Transposable element (TE) activity needs to be repressed to avoid severe genome instability and gametogenesis defects. In humans, growing evidence has implicated TE in several disorders such as cancers defining a new field of diseases called transposopathies (Wylie et al., 2016a; Wylie et al., 2016b). In the animal germline, TE activity is controlled at both transcriptional and post-transcriptional levels by small RNAs called piRNAs associated with the PIWI clade of germline Argonaute proteins (Piwi, Aub and Ago3 in Drosophila) (Brennecke et al., 2007; Gunawardane et al., 2007; Le Thomas et al., 2013; Sienski et al., 2012). piRNAs are processed from transcripts produced from specific heterochromatic loci enriched in TE fragments, called piRNA clusters (Brennecke et al., 2007; Gunawardane et al., 2007). These loci undergo non-canonical transcription, ignoring splicing and transcription termination signals, licensed by specific protein complexes such as Rhino-Deadlock-Cutoff (Mohn et al., 2014; Zhang et al., 2014) and Moonshiner-TRF2 (Andersen et al., 2017). Thus, when a new TE inserts into a naive genome, it will freely transpose until one copy gets inserted into a piRNA cluster leading to the production of homologous new TE piRNAs that will then repress transposition (Brennecke et al., 2008). In support of this idea, exogenous sequences inserted into preexisting piRNA clusters lead to the production of matching piRNAs (de Vanssay et al., 2012; Hermant et al., 2015; Marie et al., 2017; Muerdter et al., 2012; Pöyhönen et al., 2012). The specificity of the efficient repression mediated by piRNAs appears to be determined solely by the piRNA cluster sequences. Thus, it raises the question of how piRNA cluster loci are themselves specified. Histone H3 lysine nine tri-methylation (H3K9me3) that is recognized by Rhino, a paralog of heterochromatin protein HP1 (Klattenhoff et al., 2009), is a shared feature of piRNA clusters. Enrichment of H3K9me3, however, is not specific to piRNA clusters and tethering Rhino onto a transgene leads to the production of piRNAs only when both sense and antisense transcripts are produced (Zhang et al., 2014). This suggests that neither H3K9me3 marks nor having Rhino-bound is sufficient to induce piRNA production. One current model proposes that piRNAs clusters are defined and activated at each generation by the deposition in the egg of their corresponding piRNAs from the mother (Huang et al., 2017). In support of this model, we previously described the first case of a stable transgenerational epigenetic conversion known as paramutation in animals (de Vanssay et al., 2012). This phenomenon was first described in plants and defined as "an epigenetic interaction between two alleles of a locus, through which one allele induces a heritable modification of the other allele without modifying the DNA sequence" (Brink, 1956; Chandler, 2007). In our previous study, we showed that an inactive non-producing piRNA cluster of P transgene insertions inherited from the father can be converted into a piRNA-producing cluster by piRNAs inherited from the mother (de Vanssay et al., 2012). This attractive model, however, does not answer the question of how the first piRNAs were produced.

To address this paradox, we used the same BX2 cluster of seven P(lacW) transgenes, which resulted from multiple and successive P(lacW) transposition events, thus resembling the structure of natural piRNA clusters (Dorer and Henikoff, 1994). The key advantage of the BX2 locus is that it can exist in two epigenetic states for the production of germline piRNAs: 1) the inactive state (BX2^OFF) does not produce any piRNAs and thus is unable to repress the expression of homologous sequences, and 2) the active state (BX2^ON) produces abundant piRNAs that functionally repress a homologous reporter transgene in the female germline (de Vanssay et al., 2012; Hermant et al., 2015). We therefore used BX2 in an inactive state to search for conditions that would convert it into an active piRNA-producing locus, without pre-existing maternal piRNAs. In this report, we describe how culturing flies at high temperature, 29°C instead of 25°C, induces the conversion of an inactive BX2 locus (BX2^OFF) into a stable piRNA cluster exhibiting repression properties (BX2^ON). It should be noted that flies in their natural habitat exhibit this range of temperature, especially in the context of global warming. These data provide the first report of a de novo piRNA cluster establishment independent of maternal inheritance of homologous piRNAs and highlight how environmental changes can stably induce transgenerational modification of the epigenome.

Here, we report on the heritable establishment of a new piRNA cluster associated with silencing properties induced by high temperature during development. The epigenetic response to heat exposure has been studied in several model species: in Arabidopsis for instance, increasing temperature induces transcriptional activation of repetitive elements (Ito et al., 2011; Pecinka et al., 2010; Tittel-Elmer et al., 2010). Whether these changes involve chromatin modifications is not clear but none of these modifications have been found to be heritable through generations except in mutants for siRNA biogenesis where high frequency of new TE insertions was observed in the progeny of stressed plants (Ito et al., 2011). In animals, response to heat can result in modification of DNA methylation at specific loci in reef building coral (Dimond and Roberts, 2016), chicken (Yossifoff et al., 2008) and wild guinea pigs (Weyrich et al., 2016). In the latter, modifications affecting ≈50 genes are inherited in G1 progeny (Weyrich et al., 2016). The mechanisms of this heritability, however, are not yet understood. In Drosophila, heat-shock treatment of 0–2 hr embryos for one hour at 37°C or subjecting flies to osmotic stress induce phosphorylation of dATF-2 and its release from heterochromatin (Seong et al., 2011). This defective chromatin state is maintained for several generations before returning to the original state. We have tested if such stresses were able to convert BX2^OFF into BX2^ON in one generation but neither heat-shock nor osmotic stress induces BX2 conversion (Supplementary file 11), suggesting that BX2 activation does not depend on dATF-2. In a more recent paper, Fast et al. (Fast et al., 2017) found less piRNAs at 29°C than at 18°C. However, RNAseq analyses of differentially expressed genes involved in the piRNA pathway were not conclusive, as some piRNA genes (ago3, aub, zuc, armi) were more expressed at 29°C while others (shu, hsp83, Yb) were less expressed as compared to the levels at 18°C (Fast et al., 2017). Overall, the enhancement of the piRNA ping-pong amplification loop observed at 29°C was attributed to RNA secondary structures, because of a lack of specificity for any particular class of TE (Fast et al., 2017). Furthermore, in contrast to our RT-qPCR results obtained at 25°C and 29°C (Figure 4D and Figure 4—figure supplement 1A–B), AGO1 was not differentially expressed between 18°C and 29°C (Fast et al., 2017). This difference can be explained, because, in our experiments, we checked for specific spliced transcripts of AGO1 (RA, RC and RD) that originate upstream the BX2 insertion point (Figure 4C). In agreement with Fast et al., no statistical difference in the global amount of AGO1 transcripts between 25°C and 29°C is observed using primers located downstream the BX2 insertion point (Figure 4—figure supplement 1D). This suggests that the AGO1 promoter located downstream the BX2 insertion point, and responsible for the production of the RB isoform (see Figure 4), is not sensitive to temperature. Taken together, these observations show that temperature modification might induce epigenetic changes in several species but the underlying mechanisms remain largely unsolved.

Whole genome comparison of small RNA sequencing between BX2^OFF and newly BX2^ON heat-converted flies (BX2^Θ) did not reveal additional regions stably converted for piRNA production (Figure 2—figure supplement 3). This suggests that no other loci are metastable for piRNA production, that is all potential piRNA clusters are already active at 25°C. This observation raises the question of what makes BX2 locus competent for piRNA activation at high temperature. The BX2 locus is the result of successive induced transpositions of P(lacW) used to screen for white-variegating phenotype (Dorer and Henikoff, 1994). Thus, BX2 resembles natural TE clusters where TEs have the capacity to transpose into each other assembling structures named nested TE, as described in numerous genomes (Gao et al., 2012; Liu et al., 2007; Zanni et al., 2013). Some nested TE loci might have the capacity to be activated and respond to a new TE invasion. In Drosophila, BX2-like tandemly inserted transgenes were shown to be new sites of HP1 enrichment in larval salivary glands, emphasizing a heterochromatic structure at the BX2 locus in somatic cells (Fanti et al., 1998), and to cause pairing-dependent silencing (Dorer and Henikoff, 1997). When tested for repressing capacities, however, this strain is inactive for BX2 piRNA production (de Vanssay et al., 2012; Josse et al., 2008). We have shown that maternally inherited P(lacW) piRNAs are able to paramutate with complete and stable penetrance from an inactive BX2 locus into an active locus for piRNA production (de Vanssay et al., 2012). The paramutated BX2 locus appeared to be a genuine piRNA cluster since it is sensitive to a number of factors known to be involved in piRNA biogenesis such as aub, rhi, cuff, zuc (Hermant et al., 2015) and moonshiner (Figure 3—figure supplement 3). The number of transgene copies appears to be crucial in the process since smaller number of transgenes results in less somatic heterochromatinization (Dorer and Henikoff, 1994; Fanti et al., 1998), less pairing-dependent silencing (Dorer and Henikoff, 1997) and unstable paramutation (de Vanssay et al., 2012). Taken together, these data suggest that the heterochromatic structure of a cluster precedes piRNA production. This is supported by our ChIP experiments showing that H3K9me3 levels on BX2^OFF are slightly below the H3K9me3 level of piRNA producing states (BX2^ON and BX2^Θ, Figure 2D). The same observation can be made for Rhino (Figure 2E), suggesting that Rhino may be already present on the BX2^OFF locus but below the threshold required for piRNA production as suggested by Akulenko et al. (2018). Thus, a locus made of repeated sequences and being likely heterochromatic (H3K9me3, Rhino) is a necessary but not sufficient condition to specify an active piRNA cluster.

