (n = 2). (C) Line graphs showing the RNA expression (in rpm) for the 30 most expressed transposons during the indicated time points of embryogenesis. Error bars show standard deviation (n = 2). (D) Confocal fluorescent microscopy images of control w1118 embryos showing nuclei stained with DAPI and roo transcripts by RNA-FISH at the indicated embryonic stages (also see Figure 1—figure supplement 2B). Scale bar = 100 µm.

Transcriptomes of 0–2 hr after egg laying (AEL) embryos represent the maternally inherited mRNA pool. Maternal transcripts are cleared and the zygotic genome is activated (zygotic genome activation [ZGA]) around nuclear cycle 14 (NC14; 2–2.5 hr AEL), and we generated RNA-seq data spanning 1 hr intervals of development from this point up to 10 hr AEL (stage 13). For comparison, we also included two late-stage embryo time points (12–13 hr and 17–18 hr AEL) as these were times when our prior data indicated that maternal Piwi was no longer detectable in somatic nuclei (Brennecke et al., 2008). To take into account different library sizes and facilitate comparability throughout our time-course experiment that only contained two biological replicates per time point, our RNA-seq data was normalized to reads per million (rpm). We benchmarked our dataset by comparing the expression of selected embryonic genes in our RNA-seq to reported transcriptomes in FlyBase (Graveley et al., 2011). We found highly similar expression patterns of genes that are dynamically regulated during embryogenesis (Figure 1—figure supplement 1A, B). Furthermore, well-validated maternal (e.g., fs(1)N and gammaTub37C) and zygotic (e.g., Ultrabithorax [Ubx] and wingless [wg]) genes demonstrated their expected expression patterns in our datasets (Figure 1—figure supplement 2A).

We detected only very few transposon transcripts in pre-ZGA embryos (0–2 hr AEL), as might be expected from their effective silencing by the piRNA pathway in ovaries. TE expression steadily increased following ZGA and peaked between 4 and 6 hr AEL (Figure 1B), similar to what was noted in prior reports (Batut et al., 2013). At the peak, transposon RNAs correspond to ~1.7% of the total embryonic transcriptome, with levels at the later studied time points decreasing to below 1% of the overall transcriptome. Transposons often show highly dynamic spatio-temporal expression; thus, we analyzed the contribution of individual TE families to the embryonic transcriptome. Interestingly, the majority of transposon expression could be attributed to one single transposon family, roo (Figure 1C). At its peak at 4–6 hr AEL, reads derived from the roo TE accounted for more than 1% of the entire embryonic transcriptome, corresponding to more than 70% of all TE-derived reads overall. From its expression peak, roo mRNA levels declined strongly before leveling off at around 12 hr AEL. While less pronounced, other transposons, such as copia and 297, also showed dynamic expression changes during embryogenesis (Figure 1—figure supplement 2F).

The roo expression peak at 4–6 hr AEL could be due to transcription from germ cell precursors, which become transcriptionally active around 3.5 hr AEL (stage 8) (Van Doren et al., 1998; Zalokar, 1976). However, the sheer abundance of roo and other transposon transcripts argued strongly that they must emanate at least in part from somatic nuclei as these vastly outnumber the germ cell precursors. To directly test the origin of roo transcripts during embryogenesis, we performed RNA fluorescence in situ hybridization (RNA-FISH). In agreement with our RNA-seq data, roo transcripts were detected as early as stage 6 (in gastrulating embryos ~3 hr AEL) and localized predominantly to yolk cell nuclei (Figure 1D, Figure 1—figure supplement 2B). Stage 11 embryos (~5 hr AEL) showed strong roo RNA signal in somatic cells of the mesoderm, similar to earlier reports (Brönner et al., 1995; Ding and Lipshitz, 1994). In contrast, roo transcript levels were undetectable by FISH in late-stage embryos (>10 hr AEL). These data indicate a transient somatic burst of roo expression during early Drosophila development.

TEs rely on proteins encoded in their open reading frames (ORFs) for mobilization. roo is an LTR retrotransposon and, as has been proposed for gypsy in ovarian follicle cells (Kim et al., 1994; Leblanc et al., 2000; Song et al., 1997), could potentially be packaged into virion-like particles, possibly enabling infection of germ cell precursors as a propagation mechanism. To determine whether roo-encoded proteins are expressed in embryos, we mined quantitative proteomic data from three developmental intervals (Figure 1A). The first, 0–2 hr AEL, represents the time before ZGA when the proteome is derived from maternal protein deposition and zygotic translation of maternal mRNAs. The second, 5–7 hr AEL, represents an interval where zygotic roo expression had become robust, and the third, 10–12 hr AEL, is a time at which roo RNA levels had substantially declined.

In transcriptionally silent embryos (0–2 hr AEL), we detected over 6400 unique proteins. Compared to 0–2 hr embryos, the signal intensity of ~17% or 1114 proteins significantly increased (p<0.01) by over 25% in 5–7 hr AEL embryos (Figure 1—figure supplement 2C). We also detected 490 (or ~8% of) proteins that significantly decreased (p<0.01) by over 25% in 5–7 hr AEL embryos (Figure 1—figure supplement 2D). The majority of proteins (4652 or 72%), however, did not change by more than 25% between 0–2 hr and 5–7 hr AEL embryos. As with transcriptome analyses, known maternally deposited and zygotically expressed proteins showed their expected patterns of presence in the datasets.

Compared to the early time point (0–2 hr AEL), 5–7 hr AEL embryos showed significant accumulation of roo peptides (p<0.01) corresponding to its expression peak. roo encodes a single ORF (with a predicted protein weight of 272 kDa), which contains a group-specific antigen-like protein (gag), a reverse transcriptase (RT/pol), an envelope protein (env), two peptidases-like domains (Pep), and a zinc finger (Figure 1—figure supplement 2E). We detected peptides corresponding to the gag, pol, and env proteins (Figure 1—figure supplement 2E, bottom), indicating potential competence for retrotransposition. We additionally detected proteins derived from other transposons including copia and 297. Of note, roo ORFs remained detectable at 10–12 hr AEL (Figure 1—figure supplement 2F), possibly suggesting substantial stability, as this was a time at which roo mRNA levels had diminished.

The known cohort of piRNA coTGS factors is present during embryogenesis

The decline in transposon expression from 4 to 6 hr to 10–12 hr intervals of embryogenesis could potentially involve the piRNA pathway. However, piRNA-guided post-transcriptional or co-transcriptional silencing also requires a growing list of additional proteins (reviewed in Czech et al., 2018; Ozata et al., 2019). We therefore probed the expression of known piRNA pathway components during various stages of embryogenesis in our transcriptomic and proteomic datasets.

With the exception of Piwi, genes involved in coTGS were both maternally deposited and zygotically expressed during the first ~10 hr of embryogenesis (Figure 1—figure supplement 2A). Components of the PICTS complex, comprising Panx, Nxf2, Nxt1, and Ctp, showed abundant protein expression in the 5–7 hr and 10–12 hr AEL time intervals. piRNA-mediated coTGS also depends on several general chromatin modifiers, including Egg and its co-factor Wde (Osumi et al., 2019; Rangan et al., 2011; Sienski et al., 2015; Yu et al., 2015). Both of these proteins are required for heterochromatin formation in the embryo, and Egg in particular has previously been implicated in embryonic repeat silencing (Seller et al., 2019). Similar to piRNA-specific coTGS factors, proteins involved in general chromatin silencing were both maternally deposited and zygotically expressed and detected at all studied time points, as expected based on their ubiquitous functions (Figure 1—figure supplement 2A). Of note, Piwi mRNA appears to be primarily maternally deposited, with zygotic transcript levels remaining low throughout embryogenesis (Figure 1—figure supplement 2A, Figure 2A, B).

Figure 2 with 1 supplement see all
piRNA co-transcriptional gene silencing (coTGS) factors are maternally inherited and localize to somatic cells of the Drosophila embryo.

(A) Bar graphs showing Piwi RNA expression (in rpm) at the indicated time points in control w1118 embryos. Error bars show standard deviation (n = 2). (B) Bar graphs showing Piwi protein signal intensity (arbitrary units) at the indicated time points in control w1118 embryos. Error bars show standard deviation (n = 3). (C) Stand-still images from Video 1 obtained by light-sheet fluorescent live microscopy of embryos derived from parents expressing GFP-Piwi (green) and H2Av-RFP (red) for the indicated time points. Scale bar = 50 µm. (D) As in (C) but showing the transition from NC13 to NC14. Scale bar = 50 µm. (E) Confocal fluorescent microscopy images of embryos derived from females expressing GFP-AID-Piwi crossed to control w1118 males probing for GFP and DAPI. Shown are embryos at the blastoderm stage (stage 5) and late-stage embryos (>12 hr after egg laying). Scale bar = 100 µm. Zoom of the indicated regions showing developing germ cells. Scale bar = 10µm. (F) As in (E) but showing embryos derived from control w1118 females crossed to GFP-AID-Piwi males. (G) Bar graph showing small RNA-seq from Piwi immunoprecipitation of 0–8 hr control w1118 embryos (green, n = 1) or adult ovaries (orange, n = 1). Shown are antisense piRNAs of the 30 most abundant TE families in embryos as percentage of reads mapping to indicated transposons relative to all transposable element-mapping antisense piRNAs.

In contrast, we noted little or no maternal deposition and low zygotic expression of key components of the piRNA precursor expression and export machinery and of critical piRNA biogenesis factors (Figure 1—figure supplement 2A). Considered together, our expression analyses are consistent with the potential of maternally instructed Piwi protein acting through coTGS during Drosophila embryogenesis.

Components of the piRNA-guided coTGS machinery are enriched in somatic and pole cell nuclei during embryogenesis

To assess the potential role of the piRNA pathway in regulating the transposon burst during Drosophila embryogenesis, we examined the spatial and temporal expression of coTGS proteins in the developing embryo using light-sheet live fluorescence microscopy. For this purpose, we used two previously published lines carrying a modified BAC expressing either GFP-Piwi or GFP-Panx from its endogenous regulatory region (Handler et al., 2013; Sienski et al., 2015) and a GFP-Nxf2 knock-in line that we generated by CRISPR/Cas9 (Fabry et al., 2019). As Ctp and Nxt1 have many additional functions, we did not examine their localization in this study. We also crossed in a transgene carrying H2Av-RFP to enable tracking of nuclei. Pre-blastoderm stage embryos (0.5 hr AEL) were continuously imaged for >10 hr of embryogenesis. As previously reported (Brennecke et al., 2008; Mani et al., 2014; Megosh et al., 2006) and consistent with its maternal deposition, we detected GFP-Piwi during the pre-blastoderm stage (NC1–9, ~0–30 min AEL) localized to the posterior pole where it formed a crescent-like structure (Video 1, Figure 2C).

Video 1
GFP-Piwi live imaging.

Light-sheet fluorescent live microscopy of embryos derived from parents expressing GFP-Piwi (green) and H2Av-RFP (red) in developing embryos for indicated time after egg laying.

As embryogenesis progressed and somatic nuclei migrated to the surface (NC 9–14, ~1.5–3 hr AEL), Piwi localized to somatic nuclei and to the pole plasm surrounding the nuclei of germline progenitor cells, as we and others reported earlier based on immunofluorescence staining of fixed embryos (Brennecke et al., 2008; Mani et al., 2014; Megosh et al., 2006). In agreement with an earlier report (Mani et al., 2014), our dynamic data revealed that nuclear Piwi signal strongly decreased during mitotic cycles, with little fluorescence signal overlapping with H2Av-RFP during nuclear divisions (Video 1, Figure 2D). We continued to detect Piwi expression in somatic nuclei throughout the first 10 hr of embryogenesis; however, signal intensity decreased over time. This observation was consistent with transcriptomic and proteomic measurements taken over a comparable time course (Figure 2A, B).

Similar to Piwi, both Nxf2 and Panx were detected in somatic and pole cell nuclei from the syncytial blastoderm stage (Videos 2 and 3, Figure 2—figure supplement 1A–E). In contrast to Piwi, Panx and Nxf2 showed strong co-localization with H2Av-RFP during mitotic cycles (Videos 2 and 3, Figure 2—figure supplement 1F, G), suggesting that while Piwi is predominantly excluded, Nxf2 and Panx are retained in the nucleoplasm during mitosis. Consistent with our RNA-seq and TMT-MS data (Figure 2—figure supplement 1H–K), as embryogenesis progressed, Panx and Nxf2 remain detectable for several hours (>10 hr AEL), closely matching the protein expression of Piwi.

Video 2
GFP-Panx live imaging.

Light-sheet fluorescent live microscopy of embryos derived from parents expressing GFP-Panx (green) and H2Av-RFP (red) in developing embryos for indicated time after egg laying.

Video 3
GFP-Nxf2 live imaging.

Light-sheet fluorescent live microscopy of embryos derived from parents expressing GFP-Nxf2 (green) and H2Av-RFP (red) in developing embryos for indicated time after egg laying.

Piwi carries epigenetic information in the form of piRNAs (Brennecke et al., 2008; Le Thomas et al., 2014b). However, it is unclear if Piwi-piRNA complexes are assembled during oogenesis prior to maternal deposition into the embryo, or whether zygotic piRNA biogenesis and Piwi loading also occurs. We therefore analyzed the expression of GFP-tagged Piwi from reciprocal crosses with control w1118 flies by immunofluorescence staining in early and late-stage embryos. Embryos derived from females expressing GFP-Piwi showed strong maternal deposition of Piwi during early embryogenesis (Figure 2E, Figure 2—figure supplement 1L), with GFP fluorescence in later stage (>12 hr AEL) embryos restricted to the germline cells. Consistent with maternal deposition of Piwi, embryos derived from the reciprocal cross showed no GFP signal in the early embryos (Figure 2F, Figure 2—figure supplement 1M). Instead, we only observed GFP-Piwi signal in the developing gonads of late-stage embryos, likely as a result of zygotic expression. Strikingly, Piwi of zygotic origin localized exclusively to the cytoplasm of the germ cell progenitors and was not detected in nuclei, suggesting that zygotically transcribed Piwi is likely not relevant for coTGS until later in development.

The decline in embryonic transposon expression is correlated with hallmarks of piRNA-dependent co-transcriptional silencing

Piwi proteins are guided by their piRNA co-factors to recognize and co-transcriptionally silence active transposons in the Drosophila ovary (Le Thomas et al., 2013; Post et al., 2014; Sienski et al., 2015; Sienski et al., 2012; Yu et al., 2015). If this pathway were relevant in the embryonic soma, maternally deposited Piwi would require instructions to recognize embryonically expressed elements. To examine this possibility, we immunoprecipitated Piwi from 0 to 8 hr control w1118 embryos as well as from adult ovaries and sequenced the associated small RNAs. Piwi in both tissues existed in complex with 23- to 28-nt piRNAs and showed nearly indistinguishable size profiles that were biased for antisense reads (Figure 2—figure supplement 1N, O). Closer inspection by aligning the reads to transposon consensus sequences revealed similar piRNA levels for the majority of TEs; however, we detected some notable differences (Figure 2G). Piwi in ovaries showed higher levels of antisense piRNAs targeting the TEs mdg1 and 412, in agreement with the majority of these small RNAs originating from the soma-specific flam piRNA cluster (Brennecke et al., 2007; Malone et al., 2009; Zanni et al., 2013). Piwi in embryos showed high levels of antisense piRNAs targeting roo (~16% or all TE-targeting reads) and 297 (~9%), consistent with an ability of maternally deposited Piwi to potentially recognize these TEs when expressed in the embryo.

In Drosophila ovaries, coTGS depends on Piwi-mediated recruitment of the PICTS/SFiNX complex and correlates with the deposition of H3K9me3 marks at TE insertions and surrounding genomic regions (Batki et al., 2019; Eastwood et al., 2021; Fabry et al., 2019; Murano et al., 2019; Schnabl et al., 2021; Sienski et al., 2015; Yu et al., 2015; Zhao et al., 2019). Due to the poor conservation of the genomic locations of transposon insertions between different Drosophila strains, we used whole-genome sequencing (WGS) to de novo identify the TE insertion sites present in our control w1118 flies (see Materials and methods). This data enabled us to identify over 600 euchromatic transposon insertions that are absent from the dm6 reference genome, and these were used for our chromatin analyses, as most annotated insertions in the dm6 genome assembly were absent from our strain.

In order to determine the fate of transposon loci throughout embryogenesis, we performed H3K9me3 chromatin immunoprecipitation followed by sequencing (ChIP-seq) on control w1118 embryos at 2 hr intervals covering the period when transposon expression is dynamic (0–10 hr AEL) and a later time point (16–18 hr AEL) well after maternal Piwi protein was no longer detectable in somatic nuclei (Figures 1A and 2C). We included adult ovaries, which show piRNA-guided coTGS, as well as adult heads, a somatic tissue without active piRNA pathway, to compare the changes of this repressive chromatin mark across different stages and tissues of Drosophila development.

Early embryos (0–2 hr AEL) showed low levels of H3K9me3 signal at 117 euchromatic, w1118-specific roo insertions (Figure 3A). However, as development progressed, H3K9me3 levels increased with a peak at 6–10 hr AEL (Figure 3A, Figure 3—figure supplement 1A). Thus, deposition of repressive chromatin marks correlated with the RNA expression of roo, yet the maximum of H3K9me3 accumulation lagged behind the RNA expression peak by approximately 2 hr. These data are consistent with a requirement for nascent transcription for efficient conversion of a TE insertion into heterochromatin, as previously reported in yeast (Bühler et al., 2006; Shimada et al., 2016) and for the recognition of transposon loci by the piRNA pathway (Le Thomas et al., 2013; Post et al., 2014; Rozhkov et al., 2013; Sienski et al., 2015; Sienski et al., 2012; Yu et al., 2015). Of note, the deposition of repressive marks trailed the direction of transcription and showed higher signal enrichments in the regions downstream of the transposon insertions, as previously observed for piRNA-dependent silencing in cell culture systems (Fabry et al., 2019; Sienski et al., 2015; Sienski et al., 2012).

Figure 3 with 1 supplement see all
Transposon insertions targeted by piRNAs in embryos show epigenetic changes characteristic of co-transcriptional gene silencing.

(A) Heatmaps (top) and metaplots (bottom) showing H3K9me3 ChIP-seq signal (in rpm) for the indicated embryonic stages and adult tissues at 117 euchromatic, w1118-specific roo insertions (n = 2). Signal is shown within 10 kb from insertion site and sorted from 5′ to 3′. (B) UCSC genome browser screenshot showing H3K9me3 ChIP-seq signal for the indicated genes on chromosome 2R carrying w1118-specific roo and Doc insertions.

Interestingly, H3K9me3 signal at euchromatic roo insertions of 16–18 hr AEL embryos, which lacked maternal Piwi in somatic nuclei and no longer express roo, showed diminished intensities compared to earlier time intervals. Similarly, heads and ovaries, both tissues from adult flies, showed no enrichment of H3K9me3 at euchromatic roo insertions, despite the presence of a functional piRNA pathway in ovaries. Considered together, these data suggest that maternal piRNAs program a response to a burst of roo expression during embryonic development but that the deposition of H3K9me3 marks, likely directed via coTGS, no longer occurs at developmental time points and in tissues where roo is not expressed. This is consistent both with the known requirement for active transcription for targeting by Piwi and with the observed need for continuous engagement of the PICTS/SFiNX complex to maintain H3K9me3 marks on transposon loci (Batki et al., 2019; Eastwood et al., 2021; Fabry et al., 2019; Le Thomas et al., 2013; Murano et al., 2019; Post et al., 2014; Rozhkov et al., 2013; Schnabl et al., 2021; Sienski et al., 2015; Sienski et al., 2012; Yu et al., 2015; Zhao et al., 2019).

To investigate whether this mechanism is specific to roo or more general, we examined the transposon 297, which is also expressed during embryogenesis (Figure 1—figure supplement 1F) and showed high targeting potential by maternally inherited piRNAs (Figure 2G). Genomic loci in close proximity to euchromatic, w1118-specific 297 insertions (n = 20) showed dynamic deposition of H3K9me3 similar to roo (Figure 3—figure supplement 1B). However, while H3K9me3 levels at roo insertions peaked between 6 and 10 hr AEL, 297 insertions showed the maximum H3K9me3 signal intensity between 2 and 8 hr AEL, suggesting that these loci are targeted by coTGS earlier than roo insertions. In contrast, H3K9me3 occupancy at transposons such as mdg1 and 412 that were expressed during embryogenesis but lacked substantial maternal deposition of piRNAs retained low H3K9me3 levels throughout embryogenesis, though they showed a strong enrichment in ovaries (Figure 3—figure supplement 1C).

To determine whether the deposition of repressive chromatin marks at euchromatic 297 and roo insertions was specific, rather than reflecting a general trend of H3K9me3 accumulation genome-wide, we analyzed genomic regions not targeted by maternally inherited piRNAs. H3K9me3 signal at constitutive heterochromatin remained stable throughout the sampled time points (Figure 3—figure supplement 1D), while H3K9me3 levels on chromosome 4 increased steadily throughout development (Figure 3—figure supplement 1E). Of note, while ovaries showed no coTGS signature at roo insertions, other transposons, such as Doc, showed a clear accumulation of H3K9me3 marks that was absent in embryos during all assayed time points (Figure 3B). Considered together, these results are consistent with piRNA-guided chromatin modification of a subset of transposons that show activity during Drosophila embryonic development.

An auxin-inducible degron enables rapid depletion of Piwi in ovaries and early embryos

Though embryonically repressed transposons bore hallmarks of piRNA-guided heterochromatin formation, the reliance of the pathway on maternally deposited Piwi-piRNA complexes prevented a demonstration that silencing depended on the pathway through conventional genetics. Ovaries that lack key piRNA pathway silencing factors show substantial expression changes and produce morphologically altered eggs that largely fail to develop normally (Cox et al., 1998; Czech et al., 2013; Handler et al., 2013; Khurana et al., 2010; Klattenhoff et al., 2007; Klenov et al., 2011; Li et al., 2009a; Malone et al., 2009; Mani et al., 2014; Muerdter et al., 2013; Pane et al., 2007; Park et al., 2019).

To investigate the effect of Piwi depletion on Drosophila embryogenesis without affecting oogenesis, we used the auxin-inducible degron (AID) system (Nishimura et al., 2009). This protein degradation system comprised an AID-tag, fused to the protein of interest, and the plant-derived F-box protein transport inhibitor response 1 (TIR1). AID and TIR1 associate with each other in an auxin-dependent manner, with binding of TIR1 to the AID-tagged target leading to the recruitment of the cellular ubiquitination machinery and target protein degradation via the proteasome (Figure 4A). This conditional degradation system has proven effective in several model organisms including Drosophila where it was recently shown to enable degradation of the germ cell-specific protein Vasa (Bence et al., 2017).

Figure 4 with 1 supplement see all
Degradation of Piwi protein in ovaries resembles mutant phenotypes.

(A) Cartoon illustrating the Piwi protein degradation strategy using the auxin-inducible AID-TIR1 system. (B) Confocal fluorescent microscopy images showing ovary egg chambers of GFP-AID-Piwi; OsTIR1 flies fed with yeast paste containing 5 mM auxin for the indicated time (also see Figure 4—figure supplement 1A). Blue = DAPI. Green = GFP-AID-Piwi. (C) Western blot of ovaries from females treated with 5 mM auxin-containing yeast paste for the indicated time period or control females probing for Piwi and Tubulin as a loading control. (D) Bar graphs showing rp49-normalized steady-state RNA levels of the indicated transposable elements and control genes in ovaries of GFP-AID-Piwi; OsTIR1 flies fed with yeast paste containing 5 mM auxin for the indicated time. Error bars show standard deviation (n = 3). Asterisk denotes significant changes compared to control (p<0.05, unpaired t-test). (E) Bar graphs showing the percentage of embryo deformation phenotypes laid by GFP-AID-Piwi; OsTIR1 females fed with yeast paste containing 5 mM auxin for the indicated time. (F) As in (E) but showing the hatching rate in percent.

We used CRISPR/Cas9 to insert an amino-terminal GFP-AID tag at the Drosophila piwi locus and crossed these flies to a line expressing the Oryza sativa-derived TIR1 (OsTIR1) protein under the control of the ubiquitin promoter. As a proof of concept, we tested the auxin-induced degradation of Piwi in adult ovaries of flies homozygous for both GFP-AID-Piwi and OsTIR1. Feeding flies for 24 hr with 5 mM auxin-containing yeast paste was sufficient to induce complete degradation of Piwi in ovaries (Figure 4B, C, Figure 4—figure supplement 1A), and this depletion resulted in the de-repression of transposons (Figure 4D). Notably stronger changes were observed following longer treatments, possibly implying a lag between loss of piRNA pathway function and that of repressive chromatin marks. Following a 1-day treatment, embryos laid by Piwi-depleted females developed without defects and showed similar hatching rates as their control treated siblings (Figure 4E, F). Longer auxin treatments resulted in more frequent deformation of embryos that was accompanied by reduced hatching rates (Figure 4E, F), likely due to patterning defects as a result of Piwi depletion from follicle cells.

Drosophila embryos develop within a relatively impermeable chorion, and treatment of embryos directly with auxin showed little impact. However, in dechorionated embryos we observed a near complete degradation of Piwi protein following 30 min auxin treatment of embryos collected for 0–30 min AEL (Figure 5A, B). To investigate the dynamics of auxin-mediated Piwi depletion in embryos, we used light-sheet fluorescence live microscopy. Early blastoderm embryos treated with 5 mM auxin showed rapid degradation of GFP-AID-Piwi signal, which was undetectable after 25 min of treatment (Figure 5C, Video 4). Of note, the removal of maternal Piwi in this time window did not significantly affect the embryo hatching rate (Figure 4—figure supplement 1B).

Figure 5 with 1 supplement see all
Degradation of maternally deposited Piwi in embryos leads to transposon deregulation.

(A) Schematic of embryo auxin treatments and sample collection for RNA-seq and ChIP-seq experiments. (B) Western blot showing abundance of GFP-AID-Piwi fusion protein in embryos treated with 5 mM auxin for 2 hr. OsTIR1 and tubulin expression are shown as loading control. (C) Stand-still images from Video 4 obtained by light-sheet fluorescent live microscopy of embryos derived from parents expressing GFP-AID-Piwi and OsTIR1 treated with 5 mM auxin for the indicated time intervals. Scale bar = 50 µm. (D) MA plot showing base mean expression (log10 scale) of transposon RNAs relative to their fold-change (log2 scale) in GFP-AID-Piwi; OsTIR1 embryos treated with 5 mM auxin versus control (n = 3). Gray = genes, green = TEs not changed (p<0.05), red = transposable elements (TEs) significantly changed (p<0.05) and fold-change > 2. (E) MA plot showing base mean signal intensity (log10 scale) of TEs relative to the H3K9me3 ChIP-seq signal enrichment (log2 scale) in GFP-AID-Piwi; OsTIR1 embryos treated with 5 mM auxin versus control (n = 3). Gray = TEs not significantly changed (p>0.05), purple = TEs significantly changed (p<0.05), blue = TEs significantly changed (p<0.05) and fold-change < –2. (F) Heatmaps (top) and metaplots (bottom) showing H3K9me3 ChIP-seq signal (in rpm) for control embryos and 5 mM auxin-treated embryos at 205 euchromatic, degron strain-specific roo insertions (n = 3). Signal is shown within 10 kb from insertion site and sorted from 5′ to 3′. (G) Bar graphs showing H3K9me3 signal intensity (in rpm) for the indicated treatments at roo loci. Error bars show standard deviation (n = 3). Statistics were calculated with unpaired (two-sample) t-test. (H) As in (G) but showing constitutive heterochromatin. (I) As in (G) but showing chromosome 4 regions. (J) Model of piRNA-guided chromatin modification at active transposons in somatic cells of the developing Drosophila embryo.

Video 4
GFP-AID-Piwi live imaging.

Light-sheet fluorescent live microscopy of embryos derived from parents expressing GFP-AID-Piwi (green) and OsTIR1. Time course shows early embryo treated with 5 mM auxin for indicated time points.

Maternally deposited Piwi directs heterochromatin formation at active transposon insertions during early embryogenesis

We next investigated the impact of degrading maternal Piwi from early-stage embryos on transposons. Embryos derived from flies homozygous for GFP-AID-Piwi and OsTIR1 were collected across a 30 min period and treated for an additional 2.5 hr with or without 5 mM auxin before RNA extraction, generation of libraries, and differential expression analysis of the sequenced transcriptomes (Figure 5A). These embryos corresponded to 2.5–3 hr AEL, the point at which we began to observe zygotic roo transcripts (Figure 1C), and showed minimal differences between control- and auxin-treated embryos for the same set of genes used to benchmark our dataset (Figure 5—figure supplement 1A). The majority of transposons showed no significant expression change upon Piwi depletion; however, roo and 297 were significantly de-repressed (p<0.05) by more than twofold (Figure 5D), suggesting that Piwi impacts their expression during embryogenesis. Previous studies suggested that auxin in small concentrations has a negligible impact on Drosophila development (Bence et al., 2017; Trost et al., 2016), but to control for effects of auxin itself on TE regulation, we also evaluated transposon expression in auxin-treated GFP-AID-Piwi embryos that lack OsTIR1. Without OsTIR1, 2.5–3 hr embryos treated with 5 mM auxin showed no significant changes in transposon expression compared to control siblings treated with PBS (Figure 5—figure supplement 1B).

We additionally examined changes in the repressive chromatin mark H3K9me3 to determine whether these were deposited in a piRNA-dependent fashion at euchromatic roo and 297 transposon insertions. We again collected embryos for 30 min and treated with 5 mM auxin (or PBS as a negative control) for 6 hr, which yielded embryos 6–6.5 hr AEL (Figure 5A) and corresponds to the peak in H3K9me3 signal at roo insertions in control w1118 embryos (Figure 3A, Figure 3—figure supplement 1A). Piwi depletion severely impacted H3K9me3 signal over the transposon consensus sequence of roo and 297, but not that of other TEs (Figure 5E). Additionally, H3K9me3 levels at individual transposon genomic loci (see Materials and methods for identification of TE insertions in our fly stock) showed similar patterns. H3K9me3 signal in genome-wide 5 kb bins predominantly changed when roo or 297 insertions were nearby (Figure 5—figure supplement 1C). We identified 154 bins with significantly reduced (p<0.05) H3K9me3 occupancy, while only two bins showed an increase. Of the bins with lower H3K9me3 signal, 122 and 10 were within 5 kb of roo or 297 insertions, respectively, thereby illustrating the impact of Piwi on chromatin states at genomic loci specifically targeted by maternal piRNAs. Furthermore, 205 and 63 individual euchromatic, degron strain-specific TE insertions of both roo and 297, respectively, showed a strong decrease of H3K9me3 levels in Piwi-depleted embryos (Figure 5F, G, Figure 5—figure supplement 1D–F), while H3K9me3 levels at constitutive heterochromatin and on chromosome 4 were not affected (Figure 5H, I). Of note, while roo and 297 TE levels were elevated upon auxin treatment in 2.5–3 hr AEL embryos, transposon expression returned to baseline levels comparable to untreated embryos in 6–7 hr and 7–8 hr AEL time intervals (Figure 5—figure supplement 1G). Taken together, these data strongly indicate a relationship between the deposition of repressive H3K9me3 chromatin marks at transposon insertions and maternally deposited Piwi-piRNA complexes (Figure 5J).

Discussion

Here, we have examined the role of the Piwi-directed coTGS arm of the piRNA pathway during early embryogenesis in Drosophila. By far, most of our insight into the function of piRNAs has derived from studies in germ cells or in the support cells of reproductive tissues. The intriguing observation that piRNAs and their Piwi-family binding partners are maternally deposited has led to speculation regarding potential roles for piRNAs in inter- and transgenerational epigenetic inheritance. Indeed, maternal piRNAs are critical in the suppression of hybrid dysgenesis induced by paternal transmission of I- or P-elements in matings with females that lack these transposons (Brennecke et al., 2008; Khurana et al., 2011). Epigenetic modifications induced by piRNAs appear to aid in piRNA cluster definition in the germline (Akkouche et al., 2017). Additionally, maternally deposited Aub-piRNA complexes have been implicated in embryonic gene regulation (Barckmann et al., 2015; Dufourt et al., 2017; Rouget et al., 2010). Last, maternally inherited piRNAs control transposon expression in interspecies hybrids between Drosophila melanogaster and Drosophila simulans (Kelleher et al., 2012) and regulate the TE tirant in the somatic compartment of the female gonad in D. simulans (Akkouche et al., 2013). Yet, to date, the lack of mechanisms to rapidly deplete maternally deposited PIWI proteins specifically from early embryos has hampered our ability to broadly assess their zygotic roles. By fusing a chemically inducible degron to Piwi, we were able to deplete Piwi-piRNA complexes from dechorionated embryos within less than 30 min of treatment and well before the nuclear accumulation of Piwi that is observed following activation of zygotic transcription.