In the germline, piRNA clusters produce piRNAs from both strands and it was recently shown that, in most cases, transcription initiates within clusters on both strands through the interaction of Rhino and Moonshiner (Andersen et al., 2017). In few cases, however, piRNA cluster transcription may take advantage of the read-through from a flanking promoter (Andersen et al., 2017). Zhang et al. (Zhang et al., 2014) have shown that tethering Rhino onto a transgene leads to its repression but the production of piRNA depends on the presence of another transgene producing antisense RNA. Moreover, in the context of the Pld promoter deletion, a gene flanking the 42AB piRNA cluster, flies can produce Pld piRNAs only if a Pld cDNA is expressed in trans (Andersen et al., 2017). From all of these observations, a model emerges predicting that simultaneous production of sense and antisense RNA is a shared requirement for piRNA production. However, even if BX2^OFF is transcribed on both strands, without additional signals, it still remains inactive for piRNA production.

In addition to having a number of heterochromatic repeats and a double stranded transcription, the production of de novo piRNAs from BX2 requires a triggering signal. From our experiments, BX2 conversion relies on the simultaneous increase of both sense and antisense RNAs. An active role of euchromatic copies in the establishment of new piRNA clusters by high temperature appears to be consistent with what would naturally happen during the invasion of a naive genome by new TEs or when chromosomal breakages occur leading to the loss of piRNA cluster loci (Asif-Laidin et al., 2017). At first, uncontrolled euchromatic TE transposition takes place before the establishment of repression. Such repression would occur after a copy integrates into a preexisting piRNA cluster or by the generation of a new cluster made by successive insertion of nested copies. Consequently, clusters of elements cannot exist without transcriptionally active euchromatic copies. The increase of germline antisense transcripts upon stress or environmental factors, depending on the neighboring genomic environment, and the concomitant presence of numerous sense transcripts from euchromatic active copies, appear to be the starting signals for new piRNA production. These piRNAs can then be inherited at the next generation where they will stably paramutate the corresponding DNA locus with repetitive nature. At that time, the triggering signal is no longer necessary since BX2 remains activated once flies get back at 25°C. Future generations thus remember what was once considered a threat only through the legacy of maternal piRNAs.

Small RNA sequences and project have been deposited at the GEO under accession number GSE116122.

1. 1. Casier K
   2. Boivin A

   (2018) NCBI Gene Expression Omnibus

   ID GSE116122. Environmentally-induced epigenetic conversion of a piRNA cluster.

   https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?&acc=GSE116122

1. 1. Akulenko N
   2. Ryazansky S
   3. Morgunova V
   4. Komarov PA
   5. Olovnikov I
   6. Vaury C
   7. Jensen S
   8. Kalmykova A

   (2018) Transcriptional and chromatin changes accompanying de novo formation of transgenic piRNA clusters

   Rna 24:574–584.

   https://doi.org/10.1261/rna.062851.117

   • PubMed
   • Google Scholar
2. 1. Andersen PR
   2. Tirian L
   3. Vunjak M
   4. Brennecke J

   (2017) A heterochromatin-dependent transcription machinery drives piRNA expression

   Nature 549:54–59.

   https://doi.org/10.1038/nature23482

   • PubMed
   • Google Scholar
3. 1. Antoniewski C

   (2014) Computing siRNA and piRNA overlap signatures

   Methods in Molecular Biology 1173:135–146.

   https://doi.org/10.1007/978-1-4939-0931-5_12

   • PubMed
   • Google Scholar
4. 1. Asif-Laidin A
   2. Delmarre V
   3. Laurentie J
   4. Miller WJ
   5. Ronsseray S
   6. Teysset L

   (2017) Short and long-term evolutionary dynamics of subtelomeric piRNA clusters in Drosophila

   DNA Research 24:459–472.

   https://doi.org/10.1093/dnares/dsx017

   • PubMed
   • Google Scholar
5. 1. Brennecke J
   2. Aravin AA
   3. Stark A
   4. Dus M
   5. Kellis M
   6. Sachidanandam R
   7. Hannon GJ

   (2007) Discrete small RNA-generating loci as master regulators of transposon activity in Drosophila

   Cell 128:1089–1103.

   https://doi.org/10.1016/j.cell.2007.01.043

   • PubMed
   • Google Scholar
6. 1. Brennecke J
   2. Malone CD
   3. Aravin AA
   4. Sachidanandam R
   5. Stark A
   6. Hannon GJ

   (2008) An epigenetic role for maternally inherited piRNAs in transposon silencing

   Science 322:1387–1392.

   https://doi.org/10.1126/science.1165171

   • PubMed
   • Google Scholar
7. 1. Brink RA

   (1956)

   A genetic change associated with the R locus in maize which is directed and potentially reversible

   Genetics 41:872–889.

   • PubMed
   • Google Scholar
8. 1. Chandler VL

   (2007) Paramutation: from maize to mice

   Cell 128:641–645.

   https://doi.org/10.1016/j.cell.2007.02.007

   • PubMed
   • Google Scholar
9. 1. Crooks GE
   2. Hon G
   3. Chandonia JM
   4. Brenner SE

   (2004) WebLogo: a sequence logo generator

   Genome Research 14:1188–1190.

   https://doi.org/10.1101/gr.849004

   • PubMed
   • Google Scholar
10. 1. de Vanssay A
    2. Bougé AL
    3. Boivin A
    4. Hermant C
    5. Teysset L
    6. Delmarre V
    7. Antoniewski C
    8. Ronsseray S

    (2012) Paramutation in Drosophila linked to emergence of a piRNA-producing locus

    Nature 490:112–115.

    https://doi.org/10.1038/nature11416

    • PubMed
    • Google Scholar
11. 1. Dimond JL
    2. Roberts SB

    (2016) Germline DNA methylation in reef corals: patterns and potential roles in response to environmental change

    Molecular Ecology 25:1895–1904.

    https://doi.org/10.1111/mec.13414

    • PubMed
    • Google Scholar
12. 1. Dorer DR
    2. Henikoff S

    (1994) Expansions of transgene repeats cause heterochromatin formation and gene silencing in Drosophila

    Cell 77:993–1002.

    https://doi.org/10.1016/0092-8674(94)90439-1

    • PubMed
    • Google Scholar
13. 1. Dorer DR
    2. Henikoff S

    (1997)

    Transgene repeat arrays interact with distant heterochromatin and cause silencing in cis and trans

    Genetics 147:1181–1190.

    • PubMed
    • Google Scholar
14. 1. Fanti L
    2. Dorer DR
    3. Berloco M
    4. Henikoff S
    5. Pimpinelli S

    (1998) Heterochromatin protein 1 binds transgene arrays

    Chromosoma 107:286–292.

    https://doi.org/10.1007/s004120050310

    • PubMed
    • Google Scholar
15. 1. Fast I
    2. Hewel C
    3. Wester L
    4. Schumacher J
    5. Gebert D
    6. Zischler H
    7. Berger C
    8. Rosenkranz D

    (2017) Temperature-responsive miRNAs in Drosophila orchestrate adaptation to different ambient temperatures

    Rna 23:1352–1364.

    https://doi.org/10.1261/rna.061119.117

    • PubMed
    • Google Scholar
16. 1. Frydman HM
    2. Spradling AC

    (2001)

    The receptor-like tyrosine phosphatase lar is required for epithelial planar polarity and for axis determination within Drosophila ovarian follicles

    Development 128:3209–3220.