Though nuclear localization of Piwi correlates with the appearance of its potential targets, nascent transcripts of transposons, it is unclear what triggers movement of Piwi into the somatic nuclei. Notably, nuclear translocation of Piwi lags behind in germ cell precursor nuclei, and this correlates with the observation that these nuclei activate transcription of their genomes later in embryogenesis than somatic nuclei do (Van Doren et al., 1998; Zalokar, 1976). Our dynamic imaging of Piwi localization also revealed that it shuttles out of nuclei during mitosis, as previously observed (Mani et al., 2014). Since other factors of the pathway, namely components of the PICTS/SFiNX complex, remain nuclear, it is likely that Piwi is actively excluded. Several studies have previously shown that nuclear localization of Piwi is conditional upon its binding to a piRNA partners (Klenov et al., 2011; Saito et al., 2009; Yashiro et al., 2018), but we have no indication that Piwi is unloaded and reloaded during mitotic cycles. Rather, we hypothesize that another mechanism regulates the activity of the Piwi nuclear localization signal, though what purpose this might serve and whether it also occurs in germline and follicle cells or is restricted to embryogenesis remains unclear.

It has been suggested that the evolution of the abbreviated piRNA pathway in ovarian follicle cells arose as a consequence of the lifestyle adopted by gypsy family elements. These retrotransposons show their highest expression levels in the support cells surrounding the developing germline. gypsy family elements encode an envelope protein and have been shown to assemble into virus-like particles (Kim et al., 1994; Leblanc et al., 2000; Song et al., 1997). This has led to the hypothesis that their ancestral propagation strategy combined evasion of TE repression mechanisms present in germ cells with an ability to create particles that could infect the germline, where the element could insert into the genome of the developing oocyte following reverse transcription (Kim et al., 1994; Leblanc et al., 2000; Song et al., 1997).

While this remains speculative, it does provoke questions of whether a similar strategy is adopted by roo in the embryo. roo is a quite successful element, as indicated by it being the element with the highest copy number of individual insertions in our sequenced strains (9.4% of all identified TE insertions in the w1118 strain and 9.9% in our degron line). How this is achieved remains mysterious since roo expression is extremely low in the ovary. Moreover, roo does not appear to be a target of the ovarian piRNA pathway since its gonadal expression is not increased nor does its HP1a enrichment and H3K9me3 levels change in piRNA pathway mutant animals (Figure 5—figure supplement 1H, ISenti et al., 2015; Wang and Elgin, 2011). In the embryo, the expression of roo is restricted to somatic cells, especially cell lineages giving rise to the adult mesoderm. Previous studies have suggested that roo expression is activated by twist (twi) and snail (sna), which are highly expressed in the embryogenic mesoderm (Brönner et al., 1995), and this is consistent with the spatial expression pattern that we also observe. roo expresses the full repertoire of proteins needed to form virus-like particles, and its high expression levels (exceeding 1% of the transcriptome at its peak) might enable a strategy of propagation by infection in trans, even if rates of transmission to the germ cell precursors are relatively low.

Our data strongly suggests that only maternally deposited piRNAs engage Piwi in the soma of the developing embryo. Since roo is not regulated by the piRNA pathway in the ovary, evolutionary pressures must have driven the development of a set of maternal instructions that are inherited to dampen the burst of roo expression in the developing embryo. Indeed, 16% of maternally deposited piRNAs target roo. Though there are differences between the populations of piRNAs observed in ovaries as compared to embryos, this is mostly driven by the presence of follicle cell piRNAs in in samples taken from the gonad. In our small RNA analyses, we collapse all stages of oogenesis. Thus, it is not clear whether the composition of piRNA populations shifts as the ovariole matures and whether any such shifts enrich for embryonically expressed elements late in oocyte maturation. Irrespective, a set of instructions from maternal piRNA clusters clearly builds a transgenerational ability to recognize roo and other embryonically expressed elements.

Consistent with its recognition by Piwi-piRNA complexes and recruitment of the PICTS/SFiNX complex, H3K9me3 marks build at presumably active, euchromatic roo insertions as embryos progress toward stage 13 (10 hr AEL). The peak of H3K9me3 abundance lags about 2 hr behind the peak of transcription. Since we have little other information on the dynamics of piRNA-mediated silencing, it is not clear whether this is an expected observation or whether there may be mechanisms that antagonize the ability of the piRNA pathway to immediately recognize and direct heterochromatinization of expressed roo insertions. Of note, we see a shorter interval between the embryonic peak of 297 expression and its peak of H3K9me3 accumulation. Imposition of a repressive chromatin state on roo is transient during somatic development. By 13 hr AEL, H3K9me3 peaks over roo insertions have disappeared, but roo expression has not returned. The lack of H3K9me3 also correlates with the absence of critical piRNA pathway proteins in the soma. Overall, this suggests that both the expression of these TEs and the host response via small RNAs is transient. While our data provide compelling evidence of the accumulation of repressive chromatin marks at presumably actively transcribing TE insertions, it does not carry spatial information about the precise cell types affected by H3K9me3 deposition.

Though zygotic depletion of maternal Piwi during early embryogenesis does produce a statistically significant change in roo expression (roughly twofold), this transposon remains highly expressed reaching up to 1% of the entire transcriptome in control animals, despite being targeted by the piRNA pathway. This provokes the question of whether targeting of roo by the piRNA pathway is biologically relevant. In favor of this hypothesis are several observations. Roo is expressed in ovaries at very low levels, yet the hallmarks of piRNA-dependent silencing, specifically H3K9me3, are absent from euchromatic roo insertions. This strongly indicates that roo is not controlled by the piRNA pathway in this tissue. Nonetheless, ovaries produce abundant roo piRNAs, and these are overwhelmingly in the antisense orientation. Additionally, the only uni-strand cluster expressed in germ cells, cluster 20A, has collected roo insertions in the antisense orientation. These piRNAs are abundantly maternally transmitted (16% of all piRNAs in embryos) and persist throughout the time during early embryogenesis when high-level roo expression is proposed to be driven by mesodermal transcription factors. An argument against biological significance is the lack of a clearly observable phenotype in flies following embryonic depletion of maternal Piwi. However, technical limitations enable us to only measure impacts within a single generation. It is entirely possible that the fitness cost of roo occupying 2% of the embryonic transcriptome might be substantial over time or in conditions flies might experience in the wild compared to the controlled rearing conditions in the lab.

Perhaps more importantly, our study demonstrates that recognition of a locus by the piRNA pathway does not necessarily impose the creation of a mitotically heritable epigenetic state. This is consistent with observations made by many groups in follicle cells wherein heterochromatin-mediated silencing of somatic transposons requires the continuous presence of the piRNA machinery (Batki et al., 2019; Clark et al., 2017; Dönertas et al., 2013; Fabry et al., 2019; Muerdter et al., 2013; Murano et al., 2019; Ohtani et al., 2013; Saito et al., 2009; Sienski et al., 2015; Sienski et al., 2012; Zhao et al., 2019). These data are at odds with prior observations and speculation that the maintenance of silenced epigenetic states can be primed by Piwi but maintained in a Piwi-independent mechanism throughout adult life (Gu and Elgin, 2013). The prior study noted these effects after only a 50% reduction in embryonic Piwi protein or RNA, using either of two different strategies. Though our induced proteolytic degradation strategy is unlikely to completely remove all Piwi protein, Piwi was reduced to levels that are undetectable by western blotting (Figure 5B) or via the fluorescence of its fused GFP (Figure 5C), which would, if anything, be expected to produce a more profound impact. While it is difficult to reconcile our observations with the interpretation of the prior study, there were substantial differences in what was measured and in how the measurements were made (i.e., a different set of genomic loci was studied in different Drosophila strains by different methods). The prior work made use of position-effect reporters integrated into pericentromeric heterochromatin and indicated that the expression of these in adults was sensitive to Piwi depletion in the embryo. We did not examine such a reporter, and so it remains possible that H3K9me3 marks deposited in a Piwi-dependent fashion in regions adjacent to large domains of Piwi-independent H3K9 methylation might behave differently than those deposited on active, euchromatic transposons. Considerable consistency between the two studies can be found in the prior observation that HP1a occupancy in embryos did not change substantially on several transposons studied (maximum of twofold on HeT-A) (Gu and Elgin, 2013). The prior study also failed to note large-scale changes in HP1a distribution, as a proxy for methylated H3K9, and reported only very small changes in HP1a levels on a few transposon families, as assayed in larvae by ChIP-array measurements, which collapse all insertions of a given family into a single data point. The transposons that we do identify as sensitive to Piwi during early embryogenesis do not overlap with those identified in the previous study as being mildly affected by reductions in Piwi at a later developmental stage (data not shown). This is actually consistent with our observation that the effects of profound Piwi depletion on roo and other TEs are transient during embryogenesis. Thus, it seems that the data themselves diverge less between the two studies than do the conclusions drawn. Of note, another recent report found a mild upregulation of transposons in pre-ZGA embryos upon maternal depletion of Piwi; however, this work relied on germ cell-specific knockdown during late stages of oogenesis rather than direct protein depletion in the embryo, thus at least some of the observed effects could stem from TE mobilization during ovary development (Gonzalez et al., 2021).

A recent detailed and elegant study examined the patterns of H3K9me3 accumulation during early embryogenesis in Drosophila miranda (Wei et al., 2021). Though overall, deposition of H3K9me3 did not correlate with the abundance of maternally deposited piRNAs, a set of the earliest heterochromatin nucleating elements were associated with abundant piRNAs. These targeted elements had high copy numbers and showed evidence of recent transposition activity, suggesting that they were under evolutionary pressure for robust silencing both in the ovary and the soma. It should be noted that precise nucleation sites did not necessarily overlap with abundant piRNAs, suggesting that multiple silencing mechanisms might collaborate to repress these transposon families.

Considered as a whole, our data are consistent with a role for maternally deposited piRNAs in the recognition of transposon families that have focused their expression and activity during early embryogenesis. However, our data does not support a model wherein the piRNA pathway nucleates heritable patterns of heterochromatin formation that broadly pattern the epigenetic landscape of the adult Drosophila soma, and this is perhaps consistent with our failure to observe consequential developmental abnormalities upon negation of embryonic Piwi function.

Materials and methods

Key resources table
Reagent type
(species) or resource
DesignationSource or referenceIdentifiersAdditional information
Gene (Drosophila melanogaster)nxf2FlyBaseFBgn0036640
Gene (Drosophila melanogaster)panxFlyBaseFBgn0034617
Gene (Drosophila melanogaster)piwiFlyBaseFBgn0004872
Gene (Drosophila melanogaster)rooFlyBaseFBgn0043856
AntibodyAnti-GFP (chicken polyclonal)AbcamCat# ab13970
RRID:AB_300798
WB (1:5000)
IF (1:1000)
AntibodyAnti-Piwi (rabbit polyclonal)DOI: 10.1016/j.cell.2007.01.043N/AWB (1:1000)
IF (1:500)
AntibodyAnti-Tubulin (mouse monoclonal)AbcamCat# ab44928, RRID:AB_2241150WB (1:5000)
AntibodyAnti-Myc tag (rabbit polyclonal)AbcamCat# ab9106, RRID:AB_307014WB (1:1000)
AntibodyAnti-H3K4me2 (rabbit polyclonal)Merck MilliporeCat# 07-030, RRID:AB_310342
(Lot# 2971019)
ChIP (1:50)
AntibodyAnti-H3K9me3 (rabbit polyclonal)Active MotifCat# 39161, RRID:AB_2532132
(Lot# 15617003)
ChIP (1:50)
AntibodyAnti-Mouse IgG Alexa Fluor 488 (goat polyclonal)Thermo Fisher ScientificCat# A-11029, RRID:AB_2534088IF (1:500)
AntibodyAnti-Mouse IgG Alexa Fluor 555 (goat polyclonal)Thermo Fisher ScientificCat# A-21424, RRID:AB_141780IF (1:500)
AntibodyAnti-Rabbit IgG Alexa Fluor 647 (goat polyclonal)Thermo Fisher ScientificCat# A-21245, RRID:AB_2535813IF (1:500)
Commercial assay or kitRIPA Lysis and Extraction BufferThermo Fisher ScientificCat# 89901
Commercial assay or kitcOmplete, Mini, EDTA-free Protease Inhibitor CocktailSigma-AldrichCat#11836170001
Commercial assay or kitRNasin Plus RNase InhibitorPromegaCat# N2615
Commercial assay or kitBlood and Cell Culture DNA Mini KitQiagenCat# 13323
Commercial assay or kitRNeasy Mini KitQiagenCat# 74106
Commercial assay or kitNEBNext Poly(A) mRNA Magnetic Isolation ModuleNEBCat# E7490L
Commercial assay or kitNEBNext Ultra Directional RNA Library Prep Kit for IlluminaNEBCat# E7420L
Commercial assay or kitNEBNext Ultra II DNA Library Prep Kit for IlluminaNEBCat# E7645L
Commercial assay or kitInvitrogen SuperScript IV Reverse TranscriptaseThermo Fisher ScientificCat# 18090050
Commercial assay or kitIndole-3-acetic acid sodium saltSigma-AldrichCat# I5148-10G
Commercial assay or kitPierce 16% formaldehyde (w/v), methanol-freeThermo Fisher ScientificCat# 28908
Commercial assay or kitMinElute PCR Purification KitQiagenCat# 28004
Software, algorithmFijiImageJRRID:SCR_002285
Software, algorithmZeiss ZEN Imaging SoftwareZeissRRID:SCR_018163
Software, algorithmProteome Discoverer 2.1Thermo Fisher ScientificRRID:SCR_014477
Software, algorithmRRCoreTeamN/A
Software, algorithmSTARDOI:10.1093/bioinformatics/bts635RRID:SCR_015899
Software, algorithmTEMPhttps://github.com/JialiUMassWengLab/TEMPZhuang et al., 2014RRID:SCR_001788
Software, algorithmProdigalhttps://github.com/hyattpd/ProdigalN/A
Software, algorithmDEseq2DOI:10.1186/s13059-014-0550-8RRID:SCR_015687
Software, algorithmImage Studio LiteLI-CORRRID:SCR_013715

Fly stocks and handling

Request a detailed protocol

All flies were kept at 25°C on standard cornmeal or propionic food. Flies expressing GFP-Nxf2 from the endogenous locus were generated by CRISPR/Cas9 (Fabry et al., 2019). Transgenic flies carrying a BAC transgene expressing GFP-Panx and GFP-Piwi were generated by the Brennecke lab (Handler et al., 2013; Sienski et al., 2015) and obtained from the Vienna Drosophila Resource Center. Control w1118 flies were a gift from the University of Cambridge Department of Genetics Fly Facility, and flies expressing His2Av-RFP were a gift from the St Johnston lab. Flies between 3 and 14 days after hatching were used for experiments.

Generation of fly strains

Request a detailed protocol

GFP-AID-Piwi knock-in flies were generated by CRISPR/Cas9 genome engineering. Homology arms of 1 kb flanking the targeting site were cloned into pUC19 by Gibson Assembly and co-injected with pCFD3 (Addgene # 49410) containing a single-guide RNA (Port et al., 2014) into embryos expressing vas-Cas9 (Bloomington Drosophila Stock Center # 51323). Flies expressing OsTIR1 under the D. melanogaster Ubiquitin-63E promoter were generated by phiC31 integrase-mediated transgenesis by injection of plasmids containing expression cassettes for proteins into embryos of genotype ‘y w P[y[+t7.7]=nos-phiC31\int.NLS]X #12; +; P[y[+t7.7]=CaryP]attP2,’ resulting in transgene integration on chromosome 3. Microinjection and fly stock generation was carried out by the University of Cambridge Department of Genetics Fly Facility. Transgenic and knock-in flies were identified by genotyping PCRs and confirmed via Sanger sequencing.

Western blot

Request a detailed protocol

Protein concentration was measured using a Direct Detect Infrared Spectrometer (Merck). 20 µg of proteins were separated on a NuPAGE 4–12% Bis-Tris gel (Thermo Fisher Scientific). Proteins were transferred for 2 hr at 100 V, 400 mA, 100 W on an Immun-Blot Low Fluorescent PVDF Membrane (BioRad) and blocked for 1 hr in 1× LI-COR TBS Blocking Buffer (LI-COR). Primary antibodies were incubated overnight at 4°C. LI-COR secondary antibodies were incubated for 45 min at room temperature (RT) and images acquired with an Odyssey CLx scanner (LI-COR).

Drosophila ovary immunofluorescence

Request a detailed protocol

Fly ovaries were dissected in ice-cold Phosphate-buffered saline (PBS) and fixed in 4% PFA diluted in PBS for 15 min at room temperature while rotating. Following three rinses and three 10 min washing steps in PBS-Tr (0.3% Triton X-100 in PBS), ovaries were blocked for 2 hr at RT while rotating in PBS-Tr + 1% BSA. Primary antibody incubation was carried out in blocking buffer overnight at 4°C while rotating, followed by three washing steps for 10 min each in PBS-Tr. All following steps were performed in the dark. Secondary antibodies were diluted in blocking buffer and incubated overnight at 4°C while rotating. Ovaries were washed four times for 10 min in PBS-Tr and stained with 0.5 µg/ml DAPI (Thermo Fisher Scientific) for 10 min. Following two additional washing steps for 5 min in PBS, ovaries were mounted in ProLong Diamond Antifade Mountant (Thermo Fisher Scientific) and imaged on a Leica SP8 confocal microscope using a 40× Oil objective.

Drosophila embryo immunofluorescence

Request a detailed protocol

Embryos were collected and dechorionated in 50% bleach for 1 min. Embryos were transferred into 1 ml fixing solution (600 µl 4% PFA in PBS, 400 µl n-heptane) and fixed for 20 min at RT while rotating. The lower aqueous phase was removed and 600 µl methanol added. The tube was vortexed vigorously for 1 min to remove vitelline membranes. Embryos were allowed to sink to the bottom of the tube and all liquid was removed, followed by two washes with methanol for 1 min each. Embryos were stored at −20°C at least overnight or until further processing. In order to rehydrate embryos, three washes each 5 min with PBST (0.1% Tween20 in PBS) were performed and embryos blocked for 1 hr at RT in PBST + 5% BSA. Primary antibodies were incubated overnight at 4°C while rotating in blocking buffer followed by 3 washes for 15 min each with PBST. All following steps were performed in the dark. Secondary antibodies were diluted in blocking buffer and incubated at RT for 2 hr. Embryos were rinsed three times and washed two times for 15 min. Nuclei were stained with 0.5 µg/ml DAPI (Thermo Fisher Scientific) for 10 min. Following two additional washing steps for 5 min in PBS, embryos were mounted in ProLong Diamond Antifade Mountant (Thermo Fisher Scientific) and imaged on a Leica SP8 confocal microscope using a 40× Oil objective.

Combined RNA-FISH and IF in embryos

Request a detailed protocol

Embryos were collected, dechorionated, and processed as described above until secondary antibody incubation. For all steps containing BSA addition, RNAsin Plus RNase inhibitors were added (1:1000, Promega). Following secondary antibody incubation, cells were washed three times for 15 min in PBST at RT while rotating. Embryos were fixed in 4% PFA in PBST solution for 25 min and rinsed three times with PBST for 5 min each. Embryos were pre-hybridized in 100 µl hybridization buffer (50% formamide, 5× saline-sodium citrate (SSC), 9 mM citric acid pH 6.0, 0.1% Tween20, 50 µg/ml heparin, 1× Denhardt’s solution [Sigma-Aldrich], 10% dextran sulfate) for 2 hr at 65°C. Probes were hybridized in hybridization buffer supplemented with 2 nM of each FISH probe at 45°C overnight. Following washing twice with probe wash buffer (50% formamide, 5× SSC, 9 mM citric acid pH 6.0, 0.1% Tween20, 50 µg/ml heparin) for 5 min and twice for 30 min at 45°C, embryos were incubated in amplification buffer (5× SSC, 0.1% Tween20, 10% dextran sulfate) for 10 min at RT. Hairpins were prepared as described above and embryos incubated in fresh amplification buffer with 120 nM of each probe at RT overnight in the dark. Embryos were washed twice with 5× SSC for 5 min. Nuclei were stained with 0.5 µg/ml DAPI diluted in 2× SSC for 15 min. Following washing twice with 2× SSC for 10 min, embryos were mounted in ProLong Diamond Antifade Mountant (Thermo Fisher Scientific) and imaged on a Leica SP8 confocal microscope using a 40× Oil objective.

Light-sheet fluorescent microscopy (LSFM) of Drosophila embryos

Request a detailed protocol

Embryos were collected and dechorionated as described above. 1 ml of 1% low melting point (LMP) agarose was prepared and embryos transferred into capillaries (catalog # 100003476381, Brand) using a fitting plunger. Embryos were attempted to be positioned vertically in the capillary by twisting until agarose solidified. Capillaries were stored in PBS at RT until imaging. LSFM was performed on a Zeiss Lightsheet Z.1 (Carl Zeiss, Germany) at 25°C with a 20×/1.0 Plan-Apochromat water-immersion objective lens. Embryos were lowered carefully out of the capillary into the imaging chamber filled with PBS and positioned directly between the light-sheet illumination objectives (10×/0.2, left and right). Z-stack images for GFP and RFP (excitation at 488 and 561 nm, respectively) were acquired every 2 min for >10 hr with the lowest possible laser intensity (2.5% for GFP and 10% for RFP). Generated data was analyzed in Zeiss ZEN Imaging Software and Fiji (ImageJ).

ChIP-seq for Drosophila embryos

Request a detailed protocol

50 µl of embryos were collected and dechorionated as described above and transferred in 1 ml Crosslinking solution (1% formaldehyde in PBS, 50% n-heptane) and vortexed on high speed for precisely 15 min. 90 µl 2.5M glycine solution was added to quench excess formaldehyde and incubated for 5 min at RT while rotating. Embryos were allowed to sink to the bottom of the tube and all liquid was removed. Embryos were washed three times for 4 min with ice-cold buffer A (60 mM KCl, 15 mM NaCl, 4 mM MgCl2,15mM HEPES pH 7.6, 0.5% DTT, 1× PI) supplemented with 0.1% Triton X-100 (A-Tx buffer). All liquid was removed and embryos flash-frozen and stored at −80°C until further processing. Crosslinked embryos were transferred to a 2 ml Dounce homogenizer in 1 ml A-TBP (Buffer A + 0.5% Triton X-100). Following an additional washing step with A-TBP, embryos were lysed in 1 ml A-TBP using 10 strokes with a tight-fitting pestle. Lysate was centrifuged at 3200 g for 10 min at 4°C and supernatant removed. The pellet was resuspended in 1 ml Lysis buffer (15 mM HEPES, 140 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, 1% Triton, 0.5 mM DTT, 10 mM sodium butyrate, 0.1% sodium deoxycholate, 1× PI) and incubated at 4°C for 15 min while rotating. Following centrifugation at 3200 g for 10 min at 4°C, the pellet was washed twice with Lysis buffer and centrifuged again. All liquid was removed, and the pellet resuspended in 300 µl LB3 (10 mM Tris-HCl, pH 8, 100 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, 0.1% Na-Deoxycholate, 0.5% N-lauroylsarcosine, 1× PI). Sonication was carried out using the Bioruptor pico (Diagenode) for six cycles (30 s on, 30 s off settings). Debris was removed from the chromatin-containing supernatant by spinning down at full speed for 10 min at 4°C. Prepared chromatin was either frozen down in liquid nitrogen and stored at −80°C or used immediately. 5% of the chromatin fraction was flash-frozen as an input sample. 100 µl magnetic Protein A-coupled Dynabeads (Thermo Fisher Scientific) were washed three times in 1 ml blocking solution (0.2% BSA in PBS). The blocking solution was removed using a magnetic rack. 5 µl of anti-H3K9me3 or anti-H3K4me2 polyclonal antibody was diluted in 250 µl blocking solution and incubated with 100 µl washed beads by rotating at 4°C for at least 4 hr up to overnight. The supernatant was removed and beads washed three times in blocking solution as described above. The chromatin solution was added to the beads and incubated at 4°C while rotating overnight. Following four washing steps for 2 min each using ice-cold Lysis Buffer (15 mM HEPES, 140 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, 1% Triton, 0.5 mM DTT, 10 mM sodium butyrate, 0.1% sodium deoxycholate, 1× PI, 0.05% SDS), beads were washed two additional times with ice-cold 1× TE buffer. All liquid was removed and beads resuspended in 200 µl Elution buffer (50 mM Tris-HCl, pH 8; 10 mM EDTA; 1% SDS). Input samples were thawed and brought up to 200 µl with Elution buffer. Samples were transferred into 200 µl Maxymum Recovery PCR tubes (Axygen) and incubated at 65°C for 16–18 hr for reverse crosslinking. RNA contamination was removed by adding 200 µl 1× TE buffer and 8 µl of 1 mg/ml RNase A (Ambion) to ChIP and input samples followed by incubation at 37°C for 30 min. Proteins were digested using 4 µl Proteinase K (800 U/ml, NEB) and incubation at 55°C for 2 hr. Reverse crosslinked DNA was recovered using the MinElute PCR purification Kit (Qiagen) according to the manufacturer’s recommendation and eluted in 15 µl nuclease-free water. DNA recovery was verified and quantified using 1 µl for Bioanalyzer (Agilent) electrophoresis.

ChIP-seq for Drosophila ovaries and heads

Request a detailed protocol

50 Drosophila ovaries were dissected in ice-cold PBS. Heads were dislodged by pouring liquid nitrogen over whole flies in a dish followed by shaking and collecting 50 µl broken-off heads in 1.5 ml tube. Samples were homogenized in 100 µl Buffer A1 (60 mM KCl, 15 mM NaCl, 4 mM MgCl2,15 mM HEPES pH 7.6, 0.5% DTT, 0.5% Triton X-100, 1× PI) using a rotating pestle. The volume was brought up to 1 ml with buffer A1 and formaldehyde added to a final concentration of 1.8% for crosslinking. Samples were rotated for exactly 15 min at RT and glycine solution added to a final concentration of 225 mM. Samples were allowed to rotate for an additional 5 min and were centrifuged at 4000 g for 5 min at 4°C. The supernatant was removed, the pellet washed twice with buffer A1 and once with buffer A2 (140 mM NaCl, 15 mM HEPES pH 7.6, 1 mM EDTA, 0.5 mM EGTA, 1% Triton X-100, 0.5 mM DTT, 0.1% sodium deoxycholate, 10 mM sodium butyrate, 1× PI) at 4°C. The pellet was then resuspended in 100 µl A2 buffer supplemented with 1% SDS and 0.5% N-laurosylsarcosine and incubated at 4°C for 2 hr while shaking vigorously. Lysate was sonicated using the Bioruptor pico for 16 cycles (30 s on, 30 s off). The sonicated lysate was spun at full speed for 10 min at 4°C and the supernatant transferred to a new tube. The volume was brought up to 1 ml with A2 buffer supplemented with 0.1% SDS. Chromatin used for ChIP was precleared with 15 µl washed Protein A Dynabeads and incubated with antibody coated beads as described above. Further steps were performed as described above for embryo ChIP.

Piwi-IP from Drosophila ovaries and embryos for small RNA-seq

Request a detailed protocol

Piwi-piRNA complexes were isolated from ovaries or from 0 to 8 hr control w1118 embryos similar to previous reports (Hayashi et al., 2016; Mohn et al., 2015). In short, 100 µl of ovaries were dissected in PBS on ice. 100 µl of embryos were collected on grape juice agar plates and transferred to a mesh strainer. Following dechorionation in 50% bleach, embryos were washed under running tap water for at least 1 min or until bleach smell disappeared. Ovary and embryo samples were washed twice with ice-cold PBS and homogenized in 1 ml lysis buffer (10 mM HEPES pH 7.3, 150 mM NaCl, 5 mM MgCl2, 10% glycerol, 1% Triton x-100, 1 mM DTT, 1 mM EDTA, 0.1 mM PMSF, 1× PI, 1:1000 RNasin [Promega]) using a 2 ml Dounce homogenizer. Material was lysed with five strokes with a loose pestle and five strokes with a tight pestle on ice. Lysate was incubated for 1 hr at 4°C while rotating and centrifuged at full speed for 10 min to pellet debris. Supernatant was transferred to a new tube and protein concentration determined by Direct Detect (Millipore). 1 mg of lysate per immunoprecipitation was used for the following steps. 50 µl Protein A Dynabeads (Thermo Fisher Scientific) were washed with lysis buffer three times for 3 min each. Washed beads were resuspended in 400 µl lysis buffer and 5 µl anti-Piwi (Hannon Lab) or rabbit IgG antibodies (Abcam, ab37415) added. Following overnight incubation at 4°C while rotating, beads were washed three times for 5 min in 500 µl lysis buffer. Antibody-coupled beads were added to lysates and volume brought up to 1 ml with lysis buffer. The solution was incubated at 4°C overnight while rotating. Supernatant was removed and saved for quality control western blotting analysis. Beads were washed six times for 10 min with 1 ml wash buffer (10 mM HEPES pH 7.3, 150 mM NaCl, MgCl2, 10% glycerol, 1% Empigen BB Detergent [Merck], 1× PI). For the first wash, 1 µl RNasin was added to the wash buffer and tubes were changed between each wash. 10% of beads were set aside for quality control and 90% resuspended in 1 ml Trizol (Thermo Fisher Scientific) and stored at −80°C until further processing.

Whole-genome sequencing

Request a detailed protocol

100 flies were collected in a 1.5 ml tube and frozen at −80°C for at least 1 hr. High molecular weight genomic DNA was isolated using the Blood and Cell Culture DNA Mini kit (Qiagen). Flies were homogenized using a rotating pestle on ice for 1 min. 700 µl G2 and 50 µl Proteinase K (800 U/ml, NEB) were added to each tube and incubated at 50°C for 2 hr with occasional tube inversions. Tubes were spun at 5000 g for 10 min at 4°C and supernatant transferred to new tube avoiding debris. A Qiagen Genomic-tip 20/G was equilibrated with 1 ml QBT buffer and allowed to empty by gravity flow. The supernatant containing digested proteins and genomic DNA was added to the column and allowed to flow through. The column was washed three times with 1 ml QC buffer. Elution was carried out with 1 ml QF elution buffer and repeated once. Flow through was transferred to two new tubes (1 ml each) and 700 µl isopropanol added. Tubes were inverted 10 times and centrifuged at full speed for 15 min at 4°C. The pellet was washed with 70% ethanol twice and air-dried for 5 min. 25 µl RNase-free water was added and DNA resuspended by flicking tube gently several times while incubating at 37°C for 2 hr. DNA was stored at 4°C. DNA was sheered using a Covaris S220 (Covaris). 3 µg of genomic DNA was diluted in RNase-Free water and transferred to a AFA Fiber Crimp-Cap (PN520052, Covaris) microtube. Sonication was carried out with the following settings: peak incident power (W) 105, duty factor 5%, cycles per burst 200, and treatment time 80 s. This resulted in sheared DNA fragments peaking at 500 bp. DNA was recovered using the QIAquick PCR Purification Kit (Qiagen).

RNA extraction

Request a detailed protocol

RNA for RNA-seq and qRT-PCR experiments was isolated using the RNeasy Mini kit (Qiagen) with on-column DNA digestion (RNase-free DNase Set, Qiagen) according to the manufacturer’s recommendations. RNA for small RNA-seq experiments were isolated using Trizol following the manufacturer’s instructions.