    • PubMed
    • Google Scholar
17. 1. Gao C
    2. Xiao M
    3. Ren X
    4. Hayward A
    5. Yin J
    6. Wu L
    7. Fu D
    8. Li J

    (2012) Characterization and functional annotation of nested transposable elements in eukaryotic genomes

    Genomics 100:222–230.

    https://doi.org/10.1016/j.ygeno.2012.07.004

    • PubMed
    • Google Scholar
18. 1. Gunawardane LS
    2. Saito K
    3. Nishida KM
    4. Miyoshi K
    5. Kawamura Y
    6. Nagami T
    7. Siomi H
    8. Siomi MC

    (2007) A slicer-mediated mechanism for repeat-associated siRNA 5' end formation in Drosophila

    Science 315:1587–1590.

    https://doi.org/10.1126/science.1140494

    • PubMed
    • Google Scholar
19. 1. Hermant C
    2. Boivin A
    3. Teysset L
    4. Delmarre V
    5. Asif-Laidin A
    6. van den Beek M
    7. Antoniewski C
    8. Ronsseray S

    (2015) Paramutation in Drosophila requires both nuclear and cytoplasmic actors of the piRNA pathway and induces Cis-spreading of piRNA production

    Genetics 201:1381–1396.

    https://doi.org/10.1534/genetics.115.180307

    • PubMed
    • Google Scholar
20. 1. Huang X
    2. Fejes Tóth K
    3. Aravin AA

    (2017) piRNA biogenesis in Drosophila melanogaster

    Trends in Genetics 33:882–894.

    https://doi.org/10.1016/j.tig.2017.09.002

    • PubMed
    • Google Scholar
21. 1. Ito H
    2. Gaubert H
    3. Bucher E
    4. Mirouze M
    5. Vaillant I
    6. Paszkowski J

    (2011) An siRNA pathway prevents transgenerational retrotransposition in plants subjected to stress

    Nature 472:115–119.

    https://doi.org/10.1038/nature09861

    • PubMed
    • Google Scholar
22. 1. Josse T
    2. Maurel-Zaffran C
    3. de Vanssay A
    4. Teysset L
    5. Todeschini AL
    6. Delmarre V
    7. Chaminade N
    8. Anxolabéhère D
    9. Ronsseray S

    (2008) Telomeric trans-silencing in Drosophila melanogaster: tissue specificity, development and functional interactions between non-homologous telomeres

    PLoS ONE 3:e3249.

    https://doi.org/10.1371/journal.pone.0003249

    • PubMed
    • Google Scholar
23. 1. Klattenhoff C
    2. Xi H
    3. Li C
    4. Lee S
    5. Xu J
    6. Khurana JS
    7. Zhang F
    8. Schultz N
    9. Koppetsch BS
    10. Nowosielska A
    11. Seitz H
    12. Zamore PD
    13. Weng Z
    14. Theurkauf WE

    (2009) The Drosophila HP1 homolog rhino is required for transposon silencing and piRNA production by dual-strand clusters

    Cell 138:1137–1149.

    https://doi.org/10.1016/j.cell.2009.07.014

    • PubMed
    • Google Scholar
24. 1. Kofler R
    2. Senti K-A
    3. Nolte V
    4. Tobler R
    5. Schlötterer C

    (2018) Molecular dissection of a natural transposable element invasion

    Genome Research 28:824–835.

    https://doi.org/10.1101/gr.228627.117

    • Google Scholar
25. 1. Langmead B
    2. Trapnell C
    3. Pop M
    4. Salzberg SL

    (2009) Ultrafast and memory-efficient alignment of short DNA sequences to the human genome

    Genome Biology 10:R25.

    https://doi.org/10.1186/gb-2009-10-3-r25

    • PubMed
    • Google Scholar
26. 1. Le Thomas A
    2. Rogers AK
    3. Webster A
    4. Marinov GK
    5. Liao SE
    6. Perkins EM
    7. Hur JK
    8. Aravin AA
    9. Tóth KF

    (2013) Piwi induces piRNA-guided transcriptional silencing and establishment of a repressive chromatin state

    Genes & Development 27:390–399.

    https://doi.org/10.1101/gad.209841.112

    • PubMed
    • Google Scholar
27. 1. Le Thomas A
    2. Stuwe E
    3. Li S
    4. Du J
    5. Marinov G
    6. Rozhkov N
    7. Chen YC
    8. Luo Y
    9. Sachidanandam R
    10. Toth KF
    11. Patel D
    12. Aravin AA

    (2014) Transgenerationally inherited piRNAs trigger piRNA biogenesis by changing the chromatin of piRNA clusters and inducing precursor processing

    Genes & Development 28:1667–1680.

    https://doi.org/10.1101/gad.245514.114

    • PubMed
    • Google Scholar
28. 1. Lemaitre B
    2. Ronsseray S
    3. Coen D

    (1993)

    Maternal repression of the P element promoter in the germline of Drosophila melanogaster: a model for the P cytotype

    Genetics 135:149–160.

    • PubMed
    • Google Scholar
29. 1. Liu R
    2. Vitte C
    3. Ma J
    4. Mahama AA
    5. Dhliwayo T
    6. Lee M
    7. Bennetzen JL

    (2007) A GeneTrek analysis of the maize genome

    PNAS 104:11844–11849.

    https://doi.org/10.1073/pnas.0704258104

    • PubMed
    • Google Scholar
30. 1. Marie PP
    2. Ronsseray S
    3. Boivin A

    (2017) From embryo to adult: pirna-mediated silencing throughout germline development in Drosophila

    G3: Genes|Genomes|Genetics 7:505–516.

    https://doi.org/10.1534/g3.116.037291

    • Google Scholar
31. 1. Mlodzik M
    2. Hiromi Y

    (1992) Enhancer trap method in Drosophila: its application to neurobiology

    Methods in Neurosciences pp. 397–414.

    https://doi.org/10.1016/b978-0-12-185267-2.50030-1

    • Google Scholar
32. 1. Mohn F
    2. Sienski G
    3. Handler D
    4. Brennecke J

    (2014) The rhino-deadlock-cutoff complex licenses noncanonical transcription of dual-strand piRNA clusters in Drosophila

    Cell 157:1364–1379.

    https://doi.org/10.1016/j.cell.2014.04.031

    • PubMed
    • Google Scholar
33. 1. Muerdter F
    2. Olovnikov I
    3. Molaro A
    4. Rozhkov NV
    5. Czech B
    6. Gordon A
    7. Hannon GJ
    8. Aravin AA

    (2012) Production of artificial piRNAs in flies and mice

    Rna 18:42–52.

    https://doi.org/10.1261/rna.029769.111

    • Google Scholar
34. 1. Pecinka A
    2. Dinh HQ
    3. Baubec T
    4. Rosa M
    5. Lettner N
    6. Mittelsten Scheid O

    (2010) Epigenetic regulation of repetitive elements is attenuated by prolonged heat stress in arabidopsis

    The Plant Cell 22:3118–3129.

    https://doi.org/10.1105/tpc.110.078493

    • PubMed
    • Google Scholar
35. 1. Pöyhönen M
    2. de Vanssay A
    3. Delmarre V
    4. Hermant C
    5. Todeschini AL
    6. Teysset L
    7. Ronsseray S

    (2012) Homology-dependent silencing by an exogenous sequence in the Drosophila germline

    G3: Genes|Genomes|Genetics 2:331–338.

    https://doi.org/10.1534/g3.111.001925

    • Google Scholar
36. 1. Ronsseray S
    2. Anxolabéhère D
    3. Périquet G

    (1984) Hybrid dysgenesis in Drosophila melanogaster: influence of temperature on cytotype determination in the P-M system

    MGG Molecular & General Genetics 196:17–23.

    https://doi.org/10.1007/BF00334086

    • PubMed
    • Google Scholar
37. 1. Ronsseray S

    (1986) P-M system of hybrid dysgenesis inDrosophila melanogaster: thermic modifications of the cytotype can be detected for several generations

    MGG Molecular & General Genetics 205:23–27.

    https://doi.org/10.1007/BF02428028

    • Google Scholar
38. 1. Ronsseray S
    2. Lehmann M
    3. Anxolabéhère D

    (1991)

    The maternally inherited regulation of P elements in Drosophila melanogaster can be elicited by two P copies at cytological site 1A on the X chromosome

    Genetics 129:501–512.