Library preparation

Request a detailed protocol

1 µg of total RNA was used as input material for RNA-seq library preparation. The NEBNext Poly(A) mRNA magnetic Isolation Module (NEB) was used to isolate poly(A) RNAs. Libraries were generated with the NEBNext Ultra Directional RNA Library Prep kit for Illumina (NEB) according to the manufacturer's instructions. Small RNA libraries were generated as described previously (Jayaprakash et al., 2011). In short, 19- to 28-nt-sized small RNAs were purified by PAGE from Piwi-bound RNA isolated from ovaries or embryos. Next, the 3' adapter (containing four random nucleotides at the 5' end) was ligated overnight using T4 RNA ligase 2, truncated KQ (NEB). Following recovery of the products by PAGE purification, the 5' adapter (containing four random nucleotides at the 3' end) was ligated to the small RNAs using T4 RNA ligase (Abcam) for 1 hr. Small RNAs containing both adapters were recovered by PAGE purification, reverse transcribed, and PCR amplified. WGS libraries were generated using the NEBNext Ultra II DNA Library Prep kit (NEB) according to the manufacturer’s recommendation with 1 µg input material. Three PCR amplification cycles were performed. Libraries were quantified using the Library Quantification Kit for Illumina (Kapa Biosystems).

Next-generation sequencing

Request a detailed protocol

Sequencing was performed by the Genomics Core facility at CRUK CI. RNA-seq, ChIP-seq, and small RNA-seq libraries were sequenced on an Illumina HiSeq 4000 according to the manufacturer’s recommendations using single-end 50 bp runs. WGS libraries were sequenced with paired-end 150 bp runs on Illumina HiSeq 4000 or NovaSeq.

Quantitative reverse transcription polymerase chain reaction (qRT-PCR)

Request a detailed protocol

Reverse transcription was performed using the SuperScript IV reverse transcriptase Kit (Thermo Fisher Scientific) with 1 µg of total RNA. qRT-PCR was performed on a QuantStudio Real-Time PCR Light Cycler (Thermo Fisher Scientific) in technical triplicates. Expression of targets was quantified using the ddCT method (Livak and Schmittgen, 2001). Fold-change was calculated as indicated in the figure legends and normalized to rp49. All primers are listed in Supplementary file 1.

Protein isolation from whole embryos and quantitative mass spectrometry

Request a detailed protocol

100 µl of control w1118 embryos for time points 0–2 hr, 5–7 hr, and 10–12 hr AEL were collected in three biological replicates on agar plates and dechorionated. Embryos were then lysed in lysis buffer (0.1% SDS, 0.1 M triethylammonium bicarbonate [TEAB], 1× Halt Protease and Phosphatase Inhibitor [Thermo Fisher Scientific]) using a rotating pestle on ice for 2 min or until entirely homogenized. Lysate was heated for 5 min at 90°C and probe sonicated for 20 s (20% power with pulse of 1 s). Debris was pelleted by centrifugation at full speed for 10 min at 4°C and supernatant transferred to a new tube. Protein concentration was measured using Bradford Assay (BioRad). 100 µg protein was digested with trypsin overnight at 37°C. TMT chemical isobaric labeling was performed as described previously (Papachristou et al., 2018). Peptide fractions were analyzed on a Dionex Ultimate 3000 UHPLC system coupled with the nano-ESI Fusion Lumos mass spectrometer (Thermo Fisher Scientific).

Treatment of embryos for auxin-induced degradation

Request a detailed protocol

Embryos were collected for 30 min and dechorionated. Control embryos were transferred into a fine mesh strainer placed in a plastic dish and submerged in PBS. 1 M auxin solution was generated by diluting Indole-3-acetic acid (IAA), a highly permeable small molecule as recently shown for Caenorhabditis elegans embryos (Zhang et al., 2015), in water and stored protected from light at −20°C. Auxin-treated embryos were submerged in PBS with indicated auxin concentrations. Embryos were placed at 25°C for appropriate times and harvested for RNA experiments by transferring into 1 ml Trizol followed by RNA extraction. Embryos used for ChIP-seq were processed as described above.

RNA-seq and ChIP-seq analysis

Request a detailed protocol

Raw fastq files contained 50 bp reads. The first and the last two bases of all reads were trimmed using fastx_trimmer (http://hannonlab.cshl.edu/fastx_toolkit/). Reads were first aligned to the consensus sequence for all D. melanogaster transposons using STAR (Dobin et al., 2013) allowing random allocation of multimappers. Unmapped reads were further aligned to D. melanogaster genome release 6 (dm6) keeping uniquely mapping reads. Generated bam files for RNA-seq were further split in reads originating from sense and antisense genomic strands using samtools view options -f 0x10 and -F 0x10 for sense and antisense reads, respectively (Li et al., 2009c). Indexes were generated using samtools index function. Coverage files were generated using bamCoverage with normalization mode --normalizeUsing CPM (Ramírez et al., 2014) and applying a scaling factor (--scaleFactor). Scaling factors for individual files were calculated by dividing the sum of mapped reads contained in the file by the sum of all transposon and dm6 mapping reads of the corresponding library. Reads mapping to protein-coding genes were counted with htseq (Anders et al., 2015) using a feature file downloaded from Ensembl release BDGP6.22. Reads mapping to individual transposons were counted with a custom script using samtools idxstats function to extract reads mapping to individual sequences of the reference genome/transposon consensus sequence.

Count files for RNA-seq time-course experiments generated as described above were normalized to rpm to account for differences in library size and allow comparability between time points. Heatmaps displaying expression profiles of genes and transposons during embryogenesis show the mean expression values of the biological replicates, while bar graphs display the individual data points as well as the mean expression and standard deviation. Bar graphs and heatmaps were plotted in R using ggplot2.

RNA-seq experiments comparing auxin- and PBS-treated embryos of the same stage and collection were analyzed using differential expression quantification methods allowing for statistical evaluation of differences between RNA output as a direct result of auxin treatment. Differential expression analysis was performed using DESeq2 (Love et al., 2014). MA plots show base mean RNA expression across conditions and were calculated as previously described by Love and colleagues.

ChIP-seq reads were normalized by library size and rpm calculated for concatenated replicates using the deepTools2 bamCoverage function (Ramírez et al., 2016) with bin size 10 bp. MA plots displaying H3K9me3 signal intensity fold-changes between auxin-treated and control samples were calculated using DESeq2 for individual replicates (n = 3). Metaplots flanking euchromatic transposon insertion sites were calculated using computeMatrix scale-region function from deepTools2 with bin size 10 bp. All scripts used for sequencing analysis are available on GitHub (https://github.com/mhf27/hannon_roo_fabry2021, copy archived at swh:1:rev:f088572638701e0ae6f13d9e025642b9476146b5Fabry, 2021).

Small RNA-seq analysis

Request a detailed protocol

Reads from small RNA-seq libraries were adapter clipped using fastx_clipper with settings -Q33 -l 15 -a AGATCGGAAGAGCACACGTCT. The first and last four bases of adapter clipped reads were trimmed using seqtk trimfq (https://github.com/lh3/seqtkLi, 2021). Only high-quality reads with length between 19 and 31 bp were used for further analysis. Small RNAs were aligned as described above and size profiles plotted in R.

Generation of annotation files

Request a detailed protocol

Mappability track for dm6 with 50 bp resolution was calculated according to a previously published method (Derrien et al., 2012). The de novo transposon insertion calling for the homozygous control w1118 strain and our line carrying both GFP-AID-Piwi and OsTIR1 was performed using the TEMP algorithm (Zhuang et al., 2014). In brief, ~500 bp genomic DNA fragments were amplified and sequenced generating 150 bp paired-end reads, which were aligned to dm6 using BWA (Li and Durbin, 2009b). Reads with only one mate aligned to dm6 were extracted and the unmapped mate uniquely aligned to transposon consensus sequences in order to ensure correct directionality calling. Calculated insertion sites were extracted from generated GTF files if they were supported by reads on both sides (1p1). Transposon insertion files containing coordinates as well as statistical information have been submitted to GEO (GSE160778). Euchromatic regions (chr2R:6460000–25286936, chr2L:1–22160000, chr3L:1–23030000, chr3R:4200000–32079331, chrX:250000–21500000) were defined by measuring H3K9me3 signal genome-wide in sliding windows of 10 kb bins and calculating signal enrichment over input. We identified a total of 632 euchromatic TE insertions in w1118 and 1738 in our degron strain (GFP-AID-Piwi; OsTIR1). The protein database used to identify peptides from Drosophila genes and transposons was generated by merging an existing database downloaded from FlyBase (dmel-r6.24.fa) with translated ORFs of transposons. ORFs were predicted and translated using prodigal (https://github.com/hyattpd/ProdigalHyatt, 2020). ORFs with less than 300 amino acids were removed using seqtk -L 300 and the file was converted to fasta format.

Protein domain prediction

Request a detailed protocol

Functional analysis of protein sequences was performed using the InterPro web application (https://www.ebi.ac.uk/interpro/). Protein domains and families for ORF encoded by roo transcripts were predicted using default settings.

Mass spectrometry raw data processing

Request a detailed protocol

Raw data files were processed according to previous reports (Papachristou et al., 2018). Spectral .raw files were analyzed with the SequestHT search engine on Thermo Scientific Proteome Discoverer 2.1 for peptide and protein identification. Data was searched against a modified FlyBase protein database with the following parameters: precursor mass tolerance 20 ppm, fragment mass tolerance 0.5 Da. Dynamic modifications were oxidation of methionine residues (+15.995 Da) and deamidation of asparagine and glutamine (+0.984 Da), and static modifications were TMT6plex at any amino-terminus, lysine (+229.163 Da), and methylthio at cysteine (+45.988 Da). The Reporter Ion Quantifier node included a TMT 6plex (Thermo Scientific Instruments) Quantification Method, for MS3 scan events, HCD activation type, integration window tolerance 20 ppm, and integration method Most Confident Centroid. Peptides with an FDR > 1% were removed. The downstream workflow included signal-to-noise (S/N) calculation of TMT intensities. Level of confidence for peptide identifications was estimated using the Percolator node with decoy database search. Strict FDR was set at q-value < 0.01.

Bioinformatics analysis of proteomics data

Request a detailed protocol

Processed data files were analyzed as described in a previous publication (Papachristou et al., 2018) using qPLEXanalyzer in R with multimapping peptides included in the analysis. Bar graphs showing protein intensities for Piwi and volcano plots with indicated comparisons were plotted in R using ggplot2.

Quantification and statistical analysis

Request a detailed protocol

Statistical tests used for individual experiments are indicated in the figure legends. Statistical analyses applied to hatching rates, qPCR datasets, and ChIP-seq signal intensity were calculated by unpaired (two-sample) t-test. Significance of TMT mass spectrometry data was calculated according to Papachristou et al., 2018. Differential expression of RNA-seq experiments and differential enrichment of ChIP-seq experiments was calculated using DeSeq2 using adjusted p values as described in Love et al., 2014. The number of biological replicates is indicated in the figure legends.

Data availability

Raw data from high-throughput sequencing experiments are available at GEO under accession number GSE160778. Raw data from proteomics experiments are available on PRIDE with accession number PXD022409. Source data files have been provided for Figures 1 and 2.

The following data sets were generated
    1. Fabry MH
    2. Falconio FA
    3. Joud F
    4. Lythgoe EK
    5. Czech B
    6. Hannon GJ
    (2021) NCBI Gene Expression Omnibus
    ID GSE160778. Maternally inherited piRNAs direct transient heterochromatin formation at active transposons during early Drosophila embryogenesis.
    1. Fabry MH
    2. Falconio FA
    3. Joud F
    4. Lythgoe EK
    5. Czech B
    6. Hannon GJ
    (2021) PRIDE
    ID PXD022409. Maternally inherited piRNAs silence transposons during Drosophila embryogenesis.
The following previously published data sets were used

References

    1. Song SU
    2. Kurkulos M
    3. Boeke JD
    4. Corces VG
    (1997)
    Infection of the germ line by retroviral particles produced in the follicle cells: a possible mechanism for the mobilization of the gypsy retroelement of Drosophila
    Development 124:2789–2798.

Decision letter

  1. Kevin Struhl
    Senior Editor; Harvard Medical School, United States

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

This manuscript is of broad interest to readers in the fields of RNA biology, epigenetics, and early development. The authors provide an exceptionally comprehensive description of the temporal dynamics of mobile element RNA, host defense protein, and epigenetic mark abundance across fruit fly early embryo development. Pairing this descriptive work with the first application of a rapid protein degradation system to maternal proteins in the model fruit fly embryo, the authors reject a previously accepted model that the maternally-deposited protein Piwi establishes gene silencing transmitted epigenetically to later stages of development.

Decision letter after peer review:

Thank you for submitting your article "Maternally inherited piRNAs silence transposons during Drosophila embryogenesis" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Kevin Struhl as the Senior Editor. The reviewers have opted to remain anonymous.

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission. The three reviewers applauded this exceptionally comprehensive description of the temporal dynamics of transposable element RNA, of piRNA pathway proteins and of H3K9me3 across embryogenesis. The reviewers also agreed that the novel application of the auxin-degron system to maternally deposited proteins in the early embryo opens new avenues of research for the community. Finally, the inference that maternal Piwi may actually not set up epigenetic silencing in later life stages represents an important course correction for the field. You will see below, however, that the reviewers were unconvinced that the current data support the major claim of the paper articulated in title, abstract, and final model, namely, that maternal Piwi is required for transposon suppression. We welcome a revised version of the manuscript that stays closer to the data, guided by the list of essential revisions below.

Essential revisions:

New experiments/analyses

1) The reviewers were unconvinced that the data support the major claim of the paper- namely, that maternally-deposited Piwi silences roo. In wildtype embryos, the dramatic increase in roo by 4hr and especially the sharp increase between 2hr-4hr – compared to the comparatively subtle 2-fold increase upon Piwi degradation – makes careful developmental staging hugely important for inferences from the latter experimental data. An internal control to carefully calibrate the RNA-seq with progression of development is necessary. Moreover, the 297 transposon curiously shows just over 2-fold increase versus control. How can we reconcile this change with log10 mean expression is >3 on the Figure 5D but in Figure 1C, the 297 transposon RPMs are barely registering any RPMs throughout embryogenesis.

2) Related to point 1), the reviewers requested additional timepoints for the auxin-induced degradation experiment. Specifically, RNA-seq or even just qPCR well-after the timepoints reported for both roo and 297. These timeports are important for determining if these two elements still drop to original low levels by 12hr or 17hr despite Piwi degradation at the beginning of embryogenesis. If so, the major claim of the paper would be yet further undermined.

3) The reviewers appreciated the compelling loss of H3K9me3 across the genome and at roo insertions upon Piwi degradation. However, the reviewers were less convinced of the significance of H3K9me3 depletion for gene regulation. For example, in Figure 5-SuppFig-1D, the roo insertion with H3K9me3 is overlapping Hid promoter, it seems surprising that there is little change in Hid mRNA levels after auxin induced degradation and loss of H3K9me3 signal. Related, for Figure 3B, the authors should overlay their RNAseq data with the ChIP-seq tracks for lbm and Tsp42El (which are both normally expressed in the embryo). If the pattern is like what is shown on Flybase, lbm and Tsp42El expression may actually increase during embryogenesis in the same degree of H3K9me3 accumulation around the roo insertion. This pattern could oppose the model of direct or meaningful silencing by piRNAs. While there is clearly a piRNA-chromatin response at roo insertions, the effect on roo silencing may actually be quite modest, and the modesty of this effect may contribute to the perplexing lack of any later developmental phenotypes from the auxin-induced Piwi degradation.

4) The reviewers agreed that one of the most impactful contributions made by this report is the rejection of the model put forward by Gu and Elgin. To further probe what may account for the differences between the two studies, the authors could take advantage of their own data. Specifically, in Gu and Elgin, they found different TE families responded differently to the depletion of maternal Piwi – authors could pull these TE families from their genome-wide data to investigate the dynamics (expression and K9 enrichment in response to maternal piwi depletion), further addressing this discrepancy.

Language modification/softening claims/key clarifications

5) Even if the 2-fold increase in roo holds up after more rigorously controlled developmental staging, the reviewers remained unconvinced that such a subtle change (compared to the dramatic spike in roo expression WT embryos) warrants the current title (and the model in figure 5). Unless, for example, the 2-fold increase triggers additional roo transposition (as assayed by WGS), then the title/model appears overstated. Modification of this claim in the title, abstract, and discussion is required.