    • PubMed
    • Google Scholar
39. 1. Seong KH
    2. Li D
    3. Shimizu H
    4. Nakamura R
    5. Ishii S

    (2011) Inheritance of stress-induced, ATF-2-dependent epigenetic change

    Cell 145:1049–1061.

    https://doi.org/10.1016/j.cell.2011.05.029

    • PubMed
    • Google Scholar
40. 1. Sienski G
    2. Dönertas D
    3. Brennecke J

    (2012) Transcriptional silencing of transposons by piwi and maelstrom and its impact on chromatin state and gene expression

    Cell 151:964–980.

    https://doi.org/10.1016/j.cell.2012.10.040

    • PubMed
    • Google Scholar
41. 1. Tittel-Elmer M
    2. Bucher E
    3. Broger L
    4. Mathieu O
    5. Paszkowski J
    6. Vaillant I

    (2010) Stress-induced activation of heterochromatic transcription

    PLoS Genetics 6:e1001175.

    https://doi.org/10.1371/journal.pgen.1001175

    • PubMed
    • Google Scholar
42. 1. Weyrich A
    2. Lenz D
    3. Jeschek M
    4. Chung TH
    5. Rübensam K
    6. Göritz F
    7. Jewgenow K
    8. Fickel J

    (2016) Paternal intergenerational epigenetic response to heat exposure in male wild guinea pigs

    Molecular Ecology 25:1729–1740.

    https://doi.org/10.1111/mec.13494

    • PubMed
    • Google Scholar
43. 1. Wylie A
    2. Jones AE
    3. Abrams JM

    (2016a) p53 in the game of transposons

    BioEssays 38:1111–1116.

    https://doi.org/10.1002/bies.201600115

    • PubMed
    • Google Scholar
44. 1. Wylie A
    2. Jones AE
    3. D'Brot A
    4. Lu WJ
    5. Kurtz P
    6. Moran JV
    7. Rakheja D
    8. Chen KS
    9. Hammer RE
    10. Comerford SA
    11. Amatruda JF
    12. Abrams JM

    (2016b) p53 genes function to restrain mobile elements

    Genes & Development 30:64–77.

    https://doi.org/10.1101/gad.266098.115

    • PubMed
    • Google Scholar
45. 1. Yossifoff M
    2. Kisliouk T
    3. Meiri N

    (2008) Dynamic changes in DNA methylation during thermal control establishment affect CREB binding to the brain-derived neurotrophic factor promoter

    European Journal of Neuroscience 28:2267–2277.

    https://doi.org/10.1111/j.1460-9568.2008.06532.x

    • PubMed
    • Google Scholar
46. 1. Zanni V
    2. Eymery A
    3. Coiffet M
    4. Zytnicki M
    5. Luyten I
    6. Quesneville H
    7. Vaury C
    8. Jensen S

    (2013) Distribution, evolution, and diversity of retrotransposons at the flamenco locus reflect the regulatory properties of piRNA clusters

    PNAS 110:19842–19847.

    https://doi.org/10.1073/pnas.1313677110

    • PubMed
    • Google Scholar
47. 1. Zhang Z
    2. Wang J
    3. Schultz N
    4. Zhang F
    5. Parhad SS
    6. Tu S
    7. Vreven T
    8. Zamore PD
    9. Weng Z
    10. Theurkauf WE

    (2014) The HP1 homolog rhino anchors a nuclear complex that suppresses piRNA precursor splicing

    Cell 157:1353–1363.

    https://doi.org/10.1016/j.cell.2014.04.030

    • PubMed
    • Google Scholar

[Editorial note: This article has been through an editorial process in which the authors decide how to respond to the issues raised during peer review. The Reviewing Editor's assessment is that all the issues have been addressed.]

Thank you for submitting your article "Environmentally-induced epigenetic conversion of a piRNA cluster" for consideration by eLife. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by a Reviewing Editor and James Manley as the Senior Editor. The reviewers have opted to remain anonymous.

We have included the separate reviews below for your consideration. We would recommend that you first respond to outline your replies to the reviewers and what revisions you intend to make, so that we can offer further guidance. We note that all the reviewers felt that more mechanistic insight would be desirable. If you have any questions, please do not hesitate to contact us.

Separate reviews (please respond to each point):

Reviewer #1:

The authors report how high temperature (environment) can induce a DNA stretch to become a piRNA-producing locus in female flies.

Findings

1) Some (25% of the females studied) flies within a population are randomly induced to turn an inactive piece of DNA into an active locus that generates piRNAs, when grown at higher temperature (29^oC instead of normal 25^oC). This is monitored by the clever use of a b-gal reporter assay. This induction increases with generation time.

2) Such temperature-activated clusters are identical to those activated by maternally-inherited piRNAs. They present similar piRNAs produced and are decorated with the expected chromatin and piRNA-factor marks.

3) The higher temperature increased antisense RNA production from the previously-silent locus and this was channelled into the siRNA pathway and did not lead to increased piRNA production.

4) The presence of the complementary target RNA (expressed from a euchromatic locus) was essential for the high temperature-triggered piRNA generation.

Concerns

1) Abstract, Second sentence: There is no data to show that piRNA clusters are maternally defined except in flies. In mice germ cells are induced from the somatic lineage at embryonic day 7.5 and there is no possibility of maternal inheritance playing a role. If the authors want to keep this sentence, specify it is for flies.

2) Abstract, third sentence: The sentence sounds very dramatic as it means that higher temperature alone was sufficient to convert the P-transgene locus into a piRNA cluster. Perhaps the features of the genomic region (having P-transgenes), location, chromatin status etc are responsible. I would rephrase it. It is misleading as although temperature might be the trigger, this locus might already be primed and ready to go.

3) The authors make a very striking finding regarding high temperature-induced piRNA production from a locus, and link it to presence of both sense and antisense transcripts are necessary for this. This still leaves open the question how the locus was made into piRNAs. There is no explanation for this.

4) It would be good to have the actual sequence information for the different transgenes and the b-gal reporter, and the actual transcripts present in these flies. One is left in the dark as to the level of complementarity that exists between them when they are referred to as sense and antisense transcripts.

In conclusion, the observations made are striking and extremely interesting, but perhaps a bit more of molecular insights might be useful to appreciate the observations made.

Reviewer #2:

Piwi-interacting RNAs (piRNAs) produced from piRNA clusters repress transposable elements in animal gonads. It has been shown that use of piRNA clusters is determined by maternally-deposited piRNAs. In this study, Casier et al. however showed that raising flies at high temperature (29^oC) resulted in ectopic expression of piRNAs from the BX2 transgene locus in flies that had no maternally-deposited, BX2-derived piRNAs. The BX2 transgene consists of P5’, lacZ, white, a part of pBR322 (pBR) and P3’, and does not express piRNAs at normal temperature (25^oC). The authors claimed that this was the first case showing that heat treatment can activate de novo piRNA production from the BX2 locus (i.e., non-piRNA cluster) in a manner independent of maternally-deposited piRNAs.

The authors' findings also include:

1) Activation of de novo piRNA production by heat treatment was restricted to the BX2 locus.

2) BX2-derived piRNAs produced upon heat treatment showed molecular properties similar to those of BX2-derived piRNAs expressed in a manner dependent of maternally-deposited piRNAs.

3) Ectopically expressed lacZ-piRNAs from the BX2 locus were capable of silencing βGal over generations.

4) The βGal silencing effect was low at the first generation but gradually increased over generations.

5) Rhino and H3K9me3 accumulated at the BX2 locus upon heat treatment. However, this may not be the cause to ectopic expression of BX2-derived piRNAs, as the authors claimed that BX2^OFF and BX2^ON produced anyway similar amounts of sense and antisense transcripts (subsection “Epigenetic conversion at 29°C occurs at a low rate from the first generation”).

6) An increase of transcription from the AGO1 promoter at 29^oC could lead to an increase of BX2 antisense RNA transcription. The BX2 transgene was inserted in the AGO1 gene in a convergent transcription manner.

7) Activation of BX2-derived piRNA production by heat treatment required a homologous sequence.

I found the authors' finding that heat treatment induces de novo piRNA production from non-piRNA clusters potentially interesting. However, the mechanistic insights remain vague. The authors claimed that homologous sequence was required for the induction. However, to claim this, much more supportive data should be provided. The authors also found that H3K9me3 and Rhino started to accumulate at the BX2 locus upon heat treatment. However, the meaning of the accumulation remains unclear. My other concerns were indicated below. I hope that this reviewer's concerns might be helpful for the authors to revise the manuscript.

Concerns and suggestions from this reviewer:

1) The authors proposed that the interaction between the excess of BX2 antisense transcripts and the sense P(TARGET) transcripts is a prerequisite for the production of de novo piRNAs (subsection “Heat conversion requires a homologous sequence in trans”). To confirm this, I recommend the authors conducting experiments using heat treated BX2^OFF that do not have lacZ at the BX2 loci but contains P(TARGET).

2) The level of AGO1 RNA transcripts should be examined in BX2^OFF line containing P(TARGET). I do not understand why the authors used the line without P(TARGET) in this particular experiment.

3) Figure 4B: Upon heat treatment, the expression level of BX2 antisense transcripts was raised. Was this phenomenon related to Rhino and H3K9me3 accumulation observed in Figure 2DE? Experiments in Figure 2 and Figure 4 should be performed using the same fly lines.

4) Is the BX2 transgene in the BX2^OFF line containing P(TARGET) also inserted in the AGO1 gene? This should be examined. If this were the case, the authors better examine whether piRNAs were derived from the AGO1 gene.

5) The authors should test experimentally whether BX2-derived piRNAs are loaded into PIWI proteins. This is very important.

6) Figure 2B: Explain why BX2-derived piRNAs did not show 10A bias. The data shown in Figure 2C supported the idea that BX2-derived piRNAs were products of the ping-pong pathway. Then, they should have shown 10A bias, but in reality they did not.

7) Figure 2DE: The authors should examine genome-widely where in the genome H3K9me3 and Rhino accumulated upon heat treatment. It is hard to imagine that H3K9me3 and Rhino accumulation only occurred at the BX2 locus upon heat treatment.

8) The authors should examine whether H3K9me3 and Rhino accumulated at the BX2 locus in BX2^OFF line without P(TARGET) upon heat treatment.

9) This manuscript should be edited thoroughly by a native English speaker.

10) Table 1 should be removed from the main text.

Reviewer #3:

This work takes advantage of a clever system using suppression of a lacZ reporter gene as a read out indicating the production of piRNAs from the BX2 locus. Use of this system has resulted in several novel and interesting results, most notably the paramutation phenomenon. In this work, Casier et al. show that the BX2 locus is silent at 25^oC, but produces piRNAs at 29 ^oC in the presence of homologous sequence in the background.

While I have no problem with the experiments, but did wish for more insight into the mechanism, and more caution in the interpretation. In particular, entirely environmental specification of piRNA clusters implied in the Abstract is a strong claim, and as the authors show themselves, is an oversimplification. The induction they find is quite specific to their system. Specifically, there are higher levels of homologous sense and anti-sense transcript at 29 than at 25, and temperature alone doesn't change other piRNA production. Similarly, it would be useful to know if the expression of the sense transcript from P-Target is also higher at 29 that at 25. (I think this construct has a heat-shock promoter, as well as that of the P-element, which shows temperature sensitive effects.)

I wasn't entirely convinced by the suggestion that expression of natural piRNA clusters is higher at 29 ^oC. Many piRNAs are generated as a secondary byproduct of transposable element message, and many transposable elements seem to have temperature sensitive expression (possibly as a byproduct of changes in chromatin.) The piRNA dependent splicing suppression of the P-element is also, apparently, not temperature sensitive (Teixiera et al., 2017).

Minor points. It does seem important (and quick) to eliminate the possibility that there's any active P-elements in these lines. While w1118 is a classic M-type background, there are some cases of sublines of other M-types harboring P-elements via contamination (Rahman et al. Nucleic Acids Research, Volume 43, Issue 22, 15 December 2015, Pages 10655-10672). If there is active P-element, I think it is possible that transgenes (which has the TIRs necessary for P-element insertion) has been picked up and put in another piRNA cluster. It's a simple matter of a few PCRs to exclude this possibility. The Cy and CyRoi backgrounds should be tested as well, to show that they are devoid of any plus-strand transcript that could be an ongoing trigger for piRNA production.

Subsection “Epigenetic conversion at 29°C occurs at a low rate from the first generation” paragraph two. The model is entirely sensible, but the data are quite noisy, and seem like they would be consistent with any model predicting an increase over generations. Can they compare this model to, for example, a model without the “c” parameter using Akaike Information Criteria or a log-likelihood test? This would make the analysis more informative.

Subsection “Heat conversion requires a homologous sequence in trans”: The chi-square statistic and degrees of freedom should be reported as well as the p-value. (And this is very minor, but the Greek letter is usually used in place of “chi”.)

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

Thank you for resubmitting your work entitled "Environmentally-induced epigenetic conversion of a piRNA cluster" for further consideration at eLife. Your revised article has been evaluated by James Manley (Senior Editor), a Reviewing Editor, and two reviewers.

The manuscript has been improved but the referees still have significant concerns. If you decide to proceed with publication of the present version, our editorial rating, which will be displayed immediately below the Abstract, would be that "major issues remain unresolved."

The reviewer comments (on the original version and revised version) would also be published, along with your responses.

Alternatively, you might decide to undertake significant extra work to try to address the concerns. Or, you might decide to shorten the paper (highlighting the most interesting findings) and leaving extensive additional work for a future paper. You also have the option to withdraw the present paper from consideration.

The reviews follow below, and we will look forward to hearing how you propose to proceed.

Reviewer #1:

In the revised version of the manuscript the authors find an involvement of sense-antisense transcripts and Dicer-2 in piRNA biogenesis. The authors may remember that one of the early findings celebrated as a major discovery was the lack of a role for double-stranded RNAs and the double-stranded RNA cleaving enzyme Dicer, in piRNA biogenesis. If now, the authors find a role for these in piRNA biogenesis, it has to be backed with strong data.

1) Can the authors demonstrate loading of the 21 nt RNAs into Piwi proteins?

2) If its role is epigenetic conversion, then ideally it has be loaded into the nuclear Piwi.

3) Are there experiments linking Dicer-2 to loading into a Piwi protein? Are they ever found in the same complexes?

I agree that there is a strong functional/genetic data that is mysterious, but very interesting. The heat-activated piRNAs are functional as they can silence a LacZ transgene. Since this manuscript will be published in some form, I am happy to support description of this part in a revised manuscript, but without wild speculation of Dicer-2 in piRNA biogenesis (even if they examine a Dicer mutant). These attempts at explaining the molecular mechanism may be toned down. I am worried that any prominent mention of Dicer-2 in this context may only serve to foster suspicion of the interesting genetic observation.

Reviewer #3:

I still find this study clever and interesting, and think the manuscript shows some specific improvements:

The “primed” nature of the BX2 locus is clearer, and the claim in the paper is more measured and reasonably justified.

The experiment showing the Dcr2 dependence of the conversion of the BX2 from off to on goes a little way toward understanding the mechanism of piRNA cluster formation, implying that it is dependent on siRNAs.

The addition of the figure showing the model for the mechanism. The weak point of the proposed mechanism, as I see it, is the lack of evidence that the siRNAs are loaded onto PIWI, any evidence the authors can provide for this mechanism would greatly improve the paper. I can't make any suggestions that seem technically feasible, however, in light of the low conversion rate in this system.

Further clarification would also be helpful for these aspects:

Regarding the model in Figure 1—figure supplement 3: I previously suggested a statistical analysis of this model. Rather than the more detailed analysis I suggested previously, I think a reasonable compromise would be to show that there is a significant increase over generations. That is the important point here (and, in fact, it's hard to imagine a scenario where the “memory” of conversion does not play a role in the increase over generations). But, while there appears to be a trend, the data are sufficiently noisy that it would be useful just to see that the increase is significant (perhaps a linear regression on transformed data?).

Discussion opening paragraph: Please explain what you mean by the “specific spliced transcript” comment; it wasn't clear.

e.g., supplementary file 11: It would be nice to more cautious in the interpretation where there are large differences in the number of flies examined. In Figure 3, for example, I think the numbers show that 41 of 1447 females show partial or complete repression; the comparable (?) numbers is Supplementary file 11 show 0 of 32 (?) females show any repression. This is not a significant difference via Fisher's exact test.

Additional data files and statistical comments:

Please see major comments for suggestions regarding statistical analyses, particularly in cases where conclusions are drawn about the absence of repression in small sample sizes.

Reviewer #1:

[…] Concerns

1) Abstract, Second sentence: There is no data to show that piRNA clusters are maternally defined except in flies. In mice germ cells are induced from the somatic lineage at embryonic day 7.5 and there is no possibility of maternal inheritance playing a role. If the authors want to keep this sentence, specify it is for flies.

We have specified that the hypothesis of the determination of piRNA cluster by maternal inheritance of homologous piRNA is for flies.

2) Abstract, third sentence: The sentence sounds very dramatic as it means that higher temperature alone was sufficient to convert the P-transgene locus into a piRNA cluster. Perhaps the features of the genomic region (having P-transgenes), location, chromatin status etc are responsible. I would rephrase it. It is misleading as although temperature might be the trigger, this locus might already be primed and ready to go.

The reviewer #1 is completely right, our results show that the locus is ready to become a piRNA cluster and the temperature is the trigger for the activation. We have rephrased the sentence.