6) In the ovary, roo is the transposon family with the highest density of antisense piRNAs present and roo mRNA is strongly upregulated upon combined Aub/Ago3 knock-down (Senti et al., 2015). By contrast the authors state that roo is not regulated by the ovarian piRNA pathway (lines 760-763). Their statement, however, is based only on nxf2 KO flies, which inhibit specifically coTGS and not PTGS, in line with previous findings of roo being insensitive to Piwi-mediated regulation in the ovary (Théron, NAR 2018). A revision of this interpretation and referencing the papers showing potent piRNA-mediated regulation of ovarian roo transcripts by PTGS is necessary.

7) The authors' finding of 297 is interesting but needs more elaboration. Based on previous functional studies of piRNA coTGS mutants (e.g., piwi, mael, arx, mael), 297's response is categorized with TE families that are classified as "opposite categories" by authors' data and interpretation – 412 and mdg1 families. In these previous studies, similar to 412 and mdg1, 297 has burst transcription and reduced K9 in these mutants in ovarian somatic cells. I am puzzled by how to reconcile the authors' interpretation of Roo and 297 with these previous findings. Based on Figure1 – SfigF, the expression dynamics of 297 differs from that of roo. Can this be attributed to the absence of normalization by copy number?

8) Since roo is known to be potently regulated through PTGS in the ovary and Aub is maternally inherited, conclusions about somatic piRNA-mediated regulation of roo (and other PTGS-targeted transposons) requires depleting Aub in addition to Piwi. If this experiment is not feasible, a description of the limitations of Piwi depletion specifically regarding likely redundancy between embryonic coTGS and PTGS by Piwi- and Aub-piRNA complexes, respectively, is essential.

9) Images requires quantification. In Figure 2-supplement 1, for example, by only glancing at the images for piwi, panx, and nxf2 vs H2Av localization, I would not have drawn the same conclusion as the authors. Authors should quantify the co-localization of GFP and RFP foci to support their conclusion.

10) The reviewers found omissions in the methods section that require attention.

a. Given the challenges of diffusing small molecules across dechorionated embryos, additional detail about the development of the auxin systems is warranted.

b. Transposon-calling method in the w1118 genotype and the strain with the AID-tagged Piwi should be reported. In the auxin-degron study, the sequenced strain would be heterozygous, complicating TE calling with short reads. Might the uncertainty associated with calling TEs in heterozygotes have led to the confusing results in Figure 5D and E? (For Figure 5 – SFigure C-E, the H3K9me3 enrichment is not right at the boundary of 297 insertions, but a short distance from it). This is a sharp contrast to Figure 3A-B, which is more typical -are the 297 boundaries inappropriately shifted?

c. It appears that only euchromatic TEs were incorporated into the analysis – if so, please clearly state this.

d. The y-axes should be the same for Figure 2A (for piwi) and Figure 2-SFigE (for panx) and 2-SFigG (for Nxf2) to help with comparison across these factors. Same for Figure 2B and Figure 2-SF and H.

e. Details of the ChIP-seq analyses are missing. For some, the authors used rpm (e.g., Figure 3A) while at other places, authors used fold enrichment (e.g., Figure 5E-was the former not normalized to input while the latter was?)

11) What prompts the massive clearance of the H3K9me across the 177 roo insertions after 10h AEL and does this have a real link to Piwi/piRNA binding to the roo nascent transcripts? Maybe speculate on this more in the discussion?

12) Pg 9 Line 236 To determine whether roo might be competent for retrotransposition in embryos, the authors mined quantitative proteomic data for roo peptides of gag, pol and env, but this just establishes ORF expression, not the act of retrotransposition, which actually requires WGS analysis for new copies of roo TE insertions. I suggest changing to more accurate statement like "To determine whether roo mRNAs are effectively being translated during the pulse of embryonic expression, we mined quantitative proteomic data…"

13) What is the consequence of lower viability of dechorionated embryos in regard to RNA-seq and ChIP-seq analyses. When do the embryos die off and how would this affect the dynamic range of the analyses?

14) Calling the orientation of TEs from short read data is tricky. How were these data validated?

Reviewer #1:

This study provides the most solid characterization of Drosophila Piwi mRNA and protein levels throughout embryogenesis to date, with insightful data on Drosophila embryonic TE expression and H3K9m3 marks, and useful creation of a GFP-AID-Piwi fly strain that enables testing whether maternally contributed Piwi has a direct role in responding to TE expression and H3K9me3 levels in the embryos after auxin-induced Piwi degradation. I applaud the thorough analysis, well written prose, beautiful figures and movies, and well-designed experiments. All the data and reagents from this study need to be shared with the public in a revision of the manuscript that I would welcome to see.

But I have one major contention with the hard push of authors to fit the data to a desired hypothesis and mechanism which proposes that maternally-deposited Piwi/piRNAs have a direct role in "silencing" the roo transposon. This issue begins with the paper's title of "Maternally inherited piRNAs silence transposons during Drosophila embryogenesis". My critiques and interpretation of the data suggest to me that Maternal Piwi and piRNAs are 'Responding' to the major roo transposon expression burst during Drosophila embryogenesis, but the importance of a silencing role is debatable.Reviewer #2:

In this study, Fabry and co-authors combined elegant and novel genetic experiments, live imaging, and functional genomics to investigate the role of maternally deposit piRNA machinery, in particular piwi, in silencing transposable elements (TEs) in developing embryos. Surprisingly, they found that maternally deposited piwi, but not zygotically expressed piwi, plays a dominant role in silencing TE families that are predominantly active in the embryos, especially for Roo and 297. Their developmental time course analysis provides a detailed investigation for the sequential events during embryogenesis, namely the nuclear localization of piwi, the burst of TE transcription, and transcriptional silencing of TEs. While the authors' findings are exciting and seem well supported, there are some incongruencies with previous studies, and these need to be further addressed. This includes that piwi's role in maintaining, in addition to initiating, TE epigenetic silencing in somatic tissues and whether TE family 297's activities and host-directed silencing are predominantly embryonic. The varying dynamics among different TE families may worth further investigation (by performing analyses that are normalized by TE copy number) to gain a full picture of the role of piwi in suppressing not only Roo, but also other TE families. Some technical details also need further clarification. Overall, this is an important study that will further our understanding of how hosts suppress selfish genetic parasites.

Reviewer #3:

Argonaute proteins of the Piwi-clade are best known for their role in germline silencing of transposons, but both Piwi and Aub have been shown to also be highly expressed in somatic cells during early Drosophila embryogenesis raising the question of putative somatic regulatory functions. Fabry et al. address this important question by first performing a detailed characterization of endogenous expression of transposons and the transposon silencing piRNA pathway during Drosophila embryogenesis (Figures 1-3) and then an experimental test of the embryonic functions of maternally inherited Piwi/piRNA molecules using a new protocol for degron-mediated protein depletion in dechorionated embryos (Figures 4-5).

In the first part of the paper the authors describe strong embryonic expression of the roo transposon as well as embryonic expression of maternally inherited Piwi. Although this to a certain extent validates previous observations, the authors' comprehensive analyses of transposons and piRNA pathway genes in both transcriptome and proteome data adds a very valuable and novel overview of this gene regulation. In addition, the authors extend on the current knowledge in several places, for example by characterizing Piwi expression from maternal vs zygotic origin (Figure 2E-F).

In the second part of the paper the authors present an elegant implementation of auxin-mediated depletion of degron-tagged Piwi. By auxin administration to dechorionated embryos, the authors are able to deplete maternally inherited Piwi-piRNA protein complexes within 25 minutes. This is an important advance compared to previous RNAi-based methods. In the following analyses the authors describe Piwi-dependent embryonic heterochromatin formation at roo and 297 transposon insertions.

The experiments are in general well designed and controlled and the analyses are broad and comprehensive. As the authors highlight in the last paragraph of the paper, the presented data largely disproves the tested hypothesis of Piwi-piRNA-mediated somatic epigenetic gene regulation in the Drosophila embryo. This finding is important and will be of broad interest for the field.

My main reservation with the current manuscript is that I find the conclusions on embryonic transposon regulation to be unnecessarily overstated. I find that this overstatement somewhat overshadows the important findings that resolve the question of maternal Piwi-piRNA functions in embryo gene regulation.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for submitting your article "Maternally inherited piRNAs direct transient heterochromatin formation at active transposons during early Drosophila embryogenesis" for consideration by eLife.

We found the revised language much more in line with the data and were satisfied with virtually all additional analyses, particularly the new data on embryo staging. A few hanging concerns remain that I trust can be quickly addressed. Note that the absence of tracked changes in the revision document made it difficult, at least in two instances, to track stated adjustments to the text. Please point to these changes with line numbers.

Essential Revisions:

1. The referees requested that panel C from the figure for reviewers be included in the main text along with A and B (maybe as part of the supplement to figure 5? (rebuttal point 1)).

2. Please include language justifying the use of different approaches to analyzing the RNA-seq data presented in Figures 1C and 5D (rebuttal point 1).

3. Please add to the main text, possibly in the legend, that the different strains (w1118 and the degron strain) had different 297 insertion numbers/mean expression (rebuttal point 1).

4. Please point to where language referring to "further highlighted the different normalization strategies in…the text" (rebuttal point 1) is found.

5. Please point to where language referring to "Appropriate adjustments have been made to the text underlining the limitations of whole embryo approaches…" is found.

6. Given that the only major publication addressing maternal Piwi impacts on epigenetic silencing uncovered a very different result, additional language in the main text reconciling the current dataset with Gu and Elgin is still warranted. The few sentences added to the revision are not sufficient to help the reader understand the discrepancy. The cited more modest depletion of Piwi in Gu and Elgin should have more modest effects on the Piwi-regulated TEs- the absence of overlap with your more complete Piwi depletion remains to be explained. More, the discovery of no overlap between upregulated TEs in the two datasets lacks reference to the new figure (or at least a parenthetical "data not shown").

https://doi.org/10.7554/eLife.68573.sa1

Author response

Essential revisions:-

New experiments/analyses

1) The reviewers were unconvinced that the data support the major claim of the paper- namely, that maternally-deposited Piwi silences roo. In wildtype embryos, the dramatic increase in roo by 4hr and especially the sharp increase between 2hr-4hr – compared to the comparatively subtle 2-fold increase upon Piwi degradation – makes careful developmental staging hugely important for inferences from the latter experimental data. An internal control to carefully calibrate the RNA-seq with progression of development is necessary.

First, regarding the main claim of the paper, perhaps we used language that was a bit too strong, and we appreciate the feedback from the referees in that regard. We have now changed the title and calibrated that language throughout the paper. However, we don’t believe that it is appropriate to abandon entirely the notion that roo is regulated by maternally deposited piRNAs in Drosophila embryos. Roo expression is essentially restricted to embryogenesis, yet, flies produce abundant roo piRNAs, including from a specialized piRNA cluster (20A), which is the only uni-strand cluster expressed in the germline. This cluster has gathered roo insertions in predominantly the antisense orientation, as has flamenco for gypsy family elements. The evolutionary pressure to create and maintain such a locus strongly suggests a biologically relevant regulatory relationship between the piRNA pathway and roo. Roo piRNAs are the most abundant class deposited in embryos, where H3K9me3 marks are indisputably deposited on roo insertions in a Piwi-dependent fashion. Roo does show a strong spike in expression in early embryos, which is increased further in the absence of Piwi. Thus, the piRNA pathway does blunt or dampen the expression of roo during this developmental stage. Though loss of that regulation does not have an overt phenotypic impact in a single generation in optimal lab conditions, it remains a strong possibility that the selective pressure to retain recognition of roo by the piRNA pathway in embryos implies that there could be fitness costs should that link be broken. Therefore, while we have modified the language of the paper and attempted to make the points raised herein clearer, we continue to believe that our data does indicate piRNA-dependent regulation of roo.

Regarding the possibility of technical artifacts that might arise from variations in the developmental stages of the embryos at our various time points, we have taken great precautions to limit artifacts arising from female egg retention by ensuring optimal egg laying conditions as well as manually checking a portion of sampled embryos for their expected stage. However, in order to further validate our staging strategy and as requested by the referees, we downloaded publicly available data sets from Flybase profiling gene expression during the complete embryogenesis of Drosophila (Graveley et al., 2011) and compared gene expression with our sampled time points in control w1118 embryos. We found strong correlation in the expression of genes between our experiments (included as an image for the reviewers but subsequently added as Figure 1—figure supplement 1A) and the reference data (included as an image for the reviewers but subsequently added as Figure 1—figure supplement 1B). By examining well-studied genes for primarily maternally deposited transcripts such as the maternal alphaTub67C or exuperantia (exu), involved in proper localisation of bicoid mRNA during oogenesis, and early zygotic transcripts such as the mesodermal master transcription factor twist (twi), which has been shown to drive roo expression, the regulator of dorsal-ventral pattern zerknullt (zen) and disrupted underground network (dunk), involved in syncytial embryo cellularisation during stage 5 (2-3h AEL), we further validated our early time points spanning the beginning of embryogenesis and zygotic genome activation (ZGA). Of note, we found very similar expression pattern between our 2.5-3h AEL w1118 embryo timepoint compared to the degradation experiment profiling 2.5-3h AEL GFP-AID-Piwi; OsTIR1 embryos with or without auxin treatment (included as an image for the reviewers but subsequently added as Figure 5—figure supplement 1A).

Moreover, the 297 transposon curiously shows just over 2-fold increase versus control. How can we reconcile this change with log10 mean expression is >3 on the Figure 5D but in Figure 1C, the 297 transposon RPMs are barely registering any RPMs throughout embryogenesis.

Figure 1C illustrates the expression of transposon transcripts for all sampled embryo time points in our control w1118 strain. Our data is normalised to reads per million (rpm). Figure 5D shows the base mean expression (in log10) of the experiment on the x-axis and fold-change on the y-axis. The base mean is calculated using a different normalisation strategy compared to rpm (median of ratios: counts divided by sample-specific size factors determined by median ratio of gene counts relative to geometric mean per gene) as described in (Love et al., 2014). Therefore, rpm and log10 base mean are not directly comparable. Additionally, our w1118 control strain used for Figure 1C has a lower number of 297 insertions than the degron strain used for Figure 5D (27 euchromatic insertions in w1118 versus 82 insertions in the GFP-AID-Piwi; OsTIR1 line). The mean expression of 297 in w1118 is ~70 rpm while mean expression in untreated degron embryos is 167 rpm (converted from log10 base means). This discrepancy is likely due to the different transposon content in both strains. In order to avoid confusion of the reader, we further highlighted the different normalisation strategies in both the text and method sections.

2) Related to point 1), the reviewers requested additional timepoints for the auxin-induced degradation experiment. Specifically, RNA-seq or even just qPCR well-after the timepoints reported for both roo and 297. These timeports are important for determining if these two elements still drop to original low levels by 12hr or 17hr despite Piwi degradation at the beginning of embryogenesis. If so, the major claim of the paper would be yet further undermined.

We believe that we have addressed the reviewer’s concerns by performing qPCR experiments on embryos derived from our degron strain (GFP-AID-Piwi; OsTIR1) treated with or without auxin for additional time points as requested in order to assess whether roo and 297 expression levels remain elevated over control samples or return to baseline expression. By analysing expression of both roo and 297 from early embryogenesis to later stages, we were able to replicate our previous finding showing upregulation of both roo and 297 upon auxin treatment in 2.5-3h embryos. However, both TE expression levels return to baseline values at 6-7h and 7-8h AEL time points (thus we did not perform the experiment at time points as late as 12h or 17h AEL). These data suggest that transposon expression upon Piwi depletion is only affected in early embryogenesis. We included this data in our revised manuscript as Figure 5—figure supplement 1F and discuss the results in context of the overall conclusions drawn from our new data. We do disagree that this further undermines the major claim of the paper. This simply means that roo and 297 are either silenced by other means at those developmental time points or that the transcription factors necessary to drive their expression during a specific window of development are simply absent. While the transcription factors regulating 297 are not known, for roo the latter possibility is supported by the expression of twist (twi), which shows strongest expression between 2-6h AEL (also see Figure 1—figure supplement A,B). Note, some delay in roo expression is expected due to the dynamics of translation and activation of roo loci by Twi protein.