3) The authors make a very striking finding regarding high temperature-induced piRNA production from a locus, and link it to presence of both sense and antisense transcripts are necessary for this. This still leaves open the question how the locus was made into piRNAs. There is no explanation for this.

This is the major point raised by the three reviewers. The major concern in our system is the low occurrence of the epigenetic conversion (2%). It was possible to identify and quantify it through generations because of the use of the high reliability of the ßGalactosidase staining allowing us to (i) visualize the status of each eggs chambers individually at each generation, (2) study a large number of flies (Figure 1 and 3, for instance), (3) do statistics. However, observing 2% difference during the process of activation by molecular analysis appeared impossible (for example, the observation of 2% increase of piRNAs or chromatin modification through ChIP analyses is not statistically possible). Therefore, molecular analyses were performed on stable epigenetic states (BX2^OFF or BX2^ON converted by maternally inherited piRNAs or by high temperature) and after being transferred at 25°C.

In order to answer to this major point, we performed additional experiments to test if double stranded RNAs should be formed prior piRNA biogenesis to activate BX2. So, we tested whether Dcr-2 might affect the activation of BX2 by temperature. We generated flies containing BX2, Dcr-2 mutant allele and a P(TARGET) tested the BX2 activation at each generation by ßGalactosidase staining. We present results showing that Dcr-2 mutation impairs the activation of BX2 by high temperature (Figure 3—figure supplement 3). This result is consistent with the requirement of double strand RNA intermediates for BX2 conversion into a piRNA cluster. We have added these experiments at the end of the result section as well as an additional figure (Figure 5) containing a cartoon putting together all of the results in a model. We hope that these new data will satisfy the reviewer concerns on the mechanism of activation.

4) It would be good to have the actual sequence information for the different transgenes and the b-gal reporter, and the actual transcripts present in these flies. One is left in the dark as to the level of complementarity that exists between them when they are referred to as sense and antisense transcripts.

We have added the precise structure of all transgenes used in this study showing that the complementarity between the BX2 and the P(TARGET) transgenes is restricted to the lacZ gene and the 5' and 3' region of the P element required for transgenesis (new Figure 1—figure supplement 1).We have specified what are sense and antisense BX2 transcripts (they correspond to sense and antisense lacZ transcripts, respectively) in the text and in the legend of the Figure 4.

In conclusion, the observations made are striking and extremely interesting, but perhaps a bit more of molecular insights might be useful to appreciate the observations made.

We appreciate the interest of this reviewer to our work and hope that the additional genetic results obtained with others euchromatic transgenes (see below) and with Dcr-2 sufficiently improve the molecular understanding of the phenomenon we are describing.

Reviewer #2:

[…] I found the authors' finding that heat treatment induces de novo piRNA production from non-piRNA clusters potentially interesting. However, the mechanistic insights remain vague.

As previously described, we hope that the experiments with others transgenes (see below) and with Dcr-2 give more insight on the mechanistic part of the phenomenon. Based on the new results presented in this revised version, we have added a new figure presenting a model of mechanism (Figure 5).

The authors claimed that homologous sequence was required for the induction. However, to claim this, much more supportive data should be provided.

We have performed additional experiments using another target, a P-lacZ-rosy reporter transgene inserted in euchromatin and expressed in the germline (P(A92), BQ16, FBti0003435). In the presence of this transgene, BX2 was converted at 29°C at the same rate than with P(TARGET) showing that BX2 conversion of BX2 by temperature cannot be attributed to any specificity linked to the P(TARGET) insertion. We have added these results in the main text and in a supplementary table (Supplementary file 10).Moreover, the potential effect of the background has been now tested over 30 generations and no conversion event has been observed, reinforcing the specific role of the germline expressed target in the activation of BX2. The Supplementary file 9 has been completed. We also performed a new experiment using a P(A92) transgene that is expressed in the somatic follicle cells but not in the germline nurse cells. With this transgene, no conversion was observed, showing that the euchromatic homologous transgene need to be transcribed (Figure 3—figure supplement 2). This reinforces the role of the "sense" transcripts brought by the euchromatic transgene in the conversion process.

The authors also found that H3K9me3 and Rhino started to accumulate at the BX2 locus upon heat treatment. However, the meaning of the accumulation remains unclear. My other concerns were indicated below. I hope that this reviewer's concerns might be helpful for the authors to revise the manuscript.

As a matter of fact, the enrichment in H3K9me3 and Rhino on the BX2 locus can be interpreted by a progressive accumulation of all the loci present in egg chambers studied or by a progressive increase number of egg chambers activated through generation (each one of them being fully activated or not activated at all, ON/OFF conversion). Based on our results obtained by lacZ staining showing exclusively egg chambers converted or not converted with low occurrence, we favor an on/off conversion per egg chamber induced by maternal inheritance of homologous piRNA or by temperature.

Concerns and suggestions from this reviewer:

1) The authors proposed that the interaction between the excess of BX2 antisense transcripts and the sense P(TARGET) transcripts is a prerequisite for the production of de novo piRNAs (subsection “Heat conversion requires a homologous sequence in trans”). To confirm this, I recommend the authors conducting experiments using heat treated BX2^OFF that do not have lacZ at the BX2 loci but contains P(TARGET).

This could be a good idea but we cannot modify at will the BX2 locus that had been obtained through several rounds of transposition followed by thousand of flies selection (Dorer and Henik_off 1994). P-transgenes have the capacity to inserted in the genome without hot spots (except gene regions rich), therefore it seems to be very difficult to recover 7 insertions within the same location. Moreover, if 7 P(lacW) was identified in another locus, it will not have necessary the same biological properties than the BX2 locus.

2) The level of AGO1 RNA transcripts should be examined in BX2^OFF line containing P(TARGET). I do not understand why the authors used the line without P(TARGET) in this particular experiment.

The presence of the BX2 cluster may affect specific transcripts of one allele of the AGO1 gene and to be sure to measure the effect of temperature on AGO1 transcripts, we performed experiments without BX2. However, we have reproduced the measurement of AGO1 steady-state RNA level by qRT-PCR experiment in a BX2 context, with and without P(TARGET) and confirmed that the amount of AGO1 transcripts is higher at 29°C compared to 25°C. These results have been added in the Figure 2—figure supplement 2.

3) Figure 4B: Upon heat treatment, the expression level of BX2 antisense transcripts was raised. Was this phenomenon related to Rhino and H3K9me3 accumulation observed in Figure 2DE?

The two studies were addressing different questions:

The Figure 2D-E was studying the molecular changes of a stable epigenetic conversion: that is to say the BX2 locus was activated by heat treatment after several generations at 29°C and this activation is stable because it is maintained even when flies are incubated at 25°C (the classical temperature). In this context, we compared a stable OFF line and stable ON lines (converted by maternally inherited piRNAs or by temperature) for differences of chromatin marks. Statistical but weak differences were identified (Figure 2).

In the Figure 4, we were addressing the mechanism of the triggering, that is to say what is happening after one generation at 29°C. Therefore, in this experiment, we were not expecting to be able to identify chromatin differences with a 2% conversion rate.

Experiments in Figure 2 and Figure 4 should be performed using the same fly lines.

In Figure 2D-E, ChIP experiments were performed on flies stably activated by temperature. They hold the P(TARGET) since it is required for their activation and used as internal reporter of the activated state. In Figure 4, measurement of antisense lacZ transcripts from BX2 implies that all lacZ RNA should come from BX2 locus. The presence of P(TARGET) which shares lacZ sequence would impair the specific measurement of lacZ antisense coming from BX2 locus.

4) Is the BX2 transgene in the BX2^OFF line containing P(TARGET) also inserted in the AGO1 gene? This should be examined. If this were the case, the authors better examine whether piRNAs were derived from the AGO1 gene.

All BX2 transgenes, whether ON or OFF, are inserted into the AGO1 gene. We have analyzed the small RNA produced from AGO1 gene region in the BX2^OFF and BX2^ON strains: no significant piRNA amount was produced in all epigenetic contexts. This shows that AGO1 is not a natural piRNA producing locus. Later, looking for piRNAs from AGO1 gene region at 25°C and 29°C in BX2^OFF strains confirm that the AGO1 gene is not a piRNA cluster and that 29°C does not induce de novo production of small RNAs from the AGO1 gene region. These data have been added in a new Figure 2—figure supplement 2 and in Figure 4J and are discussed in the text. Figure legends have been modified as well.

5) The authors should test experimentally whether BX2-derived piRNAs are loaded into PIWI proteins. This is very important.

From the first description of piRNAs and piRNA clusters by Brennecke et al., 2007, it is known that piRNAs are loaded by the PIWI proteins allowing their biogenesis. Through these loadings, piRNAs have four main signatures that defined piRNAs: 23-29 nt, a 1U bias, a 10A bias and a 10 nt overlap between pairs of 23-29 nt small RNAs (ping-pong signature). Since this first description, several other studies used these properties in numerous organisms (Nematostella, Bombyx mori, D. simulans, marmoset, ….) and identified piRNAs. The data of Figure 2 shows these characteristics that convince us that bona fide piRNAs were produced by the new activated BX2 locus. On top of that, the strength of the transgene system used in this study allows also to provide the ultimate proof: the converted BX2^theta locus is producing small RNA that are functionally active for germline lacZ repression.

6) Figure 2B: Explain why BX2-derived piRNAs did not show 10A bias. The data shown in Figure 2C supported the idea that BX2-derived piRNAs were products of the ping-pong pathway. Then, they should have shown 10A bias, but in reality they did not.

The 10A bias is only observed in the fraction of AGO3-loaded piRNA (Brennecke et al., 2007) or can be detected in the fraction of piRNA that paired each other (De Vanssay et al., 2012). We have performed this last analysis and the results are now shown on Figure 2C and Figure 2—figure supplement 1C. The figure legends and the Materials and methods section have been completed as well.

7) Figure 2DE: The authors should examine genome-widely where in the genome H3K9me3 and Rhino accumulated upon heat treatment. It is hard to imagine that H3K9me3 and Rhino accumulation only occurred at the BX2 locus upon heat treatment.

We agree that studying whole genome effects of heat treatment could be an interesting question. However, we did not detect production of new piRNA from other loci (Figure 2—figure supplement 3). Therefore, even if modification of chromatin marks were detected elsewhere in the genome, they are not followed by a stable new piRNA production, that is the question addressed in this study. Once again, molecular analyses were performed on flies that have been activated at 29°C and then put back at 25°C. Only stable modifications can be observed. Our results showed that only BX2 is metastable for piRNA production, all other potential piRNA producing loci were already activated before heat treatment, and even if others loci are stably modified in their chromatin marks (H3K9me3 or Rhino), they did not become piRNA producing loci.

8) The authors should examine whether H3K9me3 and Rhino accumulated at the BX2 locus in BX2^OFF line without P(TARGET) upon heat treatment.

The low occurrence of epigenetic conversion (2%) does not allow enough dynamics to analyze chromatin modification. Without P(TARGET), no conversion is observed and ChIP experiment on BX2^OFF has already been performed and show less H3K9me3 or Rhino enrichment compared to BX2* or BX2^theta.

9) This manuscript should be edited thoroughly by a native English speaker.

The manuscript has been read and corrected by a native speaker (Dr Lori Pile).

10) Table 1 should be removed from the main text.

It has been done. Table 1 has been renamed Supplementary file 8

Reviewer #3:

[…] While I have no problem with the experiments, but did wish for more insight into the mechanism, and more caution in the interpretation.

As previously described, we believe that the experiment with additional transgenes and with Dcr-2 provide more insight related to the mechanistic part of the phenomenon that is illustrated by a model presented in Figure 5.

In particular, entirely enviromental specification of piRNA clusters implied in the Abstract is a strong claim, and as the authors show themselves, is an oversimplification. The induction they find is quite specific to their system. Specifically, there are higher levels of homologous sense and anti-sense transcript at 29 than at 25, and temperature alone doesn't change other piRNA production.

We agree that we oversimplified the message in the Abstract and as the first reviewer suggested, we now precisely state the role of different parameters required for the activation of BX2.

Similarly, it would be useful to know if the expression of the sense transcript from P-Target is also higher at 29 that at 25. (I think this construct has a heat-shock promoter, as well as that of the P-element, which shows temperature sensitive effects.)

We performed this experiment and the result is that P(TARGET) is less expressed at 29°C. This result has been added in the text and in a new supplementary figure (Figure 4—figure supplement 1C). As mentioned in the Materials and Methods, this transgene does not carry a heat shock promoter, it is an enhancer trap transgene inserted in the eEF5 gene, an ubiquitous translation elongation factor.

I wasn't entirely convinced by the suggestion that expression of natural piRNA clusters is higher at 29C. Many piRNAs are generated as a secondary byproduct of transposable element message, and many transposable elements seem to have temperature sensitive expression (possibly as a byproduct of changes in chromatin.)

We do not suggest that expression of natural piRNA clusters is higher at 29°C. Actually, the analysis of the major 42AB piRNA cluster shows a slight decrease of piRNA at 29°C compare to 25°C (Figure 4E). We totally agree that variations in piRNA production from natural piRNA clusters may result from several parameters including ping-pong amplification and variation in the expression of piRNA genes (as suggested by Fast et al., 2017). We have rephrased this part to clarify this point.

The piRNA dependent splicing suppression of the P-element is also, apparently, not temperature sensitive (Teixiera et al., 2017).

We note that the intron 2-3 of the P element, whose splicing is dependent of piRNA, is absent from all transgenes that have been used in this study: the P(lacW) transgene that constitutes the BX2 locus, P(TARGET) and P(A92), the two P-lacZ-rosy used as reporter transgenes.

Minor points. It does seem important (and quick) to eliminate the possibility that there's any active P-elements in these lines. While w1118 is a classic M-type background, there are some cases of sublines of other M-types harboring P-elements via contamination (Rahman et al. Nucleic Acids Research, Volume 43, Issue 22, 15 December 2015, Pages 10655-10672). If there is active P-element, I think it is possible that transgenes (which has the TIRs necessary for P-element insertion) has been picked up and put in another piRNA cluster. It's a simple matter of a few PCRs to exclude this possibility. The Cy and CyRoi backgrounds should be tested as well, to show that they are devoid of any plus-strand transcript that could be an ongoing trigger for piRNA production.

Several years ago we tested by PCR the absence of P element in the strains we used for our studies, especially in the balanced stocks commonly used. We checked again by PCR (using the primer 5'-TGATGAAATAACATAAGGTGGTCCCGTCG-3' known as P3-31 that recognized the P inverted repeated sequences) all of the stocks that we used in this study and all of the strains are devoid of natural P sequences. We propose to not add this result in the paper since it is a very specialized point but our results are available if needed. Another point to this remark is that we have established recombined lines from BX2, P(TARGET) and, in the absence of P(TARGET), they did not show activation of the BX2 locus at 29°C. If uncontrolled, natural P elements were present in our strain and responsible of the conversion, they should not all be eliminated by recombination and control lines could still show BX2 activation. Additionally, we have published that piRNA produced by P sequences only are not able to activate BX2 by paramutation (Hermant et al., 2015) demonstrating that the homology length conferred by P sequences is not sufficient to trigger BX2 conversion, even by piRNA maternal inheritance.

Subsection “Epigenetic conversion at 29°C occurs at a low rate from the first generation” paragraph two. The model is entirely sensible, but the data are quite noisy, and seem like they would be consistent with any model predicting an increase over generations. Can they compare this model to, for example, a model without the “c” parameter using Akaike Information Criteria or a log-likelihood test? This would make the analysis more informative.

We agree that the data are noisy. This is due to the high variation of occurrence of the epigenetic conversion and its sampling through generations and among replicates. The first rough observations of the phenomenon reported complete repression in ovaries after several generations. We tried to understand this effect by analyzing thousands of flies in one generation rather that dozens of flies through multiple generations and observed a small conversion frequency (≈2%, Figure 3). Then, we searched for a simple model based on the conversion frequency we observed in one generation and the fact that the ON state is transmitted to the next generation (De Vanssay et al., 2012 and Figure 3). The major problem we encountered is related to the sampling that occurred at each generation: among all egg chambers that are produced, only dozens of flies were analyzed. We think that the sampling size generates the variation we observed among replicates with a 2% conversion phenomenon. We agree with the reviewer that our data would fit with a number of models describing an increase over generations. We can remove this part if it is still considered non informative.

Subsection “Heat conversion requires a homologous sequence in trans”: The chi-square statistic and degrees of freedom should be reported as well as the p-value. (And this is very minor, but the Greek letter is usually used in place of “chi”.)

We have added in the new version the χ² value (23.35) and the degrees of freedom (2) and replaced “chi2” by “χ²” each time we used the test.

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

The reviews follow below, and we will look forward to hearing how you propose to proceed.

Reviewer #1:

In the revised version of the manuscript the authors find an involvement of sense-antisense transcripts and Dicer-2 in piRNA biogenesis. The authors may remember that one of the early findings celebrated as a major discovery was the lack of a role for double-stranded RNAs and the double-stranded RNA cleaving enzyme Dicer, in piRNA biogenesis. If now, the authors find a role for these in piRNA biogenesis, it has to be backed with strong data.