3) The reviewers appreciated the compelling loss of H3K9me3 across the genome and at roo insertions upon Piwi degradation. However, the reviewers were less convinced of the significance of H3K9me3 depletion for gene regulation. For example, in Figure 5-SuppFig-1D, the roo insertion with H3K9me3 is overlapping Hid promoter, it seems surprising that there is little change in Hid mRNA levels after auxin induced degradation and loss of H3K9me3 signal. Related, for Figure 3B, the authors should overlay their RNAseq data with the ChIP-seq tracks for lbm and Tsp42El (which are both normally expressed in the embryo). If the pattern is like what is shown on Flybase, lbm and Tsp42El expression may actually increase during embryogenesis in the same degree of H3K9me3 accumulation around the roo insertion. This pattern could oppose the model of direct or meaningful silencing by piRNAs. While there is clearly a piRNA-chromatin response at roo insertions, the effect on roo silencing may actually be quite modest, and the modesty of this effect may contribute to the perplexing lack of any later developmental phenotypes from the auxin-induced Piwi degradation.

The reviewers are correct in noting that Hid expression seems not affected by degradation of Piwi protein and subsequent loss of H3K9me3 marks during ZGA (Figure 5—figure supplement 1D). However, our experiments were performed on whole embryos. As we showed in our FISH experiments (Figure 1—figure supplement 1B), roo is primarily expressed in specific tissues associated with the developing mesoderm. H3K9me3 deposition as a response to Piwi-mediated silencing is therefore only likely to occur in tissues actually showing roo expression, as H3K9me3 is thought to occur co-transcriptionally. Therefore, expression of genes in proximity of roo insertions are only silenced if they are co-expressed in cells together with roo, which it does not appear to be. Our ChIP-seq and RNA-seq data show an average of H3K9me3 occupancy and gene expression respectively of the whole embryo. These data illustrate the advantages of “clean” systems like OSCs (cell culture) in which loss of the piRNA pathway has a uniform impact on gene expression upon loss of H3K9me3.

Gene expression of lbm and Tsp42EI followed patterns observed in previous studies (Graveley et al., 2011), which showed onset of transcription in later stages of embryogenesis (8-9h AEL). However, H3K9me3 signal accumulated at a strain-specific roo insertion in line with the overall dynamics observed in our study (Author response image 1). While accumulation of H3K9me3 was detected in our experiments, this data lacks spatial information and H3K9me3 accumulating cells might not overlap cells expressing both examined genes. These technical challenges will be overcome in the future by the development of single-cell chromatin profiling methods as well as multiomics approaches. Appropriate adjustments have been made to the text underlining the limitations of whole embryo approaches for both profiling RNA and chromatin modifications in bulk embryos.

Author response image 1
Gene expression and H3K9me3 occupancy of genes in proximity of euchromatic TE insertions.

IGV genome browser screenshot showing H3K9me3 ChIP-seq signal for the indicated genes on chromosome 2R carrying a w1118-specific roo insertion. Blue tracks show gene expression for indicated embryo time points (in rpm; n=2). Green tracks illustrate H3K9me3 occupancy at comparable time points (in rpm; n=2).

4) The reviewers agreed that one of the most impactful contributions made by this report is the rejection of the model put forward by Gu and Elgin. To further probe what may account for the differences between the two studies, the authors could take advantage of their own data. Specifically, in Gu and Elgin, they found different TE families responded differently to the depletion of maternal Piwi – authors could pull these TE families from their genome-wide data to investigate the dynamics (expression and K9 enrichment in response to maternal piwi depletion), further addressing this discrepancy.

The study by Gu and Elgin investigates the impact of Piwi reduction in early embryos on further developmental stages beyond embryogenesis using piwi null mutants. Piwi depletion in early embryos was achieved by crossing heterozygous null mutants (piwi2/+ x piwi2/+) therefore reducing, but not abolishing, maternally deposited Piwi protein (Figure 2A, Gu and Elgin, 2013). The authors analysed the impact of Piwi reduction in embryogenesis by evaluating HP1a binding and H3K9me3 levels in larvae using ChIP-array for selected genomic targets including TEs in two replicates (Figure 5B and S6, Gu and Elgin, 2013). Some transposons did not show strong HP1a enrichment in wild-type larvae (including roo, DMRP1 and XDMR), while other TEs such as gypsy5, invader3 and DIVER2 showed a higher enrichment over the input (> 2-fold). Most TEs showed an overall modest decrease in HP1a levels (69 out of 83 classes) including highlighted classes such as Bari1, Invader1, mdg1, Het-A and TART in larvae that inherited reduced Piwi levels. H3K9me3 levels at roo sequences showed no enrichment over input in wild type larvae (Figure S6, Gu and Elgin, 2013) in accordance with our observation that H3K9me3 is lost at later stages of development (Figure 3, this study). While Gu and Elgin assess the impact of maternal Piwi reduction during embryogenesis in a later larval stage, our study focuses on illuminating the impact of maternally deposited Piwi-piRNA complex on early embryogenesis. However, in order to compare both studies, we highlighted TE classes reported in the study by Gu and Elgin (using the same colour scheme used in Gu and Elgin) in our data evaluating expression profiles and H3K9me3 occupancy in treated versus untreated embryos during ZGA (Author response image 2), as requested by the reviewers. TE classes reported by Gu and Elgin as sensitive to Piwi depletion in larvae did not show significant changes in either RNA expression or H3K9me3 occupancy in our study. These results are likely due to differences in our depletion strategy targeting the entire pool of maternally deposited proteins rather than reducing Piwi levels by only 2-fold as well as the different read out (RNA-seq and ChIP-seq in our study versus HP1a ChIP-array in Gu and Elgin’s study). Additionally, different developmental stages were evaluated in both studies (early embryogenesis, this study versus larvae, Gu and Elgin) further complicating comparisons. Last, our analyses are based different transposon annotations with the data reported here using annotations containing the full length TE sequences, while the study by Gu and Elgin used separate versions for transposon bodies and LTRs annotations.

Author response image 2
Transposons sensitive to Piwi reduction in larvae showed no change in our study.

(A) MA plot showing base mean expression (log10 scale) of transposon RNAs relative to their fold-change (log2 scale) in GFP-AID-Piwi; OsTIR1 embryos treated with 5mM auxin versus control (n=3). Blue=TE classes with low HP1a enrichment in wild type larvae as reported in Gu et al., 2013. Brown=TE classes with reported HP1a enrichment >2 fold over input in wild type larvae as reported in Gu et al., 2013. Red=TE classes with reported loss of HP1a enrichment in piwi2/piwi2 mutants vs control as reported in Gu et al., 2013. (B) MA plot showing base mean signal intensity (log10 scale) of TEs relative to the H3K9me3 ChIP-seq signal enrichment (log2 scale) in GFP-AID-Piwi; OsTIR1 embryos treated with 5mM auxin versus control (n=3). Colour scheme same as in (A).

Language modification/softening claims/key clarifications

5) Even if the 2-fold increase in roo holds up after more rigorously controlled developmental staging, the reviewers remained unconvinced that such a subtle change (compared to the dramatic spike in roo expression WT embryos) warrants the current title (and the model in figure 5). Unless, for example, the 2-fold increase triggers additional roo transposition (as assayed by WGS), then the title/model appears overstated. Modification of this claim in the title, abstract, and discussion is required.

We acknowledge the reviewer’s concern about overstating the silencing capacity of a maternally inherited piRNA-dependent mechanism on transposon expression in early embryogenesis. However, our data clearly shows statistically significant upregulation of roo and 297 as well as loss of H3K9me3 during ZGA of embryogenesis upon auxin-mediated Piwi depletion. Whether this is biologically relevant remains to be determined (though see above), which we address in our discussion. Both 297 and roo steady-state RNA levels returned to baseline at later time points. This indicates that while the piRNA pathway engages transposons and apparently dampens transcription initially during embryogenesis, it fails to contain the surge of transcription following ZGA likely in response to increased abundance of transcription factors such as twist and snail further driving expression. We included our new findings (see point 2) in the revised manuscript and discuss the implications in greater detail. Additionally, we adjusted our claims/model accordingly.

We also included a paragraph in the discussion which explicitly raises whether roo regulation by piRNAs is biologically relevant, so that readers can weight the evidence for themselves:

“Though zygotic depletion of maternal Piwi during early embryogenesis does produce a statistically significant change in roo expression (roughly 2-fold), this transposon remains highly expressed reaching up to 1% of the entire transcriptome in control animals, despite being targeted by the piRNA pathway. This provokes the question of whether targeting of roo by the piRNA pathway is biologically relevant. In favour of this hypothesis are several observations. Roo is expressed in ovaries at very low levels, yet the hallmarks of piRNA-dependent silencing, specifically H3K9me3, are absent from euchromatic roo insertions. This strongly indicates that roo is not controlled by the piRNA pathway in this tissue. Nonetheless, ovaries produce abundant roo piRNAs, and these are overwhelmingly in the antisense orientation. Additionally, the only uni-strand cluster expressed in germ cells, cluster 20A, has collected roo insertions in the antisense orientation. These piRNAs are abundantly maternally transmitted (16% of all piRNAs in embryos) and persist throughout the time during early embryogenesis when high level roo expression is proposed to be driven by mesodermal transcription factors. An argument against biological significance is the lack of a clearly observable phenotype in flies following embryonic depletion of maternal Piwi. However, technical limitations enable us to only measure impacts within a single generation. It is entirely possible that the fitness cost of roo occupying 2% of the embryonic transcriptome might be substantial over time or in conditions flies might experience in the wild compared to the controlled rearing conditions in the lab.”

6) In the ovary, roo is the transposon family with the highest density of antisense piRNAs present and roo mRNA is strongly upregulated upon combined Aub/Ago3 knock-down (Senti et al., 2015). By contrast the authors state that roo is not regulated by the ovarian piRNA pathway (lines 760-763). Their statement, however, is based only on nxf2 KO flies, which inhibit specifically coTGS and not PTGS, in line with previous findings of roo being insensitive to Piwi-mediated regulation in the ovary (Théron, NAR 2018). A revision of this interpretation and referencing the papers showing potent piRNA-mediated regulation of ovarian roo transcripts by PTGS is necessary.

Indeed, Senti et al., 2015 report that a high proportion of antisense Piwi-bound piRNAs are mapping to roo in line with our study (Figure 2G, this study) as well as previous reports (Brennecke et al., 2009). Knockdown of aub and ago3 led indeed to a strong decrease in roo-mapping Piwi-bound piRNAs (Figure 2C, Senti et al.). The authors additionally show that most roo-mapping piRNAs (>90%) are antisense to the TE’s consensus sequence (Figure 2D, Senti et al.) thereby identifying the origin of these piRNAs as likely cluster-derived rather than a product of PTGS.

However, we were unable to find the basis of the reviewer’s statement that roo mRNA is strongly upregulated upon double knockdown of aub and ago3. The authors report steady-state mRNA levels for all knockdown conditions. Mining supplementary table S1, we were able to compare expression levels for roo in germline knockdowns (GLKD) of white, piwi, aub, ago3 and aub/ago3 since roo expression levels are not reported or discussed in the manuscript. roo showed a base expression (white GLKD) of 0.411 RPKM. This increased to 0.572 in piwi GLKD, 0.788 in aub GLKD, 0.864 in ago3 GLKD and 1.644 in aub/ago3 GLKD (Supplementary Table S1, Senti et al.). While the fold-change between aub/ago3 versus white GLKD is precisely 4, we argue that this is likely irrelevant due to the extremely low baseline expression in white GLKD, which is just over the detection limit. These data are in complete agreement with our reports of low expression of roo in heterozygous nxf2 mutants (0.04 RPKM) versus homozygous mutants (0.06 RPKM). We included Senti et al., 2015 in our citations as further evidence that roo is neither affected by loss of piRNA-dependent coTGS (nxf2 mutants, Figure 5—figure supplement 1G, this study; piwi GLKD Senti et al., 2015) nor by loss of PTGS (aub, ago3, aub/ago3 GLKD, Senti et al., 2015). We apologise if our conclusions based on revisiting data in Senti et al. are incorrect and would ask the referees to highlight relevant information should we have missed it.

7) The authors' finding of 297 is interesting but needs more elaboration. Based on previous functional studies of piRNA coTGS mutants (e.g., piwi, mael, arx, mael), 297's response is categorized with TE families that are classified as "opposite categories" by authors' data and interpretation – 412 and mdg1 families. In these previous studies, similar to 412 and mdg1, 297 has burst transcription and reduced K9 in these mutants in ovarian somatic cells. I am puzzled by how to reconcile the authors' interpretation of Roo and 297 with these previous findings. Based on Figure1 – SfigF, the expression dynamics of 297 differs from that of roo. Can this be attributed to the absence of normalization by copy number?

It is correct that 297 has been previously reported to be controlled by the piRNA pathway in Drosophila OSCs (Sienski et al., 2012). Loss of coTGS factors such as nxf2 or panx leads to strong transcriptional de-repression accompanied by loss of H3K9me3 at TE insertions (Sienski et al., 2015, Batki et al., 2019, Murano et al., 2019, Fabry et al., 2019, Zhao et al., 2019). Our data shows a very similar behaviour during embryogenesis. In Figure 3—figure supplement 1B we show the accumulation of H3K9me3 at euchromatic 297 insertions during embryogenesis as well as for ovaries and heads. Indeed, the chromatin state shows different dynamics compared to roo H3K9me3 profiles (Figure 3A). Unlike roo, which is not regulated by the piRNA pathway in ovaries (see point 6), 297 acquires a typical H3K9me3 signature at TE insertions in ovaries, likely as a result of coTGS in somatic sheet cells, which are the parent tissue of OSCs. These findings are consistent with the current model of coTGS in which germline regulated TEs are silenced and H3K9me3 levels increase around affected insertions. 297 showed a lower number of 297 insertions in our control w1118 strain (n=20) compared to roo insertions (n=117). Discrepancies in chromatin dynamics between the two TE classes are likely a result of different transcriptional control mechanisms. While the driver of 297 transcription is unknown, roo expression is likely to be controlled by the mesoderm-associated transcription factors twist and snail (Bronner et al., 1995), which show strong expression during early stages of embryogenesis and match the roo expression profile.

8) Since roo is known to be potently regulated through PTGS in the ovary and Aub is maternally inherited, conclusions about somatic piRNA-mediated regulation of roo (and other PTGS-targeted transposons) requires depleting Aub in addition to Piwi. If this experiment is not feasible, a description of the limitations of Piwi depletion specifically regarding likely redundancy between embryonic coTGS and PTGS by Piwi- and Aub-piRNA complexes, respectively, is essential.

Our analysis has indicated that roo is not under the control of either piRNA-dependent coTGS or PTGS in ovaries (see point 6 – but we await further information from the referee). However, Aub depletion together with Piwi is not yet feasible. Aub and Piwi are located on the same chromosome within a distance of 10kb thereby making recombination of CRISPR-mediated AID-Aub and AID-Piwi alleles highly unlikely. Engineering the Aub locus in the AID-Piwi fly could overcome these limitations, however it is not an established strategy and beyond the scope of this study.

We agree that maternally deposited Aub-piRNA complexes might serve a significant function during embryogenesis especially in regards of germline development since maternally deposited Aub primarily localises to germ plasm during embryogenesis and is not enriched in somatic nuclei (Brennecke et al., 2009). However, for these complexes to regulate embryonic transposons by PTGS, they would likely have to do so by an RNAi-like cleavage mechanism, since Ago3 is not strongly maternally deposited and we do not detect strong ping-pong signatures in embryonic piRNA populations. Of course, such signals could be swamped if Aub only performs this function in pole cells.

9) Images requires quantification. In Figure 2-supplement 1, for example, by only glancing at the images for piwi, panx, and nxf2 vs H2Av localization, I would not have drawn the same conclusion as the authors. Authors should quantify the co-localization of GFP and RFP foci to support their conclusion.

We analysed the co-localisation of Piwi, Panx, and Nxf2 with H2Av and found striking co-localisation of the three studied GFP-tagged coTGS factors and RFP-tagged H2Av as displayed by measurements of fluorescence intensity across sections for both channels in Figure 2—figure supplement 1C-E.

10) The reviewers found omissions in the methods section that require attention.

a. Given the challenges of diffusing small molecules across dechorionated embryos, additional detail about the development of the auxin systems is warranted.

The auxin-mediated degradation system has been intensively characterised in multiple animal models including C. elegans, Drosophila and mammalian systems. Auxin is a small molecule and multiple studies suggest high cellular permeability. Additionally, a recent study found high permeability of auxin through the egg and cuticle, which are well known to show poor drug efficacy due to limited permeability, even during embryogenesis (Zhang et al., 2015). We added this information as well as the appropriate citation in the methods section of the revised manuscript.

b. Transposon-calling method in the w1118 genotype and the strain with the AID-tagged Piwi should be reported. In the auxin-degron study, the sequenced strain would be heterozygous, complicating TE calling with short reads. Might the uncertainty associated with calling TEs in heterozygotes have led to the confusing results in Figure 5D and E? (For Figure 5 – SFigure C-E, the H3K9me3 enrichment is not right at the boundary of 297 insertions, but a short distance from it). This is a sharp contrast to Figure 3A-B, which is more typical -are the 297 boundaries inappropriately shifted?

Transposons were called as described in our methods section using the TEMP algorithm. We include all relevant information and a comprehensive description of our strategy in line 1141-1147 of our original manuscript. More detailed information about the strategy can be found in the TEMP publication (Zhuang et al., 2014), which we cite in the methods section. We included only high confidence transposon insertions that were supported by at least one unique read at both sides (reference genome and TE consensus sequence) in order to ensure the information on orientation is correct. To the best of our knowledge TE insertions have been accurately called using our very deeply sequenced WGS data. Files containing transposon insertions including relevant statistical measurements have been submitted as supplementary files to GEO (accession number GSE160778). We sequenced only homozygous flies of our control w1118 and degron strain (GFP-AID-Piwi; OsTIR1) and all our experiments have been performed using homozygous embryos or flies.

c. It appears that only euchromatic TEs were incorporated into the analysis – if so, please clearly state this.

Yes, this is correct. We would like to point out that our original manuscript clearly stated the use of only euchromatic transposons in our analyses in the text, figure legends, and methods section (see line 436, 457, 677, 1129 of the original manuscript). We have nevertheless highlighted this at other occasions that did not contain this information earlier.

d. The y-axes should be the same for Figure 2A (for piwi) and Figure 2-SFigE (for panx) and 2-SFigG (for Nxf2) to help with comparison across these factors. Same for Figure 2B and Figure 2-SF and H.

While we do not agree with this comment, we have updated the figures for RNA-seq and MS data using the same y-axes as requested. We would like to stress that the purpose of these panels were not to compare the expression of the highlighted genes among each other but across the studied time course.

e. Details of the ChIP-seq analyses are missing. For some, the authors used rpm (e.g., Figure 3A) while at other places, authors used fold enrichment (e.g., Figure 5E-was the former not normalized to input while the latter was?)

ChIP-seq experiments reported in Figure 3 and Figure 3—figure supplement 1 show H3K9me3 signal in reads per million (rpm) calculated for 10bp genome wide bin sizes using the deepTools2 bamCoverage function (Ramírez et al., 2016). Figure 5E reports fold-changes between H3K9me3 levels of auxin versus control treated embryos. Differential chromatin state analysis was performed using DESeq2 as reported in Love et al., 2014. We highlighted the appropriate information more clearly in the methods section and figure legends.

11) What prompts the massive clearance of the H3K9me across the 177 roo insertions after 10h AEL and does this have a real link to Piwi/piRNA binding to the roo nascent transcripts? Maybe speculate on this more in the discussion?

We added the following to the section in which this result is presented:

“This is consistent both with the known requirement for active transcription for targeting by Piwi and with the observed need for continuous engagement of PICTS to maintain H3K9me3 marks on transposon loci.”

12) Pg 9 Line 236 To determine whether roo might be competent for retrotransposition in embryos, the authors mined quantitative proteomic data for roo peptides of gag, pol and env, but this just establishes ORF expression, not the act of retrotransposition, which actually requires WGS analysis for new copies of roo TE insertions. I suggest changing to more accurate statement like "To determine whether roo mRNAs are effectively being translated during the pulse of embryonic expression, we mined quantitative proteomic data…"

We changed the wording accordingly.

13) What is the consequence of lower viability of dechorionated embryos in regard to RNA-seq and ChIP-seq analyses. When do the embryos die off and how would this affect the dynamic range of the analyses?

Dechorionation using bleach exposes the embryo surface leading to increased sensitivity to dehydration and leakage through mechanical stress. We did not observe any biases of embryos failing to develop due to dechorionation in regard to stages of development.

14) Calling the orientation of TEs from short read data is tricky. How were these data validated?

Transposon calling was performed on paired-end 150bp data. The mean fragment size of our DNA fragments used for library preparation were >500bp to allow for sequencing of fragments overlapping the reference sequence as well as the unique internal sequence of our TE consensus sequences. By allowing mapping of only unique pairs, we excluded the possibility of including fragments spanning the reference sequence and only the LTR sequence found at the beginning and end of the TE in our analysis thereby ensuring high confidence in orientation calling.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

We found the revised language much more in line with the data and were satisfied with virtually all additional analyses, particularly the new data on embryo staging. A few hanging concerns remain that I trust can be quickly addressed. Note that the absence of tracked changes in the revision document made it difficult, at least in two instances, to track stated adjustments to the text. Please point to these changes with line numbers.

Essential Revisions:

1. The referees requested that panel C from the image for the reviewers be included in the main text along with A and B (maybe as part of the supplement to figure 5? (rebuttal point 1)).

Panel C of image for the reviewers has now been added to Figure 5—figure supplement 1 as panel A and we refer to this figure in the Results section on page 22 (lines 643-647).

2. Please include language justifying the use of different approaches to analyzing the RNA-seq data presented in Figures 1C and 5D (rebuttal point 1).

Figure 1C illustrates the expression of transposon transcripts for all sampled embryo time points in our control w1118 strain. To consider different library sizes and facilitate comparability throughout our time course experiment (for which we only had two biological replicates per time point), we normalised our RNA-seq data to reads per million (rpm). We now state this on page 5, lines 177-179.

Considering the experiment using embryo collections either treated with auxin or PBS as a control was carried out with three biological replicates, we chose a different normalisation strategy to calculate differential expressed genes and transposons (Figure 5D). This allowed for use of the most commonly accepted current statistical modelling (DESeq2) of differential gene expression measurements of the impact of auxin treatment on RNA output (see also comment 4 below). We have modified the text in the main manuscript on page 22 (lines 640-643) to clearly state that we carried out differential expression analysis for the Piwi degradation experiments.

3. Please add to the main text, possibly in the legend, that the different strains (w1118 and the degron strain) had different 297 insertion numbers/mean expression (rebuttal point 1).

For clarity, we have included the number of degron strains- and w1118-specific transposon insertions, respectively, in the main text and relevant figure legends:

– Main text, page 15, lines 469-470,

– Legend of figure 3, page 16, lines 490 and 492,

– Legend of figure 3—figure supplement 1, page 17, lines 501 and 503 and 504,

– Main text, page 18, lines 531-533,

– Main text, page 22, lines 673-677,

– Legend of figure 5, page 23, line 702,

– Legend of figure 5—figure supplement 1, page 25, lines 722 and 726 and 727.

4. Please point to where language referring to "further highlighted the different normalization strategies in…the text" (rebuttal point 1) is found.

In addition to the adjustments we made in the main text of the revised manuscript (see comment 2), we also updated the methods section (lines 1233-1245) to clearly state the different normalisation strategies we used for the different datasets. The methods section now states:

“Count files for RNA-seq time course experiments generated as described above were normalised to reads per million (rpm) to account for differences in library size and allow comparability between time points. Heatmaps displaying expression profiles of genes and transposons during embryogenesis show the mean expression values of the biological replicates, while bar graphs display the individual data points as well as the mean expression and standard deviation. Bar graphs and heatmaps were plotted in R using ggplot2.

RNA-seq experiments comparing auxin- and PBS-treated embryos of the same stage and collection were analysed using differential expression quantification methods allowing for statistical evaluation of differences between RNA output as a direct result of auxin treatment. Differential expression analysis was performed using DESeq2 (Love et al., 2014). MA plots show base mean RNA expression across conditions and were calculated as previously described by Love and colleagues.”

5. Please point to where language referring to "Appropriate adjustments have been made to the text underlining the limitations of whole embryo approaches…" is found.

We have included a paragraph discussing the limitation of whole embryo approaches in the discussion on pages 27-28 (lines 837-840) of the revised manuscript:

“While our data provide compelling evidence of the accumulation of repressive chromatin marks at presumably actively transcribing TE insertions, it does not carry spatial information about the precise cell types affected by H3K9me3 deposition.”

6. Given that the only major publication addressing maternal Piwi impacts on epigenetic silencing uncovered a very different result, additional language in the main text reconciling the current dataset with Gu and Elgin is still warranted. The few sentences added to the revision are not sufficient to help the reader understand the discrepancy. The cited more modest depletion of Piwi in Gu and Elgin should have more modest effects on the Piwi-regulated TEs- the absence of overlap with your more complete Piwi depletion remains to be explained. More, the discovery of no overlap between upregulated TEs in the two datasets lacks reference to the new figure (or at least a parenthetical "data not shown").

We added text to elaborate on the discrepancy between our work and the study by Gu and Elgin in the discussion on pages 28-29 (lines 863-904). The paragraph reads:

“Perhaps more importantly, our study demonstrates that recognition of a locus by the piRNA pathway does not necessarily impose the creation of a mitotically heritable epigenetic state. […]Thus, it seems that the data themselves diverge less between the two studies than do the conclusions drawn. Of note, another recent report found a mild upregulation of transposons in pre-ZGA embryos upon maternal depletion of Piwi, however, this work relied on germ cell-specific knockdown during late stages of oogenesis rather than direct protein depletion in the embryo, thus at least some of the observed effects could stem from TE mobilisation during ovary development (Gonzalez et al., 2021).”

https://doi.org/10.7554/eLife.68573.sa2

Article and author information

Author details

  1. Martin H Fabry

    CRUK Cambridge Institute, University of Cambridge, Li Ka Shing Centre, Cambridge, United Kingdom
    Contribution
    Conceptualization, Data curation, Software, Formal analysis, Supervision, Validation, Investigation, Visualization, Methodology, Writing - original draft, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-8484-4715
  2. Federica A Falconio

    CRUK Cambridge Institute, University of Cambridge, Li Ka Shing Centre, Cambridge, United Kingdom
    Contribution
    Validation, Investigation, Writing - review and editing
    Competing interests
    No competing interests declared
  3. Fadwa Joud

    CRUK Cambridge Institute, University of Cambridge, Li Ka Shing Centre, Cambridge, United Kingdom
    Contribution
    Data curation, Formal analysis, Investigation, Visualization, Writing - review and editing
    Competing interests
    No competing interests declared
  4. Emily K Lythgoe

    CRUK Cambridge Institute, University of Cambridge, Li Ka Shing Centre, Cambridge, United Kingdom
    Contribution
    Investigation, Writing - review and editing
    Competing interests
    No competing interests declared
  5. Benjamin Czech

    CRUK Cambridge Institute, University of Cambridge, Li Ka Shing Centre, Cambridge, United Kingdom
    Contribution
    Conceptualization, Data curation, Supervision, Validation, Investigation, Visualization, Methodology, Writing - original draft, Project administration, Writing - review and editing
    For correspondence
    benjamin.czech@cruk.cam.ac.uk
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-8471-0007
  6. Gregory J Hannon

    CRUK Cambridge Institute, University of Cambridge, Li Ka Shing Centre, Cambridge, United Kingdom
    Contribution
    Conceptualization, Supervision, Funding acquisition, Methodology, Writing - original draft, Project administration, Writing - review and editing
    For correspondence
    greg.hannon@cruk.cam.ac.uk
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-4021-3898

Funding

Royal Society (RP130039)

  • Gregory J Hannon

Royal Society (RSRP\R\200001)

  • Gregory J Hannon

Cancer Research UK (A21143)

  • Gregory J Hannon

Wellcome Trust (110161/Z/15/Z)

  • Gregory J Hannon

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank Marie Bao from Life Science Editors (http://lifescienceeditors.com) for comments on the manuscript. We thank the members of the Hannon group, especially Susanne Bornelöv, for discussion and feedback on the manuscript. We thank the Cancer Research UK Cambridge Institute Bioinformatics, Genomics, Microscopy, RICS, and Proteomics Core Facilities for support, in particular Kamal Kishore, Evangelia Papachristou, and Carmen Gonzalez Tejedo. We thank the University of Cambridge Department of Genetics Fly Facility for microinjection services and fly stock generation. We thank the Vienna Drosophila Resource Center for fly stocks. We thank Daniel St Johnston for H2Av-RFP flies. This research was funded in whole, or in part, by Cancer Research UK (A21143) and the Wellcome Trust (110161/Z/15/Z). For the purpose of open access, the author has applied a CC BY public copyright license to any Author Accepted Manuscript version arising from this submission. GJH is a Royal Society Wolfson Research Professor (RP130039 and RSRP\R\200001).

Senior Editor

  1. Kevin Struhl, Harvard Medical School, United States

Publication history

  1. Received: March 19, 2021
  2. Preprint posted: May 25, 2021 (view preprint)
  3. Accepted: July 7, 2021
  4. Accepted Manuscript published: July 8, 2021 (version 1)
  5. Version of Record published: August 9, 2021 (version 2)

Copyright

© 2021, Fabry et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,593
    Page views
  • 233
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Chromosomes and Gene Expression
    2. Developmental Biology
    Benoit Roch et al.
    Research Article Updated

    We developed an Xrcc4M61R separation of function mouse line to overcome the embryonic lethality of Xrcc4-deficient mice. XRCC4M61R protein does not interact with Xlf, thus obliterating XRCC4-Xlf filament formation while preserving the ability to stabilize DNA ligase IV. X4M61R mice, which are DNA repair deficient, phenocopy the Nhej1-/- (known as Xlf -/-) setting with a minor impact on the development of the adaptive immune system. The core non-homologous end-joining (NHEJ) DNA repair factor XRCC4 is therefore not mandatory for V(D)J recombination aside from its role in stabilizing DNA ligase IV. In contrast, Xrcc4M61R mice crossed on Paxx-/-, Nhej1-/-, or Atm-/- backgrounds are severely immunocompromised, owing to aborted V(D)J recombination as in Xlf-Paxx and Xlf-Atm double Knock Out (DKO) settings. Furthermore, massive apoptosis of post-mitotic neurons causes embryonic lethality of Xrcc4M61R -Nhej1-/- double mutants. These in vivo results reveal new functional interplays between XRCC4 and PAXX, ATM and Xlf in mouse development and provide new insights into the understanding of the clinical manifestations of human XRCC4-deficient condition, in particular its absence of immune deficiency.

    1. Chromosomes and Gene Expression
    2. Computational and Systems Biology
    Zelin Liu et al.
    Tools and Resources

    Circular RNAs (circRNAs) act through multiple mechanisms via their sequence features to fine-tune gene expression networks. Due to overlapping sequences with linear cognates, identifying internal sequences of circRNAs remains a challenge, which hinders a comprehensive understanding of circRNA functions and mechanisms. Here, based on rolling circular reverse transcription (RCRT) and nanopore sequencing, we developed circFL-seq, a full-length circRNA sequencing method, to profile circRNA at the isoform level. With a customized computational pipeline to directly identify full-length sequences from rolling circular reads, we reconstructed 77,606 high-quality circRNAs from seven human cell lines and two human tissues. circFL-seq benefits from rolling circles and long-read sequencing, and the results showed more than tenfold enrichment of circRNA reads and advantages for both detection and quantification at the isoform level compared to those for short-read RNA sequencing. The concordance of the RT-qPCR and circFL-seq results for the identification of differential alternative splicing suggested wide application prospects for functional studies of internal variants in circRNAs. Moreover, the detection of fusion circRNAs at the omics scale may further expand the application of circFL-seq. Together, the accurate identification and quantification of full-length circRNAs make circFL-seq a potential tool for large-scale screening of functional circRNAs.