1) Can the authors demonstrate loading of the 21 nt RNAs into Piwi proteins?

2) If its role is epigenetic conversion, then ideally it has be loaded into the nuclear Piwi.

3) Are there experiments linking Dicer-2 to loading into a Piwi protein? Are they ever found in the same complexes?

I agree that there is a strong functional/genetic data that is mysterious, but very interesting. The heat-activated piRNAs are functional as they can silence a LacZ transgene. Since this manuscript will be published in some form, I am happy to support description of this part in a revised manuscript, but without wild speculation of Dicer-2 in piRNA biogenesis (even if they examine a Dicer mutant). These attempts at explaining the molecular mechanism may be toned down. I am worried that any prominent mention of Dicer-2 in this context may only serve to foster suspicion of the interesting genetic observation.

We agree with the first reviewer that implications of Dcr-2 in piRNA biology suggest a major breakthrough that our experiments do not definitively prove. We would like to clarify, however, that the role of Dcr-2 we proposed was not in the maintenance of piRNA production, but rather, a more opportunistic way to accidentally produce some new piRNAs. In our system, where the BX2 cluster is a peculiar genomic structure, this de novo piRNAs production would be sufficient to convert BX2 into an active piRNA cluster. Once activated, Dcr-2 is no longer required for piRNA production from BX2 as we published in Hermant et al., 2015. We agree that the biochemical experiments proposed by the first reviewer are well adapted in theory. We think, however, that the 2% conversion rate is too low to be able to observe such molecular details. For these reasons, we agree to your proposal to remove the results obtained with Dcr-2 and tone down our model presented in Figure 5. We propose to consider the hypothesis of somehow making double strand RNAs as a starting point to accidentally produce the first new piRNAs, but clearly state that much of the mechanistic aspects remain unknown, as asked by the editors.

Reviewer #3:

I still find this study clever and interesting, and think the manuscript shows some specific improvements:

The “primed” nature of the BX2 locus is clearer, and the claim in the paper is more measured and reasonably justified.

The experiment showing the Dcr2 dependence of the conversion of the BX2 from off to on goes a little way toward understanding the mechanism of piRNA cluster formation, implying that it is dependent on siRNAs.

The addition of the figure showing the model for the mechanism. The weak point of the proposed mechanism, as I see it, is the lack of evidence that the siRNAs are loaded onto PIWI, any evidence the authors can provide for this mechanism would greatly improve the paper. I can't make any suggestions that seem technically feasible, however, in light of the low conversion rate in this system.

As stated above, we agree that the biochemical experiments proposed by the reviewer are well adapted in theory but that the 2% conversion rate is too low to be able to observe such molecular details. For these reasons, we agree to your proposal to remove the results obtained with Dcr-2 and tone down our model presented in Figure 5.

Further clarification would also be helpful for these aspects:

Regarding the model in Figure 1—figure supplement 3: I previously suggested a statistical analysis of this model. Rather than the more detailed analysis I suggested previously, I think a reasonable compromise would be to show that there is a significant increase over generations. That is the important point here (and, in fact, it's hard to imagine a scenario where the “memory” of conversion does not play a role in the increase over generations). But, while there appears to be a trend, the data are sufficiently noisy that it would be useful just to see that the increase is significant (perhaps a linear regression on transformed data?).

We removed the mathematical model of progression of conversion during generation and propose a simpler description of the variability observed among lines. Mean and 95% confidence interval per generation were calculated on modified data (arcsine square root). The text (subsection “Epigenetic conversion at 29°C occurs at a low rate from the first generation”), the Figure 1—figure supplement 3B and its legend have been modified accordingly.

Discussion opening paragraph: Please explain what you mean by the “specific spliced transcript” comment; it wasn't clear.

By “specific spliced transcript”, we mean that we measured AGO1 transcripts that overlap the BX2 insertion point and that come from a distant promoter. We have added this specific description in the legend of Figure 4. There is another promoter for AGO1 that is located after the BX2 insertion point. AGO1 transcripts produced from this promoter are only slightly affected by temperature (not significantly), allowing us to conclude that it could explain the apparent discrepancy with the results published in Fast et al., 2017. We have added this analysis in Figure 4—figure supplement 1D and modified the discussion and the legend, accordingly.

e.g., supplementary file 11: It would be nice to more cautious in the interpretation where there are large differences in the number of flies examined. In Figure 3, for example, I think the numbers show that 41 of 1447 females show partial or complete repression; the comparable (?) numbers is Supplementary file 11 show 0 of 32 (?) females show any repression. This is not a significant difference via Fisher's exact test.

We would like to clarify that the results obtained with NaCl and heat shock have to be compared to those obtained at 29°C in G1 (showed in Figure 3), considering the proportion of converted egg chambers and not flies (that are analyzed in G2). For the heat shock experiment, the Χ² test was 106.03 and the p-value was 7.25 x 10^-25 (heat shock 0/3840 versus 29°C 586/21720) and for the NaCl experiment, the Χ² test was 115.93 with a p-value of 4.91 x 10^-27 (NaCl 0/4200 versus 29°C 586/21720). We have described this analysis in a clearer way in the legend of supplementary file 11 and by adding the results obtained at 29°C (from Figure 3) in the table of the revised version.

Finally, we renamed the P(TARGET) transgenes used in this study as asked by editors and specified their domain of expression to be as clear as possible for a large audience.

Author details

1. Karine Casier

   Laboratoire Biologie du Développement, UMR7622, Sorbonne Université, CNRS, Institut de Biologie Paris-Seine, Paris, France

   Contribution

   Conceptualization, Data curation, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing—review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-6825-8057

2. Valérie Delmarre

   Laboratoire Biologie du Développement, UMR7622, Sorbonne Université, CNRS, Institut de Biologie Paris-Seine, Paris, France

   Contribution

   Data curation, Validation, Investigation

   Competing interests

   No competing interests declared

3. Nathalie Gueguen

   GReD, Université Clermont Auvergne, CNRS, INSERM, BP 10448, Clermont-Ferrand, France

   Contribution

   Validation, Investigation

   Competing interests

   No competing interests declared

4. Catherine Hermant

   Laboratoire Biologie du Développement, UMR7622, Sorbonne Université, CNRS, Institut de Biologie Paris-Seine, Paris, France

   Present address

   Center for Interdisciplinary Research in Biology (CIRB), Collège de France, CNRS, INSERM, PSL Research University, Paris, France

   Contribution

   Validation, Investigation

   Competing interests

   No competing interests declared

5. Elise Viodé

   Laboratoire Biologie du Développement, UMR7622, Sorbonne Université, CNRS, Institut de Biologie Paris-Seine, Paris, France

   Contribution

   Validation, Investigation

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-5058-2710

6. Chantal Vaury

   GReD, Université Clermont Auvergne, CNRS, INSERM, BP 10448, Clermont-Ferrand, France

   Contribution

   Funding acquisition, Writing—review and editing

   Competing interests

   No competing interests declared

7. Stéphane Ronsseray

   Laboratoire Biologie du Développement, UMR7622, Sorbonne Université, CNRS, Institut de Biologie Paris-Seine, Paris, France

   Contribution

   Conceptualization, Resources, Funding acquisition, Investigation, Writing—review and editing

   Competing interests

   No competing interests declared

8. Emilie Brasset

   GReD, Université Clermont Auvergne, CNRS, INSERM, BP 10448, Clermont-Ferrand, France

   Contribution

   Validation, Investigation, Writing—review and editing

   Competing interests

   No competing interests declared

9. Laure Teysset

   Laboratoire Biologie du Développement, UMR7622, Sorbonne Université, CNRS, Institut de Biologie Paris-Seine, Paris, France

   Contribution

   Supervision, Funding acquisition, Investigation, Visualization, Project administration, Writing—review and editing

   For correspondence

   laure.teysset@upmc.fr

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-7413-1850

10. Antoine Boivin

    Laboratoire Biologie du Développement, UMR7622, Sorbonne Université, CNRS, Institut de Biologie Paris-Seine, Paris, France

    Contribution

    Conceptualization, Data curation, Formal analysis, Supervision, Validation, Investigation, Visualization, Methodology, Writing—original draft, Project administration, Writing—review and editing

    For correspondence

    antoine.boivin@upmc.fr

    Competing interests

    No competing interests declared

    "This ORCID iD identifies the author of this article:" 0000-0001-8671-1599

• 2,147

  Page views

• 283

  Downloads

• 17

  Citations

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